target-i386: Clear KVM CPUID features if KVM is disabled
[qemu/kevin.git] / blockdev.c
blob29c6561fd89dc076fcd412ab987904c37b72208c
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/throttle-groups.h"
39 #include "monitor/monitor.h"
40 #include "qemu/error-report.h"
41 #include "qemu/option.h"
42 #include "qemu/config-file.h"
43 #include "qapi/qmp/types.h"
44 #include "qapi-visit.h"
45 #include "qapi/qmp/qerror.h"
46 #include "qapi/qmp-output-visitor.h"
47 #include "qapi/util.h"
48 #include "sysemu/sysemu.h"
49 #include "block/block_int.h"
50 #include "qmp-commands.h"
51 #include "trace.h"
52 #include "sysemu/arch_init.h"
53 #include "qemu/cutils.h"
54 #include "qemu/help_option.h"
56 static QTAILQ_HEAD(, BlockDriverState) monitor_bdrv_states =
57 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states);
59 static int do_open_tray(const char *blk_name, const char *qdev_id,
60 bool force, Error **errp);
62 static const char *const if_name[IF_COUNT] = {
63 [IF_NONE] = "none",
64 [IF_IDE] = "ide",
65 [IF_SCSI] = "scsi",
66 [IF_FLOPPY] = "floppy",
67 [IF_PFLASH] = "pflash",
68 [IF_MTD] = "mtd",
69 [IF_SD] = "sd",
70 [IF_VIRTIO] = "virtio",
71 [IF_XEN] = "xen",
74 static int if_max_devs[IF_COUNT] = {
76 * Do not change these numbers! They govern how drive option
77 * index maps to unit and bus. That mapping is ABI.
79 * All controllers used to implement if=T drives need to support
80 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
81 * Otherwise, some index values map to "impossible" bus, unit
82 * values.
84 * For instance, if you change [IF_SCSI] to 255, -drive
85 * if=scsi,index=12 no longer means bus=1,unit=5, but
86 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
87 * the drive can't be set up. Regression.
89 [IF_IDE] = 2,
90 [IF_SCSI] = 7,
93 /**
94 * Boards may call this to offer board-by-board overrides
95 * of the default, global values.
97 void override_max_devs(BlockInterfaceType type, int max_devs)
99 BlockBackend *blk;
100 DriveInfo *dinfo;
102 if (max_devs <= 0) {
103 return;
106 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
107 dinfo = blk_legacy_dinfo(blk);
108 if (dinfo->type == type) {
109 fprintf(stderr, "Cannot override units-per-bus property of"
110 " the %s interface, because a drive of that type has"
111 " already been added.\n", if_name[type]);
112 g_assert_not_reached();
116 if_max_devs[type] = max_devs;
120 * We automatically delete the drive when a device using it gets
121 * unplugged. Questionable feature, but we can't just drop it.
122 * Device models call blockdev_mark_auto_del() to schedule the
123 * automatic deletion, and generic qdev code calls blockdev_auto_del()
124 * when deletion is actually safe.
126 void blockdev_mark_auto_del(BlockBackend *blk)
128 DriveInfo *dinfo = blk_legacy_dinfo(blk);
129 BlockDriverState *bs = blk_bs(blk);
130 AioContext *aio_context;
132 if (!dinfo) {
133 return;
136 if (bs) {
137 aio_context = bdrv_get_aio_context(bs);
138 aio_context_acquire(aio_context);
140 if (bs->job) {
141 block_job_cancel(bs->job);
144 aio_context_release(aio_context);
147 dinfo->auto_del = 1;
150 void blockdev_auto_del(BlockBackend *blk)
152 DriveInfo *dinfo = blk_legacy_dinfo(blk);
154 if (dinfo && dinfo->auto_del) {
155 monitor_remove_blk(blk);
156 blk_unref(blk);
161 * Returns the current mapping of how many units per bus
162 * a particular interface can support.
164 * A positive integer indicates n units per bus.
165 * 0 implies the mapping has not been established.
166 * -1 indicates an invalid BlockInterfaceType was given.
168 int drive_get_max_devs(BlockInterfaceType type)
170 if (type >= IF_IDE && type < IF_COUNT) {
171 return if_max_devs[type];
174 return -1;
177 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
179 int max_devs = if_max_devs[type];
180 return max_devs ? index / max_devs : 0;
183 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
185 int max_devs = if_max_devs[type];
186 return max_devs ? index % max_devs : index;
189 QemuOpts *drive_def(const char *optstr)
191 return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
194 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
195 const char *optstr)
197 QemuOpts *opts;
199 opts = drive_def(optstr);
200 if (!opts) {
201 return NULL;
203 if (type != IF_DEFAULT) {
204 qemu_opt_set(opts, "if", if_name[type], &error_abort);
206 if (index >= 0) {
207 qemu_opt_set_number(opts, "index", index, &error_abort);
209 if (file)
210 qemu_opt_set(opts, "file", file, &error_abort);
211 return opts;
214 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
216 BlockBackend *blk;
217 DriveInfo *dinfo;
219 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
220 dinfo = blk_legacy_dinfo(blk);
221 if (dinfo && dinfo->type == type
222 && dinfo->bus == bus && dinfo->unit == unit) {
223 return dinfo;
227 return NULL;
230 bool drive_check_orphaned(void)
232 BlockBackend *blk;
233 DriveInfo *dinfo;
234 bool rs = false;
236 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
237 dinfo = blk_legacy_dinfo(blk);
238 /* If dinfo->bdrv->dev is NULL, it has no device attached. */
239 /* Unless this is a default drive, this may be an oversight. */
240 if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
241 dinfo->type != IF_NONE) {
242 fprintf(stderr, "Warning: Orphaned drive without device: "
243 "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
244 blk_name(blk), blk_bs(blk) ? blk_bs(blk)->filename : "",
245 if_name[dinfo->type], dinfo->bus, dinfo->unit);
246 rs = true;
250 return rs;
253 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
255 return drive_get(type,
256 drive_index_to_bus_id(type, index),
257 drive_index_to_unit_id(type, index));
260 int drive_get_max_bus(BlockInterfaceType type)
262 int max_bus;
263 BlockBackend *blk;
264 DriveInfo *dinfo;
266 max_bus = -1;
267 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
268 dinfo = blk_legacy_dinfo(blk);
269 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
270 max_bus = dinfo->bus;
273 return max_bus;
276 /* Get a block device. This should only be used for single-drive devices
277 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
278 appropriate bus. */
279 DriveInfo *drive_get_next(BlockInterfaceType type)
281 static int next_block_unit[IF_COUNT];
283 return drive_get(type, 0, next_block_unit[type]++);
286 static void bdrv_format_print(void *opaque, const char *name)
288 error_printf(" %s", name);
291 typedef struct {
292 QEMUBH *bh;
293 BlockDriverState *bs;
294 } BDRVPutRefBH;
296 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
298 if (!strcmp(buf, "ignore")) {
299 return BLOCKDEV_ON_ERROR_IGNORE;
300 } else if (!is_read && !strcmp(buf, "enospc")) {
301 return BLOCKDEV_ON_ERROR_ENOSPC;
302 } else if (!strcmp(buf, "stop")) {
303 return BLOCKDEV_ON_ERROR_STOP;
304 } else if (!strcmp(buf, "report")) {
305 return BLOCKDEV_ON_ERROR_REPORT;
306 } else {
307 error_setg(errp, "'%s' invalid %s error action",
308 buf, is_read ? "read" : "write");
309 return -1;
313 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
314 Error **errp)
316 const QListEntry *entry;
317 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
318 switch (qobject_type(entry->value)) {
320 case QTYPE_QSTRING: {
321 unsigned long long length;
322 const char *str = qstring_get_str(qobject_to_qstring(entry->value));
323 if (parse_uint_full(str, &length, 10) == 0 &&
324 length > 0 && length <= UINT_MAX) {
325 block_acct_add_interval(stats, (unsigned) length);
326 } else {
327 error_setg(errp, "Invalid interval length: %s", str);
328 return false;
330 break;
333 case QTYPE_QINT: {
334 int64_t length = qint_get_int(qobject_to_qint(entry->value));
335 if (length > 0 && length <= UINT_MAX) {
336 block_acct_add_interval(stats, (unsigned) length);
337 } else {
338 error_setg(errp, "Invalid interval length: %" PRId64, length);
339 return false;
341 break;
344 default:
345 error_setg(errp, "The specification of stats-intervals is invalid");
346 return false;
349 return true;
352 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
354 /* All parameters but @opts are optional and may be set to NULL. */
355 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
356 const char **throttling_group, ThrottleConfig *throttle_cfg,
357 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
359 const char *discard;
360 Error *local_error = NULL;
361 const char *aio;
363 if (bdrv_flags) {
364 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
365 *bdrv_flags |= BDRV_O_COPY_ON_READ;
368 if ((discard = qemu_opt_get(opts, "discard")) != NULL) {
369 if (bdrv_parse_discard_flags(discard, bdrv_flags) != 0) {
370 error_setg(errp, "Invalid discard option");
371 return;
375 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
376 if (!strcmp(aio, "native")) {
377 *bdrv_flags |= BDRV_O_NATIVE_AIO;
378 } else if (!strcmp(aio, "threads")) {
379 /* this is the default */
380 } else {
381 error_setg(errp, "invalid aio option");
382 return;
387 /* disk I/O throttling */
388 if (throttling_group) {
389 *throttling_group = qemu_opt_get(opts, "throttling.group");
392 if (throttle_cfg) {
393 throttle_config_init(throttle_cfg);
394 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
395 qemu_opt_get_number(opts, "throttling.bps-total", 0);
396 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
397 qemu_opt_get_number(opts, "throttling.bps-read", 0);
398 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
399 qemu_opt_get_number(opts, "throttling.bps-write", 0);
400 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
401 qemu_opt_get_number(opts, "throttling.iops-total", 0);
402 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
403 qemu_opt_get_number(opts, "throttling.iops-read", 0);
404 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
405 qemu_opt_get_number(opts, "throttling.iops-write", 0);
407 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
408 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
409 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
410 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
411 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
412 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
413 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
414 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
415 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
416 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
417 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
418 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
420 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
421 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
422 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
423 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
424 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
425 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
426 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
427 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
428 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
429 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
430 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
431 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
433 throttle_cfg->op_size =
434 qemu_opt_get_number(opts, "throttling.iops-size", 0);
436 if (!throttle_is_valid(throttle_cfg, errp)) {
437 return;
441 if (detect_zeroes) {
442 *detect_zeroes =
443 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
444 qemu_opt_get(opts, "detect-zeroes"),
445 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX,
446 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
447 &local_error);
448 if (local_error) {
449 error_propagate(errp, local_error);
450 return;
453 if (bdrv_flags &&
454 *detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
455 !(*bdrv_flags & BDRV_O_UNMAP))
457 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
458 "without setting discard operation to unmap");
459 return;
464 /* Takes the ownership of bs_opts */
465 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
466 Error **errp)
468 const char *buf;
469 int bdrv_flags = 0;
470 int on_read_error, on_write_error;
471 bool account_invalid, account_failed;
472 bool writethrough, read_only;
473 BlockBackend *blk;
474 BlockDriverState *bs;
475 ThrottleConfig cfg;
476 int snapshot = 0;
477 Error *error = NULL;
478 QemuOpts *opts;
479 QDict *interval_dict = NULL;
480 QList *interval_list = NULL;
481 const char *id;
482 BlockdevDetectZeroesOptions detect_zeroes =
483 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
484 const char *throttling_group = NULL;
486 /* Check common options by copying from bs_opts to opts, all other options
487 * stay in bs_opts for processing by bdrv_open(). */
488 id = qdict_get_try_str(bs_opts, "id");
489 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
490 if (error) {
491 error_propagate(errp, error);
492 goto err_no_opts;
495 qemu_opts_absorb_qdict(opts, bs_opts, &error);
496 if (error) {
497 error_propagate(errp, error);
498 goto early_err;
501 if (id) {
502 qdict_del(bs_opts, "id");
505 /* extract parameters */
506 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
508 account_invalid = qemu_opt_get_bool(opts, "stats-account-invalid", true);
509 account_failed = qemu_opt_get_bool(opts, "stats-account-failed", true);
511 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
513 id = qemu_opts_id(opts);
515 qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
516 qdict_array_split(interval_dict, &interval_list);
518 if (qdict_size(interval_dict) != 0) {
519 error_setg(errp, "Invalid option stats-intervals.%s",
520 qdict_first(interval_dict)->key);
521 goto early_err;
524 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
525 &detect_zeroes, &error);
526 if (error) {
527 error_propagate(errp, error);
528 goto early_err;
531 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
532 if (is_help_option(buf)) {
533 error_printf("Supported formats:");
534 bdrv_iterate_format(bdrv_format_print, NULL);
535 error_printf("\n");
536 goto early_err;
539 if (qdict_haskey(bs_opts, "driver")) {
540 error_setg(errp, "Cannot specify both 'driver' and 'format'");
541 goto early_err;
543 qdict_put(bs_opts, "driver", qstring_from_str(buf));
546 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
547 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
548 on_write_error = parse_block_error_action(buf, 0, &error);
549 if (error) {
550 error_propagate(errp, error);
551 goto early_err;
555 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
556 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
557 on_read_error = parse_block_error_action(buf, 1, &error);
558 if (error) {
559 error_propagate(errp, error);
560 goto early_err;
564 if (snapshot) {
565 bdrv_flags |= BDRV_O_SNAPSHOT;
568 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
570 /* init */
571 if ((!file || !*file) && !qdict_size(bs_opts)) {
572 BlockBackendRootState *blk_rs;
574 blk = blk_new();
575 blk_rs = blk_get_root_state(blk);
576 blk_rs->open_flags = bdrv_flags;
577 blk_rs->read_only = read_only;
578 blk_rs->detect_zeroes = detect_zeroes;
580 QDECREF(bs_opts);
581 } else {
582 if (file && !*file) {
583 file = NULL;
586 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
587 * with other callers) rather than what we want as the real defaults.
