block: Add block job transactions
[qemu.git] / blockdev.c
blob944525806ee1b688a44532d7a86b6feb634105dd
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 "sysemu/block-backend.h"
34 #include "sysemu/blockdev.h"
35 #include "hw/block/block.h"
36 #include "block/blockjob.h"
37 #include "block/throttle-groups.h"
38 #include "monitor/monitor.h"
39 #include "qemu/error-report.h"
40 #include "qemu/option.h"
41 #include "qemu/config-file.h"
42 #include "qapi/qmp/types.h"
43 #include "qapi-visit.h"
44 #include "qapi/qmp/qerror.h"
45 #include "qapi/qmp-output-visitor.h"
46 #include "qapi/util.h"
47 #include "sysemu/sysemu.h"
48 #include "block/block_int.h"
49 #include "qmp-commands.h"
50 #include "trace.h"
51 #include "sysemu/arch_init.h"
53 static const char *const if_name[IF_COUNT] = {
54 [IF_NONE] = "none",
55 [IF_IDE] = "ide",
56 [IF_SCSI] = "scsi",
57 [IF_FLOPPY] = "floppy",
58 [IF_PFLASH] = "pflash",
59 [IF_MTD] = "mtd",
60 [IF_SD] = "sd",
61 [IF_VIRTIO] = "virtio",
62 [IF_XEN] = "xen",
65 static int if_max_devs[IF_COUNT] = {
67 * Do not change these numbers! They govern how drive option
68 * index maps to unit and bus. That mapping is ABI.
70 * All controllers used to imlement if=T drives need to support
71 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
72 * Otherwise, some index values map to "impossible" bus, unit
73 * values.
75 * For instance, if you change [IF_SCSI] to 255, -drive
76 * if=scsi,index=12 no longer means bus=1,unit=5, but
77 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
78 * the drive can't be set up. Regression.
80 [IF_IDE] = 2,
81 [IF_SCSI] = 7,
84 /**
85 * Boards may call this to offer board-by-board overrides
86 * of the default, global values.
88 void override_max_devs(BlockInterfaceType type, int max_devs)
90 BlockBackend *blk;
91 DriveInfo *dinfo;
93 if (max_devs <= 0) {
94 return;
97 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
98 dinfo = blk_legacy_dinfo(blk);
99 if (dinfo->type == type) {
100 fprintf(stderr, "Cannot override units-per-bus property of"
101 " the %s interface, because a drive of that type has"
102 " already been added.\n", if_name[type]);
103 g_assert_not_reached();
107 if_max_devs[type] = max_devs;
111 * We automatically delete the drive when a device using it gets
112 * unplugged. Questionable feature, but we can't just drop it.
113 * Device models call blockdev_mark_auto_del() to schedule the
114 * automatic deletion, and generic qdev code calls blockdev_auto_del()
115 * when deletion is actually safe.
117 void blockdev_mark_auto_del(BlockBackend *blk)
119 DriveInfo *dinfo = blk_legacy_dinfo(blk);
120 BlockDriverState *bs = blk_bs(blk);
121 AioContext *aio_context;
123 if (!dinfo) {
124 return;
127 if (bs) {
128 aio_context = bdrv_get_aio_context(bs);
129 aio_context_acquire(aio_context);
131 if (bs->job) {
132 block_job_cancel(bs->job);
135 aio_context_release(aio_context);
138 dinfo->auto_del = 1;
141 void blockdev_auto_del(BlockBackend *blk)
143 DriveInfo *dinfo = blk_legacy_dinfo(blk);
145 if (dinfo && dinfo->auto_del) {
146 blk_unref(blk);
151 * Returns the current mapping of how many units per bus
152 * a particular interface can support.
154 * A positive integer indicates n units per bus.
155 * 0 implies the mapping has not been established.
156 * -1 indicates an invalid BlockInterfaceType was given.
158 int drive_get_max_devs(BlockInterfaceType type)
160 if (type >= IF_IDE && type < IF_COUNT) {
161 return if_max_devs[type];
164 return -1;
167 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
169 int max_devs = if_max_devs[type];
170 return max_devs ? index / max_devs : 0;
173 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
175 int max_devs = if_max_devs[type];
176 return max_devs ? index % max_devs : index;
179 QemuOpts *drive_def(const char *optstr)
181 return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
184 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
185 const char *optstr)
187 QemuOpts *opts;
189 opts = drive_def(optstr);
190 if (!opts) {
191 return NULL;
193 if (type != IF_DEFAULT) {
194 qemu_opt_set(opts, "if", if_name[type], &error_abort);
196 if (index >= 0) {
197 qemu_opt_set_number(opts, "index", index, &error_abort);
199 if (file)
200 qemu_opt_set(opts, "file", file, &error_abort);
201 return opts;
204 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
206 BlockBackend *blk;
207 DriveInfo *dinfo;
209 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
210 dinfo = blk_legacy_dinfo(blk);
211 if (dinfo && dinfo->type == type
212 && dinfo->bus == bus && dinfo->unit == unit) {
213 return dinfo;
217 return NULL;
220 bool drive_check_orphaned(void)
222 BlockBackend *blk;
223 DriveInfo *dinfo;
224 bool rs = false;
226 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
227 dinfo = blk_legacy_dinfo(blk);
228 /* If dinfo->bdrv->dev is NULL, it has no device attached. */
229 /* Unless this is a default drive, this may be an oversight. */
230 if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
231 dinfo->type != IF_NONE) {
232 fprintf(stderr, "Warning: Orphaned drive without device: "
233 "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
234 blk_name(blk), blk_bs(blk) ? blk_bs(blk)->filename : "",
235 if_name[dinfo->type], dinfo->bus, dinfo->unit);
236 rs = true;
240 return rs;
243 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
245 return drive_get(type,
246 drive_index_to_bus_id(type, index),
247 drive_index_to_unit_id(type, index));
250 int drive_get_max_bus(BlockInterfaceType type)
252 int max_bus;
253 BlockBackend *blk;
254 DriveInfo *dinfo;
256 max_bus = -1;
257 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
258 dinfo = blk_legacy_dinfo(blk);
259 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
260 max_bus = dinfo->bus;
263 return max_bus;
266 /* Get a block device. This should only be used for single-drive devices
267 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
268 appropriate bus. */
269 DriveInfo *drive_get_next(BlockInterfaceType type)
271 static int next_block_unit[IF_COUNT];
273 return drive_get(type, 0, next_block_unit[type]++);
276 static void bdrv_format_print(void *opaque, const char *name)
278 error_printf(" %s", name);
281 typedef struct {
282 QEMUBH *bh;
283 BlockDriverState *bs;
284 } BDRVPutRefBH;
286 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
288 if (!strcmp(buf, "ignore")) {
289 return BLOCKDEV_ON_ERROR_IGNORE;
290 } else if (!is_read && !strcmp(buf, "enospc")) {
291 return BLOCKDEV_ON_ERROR_ENOSPC;
292 } else if (!strcmp(buf, "stop")) {
293 return BLOCKDEV_ON_ERROR_STOP;
294 } else if (!strcmp(buf, "report")) {
295 return BLOCKDEV_ON_ERROR_REPORT;
296 } else {
297 error_setg(errp, "'%s' invalid %s error action",
298 buf, is_read ? "read" : "write");
299 return -1;
303 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
305 if (throttle_conflicting(cfg)) {
306 error_setg(errp, "bps/iops/max total values and read/write values"
307 " cannot be used at the same time");
308 return false;
311 if (!throttle_is_valid(cfg)) {
312 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
313 return false;
316 if (throttle_max_is_missing_limit(cfg)) {
317 error_setg(errp, "bps_max/iops_max require corresponding"
318 " bps/iops values");
319 return false;
322 return true;
325 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
327 /* All parameters but @opts are optional and may be set to NULL. */
328 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
329 const char **throttling_group, ThrottleConfig *throttle_cfg,
330 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
332 const char *discard;
333 Error *local_error = NULL;
334 const char *aio;
336 if (bdrv_flags) {
337 if (!qemu_opt_get_bool(opts, "read-only", false)) {
338 *bdrv_flags |= BDRV_O_RDWR;
340 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
341 *bdrv_flags |= BDRV_O_COPY_ON_READ;
344 if ((discard = qemu_opt_get(opts, "discard")) != NULL) {
345 if (bdrv_parse_discard_flags(discard, bdrv_flags) != 0) {
346 error_setg(errp, "Invalid discard option");
347 return;
351 if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true)) {
352 *bdrv_flags |= BDRV_O_CACHE_WB;
354 if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_DIRECT, false)) {
355 *bdrv_flags |= BDRV_O_NOCACHE;
357 if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) {
358 *bdrv_flags |= BDRV_O_NO_FLUSH;
361 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
362 if (!strcmp(aio, "native")) {
363 *bdrv_flags |= BDRV_O_NATIVE_AIO;
364 } else if (!strcmp(aio, "threads")) {
365 /* this is the default */
366 } else {
367 error_setg(errp, "invalid aio option");
368 return;
373 /* disk I/O throttling */
374 if (throttling_group) {
375 *throttling_group = qemu_opt_get(opts, "throttling.group");
378 if (throttle_cfg) {
379 memset(throttle_cfg, 0, sizeof(*throttle_cfg));
380 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
381 qemu_opt_get_number(opts, "throttling.bps-total", 0);
382 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
383 qemu_opt_get_number(opts, "throttling.bps-read", 0);
384 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
385 qemu_opt_get_number(opts, "throttling.bps-write", 0);
386 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
387 qemu_opt_get_number(opts, "throttling.iops-total", 0);
388 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
389 qemu_opt_get_number(opts, "throttling.iops-read", 0);
390 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
391 qemu_opt_get_number(opts, "throttling.iops-write", 0);
393 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
394 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
395 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
396 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
397 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
398 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
399 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
400 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
401 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
402 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
403 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
404 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
406 throttle_cfg->op_size =
407 qemu_opt_get_number(opts, "throttling.iops-size", 0);
409 if (!check_throttle_config(throttle_cfg, errp)) {
410 return;
414 if (detect_zeroes) {
415 *detect_zeroes =
416 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
417 qemu_opt_get(opts, "detect-zeroes"),
418 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
419 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
420 &local_error);
421 if (local_error) {
422 error_propagate(errp, local_error);
423 return;
426 if (bdrv_flags &&
427 *detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
428 !(*bdrv_flags & BDRV_O_UNMAP))
430 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
431 "without setting discard operation to unmap");
432 return;
437 /* Takes the ownership of bs_opts */
438 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
439 Error **errp)
441 const char *buf;
442 int bdrv_flags = 0;
443 int on_read_error, on_write_error;
444 BlockBackend *blk;
445 BlockDriverState *bs;
446 ThrottleConfig cfg;
447 int snapshot = 0;
448 Error *error = NULL;
449 QemuOpts *opts;
450 const char *id;
451 bool has_driver_specific_opts;
452 BlockdevDetectZeroesOptions detect_zeroes =
453 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
454 const char *throttling_group = NULL;
456 /* Check common options by copying from bs_opts to opts, all other options
457 * stay in bs_opts for processing by bdrv_open(). */
458 id = qdict_get_try_str(bs_opts, "id");
459 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
