iscsi: Set request_alignment during .bdrv_refresh_limits()
[qemu/kevin.git] / blockdev.c
blob3a104a0aaacf0aa33d885e6423b8c7635bd66d93
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 *device, bool force, Error **errp);
61 static const char *const if_name[IF_COUNT] = {
62 [IF_NONE] = "none",
63 [IF_IDE] = "ide",
64 [IF_SCSI] = "scsi",
65 [IF_FLOPPY] = "floppy",
66 [IF_PFLASH] = "pflash",
67 [IF_MTD] = "mtd",
68 [IF_SD] = "sd",
69 [IF_VIRTIO] = "virtio",
70 [IF_XEN] = "xen",
73 static int if_max_devs[IF_COUNT] = {
75 * Do not change these numbers! They govern how drive option
76 * index maps to unit and bus. That mapping is ABI.
78 * All controllers used to implement if=T drives need to support
79 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
80 * Otherwise, some index values map to "impossible" bus, unit
81 * values.
83 * For instance, if you change [IF_SCSI] to 255, -drive
84 * if=scsi,index=12 no longer means bus=1,unit=5, but
85 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
86 * the drive can't be set up. Regression.
88 [IF_IDE] = 2,
89 [IF_SCSI] = 7,
92 /**
93 * Boards may call this to offer board-by-board overrides
94 * of the default, global values.
96 void override_max_devs(BlockInterfaceType type, int max_devs)
98 BlockBackend *blk;
99 DriveInfo *dinfo;
101 if (max_devs <= 0) {
102 return;
105 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
106 dinfo = blk_legacy_dinfo(blk);
107 if (dinfo->type == type) {
108 fprintf(stderr, "Cannot override units-per-bus property of"
109 " the %s interface, because a drive of that type has"
110 " already been added.\n", if_name[type]);
111 g_assert_not_reached();
115 if_max_devs[type] = max_devs;
119 * We automatically delete the drive when a device using it gets
120 * unplugged. Questionable feature, but we can't just drop it.
121 * Device models call blockdev_mark_auto_del() to schedule the
122 * automatic deletion, and generic qdev code calls blockdev_auto_del()
123 * when deletion is actually safe.
125 void blockdev_mark_auto_del(BlockBackend *blk)
127 DriveInfo *dinfo = blk_legacy_dinfo(blk);
128 BlockDriverState *bs = blk_bs(blk);
129 AioContext *aio_context;
131 if (!dinfo) {
132 return;
135 if (bs) {
136 aio_context = bdrv_get_aio_context(bs);
137 aio_context_acquire(aio_context);
139 if (bs->job) {
140 block_job_cancel(bs->job);
143 aio_context_release(aio_context);
146 dinfo->auto_del = 1;
149 void blockdev_auto_del(BlockBackend *blk)
151 DriveInfo *dinfo = blk_legacy_dinfo(blk);
153 if (dinfo && dinfo->auto_del) {
154 monitor_remove_blk(blk);
155 blk_unref(blk);
160 * Returns the current mapping of how many units per bus
161 * a particular interface can support.
163 * A positive integer indicates n units per bus.
164 * 0 implies the mapping has not been established.
165 * -1 indicates an invalid BlockInterfaceType was given.
167 int drive_get_max_devs(BlockInterfaceType type)
169 if (type >= IF_IDE && type < IF_COUNT) {
170 return if_max_devs[type];
173 return -1;
176 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
178 int max_devs = if_max_devs[type];
179 return max_devs ? index / max_devs : 0;
182 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
184 int max_devs = if_max_devs[type];
185 return max_devs ? index % max_devs : index;
188 QemuOpts *drive_def(const char *optstr)
190 return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
193 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
194 const char *optstr)
196 QemuOpts *opts;
198 opts = drive_def(optstr);
199 if (!opts) {
200 return NULL;
202 if (type != IF_DEFAULT) {
203 qemu_opt_set(opts, "if", if_name[type], &error_abort);
205 if (index >= 0) {
206 qemu_opt_set_number(opts, "index", index, &error_abort);
208 if (file)
209 qemu_opt_set(opts, "file", file, &error_abort);
210 return opts;
213 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
215 BlockBackend *blk;
216 DriveInfo *dinfo;
218 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
219 dinfo = blk_legacy_dinfo(blk);
220 if (dinfo && dinfo->type == type
221 && dinfo->bus == bus && dinfo->unit == unit) {
222 return dinfo;
226 return NULL;
229 bool drive_check_orphaned(void)
231 BlockBackend *blk;
232 DriveInfo *dinfo;
233 bool rs = false;
235 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
236 dinfo = blk_legacy_dinfo(blk);
237 /* If dinfo->bdrv->dev is NULL, it has no device attached. */
238 /* Unless this is a default drive, this may be an oversight. */
239 if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
240 dinfo->type != IF_NONE) {
241 fprintf(stderr, "Warning: Orphaned drive without device: "
242 "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
243 blk_name(blk), blk_bs(blk) ? blk_bs(blk)->filename : "",
244 if_name[dinfo->type], dinfo->bus, dinfo->unit);
245 rs = true;
249 return rs;
252 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
254 return drive_get(type,
255 drive_index_to_bus_id(type, index),
256 drive_index_to_unit_id(type, index));
259 int drive_get_max_bus(BlockInterfaceType type)
261 int max_bus;
262 BlockBackend *blk;
263 DriveInfo *dinfo;
265 max_bus = -1;
266 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
267 dinfo = blk_legacy_dinfo(blk);
268 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
269 max_bus = dinfo->bus;
272 return max_bus;
275 /* Get a block device. This should only be used for single-drive devices
276 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
277 appropriate bus. */
278 DriveInfo *drive_get_next(BlockInterfaceType type)
280 static int next_block_unit[IF_COUNT];
282 return drive_get(type, 0, next_block_unit[type]++);
285 static void bdrv_format_print(void *opaque, const char *name)
287 error_printf(" %s", name);
290 typedef struct {
291 QEMUBH *bh;
292 BlockDriverState *bs;
293 } BDRVPutRefBH;
295 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
297 if (!strcmp(buf, "ignore")) {
298 return BLOCKDEV_ON_ERROR_IGNORE;
299 } else if (!is_read && !strcmp(buf, "enospc")) {
300 return BLOCKDEV_ON_ERROR_ENOSPC;
301 } else if (!strcmp(buf, "stop")) {
302 return BLOCKDEV_ON_ERROR_STOP;
303 } else if (!strcmp(buf, "report")) {
304 return BLOCKDEV_ON_ERROR_REPORT;
305 } else {
306 error_setg(errp, "'%s' invalid %s error action",
307 buf, is_read ? "read" : "write");
308 return -1;
312 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
313 Error **errp)
315 const QListEntry *entry;
316 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
317 switch (qobject_type(entry->value)) {
319 case QTYPE_QSTRING: {
320 unsigned long long length;
321 const char *str = qstring_get_str(qobject_to_qstring(entry->value));
322 if (parse_uint_full(str, &length, 10) == 0 &&
323 length > 0 && length <= UINT_MAX) {
324 block_acct_add_interval(stats, (unsigned) length);
325 } else {
326 error_setg(errp, "Invalid interval length: %s", str);
327 return false;
329 break;
332 case QTYPE_QINT: {
333 int64_t length = qint_get_int(qobject_to_qint(entry->value));
334 if (length > 0 && length <= UINT_MAX) {
335 block_acct_add_interval(stats, (unsigned) length);
336 } else {
337 error_setg(errp, "Invalid interval length: %" PRId64, length);
338 return false;
340 break;
343 default:
344 error_setg(errp, "The specification of stats-intervals is invalid");
345 return false;
348 return true;
351 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
353 /* All parameters but @opts are optional and may be set to NULL. */
354 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
355 const char **throttling_group, ThrottleConfig *throttle_cfg,
356 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
358 const char *discard;
359 Error *local_error = NULL;
360 const char *aio;
362 if (bdrv_flags) {
363 if (!qemu_opt_get_bool(opts, "read-only", false)) {
364 *bdrv_flags |= BDRV_O_RDWR;
366 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
367 *bdrv_flags |= BDRV_O_COPY_ON_READ;
370 if ((discard = qemu_opt_get(opts, "discard")) != NULL) {
371 if (bdrv_parse_discard_flags(discard, bdrv_flags) != 0) {
372 error_setg(errp, "Invalid discard option");
373 return;
377 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
378 if (!strcmp(aio, "native")) {
379 *bdrv_flags |= BDRV_O_NATIVE_AIO;
380 } else if (!strcmp(aio, "threads")) {
381 /* this is the default */
382 } else {
383 error_setg(errp, "invalid aio option");
384 return;
389 /* disk I/O throttling */
390 if (throttling_group) {
391 *throttling_group = qemu_opt_get(opts, "throttling.group");
394 if (throttle_cfg) {
395 throttle_config_init(throttle_cfg);
396 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
397 qemu_opt_get_number(opts, "throttling.bps-total", 0);
398 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
399 qemu_opt_get_number(opts, "throttling.bps-read", 0);
400 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
401 qemu_opt_get_number(opts, "throttling.bps-write", 0);
402 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
403 qemu_opt_get_number(opts, "throttling.iops-total", 0);
404 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
405 qemu_opt_get_number(opts, "throttling.iops-read", 0);
406 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
407 qemu_opt_get_number(opts, "throttling.iops-write", 0);
409 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
410 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
411 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
412 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
413 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
414 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
415 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
416 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
417 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
418 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
419 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
420 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
422 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
423 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
424 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
425 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
426 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
427 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
428 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
429 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
430 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
431 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
432 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
433 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
435 throttle_cfg->op_size =
436 qemu_opt_get_number(opts, "throttling.iops-size", 0);
438 if (!throttle_is_valid(throttle_cfg, errp)) {
439 return;
443 if (detect_zeroes) {
444 *detect_zeroes =
445 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
446 qemu_opt_get(opts, "detect-zeroes"),
447 BLOCKDEV_DETECT_ZEROES_OPTIONS__MAX,
448 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
449 &local_error);
450 if (local_error) {
451 error_propagate(errp, local_error);
452 return;
455 if (bdrv_flags &&
456 *detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
457 !(*bdrv_flags & BDRV_O_UNMAP))
459 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
460 "without setting discard operation to unmap");
461 return;
466 /* Takes the ownership of bs_opts */
467 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
468 Error **errp)
470 const char *buf;
471 int bdrv_flags = 0;
472 int on_read_error, on_write_error;
473 bool account_invalid, account_failed;
474 bool writethrough;
475 BlockBackend *blk;
476 BlockDriverState *bs;
477 ThrottleConfig cfg;
478 int snapshot = 0;
479 Error *error = NULL;
480 QemuOpts *opts;
481 QDict *interval_dict = NULL;
482 QList *interval_list = NULL;
483 const char *id;
484 BlockdevDetectZeroesOptions detect_zeroes =
485 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
486 const char *throttling_group = NULL;
488 /* Check common options by copying from bs_opts to opts, all other options
489 * stay in bs_opts for processing by bdrv_open(). */
490 id = qdict_get_try_str(bs_opts, "id");
491 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
492 if (error) {
493 error_propagate(errp, error);
494 goto err_no_opts;
497 qemu_opts_absorb_qdict(opts, bs_opts, &error);
498 if (error) {
499 error_propagate(errp, error);
500 goto early_err;
503 if (id) {
504 qdict_del(bs_opts, "id");
507 /* extract parameters */
508 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
510 account_invalid = qemu_opt_get_bool(opts, "stats-account-invalid", true);
511 account_failed = qemu_opt_get_bool(opts, "stats-account-failed", true);
513 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
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 /* init */
569 if ((!file || !*file) && !qdict_size(bs_opts)) {
570 BlockBackendRootState *blk_rs;
572 blk = blk_new();
573 blk_rs = blk_get_root_state(blk);
574 blk_rs->open_flags = bdrv_flags;
575 blk_rs->read_only = !(bdrv_flags & BDRV_O_RDWR);
576 blk_rs->detect_zeroes = detect_zeroes;
578 QDECREF(bs_opts);
579 } else {
580 if (file && !*file) {
581 file = NULL;
584 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
585 * with other callers) rather than what we want as the real defaults.
586 * Apply the defaults here instead. */
587 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
588 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
589 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
591 if (runstate_check(RUN_STATE_INMIGRATE)) {
592 bdrv_flags |= BDRV_O_INACTIVE;
595 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
596 if (!blk) {
597 goto err_no_bs_opts;
599 bs = blk_bs(blk);
601 bs->detect_zeroes = detect_zeroes;
603 if (bdrv_key_required(bs)) {
604 autostart = 0;
607 block_acct_init(blk_get_stats(blk), account_invalid, account_failed);
609 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
610 blk_unref(blk);
611 blk = NULL;
612 goto err_no_bs_opts;
616 /* disk I/O throttling */
617 if (throttle_enabled(&cfg)) {
618 if (!throttling_group) {
619 throttling_group = blk_name(blk);
621 blk_io_limits_enable(blk, throttling_group);
622 blk_set_io_limits(blk, &cfg);
625 blk_set_enable_write_cache(blk, !writethrough);
626 blk_set_on_error(blk, on_read_error, on_write_error);
628 if (!monitor_add_blk(blk, qemu_opts_id(opts), errp)) {
629 blk_unref(blk);
630 blk = NULL;
631 goto err_no_bs_opts;
634 err_no_bs_opts:
635 qemu_opts_del(opts);
636 QDECREF(interval_dict);
637 QDECREF(interval_list);
638 return blk;
640 early_err:
641 qemu_opts_del(opts);
642 QDECREF(interval_dict);
643 QDECREF(interval_list);
644 err_no_opts:
645 QDECREF(bs_opts);
646 return NULL;
649 static QemuOptsList qemu_root_bds_opts;
651 /* Takes the ownership of bs_opts */
652 static BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
654 BlockDriverState *bs;
655 QemuOpts *opts;
656 Error *local_error = NULL;
657 BlockdevDetectZeroesOptions detect_zeroes;
658 int bdrv_flags = 0;
660 opts = qemu_opts_create(&qemu_root_bds_opts, NULL, 1, errp);
661 if (!opts) {
662 goto fail;
665 qemu_opts_absorb_qdict(opts, bs_opts, &local_error);
666 if (local_error) {
667 error_propagate(errp, local_error);
668 goto fail;
671 extract_common_blockdev_options(opts, &bdrv_flags, NULL, NULL,
672 &detect_zeroes, &local_error);
673 if (local_error) {
674 error_propagate(errp, local_error);
675 goto fail;
678 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
679 * with other callers) rather than what we want as the real defaults.