588 * Apply the defaults here instead. */
589 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
590 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
591 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
592 read_only ? "on" : "off");
593 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
595 if (runstate_check(RUN_STATE_INMIGRATE)) {
596 bdrv_flags |= BDRV_O_INACTIVE;
599 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
600 if (!blk) {
601 goto err_no_bs_opts;
603 bs = blk_bs(blk);
605 bs->detect_zeroes = detect_zeroes;
607 if (bdrv_key_required(bs)) {
608 autostart = 0;
611 block_acct_init(blk_get_stats(blk), account_invalid, account_failed);
613 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
614 blk_unref(blk);
615 blk = NULL;
616 goto err_no_bs_opts;
620 /* disk I/O throttling */
621 if (throttle_enabled(&cfg)) {
622 if (!throttling_group) {
623 throttling_group = id;
625 blk_io_limits_enable(blk, throttling_group);
626 blk_set_io_limits(blk, &cfg);
629 blk_set_enable_write_cache(blk, !writethrough);
630 blk_set_on_error(blk, on_read_error, on_write_error);
632 if (!monitor_add_blk(blk, id, errp)) {
633 blk_unref(blk);
634 blk = NULL;
635 goto err_no_bs_opts;
638 err_no_bs_opts:
639 qemu_opts_del(opts);
640 QDECREF(interval_dict);
641 QDECREF(interval_list);
642 return blk;
644 early_err:
645 qemu_opts_del(opts);
646 QDECREF(interval_dict);
647 QDECREF(interval_list);
648 err_no_opts:
649 QDECREF(bs_opts);
650 return NULL;
653 static QemuOptsList qemu_root_bds_opts;
655 /* Takes the ownership of bs_opts */
656 static BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
658 BlockDriverState *bs;
659 QemuOpts *opts;
660 Error *local_error = NULL;
661 BlockdevDetectZeroesOptions detect_zeroes;
662 int bdrv_flags = 0;
664 opts = qemu_opts_create(&qemu_root_bds_opts, NULL, 1, errp);
665 if (!opts) {
666 goto fail;
669 qemu_opts_absorb_qdict(opts, bs_opts, &local_error);
670 if (local_error) {
671 error_propagate(errp, local_error);
672 goto fail;
675 extract_common_blockdev_options(opts, &bdrv_flags, NULL, NULL,
676 &detect_zeroes, &local_error);
677 if (local_error) {
678 error_propagate(errp, local_error);
679 goto fail;
682 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
683 * with other callers) rather than what we want as the real defaults.
684 * Apply the defaults here instead. */
685 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
686 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
687 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
689 if (runstate_check(RUN_STATE_INMIGRATE)) {
690 bdrv_flags |= BDRV_O_INACTIVE;
693 bs = bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
694 if (!bs) {
695 goto fail_no_bs_opts;
698 bs->detect_zeroes = detect_zeroes;
700 fail_no_bs_opts:
701 qemu_opts_del(opts);
702 return bs;
704 fail:
705 qemu_opts_del(opts);
706 QDECREF(bs_opts);
707 return NULL;
710 void blockdev_close_all_bdrv_states(void)
712 BlockDriverState *bs, *next_bs;
714 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
715 AioContext *ctx = bdrv_get_aio_context(bs);
717 aio_context_acquire(ctx);
718 bdrv_unref(bs);
719 aio_context_release(ctx);
723 /* Iterates over the list of monitor-owned BlockDriverStates */
724 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
726 return bs ? QTAILQ_NEXT(bs, monitor_list)
727 : QTAILQ_FIRST(&monitor_bdrv_states);
730 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
731 Error **errp)
733 const char *value;
735 value = qemu_opt_get(opts, from);
736 if (value) {
737 if (qemu_opt_find(opts, to)) {
738 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
739 "same time", to, from);
740 return;
744 /* rename all items in opts */
745 while ((value = qemu_opt_get(opts, from))) {
746 qemu_opt_set(opts, to, value, &error_abort);
747 qemu_opt_unset(opts, from);
751 QemuOptsList qemu_legacy_drive_opts = {
752 .name = "drive",
753 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
754 .desc = {
756 .name = "bus",
757 .type = QEMU_OPT_NUMBER,
758 .help = "bus number",
760 .name = "unit",
761 .type = QEMU_OPT_NUMBER,
762 .help = "unit number (i.e. lun for scsi)",
764 .name = "index",
765 .type = QEMU_OPT_NUMBER,
766 .help = "index number",
768 .name = "media",
769 .type = QEMU_OPT_STRING,
770 .help = "media type (disk, cdrom)",
772 .name = "if",
773 .type = QEMU_OPT_STRING,
774 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
776 .name = "cyls",
777 .type = QEMU_OPT_NUMBER,
778 .help = "number of cylinders (ide disk geometry)",
780 .name = "heads",
781 .type = QEMU_OPT_NUMBER,
782 .help = "number of heads (ide disk geometry)",
784 .name = "secs",
785 .type = QEMU_OPT_NUMBER,
786 .help = "number of sectors (ide disk geometry)",
788 .name = "trans",
789 .type = QEMU_OPT_STRING,
790 .help = "chs translation (auto, lba, none)",
792 .name = "boot",
793 .type = QEMU_OPT_BOOL,
794 .help = "(deprecated, ignored)",
796 .name = "addr",
797 .type = QEMU_OPT_STRING,
798 .help = "pci address (virtio only)",
800 .name = "serial",
801 .type = QEMU_OPT_STRING,
802 .help = "disk serial number",
804 .name = "file",
805 .type = QEMU_OPT_STRING,
806 .help = "file name",
809 /* Options that are passed on, but have special semantics with -drive */
811 .name = BDRV_OPT_READ_ONLY,
812 .type = QEMU_OPT_BOOL,
813 .help = "open drive file as read-only",
815 .name = "rerror",
816 .type = QEMU_OPT_STRING,
817 .help = "read error action",
819 .name = "werror",
820 .type = QEMU_OPT_STRING,
821 .help = "write error action",
823 .name = "copy-on-read",
824 .type = QEMU_OPT_BOOL,
825 .help = "copy read data from backing file into image file",
828 { /* end of list */ }
832 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
834 const char *value;
835 BlockBackend *blk;
836 DriveInfo *dinfo = NULL;
837 QDict *bs_opts;
838 QemuOpts *legacy_opts;
839 DriveMediaType media = MEDIA_DISK;
840 BlockInterfaceType type;
841 int cyls, heads, secs, translation;
842 int max_devs, bus_id, unit_id, index;
843 const char *devaddr;
844 const char *werror, *rerror;
845 bool read_only = false;
846 bool copy_on_read;
847 const char *serial;
848 const char *filename;
849 Error *local_err = NULL;
850 int i;
852 /* Change legacy command line options into QMP ones */
853 static const struct {
854 const char *from;
855 const char *to;
856 } opt_renames[] = {
857 { "iops", "throttling.iops-total" },
858 { "iops_rd", "throttling.iops-read" },
859 { "iops_wr", "throttling.iops-write" },
861 { "bps", "throttling.bps-total" },
862 { "bps_rd", "throttling.bps-read" },
863 { "bps_wr", "throttling.bps-write" },
865 { "iops_max", "throttling.iops-total-max" },
866 { "iops_rd_max", "throttling.iops-read-max" },
867 { "iops_wr_max", "throttling.iops-write-max" },
869 { "bps_max", "throttling.bps-total-max" },
870 { "bps_rd_max", "throttling.bps-read-max" },
871 { "bps_wr_max", "throttling.bps-write-max" },
873 { "iops_size", "throttling.iops-size" },
875 { "group", "throttling.group" },
877 { "readonly", BDRV_OPT_READ_ONLY },
880 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
881 qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
882 &local_err);
883 if (local_err) {
884 error_report_err(local_err);
885 return NULL;
889 value = qemu_opt_get(all_opts, "cache");
890 if (value) {
891 int flags = 0;
892 bool writethrough;
894 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
895 error_report("invalid cache option");
896 return NULL;
899 /* Specific options take precedence */
900 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
901 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
902 !writethrough, &error_abort);
904 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
905 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
906 !!(flags & BDRV_O_NOCACHE), &error_abort);
908 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
909 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
910 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
912 qemu_opt_unset(all_opts, "cache");
915 /* Get a QDict for processing the options */
916 bs_opts = qdict_new();
917 qemu_opts_to_qdict(all_opts, bs_opts);
919 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
920 &error_abort);
921 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
922 if (local_err) {
923 error_report_err(local_err);
924 goto fail;
927 /* Deprecated option boot=[on|off] */
928 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
929 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
930 "ignored. Future versions will reject this parameter. Please "
931 "update your scripts.\n");
934 /* Media type */
935 value = qemu_opt_get(legacy_opts, "media");
936 if (value) {
937 if (!strcmp(value, "disk")) {
938 media = MEDIA_DISK;
939 } else if (!strcmp(value, "cdrom")) {
940 media = MEDIA_CDROM;
941 read_only = true;
942 } else {
943 error_report("'%s' invalid media", value);
944 goto fail;
948 /* copy-on-read is disabled with a warning for read-only devices */
949 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
950 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
952 if (read_only && copy_on_read) {
953 error_report("warning: disabling copy-on-read on read-only drive");
954 copy_on_read = false;
957 qdict_put(bs_opts, BDRV_OPT_READ_ONLY,
958 qstring_from_str(read_only ? "on" : "off"));
959 qdict_put(bs_opts, "copy-on-read",
960 qstring_from_str(copy_on_read ? "on" :"off"));
962 /* Controller type */
963 value = qemu_opt_get(legacy_opts, "if");
964 if (value) {
965 for (type = 0;
966 type < IF_COUNT && strcmp(value, if_name[type]);
967 type++) {
969 if (type == IF_COUNT) {
970 error_report("unsupported bus type '%s'", value);
971 goto fail;
973 } else {
974 type = block_default_type;
977 /* Geometry */
978 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
979 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
980 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
982 if (cyls || heads || secs) {
983 if (cyls < 1) {
984 error_report("invalid physical cyls number");
985 goto fail;
987 if (heads < 1) {
988 error_report("invalid physical heads number");
989 goto fail;
991 if (secs < 1) {
992 error_report("invalid physical secs number");
993 goto fail;
997 translation = BIOS_ATA_TRANSLATION_AUTO;
998 value = qemu_opt_get(legacy_opts, "trans");
999 if (value != NULL) {
1000 if (!cyls) {
1001 error_report("'%s' trans must be used with cyls, heads and secs",
1002 value);
1003 goto fail;
1005 if (!strcmp(value, "none")) {
1006 translation = BIOS_ATA_TRANSLATION_NONE;
1007 } else if (!strcmp(value, "lba")) {
1008 translation = BIOS_ATA_TRANSLATION_LBA;
1009 } else if (!strcmp(value, "large")) {
1010 translation = BIOS_ATA_TRANSLATION_LARGE;
1011 } else if (!strcmp(value, "rechs")) {
1012 translation = BIOS_ATA_TRANSLATION_RECHS;
1013 } else if (!strcmp(value, "auto")) {
1014 translation = BIOS_ATA_TRANSLATION_AUTO;
1015 } else {
1016 error_report("'%s' invalid translation type", value);
1017 goto fail;
1021 if (media == MEDIA_CDROM) {
1022 if (cyls || secs || heads) {
1023 error_report("CHS can't be set with media=cdrom");
1024 goto fail;
1028 /* Device address specified by bus/unit or index.