460 if (error) {
461 error_propagate(errp, error);
462 goto err_no_opts;
465 qemu_opts_absorb_qdict(opts, bs_opts, &error);
466 if (error) {
467 error_propagate(errp, error);
468 goto early_err;
471 if (id) {
472 qdict_del(bs_opts, "id");
475 has_driver_specific_opts = !!qdict_size(bs_opts);
477 /* extract parameters */
478 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
480 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
481 &detect_zeroes, &error);
482 if (error) {
483 error_propagate(errp, error);
484 goto early_err;
487 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
488 if (is_help_option(buf)) {
489 error_printf("Supported formats:");
490 bdrv_iterate_format(bdrv_format_print, NULL);
491 error_printf("\n");
492 goto early_err;
495 if (qdict_haskey(bs_opts, "driver")) {
496 error_setg(errp, "Cannot specify both 'driver' and 'format'");
497 goto early_err;
499 qdict_put(bs_opts, "driver", qstring_from_str(buf));
502 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
503 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
504 on_write_error = parse_block_error_action(buf, 0, &error);
505 if (error) {
506 error_propagate(errp, error);
507 goto early_err;
511 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
512 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
513 on_read_error = parse_block_error_action(buf, 1, &error);
514 if (error) {
515 error_propagate(errp, error);
516 goto early_err;
520 if (snapshot) {
521 /* always use cache=unsafe with snapshot */
522 bdrv_flags &= ~BDRV_O_CACHE_MASK;
523 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
526 /* init */
527 if ((!file || !*file) && !has_driver_specific_opts) {
528 BlockBackendRootState *blk_rs;
530 blk = blk_new(qemu_opts_id(opts), errp);
531 if (!blk) {
532 goto early_err;
535 blk_rs = blk_get_root_state(blk);
536 blk_rs->open_flags = bdrv_flags;
537 blk_rs->read_only = !(bdrv_flags & BDRV_O_RDWR);
538 blk_rs->detect_zeroes = detect_zeroes;
540 if (throttle_enabled(&cfg)) {
541 if (!throttling_group) {
542 throttling_group = blk_name(blk);
544 blk_rs->throttle_group = g_strdup(throttling_group);
545 blk_rs->throttle_state = throttle_group_incref(throttling_group);
546 blk_rs->throttle_state->cfg = cfg;
549 QDECREF(bs_opts);
550 } else {
551 if (file && !*file) {
552 file = NULL;
555 blk = blk_new_open(qemu_opts_id(opts), file, NULL, bs_opts, bdrv_flags,
556 errp);
557 if (!blk) {
558 goto err_no_bs_opts;
560 bs = blk_bs(blk);
562 bs->detect_zeroes = detect_zeroes;
564 /* disk I/O throttling */
565 if (throttle_enabled(&cfg)) {
566 if (!throttling_group) {
567 throttling_group = blk_name(blk);
569 bdrv_io_limits_enable(bs, throttling_group);
570 bdrv_set_io_limits(bs, &cfg);
573 if (bdrv_key_required(bs)) {
574 autostart = 0;
578 blk_set_on_error(blk, on_read_error, on_write_error);
580 err_no_bs_opts:
581 qemu_opts_del(opts);
582 return blk;
584 early_err:
585 qemu_opts_del(opts);
586 err_no_opts:
587 QDECREF(bs_opts);
588 return NULL;
591 static QemuOptsList qemu_root_bds_opts;
593 /* Takes the ownership of bs_opts */
594 static BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
596 BlockDriverState *bs;
597 QemuOpts *opts;
598 Error *local_error = NULL;
599 BlockdevDetectZeroesOptions detect_zeroes;
600 int ret;
601 int bdrv_flags = 0;
603 opts = qemu_opts_create(&qemu_root_bds_opts, NULL, 1, errp);
604 if (!opts) {
605 goto fail;
608 qemu_opts_absorb_qdict(opts, bs_opts, &local_error);
609 if (local_error) {
610 error_propagate(errp, local_error);
611 goto fail;
614 extract_common_blockdev_options(opts, &bdrv_flags, NULL, NULL,
615 &detect_zeroes, &local_error);
616 if (local_error) {
617 error_propagate(errp, local_error);
618 goto fail;
621 bs = NULL;
622 ret = bdrv_open(&bs, NULL, NULL, bs_opts, bdrv_flags, errp);
623 if (ret < 0) {
624 goto fail_no_bs_opts;
627 bs->detect_zeroes = detect_zeroes;
629 fail_no_bs_opts:
630 qemu_opts_del(opts);
631 return bs;
633 fail:
634 qemu_opts_del(opts);
635 QDECREF(bs_opts);
636 return NULL;
639 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
640 Error **errp)
642 const char *value;
644 value = qemu_opt_get(opts, from);
645 if (value) {
646 if (qemu_opt_find(opts, to)) {
647 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
648 "same time", to, from);
649 return;
653 /* rename all items in opts */
654 while ((value = qemu_opt_get(opts, from))) {
655 qemu_opt_set(opts, to, value, &error_abort);
656 qemu_opt_unset(opts, from);
660 QemuOptsList qemu_legacy_drive_opts = {
661 .name = "drive",
662 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
663 .desc = {
665 .name = "bus",
666 .type = QEMU_OPT_NUMBER,
667 .help = "bus number",
669 .name = "unit",
670 .type = QEMU_OPT_NUMBER,
671 .help = "unit number (i.e. lun for scsi)",
673 .name = "index",
674 .type = QEMU_OPT_NUMBER,
675 .help = "index number",
677 .name = "media",
678 .type = QEMU_OPT_STRING,
679 .help = "media type (disk, cdrom)",
681 .name = "if",
682 .type = QEMU_OPT_STRING,
683 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
685 .name = "cyls",
686 .type = QEMU_OPT_NUMBER,
687 .help = "number of cylinders (ide disk geometry)",
689 .name = "heads",
690 .type = QEMU_OPT_NUMBER,
691 .help = "number of heads (ide disk geometry)",
693 .name = "secs",
694 .type = QEMU_OPT_NUMBER,
695 .help = "number of sectors (ide disk geometry)",
697 .name = "trans",
698 .type = QEMU_OPT_STRING,
699 .help = "chs translation (auto, lba, none)",
701 .name = "boot",
702 .type = QEMU_OPT_BOOL,
703 .help = "(deprecated, ignored)",
705 .name = "addr",
706 .type = QEMU_OPT_STRING,
707 .help = "pci address (virtio only)",
709 .name = "serial",
710 .type = QEMU_OPT_STRING,
711 .help = "disk serial number",
713 .name = "file",
714 .type = QEMU_OPT_STRING,
715 .help = "file name",
718 /* Options that are passed on, but have special semantics with -drive */
720 .name = "read-only",
721 .type = QEMU_OPT_BOOL,
722 .help = "open drive file as read-only",
724 .name = "rerror",
725 .type = QEMU_OPT_STRING,
726 .help = "read error action",
728 .name = "werror",
729 .type = QEMU_OPT_STRING,
730 .help = "write error action",
732 .name = "copy-on-read",
733 .type = QEMU_OPT_BOOL,
734 .help = "copy read data from backing file into image file",
737 { /* end of list */ }
741 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
743 const char *value;
744 BlockBackend *blk;
745 DriveInfo *dinfo = NULL;
746 QDict *bs_opts;
747 QemuOpts *legacy_opts;
748 DriveMediaType media = MEDIA_DISK;
749 BlockInterfaceType type;
750 int cyls, heads, secs, translation;
751 int max_devs, bus_id, unit_id, index;
752 const char *devaddr;
753 const char *werror, *rerror;
754 bool read_only = false;
755 bool copy_on_read;
756 const char *serial;
757 const char *filename;
758 Error *local_err = NULL;
759 int i;
761 /* Change legacy command line options into QMP ones */
762 static const struct {
763 const char *from;
764 const char *to;
765 } opt_renames[] = {
766 { "iops", "throttling.iops-total" },
767 { "iops_rd", "throttling.iops-read" },
768 { "iops_wr", "throttling.iops-write" },
770 { "bps", "throttling.bps-total" },
771 { "bps_rd", "throttling.bps-read" },
772 { "bps_wr", "throttling.bps-write" },
774 { "iops_max", "throttling.iops-total-max" },
775 { "iops_rd_max", "throttling.iops-read-max" },
776 { "iops_wr_max", "throttling.iops-write-max" },
778 { "bps_max", "throttling.bps-total-max" },
779 { "bps_rd_max", "throttling.bps-read-max" },
780 { "bps_wr_max", "throttling.bps-write-max" },
782 { "iops_size", "throttling.iops-size" },
784 { "group", "throttling.group" },
786 { "readonly", "read-only" },
789 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
790 qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
791 &local_err);
792 if (local_err) {
793 error_report_err(local_err);
794 return NULL;
798 value = qemu_opt_get(all_opts, "cache");
799 if (value) {
800 int flags = 0;
802 if (bdrv_parse_cache_flags(value, &flags) != 0) {
803 error_report("invalid cache option");
804 return NULL;
807 /* Specific options take precedence */
808 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
809 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
810 !!(flags & BDRV_O_CACHE_WB), &error_abort);
812 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
813 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
814 !!(flags & BDRV_O_NOCACHE), &error_abort);
816 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
817 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
818 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
820 qemu_opt_unset(all_opts, "cache");
823 /* Get a QDict for processing the options */
824 bs_opts = qdict_new();
825 qemu_opts_to_qdict(all_opts, bs_opts);
827 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
828 &error_abort);
829 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
830 if (local_err) {
831 error_report_err(local_err);
832 goto fail;
835 /* Deprecated option boot=[on|off] */
836 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
837 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
838 "ignored. Future versions will reject this parameter. Please "
839 "update your scripts.\n");
842 /* Media type */
843 value = qemu_opt_get(legacy_opts, "media");
844 if (value) {
845 if (!strcmp(value, "disk")) {
846 media = MEDIA_DISK;
847 } else if (!strcmp(value, "cdrom")) {
848 media = MEDIA_CDROM;
849 read_only = true;
850 } else {
851 error_report("'%s' invalid media", value);
852 goto fail;
856 /* copy-on-read is disabled with a warning for read-only devices */
857 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
858 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
860 if (read_only && copy_on_read) {
861 error_report("warning: disabling copy-on-read on read-only drive");
862 copy_on_read = false;
865 qdict_put(bs_opts, "read-only",
866 qstring_from_str(read_only ? "on" : "off"));
867 qdict_put(bs_opts, "copy-on-read",
868 qstring_from_str(copy_on_read ? "on" :"off"));
870 /* Controller type */
871 value = qemu_opt_get(legacy_opts, "if");
872 if (value) {
873 for (type = 0;
874 type < IF_COUNT && strcmp(value, if_name[type]);
875 type++) {
877 if (type == IF_COUNT) {
878 error_report("unsupported bus type '%s'", value);
879 goto fail;
881 } else {
882 type = block_default_type;
885 /* Geometry */
886 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
887 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
888 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
890 if (cyls || heads || secs) {
891 if (cyls < 1) {
892 error_report("invalid physical cyls number");
893 goto fail;
895 if (heads < 1) {
896 error_report("invalid physical heads number");
897 goto fail;
899 if (secs < 1) {
900 error_report("invalid physical secs number");
901 goto fail;
905 translation = BIOS_ATA_TRANSLATION_AUTO;
906 value = qemu_opt_get(legacy_opts, "trans");
907 if (value != NULL) {
908 if (!cyls) {
909 error_report("'%s' trans must be used with cyls, heads and secs",
910 value);
911 goto fail;
913 if (!strcmp(value, "none")) {
914 translation = BIOS_ATA_TRANSLATION_NONE;
915 } else if (!strcmp(value, "lba")) {
916 translation = BIOS_ATA_TRANSLATION_LBA;
917 } else if (!strcmp(value, "large")) {
918 translation = BIOS_ATA_TRANSLATION_LARGE;
919 } else if (!strcmp(value, "rechs")) {
920 translation = BIOS_ATA_TRANSLATION_RECHS;
921 } else if (!strcmp(value, "auto")) {
922 translation = BIOS_ATA_TRANSLATION_AUTO;
923 } else {
924 error_report("'%s' invalid translation type", value);
925 goto fail;
929 if (media == MEDIA_CDROM) {
930 if (cyls || secs || heads) {
931 error_report("CHS can't be set with media=cdrom");
932 goto fail;
936 /* Device address specified by bus/unit or index.