680 * Apply the defaults here instead. */
681 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
682 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
684 if (runstate_check(RUN_STATE_INMIGRATE)) {
685 bdrv_flags |= BDRV_O_INACTIVE;
688 bs = bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
689 if (!bs) {
690 goto fail_no_bs_opts;
693 bs->detect_zeroes = detect_zeroes;
695 fail_no_bs_opts:
696 qemu_opts_del(opts);
697 return bs;
699 fail:
700 qemu_opts_del(opts);
701 QDECREF(bs_opts);
702 return NULL;
705 void blockdev_close_all_bdrv_states(void)
707 BlockDriverState *bs, *next_bs;
709 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
710 AioContext *ctx = bdrv_get_aio_context(bs);
712 aio_context_acquire(ctx);
713 bdrv_unref(bs);
714 aio_context_release(ctx);
718 /* Iterates over the list of monitor-owned BlockDriverStates */
719 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
721 return bs ? QTAILQ_NEXT(bs, monitor_list)
722 : QTAILQ_FIRST(&monitor_bdrv_states);
725 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
726 Error **errp)
728 const char *value;
730 value = qemu_opt_get(opts, from);
731 if (value) {
732 if (qemu_opt_find(opts, to)) {
733 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
734 "same time", to, from);
735 return;
739 /* rename all items in opts */
740 while ((value = qemu_opt_get(opts, from))) {
741 qemu_opt_set(opts, to, value, &error_abort);
742 qemu_opt_unset(opts, from);
746 QemuOptsList qemu_legacy_drive_opts = {
747 .name = "drive",
748 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
749 .desc = {
751 .name = "bus",
752 .type = QEMU_OPT_NUMBER,
753 .help = "bus number",
755 .name = "unit",
756 .type = QEMU_OPT_NUMBER,
757 .help = "unit number (i.e. lun for scsi)",
759 .name = "index",
760 .type = QEMU_OPT_NUMBER,
761 .help = "index number",
763 .name = "media",
764 .type = QEMU_OPT_STRING,
765 .help = "media type (disk, cdrom)",
767 .name = "if",
768 .type = QEMU_OPT_STRING,
769 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
771 .name = "cyls",
772 .type = QEMU_OPT_NUMBER,
773 .help = "number of cylinders (ide disk geometry)",
775 .name = "heads",
776 .type = QEMU_OPT_NUMBER,
777 .help = "number of heads (ide disk geometry)",
779 .name = "secs",
780 .type = QEMU_OPT_NUMBER,
781 .help = "number of sectors (ide disk geometry)",
783 .name = "trans",
784 .type = QEMU_OPT_STRING,
785 .help = "chs translation (auto, lba, none)",
787 .name = "boot",
788 .type = QEMU_OPT_BOOL,
789 .help = "(deprecated, ignored)",
791 .name = "addr",
792 .type = QEMU_OPT_STRING,
793 .help = "pci address (virtio only)",
795 .name = "serial",
796 .type = QEMU_OPT_STRING,
797 .help = "disk serial number",
799 .name = "file",
800 .type = QEMU_OPT_STRING,
801 .help = "file name",
804 /* Options that are passed on, but have special semantics with -drive */
806 .name = "read-only",
807 .type = QEMU_OPT_BOOL,
808 .help = "open drive file as read-only",
810 .name = "rerror",
811 .type = QEMU_OPT_STRING,
812 .help = "read error action",
814 .name = "werror",
815 .type = QEMU_OPT_STRING,
816 .help = "write error action",
818 .name = "copy-on-read",
819 .type = QEMU_OPT_BOOL,
820 .help = "copy read data from backing file into image file",
823 { /* end of list */ }
827 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
829 const char *value;
830 BlockBackend *blk;
831 DriveInfo *dinfo = NULL;
832 QDict *bs_opts;
833 QemuOpts *legacy_opts;
834 DriveMediaType media = MEDIA_DISK;
835 BlockInterfaceType type;
836 int cyls, heads, secs, translation;
837 int max_devs, bus_id, unit_id, index;
838 const char *devaddr;
839 const char *werror, *rerror;
840 bool read_only = false;
841 bool copy_on_read;
842 const char *serial;
843 const char *filename;
844 Error *local_err = NULL;
845 int i;
847 /* Change legacy command line options into QMP ones */
848 static const struct {
849 const char *from;
850 const char *to;
851 } opt_renames[] = {
852 { "iops", "throttling.iops-total" },
853 { "iops_rd", "throttling.iops-read" },
854 { "iops_wr", "throttling.iops-write" },
856 { "bps", "throttling.bps-total" },
857 { "bps_rd", "throttling.bps-read" },
858 { "bps_wr", "throttling.bps-write" },
860 { "iops_max", "throttling.iops-total-max" },
861 { "iops_rd_max", "throttling.iops-read-max" },
862 { "iops_wr_max", "throttling.iops-write-max" },
864 { "bps_max", "throttling.bps-total-max" },
865 { "bps_rd_max", "throttling.bps-read-max" },
866 { "bps_wr_max", "throttling.bps-write-max" },
868 { "iops_size", "throttling.iops-size" },
870 { "group", "throttling.group" },
872 { "readonly", "read-only" },
875 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
876 qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
877 &local_err);
878 if (local_err) {
879 error_report_err(local_err);
880 return NULL;
884 value = qemu_opt_get(all_opts, "cache");
885 if (value) {
886 int flags = 0;
887 bool writethrough;
889 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
890 error_report("invalid cache option");
891 return NULL;
894 /* Specific options take precedence */
895 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
896 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
897 !writethrough, &error_abort);
899 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
900 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
901 !!(flags & BDRV_O_NOCACHE), &error_abort);
903 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
904 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
905 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
907 qemu_opt_unset(all_opts, "cache");
910 /* Get a QDict for processing the options */
911 bs_opts = qdict_new();
912 qemu_opts_to_qdict(all_opts, bs_opts);
914 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
915 &error_abort);
916 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
917 if (local_err) {
918 error_report_err(local_err);
919 goto fail;
922 /* Deprecated option boot=[on|off] */
923 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
924 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
925 "ignored. Future versions will reject this parameter. Please "
926 "update your scripts.\n");
929 /* Media type */
930 value = qemu_opt_get(legacy_opts, "media");
931 if (value) {
932 if (!strcmp(value, "disk")) {
933 media = MEDIA_DISK;
934 } else if (!strcmp(value, "cdrom")) {
935 media = MEDIA_CDROM;
936 read_only = true;
937 } else {
938 error_report("'%s' invalid media", value);
939 goto fail;
943 /* copy-on-read is disabled with a warning for read-only devices */
944 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
945 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
947 if (read_only && copy_on_read) {
948 error_report("warning: disabling copy-on-read on read-only drive");
949 copy_on_read = false;
952 qdict_put(bs_opts, "read-only",
953 qstring_from_str(read_only ? "on" : "off"));
954 qdict_put(bs_opts, "copy-on-read",
955 qstring_from_str(copy_on_read ? "on" :"off"));
957 /* Controller type */
958 value = qemu_opt_get(legacy_opts, "if");
959 if (value) {
960 for (type = 0;
961 type < IF_COUNT && strcmp(value, if_name[type]);
962 type++) {
964 if (type == IF_COUNT) {
965 error_report("unsupported bus type '%s'", value);
966 goto fail;
968 } else {
969 type = block_default_type;
972 /* Geometry */
973 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
974 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
975 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
977 if (cyls || heads || secs) {
978 if (cyls < 1) {
979 error_report("invalid physical cyls number");
980 goto fail;
982 if (heads < 1) {
983 error_report("invalid physical heads number");
984 goto fail;
986 if (secs < 1) {
987 error_report("invalid physical secs number");
988 goto fail;
992 translation = BIOS_ATA_TRANSLATION_AUTO;
993 value = qemu_opt_get(legacy_opts, "trans");
994 if (value != NULL) {
995 if (!cyls) {
996 error_report("'%s' trans must be used with cyls, heads and secs",
997 value);
998 goto fail;
1000 if (!strcmp(value, "none")) {
1001 translation = BIOS_ATA_TRANSLATION_NONE;
1002 } else if (!strcmp(value, "lba")) {
1003 translation = BIOS_ATA_TRANSLATION_LBA;
1004 } else if (!strcmp(value, "large")) {
1005 translation = BIOS_ATA_TRANSLATION_LARGE;
1006 } else if (!strcmp(value, "rechs")) {
1007 translation = BIOS_ATA_TRANSLATION_RECHS;
1008 } else if (!strcmp(value, "auto")) {
1009 translation = BIOS_ATA_TRANSLATION_AUTO;
1010 } else {
1011 error_report("'%s' invalid translation type", value);
1012 goto fail;
1016 if (media == MEDIA_CDROM) {
1017 if (cyls || secs || heads) {
1018 error_report("CHS can't be set with media=cdrom");
1019 goto fail;
1023 /* Device address specified by bus/unit or index.
1024 * If none was specified, try to find the first free one. */
1025 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
1026 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
1027 index = qemu_opt_get_number(legacy_opts, "index", -1);
1029 max_devs = if_max_devs[type];
1031 if (index != -1) {
1032 if (bus_id != 0 || unit_id != -1) {
1033 error_report("index cannot be used with bus and unit");
1034 goto fail;
1036 bus_id = drive_index_to_bus_id(type, index);
1037 unit_id = drive_index_to_unit_id(type, index);
1040 if (unit_id == -1) {
1041 unit_id = 0;
1042 while (drive_get(type, bus_id, unit_id) != NULL) {
1043 unit_id++;
1044 if (max_devs && unit_id >= max_devs) {
1045 unit_id -= max_devs;
1046 bus_id++;
1051 if (max_devs && unit_id >= max_devs) {
1052 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
1053 goto fail;
1056 if (drive_get(type, bus_id, unit_id) != NULL) {
1057 error_report("drive with bus=%d, unit=%d (index=%d) exists",
1058 bus_id, unit_id, index);
1059 goto fail;
1062 /* Serial number */
1063 serial = qemu_opt_get(legacy_opts, "serial");
1065 /* no id supplied -> create one */
1066 if (qemu_opts_id(all_opts) == NULL) {
1067 char *new_id;
1068 const char *mediastr = "";
1069 if (type == IF_IDE || type == IF_SCSI) {
1070 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
1072 if (max_devs) {
1073 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
1074 mediastr, unit_id);
1075 } else {
1076 new_id = g_strdup_printf("%s%s%i", if_name[type],
1077 mediastr, unit_id);
1079 qdict_put(bs_opts, "id", qstring_from_str(new_id));
1080 g_free(new_id);
1083 /* Add virtio block device */
1084 devaddr = qemu_opt_get(legacy_opts, "addr");
1085 if (devaddr && type != IF_VIRTIO) {
1086 error_report("addr is not supported by this bus type");
1087 goto fail;
1090 if (type == IF_VIRTIO) {
1091 QemuOpts *devopts;
1092 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
1093 &error_abort);
1094 if (arch_type == QEMU_ARCH_S390X) {
1095 qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
1096 } else {
1097 qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
1099 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
1100 &error_abort);
1101 if (devaddr) {
1102 qemu_opt_set(devopts, "addr", devaddr, &error_abort);
1106 filename = qemu_opt_get(legacy_opts, "file");
1108 /* Check werror/rerror compatibility with if=... */
1109 werror = qemu_opt_get(legacy_opts, "werror");
1110 if (werror != NULL) {
1111 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
1112 type != IF_NONE) {
1113 error_report("werror is not supported by this bus type");
1114 goto fail;
1116 qdict_put(bs_opts, "werror", qstring_from_str(werror));
1119 rerror = qemu_opt_get(legacy_opts, "rerror");
1120 if (rerror != NULL) {
1121 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
1122 type != IF_NONE) {
1123 error_report("rerror is not supported by this bus type");
1124 goto fail;
1126 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
1129 /* Actual block device init: Functionality shared with blockdev-add */
1130 blk = blockdev_init(filename, bs_opts, &local_err);
1131 bs_opts = NULL;
1132 if (!blk) {
1133 if (local_err) {
1134 error_report_err(local_err);
1136 goto fail;
1137 } else {
1138 assert(!local_err);
1141 /* Create legacy DriveInfo */
1142 dinfo = g_malloc0(sizeof(*dinfo));
1143 dinfo->opts = all_opts;
1145 dinfo->cyls = cyls;
1146 dinfo->heads = heads;
1147 dinfo->secs = secs;
1148 dinfo->trans = translation;
1150 dinfo->type = type;
1151 dinfo->bus = bus_id;
1152 dinfo->unit = unit_id;
1153 dinfo->devaddr = devaddr;
1154 dinfo->serial = g_strdup(serial);
1156 blk_set_legacy_dinfo(blk, dinfo);
1158 switch(type) {
1159 case IF_IDE:
1160 case IF_SCSI:
1161 case IF_XEN:
1162 case IF_NONE:
1163 dinfo->media_cd = media == MEDIA_CDROM;
1164 break;
1165 default:
1166 break;
1169 fail:
1170 qemu_opts_del(legacy_opts);
1171 QDECREF(bs_opts);
1172 return dinfo;
1175 void hmp_commit(Monitor *mon, const QDict *qdict)
1177 const char *device = qdict_get_str(qdict, "device");
1178 BlockBackend *blk;
1179 int ret;
1181 if (!strcmp(device, "all")) {
1182 ret = blk_commit_all();
1183 } else {
1184 BlockDriverState *bs;
1185 AioContext *aio_context;
1187 blk = blk_by_name(device);
1188 if (!blk) {
1189 monitor_printf(mon, "Device '%s' not found\n", device);
1190 return;
1192 if (!blk_is_available(blk)) {
1193 monitor_printf(mon, "Device '%s' has no medium\n", device);
1194 return;
1197 bs = blk_bs(blk);
1198 aio_context = bdrv_get_aio_context(bs);
1199 aio_context_acquire(aio_context);
1201 ret = bdrv_commit(bs);
1203 aio_context_release(aio_context);
1205 if (ret < 0) {
1206 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1207 strerror(-ret));
1211 static void blockdev_do_action(TransactionAction *action, Error **errp)
1213 TransactionActionList list;
1215 list.value = action;
1216 list.next = NULL;
1217 qmp_transaction(&list, false, NULL, errp);
1220 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1221 bool has_node_name, const char *node_name,
1222 const char *snapshot_file,
1223 bool has_snapshot_node_name,
1224 const char *snapshot_node_name,
1225 bool has_format, const char *format,
1226 bool has_mode, NewImageMode mode, Error **errp)
1228 BlockdevSnapshotSync snapshot = {
1229 .has_device = has_device,
1230 .device = (char *) device,
1231 .has_node_name = has_node_name,
1232 .node_name = (char *) node_name,
1233 .snapshot_file = (char *) snapshot_file,
1234 .has_snapshot_node_name = has_snapshot_node_name,
1235 .snapshot_node_name = (char *) snapshot_node_name,
1236 .has_format = has_format,
1237 .format = (char *) format,
1238 .has_mode = has_mode,
1239 .mode = mode,
1241 TransactionAction action = {
1242 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1243 .u.blockdev_snapshot_sync.data = &snapshot,
1245 blockdev_do_action(&action, errp);
1248 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1249 Error **errp)
1251 BlockdevSnapshot snapshot_data = {
1252 .node = (char *) node,
1253 .overlay = (char *) overlay
1255 TransactionAction action = {
1256 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1257 .u.blockdev_snapshot.data = &snapshot_data,
1259 blockdev_do_action(&action, errp);
1262 void qmp_blockdev_snapshot_internal_sync(const char *device,
1263 const char *name,
1264 Error **errp)
1266 BlockdevSnapshotInternal snapshot = {
1267 .device = (char *) device,
1268 .name = (char *) name
1270 TransactionAction action = {
1271 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1272 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1274 blockdev_do_action(&action, errp);
1277 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1278 bool has_id,
1279 const char *id,
1280 bool has_name,
1281 const char *name,
1282 Error **errp)
1284 BlockDriverState *bs;
1285 BlockBackend *blk;
1286 AioContext *aio_context;
1287 QEMUSnapshotInfo sn;
1288 Error *local_err = NULL;
1289 SnapshotInfo *info = NULL;
1290 int ret;
1292 blk = blk_by_name(device);
1293 if (!blk) {
1294 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1295 "Device '%s' not found", device);
1296 return NULL;
1299 aio_context = blk_get_aio_context(blk);
1300 aio_context_acquire(aio_context);
1302 if (!has_id) {
1303 id = NULL;
1306 if (!has_name) {
1307 name = NULL;
1310 if (!id && !name) {
1311 error_setg(errp, "Name or id must be provided");
1312 goto out_aio_context;
1315 if (!blk_is_available(blk)) {
1316 error_setg(errp, "Device '%s' has no medium", device);
1317 goto out_aio_context;
1319 bs = blk_bs(blk);
1321 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1322 goto out_aio_context;
1325 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1326 if (local_err) {
1327 error_propagate(errp, local_err);
1328 goto out_aio_context;
1330 if (!ret) {
1331 error_setg(errp,
1332 "Snapshot with id '%s' and name '%s' does not exist on "
1333 "device '%s'",
1334 STR_OR_NULL(id), STR_OR_NULL(name), device);
1335 goto out_aio_context;
1338 bdrv_snapshot_delete(bs, id, name, &local_err);
1339 if (local_err) {
1340 error_propagate(errp, local_err);
1341 goto out_aio_context;
1344 aio_context_release(aio_context);
1346 info = g_new0(SnapshotInfo, 1);
1347 info->id = g_strdup(sn.id_str);
1348 info->name = g_strdup(sn.name);
1349 info->date_nsec = sn.date_nsec;
1350 info->date_sec = sn.date_sec;
1351 info->vm_state_size = sn.vm_state_size;
1352 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1353 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1355 return info;
1357 out_aio_context:
1358 aio_context_release(aio_context);
1359 return NULL;
1363 * block_dirty_bitmap_lookup:
1364 * Return a dirty bitmap (if present), after validating
1365 * the node reference and bitmap names.