1029 * If none was specified, try to find the first free one. */
1030 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
1031 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
1032 index = qemu_opt_get_number(legacy_opts, "index", -1);
1034 max_devs = if_max_devs[type];
1036 if (index != -1) {
1037 if (bus_id != 0 || unit_id != -1) {
1038 error_report("index cannot be used with bus and unit");
1039 goto fail;
1041 bus_id = drive_index_to_bus_id(type, index);
1042 unit_id = drive_index_to_unit_id(type, index);
1045 if (unit_id == -1) {
1046 unit_id = 0;
1047 while (drive_get(type, bus_id, unit_id) != NULL) {
1048 unit_id++;
1049 if (max_devs && unit_id >= max_devs) {
1050 unit_id -= max_devs;
1051 bus_id++;
1056 if (max_devs && unit_id >= max_devs) {
1057 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
1058 goto fail;
1061 if (drive_get(type, bus_id, unit_id) != NULL) {
1062 error_report("drive with bus=%d, unit=%d (index=%d) exists",
1063 bus_id, unit_id, index);
1064 goto fail;
1067 /* Serial number */
1068 serial = qemu_opt_get(legacy_opts, "serial");
1070 /* no id supplied -> create one */
1071 if (qemu_opts_id(all_opts) == NULL) {
1072 char *new_id;
1073 const char *mediastr = "";
1074 if (type == IF_IDE || type == IF_SCSI) {
1075 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
1077 if (max_devs) {
1078 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
1079 mediastr, unit_id);
1080 } else {
1081 new_id = g_strdup_printf("%s%s%i", if_name[type],
1082 mediastr, unit_id);
1084 qdict_put(bs_opts, "id", qstring_from_str(new_id));
1085 g_free(new_id);
1088 /* Add virtio block device */
1089 devaddr = qemu_opt_get(legacy_opts, "addr");
1090 if (devaddr && type != IF_VIRTIO) {
1091 error_report("addr is not supported by this bus type");
1092 goto fail;
1095 if (type == IF_VIRTIO) {
1096 QemuOpts *devopts;
1097 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
1098 &error_abort);
1099 if (arch_type == QEMU_ARCH_S390X) {
1100 qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
1101 } else {
1102 qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
1104 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
1105 &error_abort);
1106 if (devaddr) {
1107 qemu_opt_set(devopts, "addr", devaddr, &error_abort);
1111 filename = qemu_opt_get(legacy_opts, "file");
1113 /* Check werror/rerror compatibility with if=... */
1114 werror = qemu_opt_get(legacy_opts, "werror");
1115 if (werror != NULL) {
1116 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
1117 type != IF_NONE) {
1118 error_report("werror is not supported by this bus type");
1119 goto fail;
1121 qdict_put(bs_opts, "werror", qstring_from_str(werror));
1124 rerror = qemu_opt_get(legacy_opts, "rerror");
1125 if (rerror != NULL) {
1126 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
1127 type != IF_NONE) {
1128 error_report("rerror is not supported by this bus type");
1129 goto fail;
1131 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
1134 /* Actual block device init: Functionality shared with blockdev-add */
1135 blk = blockdev_init(filename, bs_opts, &local_err);
1136 bs_opts = NULL;
1137 if (!blk) {
1138 if (local_err) {
1139 error_report_err(local_err);
1141 goto fail;
1142 } else {
1143 assert(!local_err);
1146 /* Create legacy DriveInfo */
1147 dinfo = g_malloc0(sizeof(*dinfo));
1148 dinfo->opts = all_opts;
1150 dinfo->cyls = cyls;
1151 dinfo->heads = heads;
1152 dinfo->secs = secs;
1153 dinfo->trans = translation;
1155 dinfo->type = type;
1156 dinfo->bus = bus_id;
1157 dinfo->unit = unit_id;
1158 dinfo->devaddr = devaddr;
1159 dinfo->serial = g_strdup(serial);
1161 blk_set_legacy_dinfo(blk, dinfo);
1163 switch(type) {
1164 case IF_IDE:
1165 case IF_SCSI:
1166 case IF_XEN:
1167 case IF_NONE:
1168 dinfo->media_cd = media == MEDIA_CDROM;
1169 break;
1170 default:
1171 break;
1174 fail:
1175 qemu_opts_del(legacy_opts);
1176 QDECREF(bs_opts);
1177 return dinfo;
1180 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1182 BlockDriverState *bs;
1184 bs = bdrv_lookup_bs(name, name, errp);
1185 if (bs == NULL) {
1186 return NULL;
1189 if (!bdrv_is_root_node(bs)) {
1190 error_setg(errp, "Need a root block node");
1191 return NULL;
1194 if (!bdrv_is_inserted(bs)) {
1195 error_setg(errp, "Device has no medium");
1196 return NULL;
1199 return bs;
1202 static BlockBackend *qmp_get_blk(const char *blk_name, const char *qdev_id,
1203 Error **errp)
1205 BlockBackend *blk;
1207 if (!blk_name == !qdev_id) {
1208 error_setg(errp, "Need exactly one of 'device' and 'id'");
1209 return NULL;
1212 if (qdev_id) {
1213 blk = blk_by_qdev_id(qdev_id, errp);
1214 } else {
1215 blk = blk_by_name(blk_name);
1216 if (blk == NULL) {
1217 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1218 "Device '%s' not found", blk_name);
1222 return blk;
1225 void hmp_commit(Monitor *mon, const QDict *qdict)
1227 const char *device = qdict_get_str(qdict, "device");
1228 BlockBackend *blk;
1229 int ret;
1231 if (!strcmp(device, "all")) {
1232 ret = blk_commit_all();
1233 } else {
1234 BlockDriverState *bs;
1235 AioContext *aio_context;
1237 blk = blk_by_name(device);
1238 if (!blk) {
1239 monitor_printf(mon, "Device '%s' not found\n", device);
1240 return;
1242 if (!blk_is_available(blk)) {
1243 monitor_printf(mon, "Device '%s' has no medium\n", device);
1244 return;
1247 bs = blk_bs(blk);
1248 aio_context = bdrv_get_aio_context(bs);
1249 aio_context_acquire(aio_context);
1251 ret = bdrv_commit(bs);
1253 aio_context_release(aio_context);
1255 if (ret < 0) {
1256 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1257 strerror(-ret));
1261 static void blockdev_do_action(TransactionAction *action, Error **errp)
1263 TransactionActionList list;
1265 list.value = action;
1266 list.next = NULL;
1267 qmp_transaction(&list, false, NULL, errp);
1270 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1271 bool has_node_name, const char *node_name,
1272 const char *snapshot_file,
1273 bool has_snapshot_node_name,
1274 const char *snapshot_node_name,
1275 bool has_format, const char *format,
1276 bool has_mode, NewImageMode mode, Error **errp)
1278 BlockdevSnapshotSync snapshot = {
1279 .has_device = has_device,
1280 .device = (char *) device,
1281 .has_node_name = has_node_name,
1282 .node_name = (char *) node_name,
1283 .snapshot_file = (char *) snapshot_file,
1284 .has_snapshot_node_name = has_snapshot_node_name,
1285 .snapshot_node_name = (char *) snapshot_node_name,
1286 .has_format = has_format,
1287 .format = (char *) format,
1288 .has_mode = has_mode,
1289 .mode = mode,
1291 TransactionAction action = {
1292 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1293 .u.blockdev_snapshot_sync.data = &snapshot,
1295 blockdev_do_action(&action, errp);
1298 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1299 Error **errp)
1301 BlockdevSnapshot snapshot_data = {
1302 .node = (char *) node,
1303 .overlay = (char *) overlay
1305 TransactionAction action = {
1306 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1307 .u.blockdev_snapshot.data = &snapshot_data,
1309 blockdev_do_action(&action, errp);
1312 void qmp_blockdev_snapshot_internal_sync(const char *device,
1313 const char *name,
1314 Error **errp)
1316 BlockdevSnapshotInternal snapshot = {
1317 .device = (char *) device,
1318 .name = (char *) name
1320 TransactionAction action = {
1321 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1322 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1324 blockdev_do_action(&action, errp);
1327 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1328 bool has_id,
1329 const char *id,
1330 bool has_name,
1331 const char *name,
1332 Error **errp)
1334 BlockDriverState *bs;
1335 AioContext *aio_context;
1336 QEMUSnapshotInfo sn;
1337 Error *local_err = NULL;
1338 SnapshotInfo *info = NULL;
1339 int ret;
1341 bs = qmp_get_root_bs(device, errp);
1342 if (!bs) {
1343 return NULL;
1345 aio_context = bdrv_get_aio_context(bs);
1346 aio_context_acquire(aio_context);
1348 if (!has_id) {
1349 id = NULL;
1352 if (!has_name) {
1353 name = NULL;
1356 if (!id && !name) {
1357 error_setg(errp, "Name or id must be provided");
1358 goto out_aio_context;
1361 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1362 goto out_aio_context;
1365 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1366 if (local_err) {
1367 error_propagate(errp, local_err);
1368 goto out_aio_context;
1370 if (!ret) {
1371 error_setg(errp,
1372 "Snapshot with id '%s' and name '%s' does not exist on "
1373 "device '%s'",
1374 STR_OR_NULL(id), STR_OR_NULL(name), device);
1375 goto out_aio_context;
1378 bdrv_snapshot_delete(bs, id, name, &local_err);
1379 if (local_err) {
1380 error_propagate(errp, local_err);
1381 goto out_aio_context;
1384 aio_context_release(aio_context);
1386 info = g_new0(SnapshotInfo, 1);
1387 info->id = g_strdup(sn.id_str);
1388 info->name = g_strdup(sn.name);
1389 info->date_nsec = sn.date_nsec;
1390 info->date_sec = sn.date_sec;
1391 info->vm_state_size = sn.vm_state_size;
1392 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1393 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1395 return info;
1397 out_aio_context:
1398 aio_context_release(aio_context);
1399 return NULL;
1403 * block_dirty_bitmap_lookup:
1404 * Return a dirty bitmap (if present), after validating
1405 * the node reference and bitmap names.
1407 * @node: The name of the BDS node to search for bitmaps
1408 * @name: The name of the bitmap to search for
1409 * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1410 * @paio: Output pointer for aio_context acquisition, if desired. Can be NULL.
1411 * @errp: Output pointer for error information. Can be NULL.
1413 * @return: A bitmap object on success, or NULL on failure.
1415 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1416 const char *name,
1417 BlockDriverState **pbs,
1418 AioContext **paio,
1419 Error **errp)
1421 BlockDriverState *bs;
1422 BdrvDirtyBitmap *bitmap;
1423 AioContext *aio_context;
1425 if (!node) {
1426 error_setg(errp, "Node cannot be NULL");
1427 return NULL;
1429 if (!name) {
1430 error_setg(errp, "Bitmap name cannot be NULL");
1431 return NULL;
1433 bs = bdrv_lookup_bs(node, node, NULL);
1434 if (!bs) {
1435 error_setg(errp, "Node '%s' not found", node);
1436 return NULL;
1439 aio_context = bdrv_get_aio_context(bs);
1440 aio_context_acquire(aio_context);
1442 bitmap = bdrv_find_dirty_bitmap(bs, name);
1443 if (!bitmap) {
1444 error_setg(errp, "Dirty bitmap '%s' not found", name);
1445 goto fail;
1448 if (pbs) {
1449 *pbs = bs;
1451 if (paio) {
1452 *paio = aio_context;
1453 } else {
1454 aio_context_release(aio_context);
1457 return bitmap;
1459 fail:
1460 aio_context_release(aio_context);
1461 return NULL;
1464 /* New and old BlockDriverState structs for atomic group operations */
1466 typedef struct BlkActionState BlkActionState;
1469 * BlkActionOps:
1470 * Table of operations that define an Action.
1472 * @instance_size: Size of state struct, in bytes.
1473 * @prepare: Prepare the work, must NOT be NULL.
1474 * @commit: Commit the changes, can be NULL.
1475 * @abort: Abort the changes on fail, can be NULL.
1476 * @clean: Clean up resources after all transaction actions have called
1477 * commit() or abort(). Can be NULL.
1479 * Only prepare() may fail. In a single transaction, only one of commit() or
1480 * abort() will be called. clean() will always be called if it is present.
1482 typedef struct BlkActionOps {
1483 size_t instance_size;
1484 void (*prepare)(BlkActionState *common, Error **errp);
1485 void (*commit)(BlkActionState *common);
1486 void (*abort)(BlkActionState *common);
1487 void (*clean)(BlkActionState *common);
1488 } BlkActionOps;
1491 * BlkActionState:
1492 * Describes one Action's state within a Transaction.
1494 * @action: QAPI-defined enum identifying which Action to perform.
1495 * @ops: Table of ActionOps this Action can perform.
1496 * @block_job_txn: Transaction which this action belongs to.
1497 * @entry: List membership for all Actions in this Transaction.
1499 * This structure must be arranged as first member in a subclassed type,
1500 * assuming that the compiler will also arrange it to the same offsets as the
1501 * base class.