937 * If none was specified, try to find the first free one. */
938 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
939 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
940 index = qemu_opt_get_number(legacy_opts, "index", -1);
942 max_devs = if_max_devs[type];
944 if (index != -1) {
945 if (bus_id != 0 || unit_id != -1) {
946 error_report("index cannot be used with bus and unit");
947 goto fail;
949 bus_id = drive_index_to_bus_id(type, index);
950 unit_id = drive_index_to_unit_id(type, index);
953 if (unit_id == -1) {
954 unit_id = 0;
955 while (drive_get(type, bus_id, unit_id) != NULL) {
956 unit_id++;
957 if (max_devs && unit_id >= max_devs) {
958 unit_id -= max_devs;
959 bus_id++;
964 if (max_devs && unit_id >= max_devs) {
965 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
966 goto fail;
969 if (drive_get(type, bus_id, unit_id) != NULL) {
970 error_report("drive with bus=%d, unit=%d (index=%d) exists",
971 bus_id, unit_id, index);
972 goto fail;
975 /* Serial number */
976 serial = qemu_opt_get(legacy_opts, "serial");
978 /* no id supplied -> create one */
979 if (qemu_opts_id(all_opts) == NULL) {
980 char *new_id;
981 const char *mediastr = "";
982 if (type == IF_IDE || type == IF_SCSI) {
983 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
985 if (max_devs) {
986 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
987 mediastr, unit_id);
988 } else {
989 new_id = g_strdup_printf("%s%s%i", if_name[type],
990 mediastr, unit_id);
992 qdict_put(bs_opts, "id", qstring_from_str(new_id));
993 g_free(new_id);
996 /* Add virtio block device */
997 devaddr = qemu_opt_get(legacy_opts, "addr");
998 if (devaddr && type != IF_VIRTIO) {
999 error_report("addr is not supported by this bus type");
1000 goto fail;
1003 if (type == IF_VIRTIO) {
1004 QemuOpts *devopts;
1005 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
1006 &error_abort);
1007 if (arch_type == QEMU_ARCH_S390X) {
1008 qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
1009 } else {
1010 qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
1012 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
1013 &error_abort);
1014 if (devaddr) {
1015 qemu_opt_set(devopts, "addr", devaddr, &error_abort);
1019 filename = qemu_opt_get(legacy_opts, "file");
1021 /* Check werror/rerror compatibility with if=... */
1022 werror = qemu_opt_get(legacy_opts, "werror");
1023 if (werror != NULL) {
1024 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
1025 type != IF_NONE) {
1026 error_report("werror is not supported by this bus type");
1027 goto fail;
1029 qdict_put(bs_opts, "werror", qstring_from_str(werror));
1032 rerror = qemu_opt_get(legacy_opts, "rerror");
1033 if (rerror != NULL) {
1034 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
1035 type != IF_NONE) {
1036 error_report("rerror is not supported by this bus type");
1037 goto fail;
1039 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
1042 /* Actual block device init: Functionality shared with blockdev-add */
1043 blk = blockdev_init(filename, bs_opts, &local_err);
1044 bs_opts = NULL;
1045 if (!blk) {
1046 if (local_err) {
1047 error_report_err(local_err);
1049 goto fail;
1050 } else {
1051 assert(!local_err);
1054 /* Create legacy DriveInfo */
1055 dinfo = g_malloc0(sizeof(*dinfo));
1056 dinfo->opts = all_opts;
1058 dinfo->cyls = cyls;
1059 dinfo->heads = heads;
1060 dinfo->secs = secs;
1061 dinfo->trans = translation;
1063 dinfo->type = type;
1064 dinfo->bus = bus_id;
1065 dinfo->unit = unit_id;
1066 dinfo->devaddr = devaddr;
1067 dinfo->serial = g_strdup(serial);
1069 blk_set_legacy_dinfo(blk, dinfo);
1071 switch(type) {
1072 case IF_IDE:
1073 case IF_SCSI:
1074 case IF_XEN:
1075 case IF_NONE:
1076 dinfo->media_cd = media == MEDIA_CDROM;
1077 break;
1078 default:
1079 break;
1082 fail:
1083 qemu_opts_del(legacy_opts);
1084 QDECREF(bs_opts);
1085 return dinfo;
1088 void hmp_commit(Monitor *mon, const QDict *qdict)
1090 const char *device = qdict_get_str(qdict, "device");
1091 BlockBackend *blk;
1092 int ret;
1094 if (!strcmp(device, "all")) {
1095 ret = bdrv_commit_all();
1096 } else {
1097 BlockDriverState *bs;
1098 AioContext *aio_context;
1100 blk = blk_by_name(device);
1101 if (!blk) {
1102 monitor_printf(mon, "Device '%s' not found\n", device);
1103 return;
1105 if (!blk_is_available(blk)) {
1106 monitor_printf(mon, "Device '%s' has no medium\n", device);
1107 return;
1110 bs = blk_bs(blk);
1111 aio_context = bdrv_get_aio_context(bs);
1112 aio_context_acquire(aio_context);
1114 ret = bdrv_commit(bs);
1116 aio_context_release(aio_context);
1118 if (ret < 0) {
1119 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1120 strerror(-ret));
1124 static void blockdev_do_action(TransactionActionKind type, void *data,
1125 Error **errp)
1127 TransactionAction action;
1128 TransactionActionList list;
1130 action.type = type;
1131 action.u.data = data;
1132 list.value = &action;
1133 list.next = NULL;
1134 qmp_transaction(&list, errp);
1137 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1138 bool has_node_name, const char *node_name,
1139 const char *snapshot_file,
1140 bool has_snapshot_node_name,
1141 const char *snapshot_node_name,
1142 bool has_format, const char *format,
1143 bool has_mode, NewImageMode mode, Error **errp)
1145 BlockdevSnapshotSync snapshot = {
1146 .has_device = has_device,
1147 .device = (char *) device,
1148 .has_node_name = has_node_name,
1149 .node_name = (char *) node_name,
1150 .snapshot_file = (char *) snapshot_file,
1151 .has_snapshot_node_name = has_snapshot_node_name,
1152 .snapshot_node_name = (char *) snapshot_node_name,
1153 .has_format = has_format,
1154 .format = (char *) format,
1155 .has_mode = has_mode,
1156 .mode = mode,
1158 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1159 &snapshot, errp);
1162 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1163 Error **errp)
1165 BlockdevSnapshot snapshot_data = {
1166 .node = (char *) node,
1167 .overlay = (char *) overlay
1170 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1171 &snapshot_data, errp);
1174 void qmp_blockdev_snapshot_internal_sync(const char *device,
1175 const char *name,
1176 Error **errp)
1178 BlockdevSnapshotInternal snapshot = {
1179 .device = (char *) device,
1180 .name = (char *) name
1183 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1184 &snapshot, errp);
1187 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1188 bool has_id,
1189 const char *id,
1190 bool has_name,
1191 const char *name,
1192 Error **errp)
1194 BlockDriverState *bs;
1195 BlockBackend *blk;
1196 AioContext *aio_context;
1197 QEMUSnapshotInfo sn;
1198 Error *local_err = NULL;
1199 SnapshotInfo *info = NULL;
1200 int ret;
1202 blk = blk_by_name(device);
1203 if (!blk) {
1204 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1205 "Device '%s' not found", device);
1206 return NULL;
1209 aio_context = blk_get_aio_context(blk);
1210 aio_context_acquire(aio_context);
1212 if (!has_id) {
1213 id = NULL;
1216 if (!has_name) {
1217 name = NULL;
1220 if (!id && !name) {
1221 error_setg(errp, "Name or id must be provided");
1222 goto out_aio_context;
1225 if (!blk_is_available(blk)) {
1226 error_setg(errp, "Device '%s' has no medium", device);
1227 goto out_aio_context;
1229 bs = blk_bs(blk);
1231 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1232 goto out_aio_context;
1235 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1236 if (local_err) {
1237 error_propagate(errp, local_err);
1238 goto out_aio_context;
1240 if (!ret) {
1241 error_setg(errp,
1242 "Snapshot with id '%s' and name '%s' does not exist on "
1243 "device '%s'",
1244 STR_OR_NULL(id), STR_OR_NULL(name), device);
1245 goto out_aio_context;
1248 bdrv_snapshot_delete(bs, id, name, &local_err);
1249 if (local_err) {
1250 error_propagate(errp, local_err);
1251 goto out_aio_context;
1254 aio_context_release(aio_context);
1256 info = g_new0(SnapshotInfo, 1);
1257 info->id = g_strdup(sn.id_str);
1258 info->name = g_strdup(sn.name);
1259 info->date_nsec = sn.date_nsec;
1260 info->date_sec = sn.date_sec;
1261 info->vm_state_size = sn.vm_state_size;
1262 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1263 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1265 return info;
1267 out_aio_context:
1268 aio_context_release(aio_context);
1269 return NULL;
1273 * block_dirty_bitmap_lookup:
1274 * Return a dirty bitmap (if present), after validating
1275 * the node reference and bitmap names.
1277 * @node: The name of the BDS node to search for bitmaps
1278 * @name: The name of the bitmap to search for
1279 * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1280 * @paio: Output pointer for aio_context acquisition, if desired. Can be NULL.
1281 * @errp: Output pointer for error information. Can be NULL.
1283 * @return: A bitmap object on success, or NULL on failure.
1285 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1286 const char *name,
1287 BlockDriverState **pbs,
1288 AioContext **paio,
1289 Error **errp)
1291 BlockDriverState *bs;
1292 BdrvDirtyBitmap *bitmap;
1293 AioContext *aio_context;
1295 if (!node) {
1296 error_setg(errp, "Node cannot be NULL");
1297 return NULL;
1299 if (!name) {
1300 error_setg(errp, "Bitmap name cannot be NULL");
1301 return NULL;
1303 bs = bdrv_lookup_bs(node, node, NULL);
1304 if (!bs) {
1305 error_setg(errp, "Node '%s' not found", node);
1306 return NULL;
1309 aio_context = bdrv_get_aio_context(bs);
1310 aio_context_acquire(aio_context);
1312 bitmap = bdrv_find_dirty_bitmap(bs, name);
1313 if (!bitmap) {
1314 error_setg(errp, "Dirty bitmap '%s' not found", name);
1315 goto fail;
1318 if (pbs) {
1319 *pbs = bs;
1321 if (paio) {
1322 *paio = aio_context;
1323 } else {
1324 aio_context_release(aio_context);
1327 return bitmap;
1329 fail:
1330 aio_context_release(aio_context);
1331 return NULL;
1334 /* New and old BlockDriverState structs for atomic group operations */
1336 typedef struct BlkActionState BlkActionState;
1339 * BlkActionOps:
1340 * Table of operations that define an Action.
1342 * @instance_size: Size of state struct, in bytes.
1343 * @prepare: Prepare the work, must NOT be NULL.
1344 * @commit: Commit the changes, can be NULL.
1345 * @abort: Abort the changes on fail, can be NULL.
1346 * @clean: Clean up resources after all transaction actions have called
1347 * commit() or abort(). Can be NULL.
1349 * Only prepare() may fail. In a single transaction, only one of commit() or
1350 * abort() will be called. clean() will always be called if it is present.
1352 typedef struct BlkActionOps {
1353 size_t instance_size;
1354 void (*prepare)(BlkActionState *common, Error **errp);
1355 void (*commit)(BlkActionState *common);
1356 void (*abort)(BlkActionState *common);
1357 void (*clean)(BlkActionState *common);
1358 } BlkActionOps;
1361 * BlkActionState:
1362 * Describes one Action's state within a Transaction.
1364 * @action: QAPI-defined enum identifying which Action to perform.
1365 * @ops: Table of ActionOps this Action can perform.
1366 * @entry: List membership for all Actions in this Transaction.
1368 * This structure must be arranged as first member in a subclassed type,
1369 * assuming that the compiler will also arrange it to the same offsets as the
1370 * base class.