1367 * @node: The name of the BDS node to search for bitmaps
1368 * @name: The name of the bitmap to search for
1369 * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1370 * @paio: Output pointer for aio_context acquisition, if desired. Can be NULL.
1371 * @errp: Output pointer for error information. Can be NULL.
1373 * @return: A bitmap object on success, or NULL on failure.
1375 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1376 const char *name,
1377 BlockDriverState **pbs,
1378 AioContext **paio,
1379 Error **errp)
1381 BlockDriverState *bs;
1382 BdrvDirtyBitmap *bitmap;
1383 AioContext *aio_context;
1385 if (!node) {
1386 error_setg(errp, "Node cannot be NULL");
1387 return NULL;
1389 if (!name) {
1390 error_setg(errp, "Bitmap name cannot be NULL");
1391 return NULL;
1393 bs = bdrv_lookup_bs(node, node, NULL);
1394 if (!bs) {
1395 error_setg(errp, "Node '%s' not found", node);
1396 return NULL;
1399 aio_context = bdrv_get_aio_context(bs);
1400 aio_context_acquire(aio_context);
1402 bitmap = bdrv_find_dirty_bitmap(bs, name);
1403 if (!bitmap) {
1404 error_setg(errp, "Dirty bitmap '%s' not found", name);
1405 goto fail;
1408 if (pbs) {
1409 *pbs = bs;
1411 if (paio) {
1412 *paio = aio_context;
1413 } else {
1414 aio_context_release(aio_context);
1417 return bitmap;
1419 fail:
1420 aio_context_release(aio_context);
1421 return NULL;
1424 /* New and old BlockDriverState structs for atomic group operations */
1426 typedef struct BlkActionState BlkActionState;
1429 * BlkActionOps:
1430 * Table of operations that define an Action.
1432 * @instance_size: Size of state struct, in bytes.
1433 * @prepare: Prepare the work, must NOT be NULL.
1434 * @commit: Commit the changes, can be NULL.
1435 * @abort: Abort the changes on fail, can be NULL.
1436 * @clean: Clean up resources after all transaction actions have called
1437 * commit() or abort(). Can be NULL.
1439 * Only prepare() may fail. In a single transaction, only one of commit() or
1440 * abort() will be called. clean() will always be called if it is present.
1442 typedef struct BlkActionOps {
1443 size_t instance_size;
1444 void (*prepare)(BlkActionState *common, Error **errp);
1445 void (*commit)(BlkActionState *common);
1446 void (*abort)(BlkActionState *common);
1447 void (*clean)(BlkActionState *common);
1448 } BlkActionOps;
1451 * BlkActionState:
1452 * Describes one Action's state within a Transaction.
1454 * @action: QAPI-defined enum identifying which Action to perform.
1455 * @ops: Table of ActionOps this Action can perform.
1456 * @block_job_txn: Transaction which this action belongs to.
1457 * @entry: List membership for all Actions in this Transaction.
1459 * This structure must be arranged as first member in a subclassed type,
1460 * assuming that the compiler will also arrange it to the same offsets as the
1461 * base class.
1463 struct BlkActionState {
1464 TransactionAction *action;
1465 const BlkActionOps *ops;
1466 BlockJobTxn *block_job_txn;
1467 TransactionProperties *txn_props;
1468 QSIMPLEQ_ENTRY(BlkActionState) entry;
1471 /* internal snapshot private data */
1472 typedef struct InternalSnapshotState {
1473 BlkActionState common;
1474 BlockDriverState *bs;
1475 AioContext *aio_context;
1476 QEMUSnapshotInfo sn;
1477 bool created;
1478 } InternalSnapshotState;
1481 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1483 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1484 error_setg(errp,
1485 "Action '%s' does not support Transaction property "
1486 "completion-mode = %s",
1487 TransactionActionKind_lookup[s->action->type],
1488 ActionCompletionMode_lookup[s->txn_props->completion_mode]);
1489 return -1;
1491 return 0;
1494 static void internal_snapshot_prepare(BlkActionState *common,
1495 Error **errp)
1497 Error *local_err = NULL;
1498 const char *device;
1499 const char *name;
1500 BlockBackend *blk;
1501 BlockDriverState *bs;
1502 QEMUSnapshotInfo old_sn, *sn;
1503 bool ret;
1504 qemu_timeval tv;
1505 BlockdevSnapshotInternal *internal;
1506 InternalSnapshotState *state;
1507 int ret1;
1509 g_assert(common->action->type ==
1510 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1511 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1512 state = DO_UPCAST(InternalSnapshotState, common, common);
1514 /* 1. parse input */
1515 device = internal->device;
1516 name = internal->name;
1518 /* 2. check for validation */
1519 if (action_check_completion_mode(common, errp) < 0) {
1520 return;
1523 blk = blk_by_name(device);
1524 if (!blk) {
1525 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1526 "Device '%s' not found", device);
1527 return;
1530 /* AioContext is released in .clean() */
1531 state->aio_context = blk_get_aio_context(blk);
1532 aio_context_acquire(state->aio_context);
1534 if (!blk_is_available(blk)) {
1535 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1536 return;
1538 bs = blk_bs(blk);
1540 state->bs = bs;
1541 bdrv_drained_begin(bs);
1543 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1544 return;
1547 if (bdrv_is_read_only(bs)) {
1548 error_setg(errp, "Device '%s' is read only", device);
1549 return;
1552 if (!bdrv_can_snapshot(bs)) {
1553 error_setg(errp, "Block format '%s' used by device '%s' "
1554 "does not support internal snapshots",
1555 bs->drv->format_name, device);
1556 return;
1559 if (!strlen(name)) {
1560 error_setg(errp, "Name is empty");
1561 return;
1564 /* check whether a snapshot with name exist */
1565 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1566 &local_err);
1567 if (local_err) {
1568 error_propagate(errp, local_err);
1569 return;
1570 } else if (ret) {
1571 error_setg(errp,
1572 "Snapshot with name '%s' already exists on device '%s'",
1573 name, device);
1574 return;
1577 /* 3. take the snapshot */
1578 sn = &state->sn;
1579 pstrcpy(sn->name, sizeof(sn->name), name);
1580 qemu_gettimeofday(&tv);
1581 sn->date_sec = tv.tv_sec;
1582 sn->date_nsec = tv.tv_usec * 1000;
1583 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1585 ret1 = bdrv_snapshot_create(bs, sn);
1586 if (ret1 < 0) {
1587 error_setg_errno(errp, -ret1,
1588 "Failed to create snapshot '%s' on device '%s'",
1589 name, device);
1590 return;
1593 /* 4. succeed, mark a snapshot is created */
1594 state->created = true;
1597 static void internal_snapshot_abort(BlkActionState *common)
1599 InternalSnapshotState *state =
1600 DO_UPCAST(InternalSnapshotState, common, common);
1601 BlockDriverState *bs = state->bs;
1602 QEMUSnapshotInfo *sn = &state->sn;
1603 Error *local_error = NULL;
1605 if (!state->created) {
1606 return;
1609 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1610 error_reportf_err(local_error,
1611 "Failed to delete snapshot with id '%s' and "
1612 "name '%s' on device '%s' in abort: ",
1613 sn->id_str, sn->name,
1614 bdrv_get_device_name(bs));
1618 static void internal_snapshot_clean(BlkActionState *common)
1620 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1621 common, common);
1623 if (state->aio_context) {
1624 if (state->bs) {
1625 bdrv_drained_end(state->bs);
1627 aio_context_release(state->aio_context);
1631 /* external snapshot private data */
1632 typedef struct ExternalSnapshotState {
1633 BlkActionState common;
1634 BlockDriverState *old_bs;
1635 BlockDriverState *new_bs;
1636 AioContext *aio_context;
1637 } ExternalSnapshotState;
1639 static void external_snapshot_prepare(BlkActionState *common,
1640 Error **errp)
1642 int flags = 0;
1643 QDict *options = NULL;
1644 Error *local_err = NULL;
1645 /* Device and node name of the image to generate the snapshot from */
1646 const char *device;
1647 const char *node_name;
1648 /* Reference to the new image (for 'blockdev-snapshot') */
1649 const char *snapshot_ref;
1650 /* File name of the new image (for 'blockdev-snapshot-sync') */
1651 const char *new_image_file;
1652 ExternalSnapshotState *state =
1653 DO_UPCAST(ExternalSnapshotState, common, common);
1654 TransactionAction *action = common->action;
1656 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1657 * purpose but a different set of parameters */
1658 switch (action->type) {
1659 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1661 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1662 device = s->node;
1663 node_name = s->node;
1664 new_image_file = NULL;
1665 snapshot_ref = s->overlay;
1667 break;
1668 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1670 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1671 device = s->has_device ? s->device : NULL;
1672 node_name = s->has_node_name ? s->node_name : NULL;
1673 new_image_file = s->snapshot_file;
1674 snapshot_ref = NULL;
1676 break;
1677 default:
1678 g_assert_not_reached();
1681 /* start processing */
1682 if (action_check_completion_mode(common, errp) < 0) {
1683 return;
1686 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1687 if (!state->old_bs) {
1688 return;
1691 /* Acquire AioContext now so any threads operating on old_bs stop */
1692 state->aio_context = bdrv_get_aio_context(state->old_bs);
1693 aio_context_acquire(state->aio_context);
1694 bdrv_drained_begin(state->old_bs);
1696 if (!bdrv_is_inserted(state->old_bs)) {
1697 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1698 return;
1701 if (bdrv_op_is_blocked(state->old_bs,
1702 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1703 return;
1706 if (!bdrv_is_read_only(state->old_bs)) {
1707 if (bdrv_flush(state->old_bs)) {
1708 error_setg(errp, QERR_IO_ERROR);
1709 return;
1713 if (!bdrv_is_first_non_filter(state->old_bs)) {
1714 error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1715 return;
1718 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1719 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1720 const char *format = s->has_format ? s->format : "qcow2";
1721 enum NewImageMode mode;
1722 const char *snapshot_node_name =
1723 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1725 if (node_name && !snapshot_node_name) {
1726 error_setg(errp, "New snapshot node name missing");
1727 return;
1730 if (snapshot_node_name &&
1731 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1732 error_setg(errp, "New snapshot node name already in use");
1733 return;
1736 flags = state->old_bs->open_flags;
1737 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ);
1739 /* create new image w/backing file */
1740 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1741 if (mode != NEW_IMAGE_MODE_EXISTING) {
1742 int64_t size = bdrv_getlength(state->old_bs);
1743 if (size < 0) {
1744 error_setg_errno(errp, -size, "bdrv_getlength failed");
1745 return;
1747 bdrv_img_create(new_image_file, format,
1748 state->old_bs->filename,
1749 state->old_bs->drv->format_name,
1750 NULL, size, flags, &local_err, false);
1751 if (local_err) {
1752 error_propagate(errp, local_err);
1753 return;
1757 options = qdict_new();
1758 if (s->has_snapshot_node_name) {
1759 qdict_put(options, "node-name",
1760 qstring_from_str(snapshot_node_name));
1762 qdict_put(options, "driver", qstring_from_str(format));
1764 flags |= BDRV_O_NO_BACKING;
1767 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1768 errp);
1769 /* We will manually add the backing_hd field to the bs later */
1770 if (!state->new_bs) {
1771 return;
1774 if (bdrv_has_blk(state->new_bs)) {
1775 error_setg(errp, "The snapshot is already in use by %s",
1776 bdrv_get_parent_name(state->new_bs));
1777 return;
1780 if (bdrv_op_is_blocked(state->new_bs, BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
1781 errp)) {
1782 return;
1785 if (state->new_bs->backing != NULL) {
1786 error_setg(errp, "The snapshot already has a backing image");
1787 return;
1790 if (!state->new_bs->drv->supports_backing) {
1791 error_setg(errp, "The snapshot does not support backing images");
1795 static void external_snapshot_commit(BlkActionState *common)
1797 ExternalSnapshotState *state =
1798 DO_UPCAST(ExternalSnapshotState, common, common);
