qemu-ga: execute fsfreeze-freeze in reverse order of mounts
[qemu.git] / blockdev.c
blob8c83f6f1cab336d9aecf48ee8b516758f92cfdb6
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/blockdev.h"
34 #include "hw/block/block.h"
35 #include "block/blockjob.h"
36 #include "monitor/monitor.h"
37 #include "qapi/qmp/qerror.h"
38 #include "qemu/option.h"
39 #include "qemu/config-file.h"
40 #include "qapi/qmp/types.h"
41 #include "sysemu/sysemu.h"
42 #include "block/block_int.h"
43 #include "qmp-commands.h"
44 #include "trace.h"
45 #include "sysemu/arch_init.h"
47 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
48 extern QemuOptsList qemu_common_drive_opts;
50 static const char *const if_name[IF_COUNT] = {
51 [IF_NONE] = "none",
52 [IF_IDE] = "ide",
53 [IF_SCSI] = "scsi",
54 [IF_FLOPPY] = "floppy",
55 [IF_PFLASH] = "pflash",
56 [IF_MTD] = "mtd",
57 [IF_SD] = "sd",
58 [IF_VIRTIO] = "virtio",
59 [IF_XEN] = "xen",
62 static const int if_max_devs[IF_COUNT] = {
64 * Do not change these numbers! They govern how drive option
65 * index maps to unit and bus. That mapping is ABI.
67 * All controllers used to imlement if=T drives need to support
68 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
69 * Otherwise, some index values map to "impossible" bus, unit
70 * values.
72 * For instance, if you change [IF_SCSI] to 255, -drive
73 * if=scsi,index=12 no longer means bus=1,unit=5, but
74 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
75 * the drive can't be set up. Regression.
77 [IF_IDE] = 2,
78 [IF_SCSI] = 7,
82 * We automatically delete the drive when a device using it gets
83 * unplugged. Questionable feature, but we can't just drop it.
84 * Device models call blockdev_mark_auto_del() to schedule the
85 * automatic deletion, and generic qdev code calls blockdev_auto_del()
86 * when deletion is actually safe.
88 void blockdev_mark_auto_del(BlockDriverState *bs)
90 DriveInfo *dinfo = drive_get_by_blockdev(bs);
92 if (bs->job) {
93 block_job_cancel(bs->job);
95 if (dinfo) {
96 dinfo->auto_del = 1;
100 void blockdev_auto_del(BlockDriverState *bs)
102 DriveInfo *dinfo = drive_get_by_blockdev(bs);
104 if (dinfo && dinfo->auto_del) {
105 drive_put_ref(dinfo);
109 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
111 int max_devs = if_max_devs[type];
112 return max_devs ? index / max_devs : 0;
115 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
117 int max_devs = if_max_devs[type];
118 return max_devs ? index % max_devs : index;
121 QemuOpts *drive_def(const char *optstr)
123 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
126 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
127 const char *optstr)
129 QemuOpts *opts;
130 char buf[32];
132 opts = drive_def(optstr);
133 if (!opts) {
134 return NULL;
136 if (type != IF_DEFAULT) {
137 qemu_opt_set(opts, "if", if_name[type]);
139 if (index >= 0) {
140 snprintf(buf, sizeof(buf), "%d", index);
141 qemu_opt_set(opts, "index", buf);
143 if (file)
144 qemu_opt_set(opts, "file", file);
145 return opts;
148 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
150 DriveInfo *dinfo;
152 /* seek interface, bus and unit */
154 QTAILQ_FOREACH(dinfo, &drives, next) {
155 if (dinfo->type == type &&
156 dinfo->bus == bus &&
157 dinfo->unit == unit)
158 return dinfo;
161 return NULL;
164 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
166 return drive_get(type,
167 drive_index_to_bus_id(type, index),
168 drive_index_to_unit_id(type, index));
171 int drive_get_max_bus(BlockInterfaceType type)
173 int max_bus;
174 DriveInfo *dinfo;
176 max_bus = -1;
177 QTAILQ_FOREACH(dinfo, &drives, next) {
178 if(dinfo->type == type &&
179 dinfo->bus > max_bus)
180 max_bus = dinfo->bus;
182 return max_bus;
185 /* Get a block device. This should only be used for single-drive devices
186 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
187 appropriate bus. */
188 DriveInfo *drive_get_next(BlockInterfaceType type)
190 static int next_block_unit[IF_COUNT];
192 return drive_get(type, 0, next_block_unit[type]++);
195 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
197 DriveInfo *dinfo;
199 QTAILQ_FOREACH(dinfo, &drives, next) {
200 if (dinfo->bdrv == bs) {
201 return dinfo;
204 return NULL;
207 static void bdrv_format_print(void *opaque, const char *name)
209 error_printf(" %s", name);
212 static void drive_uninit(DriveInfo *dinfo)
214 qemu_opts_del(dinfo->opts);
215 bdrv_unref(dinfo->bdrv);
216 g_free(dinfo->id);
217 QTAILQ_REMOVE(&drives, dinfo, next);
218 g_free(dinfo->serial);
219 g_free(dinfo);
222 void drive_put_ref(DriveInfo *dinfo)
224 assert(dinfo->refcount);
225 if (--dinfo->refcount == 0) {
226 drive_uninit(dinfo);
230 void drive_get_ref(DriveInfo *dinfo)
232 dinfo->refcount++;
235 typedef struct {
236 QEMUBH *bh;
237 BlockDriverState *bs;
238 } BDRVPutRefBH;
240 static void bdrv_put_ref_bh(void *opaque)
242 BDRVPutRefBH *s = opaque;
244 bdrv_unref(s->bs);
245 qemu_bh_delete(s->bh);
246 g_free(s);
250 * Release a BDS reference in a BH
252 * It is not safe to use bdrv_unref() from a callback function when the callers
253 * still need the BlockDriverState. In such cases we schedule a BH to release
254 * the reference.