1372 struct BlkActionState {
1373 TransactionAction *action;
1374 const BlkActionOps *ops;
1375 QSIMPLEQ_ENTRY(BlkActionState) entry;
1378 /* internal snapshot private data */
1379 typedef struct InternalSnapshotState {
1380 BlkActionState common;
1381 BlockDriverState *bs;
1382 AioContext *aio_context;
1383 QEMUSnapshotInfo sn;
1384 bool created;
1385 } InternalSnapshotState;
1387 static void internal_snapshot_prepare(BlkActionState *common,
1388 Error **errp)
1390 Error *local_err = NULL;
1391 const char *device;
1392 const char *name;
1393 BlockBackend *blk;
1394 BlockDriverState *bs;
1395 QEMUSnapshotInfo old_sn, *sn;
1396 bool ret;
1397 qemu_timeval tv;
1398 BlockdevSnapshotInternal *internal;
1399 InternalSnapshotState *state;
1400 int ret1;
1402 g_assert(common->action->type ==
1403 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1404 internal = common->action->u.blockdev_snapshot_internal_sync;
1405 state = DO_UPCAST(InternalSnapshotState, common, common);
1407 /* 1. parse input */
1408 device = internal->device;
1409 name = internal->name;
1411 /* 2. check for validation */
1412 blk = blk_by_name(device);
1413 if (!blk) {
1414 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1415 "Device '%s' not found", device);
1416 return;
1419 /* AioContext is released in .clean() */
1420 state->aio_context = blk_get_aio_context(blk);
1421 aio_context_acquire(state->aio_context);
1423 if (!blk_is_available(blk)) {
1424 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1425 return;
1427 bs = blk_bs(blk);
1429 state->bs = bs;
1430 bdrv_drained_begin(bs);
1432 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1433 return;
1436 if (bdrv_is_read_only(bs)) {
1437 error_setg(errp, "Device '%s' is read only", device);
1438 return;
1441 if (!bdrv_can_snapshot(bs)) {
1442 error_setg(errp, "Block format '%s' used by device '%s' "
1443 "does not support internal snapshots",
1444 bs->drv->format_name, device);
1445 return;
1448 if (!strlen(name)) {
1449 error_setg(errp, "Name is empty");
1450 return;
1453 /* check whether a snapshot with name exist */
1454 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1455 &local_err);
1456 if (local_err) {
1457 error_propagate(errp, local_err);
1458 return;
1459 } else if (ret) {
1460 error_setg(errp,
1461 "Snapshot with name '%s' already exists on device '%s'",
1462 name, device);
1463 return;
1466 /* 3. take the snapshot */
1467 sn = &state->sn;
1468 pstrcpy(sn->name, sizeof(sn->name), name);
1469 qemu_gettimeofday(&tv);
1470 sn->date_sec = tv.tv_sec;
1471 sn->date_nsec = tv.tv_usec * 1000;
1472 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1474 ret1 = bdrv_snapshot_create(bs, sn);
1475 if (ret1 < 0) {
1476 error_setg_errno(errp, -ret1,
1477 "Failed to create snapshot '%s' on device '%s'",
1478 name, device);
1479 return;
1482 /* 4. succeed, mark a snapshot is created */
1483 state->created = true;
1486 static void internal_snapshot_abort(BlkActionState *common)
1488 InternalSnapshotState *state =
1489 DO_UPCAST(InternalSnapshotState, common, common);
1490 BlockDriverState *bs = state->bs;
1491 QEMUSnapshotInfo *sn = &state->sn;
1492 Error *local_error = NULL;
1494 if (!state->created) {
1495 return;
1498 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1499 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1500 "device '%s' in abort: %s",
1501 sn->id_str,
1502 sn->name,
1503 bdrv_get_device_name(bs),
1504 error_get_pretty(local_error));
1505 error_free(local_error);
1509 static void internal_snapshot_clean(BlkActionState *common)
1511 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1512 common, common);
1514 if (state->aio_context) {
1515 if (state->bs) {
1516 bdrv_drained_end(state->bs);
1518 aio_context_release(state->aio_context);
1522 /* external snapshot private data */
1523 typedef struct ExternalSnapshotState {
1524 BlkActionState common;
1525 BlockDriverState *old_bs;
1526 BlockDriverState *new_bs;
1527 AioContext *aio_context;
1528 } ExternalSnapshotState;
1530 static void external_snapshot_prepare(BlkActionState *common,
1531 Error **errp)
1533 int flags = 0, ret;
1534 QDict *options = NULL;
1535 Error *local_err = NULL;
1536 /* Device and node name of the image to generate the snapshot from */
1537 const char *device;
1538 const char *node_name;
1539 /* Reference to the new image (for 'blockdev-snapshot') */
1540 const char *snapshot_ref;
1541 /* File name of the new image (for 'blockdev-snapshot-sync') */
1542 const char *new_image_file;
1543 ExternalSnapshotState *state =
1544 DO_UPCAST(ExternalSnapshotState, common, common);
1545 TransactionAction *action = common->action;
1547 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1548 * purpose but a different set of parameters */
1549 switch (action->type) {
1550 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1552 BlockdevSnapshot *s = action->u.blockdev_snapshot;
1553 device = s->node;
1554 node_name = s->node;
1555 new_image_file = NULL;
1556 snapshot_ref = s->overlay;
1558 break;
1559 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1561 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync;
1562 device = s->has_device ? s->device : NULL;
1563 node_name = s->has_node_name ? s->node_name : NULL;
1564 new_image_file = s->snapshot_file;
1565 snapshot_ref = NULL;
1567 break;
1568 default:
1569 g_assert_not_reached();
1572 /* start processing */
1573 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1574 if (!state->old_bs) {
1575 return;
1578 /* Acquire AioContext now so any threads operating on old_bs stop */
1579 state->aio_context = bdrv_get_aio_context(state->old_bs);
1580 aio_context_acquire(state->aio_context);
1581 bdrv_drained_begin(state->old_bs);
1583 if (!bdrv_is_inserted(state->old_bs)) {
1584 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1585 return;
1588 if (bdrv_op_is_blocked(state->old_bs,
1589 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1590 return;
1593 if (!bdrv_is_read_only(state->old_bs)) {
1594 if (bdrv_flush(state->old_bs)) {
1595 error_setg(errp, QERR_IO_ERROR);
1596 return;
1600 if (!bdrv_is_first_non_filter(state->old_bs)) {
1601 error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1602 return;
1605 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1606 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync;
1607 const char *format = s->has_format ? s->format : "qcow2";
1608 enum NewImageMode mode;
1609 const char *snapshot_node_name =
1610 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1612 if (node_name && !snapshot_node_name) {
1613 error_setg(errp, "New snapshot node name missing");
1614 return;
1617 if (snapshot_node_name &&
1618 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1619 error_setg(errp, "New snapshot node name already in use");
1620 return;
1623 flags = state->old_bs->open_flags;
1625 /* create new image w/backing file */
1626 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1627 if (mode != NEW_IMAGE_MODE_EXISTING) {
1628 bdrv_img_create(new_image_file, format,
1629 state->old_bs->filename,
1630 state->old_bs->drv->format_name,
1631 NULL, -1, flags, &local_err, false);
1632 if (local_err) {
1633 error_propagate(errp, local_err);
1634 return;
1638 options = qdict_new();
1639 if (s->has_snapshot_node_name) {
1640 qdict_put(options, "node-name",
1641 qstring_from_str(snapshot_node_name));
1643 qdict_put(options, "driver", qstring_from_str(format));
1645 flags |= BDRV_O_NO_BACKING;
1648 assert(state->new_bs == NULL);
1649 ret = bdrv_open(&state->new_bs, new_image_file, snapshot_ref, options,
1650 flags, errp);
1651 /* We will manually add the backing_hd field to the bs later */
1652 if (ret != 0) {
1653 return;
1656 if (state->new_bs->blk != NULL) {
1657 error_setg(errp, "The snapshot is already in use by %s",
1658 blk_name(state->new_bs->blk));
1659 return;
1662 if (bdrv_op_is_blocked(state->new_bs, BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT,
1663 errp)) {
1664 return;
1667 if (state->new_bs->backing != NULL) {
1668 error_setg(errp, "The snapshot already has a backing image");
1669 return;
1672 if (!state->new_bs->drv->supports_backing) {
1673 error_setg(errp, "The snapshot does not support backing images");
1677 static void external_snapshot_commit(BlkActionState *common)
1679 ExternalSnapshotState *state =
1680 DO_UPCAST(ExternalSnapshotState, common, common);
1682 bdrv_set_aio_context(state->new_bs, state->aio_context);
1684 /* This removes our old bs and adds the new bs */
1685 bdrv_append(state->new_bs, state->old_bs);
1686 /* We don't need (or want) to use the transactional
1687 * bdrv_reopen_multiple() across all the entries at once, because we
1688 * don't want to abort all of them if one of them fails the reopen */
1689 bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR,
1690 NULL);
1693 static void external_snapshot_abort(BlkActionState *common)
1695 ExternalSnapshotState *state =
1696 DO_UPCAST(ExternalSnapshotState, common, common);
1697 if (state->new_bs) {
1698 bdrv_unref(state->new_bs);
1702 static void external_snapshot_clean(BlkActionState *common)
1704 ExternalSnapshotState *state =
1705 DO_UPCAST(ExternalSnapshotState, common, common);
1706 if (state->aio_context) {
1707 bdrv_drained_end(state->old_bs);
1708 aio_context_release(state->aio_context);
1712 typedef struct DriveBackupState {
1713 BlkActionState common;
1714 BlockDriverState *bs;
1715 AioContext *aio_context;
1716 BlockJob *job;
1717 } DriveBackupState;
1719 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1721 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1722 BlockBackend *blk;
1723 DriveBackup *backup;
1724 Error *local_err = NULL;
1726 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1727 backup = common->action->u.drive_backup;
1729 blk = blk_by_name(backup->device);
1730 if (!blk) {
1731 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1732 "Device '%s' not found", backup->device);
1733 return;
1736 if (!blk_is_available(blk)) {
1737 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1738 return;
1741 /* AioContext is released in .clean() */
1742 state->aio_context = blk_get_aio_context(blk);
1743 aio_context_acquire(state->aio_context);
1744 bdrv_drained_begin(blk_bs(blk));
1745 state->bs = blk_bs(blk);
1747 qmp_drive_backup(backup->device, backup->target,
1748 backup->has_format, backup->format,
1749 backup->sync,
1750 backup->has_mode, backup->mode,
1751 backup->has_speed, backup->speed,
1752 backup->has_bitmap, backup->bitmap,
1753 backup->has_on_source_error, backup->on_source_error,
1754 backup->has_on_target_error, backup->on_target_error,
1755 &local_err);
1756 if (local_err) {
1757 error_propagate(errp, local_err);
1758 return;
1761 state->job = state->bs->job;
1764 static void drive_backup_abort(BlkActionState *common)
1766 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1767 BlockDriverState *bs = state->bs;
1769 /* Only cancel if it's the job we started */
1770 if (bs && bs->job && bs->job == state->job) {
1771 block_job_cancel_sync(bs->job);
1775 static void drive_backup_clean(BlkActionState *common)
1777 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1779 if (state->aio_context) {
1780 bdrv_drained_end(state->bs);
1781 aio_context_release(state->aio_context);
1785 typedef struct BlockdevBackupState {
1786 BlkActionState common;
1787 BlockDriverState *bs;
1788 BlockJob *job;
1789 AioContext *aio_context;
1790 } BlockdevBackupState;
1792 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1794 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1795 BlockdevBackup *backup;
1796 BlockBackend *blk, *target;
1797 Error *local_err = NULL;
1799 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1800 backup = common->action->u.blockdev_backup;
1802 blk = blk_by_name(backup->device);
1803 if (!blk) {
1804 error_setg(errp, "Device '%s' not found", backup->device);
1805 return;
1808 if (!blk_is_available(blk)) {
1809 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1810 return;
1813 target = blk_by_name(backup->target);
1814 if (!target) {
1815 error_setg(errp, "Device '%s' not found", backup->target);
1816 return;
1819 /* AioContext is released in .clean() */
1820 state->aio_context = blk_get_aio_context(blk);
1821 if (state->aio_context != blk_get_aio_context(target)) {
1822 state->aio_context = NULL;
1823 error_setg(errp, "Backup between two IO threads is not implemented");
1824 return;
1826 aio_context_acquire(state->aio_context);
1827 state->bs = blk_bs(blk);
1828 bdrv_drained_begin(state->bs);
1830 qmp_blockdev_backup(backup->device, backup->target,
1831 backup->sync,
1832 backup->has_speed, backup->speed,
1833 backup->has_on_source_error, backup->on_source_error,
1834 backup->has_on_target_error, backup->on_target_error,
1835 &local_err);
1836 if (local_err) {
1837 error_propagate(errp, local_err);
1838 return;
1841 state->job = state->bs->job;
1844 static void blockdev_backup_abort(BlkActionState *common)
1846 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1847 BlockDriverState *bs = state->bs;
1849 /* Only cancel if it's the job we started */
1850 if (bs && bs->job && bs->job == state->job) {
1851 block_job_cancel_sync(bs->job);
1855 static void blockdev_backup_clean(BlkActionState *common)
1857 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1859 if (state->aio_context) {
1860 bdrv_drained_end(state->bs);
1861 aio_context_release(state->aio_context);
1865 typedef struct BlockDirtyBitmapState {
1866 BlkActionState common;
1867 BdrvDirtyBitmap *bitmap;
1868 BlockDriverState *bs;
1869 AioContext *aio_context;
1870 HBitmap *backup;
1871 bool prepared;
1872 } BlockDirtyBitmapState;
1874 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
1875 Error **errp)
1877 Error *local_err = NULL;
1878 BlockDirtyBitmapAdd *action;
1879 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1880 common, common);
1882 action = common->action->u.block_dirty_bitmap_add;
1883 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1884 qmp_block_dirty_bitmap_add(action->node, action->name,
1885 action->has_granularity, action->granularity,
1886 &local_err);
1888 if (!local_err) {
1889 state->prepared = true;
1890 } else {
1891 error_propagate(errp, local_err);
1895 static void block_dirty_bitmap_add_abort(BlkActionState *common)
1897 BlockDirtyBitmapAdd *action;
1898 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1899 common, common);
1901 action = common->action->u.block_dirty_bitmap_add;
1902 /* Should not be able to fail: IF the bitmap was added via .prepare(),
1903 * then the node reference and bitmap name must have been valid.