1503 struct BlkActionState {
1504 TransactionAction *action;
1505 const BlkActionOps *ops;
1506 BlockJobTxn *block_job_txn;
1507 TransactionProperties *txn_props;
1508 QSIMPLEQ_ENTRY(BlkActionState) entry;
1511 /* internal snapshot private data */
1512 typedef struct InternalSnapshotState {
1513 BlkActionState common;
1514 BlockDriverState *bs;
1515 AioContext *aio_context;
1516 QEMUSnapshotInfo sn;
1517 bool created;
1518 } InternalSnapshotState;
1521 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1523 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1524 error_setg(errp,
1525 "Action '%s' does not support Transaction property "
1526 "completion-mode = %s",
1527 TransactionActionKind_lookup[s->action->type],
1528 ActionCompletionMode_lookup[s->txn_props->completion_mode]);
1529 return -1;
1531 return 0;
1534 static void internal_snapshot_prepare(BlkActionState *common,
1535 Error **errp)
1537 Error *local_err = NULL;
1538 const char *device;
1539 const char *name;
1540 BlockDriverState *bs;
1541 QEMUSnapshotInfo old_sn, *sn;
1542 bool ret;
1543 qemu_timeval tv;
1544 BlockdevSnapshotInternal *internal;
1545 InternalSnapshotState *state;
1546 int ret1;
1548 g_assert(common->action->type ==
1549 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1550 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1551 state = DO_UPCAST(InternalSnapshotState, common, common);
1553 /* 1. parse input */
1554 device = internal->device;
1555 name = internal->name;
1557 /* 2. check for validation */
1558 if (action_check_completion_mode(common, errp) < 0) {
1559 return;
1562 bs = qmp_get_root_bs(device, errp);
1563 if (!bs) {
1564 return;
1567 /* AioContext is released in .clean() */
1568 state->aio_context = bdrv_get_aio_context(bs);
1569 aio_context_acquire(state->aio_context);
1571 state->bs = bs;
1572 bdrv_drained_begin(bs);
1574 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1575 return;
1578 if (bdrv_is_read_only(bs)) {
1579 error_setg(errp, "Device '%s' is read only", device);
1580 return;
1583 if (!bdrv_can_snapshot(bs)) {
1584 error_setg(errp, "Block format '%s' used by device '%s' "
1585 "does not support internal snapshots",
1586 bs->drv->format_name, device);
1587 return;
1590 if (!strlen(name)) {
1591 error_setg(errp, "Name is empty");
1592 return;
1595 /* check whether a snapshot with name exist */
1596 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1597 &local_err);
1598 if (local_err) {
1599 error_propagate(errp, local_err);
1600 return;
1601 } else if (ret) {
1602 error_setg(errp,
1603 "Snapshot with name '%s' already exists on device '%s'",
1604 name, device);
1605 return;
1608 /* 3. take the snapshot */
1609 sn = &state->sn;
1610 pstrcpy(sn->name, sizeof(sn->name), name);
1611 qemu_gettimeofday(&tv);
1612 sn->date_sec = tv.tv_sec;
1613 sn->date_nsec = tv.tv_usec * 1000;
1614 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1616 ret1 = bdrv_snapshot_create(bs, sn);
1617 if (ret1 < 0) {
1618 error_setg_errno(errp, -ret1,
1619 "Failed to create snapshot '%s' on device '%s'",
1620 name, device);
1621 return;
1624 /* 4. succeed, mark a snapshot is created */
1625 state->created = true;
1628 static void internal_snapshot_abort(BlkActionState *common)
1630 InternalSnapshotState *state =
1631 DO_UPCAST(InternalSnapshotState, common, common);
1632 BlockDriverState *bs = state->bs;
1633 QEMUSnapshotInfo *sn = &state->sn;
1634 Error *local_error = NULL;
1636 if (!state->created) {
1637 return;
1640 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1641 error_reportf_err(local_error,
1642 "Failed to delete snapshot with id '%s' and "
1643 "name '%s' on device '%s' in abort: ",
1644 sn->id_str, sn->name,
1645 bdrv_get_device_name(bs));
1649 static void internal_snapshot_clean(BlkActionState *common)
1651 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1652 common, common);
1654 if (state->aio_context) {
1655 if (state->bs) {
1656 bdrv_drained_end(state->bs);
1658 aio_context_release(state->aio_context);
1662 /* external snapshot private data */
1663 typedef struct ExternalSnapshotState {
1664 BlkActionState common;
1665 BlockDriverState *old_bs;
1666 BlockDriverState *new_bs;
1667 AioContext *aio_context;
1668 } ExternalSnapshotState;
1670 static void external_snapshot_prepare(BlkActionState *common,
1671 Error **errp)
1673 int flags = 0;
1674 QDict *options = NULL;
1675 Error *local_err = NULL;
1676 /* Device and node name of the image to generate the snapshot from */
1677 const char *device;
1678 const char *node_name;
1679 /* Reference to the new image (for 'blockdev-snapshot') */
1680 const char *snapshot_ref;
1681 /* File name of the new image (for 'blockdev-snapshot-sync') */
1682 const char *new_image_file;
1683 ExternalSnapshotState *state =
1684 DO_UPCAST(ExternalSnapshotState, common, common);
1685 TransactionAction *action = common->action;
1687 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1688 * purpose but a different set of parameters */
1689 switch (action->type) {
1690 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1692 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1693 device = s->node;
1694 node_name = s->node;
1695 new_image_file = NULL;
1696 snapshot_ref = s->overlay;
1698 break;
1699 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1701 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1702 device = s->has_device ? s->device : NULL;
1703 node_name = s->has_node_name ? s->node_name : NULL;
1704 new_image_file = s->snapshot_file;
1705 snapshot_ref = NULL;
1707 break;
1708 default:
1709 g_assert_not_reached();
1712 /* start processing */
1713 if (action_check_completion_mode(common, errp) < 0) {
1714 return;
1717 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1718 if (!state->old_bs) {
1719 return;
1722 /* Acquire AioContext now so any threads operating on old_bs stop */
1723 state->aio_context = bdrv_get_aio_context(state->old_bs);
1724 aio_context_acquire(state->aio_context);
1725 bdrv_drained_begin(state->old_bs);
1727 if (!bdrv_is_inserted(state->old_bs)) {
1728 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1729 return;
1732 if (bdrv_op_is_blocked(state->old_bs,
1733 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1734 return;
1737 if (!bdrv_is_read_only(state->old_bs)) {
1738 if (bdrv_flush(state->old_bs)) {
1739 error_setg(errp, QERR_IO_ERROR);
1740 return;
1744 if (!bdrv_is_first_non_filter(state->old_bs)) {
1745 error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1746 return;
1749 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1750 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1751 const char *format = s->has_format ? s->format : "qcow2";
1752 enum NewImageMode mode;
1753 const char *snapshot_node_name =
1754 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1756 if (node_name && !snapshot_node_name) {
1757 error_setg(errp, "New snapshot node name missing");
1758 return;
1761 if (snapshot_node_name &&
1762 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1763 error_setg(errp, "New snapshot node name already in use");
1764 return;
1767 flags = state->old_bs->open_flags;
1768 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ);
1770 /* create new image w/backing file */
1771 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1772 if (mode != NEW_IMAGE_MODE_EXISTING) {
1773 int64_t size = bdrv_getlength(state->old_bs);
1774 if (size < 0) {
1775 error_setg_errno(errp, -size, "bdrv_getlength failed");
1776 return;
1778 bdrv_img_create(new_image_file, format,
1779 state->old_bs->filename,
1780 state->old_bs->drv->format_name,
1781 NULL, size, flags, &local_err, false);
1782 if (local_err) {
1783 error_propagate(errp, local_err);
1784 return;
1788 options = qdict_new();
1789 if (s->has_snapshot_node_name) {
1790 qdict_put(options, "node-name",
1791 qstring_from_str(snapshot_node_name));
1793 qdict_put(options, "driver", qstring_from_str(format));
1795 flags |= BDRV_O_NO_BACKING;
1798 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1799 errp);
1800 /* We will manually add the backing_hd field to the bs later */
1801 if (!state->new_bs) {
1802 return;
1805 if (bdrv_has_blk(state->new_bs)) {
1806 error_setg(errp, "The snapshot is already in use");
1807 return;
1810 if (bdrv_op_is_blocked(state->new_bs, BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
1811 errp)) {
1812 return;
1815 if (state->new_bs->backing != NULL) {
1816 error_setg(errp, "The snapshot already has a backing image");
1817 return;
1820 if (!state->new_bs->drv->supports_backing) {
1821 error_setg(errp, "The snapshot does not support backing images");
1825 static void external_snapshot_commit(BlkActionState *common)
1827 ExternalSnapshotState *state =
1828 DO_UPCAST(ExternalSnapshotState, common, common);
1830 bdrv_set_aio_context(state->new_bs, state->aio_context);
1832 /* This removes our old bs and adds the new bs */
1833 bdrv_append(state->new_bs, state->old_bs);
1834 /* We don't need (or want) to use the transactional
1835 * bdrv_reopen_multiple() across all the entries at once, because we
1836 * don't want to abort all of them if one of them fails the reopen */
1837 if (!state->old_bs->copy_on_read) {
1838 bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR,
1839 NULL);
1843 static void external_snapshot_abort(BlkActionState *common)
1845 ExternalSnapshotState *state =
1846 DO_UPCAST(ExternalSnapshotState, common, common);
1847 if (state->new_bs) {
1848 bdrv_unref(state->new_bs);
1852 static void external_snapshot_clean(BlkActionState *common)
1854 ExternalSnapshotState *state =
1855 DO_UPCAST(ExternalSnapshotState, common, common);
1856 if (state->aio_context) {
1857 bdrv_drained_end(state->old_bs);
1858 aio_context_release(state->aio_context);
1862 typedef struct DriveBackupState {
1863 BlkActionState common;
1864 BlockDriverState *bs;
1865 AioContext *aio_context;
1866 BlockJob *job;
1867 } DriveBackupState;
1869 static void do_drive_backup(DriveBackup *backup, BlockJobTxn *txn,
1870 Error **errp);
1872 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1874 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1875 BlockDriverState *bs;
1876 DriveBackup *backup;
1877 Error *local_err = NULL;
1879 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1880 backup = common->action->u.drive_backup.data;
1882 bs = qmp_get_root_bs(backup->device, errp);
1883 if (!bs) {
1884 return;
1887 /* AioContext is released in .clean() */
1888 state->aio_context = bdrv_get_aio_context(bs);
1889 aio_context_acquire(state->aio_context);
1890 bdrv_drained_begin(bs);
1891 state->bs = bs;
1893 do_drive_backup(backup, common->block_job_txn, &local_err);
1894 if (local_err) {
1895 error_propagate(errp, local_err);
1896 return;
1899 state->job = state->bs->job;
1902 static void drive_backup_abort(BlkActionState *common)
1904 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1905 BlockDriverState *bs = state->bs;
1907 /* Only cancel if it's the job we started */
1908 if (bs && bs->job && bs->job == state->job) {
1909 block_job_cancel_sync(bs->job);
1913 static void drive_backup_clean(BlkActionState *common)
1915 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1917 if (state->aio_context) {
1918 bdrv_drained_end(state->bs);
1919 aio_context_release(state->aio_context);
1923 typedef struct BlockdevBackupState {
1924 BlkActionState common;
1925 BlockDriverState *bs;
1926 BlockJob *job;
1927 AioContext *aio_context;
1928 } BlockdevBackupState;
1930 static void do_blockdev_backup(BlockdevBackup *backup, BlockJobTxn *txn,
1931 Error **errp);
1933 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1935 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1936 BlockdevBackup *backup;
1937 BlockDriverState *bs, *target;
1938 Error *local_err = NULL;
1940 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1941 backup = common->action->u.blockdev_backup.data;
1943 bs = qmp_get_root_bs(backup->device, errp);
1944 if (!bs) {
1945 return;
1948 target = bdrv_lookup_bs(backup->target, backup->target, errp);
1949 if (!target) {
1950 return;
1953 /* AioContext is released in .clean() */
1954 state->aio_context = bdrv_get_aio_context(bs);
1955 if (state->aio_context != bdrv_get_aio_context(target)) {
1956 state->aio_context = NULL;
1957 error_setg(errp, "Backup between two IO threads is not implemented");
1958 return;
1960 aio_context_acquire(state->aio_context);
1961 state->bs = bs;
1962 bdrv_drained_begin(state->bs);
1964 do_blockdev_backup(backup, common->block_job_txn, &local_err);
1965 if (local_err) {
1966 error_propagate(errp, local_err);
1967 return;
1970 state->job = state->bs->job;
1973 static void blockdev_backup_abort(BlkActionState *common)
1975 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1976 BlockDriverState *bs = state->bs;
1978 /* Only cancel if it's the job we started */
1979 if (bs && bs->job && bs->job == state->job) {
1980 block_job_cancel_sync(bs->job);
1984 static void blockdev_backup_clean(BlkActionState *common)
1986 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1988 if (state->aio_context) {
1989 bdrv_drained_end(state->bs);
1990 aio_context_release(state->aio_context);
1994 typedef struct BlockDirtyBitmapState {
1995 BlkActionState common;
1996 BdrvDirtyBitmap *bitmap;
1997 BlockDriverState *bs;
1998 AioContext *aio_context;
1999 HBitmap *backup;
2000 bool prepared;
2001 } BlockDirtyBitmapState;
2003 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
2004 Error **errp)
2006 Error *local_err = NULL;
2007 BlockDirtyBitmapAdd *action;
2008 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2009 common, common);
2011 if (action_check_completion_mode(common, errp) < 0) {
2012 return;
2015 action = common->action->u.block_dirty_bitmap_add.data;
2016 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2017 qmp_block_dirty_bitmap_add(action->node, action->name,
2018 action->has_granularity, action->granularity,
2019 &local_err);
2021 if (!local_err) {
2022 state->prepared = true;
2023 } else {
2024 error_propagate(errp, local_err);
2028 static void block_dirty_bitmap_add_abort(BlkActionState *common)
2030 BlockDirtyBitmapAdd *action;
2031 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2032 common, common);
2034 action = common->action->u.block_dirty_bitmap_add.data;
2035 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2036 * then the node reference and bitmap name must have been valid.
2038 if (state->prepared) {
2039 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2043 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2044 Error **errp)
2046 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2047 common, common);
2048 BlockDirtyBitmap *action;
2050 if (action_check_completion_mode(common, errp) < 0) {
2051 return;
2054 action = common->action->u.block_dirty_bitmap_clear.data;
2055 state->bitmap = block_dirty_bitmap_lookup(action->node,
2056 action->name,
2057 &state->bs,
2058 &state->aio_context,
2059 errp);
2060 if (!state->bitmap) {
2061 return;
2064 if (bdrv_dirty_bitmap_frozen(state->bitmap)) {
2065 error_setg(errp, "Cannot modify a frozen bitmap");
2066 return;
2067 } else if (!bdrv_dirty_bitmap_enabled(state->bitmap)) {
2068 error_setg(errp, "Cannot clear a disabled bitmap");
2069 return;
2072 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2073 /* AioContext is released in .clean() */
2076 static void block_dirty_bitmap_clear_abort(BlkActionState *common)
2078 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2079 common, common);
2081 bdrv_undo_clear_dirty_bitmap(state->bitmap, state->backup);
2084 static void block_dirty_bitmap_clear_commit(BlkActionState *common)
2086 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2087 common, common);
2089 hbitmap_free(state->backup);
2092 static void block_dirty_bitmap_clear_clean(BlkActionState *common)
2094 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2095 common, common);
2097 if (state->aio_context) {
2098 aio_context_release(state->aio_context);
2102 static void abort_prepare(BlkActionState *common, Error **errp)
2104 error_setg(errp, "Transaction aborted using Abort action");
2107 static void abort_commit(BlkActionState *common)
2109 g_assert_not_reached(); /* this action never succeeds */
2112 static const BlkActionOps actions[] = {
2113 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2114 .instance_size = sizeof(ExternalSnapshotState),
2115 .prepare = external_snapshot_prepare,
2116 .commit = external_snapshot_commit,
2117 .abort = external_snapshot_abort,
2118 .clean = external_snapshot_clean,
2120 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2121 .instance_size = sizeof(ExternalSnapshotState),
2122 .prepare = external_snapshot_prepare,
2123 .commit = external_snapshot_commit,
2124 .abort = external_snapshot_abort,
2125 .clean = external_snapshot_clean,
2127 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2128 .instance_size = sizeof(DriveBackupState),
2129 .prepare = drive_backup_prepare,
2130 .abort = drive_backup_abort,
2131 .clean = drive_backup_clean,
2133 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2134 .instance_size = sizeof(BlockdevBackupState),
2135 .prepare = blockdev_backup_prepare,
2136 .abort = blockdev_backup_abort,
2137 .clean = blockdev_backup_clean,
2139 [TRANSACTION_ACTION_KIND_ABORT] = {
2140 .instance_size = sizeof(BlkActionState),
2141 .prepare = abort_prepare,
2142 .commit = abort_commit,
2144 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2145 .instance_size = sizeof(InternalSnapshotState),
2146 .prepare = internal_snapshot_prepare,
2147 .abort = internal_snapshot_abort,
2148 .clean = internal_snapshot_clean,
2150 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2151 .instance_size = sizeof(BlockDirtyBitmapState),
2152 .prepare = block_dirty_bitmap_add_prepare,
2153 .abort = block_dirty_bitmap_add_abort,
2155 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2156 .instance_size = sizeof(BlockDirtyBitmapState),
2157 .prepare = block_dirty_bitmap_clear_prepare,
2158 .commit = block_dirty_bitmap_clear_commit,
2159 .abort = block_dirty_bitmap_clear_abort,
2160 .clean = block_dirty_bitmap_clear_clean,
2165 * Allocate a TransactionProperties structure if necessary, and fill
2166 * that structure with desired defaults if they are unset.