1800 bdrv_set_aio_context(state->new_bs, state->aio_context);
1802 /* This removes our old bs and adds the new bs */
1803 bdrv_append(state->new_bs, state->old_bs);
1804 /* We don't need (or want) to use the transactional
1805 * bdrv_reopen_multiple() across all the entries at once, because we
1806 * don't want to abort all of them if one of them fails the reopen */
1807 if (!state->old_bs->copy_on_read) {
1808 bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR,
1809 NULL);
1813 static void external_snapshot_abort(BlkActionState *common)
1815 ExternalSnapshotState *state =
1816 DO_UPCAST(ExternalSnapshotState, common, common);
1817 if (state->new_bs) {
1818 bdrv_unref(state->new_bs);
1822 static void external_snapshot_clean(BlkActionState *common)
1824 ExternalSnapshotState *state =
1825 DO_UPCAST(ExternalSnapshotState, common, common);
1826 if (state->aio_context) {
1827 bdrv_drained_end(state->old_bs);
1828 aio_context_release(state->aio_context);
1832 typedef struct DriveBackupState {
1833 BlkActionState common;
1834 BlockDriverState *bs;
1835 AioContext *aio_context;
1836 BlockJob *job;
1837 } DriveBackupState;
1839 static void do_drive_backup(const char *device, const char *target,
1840 bool has_format, const char *format,
1841 enum MirrorSyncMode sync,
1842 bool has_mode, enum NewImageMode mode,
1843 bool has_speed, int64_t speed,
1844 bool has_bitmap, const char *bitmap,
1845 bool has_on_source_error,
1846 BlockdevOnError on_source_error,
1847 bool has_on_target_error,
1848 BlockdevOnError on_target_error,
1849 BlockJobTxn *txn, Error **errp);
1851 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1853 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1854 BlockBackend *blk;
1855 DriveBackup *backup;
1856 Error *local_err = NULL;
1858 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1859 backup = common->action->u.drive_backup.data;
1861 blk = blk_by_name(backup->device);
1862 if (!blk) {
1863 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1864 "Device '%s' not found", backup->device);
1865 return;
1868 if (!blk_is_available(blk)) {
1869 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1870 return;
1873 /* AioContext is released in .clean() */
1874 state->aio_context = blk_get_aio_context(blk);
1875 aio_context_acquire(state->aio_context);
1876 bdrv_drained_begin(blk_bs(blk));
1877 state->bs = blk_bs(blk);
1879 do_drive_backup(backup->device, backup->target,
1880 backup->has_format, backup->format,
1881 backup->sync,
1882 backup->has_mode, backup->mode,
1883 backup->has_speed, backup->speed,
1884 backup->has_bitmap, backup->bitmap,
1885 backup->has_on_source_error, backup->on_source_error,
1886 backup->has_on_target_error, backup->on_target_error,
1887 common->block_job_txn, &local_err);
1888 if (local_err) {
1889 error_propagate(errp, local_err);
1890 return;
1893 state->job = state->bs->job;
1896 static void drive_backup_abort(BlkActionState *common)
1898 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1899 BlockDriverState *bs = state->bs;
1901 /* Only cancel if it's the job we started */
1902 if (bs && bs->job && bs->job == state->job) {
1903 block_job_cancel_sync(bs->job);
1907 static void drive_backup_clean(BlkActionState *common)
1909 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1911 if (state->aio_context) {
1912 bdrv_drained_end(state->bs);
1913 aio_context_release(state->aio_context);
1917 typedef struct BlockdevBackupState {
1918 BlkActionState common;
1919 BlockDriverState *bs;
1920 BlockJob *job;
1921 AioContext *aio_context;
1922 } BlockdevBackupState;
1924 static void do_blockdev_backup(const char *device, const char *target,
1925 enum MirrorSyncMode sync,
1926 bool has_speed, int64_t speed,
1927 bool has_on_source_error,
1928 BlockdevOnError on_source_error,
1929 bool has_on_target_error,
1930 BlockdevOnError on_target_error,
1931 BlockJobTxn *txn, Error **errp);
1933 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1935 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1936 BlockdevBackup *backup;
1937 BlockBackend *blk, *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 blk = blk_by_name(backup->device);
1944 if (!blk) {
1945 error_setg(errp, "Device '%s' not found", backup->device);
1946 return;
1949 if (!blk_is_available(blk)) {
1950 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1951 return;
1954 target = blk_by_name(backup->target);
1955 if (!target) {
1956 error_setg(errp, "Device '%s' not found", backup->target);
1957 return;
1960 /* AioContext is released in .clean() */
1961 state->aio_context = blk_get_aio_context(blk);
1962 if (state->aio_context != blk_get_aio_context(target)) {
1963 state->aio_context = NULL;
1964 error_setg(errp, "Backup between two IO threads is not implemented");
1965 return;
1967 aio_context_acquire(state->aio_context);
1968 state->bs = blk_bs(blk);
1969 bdrv_drained_begin(state->bs);
1971 do_blockdev_backup(backup->device, backup->target,
1972 backup->sync,
1973 backup->has_speed, backup->speed,
1974 backup->has_on_source_error, backup->on_source_error,
1975 backup->has_on_target_error, backup->on_target_error,
1976 common->block_job_txn, &local_err);
1977 if (local_err) {
1978 error_propagate(errp, local_err);
1979 return;
1982 state->job = state->bs->job;
1985 static void blockdev_backup_abort(BlkActionState *common)
1987 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1988 BlockDriverState *bs = state->bs;
1990 /* Only cancel if it's the job we started */
1991 if (bs && bs->job && bs->job == state->job) {
1992 block_job_cancel_sync(bs->job);
1996 static void blockdev_backup_clean(BlkActionState *common)
1998 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
2000 if (state->aio_context) {
2001 bdrv_drained_end(state->bs);
2002 aio_context_release(state->aio_context);
2006 typedef struct BlockDirtyBitmapState {
2007 BlkActionState common;
2008 BdrvDirtyBitmap *bitmap;
2009 BlockDriverState *bs;
2010 AioContext *aio_context;
2011 HBitmap *backup;
2012 bool prepared;
2013 } BlockDirtyBitmapState;
2015 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
2016 Error **errp)
2018 Error *local_err = NULL;
2019 BlockDirtyBitmapAdd *action;
2020 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2021 common, common);
2023 if (action_check_completion_mode(common, errp) < 0) {
2024 return;
2027 action = common->action->u.block_dirty_bitmap_add.data;
2028 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2029 qmp_block_dirty_bitmap_add(action->node, action->name,
2030 action->has_granularity, action->granularity,
2031 &local_err);
2033 if (!local_err) {
2034 state->prepared = true;
2035 } else {
2036 error_propagate(errp, local_err);
2040 static void block_dirty_bitmap_add_abort(BlkActionState *common)
2042 BlockDirtyBitmapAdd *action;
2043 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2044 common, common);
2046 action = common->action->u.block_dirty_bitmap_add.data;
2047 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2048 * then the node reference and bitmap name must have been valid.
2050 if (state->prepared) {
2051 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2055 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2056 Error **errp)
2058 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2059 common, common);
2060 BlockDirtyBitmap *action;
2062 if (action_check_completion_mode(common, errp) < 0) {
2063 return;
2066 action = common->action->u.block_dirty_bitmap_clear.data;
2067 state->bitmap = block_dirty_bitmap_lookup(action->node,
2068 action->name,
2069 &state->bs,
2070 &state->aio_context,
2071 errp);
2072 if (!state->bitmap) {
2073 return;
2076 if (bdrv_dirty_bitmap_frozen(state->bitmap)) {
2077 error_setg(errp, "Cannot modify a frozen bitmap");
2078 return;
2079 } else if (!bdrv_dirty_bitmap_enabled(state->bitmap)) {
2080 error_setg(errp, "Cannot clear a disabled bitmap");
2081 return;
2084 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2085 /* AioContext is released in .clean() */
2088 static void block_dirty_bitmap_clear_abort(BlkActionState *common)
2090 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2091 common, common);
2093 bdrv_undo_clear_dirty_bitmap(state->bitmap, state->backup);
2096 static void block_dirty_bitmap_clear_commit(BlkActionState *common)
2098 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2099 common, common);
2101 hbitmap_free(state->backup);
2104 static void block_dirty_bitmap_clear_clean(BlkActionState *common)
2106 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2107 common, common);
2109 if (state->aio_context) {
2110 aio_context_release(state->aio_context);
2114 static void abort_prepare(BlkActionState *common, Error **errp)
2116 error_setg(errp, "Transaction aborted using Abort action");
2119 static void abort_commit(BlkActionState *common)
2121 g_assert_not_reached(); /* this action never succeeds */
2124 static const BlkActionOps actions[] = {
2125 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2126 .instance_size = sizeof(ExternalSnapshotState),
2127 .prepare = external_snapshot_prepare,
2128 .commit = external_snapshot_commit,
2129 .abort = external_snapshot_abort,
2130 .clean = external_snapshot_clean,
2132 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2133 .instance_size = sizeof(ExternalSnapshotState),
2134 .prepare = external_snapshot_prepare,
2135 .commit = external_snapshot_commit,
2136 .abort = external_snapshot_abort,
2137 .clean = external_snapshot_clean,
2139 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2140 .instance_size = sizeof(DriveBackupState),
2141 .prepare = drive_backup_prepare,
2142 .abort = drive_backup_abort,
2143 .clean = drive_backup_clean,
2145 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2146 .instance_size = sizeof(BlockdevBackupState),
2147 .prepare = blockdev_backup_prepare,
2148 .abort = blockdev_backup_abort,
2149 .clean = blockdev_backup_clean,
2151 [TRANSACTION_ACTION_KIND_ABORT] = {
2152 .instance_size = sizeof(BlkActionState),
2153 .prepare = abort_prepare,
2154 .commit = abort_commit,
2156 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2157 .instance_size = sizeof(InternalSnapshotState),
2158 .prepare = internal_snapshot_prepare,
2159 .abort = internal_snapshot_abort,
2160 .clean = internal_snapshot_clean,
2162 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2163 .instance_size = sizeof(BlockDirtyBitmapState),
2164 .prepare = block_dirty_bitmap_add_prepare,
2165 .abort = block_dirty_bitmap_add_abort,
2167 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2168 .instance_size = sizeof(BlockDirtyBitmapState),
2169 .prepare = block_dirty_bitmap_clear_prepare,
2170 .commit = block_dirty_bitmap_clear_commit,
2171 .abort = block_dirty_bitmap_clear_abort,
2172 .clean = block_dirty_bitmap_clear_clean,
2177 * Allocate a TransactionProperties structure if necessary, and fill
2178 * that structure with desired defaults if they are unset.
2180 static TransactionProperties *get_transaction_properties(
2181 TransactionProperties *props)
2183 if (!props) {
2184 props = g_new0(TransactionProperties, 1);
2187 if (!props->has_completion_mode) {
2188 props->has_completion_mode = true;
2189 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2192 return props;
2196 * 'Atomic' group operations. The operations are performed as a set, and if
2197 * any fail then we roll back all operations in the group.
2199 void qmp_transaction(TransactionActionList *dev_list,
2200 bool has_props,
2201 struct TransactionProperties *props,
2202 Error **errp)
2204 TransactionActionList *dev_entry = dev_list;
2205 BlockJobTxn *block_job_txn = NULL;
2206 BlkActionState *state, *next;
2207 Error *local_err = NULL;
2209 QSIMPLEQ_HEAD(snap_bdrv_states, BlkActionState) snap_bdrv_states;
2210 QSIMPLEQ_INIT(&snap_bdrv_states);
2212 /* Does this transaction get canceled as a group on failure?
2213 * If not, we don't really need to make a BlockJobTxn.