256 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
258 BDRVPutRefBH *s;
260 s = g_new(BDRVPutRefBH, 1);
261 s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
262 s->bs = bs;
263 qemu_bh_schedule(s->bh);
266 static int parse_block_error_action(const char *buf, bool is_read)
268 if (!strcmp(buf, "ignore")) {
269 return BLOCKDEV_ON_ERROR_IGNORE;
270 } else if (!is_read && !strcmp(buf, "enospc")) {
271 return BLOCKDEV_ON_ERROR_ENOSPC;
272 } else if (!strcmp(buf, "stop")) {
273 return BLOCKDEV_ON_ERROR_STOP;
274 } else if (!strcmp(buf, "report")) {
275 return BLOCKDEV_ON_ERROR_REPORT;
276 } else {
277 error_report("'%s' invalid %s error action",
278 buf, is_read ? "read" : "write");
279 return -1;
283 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
285 if (throttle_conflicting(cfg)) {
286 error_setg(errp, "bps/iops/max total values and read/write values"
287 " cannot be used at the same time");
288 return false;
291 if (!throttle_is_valid(cfg)) {
292 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
293 return false;
296 return true;
299 static DriveInfo *blockdev_init(QemuOpts *all_opts,
300 BlockInterfaceType block_default_type)
302 const char *buf;
303 const char *file = NULL;
304 const char *serial;
305 const char *mediastr = "";
306 BlockInterfaceType type;
307 enum { MEDIA_DISK, MEDIA_CDROM } media;
308 int bus_id, unit_id;
309 int cyls, heads, secs, translation;
310 int max_devs;
311 int index;
312 int ro = 0;
313 int bdrv_flags = 0;
314 int on_read_error, on_write_error;
315 const char *devaddr;
316 DriveInfo *dinfo;
317 ThrottleConfig cfg;
318 int snapshot = 0;
319 bool copy_on_read;
320 int ret;
321 Error *error = NULL;
322 QemuOpts *opts;
323 QDict *bs_opts;
324 const char *id;
325 bool has_driver_specific_opts;
326 BlockDriver *drv = NULL;
328 translation = BIOS_ATA_TRANSLATION_AUTO;
329 media = MEDIA_DISK;
331 /* Check common options by copying from all_opts to opts, all other options
332 * are stored in bs_opts. */
333 id = qemu_opts_id(all_opts);
334 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
335 if (error_is_set(&error)) {
336 qerror_report_err(error);
337 error_free(error);
338 return NULL;
341 bs_opts = qdict_new();
342 qemu_opts_to_qdict(all_opts, bs_opts);
343 qemu_opts_absorb_qdict(opts, bs_opts, &error);
344 if (error_is_set(&error)) {
345 qerror_report_err(error);
346 error_free(error);
347 return NULL;
350 if (id) {
351 qdict_del(bs_opts, "id");
354 has_driver_specific_opts = !!qdict_size(bs_opts);
356 /* extract parameters */
357 bus_id = qemu_opt_get_number(opts, "bus", 0);
358 unit_id = qemu_opt_get_number(opts, "unit", -1);
359 index = qemu_opt_get_number(opts, "index", -1);
361 cyls = qemu_opt_get_number(opts, "cyls", 0);
362 heads = qemu_opt_get_number(opts, "heads", 0);
363 secs = qemu_opt_get_number(opts, "secs", 0);
365 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
366 ro = qemu_opt_get_bool(opts, "read-only", 0);
367 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
369 file = qemu_opt_get(opts, "file");
370 serial = qemu_opt_get(opts, "serial");
372 if ((buf = qemu_opt_get(opts, "if")) != NULL) {
373 for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
375 if (type == IF_COUNT) {
376 error_report("unsupported bus type '%s'", buf);
377 return NULL;
379 } else {
380 type = block_default_type;
383 max_devs = if_max_devs[type];
385 if (cyls || heads || secs) {
386 if (cyls < 1) {
387 error_report("invalid physical cyls number");
388 return NULL;
390 if (heads < 1) {
391 error_report("invalid physical heads number");
392 return NULL;
394 if (secs < 1) {
395 error_report("invalid physical secs number");
396 return NULL;
400 if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
401 if (!cyls) {
402 error_report("'%s' trans must be used with cyls, heads and secs",
403 buf);
404 return NULL;
406 if (!strcmp(buf, "none"))
407 translation = BIOS_ATA_TRANSLATION_NONE;
408 else if (!strcmp(buf, "lba"))
409 translation = BIOS_ATA_TRANSLATION_LBA;
410 else if (!strcmp(buf, "auto"))
411 translation = BIOS_ATA_TRANSLATION_AUTO;
412 else {
413 error_report("'%s' invalid translation type", buf);
414 return NULL;
418 if ((buf = qemu_opt_get(opts, "media")) != NULL) {
419 if (!strcmp(buf, "disk")) {
420 media = MEDIA_DISK;
421 } else if (!strcmp(buf, "cdrom")) {
422 if (cyls || secs || heads) {
423 error_report("CHS can't be set with media=%s", buf);
424 return NULL;
426 media = MEDIA_CDROM;
427 } else {
428 error_report("'%s' invalid media", buf);
429 return NULL;
433 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
434 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
435 error_report("invalid discard option");
436 return NULL;
440 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
441 bdrv_flags |= BDRV_O_CACHE_WB;
443 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
444 bdrv_flags |= BDRV_O_NOCACHE;
446 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
447 bdrv_flags |= BDRV_O_NO_FLUSH;
450 #ifdef CONFIG_LINUX_AIO
451 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
452 if (!strcmp(buf, "native")) {
453 bdrv_flags |= BDRV_O_NATIVE_AIO;
454 } else if (!strcmp(buf, "threads")) {
455 /* this is the default */
456 } else {
457 error_report("invalid aio option");
458 return NULL;
461 #endif
463 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
464 if (is_help_option(buf)) {
465 error_printf("Supported formats:");
466 bdrv_iterate_format(bdrv_format_print, NULL);
467 error_printf("\n");
468 return NULL;
471 drv = bdrv_find_whitelisted_format(buf, ro);
472 if (!drv) {
473 if (!ro && bdrv_find_whitelisted_format(buf, !ro)) {
474 error_report("'%s' can be only used as read-only device.", buf);
475 } else {
476 error_report("'%s' invalid format", buf);
478 return NULL;
482 /* disk I/O throttling */
483 memset(&cfg, 0, sizeof(cfg));
484 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
485 qemu_opt_get_number(opts, "throttling.bps-total", 0);
486 cfg.buckets[THROTTLE_BPS_READ].avg =
487 qemu_opt_get_number(opts, "throttling.bps-read", 0);
488 cfg.buckets[THROTTLE_BPS_WRITE].avg =
489 qemu_opt_get_number(opts, "throttling.bps-write", 0);
490 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
491 qemu_opt_get_number(opts, "throttling.iops-total", 0);
492 cfg.buckets[THROTTLE_OPS_READ].avg =
493 qemu_opt_get_number(opts, "throttling.iops-read", 0);
494 cfg.buckets[THROTTLE_OPS_WRITE].avg =
495 qemu_opt_get_number(opts, "throttling.iops-write", 0);
497 cfg.buckets[THROTTLE_BPS_TOTAL].max =
498 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
499 cfg.buckets[THROTTLE_BPS_READ].max =
500 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
501 cfg.buckets[THROTTLE_BPS_WRITE].max =
502 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
503 cfg.buckets[THROTTLE_OPS_TOTAL].max =
504 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
505 cfg.buckets[THROTTLE_OPS_READ].max =
506 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
507 cfg.buckets[THROTTLE_OPS_WRITE].max =
508 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
510 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
512 if (!check_throttle_config(&cfg, &error)) {
513 error_report("%s", error_get_pretty(error));
514 error_free(error);
