balloon: re-enable balloon stats
[qemu.git] / blockdev.c
blob030070b607d107385498e1c427dc959257a64d1f
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.
8 */
10 #include "sysemu/blockdev.h"
11 #include "hw/block-common.h"
12 #include "block/blockjob.h"
13 #include "monitor/monitor.h"
14 #include "qapi/qmp/qerror.h"
15 #include "qemu/option.h"
16 #include "qemu/config-file.h"
17 #include "qapi/qmp/types.h"
18 #include "sysemu/sysemu.h"
19 #include "block/block_int.h"
20 #include "qmp-commands.h"
21 #include "trace.h"
22 #include "sysemu/arch_init.h"
24 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
26 static const char *const if_name[IF_COUNT] = {
27 [IF_NONE] = "none",
28 [IF_IDE] = "ide",
29 [IF_SCSI] = "scsi",
30 [IF_FLOPPY] = "floppy",
31 [IF_PFLASH] = "pflash",
32 [IF_MTD] = "mtd",
33 [IF_SD] = "sd",
34 [IF_VIRTIO] = "virtio",
35 [IF_XEN] = "xen",
38 static const int if_max_devs[IF_COUNT] = {
40 * Do not change these numbers! They govern how drive option
41 * index maps to unit and bus. That mapping is ABI.
43 * All controllers used to imlement if=T drives need to support
44 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
45 * Otherwise, some index values map to "impossible" bus, unit
46 * values.
48 * For instance, if you change [IF_SCSI] to 255, -drive
49 * if=scsi,index=12 no longer means bus=1,unit=5, but
50 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
51 * the drive can't be set up. Regression.
53 [IF_IDE] = 2,
54 [IF_SCSI] = 7,
58 * We automatically delete the drive when a device using it gets
59 * unplugged. Questionable feature, but we can't just drop it.
60 * Device models call blockdev_mark_auto_del() to schedule the
61 * automatic deletion, and generic qdev code calls blockdev_auto_del()
62 * when deletion is actually safe.
64 void blockdev_mark_auto_del(BlockDriverState *bs)
66 DriveInfo *dinfo = drive_get_by_blockdev(bs);
68 if (bs->job) {
69 block_job_cancel(bs->job);
71 if (dinfo) {
72 dinfo->auto_del = 1;
76 void blockdev_auto_del(BlockDriverState *bs)
78 DriveInfo *dinfo = drive_get_by_blockdev(bs);
80 if (dinfo && dinfo->auto_del) {
81 drive_put_ref(dinfo);
85 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
87 int max_devs = if_max_devs[type];
88 return max_devs ? index / max_devs : 0;
91 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
93 int max_devs = if_max_devs[type];
94 return max_devs ? index % max_devs : index;
97 QemuOpts *drive_def(const char *optstr)
99 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
102 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
103 const char *optstr)
105 QemuOpts *opts;
106 char buf[32];
108 opts = drive_def(optstr);
109 if (!opts) {
110 return NULL;
112 if (type != IF_DEFAULT) {
113 qemu_opt_set(opts, "if", if_name[type]);
115 if (index >= 0) {
116 snprintf(buf, sizeof(buf), "%d", index);
117 qemu_opt_set(opts, "index", buf);
119 if (file)
120 qemu_opt_set(opts, "file", file);
121 return opts;
124 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
126 DriveInfo *dinfo;
128 /* seek interface, bus and unit */
130 QTAILQ_FOREACH(dinfo, &drives, next) {
131 if (dinfo->type == type &&
132 dinfo->bus == bus &&
133 dinfo->unit == unit)
134 return dinfo;
137 return NULL;
140 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
142 return drive_get(type,
143 drive_index_to_bus_id(type, index),
144 drive_index_to_unit_id(type, index));
147 int drive_get_max_bus(BlockInterfaceType type)
149 int max_bus;
150 DriveInfo *dinfo;
152 max_bus = -1;
153 QTAILQ_FOREACH(dinfo, &drives, next) {
154 if(dinfo->type == type &&
155 dinfo->bus > max_bus)
156 max_bus = dinfo->bus;
158 return max_bus;
161 /* Get a block device. This should only be used for single-drive devices
162 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
163 appropriate bus. */
164 DriveInfo *drive_get_next(BlockInterfaceType type)
166 static int next_block_unit[IF_COUNT];
168 return drive_get(type, 0, next_block_unit[type]++);
171 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
173 DriveInfo *dinfo;
175 QTAILQ_FOREACH(dinfo, &drives, next) {
176 if (dinfo->bdrv == bs) {
177 return dinfo;
180 return NULL;
183 static void bdrv_format_print(void *opaque, const char *name)
185 error_printf(" %s", name);
188 static void drive_uninit(DriveInfo *dinfo)
190 qemu_opts_del(dinfo->opts);
191 bdrv_delete(dinfo->bdrv);
192 g_free(dinfo->id);
193 QTAILQ_REMOVE(&drives, dinfo, next);
194 g_free(dinfo);
197 void drive_put_ref(DriveInfo *dinfo)
199 assert(dinfo->refcount);
200 if (--dinfo->refcount == 0) {
201 drive_uninit(dinfo);
205 void drive_get_ref(DriveInfo *dinfo)
207 dinfo->refcount++;
210 typedef struct {
211 QEMUBH *bh;
212 DriveInfo *dinfo;
213 } DrivePutRefBH;
215 static void drive_put_ref_bh(void *opaque)
217 DrivePutRefBH *s = opaque;
219 drive_put_ref(s->dinfo);
220 qemu_bh_delete(s->bh);
221 g_free(s);
225 * Release a drive reference in a BH
227 * It is not possible to use drive_put_ref() from a callback function when the
228 * callers still need the drive. In such cases we schedule a BH to release the
229 * reference.