2215 props = get_transaction_properties(props);
2216 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2217 block_job_txn = block_job_txn_new();
2220 /* drain all i/o before any operations */
2221 bdrv_drain_all();
2223 /* We don't do anything in this loop that commits us to the operations */
2224 while (NULL != dev_entry) {
2225 TransactionAction *dev_info = NULL;
2226 const BlkActionOps *ops;
2228 dev_info = dev_entry->value;
2229 dev_entry = dev_entry->next;
2231 assert(dev_info->type < ARRAY_SIZE(actions));
2233 ops = &actions[dev_info->type];
2234 assert(ops->instance_size > 0);
2236 state = g_malloc0(ops->instance_size);
2237 state->ops = ops;
2238 state->action = dev_info;
2239 state->block_job_txn = block_job_txn;
2240 state->txn_props = props;
2241 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2243 state->ops->prepare(state, &local_err);
2244 if (local_err) {
2245 error_propagate(errp, local_err);
2246 goto delete_and_fail;
2250 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
2251 if (state->ops->commit) {
2252 state->ops->commit(state);
2256 /* success */
2257 goto exit;
2259 delete_and_fail:
2260 /* failure, and it is all-or-none; roll back all operations */
2261 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
2262 if (state->ops->abort) {
2263 state->ops->abort(state);
2266 exit:
2267 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2268 if (state->ops->clean) {
2269 state->ops->clean(state);
2271 g_free(state);
2273 if (!has_props) {
2274 qapi_free_TransactionProperties(props);
2276 block_job_txn_unref(block_job_txn);
2279 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
2281 Error *local_err = NULL;
2282 int rc;
2284 if (!has_force) {
2285 force = false;
2288 rc = do_open_tray(device, force, &local_err);
2289 if (rc && rc != -ENOSYS) {
2290 error_propagate(errp, local_err);
2291 return;
2293 error_free(local_err);
2295 qmp_x_blockdev_remove_medium(device, errp);
2298 void qmp_block_passwd(bool has_device, const char *device,
2299 bool has_node_name, const char *node_name,
2300 const char *password, Error **errp)
2302 Error *local_err = NULL;
2303 BlockDriverState *bs;
2304 AioContext *aio_context;
2306 bs = bdrv_lookup_bs(has_device ? device : NULL,
2307 has_node_name ? node_name : NULL,
2308 &local_err);
2309 if (local_err) {
2310 error_propagate(errp, local_err);
2311 return;
2314 aio_context = bdrv_get_aio_context(bs);
2315 aio_context_acquire(aio_context);
2317 bdrv_add_key(bs, password, errp);
2319 aio_context_release(aio_context);
2323 * Attempt to open the tray of @device.
2324 * If @force, ignore its tray lock.
2325 * Else, if the tray is locked, don't open it, but ask the guest to open it.
2326 * On error, store an error through @errp and return -errno.
2327 * If @device does not exist, return -ENODEV.
2328 * If it has no removable media, return -ENOTSUP.
2329 * If it has no tray, return -ENOSYS.
2330 * If the guest was asked to open the tray, return -EINPROGRESS.
2331 * Else, return 0.
2333 static int do_open_tray(const char *device, bool force, Error **errp)
2335 BlockBackend *blk;
2336 bool locked;
2338 blk = blk_by_name(device);
2339 if (!blk) {
2340 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2341 "Device '%s' not found", device);
2342 return -ENODEV;
2345 if (!blk_dev_has_removable_media(blk)) {
2346 error_setg(errp, "Device '%s' is not removable", device);
2347 return -ENOTSUP;
2350 if (!blk_dev_has_tray(blk)) {
2351 error_setg(errp, "Device '%s' does not have a tray", device);
2352 return -ENOSYS;
2355 if (blk_dev_is_tray_open(blk)) {
2356 return 0;
2359 locked = blk_dev_is_medium_locked(blk);
2360 if (locked) {
2361 blk_dev_eject_request(blk, force);
2364 if (!locked || force) {
2365 blk_dev_change_media_cb(blk, false);
2368 if (locked && !force) {
2369 error_setg(errp, "Device '%s' is locked and force was not specified, "
2370 "wait for tray to open and try again", device);
2371 return -EINPROGRESS;
2374 return 0;
2377 void qmp_blockdev_open_tray(const char *device, bool has_force, bool force,
2378 Error **errp)
2380 Error *local_err = NULL;
2381 int rc;
2383 if (!has_force) {
2384 force = false;
2386 rc = do_open_tray(device, force, &local_err);
2387 if (rc && rc != -ENOSYS && rc != -EINPROGRESS) {
2388 error_propagate(errp, local_err);
2389 return;
2391 error_free(local_err);
2394 void qmp_blockdev_close_tray(const char *device, Error **errp)
2396 BlockBackend *blk;
2398 blk = blk_by_name(device);
2399 if (!blk) {
2400 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2401 "Device '%s' not found", device);
2402 return;
2405 if (!blk_dev_has_removable_media(blk)) {
2406 error_setg(errp, "Device '%s' is not removable", device);
2407 return;
2410 if (!blk_dev_has_tray(blk)) {
2411 /* Ignore this command on tray-less devices */
2412 return;
2415 if (!blk_dev_is_tray_open(blk)) {
2416 return;
2419 blk_dev_change_media_cb(blk, true);
2422 void qmp_x_blockdev_remove_medium(const char *device, Error **errp)
2424 BlockBackend *blk;
2425 BlockDriverState *bs;
2426 AioContext *aio_context;
2427 bool has_device;
2429 blk = blk_by_name(device);
2430 if (!blk) {
2431 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2432 "Device '%s' not found", device);
2433 return;
2436 /* For BBs without a device, we can exchange the BDS tree at will */
2437 has_device = blk_get_attached_dev(blk);
2439 if (has_device && !blk_dev_has_removable_media(blk)) {
2440 error_setg(errp, "Device '%s' is not removable", device);
2441 return;
2444 if (has_device && blk_dev_has_tray(blk) && !blk_dev_is_tray_open(blk)) {
2445 error_setg(errp, "Tray of device '%s' is not open", device);
2446 return;
2449 bs = blk_bs(blk);
2450 if (!bs) {
2451 return;
2454 aio_context = bdrv_get_aio_context(bs);
2455 aio_context_acquire(aio_context);
2457 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
2458 goto out;
2461 blk_remove_bs(blk);
2463 if (!blk_dev_has_tray(blk)) {
2464 /* For tray-less devices, blockdev-open-tray is a no-op (or may not be
2465 * called at all); therefore, the medium needs to be ejected here.
2466 * Do it after blk_remove_bs() so blk_is_inserted(blk) returns the @load
2467 * value passed here (i.e. false). */
2468 blk_dev_change_media_cb(blk, false);
2471 out:
2472 aio_context_release(aio_context);
2475 static void qmp_blockdev_insert_anon_medium(const char *device,
2476 BlockDriverState *bs, Error **errp)
2478 BlockBackend *blk;
2479 bool has_device;
2481 blk = blk_by_name(device);
2482 if (!blk) {
2483 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2484 "Device '%s' not found", device);
2485 return;
2488 /* For BBs without a device, we can exchange the BDS tree at will */
2489 has_device = blk_get_attached_dev(blk);
2491 if (has_device && !blk_dev_has_removable_media(blk)) {
2492 error_setg(errp, "Device '%s' is not removable", device);
2493 return;
2496 if (has_device && blk_dev_has_tray(blk) && !blk_dev_is_tray_open(blk)) {
2497 error_setg(errp, "Tray of device '%s' is not open", device);
2498 return;
2501 if (blk_bs(blk)) {
2502 error_setg(errp, "There already is a medium in device '%s'", device);
2503 return;
2506 blk_insert_bs(blk, bs);
2508 if (!blk_dev_has_tray(blk)) {
2509 /* For tray-less devices, blockdev-close-tray is a no-op (or may not be
2510 * called at all); therefore, the medium needs to be pushed into the
2511 * slot here.
2512 * Do it after blk_insert_bs() so blk_is_inserted(blk) returns the @load
2513 * value passed here (i.e. true). */
2514 blk_dev_change_media_cb(blk, true);
2518 void qmp_x_blockdev_insert_medium(const char *device, const char *node_name,
2519 Error **errp)
2521 BlockDriverState *bs;
2523 bs = bdrv_find_node(node_name);
2524 if (!bs) {
2525 error_setg(errp, "Node '%s' not found", node_name);
2526 return;
2529 if (bdrv_has_blk(bs)) {
2530 error_setg(errp, "Node '%s' is already in use by '%s'", node_name,
2531 bdrv_get_parent_name(bs));
2532 return;
2535 qmp_blockdev_insert_anon_medium(device, bs, errp);
2538 void qmp_blockdev_change_medium(const char *device, const char *filename,
2539 bool has_format, const char *format,
2540 bool has_read_only,
2541 BlockdevChangeReadOnlyMode read_only,
2542 Error **errp)
2544 BlockBackend *blk;
2545 BlockDriverState *medium_bs = NULL;
2546 int bdrv_flags;
2547 int rc;
2548 QDict *options = NULL;
2549 Error *err = NULL;
2551 blk = blk_by_name(device);
2552 if (!blk) {
2553 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2554 "Device '%s' not found", device);
2555 goto fail;
2558 if (blk_bs(blk)) {
2559 blk_update_root_state(blk);
2562 bdrv_flags = blk_get_open_flags_from_root_state(blk);
2563 bdrv_flags &= ~(BDRV_O_TEMPORARY | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING |
2564 BDRV_O_PROTOCOL);
2566 if (!has_read_only) {
2567 read_only = BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN;
2570 switch (read_only) {
2571 case BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN:
2572 break;
2574 case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_ONLY:
2575 bdrv_flags &= ~BDRV_O_RDWR;
2576 break;
2578 case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_WRITE:
2579 bdrv_flags |= BDRV_O_RDWR;
2580 break;
2582 default:
2583 abort();
2586 if (has_format) {
2587 options = qdict_new();
2588 qdict_put(options, "driver", qstring_from_str(format));
2591 medium_bs = bdrv_open(filename, NULL, options, bdrv_flags, errp);
2592 if (!medium_bs) {
2593 goto fail;
2596 bdrv_add_key(medium_bs, NULL, &err);
2597 if (err) {
2598 error_propagate(errp, err);
2599 goto fail;
2602 rc = do_open_tray(device, false, &err);
2603 if (rc && rc != -ENOSYS) {
2604 error_propagate(errp, err);
2605 goto fail;
2607 error_free(err);
2608 err = NULL;
2610 qmp_x_blockdev_remove_medium(device, &err);
2611 if (err) {
2612 error_propagate(errp, err);
2613 goto fail;
2616 qmp_blockdev_insert_anon_medium(device, medium_bs, &err);
2617 if (err) {
2618 error_propagate(errp, err);
2619 goto fail;
2622 blk_apply_root_state(blk, medium_bs);
2624 qmp_blockdev_close_tray(device, errp);
2626 fail:
2627 /* If the medium has been inserted, the device has its own reference, so
2628 * ours must be relinquished; and if it has not been inserted successfully,
2629 * the reference must be relinquished anyway */
2630 bdrv_unref(medium_bs);
2633 /* throttling disk I/O limits */
2634 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
2635 int64_t bps_wr,
2636 int64_t iops,
2637 int64_t iops_rd,
2638 int64_t iops_wr,
2639 bool has_bps_max,
2640 int64_t bps_max,
2641 bool has_bps_rd_max,
2642 int64_t bps_rd_max,
2643 bool has_bps_wr_max,
2644 int64_t bps_wr_max,
2645 bool has_iops_max,
2646 int64_t iops_max,
2647 bool has_iops_rd_max,
2648 int64_t iops_rd_max,
2649 bool has_iops_wr_max,
2650 int64_t iops_wr_max,
2651 bool has_bps_max_length,
2652 int64_t bps_max_length,
2653 bool has_bps_rd_max_length,
2654 int64_t bps_rd_max_length,
2655 bool has_bps_wr_max_length,
2656 int64_t bps_wr_max_length,
2657 bool has_iops_max_length,
2658 int64_t iops_max_length,
2659 bool has_iops_rd_max_length,
2660 int64_t iops_rd_max_length,
2661 bool has_iops_wr_max_length,
2662 int64_t iops_wr_max_length,
2663 bool has_iops_size,
2664 int64_t iops_size,
2665 bool has_group,
2666 const char *group, Error **errp)
2668 ThrottleConfig cfg;
2669 BlockDriverState *bs;
2670 BlockBackend *blk;
2671 AioContext *aio_context;
2673 blk = blk_by_name(device);
2674 if (!blk) {
2675 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2676 "Device '%s' not found", device);
2677 return;
2680 aio_context = blk_get_aio_context(blk);
2681 aio_context_acquire(aio_context);
2683 bs = blk_bs(blk);
2684 if (!bs) {
2685 error_setg(errp, "Device '%s' has no medium", device);
2686 goto out;
2689 throttle_config_init(&cfg);
2690 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
2691 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
2692 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
2694 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
2695 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
2696 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
2698 if (has_bps_max) {
2699 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
2701 if (has_bps_rd_max) {
2702 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
2704 if (has_bps_wr_max) {
2705 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
2707 if (has_iops_max) {
2708 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
2710 if (has_iops_rd_max) {
2711 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
2713 if (has_iops_wr_max) {
2714 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
2717 if (has_bps_max_length) {
2718 cfg.buckets[THROTTLE_BPS_TOTAL].burst_length = bps_max_length;
2720 if (has_bps_rd_max_length) {
2721 cfg.buckets[THROTTLE_BPS_READ].burst_length = bps_rd_max_length;
2723 if (has_bps_wr_max_length) {
2724 cfg.buckets[THROTTLE_BPS_WRITE].burst_length = bps_wr_max_length;
2726 if (has_iops_max_length) {
2727 cfg.buckets[THROTTLE_OPS_TOTAL].burst_length = iops_max_length;
2729 if (has_iops_rd_max_length) {
2730 cfg.buckets[THROTTLE_OPS_READ].burst_length = iops_rd_max_length;
2732 if (has_iops_wr_max_length) {
2733 cfg.buckets[THROTTLE_OPS_WRITE].burst_length = iops_wr_max_length;
2736 if (has_iops_size) {
2737 cfg.op_size = iops_size;
2740 if (!throttle_is_valid(&cfg, errp)) {
2741 goto out;
2744 if (throttle_enabled(&cfg)) {
2745 /* Enable I/O limits if they're not enabled yet, otherwise
2746 * just update the throttling group. */
2747 if (!blk_get_public(blk)->throttle_state) {
2748 blk_io_limits_enable(blk, has_group ? group : device);
2749 } else if (has_group) {
2750 blk_io_limits_update_group(blk, group);
2752 /* Set the new throttling configuration */
2753 blk_set_io_limits(blk, &cfg);
2754 } else if (blk_get_public(blk)->throttle_state) {
2755 /* If all throttling settings are set to 0, disable I/O limits */
2756 blk_io_limits_disable(blk);
2759 out:
2760 aio_context_release(aio_context);
2763 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2764 bool has_granularity, uint32_t granularity,
2765 Error **errp)
2767 AioContext *aio_context;
2768 BlockDriverState *bs;
2770 if (!name || name[0] == '\0') {
2771 error_setg(errp, "Bitmap name cannot be empty");
2772 return;
2775 bs = bdrv_lookup_bs(node, node, errp);
2776 if (!bs) {
2777 return;
2780 aio_context = bdrv_get_aio_context(bs);
2781 aio_context_acquire(aio_context);
2783 if (has_granularity) {
2784 if (granularity < 512 || !is_power_of_2(granularity)) {
2785 error_setg(errp, "Granularity must be power of 2 "
2786 "and at least 512");
2787 goto out;
2789 } else {
2790 /* Default to cluster size, if available: */
2791 granularity = bdrv_get_default_bitmap_granularity(bs);
2794 bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2796 out:
2797 aio_context_release(aio_context);
2800 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2801 Error **errp)
2803 AioContext *aio_context;
2804 BlockDriverState *bs;
2805 BdrvDirtyBitmap *bitmap;
2807 bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2808 if (!bitmap || !bs) {
2809 return;
2812 if (bdrv_dirty_bitmap_frozen(bitmap)) {
2813 error_setg(errp,
2814 "Bitmap '%s' is currently frozen and cannot be removed",
2815 name);
2816 goto out;
2818 bdrv_dirty_bitmap_make_anon(bitmap);
2819 bdrv_release_dirty_bitmap(bs, bitmap);
2821 out:
2822 aio_context_release(aio_context);
2826 * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2827 * immediately after a full backup operation.