515 return NULL;
518 if (qemu_opt_get(opts, "boot") != NULL) {
519 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
520 "ignored. Future versions will reject this parameter. Please "
521 "update your scripts.\n");
524 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
525 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
526 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
527 error_report("werror is not supported by this bus type");
528 return NULL;
531 on_write_error = parse_block_error_action(buf, 0);
532 if (on_write_error < 0) {
533 return NULL;
537 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
538 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
539 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
540 error_report("rerror is not supported by this bus type");
541 return NULL;
544 on_read_error = parse_block_error_action(buf, 1);
545 if (on_read_error < 0) {
546 return NULL;
550 if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
551 if (type != IF_VIRTIO) {
552 error_report("addr is not supported by this bus type");
553 return NULL;
557 /* compute bus and unit according index */
559 if (index != -1) {
560 if (bus_id != 0 || unit_id != -1) {
561 error_report("index cannot be used with bus and unit");
562 return NULL;
564 bus_id = drive_index_to_bus_id(type, index);
565 unit_id = drive_index_to_unit_id(type, index);
568 /* if user doesn't specify a unit_id,
569 * try to find the first free
572 if (unit_id == -1) {
573 unit_id = 0;
574 while (drive_get(type, bus_id, unit_id) != NULL) {
575 unit_id++;
576 if (max_devs && unit_id >= max_devs) {
577 unit_id -= max_devs;
578 bus_id++;
583 /* check unit id */
585 if (max_devs && unit_id >= max_devs) {
586 error_report("unit %d too big (max is %d)",
587 unit_id, max_devs - 1);
588 return NULL;
592 * catch multiple definitions
595 if (drive_get(type, bus_id, unit_id) != NULL) {
596 error_report("drive with bus=%d, unit=%d (index=%d) exists",
597 bus_id, unit_id, index);
598 return NULL;
601 /* init */
603 dinfo = g_malloc0(sizeof(*dinfo));
604 if ((buf = qemu_opts_id(opts)) != NULL) {
605 dinfo->id = g_strdup(buf);
606 } else {
607 /* no id supplied -> create one */
608 dinfo->id = g_malloc0(32);
609 if (type == IF_IDE || type == IF_SCSI)
610 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
611 if (max_devs)
612 snprintf(dinfo->id, 32, "%s%i%s%i",
613 if_name[type], bus_id, mediastr, unit_id);
614 else
615 snprintf(dinfo->id, 32, "%s%s%i",
616 if_name[type], mediastr, unit_id);
618 dinfo->bdrv = bdrv_new(dinfo->id);
619 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
620 dinfo->bdrv->read_only = ro;
621 dinfo->devaddr = devaddr;
622 dinfo->type = type;
623 dinfo->bus = bus_id;
624 dinfo->unit = unit_id;
625 dinfo->cyls = cyls;
626 dinfo->heads = heads;
627 dinfo->secs = secs;
628 dinfo->trans = translation;
629 dinfo->opts = all_opts;
630 dinfo->refcount = 1;
631 if (serial != NULL) {
632 dinfo->serial = g_strdup(serial);
634 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
636 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
638 /* disk I/O throttling */
639 if (throttle_enabled(&cfg)) {
640 bdrv_io_limits_enable(dinfo->bdrv);
641 bdrv_set_io_limits(dinfo->bdrv, &cfg);
644 switch(type) {
645 case IF_IDE:
646 case IF_SCSI:
647 case IF_XEN:
648 case IF_NONE:
649 dinfo->media_cd = media == MEDIA_CDROM;
650 break;
651 case IF_SD:
652 case IF_FLOPPY:
653 case IF_PFLASH:
654 case IF_MTD:
655 break;
656 case IF_VIRTIO:
658 /* add virtio block device */
659 QemuOpts *devopts;
660 devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
661 if (arch_type == QEMU_ARCH_S390X) {
662 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
663 } else {
664 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
666 qemu_opt_set(devopts, "drive", dinfo->id);
667 if (devaddr)
668 qemu_opt_set(devopts, "addr", devaddr);
669 break;
671 default:
672 abort();
674 if (!file || !*file) {
675 if (has_driver_specific_opts) {
676 file = NULL;
677 } else {
678 return dinfo;
681 if (snapshot) {
682 /* always use cache=unsafe with snapshot */
683 bdrv_flags &= ~BDRV_O_CACHE_MASK;
684 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
687 if (copy_on_read) {
688 bdrv_flags |= BDRV_O_COPY_ON_READ;
691 if (runstate_check(RUN_STATE_INMIGRATE)) {
692 bdrv_flags |= BDRV_O_INCOMING;
695 if (media == MEDIA_CDROM) {
696 /* CDROM is fine for any interface, don't check. */
697 ro = 1;
698 } else if (ro == 1) {
699 if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
700 type != IF_NONE && type != IF_PFLASH) {
701 error_report("read-only not supported by this bus type");
702 goto err;
706 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
708 if (ro && copy_on_read) {
709 error_report("warning: disabling copy_on_read on read-only drive");
712 QINCREF(bs_opts);
713 ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv, &error);
715 if (ret < 0) {
716 error_report("could not open disk image %s: %s",
717 file ?: dinfo->id, error_get_pretty(error));
718 goto err;
721 if (bdrv_key_required(dinfo->bdrv))
722 autostart = 0;
724 QDECREF(bs_opts);
725 qemu_opts_del(opts);
727 return dinfo;
729 err:
730 qemu_opts_del(opts);
731 QDECREF(bs_opts);
732 bdrv_unref(dinfo->bdrv);
733 g_free(dinfo->id);
734 QTAILQ_REMOVE(&drives, dinfo, next);
735 g_free(dinfo);
736 return NULL;
739 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
741 const char *value;
743 value = qemu_opt_get(opts, from);
744 if (value) {
745 qemu_opt_set(opts, to, value);
746 qemu_opt_unset(opts, from);
750 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
752 const char *value;
754 /* Change legacy command line options into QMP ones */
755 qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
756 qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
757 qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
759 qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
760 qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
761 qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
763 qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
764 qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
765 qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
767 qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
768 qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
769 qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
771 qemu_opt_rename(all_opts,
772 "iops_size", "throttling.iops-size");
774 qemu_opt_rename(all_opts, "readonly", "read-only");
776 value = qemu_opt_get(all_opts, "cache");
777 if (value) {
778 int flags = 0;
780 if (bdrv_parse_cache_flags(value, &flags) != 0) {
781 error_report("invalid cache option");
782 return NULL;
785 /* Specific options take precedence */
786 if (!qemu_opt_get(all_opts, "cache.writeback")) {
787 qemu_opt_set_bool(all_opts, "cache.writeback",
788 !!(flags & BDRV_O_CACHE_WB));
790 if (!qemu_opt_get(all_opts, "cache.direct")) {
791 qemu_opt_set_bool(all_opts, "cache.direct",
792 !!(flags & BDRV_O_NOCACHE));
794 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
795 qemu_opt_set_bool(all_opts, "cache.no-flush",
796 !!(flags & BDRV_O_NO_FLUSH));
798 qemu_opt_unset(all_opts, "cache");
801 return blockdev_init(all_opts, block_default_type);