2168 static TransactionProperties *get_transaction_properties(
2169 TransactionProperties *props)
2171 if (!props) {
2172 props = g_new0(TransactionProperties, 1);
2175 if (!props->has_completion_mode) {
2176 props->has_completion_mode = true;
2177 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2180 return props;
2184 * 'Atomic' group operations. The operations are performed as a set, and if
2185 * any fail then we roll back all operations in the group.
2187 void qmp_transaction(TransactionActionList *dev_list,
2188 bool has_props,
2189 struct TransactionProperties *props,
2190 Error **errp)
2192 TransactionActionList *dev_entry = dev_list;
2193 BlockJobTxn *block_job_txn = NULL;
2194 BlkActionState *state, *next;
2195 Error *local_err = NULL;
2197 QSIMPLEQ_HEAD(snap_bdrv_states, BlkActionState) snap_bdrv_states;
2198 QSIMPLEQ_INIT(&snap_bdrv_states);
2200 /* Does this transaction get canceled as a group on failure?
2201 * If not, we don't really need to make a BlockJobTxn.
2203 props = get_transaction_properties(props);
2204 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2205 block_job_txn = block_job_txn_new();
2208 /* drain all i/o before any operations */
2209 bdrv_drain_all();
2211 /* We don't do anything in this loop that commits us to the operations */
2212 while (NULL != dev_entry) {
2213 TransactionAction *dev_info = NULL;
2214 const BlkActionOps *ops;
2216 dev_info = dev_entry->value;
2217 dev_entry = dev_entry->next;
2219 assert(dev_info->type < ARRAY_SIZE(actions));
2221 ops = &actions[dev_info->type];
2222 assert(ops->instance_size > 0);
2224 state = g_malloc0(ops->instance_size);
2225 state->ops = ops;
2226 state->action = dev_info;
2227 state->block_job_txn = block_job_txn;
2228 state->txn_props = props;
2229 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2231 state->ops->prepare(state, &local_err);
2232 if (local_err) {
2233 error_propagate(errp, local_err);
2234 goto delete_and_fail;
2238 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
2239 if (state->ops->commit) {
2240 state->ops->commit(state);
2244 /* success */
2245 goto exit;
2247 delete_and_fail:
2248 /* failure, and it is all-or-none; roll back all operations */
2249 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
2250 if (state->ops->abort) {
2251 state->ops->abort(state);
2254 exit:
2255 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2256 if (state->ops->clean) {
2257 state->ops->clean(state);
2259 g_free(state);
2261 if (!has_props) {
2262 qapi_free_TransactionProperties(props);
2264 block_job_txn_unref(block_job_txn);
2267 void qmp_eject(bool has_device, const char *device,
2268 bool has_id, const char *id,
2269 bool has_force, bool force, Error **errp)
2271 Error *local_err = NULL;
2272 int rc;
2274 if (!has_force) {
2275 force = false;
2278 rc = do_open_tray(has_device ? device : NULL,
2279 has_id ? id : NULL,
2280 force, &local_err);
2281 if (rc && rc != -ENOSYS) {
2282 error_propagate(errp, local_err);
2283 return;
2285 error_free(local_err);
2287 qmp_x_blockdev_remove_medium(has_device, device, has_id, id, errp);
2290 void qmp_block_passwd(bool has_device, const char *device,
2291 bool has_node_name, const char *node_name,
2292 const char *password, Error **errp)
2294 Error *local_err = NULL;
2295 BlockDriverState *bs;
2296 AioContext *aio_context;
2298 bs = bdrv_lookup_bs(has_device ? device : NULL,
2299 has_node_name ? node_name : NULL,
2300 &local_err);
2301 if (local_err) {
2302 error_propagate(errp, local_err);
2303 return;
2306 aio_context = bdrv_get_aio_context(bs);
2307 aio_context_acquire(aio_context);
2309 bdrv_add_key(bs, password, errp);
2311 aio_context_release(aio_context);
2315 * Attempt to open the tray of @device.
2316 * If @force, ignore its tray lock.
2317 * Else, if the tray is locked, don't open it, but ask the guest to open it.
2318 * On error, store an error through @errp and return -errno.
2319 * If @device does not exist, return -ENODEV.
2320 * If it has no removable media, return -ENOTSUP.
2321 * If it has no tray, return -ENOSYS.
2322 * If the guest was asked to open the tray, return -EINPROGRESS.
2323 * Else, return 0.
2325 static int do_open_tray(const char *blk_name, const char *qdev_id,
2326 bool force, Error **errp)
2328 BlockBackend *blk;
2329 const char *device = qdev_id ?: blk_name;
2330 bool locked;
2332 blk = qmp_get_blk(blk_name, qdev_id, errp);
2333 if (!blk) {
2334 return -ENODEV;
2337 if (!blk_dev_has_removable_media(blk)) {
2338 error_setg(errp, "Device '%s' is not removable", device);
2339 return -ENOTSUP;
2342 if (!blk_dev_has_tray(blk)) {
2343 error_setg(errp, "Device '%s' does not have a tray", device);
2344 return -ENOSYS;
2347 if (blk_dev_is_tray_open(blk)) {
2348 return 0;
2351 locked = blk_dev_is_medium_locked(blk);
2352 if (locked) {
2353 blk_dev_eject_request(blk, force);
2356 if (!locked || force) {
2357 blk_dev_change_media_cb(blk, false);
2360 if (locked && !force) {
2361 error_setg(errp, "Device '%s' is locked and force was not specified, "
2362 "wait for tray to open and try again", device);
2363 return -EINPROGRESS;
2366 return 0;
2369 void qmp_blockdev_open_tray(bool has_device, const char *device,
2370 bool has_id, const char *id,
2371 bool has_force, bool force,
2372 Error **errp)
2374 Error *local_err = NULL;
2375 int rc;
2377 if (!has_force) {
2378 force = false;
2380 rc = do_open_tray(has_device ? device : NULL,
2381 has_id ? id : NULL,
2382 force, &local_err);
2383 if (rc && rc != -ENOSYS && rc != -EINPROGRESS) {
2384 error_propagate(errp, local_err);
2385 return;
2387 error_free(local_err);
2390 void qmp_blockdev_close_tray(bool has_device, const char *device,
2391 bool has_id, const char *id,
2392 Error **errp)
2394 BlockBackend *blk;
2396 device = has_device ? device : NULL;
2397 id = has_id ? id : NULL;
2399 blk = qmp_get_blk(device, id, errp);
2400 if (!blk) {
2401 return;
2404 if (!blk_dev_has_removable_media(blk)) {
2405 error_setg(errp, "Device '%s' is not removable", device ?: id);
2406 return;
2409 if (!blk_dev_has_tray(blk)) {
2410 /* Ignore this command on tray-less devices */
2411 return;
2414 if (!blk_dev_is_tray_open(blk)) {
2415 return;
2418 blk_dev_change_media_cb(blk, true);
2421 void qmp_x_blockdev_remove_medium(bool has_device, const char *device,
2422 bool has_id, const char *id, Error **errp)
2424 BlockBackend *blk;
2425 BlockDriverState *bs;
2426 AioContext *aio_context;
2427 bool has_attached_device;
2429 device = has_device ? device : NULL;
2430 id = has_id ? id : NULL;
2432 blk = qmp_get_blk(device, id, errp);
2433 if (!blk) {
2434 return;
2437 /* For BBs without a device, we can exchange the BDS tree at will */
2438 has_attached_device = blk_get_attached_dev(blk);
2440 if (has_attached_device && !blk_dev_has_removable_media(blk)) {
2441 error_setg(errp, "Device '%s' is not removable", device ?: id);
2442 return;
2445 if (has_attached_device && blk_dev_has_tray(blk) &&
2446 !blk_dev_is_tray_open(blk))
2448 error_setg(errp, "Tray of device '%s' is not open", device ?: id);
2449 return;
2452 bs = blk_bs(blk);
2453 if (!bs) {
2454 return;
2457 aio_context = bdrv_get_aio_context(bs);
2458 aio_context_acquire(aio_context);
2460 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
2461 goto out;
2464 blk_remove_bs(blk);
2466 if (!blk_dev_has_tray(blk)) {
2467 /* For tray-less devices, blockdev-open-tray is a no-op (or may not be
2468 * called at all); therefore, the medium needs to be ejected here.
2469 * Do it after blk_remove_bs() so blk_is_inserted(blk) returns the @load
2470 * value passed here (i.e. false). */
2471 blk_dev_change_media_cb(blk, false);
2474 out:
2475 aio_context_release(aio_context);
2478 static void qmp_blockdev_insert_anon_medium(BlockBackend *blk,
2479 BlockDriverState *bs, Error **errp)
2481 bool has_device;
2483 /* For BBs without a device, we can exchange the BDS tree at will */
2484 has_device = blk_get_attached_dev(blk);
2486 if (has_device && !blk_dev_has_removable_media(blk)) {
2487 error_setg(errp, "Device is not removable");
2488 return;
2491 if (has_device && blk_dev_has_tray(blk) && !blk_dev_is_tray_open(blk)) {
2492 error_setg(errp, "Tray of the device is not open");
2493 return;
2496 if (blk_bs(blk)) {
2497 error_setg(errp, "There already is a medium in the device");
2498 return;
2501 blk_insert_bs(blk, bs);
2503 if (!blk_dev_has_tray(blk)) {
2504 /* For tray-less devices, blockdev-close-tray is a no-op (or may not be
2505 * called at all); therefore, the medium needs to be pushed into the
2506 * slot here.