2829 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2830 Error **errp)
2832 AioContext *aio_context;
2833 BdrvDirtyBitmap *bitmap;
2834 BlockDriverState *bs;
2836 bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2837 if (!bitmap || !bs) {
2838 return;
2841 if (bdrv_dirty_bitmap_frozen(bitmap)) {
2842 error_setg(errp,
2843 "Bitmap '%s' is currently frozen and cannot be modified",
2844 name);
2845 goto out;
2846 } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
2847 error_setg(errp,
2848 "Bitmap '%s' is currently disabled and cannot be cleared",
2849 name);
2850 goto out;
2853 bdrv_clear_dirty_bitmap(bitmap, NULL);
2855 out:
2856 aio_context_release(aio_context);
2859 void hmp_drive_del(Monitor *mon, const QDict *qdict)
2861 const char *id = qdict_get_str(qdict, "id");
2862 BlockBackend *blk;
2863 BlockDriverState *bs;
2864 AioContext *aio_context;
2865 Error *local_err = NULL;
2867 bs = bdrv_find_node(id);
2868 if (bs) {
2869 qmp_x_blockdev_del(false, NULL, true, id, &local_err);
2870 if (local_err) {
2871 error_report_err(local_err);
2873 return;
2876 blk = blk_by_name(id);
2877 if (!blk) {
2878 error_report("Device '%s' not found", id);
2879 return;
2882 if (!blk_legacy_dinfo(blk)) {
2883 error_report("Deleting device added with blockdev-add"
2884 " is not supported");
2885 return;
2888 aio_context = blk_get_aio_context(blk);
2889 aio_context_acquire(aio_context);
2891 bs = blk_bs(blk);
2892 if (bs) {
2893 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2894 error_report_err(local_err);
2895 aio_context_release(aio_context);
2896 return;
2899 blk_remove_bs(blk);
2902 /* Make the BlockBackend and the attached BlockDriverState anonymous */
2903 monitor_remove_blk(blk);
2905 /* If this BlockBackend has a device attached to it, its refcount will be
2906 * decremented when the device is removed; otherwise we have to do so here.
2908 if (blk_get_attached_dev(blk)) {
2909 /* Further I/O must not pause the guest */
2910 blk_set_on_error(blk, BLOCKDEV_ON_ERROR_REPORT,
2911 BLOCKDEV_ON_ERROR_REPORT);
2912 } else {
2913 blk_unref(blk);
2916 aio_context_release(aio_context);
2919 void qmp_block_resize(bool has_device, const char *device,
2920 bool has_node_name, const char *node_name,
2921 int64_t size, Error **errp)
2923 Error *local_err = NULL;
2924 BlockDriverState *bs;
2925 AioContext *aio_context;
2926 int ret;
2928 bs = bdrv_lookup_bs(has_device ? device : NULL,
2929 has_node_name ? node_name : NULL,
2930 &local_err);
2931 if (local_err) {
2932 error_propagate(errp, local_err);
2933 return;
2936 aio_context = bdrv_get_aio_context(bs);
2937 aio_context_acquire(aio_context);
2939 if (!bdrv_is_first_non_filter(bs)) {
2940 error_setg(errp, QERR_FEATURE_DISABLED, "resize");
2941 goto out;
2944 if (size < 0) {
2945 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2946 goto out;
2949 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2950 error_setg(errp, QERR_DEVICE_IN_USE, device);
2951 goto out;
2954 /* complete all in-flight operations before resizing the device */
2955 bdrv_drain_all();
2957 ret = bdrv_truncate(bs, size);
2958 switch (ret) {
2959 case 0:
2960 break;
2961 case -ENOMEDIUM:
2962 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2963 break;
2964 case -ENOTSUP:
2965 error_setg(errp, QERR_UNSUPPORTED);
2966 break;
2967 case -EACCES:
2968 error_setg(errp, "Device '%s' is read only", device);
2969 break;
2970 case -EBUSY:
2971 error_setg(errp, QERR_DEVICE_IN_USE, device);
2972 break;
2973 default:
2974 error_setg_errno(errp, -ret, "Could not resize");
2975 break;
2978 out:
2979 aio_context_release(aio_context);
2982 static void block_job_cb(void *opaque, int ret)
2984 /* Note that this function may be executed from another AioContext besides
2985 * the QEMU main loop. If you need to access anything that assumes the
2986 * QEMU global mutex, use a BH or introduce a mutex.
2989 BlockDriverState *bs = opaque;
2990 const char *msg = NULL;
2992 trace_block_job_cb(bs, bs->job, ret);
2994 assert(bs->job);
2996 if (ret < 0) {
2997 msg = strerror(-ret);
3000 if (block_job_is_cancelled(bs->job)) {
3001 block_job_event_cancelled(bs->job);
3002 } else {
3003 block_job_event_completed(bs->job, msg);
3007 void qmp_block_stream(const char *device,
3008 bool has_base, const char *base,
3009 bool has_backing_file, const char *backing_file,
3010 bool has_speed, int64_t speed,
3011 bool has_on_error, BlockdevOnError on_error,
3012 Error **errp)
3014 BlockBackend *blk;
3015 BlockDriverState *bs;
3016 BlockDriverState *base_bs = NULL;
3017 AioContext *aio_context;
3018 Error *local_err = NULL;
3019 const char *base_name = NULL;
3021 if (!has_on_error) {
3022 on_error = BLOCKDEV_ON_ERROR_REPORT;
3025 blk = blk_by_name(device);
3026 if (!blk) {
3027 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3028 "Device '%s' not found", device);
3029 return;
3032 aio_context = blk_get_aio_context(blk);
3033 aio_context_acquire(aio_context);
3035 if (!blk_is_available(blk)) {
3036 error_setg(errp, "Device '%s' has no medium", device);
3037 goto out;
3039 bs = blk_bs(blk);
3041 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
3042 goto out;
3045 if (has_base) {
3046 base_bs = bdrv_find_backing_image(bs, base);
3047 if (base_bs == NULL) {
3048 error_setg(errp, QERR_BASE_NOT_FOUND, base);
3049 goto out;
3051 assert(bdrv_get_aio_context(base_bs) == aio_context);
3052 base_name = base;
3055 /* if we are streaming the entire chain, the result will have no backing
3056 * file, and specifying one is therefore an error */
3057 if (base_bs == NULL && has_backing_file) {
3058 error_setg(errp, "backing file specified, but streaming the "
3059 "entire chain");
3060 goto out;
3063 /* backing_file string overrides base bs filename */
3064 base_name = has_backing_file ? backing_file : base_name;
3066 stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
3067 on_error, block_job_cb, bs, &local_err);
3068 if (local_err) {
3069 error_propagate(errp, local_err);
3070 goto out;
3073 trace_qmp_block_stream(bs, bs->job);
3075 out:
3076 aio_context_release(aio_context);
3079 void qmp_block_commit(const char *device,
3080 bool has_base, const char *base,
3081 bool has_top, const char *top,
3082 bool has_backing_file, const char *backing_file,
3083 bool has_speed, int64_t speed,
3084 Error **errp)
3086 BlockBackend *blk;
3087 BlockDriverState *bs;
3088 BlockDriverState *base_bs, *top_bs;
3089 AioContext *aio_context;
3090 Error *local_err = NULL;
3091 /* This will be part of the QMP command, if/when the
3092 * BlockdevOnError change for blkmirror makes it in
3094 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
3096 if (!has_speed) {
3097 speed = 0;
3100 /* Important Note:
3101 * libvirt relies on the DeviceNotFound error class in order to probe for
3102 * live commit feature versions; for this to work, we must make sure to
3103 * perform the device lookup before any generic errors that may occur in a
3104 * scenario in which all optional arguments are omitted. */
3105 blk = blk_by_name(device);
3106 if (!blk) {
3107 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3108 "Device '%s' not found", device);
3109 return;
3112 aio_context = blk_get_aio_context(blk);
3113 aio_context_acquire(aio_context);
3115 if (!blk_is_available(blk)) {
3116 error_setg(errp, "Device '%s' has no medium", device);
3117 goto out;
3119 bs = blk_bs(blk);
3121 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
3122 goto out;
3125 /* default top_bs is the active layer */
3126 top_bs = bs;
3128 if (has_top && top) {
3129 if (strcmp(bs->filename, top) != 0) {
3130 top_bs = bdrv_find_backing_image(bs, top);
3134 if (top_bs == NULL) {
3135 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
3136 goto out;
3139 assert(bdrv_get_aio_context(top_bs) == aio_context);
3141 if (has_base && base) {
3142 base_bs = bdrv_find_backing_image(top_bs, base);
3143 } else {
3144 base_bs = bdrv_find_base(top_bs);
3147 if (base_bs == NULL) {
3148 error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
3149 goto out;
3152 assert(bdrv_get_aio_context(base_bs) == aio_context);
3154 if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
3155 goto out;
3158 /* Do not allow attempts to commit an image into itself */
3159 if (top_bs == base_bs) {
3160 error_setg(errp, "cannot commit an image into itself");
3161 goto out;
3164 if (top_bs == bs) {
3165 if (has_backing_file) {
3166 error_setg(errp, "'backing-file' specified,"
3167 " but 'top' is the active layer");
3168 goto out;
3170 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
3171 bs, &local_err);
3172 } else {
3173 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
3174 has_backing_file ? backing_file : NULL, &local_err);
3176 if (local_err != NULL) {
3177 error_propagate(errp, local_err);
3178 goto out;
3181 out:
3182 aio_context_release(aio_context);
3185 static void do_drive_backup(const char *device, const char *target,
3186 bool has_format, const char *format,
3187 enum MirrorSyncMode sync,
3188 bool has_mode, enum NewImageMode mode,
3189 bool has_speed, int64_t speed,
3190 bool has_bitmap, const char *bitmap,
3191 bool has_on_source_error,
3192 BlockdevOnError on_source_error,
3193 bool has_on_target_error,
3194 BlockdevOnError on_target_error,
3195 BlockJobTxn *txn, Error **errp)
3197 BlockBackend *blk;
3198 BlockDriverState *bs;
3199 BlockDriverState *target_bs;
3200 BlockDriverState *source = NULL;
3201 BdrvDirtyBitmap *bmap = NULL;
3202 AioContext *aio_context;
3203 QDict *options = NULL;
3204 Error *local_err = NULL;
3205 int flags;
3206 int64_t size;
3208 if (!has_speed) {
3209 speed = 0;
3211 if (!has_on_source_error) {
3212 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3214 if (!has_on_target_error) {
3215 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3217 if (!has_mode) {
3218 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3221 blk = blk_by_name(device);
3222 if (!blk) {
3223 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3224 "Device '%s' not found", device);
3225 return;
3228 aio_context = blk_get_aio_context(blk);
3229 aio_context_acquire(aio_context);
3231 /* Although backup_run has this check too, we need to use bs->drv below, so
3232 * do an early check redundantly. */
3233 if (!blk_is_available(blk)) {
3234 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
3235 goto out;
3237 bs = blk_bs(blk);
3239 if (!has_format) {
3240 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
3243 /* Early check to avoid creating target */
3244 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
3245 goto out;
3248 flags = bs->open_flags | BDRV_O_RDWR;
3250 /* See if we have a backing HD we can use to create our new image
3251 * on top of. */
3252 if (sync == MIRROR_SYNC_MODE_TOP) {
3253 source = backing_bs(bs);
3254 if (!source) {
3255 sync = MIRROR_SYNC_MODE_FULL;
3258 if (sync == MIRROR_SYNC_MODE_NONE) {
3259 source = bs;
3262 size = bdrv_getlength(bs);
3263 if (size < 0) {
3264 error_setg_errno(errp, -size, "bdrv_getlength failed");
3265 goto out;
3268 if (mode != NEW_IMAGE_MODE_EXISTING) {
3269 assert(format);
3270 if (source) {
3271 bdrv_img_create(target, format, source->filename,
3272 source->drv->format_name, NULL,
3273 size, flags, &local_err, false);
3274 } else {
3275 bdrv_img_create(target, format, NULL, NULL, NULL,
3276 size, flags, &local_err, false);
3280 if (local_err) {
3281 error_propagate(errp, local_err);
3282 goto out;
3285 if (format) {
3286 options = qdict_new();
3287 qdict_put(options, "driver", qstring_from_str(format));
3290 target_bs = bdrv_open(target, NULL, options, flags, errp);
3291 if (!target_bs) {
3292 goto out;
3295 bdrv_set_aio_context(target_bs, aio_context);
3297 if (has_bitmap) {
3298 bmap = bdrv_find_dirty_bitmap(bs, bitmap);
3299 if (!bmap) {
3300 error_setg(errp, "Bitmap '%s' could not be found", bitmap);
3301 bdrv_unref(target_bs);
3302 goto out;
3306 backup_start(bs, target_bs, speed, sync, bmap,
3307 on_source_error, on_target_error,
3308 block_job_cb, bs, txn, &local_err);
3309 bdrv_unref(target_bs);
3310 if (local_err != NULL) {
3311 error_propagate(errp, local_err);
3312 goto out;
3315 out:
3316 aio_context_release(aio_context);
3319 void qmp_drive_backup(const char *device, const char *target,
3320 bool has_format, const char *format,
3321 enum MirrorSyncMode sync,
3322 bool has_mode, enum NewImageMode mode,
3323 bool has_speed, int64_t speed,
3324 bool has_bitmap, const char *bitmap,
3325 bool has_on_source_error, BlockdevOnError on_source_error,
3326 bool has_on_target_error, BlockdevOnError on_target_error,
3327 Error **errp)
3329 return do_drive_backup(device, target, has_format, format, sync,
3330 has_mode, mode, has_speed, speed,
3331 has_bitmap, bitmap,
3332 has_on_source_error, on_source_error,
3333 has_on_target_error, on_target_error,
3334 NULL, errp);
3337 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
3339 return bdrv_named_nodes_list(errp);
3342 void do_blockdev_backup(const char *device, const char *target,
3343 enum MirrorSyncMode sync,
3344 bool has_speed, int64_t speed,
3345 bool has_on_source_error,
3346 BlockdevOnError on_source_error,
3347 bool has_on_target_error,
3348 BlockdevOnError on_target_error,
3349 BlockJobTxn *txn, Error **errp)
3351 BlockBackend *blk;
3352 BlockDriverState *bs;
3353 BlockDriverState *target_bs;
3354 Error *local_err = NULL;
3355 AioContext *aio_context;
3357 if (!has_speed) {
3358 speed = 0;
3360 if (!has_on_source_error) {
3361 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3363 if (!has_on_target_error) {
3364 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3367 blk = blk_by_name(device);
3368 if (!blk) {
3369 error_setg(errp, "Device '%s' not found", device);
3370 return;
3373 aio_context = blk_get_aio_context(blk);
3374 aio_context_acquire(aio_context);
3376 if (!blk_is_available(blk)) {
3377 error_setg(errp, "Device '%s' has no medium", device);
3378 goto out;
3380 bs = blk_bs(blk);
3382 target_bs = bdrv_lookup_bs(target, target, errp);
3383 if (!target_bs) {
3384 goto out;
3387 if (bdrv_get_aio_context(target_bs) != aio_context) {
3388 if (!bdrv_has_blk(target_bs)) {
3389 /* The target BDS is not attached, we can safely move it to another
3390 * AioContext. */
3391 bdrv_set_aio_context(target_bs, aio_context);
3392 } else {
3393 error_setg(errp, "Target is attached to a different thread from "
3394 "source.");
3395 goto out;
3398 backup_start(bs, target_bs, speed, sync, NULL, on_source_error,
3399 on_target_error, block_job_cb, bs, txn, &local_err);
3400 if (local_err != NULL) {
3401 error_propagate(errp, local_err);
3403 out:
3404 aio_context_release(aio_context);
3407 void qmp_blockdev_backup(const char *device, const char *target,
3408 enum MirrorSyncMode sync,
3409 bool has_speed, int64_t speed,
3410 bool has_on_source_error,
3411 BlockdevOnError on_source_error,
3412 bool has_on_target_error,
3413 BlockdevOnError on_target_error,
3414 Error **errp)
3416 do_blockdev_backup(device, target, sync, has_speed, speed,
3417 has_on_source_error, on_source_error,
3418 has_on_target_error, on_target_error,
3419 NULL, errp);
3422 /* Parameter check and block job starting for drive mirroring.