804 void do_commit(Monitor *mon, const QDict *qdict)
806 const char *device = qdict_get_str(qdict, "device");
807 BlockDriverState *bs;
808 int ret;
810 if (!strcmp(device, "all")) {
811 ret = bdrv_commit_all();
812 } else {
813 bs = bdrv_find(device);
814 if (!bs) {
815 monitor_printf(mon, "Device '%s' not found\n", device);
816 return;
818 ret = bdrv_commit(bs);
820 if (ret < 0) {
821 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
822 strerror(-ret));
826 static void blockdev_do_action(int kind, void *data, Error **errp)
828 TransactionAction action;
829 TransactionActionList list;
831 action.kind = kind;
832 action.data = data;
833 list.value = &action;
834 list.next = NULL;
835 qmp_transaction(&list, errp);
838 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
839 bool has_format, const char *format,
840 bool has_mode, enum NewImageMode mode,
841 Error **errp)
843 BlockdevSnapshot snapshot = {
844 .device = (char *) device,
845 .snapshot_file = (char *) snapshot_file,
846 .has_format = has_format,
847 .format = (char *) format,
848 .has_mode = has_mode,
849 .mode = mode,
851 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
852 &snapshot, errp);
855 void qmp_blockdev_snapshot_internal_sync(const char *device,
856 const char *name,
857 Error **errp)
859 BlockdevSnapshotInternal snapshot = {
860 .device = (char *) device,
861 .name = (char *) name
864 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
865 &snapshot, errp);
868 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
869 bool has_id,
870 const char *id,
871 bool has_name,
872 const char *name,
873 Error **errp)
875 BlockDriverState *bs = bdrv_find(device);
876 QEMUSnapshotInfo sn;
877 Error *local_err = NULL;
878 SnapshotInfo *info = NULL;
879 int ret;
881 if (!bs) {
882 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
883 return NULL;
886 if (!has_id) {
887 id = NULL;
890 if (!has_name) {
891 name = NULL;
894 if (!id && !name) {
895 error_setg(errp, "Name or id must be provided");
896 return NULL;
899 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
900 if (error_is_set(&local_err)) {
901 error_propagate(errp, local_err);
902 return NULL;
904 if (!ret) {
905 error_setg(errp,
906 "Snapshot with id '%s' and name '%s' does not exist on "
907 "device '%s'",
908 STR_OR_NULL(id), STR_OR_NULL(name), device);
909 return NULL;
912 bdrv_snapshot_delete(bs, id, name, &local_err);
913 if (error_is_set(&local_err)) {
914 error_propagate(errp, local_err);
915 return NULL;
918 info = g_malloc0(sizeof(SnapshotInfo));
919 info->id = g_strdup(sn.id_str);
920 info->name = g_strdup(sn.name);
921 info->date_nsec = sn.date_nsec;
922 info->date_sec = sn.date_sec;
923 info->vm_state_size = sn.vm_state_size;
924 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
925 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
927 return info;
930 /* New and old BlockDriverState structs for group snapshots */
932 typedef struct BlkTransactionState BlkTransactionState;
934 /* Only prepare() may fail. In a single transaction, only one of commit() or
935 abort() will be called, clean() will always be called if it present. */
936 typedef struct BdrvActionOps {
937 /* Size of state struct, in bytes. */
938 size_t instance_size;
939 /* Prepare the work, must NOT be NULL. */
940 void (*prepare)(BlkTransactionState *common, Error **errp);
941 /* Commit the changes, can be NULL. */
942 void (*commit)(BlkTransactionState *common);
943 /* Abort the changes on fail, can be NULL. */
944 void (*abort)(BlkTransactionState *common);
945 /* Clean up resource in the end, can be NULL. */
946 void (*clean)(BlkTransactionState *common);
947 } BdrvActionOps;
950 * This structure must be arranged as first member in child type, assuming
951 * that compiler will also arrange it to the same address with parent instance.
952 * Later it will be used in free().
954 struct BlkTransactionState {
955 TransactionAction *action;
956 const BdrvActionOps *ops;
957 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
960 /* internal snapshot private data */
961 typedef struct InternalSnapshotState {
962 BlkTransactionState common;
963 BlockDriverState *bs;
964 QEMUSnapshotInfo sn;
965 } InternalSnapshotState;
967 static void internal_snapshot_prepare(BlkTransactionState *common,
968 Error **errp)
970 const char *device;
971 const char *name;
972 BlockDriverState *bs;
973 QEMUSnapshotInfo old_sn, *sn;
974 bool ret;
975 qemu_timeval tv;
976 BlockdevSnapshotInternal *internal;
977 InternalSnapshotState *state;
978 int ret1;
980 g_assert(common->action->kind ==
981 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
982 internal = common->action->blockdev_snapshot_internal_sync;
983 state = DO_UPCAST(InternalSnapshotState, common, common);
985 /* 1. parse input */
986 device = internal->device;
987 name = internal->name;
989 /* 2. check for validation */
990 bs = bdrv_find(device);
991 if (!bs) {
992 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
993 return;
996 if (!bdrv_is_inserted(bs)) {
997 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
998 return;
1001 if (bdrv_is_read_only(bs)) {
1002 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1003 return;
1006 if (!bdrv_can_snapshot(bs)) {
1007 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1008 bs->drv->format_name, device, "internal snapshot");
1009 return;
1012 if (!strlen(name)) {
1013 error_setg(errp, "Name is empty");
1014 return;
1017 /* check whether a snapshot with name exist */
1018 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn, errp);
1019 if (error_is_set(errp)) {
1020 return;
1021 } else if (ret) {
1022 error_setg(errp,
1023 "Snapshot with name '%s' already exists on device '%s'",
1024 name, device);
1025 return;
1028 /* 3. take the snapshot */
1029 sn = &state->sn;
1030 pstrcpy(sn->name, sizeof(sn->name), name);
1031 qemu_gettimeofday(&tv);
1032 sn->date_sec = tv.tv_sec;
1033 sn->date_nsec = tv.tv_usec * 1000;
1034 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1036 ret1 = bdrv_snapshot_create(bs, sn);
1037 if (ret1 < 0) {
1038 error_setg_errno(errp, -ret1,
1039 "Failed to create snapshot '%s' on device '%s'",
1040 name, device);
1041 return;
1044 /* 4. succeed, mark a snapshot is created */
1045 state->bs = bs;
1048 static void internal_snapshot_abort(BlkTransactionState *common)
1050 InternalSnapshotState *state =
1051 DO_UPCAST(InternalSnapshotState, common, common);
1052 BlockDriverState *bs = state->bs;
1053 QEMUSnapshotInfo *sn = &state->sn;
1054 Error *local_error = NULL;
1056 if (!bs) {
1057 return;
1060 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1061 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1062 "device '%s' in abort: %s",
1063 sn->id_str,
1064 sn->name,
1065 bdrv_get_device_name(bs),
1066 error_get_pretty(local_error));
1067 error_free(local_error);
1071 /* external snapshot private data */
1072 typedef struct ExternalSnapshotState {
1073 BlkTransactionState common;
1074 BlockDriverState *old_bs;
1075 BlockDriverState *new_bs;
1076 } ExternalSnapshotState;
1078 static void external_snapshot_prepare(BlkTransactionState *common,
1079 Error **errp)
1081 BlockDriver *drv;
1082 int flags, ret;
1083 Error *local_err = NULL;