2507 * Do it after blk_insert_bs() so blk_is_inserted(blk) returns the @load
2508 * value passed here (i.e. true). */
2509 blk_dev_change_media_cb(blk, true);
2513 void qmp_x_blockdev_insert_medium(bool has_device, const char *device,
2514 bool has_id, const char *id,
2515 const char *node_name, Error **errp)
2517 BlockBackend *blk;
2518 BlockDriverState *bs;
2520 blk = qmp_get_blk(has_device ? device : NULL,
2521 has_id ? id : NULL,
2522 errp);
2523 if (!blk) {
2524 return;
2527 bs = bdrv_find_node(node_name);
2528 if (!bs) {
2529 error_setg(errp, "Node '%s' not found", node_name);
2530 return;
2533 if (bdrv_has_blk(bs)) {
2534 error_setg(errp, "Node '%s' is already in use", node_name);
2535 return;
2538 qmp_blockdev_insert_anon_medium(blk, bs, errp);
2541 void qmp_blockdev_change_medium(bool has_device, const char *device,
2542 bool has_id, const char *id,
2543 const char *filename,
2544 bool has_format, const char *format,
2545 bool has_read_only,
2546 BlockdevChangeReadOnlyMode read_only,
2547 Error **errp)
2549 BlockBackend *blk;
2550 BlockDriverState *medium_bs = NULL;
2551 int bdrv_flags;
2552 int rc;
2553 QDict *options = NULL;
2554 Error *err = NULL;
2556 blk = qmp_get_blk(has_device ? device : NULL,
2557 has_id ? id : NULL,
2558 errp);
2559 if (!blk) {
2560 goto fail;
2563 if (blk_bs(blk)) {
2564 blk_update_root_state(blk);
2567 bdrv_flags = blk_get_open_flags_from_root_state(blk);
2568 bdrv_flags &= ~(BDRV_O_TEMPORARY | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING |
2569 BDRV_O_PROTOCOL);
2571 if (!has_read_only) {
2572 read_only = BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN;
2575 switch (read_only) {
2576 case BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN:
2577 break;
2579 case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_ONLY:
2580 bdrv_flags &= ~BDRV_O_RDWR;
2581 break;
2583 case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_WRITE:
2584 bdrv_flags |= BDRV_O_RDWR;
2585 break;
2587 default:
2588 abort();
2591 if (has_format) {
2592 options = qdict_new();
2593 qdict_put(options, "driver", qstring_from_str(format));
2596 medium_bs = bdrv_open(filename, NULL, options, bdrv_flags, errp);
2597 if (!medium_bs) {
2598 goto fail;
2601 bdrv_add_key(medium_bs, NULL, &err);
2602 if (err) {
2603 error_propagate(errp, err);
2604 goto fail;
2607 rc = do_open_tray(has_device ? device : NULL,
2608 has_id ? id : NULL,
2609 false, &err);
2610 if (rc && rc != -ENOSYS) {
2611 error_propagate(errp, err);
2612 goto fail;
2614 error_free(err);
2615 err = NULL;
2617 qmp_x_blockdev_remove_medium(has_device, device, has_id, id, errp);
2618 if (err) {
2619 error_propagate(errp, err);
2620 goto fail;
2623 qmp_blockdev_insert_anon_medium(blk, medium_bs, &err);
2624 if (err) {
2625 error_propagate(errp, err);
2626 goto fail;
2629 blk_apply_root_state(blk, medium_bs);
2631 qmp_blockdev_close_tray(has_device, device, has_id, id, errp);
2633 fail:
2634 /* If the medium has been inserted, the device has its own reference, so
2635 * ours must be relinquished; and if it has not been inserted successfully,
2636 * the reference must be relinquished anyway */
2637 bdrv_unref(medium_bs);
2640 /* throttling disk I/O limits */
2641 void qmp_block_set_io_throttle(BlockIOThrottle *arg, Error **errp)
2643 ThrottleConfig cfg;
2644 BlockDriverState *bs;
2645 BlockBackend *blk;
2646 AioContext *aio_context;
2648 blk = qmp_get_blk(arg->has_device ? arg->device : NULL,
2649 arg->has_id ? arg->id : NULL,
2650 errp);
2651 if (!blk) {
2652 return;
2655 aio_context = blk_get_aio_context(blk);
2656 aio_context_acquire(aio_context);
2658 bs = blk_bs(blk);
2659 if (!bs) {
2660 error_setg(errp, "Device has no medium");
2661 goto out;
2664 throttle_config_init(&cfg);
2665 cfg.buckets[THROTTLE_BPS_TOTAL].avg = arg->bps;
2666 cfg.buckets[THROTTLE_BPS_READ].avg = arg->bps_rd;
2667 cfg.buckets[THROTTLE_BPS_WRITE].avg = arg->bps_wr;
2669 cfg.buckets[THROTTLE_OPS_TOTAL].avg = arg->iops;
2670 cfg.buckets[THROTTLE_OPS_READ].avg = arg->iops_rd;
2671 cfg.buckets[THROTTLE_OPS_WRITE].avg = arg->iops_wr;
2673 if (arg->has_bps_max) {
2674 cfg.buckets[THROTTLE_BPS_TOTAL].max = arg->bps_max;
2676 if (arg->has_bps_rd_max) {
2677 cfg.buckets[THROTTLE_BPS_READ].max = arg->bps_rd_max;
2679 if (arg->has_bps_wr_max) {
2680 cfg.buckets[THROTTLE_BPS_WRITE].max = arg->bps_wr_max;
2682 if (arg->has_iops_max) {
2683 cfg.buckets[THROTTLE_OPS_TOTAL].max = arg->iops_max;
2685 if (arg->has_iops_rd_max) {
2686 cfg.buckets[THROTTLE_OPS_READ].max = arg->iops_rd_max;
2688 if (arg->has_iops_wr_max) {
2689 cfg.buckets[THROTTLE_OPS_WRITE].max = arg->iops_wr_max;
2692 if (arg->has_bps_max_length) {
2693 cfg.buckets[THROTTLE_BPS_TOTAL].burst_length = arg->bps_max_length;
2695 if (arg->has_bps_rd_max_length) {
2696 cfg.buckets[THROTTLE_BPS_READ].burst_length = arg->bps_rd_max_length;
2698 if (arg->has_bps_wr_max_length) {
2699 cfg.buckets[THROTTLE_BPS_WRITE].burst_length = arg->bps_wr_max_length;
2701 if (arg->has_iops_max_length) {
2702 cfg.buckets[THROTTLE_OPS_TOTAL].burst_length = arg->iops_max_length;
2704 if (arg->has_iops_rd_max_length) {
2705 cfg.buckets[THROTTLE_OPS_READ].burst_length = arg->iops_rd_max_length;
2707 if (arg->has_iops_wr_max_length) {
2708 cfg.buckets[THROTTLE_OPS_WRITE].burst_length = arg->iops_wr_max_length;
2711 if (arg->has_iops_size) {
2712 cfg.op_size = arg->iops_size;
2715 if (!throttle_is_valid(&cfg, errp)) {
2716 goto out;
2719 if (throttle_enabled(&cfg)) {
2720 /* Enable I/O limits if they're not enabled yet, otherwise
2721 * just update the throttling group. */
2722 if (!blk_get_public(blk)->throttle_state) {
2723 blk_io_limits_enable(blk,
2724 arg->has_group ? arg->group :
2725 arg->has_device ? arg->device :
2726 arg->id);
2727 } else if (arg->has_group) {
2728 blk_io_limits_update_group(blk, arg->group);
2730 /* Set the new throttling configuration */
2731 blk_set_io_limits(blk, &cfg);
2732 } else if (blk_get_public(blk)->throttle_state) {
2733 /* If all throttling settings are set to 0, disable I/O limits */
2734 blk_io_limits_disable(blk);
2737 out:
2738 aio_context_release(aio_context);
2741 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2742 bool has_granularity, uint32_t granularity,
2743 Error **errp)
2745 AioContext *aio_context;
2746 BlockDriverState *bs;
2748 if (!name || name[0] == '\0') {
2749 error_setg(errp, "Bitmap name cannot be empty");
2750 return;
2753 bs = bdrv_lookup_bs(node, node, errp);
2754 if (!bs) {
2755 return;
2758 aio_context = bdrv_get_aio_context(bs);
2759 aio_context_acquire(aio_context);
2761 if (has_granularity) {
2762 if (granularity < 512 || !is_power_of_2(granularity)) {
2763 error_setg(errp, "Granularity must be power of 2 "
2764 "and at least 512");
2765 goto out;
2767 } else {
2768 /* Default to cluster size, if available: */
2769 granularity = bdrv_get_default_bitmap_granularity(bs);
2772 bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2774 out:
2775 aio_context_release(aio_context);
2778 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2779 Error **errp)
2781 AioContext *aio_context;
2782 BlockDriverState *bs;
2783 BdrvDirtyBitmap *bitmap;
2785 bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2786 if (!bitmap || !bs) {
2787 return;
2790 if (bdrv_dirty_bitmap_frozen(bitmap)) {
2791 error_setg(errp,
2792 "Bitmap '%s' is currently frozen and cannot be removed",
2793 name);
2794 goto out;
2796 bdrv_dirty_bitmap_make_anon(bitmap);
2797 bdrv_release_dirty_bitmap(bs, bitmap);
2799 out:
2800 aio_context_release(aio_context);
2804 * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2805 * immediately after a full backup operation.
2807 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2808 Error **errp)
2810 AioContext *aio_context;
2811 BdrvDirtyBitmap *bitmap;
2812 BlockDriverState *bs;
2814 bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2815 if (!bitmap || !bs) {
2816 return;
2819 if (bdrv_dirty_bitmap_frozen(bitmap)) {
2820 error_setg(errp,
2821 "Bitmap '%s' is currently frozen and cannot be modified",
2822 name);
2823 goto out;
2824 } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
2825 error_setg(errp,
2826 "Bitmap '%s' is currently disabled and cannot be cleared",
2827 name);
2828 goto out;
2831 bdrv_clear_dirty_bitmap(bitmap, NULL);
2833 out:
2834 aio_context_release(aio_context);
2837 void hmp_drive_del(Monitor *mon, const QDict *qdict)
2839 const char *id = qdict_get_str(qdict, "id");
2840 BlockBackend *blk;
2841 BlockDriverState *bs;
2842 AioContext *aio_context;
2843 Error *local_err = NULL;
2845 bs = bdrv_find_node(id);
2846 if (bs) {
2847 qmp_x_blockdev_del(id, &local_err);
2848 if (local_err) {
2849 error_report_err(local_err);
2851 return;
2854 blk = blk_by_name(id);
2855 if (!blk) {
2856 error_report("Device '%s' not found", id);
2857 return;
2860 if (!blk_legacy_dinfo(blk)) {
2861 error_report("Deleting device added with blockdev-add"
2862 " is not supported");
2863 return;
2866 aio_context = blk_get_aio_context(blk);
2867 aio_context_acquire(aio_context);
2869 bs = blk_bs(blk);
2870 if (bs) {
2871 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2872 error_report_err(local_err);
2873 aio_context_release(aio_context);
2874 return;
2877 blk_remove_bs(blk);
2880 /* Make the BlockBackend and the attached BlockDriverState anonymous */
2881 monitor_remove_blk(blk);
2883 /* If this BlockBackend has a device attached to it, its refcount will be
2884 * decremented when the device is removed; otherwise we have to do so here.
2886 if (blk_get_attached_dev(blk)) {
2887 /* Further I/O must not pause the guest */
2888 blk_set_on_error(blk, BLOCKDEV_ON_ERROR_REPORT,
2889 BLOCKDEV_ON_ERROR_REPORT);
2890 } else {
2891 blk_unref(blk);
2894 aio_context_release(aio_context);
2897 void qmp_block_resize(bool has_device, const char *device,
2898 bool has_node_name, const char *node_name,
2899 int64_t size, Error **errp)
2901 Error *local_err = NULL;
2902 BlockDriverState *bs;
2903 AioContext *aio_context;
2904 int ret;
2906 bs = bdrv_lookup_bs(has_device ? device : NULL,
2907 has_node_name ? node_name : NULL,
2908 &local_err);
2909 if (local_err) {
2910 error_propagate(errp, local_err);
2911 return;
2914 aio_context = bdrv_get_aio_context(bs);
2915 aio_context_acquire(aio_context);
2917 if (!bdrv_is_first_non_filter(bs)) {
2918 error_setg(errp, QERR_FEATURE_DISABLED, "resize");
2919 goto out;
2922 if (size < 0) {
2923 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2924 goto out;
2927 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2928 error_setg(errp, QERR_DEVICE_IN_USE, device);
2929 goto out;
2932 /* complete all in-flight operations before resizing the device */
2933 bdrv_drain_all();
2935 ret = bdrv_truncate(bs, size);
2936 switch (ret) {
2937 case 0:
2938 break;
2939 case -ENOMEDIUM:
2940 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2941 break;
2942 case -ENOTSUP:
2943 error_setg(errp, QERR_UNSUPPORTED);
2944 break;
2945 case -EACCES:
2946 error_setg(errp, "Device '%s' is read only", device);
2947 break;
2948 case -EBUSY:
2949 error_setg(errp, QERR_DEVICE_IN_USE, device);
2950 break;
2951 default:
2952 error_setg_errno(errp, -ret, "Could not resize");
2953 break;
2956 out:
2957 aio_context_release(aio_context);
2960 static void block_job_cb(void *opaque, int ret)
2962 /* Note that this function may be executed from another AioContext besides
2963 * the QEMU main loop. If you need to access anything that assumes the
2964 * QEMU global mutex, use a BH or introduce a mutex.
2967 BlockDriverState *bs = opaque;
2968 const char *msg = NULL;
2970 trace_block_job_cb(bs, bs->job, ret);
2972 assert(bs->job);
2974 if (ret < 0) {
2975 msg = strerror(-ret);
2978 if (block_job_is_cancelled(bs->job)) {
2979 block_job_event_cancelled(bs->job);
2980 } else {
2981 block_job_event_completed(bs->job, msg);
2985 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
2986 bool has_base, const char *base,
2987 bool has_backing_file, const char *backing_file,
2988 bool has_speed, int64_t speed,
2989 bool has_on_error, BlockdevOnError on_error,
2990 Error **errp)
2992 BlockDriverState *bs;
2993 BlockDriverState *base_bs = NULL;
2994 AioContext *aio_context;
2995 Error *local_err = NULL;
2996 const char *base_name = NULL;
2998 if (!has_on_error) {
2999 on_error = BLOCKDEV_ON_ERROR_REPORT;
3002 bs = qmp_get_root_bs(device, errp);
3003 if (!bs) {
3004 return;
3007 aio_context = bdrv_get_aio_context(bs);
3008 aio_context_acquire(aio_context);
3010 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
3011 goto out;
3014 if (has_base) {
3015 base_bs = bdrv_find_backing_image(bs, base);
3016 if (base_bs == NULL) {
3017 error_setg(errp, QERR_BASE_NOT_FOUND, base);
3018 goto out;
3020 assert(bdrv_get_aio_context(base_bs) == aio_context);
3021 base_name = base;
3024 /* if we are streaming the entire chain, the result will have no backing
3025 * file, and specifying one is therefore an error */
3026 if (base_bs == NULL && has_backing_file) {
3027 error_setg(errp, "backing file specified, but streaming the "
3028 "entire chain");
3029 goto out;
3032 /* backing_file string overrides base bs filename */
3033 base_name = has_backing_file ? backing_file : base_name;
3035 stream_start(has_job_id ? job_id : NULL, bs, base_bs, base_name,
3036 has_speed ? speed : 0, on_error, block_job_cb, bs, &local_err);
3037 if (local_err) {
3038 error_propagate(errp, local_err);
3039 goto out;
3042 trace_qmp_block_stream(bs, bs->job);
3044 out:
3045 aio_context_release(aio_context);
3048 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
3049 bool has_base, const char *base,
3050 bool has_top, const char *top,
3051 bool has_backing_file, const char *backing_file,
3052 bool has_speed, int64_t speed,
3053 Error **errp)
3055 BlockDriverState *bs;
3056 BlockDriverState *base_bs, *top_bs;
3057 AioContext *aio_context;
3058 Error *local_err = NULL;
3059 /* This will be part of the QMP command, if/when the
3060 * BlockdevOnError change for blkmirror makes it in
3062 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
3064 if (!has_speed) {
3065 speed = 0;
3068 /* Important Note:
3069 * libvirt relies on the DeviceNotFound error class in order to probe for
3070 * live commit feature versions; for this to work, we must make sure to
3071 * perform the device lookup before any generic errors that may occur in a
3072 * scenario in which all optional arguments are omitted. */
3073 bs = qmp_get_root_bs(device, &local_err);
3074 if (!bs) {
3075 bs = bdrv_lookup_bs(device, device, NULL);
3076 if (!bs) {
3077 error_free(local_err);
3078 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3079 "Device '%s' not found", device);
3080 } else {
3081 error_propagate(errp, local_err);
3083 return;
3086 aio_context = bdrv_get_aio_context(bs);
3087 aio_context_acquire(aio_context);
3089 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
3090 goto out;
3093 /* default top_bs is the active layer */
3094 top_bs = bs;
3096 if (has_top && top) {
3097 if (strcmp(bs->filename, top) != 0) {
3098 top_bs = bdrv_find_backing_image(bs, top);
3102 if (top_bs == NULL) {
3103 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
3104 goto out;
3107 assert(bdrv_get_aio_context(top_bs) == aio_context);
3109 if (has_base && base) {
3110 base_bs = bdrv_find_backing_image(top_bs, base);
3111 } else {
3112 base_bs = bdrv_find_base(top_bs);
3115 if (base_bs == NULL) {
3116 error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
3117 goto out;
3120 assert(bdrv_get_aio_context(base_bs) == aio_context);
3122 if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
3123 goto out;
3126 /* Do not allow attempts to commit an image into itself */
3127 if (top_bs == base_bs) {
3128 error_setg(errp, "cannot commit an image into itself");
3129 goto out;
3132 if (top_bs == bs) {
3133 if (has_backing_file) {
3134 error_setg(errp, "'backing-file' specified,"
3135 " but 'top' is the active layer");
3136 goto out;
3138 commit_active_start(has_job_id ? job_id : NULL, bs, base_bs, speed,
3139 on_error, block_job_cb, bs, &local_err, false);
3140 } else {
3141 commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, speed,
3142 on_error, block_job_cb, bs,
3143 has_backing_file ? backing_file : NULL, &local_err);
3145 if (local_err != NULL) {
3146 error_propagate(errp, local_err);
3147 goto out;
3150 out:
3151 aio_context_release(aio_context);
3154 static void do_drive_backup(DriveBackup *backup, BlockJobTxn *txn, Error **errp)
3156 BlockDriverState *bs;
3157 BlockDriverState *target_bs;
3158 BlockDriverState *source = NULL;
3159 BdrvDirtyBitmap *bmap = NULL;
3160 AioContext *aio_context;
3161 QDict *options = NULL;
3162 Error *local_err = NULL;
3163 int flags;
3164 int64_t size;
3166 if (!backup->has_speed) {
3167 backup->speed = 0;
3169 if (!backup->has_on_source_error) {
3170 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3172 if (!backup->has_on_target_error) {
3173 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3175 if (!backup->has_mode) {
3176 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3178 if (!backup->has_job_id) {
3179 backup->job_id = NULL;
3181 if (!backup->has_compress) {
3182 backup->compress = false;
3185 bs = qmp_get_root_bs(backup->device, errp);
3186 if (!bs) {
3187 return;
3190 aio_context = bdrv_get_aio_context(bs);
3191 aio_context_acquire(aio_context);
3193 if (!backup->has_format) {
3194 backup->format = backup->mode == NEW_IMAGE_MODE_EXISTING ?