3423 * Caller should hold @device and @target's aio context (must be the same).
3425 static void blockdev_mirror_common(BlockDriverState *bs,
3426 BlockDriverState *target,
3427 bool has_replaces, const char *replaces,
3428 enum MirrorSyncMode sync,
3429 BlockMirrorBackingMode backing_mode,
3430 bool has_speed, int64_t speed,
3431 bool has_granularity, uint32_t granularity,
3432 bool has_buf_size, int64_t buf_size,
3433 bool has_on_source_error,
3434 BlockdevOnError on_source_error,
3435 bool has_on_target_error,
3436 BlockdevOnError on_target_error,
3437 bool has_unmap, bool unmap,
3438 Error **errp)
3441 if (!has_speed) {
3442 speed = 0;
3444 if (!has_on_source_error) {
3445 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3447 if (!has_on_target_error) {
3448 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3450 if (!has_granularity) {
3451 granularity = 0;
3453 if (!has_buf_size) {
3454 buf_size = 0;
3456 if (!has_unmap) {
3457 unmap = true;
3460 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3461 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3462 "a value in range [512B, 64MB]");
3463 return;
3465 if (granularity & (granularity - 1)) {
3466 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3467 "power of 2");
3468 return;
3471 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3472 return;
3474 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3475 return;
3478 if (!bs->backing && sync == MIRROR_SYNC_MODE_TOP) {
3479 sync = MIRROR_SYNC_MODE_FULL;
3482 /* pass the node name to replace to mirror start since it's loose coupling
3483 * and will allow to check whether the node still exist at mirror completion
3485 mirror_start(bs, target,
3486 has_replaces ? replaces : NULL,
3487 speed, granularity, buf_size, sync, backing_mode,
3488 on_source_error, on_target_error, unmap,
3489 block_job_cb, bs, errp);
3492 void qmp_drive_mirror(const char *device, const char *target,
3493 bool has_format, const char *format,
3494 bool has_node_name, const char *node_name,
3495 bool has_replaces, const char *replaces,
3496 enum MirrorSyncMode sync,
3497 bool has_mode, enum NewImageMode mode,
3498 bool has_speed, int64_t speed,
3499 bool has_granularity, uint32_t granularity,
3500 bool has_buf_size, int64_t buf_size,
3501 bool has_on_source_error, BlockdevOnError on_source_error,
3502 bool has_on_target_error, BlockdevOnError on_target_error,
3503 bool has_unmap, bool unmap,
3504 Error **errp)
3506 BlockDriverState *bs;
3507 BlockBackend *blk;
3508 BlockDriverState *source, *target_bs;
3509 AioContext *aio_context;
3510 BlockMirrorBackingMode backing_mode;
3511 Error *local_err = NULL;
3512 QDict *options = NULL;
3513 int flags;
3514 int64_t size;
3516 blk = blk_by_name(device);
3517 if (!blk) {
3518 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3519 "Device '%s' not found", device);
3520 return;
3523 aio_context = blk_get_aio_context(blk);
3524 aio_context_acquire(aio_context);
3526 if (!blk_is_available(blk)) {
3527 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
3528 goto out;
3530 bs = blk_bs(blk);
3531 if (!has_mode) {
3532 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3535 if (!has_format) {
3536 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
3539 flags = bs->open_flags | BDRV_O_RDWR;
3540 source = backing_bs(bs);
3541 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
3542 sync = MIRROR_SYNC_MODE_FULL;
3544 if (sync == MIRROR_SYNC_MODE_NONE) {
3545 source = bs;
3548 size = bdrv_getlength(bs);
3549 if (size < 0) {
3550 error_setg_errno(errp, -size, "bdrv_getlength failed");
3551 goto out;
3554 if (has_replaces) {
3555 BlockDriverState *to_replace_bs;
3556 AioContext *replace_aio_context;
3557 int64_t replace_size;
3559 if (!has_node_name) {
3560 error_setg(errp, "a node-name must be provided when replacing a"
3561 " named node of the graph");
3562 goto out;
3565 to_replace_bs = check_to_replace_node(bs, replaces, &local_err);
3567 if (!to_replace_bs) {
3568 error_propagate(errp, local_err);
3569 goto out;
3572 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3573 aio_context_acquire(replace_aio_context);
3574 replace_size = bdrv_getlength(to_replace_bs);
3575 aio_context_release(replace_aio_context);
3577 if (size != replace_size) {
3578 error_setg(errp, "cannot replace image with a mirror image of "
3579 "different size");
3580 goto out;
3584 if (mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3585 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3586 } else {
3587 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3590 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
3591 && mode != NEW_IMAGE_MODE_EXISTING)
3593 /* create new image w/o backing file */
3594 assert(format);
3595 bdrv_img_create(target, format,
3596 NULL, NULL, NULL, size, flags, &local_err, false);
3597 } else {
3598 switch (mode) {
3599 case NEW_IMAGE_MODE_EXISTING:
3600 break;
3601 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3602 /* create new image with backing file */
3603 bdrv_img_create(target, format,
3604 source->filename,
3605 source->drv->format_name,
3606 NULL, size, flags, &local_err, false);
3607 break;
3608 default:
3609 abort();
3613 if (local_err) {
3614 error_propagate(errp, local_err);
3615 goto out;
3618 options = qdict_new();
3619 if (has_node_name) {
3620 qdict_put(options, "node-name", qstring_from_str(node_name));
3622 if (format) {
3623 qdict_put(options, "driver", qstring_from_str(format));
3626 /* Mirroring takes care of copy-on-write using the source's backing
3627 * file.
3629 target_bs = bdrv_open(target, NULL, options, flags | BDRV_O_NO_BACKING,
3630 errp);
3631 if (!target_bs) {
3632 goto out;
3635 bdrv_set_aio_context(target_bs, aio_context);
3637 blockdev_mirror_common(bs, target_bs,
3638 has_replaces, replaces, sync, backing_mode,
3639 has_speed, speed,
3640 has_granularity, granularity,
3641 has_buf_size, buf_size,
3642 has_on_source_error, on_source_error,
3643 has_on_target_error, on_target_error,
3644 has_unmap, unmap,
3645 &local_err);
3646 bdrv_unref(target_bs);
3647 error_propagate(errp, local_err);
3648 out:
3649 aio_context_release(aio_context);
3652 void qmp_blockdev_mirror(const char *device, const char *target,
3653 bool has_replaces, const char *replaces,
3654 MirrorSyncMode sync,
3655 bool has_speed, int64_t speed,
3656 bool has_granularity, uint32_t granularity,
3657 bool has_buf_size, int64_t buf_size,
3658 bool has_on_source_error,
3659 BlockdevOnError on_source_error,
3660 bool has_on_target_error,
3661 BlockdevOnError on_target_error,
3662 Error **errp)
3664 BlockDriverState *bs;
3665 BlockBackend *blk;
3666 BlockDriverState *target_bs;
3667 AioContext *aio_context;
3668 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3669 Error *local_err = NULL;
3671 blk = blk_by_name(device);
3672 if (!blk) {
3673 error_setg(errp, "Device '%s' not found", device);
3674 return;
3676 bs = blk_bs(blk);
3678 if (!bs) {
3679 error_setg(errp, "Device '%s' has no media", device);
3680 return;
3683 target_bs = bdrv_lookup_bs(target, target, errp);
3684 if (!target_bs) {
3685 return;
3688 aio_context = bdrv_get_aio_context(bs);
3689 aio_context_acquire(aio_context);
3691 bdrv_set_aio_context(target_bs, aio_context);
3693 blockdev_mirror_common(bs, target_bs,
3694 has_replaces, replaces, sync, backing_mode,
3695 has_speed, speed,
3696 has_granularity, granularity,
3697 has_buf_size, buf_size,
3698 has_on_source_error, on_source_error,
3699 has_on_target_error, on_target_error,
3700 true, true,
3701 &local_err);
3702 error_propagate(errp, local_err);
3704 aio_context_release(aio_context);
3707 /* Get the block job for a given device name and acquire its AioContext */
3708 static BlockJob *find_block_job(const char *device, AioContext **aio_context,
3709 Error **errp)
3711 BlockBackend *blk;
3712 BlockDriverState *bs;
3714 *aio_context = NULL;
3716 blk = blk_by_name(device);
3717 if (!blk) {
3718 goto notfound;
3721 *aio_context = blk_get_aio_context(blk);
3722 aio_context_acquire(*aio_context);
3724 if (!blk_is_available(blk)) {
3725 goto notfound;
3727 bs = blk_bs(blk);
3729 if (!bs->job) {
3730 goto notfound;
3733 return bs->job;
3735 notfound:
3736 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3737 "No active block job on device '%s'", device);
3738 if (*aio_context) {
3739 aio_context_release(*aio_context);
3740 *aio_context = NULL;
3742 return NULL;
3745 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3747 AioContext *aio_context;
3748 BlockJob *job = find_block_job(device, &aio_context, errp);
3750 if (!job) {
3751 return;
3754 block_job_set_speed(job, speed, errp);
3755 aio_context_release(aio_context);
3758 void qmp_block_job_cancel(const char *device,
3759 bool has_force, bool force, Error **errp)
3761 AioContext *aio_context;
3762 BlockJob *job = find_block_job(device, &aio_context, errp);
3764 if (!job) {
3765 return;
3768 if (!has_force) {
3769 force = false;
3772 if (job->user_paused && !force) {
3773 error_setg(errp, "The block job for device '%s' is currently paused",
3774 device);
3775 goto out;
3778 trace_qmp_block_job_cancel(job);
3779 block_job_cancel(job);
3780 out:
3781 aio_context_release(aio_context);
3784 void qmp_block_job_pause(const char *device, Error **errp)
3786 AioContext *aio_context;
3787 BlockJob *job = find_block_job(device, &aio_context, errp);
3789 if (!job || job->user_paused) {
3790 return;
3793 job->user_paused = true;
3794 trace_qmp_block_job_pause(job);
3795 block_job_pause(job);
3796 aio_context_release(aio_context);
3799 void qmp_block_job_resume(const char *device, Error **errp)
3801 AioContext *aio_context;
3802 BlockJob *job = find_block_job(device, &aio_context, errp);
3804 if (!job || !job->user_paused) {
3805 return;
3808 job->user_paused = false;
3809 trace_qmp_block_job_resume(job);
3810 block_job_iostatus_reset(job);
3811 block_job_resume(job);
3812 aio_context_release(aio_context);
3815 void qmp_block_job_complete(const char *device, Error **errp)
3817 AioContext *aio_context;
3818 BlockJob *job = find_block_job(device, &aio_context, errp);
3820 if (!job) {
3821 return;
3824 trace_qmp_block_job_complete(job);
3825 block_job_complete(job, errp);
3826 aio_context_release(aio_context);
3829 void qmp_change_backing_file(const char *device,
3830 const char *image_node_name,
3831 const char *backing_file,
3832 Error **errp)
3834 BlockBackend *blk;
3835 BlockDriverState *bs = NULL;
3836 AioContext *aio_context;
3837 BlockDriverState *image_bs = NULL;
3838 Error *local_err = NULL;
3839 bool ro;
3840 int open_flags;
3841 int ret;
3843 blk = blk_by_name(device);
3844 if (!blk) {
3845 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3846 "Device '%s' not found", device);
3847 return;
3850 aio_context = blk_get_aio_context(blk);
3851 aio_context_acquire(aio_context);
3853 if (!blk_is_available(blk)) {
3854 error_setg(errp, "Device '%s' has no medium", device);
3855 goto out;
3857 bs = blk_bs(blk);
3859 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3860 if (local_err) {
3861 error_propagate(errp, local_err);
3862 goto out;
3865 if (!image_bs) {
3866 error_setg(errp, "image file not found");
3867 goto out;
3870 if (bdrv_find_base(image_bs) == image_bs) {
3871 error_setg(errp, "not allowing backing file change on an image "
3872 "without a backing file");
3873 goto out;
3876 /* even though we are not necessarily operating on bs, we need it to
3877 * determine if block ops are currently prohibited on the chain */
3878 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3879 goto out;
3882 /* final sanity check */
3883 if (!bdrv_chain_contains(bs, image_bs)) {
3884 error_setg(errp, "'%s' and image file are not in the same chain",
3885 device);
3886 goto out;
3889 /* if not r/w, reopen to make r/w */
3890 open_flags = image_bs->open_flags;
3891 ro = bdrv_is_read_only(image_bs);
3893 if (ro) {
3894 bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
3895 if (local_err) {
3896 error_propagate(errp, local_err);
3897 goto out;
3901 ret = bdrv_change_backing_file(image_bs, backing_file,
3902 image_bs->drv ? image_bs->drv->format_name : "");
3904 if (ret < 0) {
3905 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3906 backing_file);
3907 /* don't exit here, so we can try to restore open flags if
3908 * appropriate */
3911 if (ro) {
3912 bdrv_reopen(image_bs, open_flags, &local_err);
3913 error_propagate(errp, local_err);
3916 out:
3917 aio_context_release(aio_context);
3920 void hmp_drive_add_node(Monitor *mon, const char *optstr)
3922 QemuOpts *opts;
3923 QDict *qdict;
3924 Error *local_err = NULL;
3926 opts = qemu_opts_parse_noisily(&qemu_drive_opts, optstr, false);
3927 if (!opts) {
3928 return;
3931 qdict = qemu_opts_to_qdict(opts, NULL);
3933 if (!qdict_get_try_str(qdict, "node-name")) {
3934 QDECREF(qdict);
3935 error_report("'node-name' needs to be specified");
3936 goto out;
3939 BlockDriverState *bs = bds_tree_init(qdict, &local_err);
3940 if (!bs) {
3941 error_report_err(local_err);
3942 goto out;
3945 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
3947 out:
3948 qemu_opts_del(opts);
3951 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3953 QmpOutputVisitor *ov = qmp_output_visitor_new();
3954 BlockDriverState *bs;
3955 BlockBackend *blk = NULL;
3956 QObject *obj;
3957 QDict *qdict;
3958 Error *local_err = NULL;
3960 /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
3961 * cache.direct=false instead of silently switching to aio=threads, except
3962 * when called from drive_new().