1084 const char *device;
1085 const char *new_image_file;
1086 const char *format = "qcow2";
1087 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1088 ExternalSnapshotState *state =
1089 DO_UPCAST(ExternalSnapshotState, common, common);
1090 TransactionAction *action = common->action;
1092 /* get parameters */
1093 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1095 device = action->blockdev_snapshot_sync->device;
1096 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1097 if (action->blockdev_snapshot_sync->has_format) {
1098 format = action->blockdev_snapshot_sync->format;
1100 if (action->blockdev_snapshot_sync->has_mode) {
1101 mode = action->blockdev_snapshot_sync->mode;
1104 /* start processing */
1105 drv = bdrv_find_format(format);
1106 if (!drv) {
1107 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1108 return;
1111 state->old_bs = bdrv_find(device);
1112 if (!state->old_bs) {
1113 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1114 return;
1117 if (!bdrv_is_inserted(state->old_bs)) {
1118 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1119 return;
1122 if (bdrv_in_use(state->old_bs)) {
1123 error_set(errp, QERR_DEVICE_IN_USE, device);
1124 return;
1127 if (!bdrv_is_read_only(state->old_bs)) {
1128 if (bdrv_flush(state->old_bs)) {
1129 error_set(errp, QERR_IO_ERROR);
1130 return;
1134 flags = state->old_bs->open_flags;
1136 /* create new image w/backing file */
1137 if (mode != NEW_IMAGE_MODE_EXISTING) {
1138 bdrv_img_create(new_image_file, format,
1139 state->old_bs->filename,
1140 state->old_bs->drv->format_name,
1141 NULL, -1, flags, &local_err, false);
1142 if (error_is_set(&local_err)) {
1143 error_propagate(errp, local_err);
1144 return;
1148 /* We will manually add the backing_hd field to the bs later */
1149 state->new_bs = bdrv_new("");
1150 /* TODO Inherit bs->options or only take explicit options with an
1151 * extended QMP command? */
1152 ret = bdrv_open(state->new_bs, new_image_file, NULL,
1153 flags | BDRV_O_NO_BACKING, drv, &local_err);
1154 if (ret != 0) {
1155 error_propagate(errp, local_err);
1159 static void external_snapshot_commit(BlkTransactionState *common)
1161 ExternalSnapshotState *state =
1162 DO_UPCAST(ExternalSnapshotState, common, common);
1164 /* This removes our old bs and adds the new bs */
1165 bdrv_append(state->new_bs, state->old_bs);
1166 /* We don't need (or want) to use the transactional
1167 * bdrv_reopen_multiple() across all the entries at once, because we
1168 * don't want to abort all of them if one of them fails the reopen */
1169 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1170 NULL);
1173 static void external_snapshot_abort(BlkTransactionState *common)
1175 ExternalSnapshotState *state =
1176 DO_UPCAST(ExternalSnapshotState, common, common);
1177 if (state->new_bs) {
1178 bdrv_unref(state->new_bs);
1182 typedef struct DriveBackupState {
1183 BlkTransactionState common;
1184 BlockDriverState *bs;
1185 BlockJob *job;
1186 } DriveBackupState;
1188 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1190 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1191 DriveBackup *backup;
1192 Error *local_err = NULL;
1194 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1195 backup = common->action->drive_backup;
1197 qmp_drive_backup(backup->device, backup->target,
1198 backup->has_format, backup->format,
1199 backup->sync,
1200 backup->has_mode, backup->mode,
1201 backup->has_speed, backup->speed,
1202 backup->has_on_source_error, backup->on_source_error,
1203 backup->has_on_target_error, backup->on_target_error,
1204 &local_err);
1205 if (error_is_set(&local_err)) {
1206 error_propagate(errp, local_err);
1207 state->bs = NULL;
1208 state->job = NULL;
1209 return;
1212 state->bs = bdrv_find(backup->device);
1213 state->job = state->bs->job;
1216 static void drive_backup_abort(BlkTransactionState *common)
1218 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1219 BlockDriverState *bs = state->bs;
1221 /* Only cancel if it's the job we started */
1222 if (bs && bs->job && bs->job == state->job) {
1223 block_job_cancel_sync(bs->job);
1227 static void abort_prepare(BlkTransactionState *common, Error **errp)
1229 error_setg(errp, "Transaction aborted using Abort action");
1232 static void abort_commit(BlkTransactionState *common)
1234 g_assert_not_reached(); /* this action never succeeds */
1237 static const BdrvActionOps actions[] = {
1238 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1239 .instance_size = sizeof(ExternalSnapshotState),
1240 .prepare = external_snapshot_prepare,
1241 .commit = external_snapshot_commit,
1242 .abort = external_snapshot_abort,
1244 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1245 .instance_size = sizeof(DriveBackupState),
1246 .prepare = drive_backup_prepare,
1247 .abort = drive_backup_abort,
1249 [TRANSACTION_ACTION_KIND_ABORT] = {
1250 .instance_size = sizeof(BlkTransactionState),
1251 .prepare = abort_prepare,
1252 .commit = abort_commit,
1254 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1255 .instance_size = sizeof(InternalSnapshotState),
1256 .prepare = internal_snapshot_prepare,
1257 .abort = internal_snapshot_abort,
1262 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1263 * then we do not pivot any of the devices in the group, and abandon the
1264 * snapshots
1266 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1268 TransactionActionList *dev_entry = dev_list;
1269 BlkTransactionState *state, *next;
1270 Error *local_err = NULL;
1272 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1273 QSIMPLEQ_INIT(&snap_bdrv_states);
1275 /* drain all i/o before any snapshots */
1276 bdrv_drain_all();
1278 /* We don't do anything in this loop that commits us to the snapshot */
1279 while (NULL != dev_entry) {
1280 TransactionAction *dev_info = NULL;
1281 const BdrvActionOps *ops;
1283 dev_info = dev_entry->value;
1284 dev_entry = dev_entry->next;
1286 assert(dev_info->kind < ARRAY_SIZE(actions));
1288 ops = &actions[dev_info->kind];
1289 assert(ops->instance_size > 0);
1291 state = g_malloc0(ops->instance_size);
1292 state->ops = ops;
1293 state->action = dev_info;
1294 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1296 state->ops->prepare(state, &local_err);
1297 if (error_is_set(&local_err)) {
1298 error_propagate(errp, local_err);
1299 goto delete_and_fail;
1303 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1304 if (state->ops->commit) {
1305 state->ops->commit(state);
1309 /* success */
1310 goto exit;
1312 delete_and_fail:
1314 * failure, and it is all-or-none; abandon each new bs, and keep using
1315 * the original bs for all images
1317 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1318 if (state->ops->abort) {
1319 state->ops->abort(state);
1322 exit:
1323 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1324 if (state->ops->clean) {
1325 state->ops->clean(state);
1327 g_free(state);
1332 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1334 if (bdrv_in_use(bs)) {
1335 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1336 return;
1338 if (!bdrv_dev_has_removable_media(bs)) {
1339 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1340 return;
1343 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1344 bdrv_dev_eject_request(bs, force);