3195 NULL : (char*) bs->drv->format_name;
3198 /* Early check to avoid creating target */
3199 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
3200 goto out;
3203 flags = bs->open_flags | BDRV_O_RDWR;
3205 /* See if we have a backing HD we can use to create our new image
3206 * on top of. */
3207 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
3208 source = backing_bs(bs);
3209 if (!source) {
3210 backup->sync = MIRROR_SYNC_MODE_FULL;
3213 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
3214 source = bs;
3217 size = bdrv_getlength(bs);
3218 if (size < 0) {
3219 error_setg_errno(errp, -size, "bdrv_getlength failed");
3220 goto out;
3223 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
3224 assert(backup->format);
3225 if (source) {
3226 bdrv_img_create(backup->target, backup->format, source->filename,
3227 source->drv->format_name, NULL,
3228 size, flags, &local_err, false);
3229 } else {
3230 bdrv_img_create(backup->target, backup->format, NULL, NULL, NULL,
3231 size, flags, &local_err, false);
3235 if (local_err) {
3236 error_propagate(errp, local_err);
3237 goto out;
3240 if (backup->format) {
3241 options = qdict_new();
3242 qdict_put(options, "driver", qstring_from_str(backup->format));
3245 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
3246 if (!target_bs) {
3247 goto out;
3250 bdrv_set_aio_context(target_bs, aio_context);
3252 if (backup->has_bitmap) {
3253 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
3254 if (!bmap) {
3255 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
3256 bdrv_unref(target_bs);
3257 goto out;
3261 backup_start(backup->job_id, bs, target_bs, backup->speed, backup->sync,
3262 bmap, backup->compress, backup->on_source_error,
3263 backup->on_target_error, block_job_cb, bs, txn, &local_err);
3264 bdrv_unref(target_bs);
3265 if (local_err != NULL) {
3266 error_propagate(errp, local_err);
3267 goto out;
3270 out:
3271 aio_context_release(aio_context);
3274 void qmp_drive_backup(DriveBackup *arg, Error **errp)
3276 return do_drive_backup(arg, NULL, errp);
3279 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
3281 return bdrv_named_nodes_list(errp);
3284 void do_blockdev_backup(BlockdevBackup *backup, BlockJobTxn *txn, Error **errp)
3286 BlockDriverState *bs;
3287 BlockDriverState *target_bs;
3288 Error *local_err = NULL;
3289 AioContext *aio_context;
3291 if (!backup->has_speed) {
3292 backup->speed = 0;
3294 if (!backup->has_on_source_error) {
3295 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3297 if (!backup->has_on_target_error) {
3298 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3300 if (!backup->has_job_id) {
3301 backup->job_id = NULL;
3303 if (!backup->has_compress) {
3304 backup->compress = false;
3307 bs = qmp_get_root_bs(backup->device, errp);
3308 if (!bs) {
3309 return;
3312 aio_context = bdrv_get_aio_context(bs);
3313 aio_context_acquire(aio_context);
3315 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
3316 if (!target_bs) {
3317 goto out;
3320 if (bdrv_get_aio_context(target_bs) != aio_context) {
3321 if (!bdrv_has_blk(target_bs)) {
3322 /* The target BDS is not attached, we can safely move it to another
3323 * AioContext. */
3324 bdrv_set_aio_context(target_bs, aio_context);
3325 } else {
3326 error_setg(errp, "Target is attached to a different thread from "
3327 "source.");
3328 goto out;
3331 backup_start(backup->job_id, bs, target_bs, backup->speed, backup->sync,
3332 NULL, backup->compress, backup->on_source_error,
3333 backup->on_target_error, block_job_cb, bs, txn, &local_err);
3334 if (local_err != NULL) {
3335 error_propagate(errp, local_err);
3337 out:
3338 aio_context_release(aio_context);
3341 void qmp_blockdev_backup(BlockdevBackup *arg, Error **errp)
3343 do_blockdev_backup(arg, NULL, errp);
3346 /* Parameter check and block job starting for drive mirroring.
3347 * Caller should hold @device and @target's aio context (must be the same).
3349 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
3350 BlockDriverState *target,
3351 bool has_replaces, const char *replaces,
3352 enum MirrorSyncMode sync,
3353 BlockMirrorBackingMode backing_mode,
3354 bool has_speed, int64_t speed,
3355 bool has_granularity, uint32_t granularity,
3356 bool has_buf_size, int64_t buf_size,
3357 bool has_on_source_error,
3358 BlockdevOnError on_source_error,
3359 bool has_on_target_error,
3360 BlockdevOnError on_target_error,
3361 bool has_unmap, bool unmap,
3362 Error **errp)
3365 if (!has_speed) {
3366 speed = 0;
3368 if (!has_on_source_error) {
3369 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3371 if (!has_on_target_error) {
3372 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3374 if (!has_granularity) {
3375 granularity = 0;
3377 if (!has_buf_size) {
3378 buf_size = 0;
3380 if (!has_unmap) {
3381 unmap = true;
3384 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3385 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3386 "a value in range [512B, 64MB]");
3387 return;
3389 if (granularity & (granularity - 1)) {
3390 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3391 "power of 2");
3392 return;
3395 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3396 return;
3398 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3399 return;
3402 if (!bs->backing && sync == MIRROR_SYNC_MODE_TOP) {
3403 sync = MIRROR_SYNC_MODE_FULL;
3406 /* pass the node name to replace to mirror start since it's loose coupling
3407 * and will allow to check whether the node still exist at mirror completion
3409 mirror_start(job_id, bs, target,
3410 has_replaces ? replaces : NULL,
3411 speed, granularity, buf_size, sync, backing_mode,
3412 on_source_error, on_target_error, unmap,
3413 block_job_cb, bs, errp);
3416 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3418 BlockDriverState *bs;
3419 BlockDriverState *source, *target_bs;
3420 AioContext *aio_context;
3421 BlockMirrorBackingMode backing_mode;
3422 Error *local_err = NULL;
3423 QDict *options = NULL;
3424 int flags;
3425 int64_t size;
3426 const char *format = arg->format;
3428 bs = qmp_get_root_bs(arg->device, errp);
3429 if (!bs) {
3430 return;
3433 aio_context = bdrv_get_aio_context(bs);
3434 aio_context_acquire(aio_context);
3436 if (!arg->has_mode) {
3437 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3440 if (!arg->has_format) {
3441 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3442 ? NULL : bs->drv->format_name);
3445 flags = bs->open_flags | BDRV_O_RDWR;
3446 source = backing_bs(bs);
3447 if (!source && arg->sync == MIRROR_SYNC_MODE_TOP) {
3448 arg->sync = MIRROR_SYNC_MODE_FULL;
3450 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3451 source = bs;
3454 size = bdrv_getlength(bs);
3455 if (size < 0) {
3456 error_setg_errno(errp, -size, "bdrv_getlength failed");
3457 goto out;
3460 if (arg->has_replaces) {
3461 BlockDriverState *to_replace_bs;
3462 AioContext *replace_aio_context;
3463 int64_t replace_size;
3465 if (!arg->has_node_name) {
3466 error_setg(errp, "a node-name must be provided when replacing a"
3467 " named node of the graph");
3468 goto out;
3471 to_replace_bs = check_to_replace_node(bs, arg->replaces, &local_err);
3473 if (!to_replace_bs) {
3474 error_propagate(errp, local_err);
3475 goto out;
3478 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3479 aio_context_acquire(replace_aio_context);
3480 replace_size = bdrv_getlength(to_replace_bs);
3481 aio_context_release(replace_aio_context);
3483 if (size != replace_size) {
3484 error_setg(errp, "cannot replace image with a mirror image of "
3485 "different size");
3486 goto out;
3490 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3491 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3492 } else {
3493 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3496 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !source)
3497 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3499 /* create new image w/o backing file */
3500 assert(format);
3501 bdrv_img_create(arg->target, format,
3502 NULL, NULL, NULL, size, flags, &local_err, false);
3503 } else {
3504 switch (arg->mode) {
3505 case NEW_IMAGE_MODE_EXISTING:
3506 break;
3507 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3508 /* create new image with backing file */
3509 bdrv_img_create(arg->target, format,
3510 source->filename,
3511 source->drv->format_name,
3512 NULL, size, flags, &local_err, false);
3513 break;
3514 default:
3515 abort();
3519 if (local_err) {
3520 error_propagate(errp, local_err);
3521 goto out;
3524 options = qdict_new();
3525 if (arg->has_node_name) {
3526 qdict_put(options, "node-name", qstring_from_str(arg->node_name));
3528 if (format) {
3529 qdict_put(options, "driver", qstring_from_str(format));
3532 /* Mirroring takes care of copy-on-write using the source's backing
3533 * file.