1905 if (state->prepared) {
1906 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
1910 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
1911 Error **errp)
1913 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1914 common, common);
1915 BlockDirtyBitmap *action;
1917 action = common->action->u.block_dirty_bitmap_clear;
1918 state->bitmap = block_dirty_bitmap_lookup(action->node,
1919 action->name,
1920 &state->bs,
1921 &state->aio_context,
1922 errp);
1923 if (!state->bitmap) {
1924 return;
1927 if (bdrv_dirty_bitmap_frozen(state->bitmap)) {
1928 error_setg(errp, "Cannot modify a frozen bitmap");
1929 return;
1930 } else if (!bdrv_dirty_bitmap_enabled(state->bitmap)) {
1931 error_setg(errp, "Cannot clear a disabled bitmap");
1932 return;
1935 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
1936 /* AioContext is released in .clean() */
1939 static void block_dirty_bitmap_clear_abort(BlkActionState *common)
1941 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1942 common, common);
1944 bdrv_undo_clear_dirty_bitmap(state->bitmap, state->backup);
1947 static void block_dirty_bitmap_clear_commit(BlkActionState *common)
1949 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1950 common, common);
1952 hbitmap_free(state->backup);
1955 static void block_dirty_bitmap_clear_clean(BlkActionState *common)
1957 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1958 common, common);
1960 if (state->aio_context) {
1961 aio_context_release(state->aio_context);
1965 static void abort_prepare(BlkActionState *common, Error **errp)
1967 error_setg(errp, "Transaction aborted using Abort action");
1970 static void abort_commit(BlkActionState *common)
1972 g_assert_not_reached(); /* this action never succeeds */
1975 static const BlkActionOps actions[] = {
1976 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
1977 .instance_size = sizeof(ExternalSnapshotState),
1978 .prepare = external_snapshot_prepare,
1979 .commit = external_snapshot_commit,
1980 .abort = external_snapshot_abort,
1982 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1983 .instance_size = sizeof(ExternalSnapshotState),
1984 .prepare = external_snapshot_prepare,
1985 .commit = external_snapshot_commit,
1986 .abort = external_snapshot_abort,
1987 .clean = external_snapshot_clean,
1989 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1990 .instance_size = sizeof(DriveBackupState),
1991 .prepare = drive_backup_prepare,
1992 .abort = drive_backup_abort,
1993 .clean = drive_backup_clean,
1995 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
1996 .instance_size = sizeof(BlockdevBackupState),
1997 .prepare = blockdev_backup_prepare,
1998 .abort = blockdev_backup_abort,
1999 .clean = blockdev_backup_clean,
2001 [TRANSACTION_ACTION_KIND_ABORT] = {
2002 .instance_size = sizeof(BlkActionState),
2003 .prepare = abort_prepare,
2004 .commit = abort_commit,
2006 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2007 .instance_size = sizeof(InternalSnapshotState),
2008 .prepare = internal_snapshot_prepare,
2009 .abort = internal_snapshot_abort,
2010 .clean = internal_snapshot_clean,
2012 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2013 .instance_size = sizeof(BlockDirtyBitmapState),
2014 .prepare = block_dirty_bitmap_add_prepare,
2015 .abort = block_dirty_bitmap_add_abort,
2017 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2018 .instance_size = sizeof(BlockDirtyBitmapState),
2019 .prepare = block_dirty_bitmap_clear_prepare,
2020 .commit = block_dirty_bitmap_clear_commit,
2021 .abort = block_dirty_bitmap_clear_abort,
2022 .clean = block_dirty_bitmap_clear_clean,
2027 * 'Atomic' group operations. The operations are performed as a set, and if
2028 * any fail then we roll back all operations in the group.
2030 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
2032 TransactionActionList *dev_entry = dev_list;
2033 BlkActionState *state, *next;
2034 Error *local_err = NULL;
2036 QSIMPLEQ_HEAD(snap_bdrv_states, BlkActionState) snap_bdrv_states;
2037 QSIMPLEQ_INIT(&snap_bdrv_states);
2039 /* drain all i/o before any operations */
2040 bdrv_drain_all();
2042 /* We don't do anything in this loop that commits us to the operations */
2043 while (NULL != dev_entry) {
2044 TransactionAction *dev_info = NULL;
2045 const BlkActionOps *ops;
2047 dev_info = dev_entry->value;
2048 dev_entry = dev_entry->next;
2050 assert(dev_info->type < ARRAY_SIZE(actions));
2052 ops = &actions[dev_info->type];
2053 assert(ops->instance_size > 0);
2055 state = g_malloc0(ops->instance_size);
2056 state->ops = ops;
2057 state->action = dev_info;
2058 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2060 state->ops->prepare(state, &local_err);
2061 if (local_err) {
2062 error_propagate(errp, local_err);
2063 goto delete_and_fail;
2067 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
2068 if (state->ops->commit) {
2069 state->ops->commit(state);
2073 /* success */
2074 goto exit;
2076 delete_and_fail:
2077 /* failure, and it is all-or-none; roll back all operations */
2078 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
2079 if (state->ops->abort) {
2080 state->ops->abort(state);
2083 exit:
2084 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2085 if (state->ops->clean) {
2086 state->ops->clean(state);
2088 g_free(state);
2092 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
2094 Error *local_err = NULL;
2096 qmp_blockdev_open_tray(device, has_force, force, &local_err);
2097 if (local_err) {
2098 error_propagate(errp, local_err);
2099 return;
2102 qmp_blockdev_remove_medium(device, errp);
2105 void qmp_block_passwd(bool has_device, const char *device,
2106 bool has_node_name, const char *node_name,
2107 const char *password, Error **errp)
2109 Error *local_err = NULL;
2110 BlockDriverState *bs;
2111 AioContext *aio_context;
2113 bs = bdrv_lookup_bs(has_device ? device : NULL,
2114 has_node_name ? node_name : NULL,
2115 &local_err);
2116 if (local_err) {
2117 error_propagate(errp, local_err);
2118 return;
2121 aio_context = bdrv_get_aio_context(bs);
2122 aio_context_acquire(aio_context);
2124 bdrv_add_key(bs, password, errp);
2126 aio_context_release(aio_context);
2129 void qmp_blockdev_open_tray(const char *device, bool has_force, bool force,
2130 Error **errp)
2132 BlockBackend *blk;
2133 bool locked;
2135 if (!has_force) {
2136 force = false;
2139 blk = blk_by_name(device);
2140 if (!blk) {
2141 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2142 "Device '%s' not found", device);
2143 return;
2146 if (!blk_dev_has_removable_media(blk)) {
2147 error_setg(errp, "Device '%s' is not removable", device);
2148 return;
2151 if (blk_dev_is_tray_open(blk)) {
2152 return;
2155 locked = blk_dev_is_medium_locked(blk);
2156 if (locked) {
2157 blk_dev_eject_request(blk, force);
2160 if (!locked || force) {
2161 blk_dev_change_media_cb(blk, false);
2165 void qmp_blockdev_close_tray(const char *device, Error **errp)
2167 BlockBackend *blk;
2169 blk = blk_by_name(device);
2170 if (!blk) {
2171 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2172 "Device '%s' not found", device);
2173 return;
2176 if (!blk_dev_has_removable_media(blk)) {
2177 error_setg(errp, "Device '%s' is not removable", device);
2178 return;
2181 if (!blk_dev_is_tray_open(blk)) {
2182 return;
2185 blk_dev_change_media_cb(blk, true);
2188 void qmp_blockdev_remove_medium(const char *device, Error **errp)
2190 BlockBackend *blk;
2191 BlockDriverState *bs;
2192 AioContext *aio_context;
2193 bool has_device;
2195 blk = blk_by_name(device);
2196 if (!blk) {
2197 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2198 "Device '%s' not found", device);
2199 return;
2202 /* For BBs without a device, we can exchange the BDS tree at will */
2203 has_device = blk_get_attached_dev(blk);
2205 if (has_device && !blk_dev_has_removable_media(blk)) {
2206 error_setg(errp, "Device '%s' is not removable", device);
2207 return;
2210 if (has_device && !blk_dev_is_tray_open(blk)) {
2211 error_setg(errp, "Tray of device '%s' is not open", device);
2212 return;
2215 bs = blk_bs(blk);
2216 if (!bs) {
2217 return;
2220 aio_context = bdrv_get_aio_context(bs);
2221 aio_context_acquire(aio_context);
2223 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
2224 goto out;
2227 /* This follows the convention established by bdrv_make_anon() */
2228 if (bs->device_list.tqe_prev) {
2229 QTAILQ_REMOVE(&bdrv_states, bs, device_list);
2230 bs->device_list.tqe_prev = NULL;
2233 blk_remove_bs(blk);
2235 out:
2236 aio_context_release(aio_context);
2239 static void qmp_blockdev_insert_anon_medium(const char *device,
2240 BlockDriverState *bs, Error **errp)
2242 BlockBackend *blk;
2243 bool has_device;
2245 blk = blk_by_name(device);
2246 if (!blk) {
2247 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2248 "Device '%s' not found", device);
2249 return;
2252 /* For BBs without a device, we can exchange the BDS tree at will */
2253 has_device = blk_get_attached_dev(blk);
2255 if (has_device && !blk_dev_has_removable_media(blk)) {
2256 error_setg(errp, "Device '%s' is not removable", device);
2257 return;
2260 if (has_device && !blk_dev_is_tray_open(blk)) {
2261 error_setg(errp, "Tray of device '%s' is not open", device);
2262 return;
2265 if (blk_bs(blk)) {
2266 error_setg(errp, "There already is a medium in device '%s'", device);
2267 return;
2270 blk_insert_bs(blk, bs);
2272 QTAILQ_INSERT_TAIL(&bdrv_states, bs, device_list);
2275 void qmp_blockdev_insert_medium(const char *device, const char *node_name,
2276 Error **errp)
2278 BlockDriverState *bs;
2280 bs = bdrv_find_node(node_name);
2281 if (!bs) {
2282 error_setg(errp, "Node '%s' not found", node_name);
2283 return;
2286 if (bs->blk) {
2287 error_setg(errp, "Node '%s' is already in use by '%s'", node_name,
2288 blk_name(bs->blk));
2289 return;
2292 qmp_blockdev_insert_anon_medium(device, bs, errp);
2295 void qmp_blockdev_change_medium(const char *device, const char *filename,
2296 bool has_format, const char *format,
2297 bool has_read_only,
2298 BlockdevChangeReadOnlyMode read_only,
2299 Error **errp)
2301 BlockBackend *blk;
2302 BlockDriverState *medium_bs = NULL;
2303 int bdrv_flags, ret;
2304 QDict *options = NULL;
2305 Error *err = NULL;
2307 blk = blk_by_name(device);
2308 if (!blk) {
2309 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2310 "Device '%s' not found", device);
2311 goto fail;
2314 if (blk_bs(blk)) {
2315 blk_update_root_state(blk);
2318 bdrv_flags = blk_get_open_flags_from_root_state(blk);
2320 if (!has_read_only) {
2321 read_only = BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN;
2324 switch (read_only) {
2325 case BLOCKDEV_CHANGE_READ_ONLY_MODE_RETAIN:
2326 break;
2328 case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_ONLY:
2329 bdrv_flags &= ~BDRV_O_RDWR;
2330 break;
2332 case BLOCKDEV_CHANGE_READ_ONLY_MODE_READ_WRITE:
2333 bdrv_flags |= BDRV_O_RDWR;
2334 break;
2336 default:
2337 abort();
2340 if (has_format) {
2341 options = qdict_new();
2342 qdict_put(options, "driver", qstring_from_str(format));
2345 assert(!medium_bs);
2346 ret = bdrv_open(&medium_bs, filename, NULL, options, bdrv_flags, errp);
2347 if (ret < 0) {
2348 goto fail;
2351 blk_apply_root_state(blk, medium_bs);
2353 bdrv_add_key(medium_bs, NULL, &err);
2354 if (err) {
2355 error_propagate(errp, err);
2356 goto fail;
2359 qmp_blockdev_open_tray(device, false, false, &err);
2360 if (err) {
2361 error_propagate(errp, err);
2362 goto fail;
2365 qmp_blockdev_remove_medium(device, &err);
2366 if (err) {
2367 error_propagate(errp, err);
2368 goto fail;
2371 qmp_blockdev_insert_anon_medium(device, medium_bs, &err);
2372 if (err) {
2373 error_propagate(errp, err);
2374 goto fail;
2377 qmp_blockdev_close_tray(device, errp);
2379 fail:
2380 /* If the medium has been inserted, the device has its own reference, so
2381 * ours must be relinquished; and if it has not been inserted successfully,
2382 * the reference must be relinquished anyway */
2383 bdrv_unref(medium_bs);
2386 /* throttling disk I/O limits */
2387 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
2388 int64_t bps_wr,
2389 int64_t iops,
2390 int64_t iops_rd,
2391 int64_t iops_wr,
2392 bool has_bps_max,
2393 int64_t bps_max,
2394 bool has_bps_rd_max,
2395 int64_t bps_rd_max,
2396 bool has_bps_wr_max,
2397 int64_t bps_wr_max,
2398 bool has_iops_max,
2399 int64_t iops_max,
2400 bool has_iops_rd_max,
2401 int64_t iops_rd_max,
2402 bool has_iops_wr_max,
2403 int64_t iops_wr_max,
2404 bool has_iops_size,
2405 int64_t iops_size,
2406 bool has_group,
2407 const char *group, Error **errp)
2409 ThrottleConfig cfg;
2410 BlockDriverState *bs;
2411 BlockBackend *blk;
2412 AioContext *aio_context;
2414 blk = blk_by_name(device);
2415 if (!blk) {
2416 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2417 "Device '%s' not found", device);
2418 return;
2421 aio_context = blk_get_aio_context(blk);
2422 aio_context_acquire(aio_context);
2424 bs = blk_bs(blk);
2425 if (!bs) {
2426 error_setg(errp, "Device '%s' has no medium", device);
2427 goto out;
2430 memset(&cfg, 0, sizeof(cfg));
2431 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
2432 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
2433 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
2435 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
2436 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
2437 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
2439 if (has_bps_max) {
2440 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
2442 if (has_bps_rd_max) {
2443 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
2445 if (has_bps_wr_max) {
2446 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
2448 if (has_iops_max) {
2449 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
2451 if (has_iops_rd_max) {
2452 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
2454 if (has_iops_wr_max) {
2455 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
2458 if (has_iops_size) {
2459 cfg.op_size = iops_size;
2462 if (!check_throttle_config(&cfg, errp)) {
2463 goto out;
2466 if (throttle_enabled(&cfg)) {
2467 /* Enable I/O limits if they're not enabled yet, otherwise
2468 * just update the throttling group. */
2469 if (!bs->throttle_state) {
2470 bdrv_io_limits_enable(bs, has_group ? group : device);
2471 } else if (has_group) {
2472 bdrv_io_limits_update_group(bs, group);
2474 /* Set the new throttling configuration */
2475 bdrv_set_io_limits(bs, &cfg);
2476 } else if (bs->throttle_state) {
2477 /* If all throttling settings are set to 0, disable I/O limits */
2478 bdrv_io_limits_disable(bs);
2481 out:
2482 aio_context_release(aio_context);
2485 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2486 bool has_granularity, uint32_t granularity,
2487 Error **errp)
2489 AioContext *aio_context;
2490 BlockDriverState *bs;
2492 if (!name || name[0] == '\0') {
2493 error_setg(errp, "Bitmap name cannot be empty");
2494 return;
2497 bs = bdrv_lookup_bs(node, node, errp);
2498 if (!bs) {
2499 return;
2502 aio_context = bdrv_get_aio_context(bs);
2503 aio_context_acquire(aio_context);
2505 if (has_granularity) {
2506 if (granularity < 512 || !is_power_of_2(granularity)) {
2507 error_setg(errp, "Granularity must be power of 2 "
2508 "and at least 512");
2509 goto out;
2511 } else {
2512 /* Default to cluster size, if available: */
2513 granularity = bdrv_get_default_bitmap_granularity(bs);
2516 bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2518 out:
2519 aio_context_release(aio_context);
2522 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2523 Error **errp)
2525 AioContext *aio_context;
2526 BlockDriverState *bs;
2527 BdrvDirtyBitmap *bitmap;
2529 bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2530 if (!bitmap || !bs) {
2531 return;
2534 if (bdrv_dirty_bitmap_frozen(bitmap)) {
2535 error_setg(errp,
2536 "Bitmap '%s' is currently frozen and cannot be removed",
2537 name);
2538 goto out;
2540 bdrv_dirty_bitmap_make_anon(bitmap);
2541 bdrv_release_dirty_bitmap(bs, bitmap);
2543 out:
2544 aio_context_release(aio_context);
2548 * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2549 * immediately after a full backup operation.