1345 if (!force) {
1346 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1347 return;
1351 bdrv_close(bs);
1354 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1356 BlockDriverState *bs;
1358 bs = bdrv_find(device);
1359 if (!bs) {
1360 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1361 return;
1364 eject_device(bs, force, errp);
1367 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1369 BlockDriverState *bs;
1370 int err;
1372 bs = bdrv_find(device);
1373 if (!bs) {
1374 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1375 return;
1378 err = bdrv_set_key(bs, password);
1379 if (err == -EINVAL) {
1380 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1381 return;
1382 } else if (err < 0) {
1383 error_set(errp, QERR_INVALID_PASSWORD);
1384 return;
1388 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1389 int bdrv_flags, BlockDriver *drv,
1390 const char *password, Error **errp)
1392 Error *local_err = NULL;
1393 int ret;
1395 ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv, &local_err);
1396 if (ret < 0) {
1397 error_propagate(errp, local_err);
1398 return;
1401 if (bdrv_key_required(bs)) {
1402 if (password) {
1403 if (bdrv_set_key(bs, password) < 0) {
1404 error_set(errp, QERR_INVALID_PASSWORD);
1406 } else {
1407 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1408 bdrv_get_encrypted_filename(bs));
1410 } else if (password) {
1411 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1415 void qmp_change_blockdev(const char *device, const char *filename,
1416 bool has_format, const char *format, Error **errp)
1418 BlockDriverState *bs;
1419 BlockDriver *drv = NULL;
1420 int bdrv_flags;
1421 Error *err = NULL;
1423 bs = bdrv_find(device);
1424 if (!bs) {
1425 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1426 return;
1429 if (format) {
1430 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1431 if (!drv) {
1432 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1433 return;
1437 eject_device(bs, 0, &err);
1438 if (error_is_set(&err)) {
1439 error_propagate(errp, err);
1440 return;
1443 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1444 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1446 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1449 /* throttling disk I/O limits */
1450 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1451 int64_t bps_wr,
1452 int64_t iops,
1453 int64_t iops_rd,
1454 int64_t iops_wr,
1455 bool has_bps_max,
1456 int64_t bps_max,
1457 bool has_bps_rd_max,
1458 int64_t bps_rd_max,
1459 bool has_bps_wr_max,
1460 int64_t bps_wr_max,
1461 bool has_iops_max,
1462 int64_t iops_max,
1463 bool has_iops_rd_max,
1464 int64_t iops_rd_max,
1465 bool has_iops_wr_max,
1466 int64_t iops_wr_max,
1467 bool has_iops_size,
1468 int64_t iops_size, Error **errp)
1470 ThrottleConfig cfg;
1471 BlockDriverState *bs;
1473 bs = bdrv_find(device);
1474 if (!bs) {
1475 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1476 return;
1479 memset(&cfg, 0, sizeof(cfg));
1480 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1481 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1482 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1484 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1485 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1486 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1488 if (has_bps_max) {
1489 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1491 if (has_bps_rd_max) {
1492 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1494 if (has_bps_wr_max) {
1495 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1497 if (has_iops_max) {
1498 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1500 if (has_iops_rd_max) {
1501 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1503 if (has_iops_wr_max) {
1504 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1507 if (has_iops_size) {
1508 cfg.op_size = iops_size;
1511 if (!check_throttle_config(&cfg, errp)) {
1512 return;
1515 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1516 bdrv_io_limits_enable(bs);
1517 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1518 bdrv_io_limits_disable(bs);
1521 if (bs->io_limits_enabled) {
1522 bdrv_set_io_limits(bs, &cfg);
1526 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1528 const char *id = qdict_get_str(qdict, "id");
1529 BlockDriverState *bs;
1531 bs = bdrv_find(id);
1532 if (!bs) {
1533 qerror_report(QERR_DEVICE_NOT_FOUND, id);
1534 return -1;
1536 if (bdrv_in_use(bs)) {
1537 qerror_report(QERR_DEVICE_IN_USE, id);
1538 return -1;
1541 /* quiesce block driver; prevent further io */
1542 bdrv_drain_all();
1543 bdrv_flush(bs);
1544 bdrv_close(bs);
1546 /* if we have a device attached to this BlockDriverState
1547 * then we need to make the drive anonymous until the device
1548 * can be removed. If this is a drive with no device backing
1549 * then we can just get rid of the block driver state right here.
1551 if (bdrv_get_attached_dev(bs)) {
1552 bdrv_make_anon(bs);
1554 /* Further I/O must not pause the guest */
1555 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1556 BLOCKDEV_ON_ERROR_REPORT);
1557 } else {
1558 drive_uninit(drive_get_by_blockdev(bs));
1561 return 0;
1564 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1566 BlockDriverState *bs;
1567 int ret;
1569 bs = bdrv_find(device);
1570 if (!bs) {
1571 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1572 return;
1575 if (size < 0) {
1576 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1577 return;
1580 /* complete all in-flight operations before resizing the device */
1581 bdrv_drain_all();
1583 ret = bdrv_truncate(bs, size);
1584 switch (ret) {
1585 case 0:
1586 break;
1587 case -ENOMEDIUM:
1588 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1589 break;
1590 case -ENOTSUP:
1591 error_set(errp, QERR_UNSUPPORTED);
1592 break;
1593 case -EACCES:
1594 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1595 break;
1596 case -EBUSY:
1597 error_set(errp, QERR_DEVICE_IN_USE, device);
1598 break;
1599 default:
1600 error_setg_errno(errp, -ret, "Could not resize");
1601 break;
1605 static void block_job_cb(void *opaque, int ret)
1607 BlockDriverState *bs = opaque;
1608 QObject *obj;
1610 trace_block_job_cb(bs, bs->job, ret);
1612 assert(bs->job);
1613 obj = qobject_from_block_job(bs->job);
1614 if (ret < 0) {
1615 QDict *dict = qobject_to_qdict(obj);
1616 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1619 if (block_job_is_cancelled(bs->job)) {
1620 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1621 } else {
1622 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1624 qobject_decref(obj);
1626 bdrv_put_ref_bh_schedule(bs);
1629 void qmp_block_stream(const char *device, bool has_base,
1630 const char *base, bool has_speed, int64_t speed,
1631 bool has_on_error, BlockdevOnError on_error,
1632 Error **errp)
1634 BlockDriverState *bs;
1635 BlockDriverState *base_bs = NULL;
1636 Error *local_err = NULL;
1638 if (!has_on_error) {
1639 on_error = BLOCKDEV_ON_ERROR_REPORT;
1642 bs = bdrv_find(device);
1643 if (!bs) {
1644 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1645 return;
1648 if (base) {
1649 base_bs = bdrv_find_backing_image(bs, base);