3964 * For now, simply forbidding the combination for all drivers will do. */
3965 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
3966 bool direct = options->has_cache &&
3967 options->cache->has_direct &&
3968 options->cache->direct;
3969 if (!direct) {
3970 error_setg(errp, "aio=native requires cache.direct=true");
3971 goto fail;
3975 visit_type_BlockdevOptions(qmp_output_get_visitor(ov), NULL, &options,
3976 &local_err);
3977 if (local_err) {
3978 error_propagate(errp, local_err);
3979 goto fail;
3982 obj = qmp_output_get_qobject(ov);
3983 qdict = qobject_to_qdict(obj);
3985 qdict_flatten(qdict);
3987 if (options->has_id) {
3988 blk = blockdev_init(NULL, qdict, &local_err);
3989 if (local_err) {
3990 error_propagate(errp, local_err);
3991 goto fail;
3994 bs = blk_bs(blk);
3995 } else {
3996 if (!qdict_get_try_str(qdict, "node-name")) {
3997 error_setg(errp, "'id' and/or 'node-name' need to be specified for "
3998 "the root node");
3999 goto fail;
4002 bs = bds_tree_init(qdict, errp);
4003 if (!bs) {
4004 goto fail;
4007 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
4010 if (bs && bdrv_key_required(bs)) {
4011 if (blk) {
4012 monitor_remove_blk(blk);
4013 blk_unref(blk);
4014 } else {
4015 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
4016 bdrv_unref(bs);
4018 error_setg(errp, "blockdev-add doesn't support encrypted devices");
4019 goto fail;
4022 fail:
4023 qmp_output_visitor_cleanup(ov);
4026 void qmp_x_blockdev_del(bool has_id, const char *id,
4027 bool has_node_name, const char *node_name, Error **errp)
4029 AioContext *aio_context;
4030 BlockBackend *blk;
4031 BlockDriverState *bs;
4033 if (has_id && has_node_name) {
4034 error_setg(errp, "Only one of id and node-name must be specified");
4035 return;
4036 } else if (!has_id && !has_node_name) {
4037 error_setg(errp, "No block device specified");
4038 return;
4041 if (has_id) {
4042 /* blk_by_name() never returns a BB that is not owned by the monitor */
4043 blk = blk_by_name(id);
4044 if (!blk) {
4045 error_setg(errp, "Cannot find block backend %s", id);
4046 return;
4048 if (blk_legacy_dinfo(blk)) {
4049 error_setg(errp, "Deleting block backend added with drive-add"
4050 " is not supported");
4051 return;
4053 if (blk_get_refcnt(blk) > 1) {
4054 error_setg(errp, "Block backend %s is in use", id);
4055 return;
4057 bs = blk_bs(blk);
4058 aio_context = blk_get_aio_context(blk);
4059 } else {
4060 blk = NULL;
4061 bs = bdrv_find_node(node_name);
4062 if (!bs) {
4063 error_setg(errp, "Cannot find node %s", node_name);
4064 return;
4066 if (bdrv_has_blk(bs)) {
4067 error_setg(errp, "Node %s is in use by %s",
4068 node_name, bdrv_get_parent_name(bs));
4069 return;
4071 aio_context = bdrv_get_aio_context(bs);
4074 aio_context_acquire(aio_context);
4076 if (bs) {
4077 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
4078 goto out;
4081 if (!blk && !bs->monitor_list.tqe_prev) {
4082 error_setg(errp, "Node %s is not owned by the monitor",
4083 bs->node_name);
4084 goto out;
4087 if (bs->refcnt > 1) {
4088 error_setg(errp, "Block device %s is in use",
4089 bdrv_get_device_or_node_name(bs));
4090 goto out;
4094 if (blk) {
4095 monitor_remove_blk(blk);
4096 blk_unref(blk);
4097 } else {
4098 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
4099 bdrv_unref(bs);
4102 out:
4103 aio_context_release(aio_context);
4106 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
4107 const char *child_name)
4109 BdrvChild *child;
4111 QLIST_FOREACH(child, &parent_bs->children, next) {
4112 if (strcmp(child->name, child_name) == 0) {
4113 return child;
4117 return NULL;
4120 void qmp_x_blockdev_change(const char *parent, bool has_child,
4121 const char *child, bool has_node,
4122 const char *node, Error **errp)
4124 BlockDriverState *parent_bs, *new_bs = NULL;
4125 BdrvChild *p_child;
4127 parent_bs = bdrv_lookup_bs(parent, parent, errp);
4128 if (!parent_bs) {
4129 return;
4132 if (has_child == has_node) {
4133 if (has_child) {
4134 error_setg(errp, "The parameters child and node are in conflict");
4135 } else {
4136 error_setg(errp, "Either child or node must be specified");
4138 return;
4141 if (has_child) {
4142 p_child = bdrv_find_child(parent_bs, child);
4143 if (!p_child) {
4144 error_setg(errp, "Node '%s' does not have child '%s'",
4145 parent, child);
4146 return;
4148 bdrv_del_child(parent_bs, p_child, errp);
4151 if (has_node) {
4152 new_bs = bdrv_find_node(node);
4153 if (!new_bs) {
4154 error_setg(errp, "Node '%s' not found", node);
4155 return;
4157 bdrv_add_child(parent_bs, new_bs, errp);
4161 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
4163 BlockJobInfoList *head = NULL, **p_next = &head;
4164 BlockJob *job;
4166 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
4167 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
4168 AioContext *aio_context = blk_get_aio_context(job->blk);
4170 aio_context_acquire(aio_context);
4171 elem->value = block_job_query(job);
4172 aio_context_release(aio_context);
4174 *p_next = elem;
4175 p_next = &elem->next;
4178 return head;
4181 QemuOptsList qemu_common_drive_opts = {
4182 .name = "drive",
4183 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
4184 .desc = {
4186 .name = "snapshot",
4187 .type = QEMU_OPT_BOOL,
4188 .help = "enable/disable snapshot mode",
4190 .name = "discard",
4191 .type = QEMU_OPT_STRING,
4192 .help = "discard operation (ignore/off, unmap/on)",
4194 .name = "aio",
4195 .type = QEMU_OPT_STRING,
4196 .help = "host AIO implementation (threads, native)",
4198 .name = BDRV_OPT_CACHE_WB,
4199 .type = QEMU_OPT_BOOL,
4200 .help = "Enable writeback mode",
4202 .name = "format",
4203 .type = QEMU_OPT_STRING,
4204 .help = "disk format (raw, qcow2, ...)",
4206 .name = "rerror",
4207 .type = QEMU_OPT_STRING,
4208 .help = "read error action",
4210 .name = "werror",
4211 .type = QEMU_OPT_STRING,
4212 .help = "write error action",
4214 .name = "read-only",
4215 .type = QEMU_OPT_BOOL,
4216 .help = "open drive file as read-only",
4218 .name = "throttling.iops-total",
4219 .type = QEMU_OPT_NUMBER,
4220 .help = "limit total I/O operations per second",
4222 .name = "throttling.iops-read",
4223 .type = QEMU_OPT_NUMBER,
4224 .help = "limit read operations per second",
4226 .name = "throttling.iops-write",
4227 .type = QEMU_OPT_NUMBER,
4228 .help = "limit write operations per second",
4230 .name = "throttling.bps-total",
4231 .type = QEMU_OPT_NUMBER,
4232 .help = "limit total bytes per second",
4234 .name = "throttling.bps-read",
4235 .type = QEMU_OPT_NUMBER,
4236 .help = "limit read bytes per second",
4238 .name = "throttling.bps-write",
4239 .type = QEMU_OPT_NUMBER,
4240 .help = "limit write bytes per second",
4242 .name = "throttling.iops-total-max",
4243 .type = QEMU_OPT_NUMBER,
4244 .help = "I/O operations burst",
4246 .name = "throttling.iops-read-max",
4247 .type = QEMU_OPT_NUMBER,
4248 .help = "I/O operations read burst",
4250 .name = "throttling.iops-write-max",
4251 .type = QEMU_OPT_NUMBER,
4252 .help = "I/O operations write burst",
4254 .name = "throttling.bps-total-max",
4255 .type = QEMU_OPT_NUMBER,
4256 .help = "total bytes burst",
4258 .name = "throttling.bps-read-max",
4259 .type = QEMU_OPT_NUMBER,
4260 .help = "total bytes read burst",
4262 .name = "throttling.bps-write-max",
4263 .type = QEMU_OPT_NUMBER,
4264 .help = "total bytes write burst",
4266 .name = "throttling.iops-total-max-length",
4267 .type = QEMU_OPT_NUMBER,
4268 .help = "length of the iops-total-max burst period, in seconds",
4270 .name = "throttling.iops-read-max-length",
4271 .type = QEMU_OPT_NUMBER,
4272 .help = "length of the iops-read-max burst period, in seconds",
4274 .name = "throttling.iops-write-max-length",
4275 .type = QEMU_OPT_NUMBER,
4276 .help = "length of the iops-write-max burst period, in seconds",
4278 .name = "throttling.bps-total-max-length",
4279 .type = QEMU_OPT_NUMBER,
4280 .help = "length of the bps-total-max burst period, in seconds",
4282 .name = "throttling.bps-read-max-length",
4283 .type = QEMU_OPT_NUMBER,
4284 .help = "length of the bps-read-max burst period, in seconds",
4286 .name = "throttling.bps-write-max-length",
4287 .type = QEMU_OPT_NUMBER,
4288 .help = "length of the bps-write-max burst period, in seconds",
4290 .name = "throttling.iops-size",
4291 .type = QEMU_OPT_NUMBER,
4292 .help = "when limiting by iops max size of an I/O in bytes",
4294 .name = "throttling.group",
4295 .type = QEMU_OPT_STRING,
4296 .help = "name of the block throttling group",
4298 .name = "copy-on-read",
4299 .type = QEMU_OPT_BOOL,
4300 .help = "copy read data from backing file into image file",
4302 .name = "detect-zeroes",
4303 .type = QEMU_OPT_STRING,
4304 .help = "try to optimize zero writes (off, on, unmap)",
4306 .name = "stats-account-invalid",
4307 .type = QEMU_OPT_BOOL,
4308 .help = "whether to account for invalid I/O operations "
4309 "in the statistics",
4311 .name = "stats-account-failed",
4312 .type = QEMU_OPT_BOOL,
4313 .help = "whether to account for failed I/O operations "
4314 "in the statistics",
4316 { /* end of list */ }
4320 static QemuOptsList qemu_root_bds_opts = {
4321 .name = "root-bds",
4322 .head = QTAILQ_HEAD_INITIALIZER(qemu_root_bds_opts.head),
4323 .desc = {
4325 .name = "discard",
4326 .type = QEMU_OPT_STRING,
4327 .help = "discard operation (ignore/off, unmap/on)",
4329 .name = "aio",
4330 .type = QEMU_OPT_STRING,
4331 .help = "host AIO implementation (threads, native)",
4333 .name = "read-only",
4334 .type = QEMU_OPT_BOOL,
4335 .help = "open drive file as read-only",
4337 .name = "copy-on-read",
4338 .type = QEMU_OPT_BOOL,
4339 .help = "copy read data from backing file into image file",
4341 .name = "detect-zeroes",
4342 .type = QEMU_OPT_STRING,
4343 .help = "try to optimize zero writes (off, on, unmap)",
4345 { /* end of list */ }
4349 QemuOptsList qemu_drive_opts = {
4350 .name = "drive",
4351 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
4352 .desc = {
4354 * no elements => accept any params
4355 * validation will happen later
4357 { /* end of list */ }