2551 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2552 Error **errp)
2554 AioContext *aio_context;
2555 BdrvDirtyBitmap *bitmap;
2556 BlockDriverState *bs;
2558 bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2559 if (!bitmap || !bs) {
2560 return;
2563 if (bdrv_dirty_bitmap_frozen(bitmap)) {
2564 error_setg(errp,
2565 "Bitmap '%s' is currently frozen and cannot be modified",
2566 name);
2567 goto out;
2568 } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
2569 error_setg(errp,
2570 "Bitmap '%s' is currently disabled and cannot be cleared",
2571 name);
2572 goto out;
2575 bdrv_clear_dirty_bitmap(bitmap, NULL);
2577 out:
2578 aio_context_release(aio_context);
2581 void hmp_drive_del(Monitor *mon, const QDict *qdict)
2583 const char *id = qdict_get_str(qdict, "id");
2584 BlockBackend *blk;
2585 BlockDriverState *bs;
2586 AioContext *aio_context;
2587 Error *local_err = NULL;
2589 blk = blk_by_name(id);
2590 if (!blk) {
2591 error_report("Device '%s' not found", id);
2592 return;
2595 if (!blk_legacy_dinfo(blk)) {
2596 error_report("Deleting device added with blockdev-add"
2597 " is not supported");
2598 return;
2601 aio_context = blk_get_aio_context(blk);
2602 aio_context_acquire(aio_context);
2604 bs = blk_bs(blk);
2605 if (bs) {
2606 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2607 error_report_err(local_err);
2608 aio_context_release(aio_context);
2609 return;
2612 bdrv_close(bs);
2615 /* if we have a device attached to this BlockDriverState
2616 * then we need to make the drive anonymous until the device
2617 * can be removed. If this is a drive with no device backing
2618 * then we can just get rid of the block driver state right here.
2620 if (blk_get_attached_dev(blk)) {
2621 blk_hide_on_behalf_of_hmp_drive_del(blk);
2622 /* Further I/O must not pause the guest */
2623 blk_set_on_error(blk, BLOCKDEV_ON_ERROR_REPORT,
2624 BLOCKDEV_ON_ERROR_REPORT);
2625 } else {
2626 blk_unref(blk);
2629 aio_context_release(aio_context);
2632 void qmp_block_resize(bool has_device, const char *device,
2633 bool has_node_name, const char *node_name,
2634 int64_t size, Error **errp)
2636 Error *local_err = NULL;
2637 BlockDriverState *bs;
2638 AioContext *aio_context;
2639 int ret;
2641 bs = bdrv_lookup_bs(has_device ? device : NULL,
2642 has_node_name ? node_name : NULL,
2643 &local_err);
2644 if (local_err) {
2645 error_propagate(errp, local_err);
2646 return;
2649 aio_context = bdrv_get_aio_context(bs);
2650 aio_context_acquire(aio_context);
2652 if (!bdrv_is_first_non_filter(bs)) {
2653 error_setg(errp, QERR_FEATURE_DISABLED, "resize");
2654 goto out;
2657 if (size < 0) {
2658 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2659 goto out;
2662 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2663 error_setg(errp, QERR_DEVICE_IN_USE, device);
2664 goto out;
2667 /* complete all in-flight operations before resizing the device */
2668 bdrv_drain_all();
2670 ret = bdrv_truncate(bs, size);
2671 switch (ret) {
2672 case 0:
2673 break;
2674 case -ENOMEDIUM:
2675 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2676 break;
2677 case -ENOTSUP:
2678 error_setg(errp, QERR_UNSUPPORTED);
2679 break;
2680 case -EACCES:
2681 error_setg(errp, "Device '%s' is read only", device);
2682 break;
2683 case -EBUSY:
2684 error_setg(errp, QERR_DEVICE_IN_USE, device);
2685 break;
2686 default:
2687 error_setg_errno(errp, -ret, "Could not resize");
2688 break;
2691 out:
2692 aio_context_release(aio_context);
2695 static void block_job_cb(void *opaque, int ret)
2697 /* Note that this function may be executed from another AioContext besides
2698 * the QEMU main loop. If you need to access anything that assumes the
2699 * QEMU global mutex, use a BH or introduce a mutex.
2702 BlockDriverState *bs = opaque;
2703 const char *msg = NULL;
2705 trace_block_job_cb(bs, bs->job, ret);
2707 assert(bs->job);
2709 if (ret < 0) {
2710 msg = strerror(-ret);
2713 if (block_job_is_cancelled(bs->job)) {
2714 block_job_event_cancelled(bs->job);
2715 } else {
2716 block_job_event_completed(bs->job, msg);
2720 void qmp_block_stream(const char *device,
2721 bool has_base, const char *base,
2722 bool has_backing_file, const char *backing_file,
2723 bool has_speed, int64_t speed,
2724 bool has_on_error, BlockdevOnError on_error,
2725 Error **errp)
2727 BlockBackend *blk;
2728 BlockDriverState *bs;
2729 BlockDriverState *base_bs = NULL;
2730 AioContext *aio_context;
2731 Error *local_err = NULL;
2732 const char *base_name = NULL;
2734 if (!has_on_error) {
2735 on_error = BLOCKDEV_ON_ERROR_REPORT;
2738 blk = blk_by_name(device);
2739 if (!blk) {
2740 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2741 "Device '%s' not found", device);
2742 return;
2745 aio_context = blk_get_aio_context(blk);
2746 aio_context_acquire(aio_context);
2748 if (!blk_is_available(blk)) {
2749 error_setg(errp, "Device '%s' has no medium", device);
2750 goto out;
2752 bs = blk_bs(blk);
2754 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
2755 goto out;
2758 if (has_base) {
2759 base_bs = bdrv_find_backing_image(bs, base);
2760 if (base_bs == NULL) {
2761 error_setg(errp, QERR_BASE_NOT_FOUND, base);
2762 goto out;
2764 assert(bdrv_get_aio_context(base_bs) == aio_context);
2765 base_name = base;
2768 /* if we are streaming the entire chain, the result will have no backing
2769 * file, and specifying one is therefore an error */
2770 if (base_bs == NULL && has_backing_file) {
2771 error_setg(errp, "backing file specified, but streaming the "
2772 "entire chain");
2773 goto out;
2776 /* backing_file string overrides base bs filename */
2777 base_name = has_backing_file ? backing_file : base_name;
2779 stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2780 on_error, block_job_cb, bs, &local_err);
2781 if (local_err) {
2782 error_propagate(errp, local_err);
2783 goto out;
2786 trace_qmp_block_stream(bs, bs->job);
2788 out:
2789 aio_context_release(aio_context);
2792 void qmp_block_commit(const char *device,
2793 bool has_base, const char *base,
2794 bool has_top, const char *top,
2795 bool has_backing_file, const char *backing_file,
2796 bool has_speed, int64_t speed,
2797 Error **errp)
2799 BlockBackend *blk;
2800 BlockDriverState *bs;
2801 BlockDriverState *base_bs, *top_bs;
2802 AioContext *aio_context;
2803 Error *local_err = NULL;
2804 /* This will be part of the QMP command, if/when the
2805 * BlockdevOnError change for blkmirror makes it in
2807 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2809 if (!has_speed) {
2810 speed = 0;
2813 /* Important Note:
2814 * libvirt relies on the DeviceNotFound error class in order to probe for
2815 * live commit feature versions; for this to work, we must make sure to
2816 * perform the device lookup before any generic errors that may occur in a
2817 * scenario in which all optional arguments are omitted. */
2818 blk = blk_by_name(device);
2819 if (!blk) {
2820 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2821 "Device '%s' not found", device);
2822 return;
2825 aio_context = blk_get_aio_context(blk);
2826 aio_context_acquire(aio_context);
2828 if (!blk_is_available(blk)) {
2829 error_setg(errp, "Device '%s' has no medium", device);
2830 goto out;
2832 bs = blk_bs(blk);
2834 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2835 goto out;
2838 /* default top_bs is the active layer */
2839 top_bs = bs;
2841 if (has_top && top) {
2842 if (strcmp(bs->filename, top) != 0) {
2843 top_bs = bdrv_find_backing_image(bs, top);
2847 if (top_bs == NULL) {
2848 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2849 goto out;
2852 assert(bdrv_get_aio_context(top_bs) == aio_context);
2854 if (has_base && base) {
2855 base_bs = bdrv_find_backing_image(top_bs, base);
2856 } else {
2857 base_bs = bdrv_find_base(top_bs);
2860 if (base_bs == NULL) {
2861 error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2862 goto out;
2865 assert(bdrv_get_aio_context(base_bs) == aio_context);
2867 if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2868 goto out;
2871 /* Do not allow attempts to commit an image into itself */
2872 if (top_bs == base_bs) {
2873 error_setg(errp, "cannot commit an image into itself");
2874 goto out;
2877 if (top_bs == bs) {
2878 if (has_backing_file) {
2879 error_setg(errp, "'backing-file' specified,"
2880 " but 'top' is the active layer");
2881 goto out;
2883 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2884 bs, &local_err);
2885 } else {
2886 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2887 has_backing_file ? backing_file : NULL, &local_err);
2889 if (local_err != NULL) {
2890 error_propagate(errp, local_err);
2891 goto out;
2894 out:
2895 aio_context_release(aio_context);
2898 void qmp_drive_backup(const char *device, const char *target,
2899 bool has_format, const char *format,
2900 enum MirrorSyncMode sync,
2901 bool has_mode, enum NewImageMode mode,
2902 bool has_speed, int64_t speed,
2903 bool has_bitmap, const char *bitmap,