1650 if (base_bs == NULL) {
1651 error_set(errp, QERR_BASE_NOT_FOUND, base);
1652 return;
1656 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1657 on_error, block_job_cb, bs, &local_err);
1658 if (error_is_set(&local_err)) {
1659 error_propagate(errp, local_err);
1660 return;
1663 trace_qmp_block_stream(bs, bs->job);
1666 void qmp_block_commit(const char *device,
1667 bool has_base, const char *base, const char *top,
1668 bool has_speed, int64_t speed,
1669 Error **errp)
1671 BlockDriverState *bs;
1672 BlockDriverState *base_bs, *top_bs;
1673 Error *local_err = NULL;
1674 /* This will be part of the QMP command, if/when the
1675 * BlockdevOnError change for blkmirror makes it in
1677 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1679 /* drain all i/o before commits */
1680 bdrv_drain_all();
1682 bs = bdrv_find(device);
1683 if (!bs) {
1684 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1685 return;
1688 /* default top_bs is the active layer */
1689 top_bs = bs;
1691 if (top) {
1692 if (strcmp(bs->filename, top) != 0) {
1693 top_bs = bdrv_find_backing_image(bs, top);
1697 if (top_bs == NULL) {
1698 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1699 return;
1702 if (has_base && base) {
1703 base_bs = bdrv_find_backing_image(top_bs, base);
1704 } else {
1705 base_bs = bdrv_find_base(top_bs);
1708 if (base_bs == NULL) {
1709 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1710 return;
1713 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1714 &local_err);
1715 if (local_err != NULL) {
1716 error_propagate(errp, local_err);
1717 return;
1721 void qmp_drive_backup(const char *device, const char *target,
1722 bool has_format, const char *format,
1723 enum MirrorSyncMode sync,
1724 bool has_mode, enum NewImageMode mode,
1725 bool has_speed, int64_t speed,
1726 bool has_on_source_error, BlockdevOnError on_source_error,
1727 bool has_on_target_error, BlockdevOnError on_target_error,
1728 Error **errp)
1730 BlockDriverState *bs;
1731 BlockDriverState *target_bs;
1732 BlockDriverState *source = NULL;
1733 BlockDriver *drv = NULL;
1734 Error *local_err = NULL;
1735 int flags;
1736 int64_t size;
1737 int ret;
1739 if (!has_speed) {
1740 speed = 0;
1742 if (!has_on_source_error) {
1743 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1745 if (!has_on_target_error) {
1746 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1748 if (!has_mode) {
1749 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1752 bs = bdrv_find(device);
1753 if (!bs) {
1754 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1755 return;
1758 if (!bdrv_is_inserted(bs)) {
1759 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1760 return;
1763 if (!has_format) {
1764 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1766 if (format) {
1767 drv = bdrv_find_format(format);
1768 if (!drv) {
1769 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1770 return;
1774 if (bdrv_in_use(bs)) {
1775 error_set(errp, QERR_DEVICE_IN_USE, device);
1776 return;
1779 flags = bs->open_flags | BDRV_O_RDWR;
1781 /* See if we have a backing HD we can use to create our new image
1782 * on top of. */
1783 if (sync == MIRROR_SYNC_MODE_TOP) {
1784 source = bs->backing_hd;
1785 if (!source) {
1786 sync = MIRROR_SYNC_MODE_FULL;
1789 if (sync == MIRROR_SYNC_MODE_NONE) {
1790 source = bs;
1793 size = bdrv_getlength(bs);
1794 if (size < 0) {
1795 error_setg_errno(errp, -size, "bdrv_getlength failed");
1796 return;
1799 if (mode != NEW_IMAGE_MODE_EXISTING) {
1800 assert(format && drv);
1801 if (source) {
1802 bdrv_img_create(target, format, source->filename,
1803 source->drv->format_name, NULL,
1804 size, flags, &local_err, false);
1805 } else {
1806 bdrv_img_create(target, format, NULL, NULL, NULL,
1807 size, flags, &local_err, false);
1811 if (error_is_set(&local_err)) {
1812 error_propagate(errp, local_err);
1813 return;
1816 target_bs = bdrv_new("");
1817 ret = bdrv_open(target_bs, target, NULL, flags, drv, &local_err);
1818 if (ret < 0) {
1819 bdrv_unref(target_bs);
1820 error_propagate(errp, local_err);
1821 return;
1824 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1825 block_job_cb, bs, &local_err);
1826 if (local_err != NULL) {
1827 bdrv_unref(target_bs);
1828 error_propagate(errp, local_err);
1829 return;
1833 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1835 void qmp_drive_mirror(const char *device, const char *target,
1836 bool has_format, const char *format,
1837 enum MirrorSyncMode sync,
1838 bool has_mode, enum NewImageMode mode,
1839 bool has_speed, int64_t speed,
1840 bool has_granularity, uint32_t granularity,
1841 bool has_buf_size, int64_t buf_size,
1842 bool has_on_source_error, BlockdevOnError on_source_error,
1843 bool has_on_target_error, BlockdevOnError on_target_error,
1844 Error **errp)
1846 BlockDriverState *bs;
1847 BlockDriverState *source, *target_bs;
1848 BlockDriver *drv = NULL;
1849 Error *local_err = NULL;
1850 int flags;
1851 int64_t size;
1852 int ret;
1854 if (!has_speed) {
1855 speed = 0;
1857 if (!has_on_source_error) {
1858 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1860 if (!has_on_target_error) {
1861 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1863 if (!has_mode) {
1864 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1866 if (!has_granularity) {
1867 granularity = 0;
1869 if (!has_buf_size) {
1870 buf_size = DEFAULT_MIRROR_BUF_SIZE;
1873 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1874 error_set(errp, QERR_INVALID_PARAMETER, device);
1875 return;
1877 if (granularity & (granularity - 1)) {
1878 error_set(errp, QERR_INVALID_PARAMETER, device);
1879 return;
1882 bs = bdrv_find(device);
1883 if (!bs) {
1884 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1885 return;
1888 if (!bdrv_is_inserted(bs)) {
1889 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1890 return;
1893 if (!has_format) {
1894 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1896 if (format) {
1897 drv = bdrv_find_format(format);
1898 if (!drv) {
1899 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1900 return;
1904 if (bdrv_in_use(bs)) {
1905 error_set(errp, QERR_DEVICE_IN_USE, device);
1906 return;
1909 flags = bs->open_flags | BDRV_O_RDWR;
1910 source = bs->backing_hd;
1911 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1912 sync = MIRROR_SYNC_MODE_FULL;
1915 size = bdrv_getlength(bs);
1916 if (size < 0) {
1917 error_setg_errno(errp, -size, "bdrv_getlength failed");
1918 return;
1921 if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1922 /* create new image w/o backing file */
1923 assert(format && drv);
1924 bdrv_img_create(target, format,
1925 NULL, NULL, NULL, size, flags, &local_err, false);
1926 } else {
1927 switch (mode) {
1928 case NEW_IMAGE_MODE_EXISTING:
1929 break;
1930 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1931 /* create new image with backing file */
1932 bdrv_img_create(target, format,
1933 source->filename,
1934 source->drv->format_name,
1935 NULL, size, flags, &local_err, false);
1936 break;
1937 default:
1938 abort();
1942 if (error_is_set(&local_err)) {
1943 error_propagate(errp, local_err);
1944 return;
1947 /* Mirroring takes care of copy-on-write using the source's backing
1948 * file.