3535 target_bs = bdrv_open(arg->target, NULL, options,
3536 flags | BDRV_O_NO_BACKING, errp);
3537 if (!target_bs) {
3538 goto out;
3541 bdrv_set_aio_context(target_bs, aio_context);
3543 blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3544 arg->has_replaces, arg->replaces, arg->sync,
3545 backing_mode, arg->has_speed, arg->speed,
3546 arg->has_granularity, arg->granularity,
3547 arg->has_buf_size, arg->buf_size,
3548 arg->has_on_source_error, arg->on_source_error,
3549 arg->has_on_target_error, arg->on_target_error,
3550 arg->has_unmap, arg->unmap,
3551 &local_err);
3552 bdrv_unref(target_bs);
3553 error_propagate(errp, local_err);
3554 out:
3555 aio_context_release(aio_context);
3558 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3559 const char *device, const char *target,
3560 bool has_replaces, const char *replaces,
3561 MirrorSyncMode sync,
3562 bool has_speed, int64_t speed,
3563 bool has_granularity, uint32_t granularity,
3564 bool has_buf_size, int64_t buf_size,
3565 bool has_on_source_error,
3566 BlockdevOnError on_source_error,
3567 bool has_on_target_error,
3568 BlockdevOnError on_target_error,
3569 Error **errp)
3571 BlockDriverState *bs;
3572 BlockDriverState *target_bs;
3573 AioContext *aio_context;
3574 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3575 Error *local_err = NULL;
3577 bs = qmp_get_root_bs(device, errp);
3578 if (!bs) {
3579 return;
3582 target_bs = bdrv_lookup_bs(target, target, errp);
3583 if (!target_bs) {
3584 return;
3587 aio_context = bdrv_get_aio_context(bs);
3588 aio_context_acquire(aio_context);
3590 bdrv_set_aio_context(target_bs, aio_context);
3592 blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
3593 has_replaces, replaces, sync, backing_mode,
3594 has_speed, speed,
3595 has_granularity, granularity,
3596 has_buf_size, buf_size,
3597 has_on_source_error, on_source_error,
3598 has_on_target_error, on_target_error,
3599 true, true,
3600 &local_err);
3601 error_propagate(errp, local_err);
3603 aio_context_release(aio_context);
3606 /* Get a block job using its ID and acquire its AioContext */
3607 static BlockJob *find_block_job(const char *id, AioContext **aio_context,
3608 Error **errp)
3610 BlockJob *job;
3612 assert(id != NULL);
3614 *aio_context = NULL;
3616 job = block_job_get(id);
3618 if (!job) {
3619 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3620 "Block job '%s' not found", id);
3621 return NULL;
3624 *aio_context = blk_get_aio_context(job->blk);
3625 aio_context_acquire(*aio_context);
3627 return job;
3630 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3632 AioContext *aio_context;
3633 BlockJob *job = find_block_job(device, &aio_context, errp);
3635 if (!job) {
3636 return;
3639 block_job_set_speed(job, speed, errp);
3640 aio_context_release(aio_context);
3643 void qmp_block_job_cancel(const char *device,
3644 bool has_force, bool force, Error **errp)
3646 AioContext *aio_context;
3647 BlockJob *job = find_block_job(device, &aio_context, errp);
3649 if (!job) {
3650 return;
3653 if (!has_force) {
3654 force = false;
3657 if (job->user_paused && !force) {
3658 error_setg(errp, "The block job for device '%s' is currently paused",
3659 device);
3660 goto out;
3663 trace_qmp_block_job_cancel(job);
3664 block_job_cancel(job);
3665 out:
3666 aio_context_release(aio_context);
3669 void qmp_block_job_pause(const char *device, Error **errp)
3671 AioContext *aio_context;
3672 BlockJob *job = find_block_job(device, &aio_context, errp);
3674 if (!job || job->user_paused) {
3675 return;
3678 job->user_paused = true;
3679 trace_qmp_block_job_pause(job);
3680 block_job_pause(job);
3681 aio_context_release(aio_context);
3684 void qmp_block_job_resume(const char *device, Error **errp)
3686 AioContext *aio_context;
3687 BlockJob *job = find_block_job(device, &aio_context, errp);
3689 if (!job || !job->user_paused) {
3690 return;
3693 job->user_paused = false;
3694 trace_qmp_block_job_resume(job);
3695 block_job_iostatus_reset(job);
3696 block_job_resume(job);
3697 aio_context_release(aio_context);
3700 void qmp_block_job_complete(const char *device, Error **errp)
3702 AioContext *aio_context;
3703 BlockJob *job = find_block_job(device, &aio_context, errp);
3705 if (!job) {
3706 return;
3709 trace_qmp_block_job_complete(job);
3710 block_job_complete(job, errp);
3711 aio_context_release(aio_context);
3714 void qmp_change_backing_file(const char *device,
3715 const char *image_node_name,
3716 const char *backing_file,
3717 Error **errp)
3719 BlockDriverState *bs = NULL;
3720 AioContext *aio_context;
3721 BlockDriverState *image_bs = NULL;
3722 Error *local_err = NULL;
3723 bool ro;
3724 int open_flags;
3725 int ret;
3727 bs = qmp_get_root_bs(device, errp);
3728 if (!bs) {
3729 return;
3732 aio_context = bdrv_get_aio_context(bs);
3733 aio_context_acquire(aio_context);
3735 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3736 if (local_err) {
3737 error_propagate(errp, local_err);
3738 goto out;
3741 if (!image_bs) {
3742 error_setg(errp, "image file not found");
3743 goto out;
3746 if (bdrv_find_base(image_bs) == image_bs) {
3747 error_setg(errp, "not allowing backing file change on an image "
3748 "without a backing file");
3749 goto out;
3752 /* even though we are not necessarily operating on bs, we need it to
3753 * determine if block ops are currently prohibited on the chain */
3754 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3755 goto out;
3758 /* final sanity check */
3759 if (!bdrv_chain_contains(bs, image_bs)) {
3760 error_setg(errp, "'%s' and image file are not in the same chain",
3761 device);
3762 goto out;
3765 /* if not r/w, reopen to make r/w */
3766 open_flags = image_bs->open_flags;
3767 ro = bdrv_is_read_only(image_bs);
3769 if (ro) {
3770 bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
3771 if (local_err) {
3772 error_propagate(errp, local_err);
3773 goto out;
3777 ret = bdrv_change_backing_file(image_bs, backing_file,
3778 image_bs->drv ? image_bs->drv->format_name : "");
3780 if (ret < 0) {
3781 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3782 backing_file);
3783 /* don't exit here, so we can try to restore open flags if
3784 * appropriate */
3787 if (ro) {
3788 bdrv_reopen(image_bs, open_flags, &local_err);
3789 error_propagate(errp, local_err);
3792 out:
3793 aio_context_release(aio_context);
3796 void hmp_drive_add_node(Monitor *mon, const char *optstr)
3798 QemuOpts *opts;
3799 QDict *qdict;
3800 Error *local_err = NULL;
3802 opts = qemu_opts_parse_noisily(&qemu_drive_opts, optstr, false);
3803 if (!opts) {
3804 return;
3807 qdict = qemu_opts_to_qdict(opts, NULL);
3809 if (!qdict_get_try_str(qdict, "node-name")) {
3810 QDECREF(qdict);
3811 error_report("'node-name' needs to be specified");
3812 goto out;
3815 BlockDriverState *bs = bds_tree_init(qdict, &local_err);
3816 if (!bs) {
3817 error_report_err(local_err);
3818 goto out;
3821 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
3823 out:
3824 qemu_opts_del(opts);
3827 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3829 BlockDriverState *bs;
3830 QObject *obj;
3831 Visitor *v = qmp_output_visitor_new(&obj);
3832 QDict *qdict;
3833 Error *local_err = NULL;
3835 /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
3836 * cache.direct=false instead of silently switching to aio=threads, except
3837 * when called from drive_new().
3839 * For now, simply forbidding the combination for all drivers will do. */
3840 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
3841 bool direct = options->has_cache &&
3842 options->cache->has_direct &&
3843 options->cache->direct;
3844 if (!direct) {
3845 error_setg(errp, "aio=native requires cache.direct=true");
3846 goto fail;
3850 visit_type_BlockdevOptions(v, NULL, &options, &local_err);
3851 if (local_err) {
3852 error_propagate(errp, local_err);
3853 goto fail;
3856 visit_complete(v, &obj);
3857 qdict = qobject_to_qdict(obj);
3859 qdict_flatten(qdict);
3861 if (!qdict_get_try_str(qdict, "node-name")) {
3862 error_setg(errp, "'node-name' must be specified for the root node");
3863 goto fail;
3866 bs = bds_tree_init(qdict, errp);
3867 if (!bs) {
3868 goto fail;
3871 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
3873 if (bs && bdrv_key_required(bs)) {
3874 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3875 bdrv_unref(bs);
3876 error_setg(errp, "blockdev-add doesn't support encrypted devices");
3877 goto fail;
3880 fail:
3881 visit_free(v);
3884 void qmp_x_blockdev_del(const char *node_name, Error **errp)
3886 AioContext *aio_context;
3887 BlockDriverState *bs;
3889 bs = bdrv_find_node(node_name);
3890 if (!bs) {
3891 error_setg(errp, "Cannot find node %s", node_name);
3892 return;
3894 if (bdrv_has_blk(bs)) {
3895 error_setg(errp, "Node %s is in use", node_name);
3896 return;
3898 aio_context = bdrv_get_aio_context(bs);
3899 aio_context_acquire(aio_context);
3901 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3902 goto out;
3905 if (!bs->monitor_list.tqe_prev) {
3906 error_setg(errp, "Node %s is not owned by the monitor",
3907 bs->node_name);
3908 goto out;
3911 if (bs->refcnt > 1) {
3912 error_setg(errp, "Block device %s is in use",
3913 bdrv_get_device_or_node_name(bs));
3914 goto out;
3917 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3918 bdrv_unref(bs);
3920 out:
3921 aio_context_release(aio_context);
3924 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3925 const char *child_name)
3927 BdrvChild *child;
3929 QLIST_FOREACH(child, &parent_bs->children, next) {
3930 if (strcmp(child->name, child_name) == 0) {
3931 return child;
3935 return NULL;
3938 void qmp_x_blockdev_change(const char *parent, bool has_child,
3939 const char *child, bool has_node,
3940 const char *node, Error **errp)
3942 BlockDriverState *parent_bs, *new_bs = NULL;
3943 BdrvChild *p_child;
3945 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3946 if (!parent_bs) {
3947 return;
3950 if (has_child == has_node) {
3951 if (has_child) {
3952 error_setg(errp, "The parameters child and node are in conflict");
3953 } else {
3954 error_setg(errp, "Either child or node must be specified");
3956 return;
3959 if (has_child) {
3960 p_child = bdrv_find_child(parent_bs, child);
3961 if (!p_child) {
3962 error_setg(errp, "Node '%s' does not have child '%s'",
3963 parent, child);
3964 return;
3966 bdrv_del_child(parent_bs, p_child, errp);
3969 if (has_node) {
3970 new_bs = bdrv_find_node(node);
3971 if (!new_bs) {
3972 error_setg(errp, "Node '%s' not found", node);
3973 return;
3975 bdrv_add_child(parent_bs, new_bs, errp);
3979 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3981 BlockJobInfoList *head = NULL, **p_next = &head;
3982 BlockJob *job;
3984 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
3985 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
3986 AioContext *aio_context = blk_get_aio_context(job->blk);
3988 aio_context_acquire(aio_context);
3989 elem->value = block_job_query(job);
3990 aio_context_release(aio_context);
3992 *p_next = elem;
3993 p_next = &elem->next;
3996 return head;
3999 QemuOptsList qemu_common_drive_opts = {
4000 .name = "drive",
4001 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
4002 .desc = {
4004 .name = "snapshot",
4005 .type = QEMU_OPT_BOOL,
4006 .help = "enable/disable snapshot mode",
4008 .name = "discard",
4009 .type = QEMU_OPT_STRING,
4010 .help = "discard operation (ignore/off, unmap/on)",
4012 .name = "aio",
4013 .type = QEMU_OPT_STRING,
4014 .help = "host AIO implementation (threads, native)",
4016 .name = BDRV_OPT_CACHE_WB,
4017 .type = QEMU_OPT_BOOL,
4018 .help = "Enable writeback mode",
4020 .name = "format",
4021 .type = QEMU_OPT_STRING,
4022 .help = "disk format (raw, qcow2, ...)",
4024 .name = "rerror",
4025 .type = QEMU_OPT_STRING,
4026 .help = "read error action",
4028 .name = "werror",
4029 .type = QEMU_OPT_STRING,
4030 .help = "write error action",
4032 .name = BDRV_OPT_READ_ONLY,
4033 .type = QEMU_OPT_BOOL,
4034 .help = "open drive file as read-only",
4036 .name = "throttling.iops-total",
4037 .type = QEMU_OPT_NUMBER,
4038 .help = "limit total I/O operations per second",
4040 .name = "throttling.iops-read",
4041 .type = QEMU_OPT_NUMBER,
4042 .help = "limit read operations per second",
4044 .name = "throttling.iops-write",
4045 .type = QEMU_OPT_NUMBER,
4046 .help = "limit write operations per second",
4048 .name = "throttling.bps-total",
4049 .type = QEMU_OPT_NUMBER,
4050 .help = "limit total bytes per second",
4052 .name = "throttling.bps-read",
4053 .type = QEMU_OPT_NUMBER,
4054 .help = "limit read bytes per second",
4056 .name = "throttling.bps-write",
4057 .type = QEMU_OPT_NUMBER,
4058 .help = "limit write bytes per second",
4060 .name = "throttling.iops-total-max",
4061 .type = QEMU_OPT_NUMBER,
4062 .help = "I/O operations burst",
4064 .name = "throttling.iops-read-max",
4065 .type = QEMU_OPT_NUMBER,
4066 .help = "I/O operations read burst",
4068 .name = "throttling.iops-write-max",
4069 .type = QEMU_OPT_NUMBER,
4070 .help = "I/O operations write burst",
4072 .name = "throttling.bps-total-max",
4073 .type = QEMU_OPT_NUMBER,
4074 .help = "total bytes burst",
4076 .name = "throttling.bps-read-max",
4077 .type = QEMU_OPT_NUMBER,
4078 .help = "total bytes read burst",
4080 .name = "throttling.bps-write-max",
4081 .type = QEMU_OPT_NUMBER,
4082 .help = "total bytes write burst",
4084 .name = "throttling.iops-total-max-length",
4085 .type = QEMU_OPT_NUMBER,
4086 .help = "length of the iops-total-max burst period, in seconds",
4088 .name = "throttling.iops-read-max-length",
4089 .type = QEMU_OPT_NUMBER,
4090 .help = "length of the iops-read-max burst period, in seconds",
4092 .name = "throttling.iops-write-max-length",
4093 .type = QEMU_OPT_NUMBER,
4094 .help = "length of the iops-write-max burst period, in seconds",
4096 .name = "throttling.bps-total-max-length",
4097 .type = QEMU_OPT_NUMBER,
4098 .help = "length of the bps-total-max burst period, in seconds",
4100 .name = "throttling.bps-read-max-length",
4101 .type = QEMU_OPT_NUMBER,
4102 .help = "length of the bps-read-max burst period, in seconds",
4104 .name = "throttling.bps-write-max-length",
4105 .type = QEMU_OPT_NUMBER,
4106 .help = "length of the bps-write-max burst period, in seconds",
4108 .name = "throttling.iops-size",
4109 .type = QEMU_OPT_NUMBER,
4110 .help = "when limiting by iops max size of an I/O in bytes",
4112 .name = "throttling.group",
4113 .type = QEMU_OPT_STRING,
4114 .help = "name of the block throttling group",
4116 .name = "copy-on-read",
4117 .type = QEMU_OPT_BOOL,
4118 .help = "copy read data from backing file into image file",
4120 .name = "detect-zeroes",
4121 .type = QEMU_OPT_STRING,
4122 .help = "try to optimize zero writes (off, on, unmap)",
4124 .name = "stats-account-invalid",
4125 .type = QEMU_OPT_BOOL,
4126 .help = "whether to account for invalid I/O operations "
4127 "in the statistics",
4129 .name = "stats-account-failed",
4130 .type = QEMU_OPT_BOOL,
4131 .help = "whether to account for failed I/O operations "
4132 "in the statistics",
4134 { /* end of list */ }
4138 static QemuOptsList qemu_root_bds_opts = {
4139 .name = "root-bds",
4140 .head = QTAILQ_HEAD_INITIALIZER(qemu_root_bds_opts.head),
4141 .desc = {
4143 .name = "discard",
4144 .type = QEMU_OPT_STRING,
4145 .help = "discard operation (ignore/off, unmap/on)",
4147 .name = "aio",
4148 .type = QEMU_OPT_STRING,
4149 .help = "host AIO implementation (threads, native)",
4151 .name = "copy-on-read",
4152 .type = QEMU_OPT_BOOL,
4153 .help = "copy read data from backing file into image file",
4155 .name = "detect-zeroes",
4156 .type = QEMU_OPT_STRING,
4157 .help = "try to optimize zero writes (off, on, unmap)",
4159 { /* end of list */ }
4163 QemuOptsList qemu_drive_opts = {
4164 .name = "drive",
4165 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
4166 .desc = {
4168 * no elements => accept any params
4169 * validation will happen later
4171 { /* end of list */ }