2904 bool has_on_source_error, BlockdevOnError on_source_error,
2905 bool has_on_target_error, BlockdevOnError on_target_error,
2906 Error **errp)
2908 BlockBackend *blk;
2909 BlockDriverState *bs;
2910 BlockDriverState *target_bs;
2911 BlockDriverState *source = NULL;
2912 BdrvDirtyBitmap *bmap = NULL;
2913 AioContext *aio_context;
2914 QDict *options = NULL;
2915 Error *local_err = NULL;
2916 int flags;
2917 int64_t size;
2918 int ret;
2920 if (!has_speed) {
2921 speed = 0;
2923 if (!has_on_source_error) {
2924 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2926 if (!has_on_target_error) {
2927 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2929 if (!has_mode) {
2930 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2933 blk = blk_by_name(device);
2934 if (!blk) {
2935 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2936 "Device '%s' not found", device);
2937 return;
2940 aio_context = blk_get_aio_context(blk);
2941 aio_context_acquire(aio_context);
2943 /* Although backup_run has this check too, we need to use bs->drv below, so
2944 * do an early check redundantly. */
2945 if (!blk_is_available(blk)) {
2946 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2947 goto out;
2949 bs = blk_bs(blk);
2951 if (!has_format) {
2952 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2955 /* Early check to avoid creating target */
2956 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2957 goto out;
2960 flags = bs->open_flags | BDRV_O_RDWR;
2962 /* See if we have a backing HD we can use to create our new image
2963 * on top of. */
2964 if (sync == MIRROR_SYNC_MODE_TOP) {
2965 source = backing_bs(bs);
2966 if (!source) {
2967 sync = MIRROR_SYNC_MODE_FULL;
2970 if (sync == MIRROR_SYNC_MODE_NONE) {
2971 source = bs;
2974 size = bdrv_getlength(bs);
2975 if (size < 0) {
2976 error_setg_errno(errp, -size, "bdrv_getlength failed");
2977 goto out;
2980 if (mode != NEW_IMAGE_MODE_EXISTING) {
2981 assert(format);
2982 if (source) {
2983 bdrv_img_create(target, format, source->filename,
2984 source->drv->format_name, NULL,
2985 size, flags, &local_err, false);
2986 } else {
2987 bdrv_img_create(target, format, NULL, NULL, NULL,
2988 size, flags, &local_err, false);
2992 if (local_err) {
2993 error_propagate(errp, local_err);
2994 goto out;
2997 if (format) {
2998 options = qdict_new();
2999 qdict_put(options, "driver", qstring_from_str(format));
3002 target_bs = NULL;
3003 ret = bdrv_open(&target_bs, target, NULL, options, flags, &local_err);
3004 if (ret < 0) {
3005 error_propagate(errp, local_err);
3006 goto out;
3009 bdrv_set_aio_context(target_bs, aio_context);
3011 if (has_bitmap) {
3012 bmap = bdrv_find_dirty_bitmap(bs, bitmap);
3013 if (!bmap) {
3014 error_setg(errp, "Bitmap '%s' could not be found", bitmap);
3015 goto out;
3019 backup_start(bs, target_bs, speed, sync, bmap,
3020 on_source_error, on_target_error,
3021 block_job_cb, bs, &local_err);
3022 if (local_err != NULL) {
3023 bdrv_unref(target_bs);
3024 error_propagate(errp, local_err);
3025 goto out;
3028 out:
3029 aio_context_release(aio_context);
3032 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
3034 return bdrv_named_nodes_list(errp);
3037 void qmp_blockdev_backup(const char *device, const char *target,
3038 enum MirrorSyncMode sync,
3039 bool has_speed, int64_t speed,
3040 bool has_on_source_error,
3041 BlockdevOnError on_source_error,
3042 bool has_on_target_error,
3043 BlockdevOnError on_target_error,
3044 Error **errp)
3046 BlockBackend *blk, *target_blk;
3047 BlockDriverState *bs;
3048 BlockDriverState *target_bs;
3049 Error *local_err = NULL;
3050 AioContext *aio_context;
3052 if (!has_speed) {
3053 speed = 0;
3055 if (!has_on_source_error) {
3056 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3058 if (!has_on_target_error) {
3059 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3062 blk = blk_by_name(device);
3063 if (!blk) {
3064 error_setg(errp, "Device '%s' not found", device);
3065 return;
3068 aio_context = blk_get_aio_context(blk);
3069 aio_context_acquire(aio_context);
3071 if (!blk_is_available(blk)) {
3072 error_setg(errp, "Device '%s' has no medium", device);
3073 goto out;
3075 bs = blk_bs(blk);
3077 target_blk = blk_by_name(target);
3078 if (!target_blk) {
3079 error_setg(errp, "Device '%s' not found", target);
3080 goto out;
3083 if (!blk_is_available(target_blk)) {
3084 error_setg(errp, "Device '%s' has no medium", target);
3085 goto out;
3087 target_bs = blk_bs(target_blk);
3089 bdrv_ref(target_bs);
3090 bdrv_set_aio_context(target_bs, aio_context);
3091 backup_start(bs, target_bs, speed, sync, NULL, on_source_error,
3092 on_target_error, block_job_cb, bs, &local_err);
3093 if (local_err != NULL) {
3094 bdrv_unref(target_bs);
3095 error_propagate(errp, local_err);
3097 out:
3098 aio_context_release(aio_context);
3101 void qmp_drive_mirror(const char *device, const char *target,
3102 bool has_format, const char *format,
3103 bool has_node_name, const char *node_name,
3104 bool has_replaces, const char *replaces,
3105 enum MirrorSyncMode sync,
3106 bool has_mode, enum NewImageMode mode,
3107 bool has_speed, int64_t speed,
3108 bool has_granularity, uint32_t granularity,
3109 bool has_buf_size, int64_t buf_size,
3110 bool has_on_source_error, BlockdevOnError on_source_error,
3111 bool has_on_target_error, BlockdevOnError on_target_error,
3112 bool has_unmap, bool unmap,
3113 Error **errp)
3115 BlockBackend *blk;
3116 BlockDriverState *bs;
3117 BlockDriverState *source, *target_bs;
3118 AioContext *aio_context;
3119 Error *local_err = NULL;
3120 QDict *options;
3121 int flags;
3122 int64_t size;
3123 int ret;
3125 if (!has_speed) {
3126 speed = 0;
3128 if (!has_on_source_error) {
3129 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3131 if (!has_on_target_error) {
3132 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3134 if (!has_mode) {
3135 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3137 if (!has_granularity) {
3138 granularity = 0;
3140 if (!has_buf_size) {
3141 buf_size = 0;
3143 if (!has_unmap) {
3144 unmap = true;
3147 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3148 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3149 "a value in range [512B, 64MB]");
3150 return;
3152 if (granularity & (granularity - 1)) {
3153 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3154 "power of 2");
3155 return;
3158 blk = blk_by_name(device);
3159 if (!blk) {
3160 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3161 "Device '%s' not found", device);
3162 return;
3165 aio_context = blk_get_aio_context(blk);
3166 aio_context_acquire(aio_context);
3168 if (!blk_is_available(blk)) {
3169 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
3170 goto out;
3172 bs = blk_bs(blk);
3174 if (!has_format) {
3175 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
3178 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
3179 goto out;
3182 flags = bs->open_flags | BDRV_O_RDWR;
3183 source = backing_bs(bs);
3184 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
3185 sync = MIRROR_SYNC_MODE_FULL;
3187 if (sync == MIRROR_SYNC_MODE_NONE) {
3188 source = bs;
3191 size = bdrv_getlength(bs);
3192 if (size < 0) {
3193 error_setg_errno(errp, -size, "bdrv_getlength failed");
3194 goto out;
3197 if (has_replaces) {
3198 BlockDriverState *to_replace_bs;
3199 AioContext *replace_aio_context;
3200 int64_t replace_size;
3202 if (!has_node_name) {
3203 error_setg(errp, "a node-name must be provided when replacing a"
3204 " named node of the graph");
3205 goto out;
3208 to_replace_bs = check_to_replace_node(bs, replaces, &local_err);
3210 if (!to_replace_bs) {
3211 error_propagate(errp, local_err);
3212 goto out;
3215 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3216 aio_context_acquire(replace_aio_context);
3217 replace_size = bdrv_getlength(to_replace_bs);
3218 aio_context_release(replace_aio_context);
3220 if (size != replace_size) {
3221 error_setg(errp, "cannot replace image with a mirror image of "
3222 "different size");
3223 goto out;
3227 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
3228 && mode != NEW_IMAGE_MODE_EXISTING)
3230 /* create new image w/o backing file */
3231 assert(format);
3232 bdrv_img_create(target, format,
3233 NULL, NULL, NULL, size, flags, &local_err, false);
3234 } else {
3235 switch (mode) {
3236 case NEW_IMAGE_MODE_EXISTING:
3237 break;
3238 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3239 /* create new image with backing file */
3240 bdrv_img_create(target, format,
3241 source->filename,
3242 source->drv->format_name,
3243 NULL, size, flags, &local_err, false);
3244 break;
3245 default:
3246 abort();
3250 if (local_err) {
3251 error_propagate(errp, local_err);
3252 goto out;
3255 options = qdict_new();
3256 if (has_node_name) {
3257 qdict_put(options, "node-name", qstring_from_str(node_name));
3259 if (format) {
3260 qdict_put(options, "driver", qstring_from_str(format));
3263 /* Mirroring takes care of copy-on-write using the source's backing
3264 * file.