1950 target_bs = bdrv_new("");
1951 ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv,
1952 &local_err);
1953 if (ret < 0) {
1954 bdrv_unref(target_bs);
1955 error_propagate(errp, local_err);
1956 return;
1959 mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1960 on_source_error, on_target_error,
1961 block_job_cb, bs, &local_err);
1962 if (local_err != NULL) {
1963 bdrv_unref(target_bs);
1964 error_propagate(errp, local_err);
1965 return;
1969 static BlockJob *find_block_job(const char *device)
1971 BlockDriverState *bs;
1973 bs = bdrv_find(device);
1974 if (!bs || !bs->job) {
1975 return NULL;
1977 return bs->job;
1980 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1982 BlockJob *job = find_block_job(device);
1984 if (!job) {
1985 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1986 return;
1989 block_job_set_speed(job, speed, errp);
1992 void qmp_block_job_cancel(const char *device,
1993 bool has_force, bool force, Error **errp)
1995 BlockJob *job = find_block_job(device);
1997 if (!has_force) {
1998 force = false;
2001 if (!job) {
2002 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2003 return;
2005 if (job->paused && !force) {
2006 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
2007 return;
2010 trace_qmp_block_job_cancel(job);
2011 block_job_cancel(job);
2014 void qmp_block_job_pause(const char *device, Error **errp)
2016 BlockJob *job = find_block_job(device);
2018 if (!job) {
2019 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2020 return;
2023 trace_qmp_block_job_pause(job);
2024 block_job_pause(job);
2027 void qmp_block_job_resume(const char *device, Error **errp)
2029 BlockJob *job = find_block_job(device);
2031 if (!job) {
2032 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2033 return;
2036 trace_qmp_block_job_resume(job);
2037 block_job_resume(job);
2040 void qmp_block_job_complete(const char *device, Error **errp)
2042 BlockJob *job = find_block_job(device);
2044 if (!job) {
2045 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2046 return;
2049 trace_qmp_block_job_complete(job);
2050 block_job_complete(job, errp);
2053 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2055 BlockJobInfoList **prev = opaque;
2056 BlockJob *job = bs->job;
2058 if (job) {
2059 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2060 elem->value = block_job_query(bs->job);
2061 (*prev)->next = elem;
2062 *prev = elem;
2066 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2068 /* Dummy is a fake list element for holding the head pointer */
2069 BlockJobInfoList dummy = {};
2070 BlockJobInfoList *prev = &dummy;
2071 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2072 return dummy.next;
2075 QemuOptsList qemu_common_drive_opts = {
2076 .name = "drive",
2077 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2078 .desc = {
2080 .name = "bus",
2081 .type = QEMU_OPT_NUMBER,
2082 .help = "bus number",
2084 .name = "unit",
2085 .type = QEMU_OPT_NUMBER,
2086 .help = "unit number (i.e. lun for scsi)",
2088 .name = "if",
2089 .type = QEMU_OPT_STRING,
2090 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
2092 .name = "index",
2093 .type = QEMU_OPT_NUMBER,
2094 .help = "index number",
2096 .name = "cyls",
2097 .type = QEMU_OPT_NUMBER,
2098 .help = "number of cylinders (ide disk geometry)",
2100 .name = "heads",
2101 .type = QEMU_OPT_NUMBER,
2102 .help = "number of heads (ide disk geometry)",
2104 .name = "secs",
2105 .type = QEMU_OPT_NUMBER,
2106 .help = "number of sectors (ide disk geometry)",
2108 .name = "trans",
2109 .type = QEMU_OPT_STRING,
2110 .help = "chs translation (auto, lba. none)",
2112 .name = "media",
2113 .type = QEMU_OPT_STRING,
2114 .help = "media type (disk, cdrom)",
2116 .name = "snapshot",
2117 .type = QEMU_OPT_BOOL,
2118 .help = "enable/disable snapshot mode",
2120 .name = "file",
2121 .type = QEMU_OPT_STRING,
2122 .help = "disk image",
2124 .name = "discard",
2125 .type = QEMU_OPT_STRING,
2126 .help = "discard operation (ignore/off, unmap/on)",
2128 .name = "cache.writeback",
2129 .type = QEMU_OPT_BOOL,
2130 .help = "enables writeback mode for any caches",
2132 .name = "cache.direct",
2133 .type = QEMU_OPT_BOOL,
2134 .help = "enables use of O_DIRECT (bypass the host page cache)",
2136 .name = "cache.no-flush",
2137 .type = QEMU_OPT_BOOL,
2138 .help = "ignore any flush requests for the device",
2140 .name = "aio",
2141 .type = QEMU_OPT_STRING,
2142 .help = "host AIO implementation (threads, native)",
2144 .name = "format",
2145 .type = QEMU_OPT_STRING,
2146 .help = "disk format (raw, qcow2, ...)",
2148 .name = "serial",
2149 .type = QEMU_OPT_STRING,
2150 .help = "disk serial number",
2152 .name = "rerror",
2153 .type = QEMU_OPT_STRING,
2154 .help = "read error action",
2156 .name = "werror",
2157 .type = QEMU_OPT_STRING,
2158 .help = "write error action",
2160 .name = "addr",
2161 .type = QEMU_OPT_STRING,
2162 .help = "pci address (virtio only)",
2164 .name = "read-only",
2165 .type = QEMU_OPT_BOOL,
2166 .help = "open drive file as read-only",
2168 .name = "throttling.iops-total",
2169 .type = QEMU_OPT_NUMBER,
2170 .help = "limit total I/O operations per second",
2172 .name = "throttling.iops-read",
2173 .type = QEMU_OPT_NUMBER,
2174 .help = "limit read operations per second",
2176 .name = "throttling.iops-write",
2177 .type = QEMU_OPT_NUMBER,
2178 .help = "limit write operations per second",
2180 .name = "throttling.bps-total",
2181 .type = QEMU_OPT_NUMBER,
2182 .help = "limit total bytes per second",
2184 .name = "throttling.bps-read",
2185 .type = QEMU_OPT_NUMBER,
2186 .help = "limit read bytes per second",
2188 .name = "throttling.bps-write",
2189 .type = QEMU_OPT_NUMBER,
2190 .help = "limit write bytes per second",
2192 .name = "throttling.iops-total-max",
2193 .type = QEMU_OPT_NUMBER,
2194 .help = "I/O operations burst",
2196 .name = "throttling.iops-read-max",
2197 .type = QEMU_OPT_NUMBER,
2198 .help = "I/O operations read burst",
2200 .name = "throttling.iops-write-max",
2201 .type = QEMU_OPT_NUMBER,
2202 .help = "I/O operations write burst",
2204 .name = "throttling.bps-total-max",
2205 .type = QEMU_OPT_NUMBER,
2206 .help = "total bytes burst",
2208 .name = "throttling.bps-read-max",
2209 .type = QEMU_OPT_NUMBER,
2210 .help = "total bytes read burst",
2212 .name = "throttling.bps-write-max",
2213 .type = QEMU_OPT_NUMBER,
2214 .help = "total bytes write burst",
2216 .name = "throttling.iops-size",
2217 .type = QEMU_OPT_NUMBER,
2218 .help = "when limiting by iops max size of an I/O in bytes",
2220 .name = "copy-on-read",
2221 .type = QEMU_OPT_BOOL,
2222 .help = "copy read data from backing file into image file",
2224 .name = "boot",
2225 .type = QEMU_OPT_BOOL,
2226 .help = "(deprecated, ignored)",
2228 { /* end of list */ }
2232 QemuOptsList qemu_drive_opts = {
2233 .name = "drive",
2234 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2235 .desc = {
2237 * no elements => accept any params
2238 * validation will happen later
2240 { /* end of list */ }