3266 target_bs = NULL;
3267 ret = bdrv_open(&target_bs, target, NULL, options,
3268 flags | BDRV_O_NO_BACKING, &local_err);
3269 if (ret < 0) {
3270 error_propagate(errp, local_err);
3271 goto out;
3274 bdrv_set_aio_context(target_bs, aio_context);
3276 /* pass the node name to replace to mirror start since it's loose coupling
3277 * and will allow to check whether the node still exist at mirror completion
3279 mirror_start(bs, target_bs,
3280 has_replaces ? replaces : NULL,
3281 speed, granularity, buf_size, sync,
3282 on_source_error, on_target_error,
3283 unmap,
3284 block_job_cb, bs, &local_err);
3285 if (local_err != NULL) {
3286 bdrv_unref(target_bs);
3287 error_propagate(errp, local_err);
3288 goto out;
3291 out:
3292 aio_context_release(aio_context);
3295 /* Get the block job for a given device name and acquire its AioContext */
3296 static BlockJob *find_block_job(const char *device, AioContext **aio_context,
3297 Error **errp)
3299 BlockBackend *blk;
3300 BlockDriverState *bs;
3302 *aio_context = NULL;
3304 blk = blk_by_name(device);
3305 if (!blk) {
3306 goto notfound;
3309 *aio_context = blk_get_aio_context(blk);
3310 aio_context_acquire(*aio_context);
3312 if (!blk_is_available(blk)) {
3313 goto notfound;
3315 bs = blk_bs(blk);
3317 if (!bs->job) {
3318 goto notfound;
3321 return bs->job;
3323 notfound:
3324 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3325 "No active block job on device '%s'", device);
3326 if (*aio_context) {
3327 aio_context_release(*aio_context);
3328 *aio_context = NULL;
3330 return NULL;
3333 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3335 AioContext *aio_context;
3336 BlockJob *job = find_block_job(device, &aio_context, errp);
3338 if (!job) {
3339 return;
3342 block_job_set_speed(job, speed, errp);
3343 aio_context_release(aio_context);
3346 void qmp_block_job_cancel(const char *device,
3347 bool has_force, bool force, Error **errp)
3349 AioContext *aio_context;
3350 BlockJob *job = find_block_job(device, &aio_context, errp);
3352 if (!job) {
3353 return;
3356 if (!has_force) {
3357 force = false;
3360 if (job->user_paused && !force) {
3361 error_setg(errp, "The block job for device '%s' is currently paused",
3362 device);
3363 goto out;
3366 trace_qmp_block_job_cancel(job);
3367 block_job_cancel(job);
3368 out:
3369 aio_context_release(aio_context);
3372 void qmp_block_job_pause(const char *device, Error **errp)
3374 AioContext *aio_context;
3375 BlockJob *job = find_block_job(device, &aio_context, errp);
3377 if (!job || job->user_paused) {
3378 return;
3381 job->user_paused = true;
3382 trace_qmp_block_job_pause(job);
3383 block_job_pause(job);
3384 aio_context_release(aio_context);
3387 void qmp_block_job_resume(const char *device, Error **errp)
3389 AioContext *aio_context;
3390 BlockJob *job = find_block_job(device, &aio_context, errp);
3392 if (!job || !job->user_paused) {
3393 return;
3396 job->user_paused = false;
3397 trace_qmp_block_job_resume(job);
3398 block_job_resume(job);
3399 aio_context_release(aio_context);
3402 void qmp_block_job_complete(const char *device, Error **errp)
3404 AioContext *aio_context;
3405 BlockJob *job = find_block_job(device, &aio_context, errp);
3407 if (!job) {
3408 return;
3411 trace_qmp_block_job_complete(job);
3412 block_job_complete(job, errp);
3413 aio_context_release(aio_context);
3416 void qmp_change_backing_file(const char *device,
3417 const char *image_node_name,
3418 const char *backing_file,
3419 Error **errp)
3421 BlockBackend *blk;
3422 BlockDriverState *bs = NULL;
3423 AioContext *aio_context;
3424 BlockDriverState *image_bs = NULL;
3425 Error *local_err = NULL;
3426 bool ro;
3427 int open_flags;
3428 int ret;
3430 blk = blk_by_name(device);
3431 if (!blk) {
3432 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3433 "Device '%s' not found", device);
3434 return;
3437 aio_context = blk_get_aio_context(blk);
3438 aio_context_acquire(aio_context);
3440 if (!blk_is_available(blk)) {
3441 error_setg(errp, "Device '%s' has no medium", device);
3442 goto out;
3444 bs = blk_bs(blk);
3446 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3447 if (local_err) {
3448 error_propagate(errp, local_err);
3449 goto out;
3452 if (!image_bs) {
3453 error_setg(errp, "image file not found");
3454 goto out;
3457 if (bdrv_find_base(image_bs) == image_bs) {
3458 error_setg(errp, "not allowing backing file change on an image "
3459 "without a backing file");
3460 goto out;
3463 /* even though we are not necessarily operating on bs, we need it to
3464 * determine if block ops are currently prohibited on the chain */
3465 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3466 goto out;
3469 /* final sanity check */
3470 if (!bdrv_chain_contains(bs, image_bs)) {
3471 error_setg(errp, "'%s' and image file are not in the same chain",
3472 device);
3473 goto out;
3476 /* if not r/w, reopen to make r/w */
3477 open_flags = image_bs->open_flags;
3478 ro = bdrv_is_read_only(image_bs);
3480 if (ro) {
3481 bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
3482 if (local_err) {
3483 error_propagate(errp, local_err);
3484 goto out;
3488 ret = bdrv_change_backing_file(image_bs, backing_file,
3489 image_bs->drv ? image_bs->drv->format_name : "");
3491 if (ret < 0) {
3492 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3493 backing_file);
3494 /* don't exit here, so we can try to restore open flags if
3495 * appropriate */
3498 if (ro) {
3499 bdrv_reopen(image_bs, open_flags, &local_err);
3500 if (local_err) {
3501 error_propagate(errp, local_err); /* will preserve prior errp */
3505 out:
3506 aio_context_release(aio_context);
3509 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3511 QmpOutputVisitor *ov = qmp_output_visitor_new();
3512 BlockDriverState *bs;
3513 BlockBackend *blk = NULL;
3514 QObject *obj;
3515 QDict *qdict;
3516 Error *local_err = NULL;
3518 /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
3519 * cache.direct=false instead of silently switching to aio=threads, except
3520 * when called from drive_new().
3522 * For now, simply forbidding the combination for all drivers will do. */
3523 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
3524 bool direct = options->has_cache &&
3525 options->cache->has_direct &&
3526 options->cache->direct;
3527 if (!direct) {
3528 error_setg(errp, "aio=native requires cache.direct=true");
3529 goto fail;
3533 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
3534 &options, NULL, &local_err);
3535 if (local_err) {
3536 error_propagate(errp, local_err);
3537 goto fail;
3540 obj = qmp_output_get_qobject(ov);
3541 qdict = qobject_to_qdict(obj);
3543 qdict_flatten(qdict);
3545 if (options->has_id) {
3546 blk = blockdev_init(NULL, qdict, &local_err);
3547 if (local_err) {
3548 error_propagate(errp, local_err);
3549 goto fail;
3552 bs = blk_bs(blk);
3553 } else {
3554 if (!qdict_get_try_str(qdict, "node-name")) {
3555 error_setg(errp, "'id' and/or 'node-name' need to be specified for "
3556 "the root node");
3557 goto fail;
3560 bs = bds_tree_init(qdict, errp);
3561 if (!bs) {
3562 goto fail;
3566 if (bs && bdrv_key_required(bs)) {
3567 if (blk) {
3568 blk_unref(blk);
3569 } else {
3570 bdrv_unref(bs);
3572 error_setg(errp, "blockdev-add doesn't support encrypted devices");
3573 goto fail;
3576 fail:
3577 qmp_output_visitor_cleanup(ov);
3580 void qmp_x_blockdev_del(bool has_id, const char *id,
3581 bool has_node_name, const char *node_name, Error **errp)
3583 AioContext *aio_context;
3584 BlockBackend *blk;
3585 BlockDriverState *bs;
3587 if (has_id && has_node_name) {
3588 error_setg(errp, "Only one of id and node-name must be specified");
3589 return;
3590 } else if (!has_id && !has_node_name) {
3591 error_setg(errp, "No block device specified");
3592 return;
3595 if (has_id) {
3596 blk = blk_by_name(id);
3597 if (!blk) {
3598 error_setg(errp, "Cannot find block backend %s", id);
3599 return;
3601 if (blk_get_refcnt(blk) > 1) {
3602 error_setg(errp, "Block backend %s is in use", id);
3603 return;
3605 bs = blk_bs(blk);
3606 aio_context = blk_get_aio_context(blk);
3607 } else {
3608 bs = bdrv_find_node(node_name);
3609 if (!bs) {
3610 error_setg(errp, "Cannot find node %s", node_name);
3611 return;
3613 blk = bs->blk;
3614 if (blk) {
3615 error_setg(errp, "Node %s is in use by %s",
3616 node_name, blk_name(blk));
3617 return;
3619 aio_context = bdrv_get_aio_context(bs);
3622 aio_context_acquire(aio_context);
3624 if (bs) {
3625 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3626 goto out;
3629 if (bs->refcnt > 1 || !QLIST_EMPTY(&bs->parents)) {
3630 error_setg(errp, "Block device %s is in use",
3631 bdrv_get_device_or_node_name(bs));
3632 goto out;
3636 if (blk) {
3637 blk_unref(blk);
3638 } else {
3639 bdrv_unref(bs);
3642 out:
3643 aio_context_release(aio_context);
3646 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3648 BlockJobInfoList *head = NULL, **p_next = &head;
3649 BlockDriverState *bs;
3651 for (bs = bdrv_next(NULL); bs; bs = bdrv_next(bs)) {
3652 AioContext *aio_context = bdrv_get_aio_context(bs);
3654 aio_context_acquire(aio_context);
3656 if (bs->job) {
3657 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
3658 elem->value = block_job_query(bs->job);
3659 *p_next = elem;
3660 p_next = &elem->next;
3663 aio_context_release(aio_context);
3666 return head;
3669 QemuOptsList qemu_common_drive_opts = {
3670 .name = "drive",
3671 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3672 .desc = {
3674 .name = "snapshot",
3675 .type = QEMU_OPT_BOOL,
3676 .help = "enable/disable snapshot mode",
3678 .name = "discard",
3679 .type = QEMU_OPT_STRING,
3680 .help = "discard operation (ignore/off, unmap/on)",
3682 .name = BDRV_OPT_CACHE_WB,
3683 .type = QEMU_OPT_BOOL,
3684 .help = "enables writeback mode for any caches",
3686 .name = BDRV_OPT_CACHE_DIRECT,
3687 .type = QEMU_OPT_BOOL,
3688 .help = "enables use of O_DIRECT (bypass the host page cache)",
3690 .name = BDRV_OPT_CACHE_NO_FLUSH,
3691 .type = QEMU_OPT_BOOL,
3692 .help = "ignore any flush requests for the device",
3694 .name = "aio",
3695 .type = QEMU_OPT_STRING,
3696 .help = "host AIO implementation (threads, native)",
3698 .name = "format",
3699 .type = QEMU_OPT_STRING,
3700 .help = "disk format (raw, qcow2, ...)",
3702 .name = "rerror",
3703 .type = QEMU_OPT_STRING,
3704 .help = "read error action",
3706 .name = "werror",
3707 .type = QEMU_OPT_STRING,
3708 .help = "write error action",
3710 .name = "read-only",
3711 .type = QEMU_OPT_BOOL,
3712 .help = "open drive file as read-only",
3714 .name = "throttling.iops-total",
3715 .type = QEMU_OPT_NUMBER,
3716 .help = "limit total I/O operations per second",
3718 .name = "throttling.iops-read",
3719 .type = QEMU_OPT_NUMBER,
3720 .help = "limit read operations per second",
3722 .name = "throttling.iops-write",
3723 .type = QEMU_OPT_NUMBER,
3724 .help = "limit write operations per second",
3726 .name = "throttling.bps-total",
3727 .type = QEMU_OPT_NUMBER,
3728 .help = "limit total bytes per second",
3730 .name = "throttling.bps-read",
3731 .type = QEMU_OPT_NUMBER,
3732 .help = "limit read bytes per second",
3734 .name = "throttling.bps-write",
3735 .type = QEMU_OPT_NUMBER,
3736 .help = "limit write bytes per second",
3738 .name = "throttling.iops-total-max",
3739 .type = QEMU_OPT_NUMBER,
3740 .help = "I/O operations burst",
3742 .name = "throttling.iops-read-max",
3743 .type = QEMU_OPT_NUMBER,
3744 .help = "I/O operations read burst",
3746 .name = "throttling.iops-write-max",
3747 .type = QEMU_OPT_NUMBER,
3748 .help = "I/O operations write burst",
3750 .name = "throttling.bps-total-max",
3751 .type = QEMU_OPT_NUMBER,
3752 .help = "total bytes burst",
3754 .name = "throttling.bps-read-max",
3755 .type = QEMU_OPT_NUMBER,
3756 .help = "total bytes read burst",
3758 .name = "throttling.bps-write-max",
3759 .type = QEMU_OPT_NUMBER,
3760 .help = "total bytes write burst",
3762 .name = "throttling.iops-size",
3763 .type = QEMU_OPT_NUMBER,
3764 .help = "when limiting by iops max size of an I/O in bytes",
3766 .name = "throttling.group",
3767 .type = QEMU_OPT_STRING,
3768 .help = "name of the block throttling group",
3770 .name = "copy-on-read",
3771 .type = QEMU_OPT_BOOL,
3772 .help = "copy read data from backing file into image file",
3774 .name = "detect-zeroes",
3775 .type = QEMU_OPT_STRING,
3776 .help = "try to optimize zero writes (off, on, unmap)",
3778 { /* end of list */ }
3782 static QemuOptsList qemu_root_bds_opts = {
3783 .name = "root-bds",
3784 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3785 .desc = {
3787 .name = "discard",
3788 .type = QEMU_OPT_STRING,
3789 .help = "discard operation (ignore/off, unmap/on)",
3791 .name = "cache.writeback",
3792 .type = QEMU_OPT_BOOL,
3793 .help = "enables writeback mode for any caches",
3795 .name = "cache.direct",
3796 .type = QEMU_OPT_BOOL,
3797 .help = "enables use of O_DIRECT (bypass the host page cache)",
3799 .name = "cache.no-flush",
3800 .type = QEMU_OPT_BOOL,
3801 .help = "ignore any flush requests for the device",
3803 .name = "aio",
3804 .type = QEMU_OPT_STRING,
3805 .help = "host AIO implementation (threads, native)",
3807 .name = "read-only",
3808 .type = QEMU_OPT_BOOL,
3809 .help = "open drive file as read-only",
3811 .name = "copy-on-read",
3812 .type = QEMU_OPT_BOOL,
3813 .help = "copy read data from backing file into image file",
3815 .name = "detect-zeroes",
3816 .type = QEMU_OPT_STRING,
3817 .help = "try to optimize zero writes (off, on, unmap)",
3819 { /* end of list */ }
3823 QemuOptsList qemu_drive_opts = {
3824 .name = "drive",
3825 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3826 .desc = {
3828 * no elements => accept any params
3829 * validation will happen later
3831 { /* end of list */ }