231 static void drive_put_ref_bh_schedule(DriveInfo *dinfo)
233 DrivePutRefBH *s;
235 s = g_new(DrivePutRefBH, 1);
236 s->bh = qemu_bh_new(drive_put_ref_bh, s);
237 s->dinfo = dinfo;
238 qemu_bh_schedule(s->bh);
241 static int parse_block_error_action(const char *buf, bool is_read)
243 if (!strcmp(buf, "ignore")) {
244 return BLOCKDEV_ON_ERROR_IGNORE;
245 } else if (!is_read && !strcmp(buf, "enospc")) {
246 return BLOCKDEV_ON_ERROR_ENOSPC;
247 } else if (!strcmp(buf, "stop")) {
248 return BLOCKDEV_ON_ERROR_STOP;
249 } else if (!strcmp(buf, "report")) {
250 return BLOCKDEV_ON_ERROR_REPORT;
251 } else {
252 error_report("'%s' invalid %s error action",
253 buf, is_read ? "read" : "write");
254 return -1;
258 static bool do_check_io_limits(BlockIOLimit *io_limits)
260 bool bps_flag;
261 bool iops_flag;
263 assert(io_limits);
265 bps_flag = (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] != 0)
266 && ((io_limits->bps[BLOCK_IO_LIMIT_READ] != 0)
267 || (io_limits->bps[BLOCK_IO_LIMIT_WRITE] != 0));
268 iops_flag = (io_limits->iops[BLOCK_IO_LIMIT_TOTAL] != 0)
269 && ((io_limits->iops[BLOCK_IO_LIMIT_READ] != 0)
270 || (io_limits->iops[BLOCK_IO_LIMIT_WRITE] != 0));
271 if (bps_flag || iops_flag) {
272 return false;
275 return true;
278 DriveInfo *drive_init(QemuOpts *opts, BlockInterfaceType block_default_type)
280 const char *buf;
281 const char *file = NULL;
282 const char *serial;
283 const char *mediastr = "";
284 BlockInterfaceType type;
285 enum { MEDIA_DISK, MEDIA_CDROM } media;
286 int bus_id, unit_id;
287 int cyls, heads, secs, translation;
288 BlockDriver *drv = NULL;
289 int max_devs;
290 int index;
291 int ro = 0;
292 int bdrv_flags = 0;
293 int on_read_error, on_write_error;
294 const char *devaddr;
295 DriveInfo *dinfo;
296 BlockIOLimit io_limits;
297 int snapshot = 0;
298 bool copy_on_read;
299 int ret;
301 translation = BIOS_ATA_TRANSLATION_AUTO;
302 media = MEDIA_DISK;
304 /* extract parameters */
305 bus_id = qemu_opt_get_number(opts, "bus", 0);
306 unit_id = qemu_opt_get_number(opts, "unit", -1);
307 index = qemu_opt_get_number(opts, "index", -1);
309 cyls = qemu_opt_get_number(opts, "cyls", 0);
310 heads = qemu_opt_get_number(opts, "heads", 0);
311 secs = qemu_opt_get_number(opts, "secs", 0);
313 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
314 ro = qemu_opt_get_bool(opts, "readonly", 0);
315 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
317 file = qemu_opt_get(opts, "file");
318 serial = qemu_opt_get(opts, "serial");
320 if ((buf = qemu_opt_get(opts, "if")) != NULL) {
321 for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
323 if (type == IF_COUNT) {
324 error_report("unsupported bus type '%s'", buf);
325 return NULL;
327 } else {
328 type = block_default_type;
331 max_devs = if_max_devs[type];
333 if (cyls || heads || secs) {
334 if (cyls < 1) {
335 error_report("invalid physical cyls number");
336 return NULL;
338 if (heads < 1) {
339 error_report("invalid physical heads number");
340 return NULL;
342 if (secs < 1) {
343 error_report("invalid physical secs number");
344 return NULL;
348 if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
349 if (!cyls) {
350 error_report("'%s' trans must be used with cyls, heads and secs",
351 buf);
352 return NULL;
354 if (!strcmp(buf, "none"))
355 translation = BIOS_ATA_TRANSLATION_NONE;
356 else if (!strcmp(buf, "lba"))
357 translation = BIOS_ATA_TRANSLATION_LBA;
358 else if (!strcmp(buf, "auto"))
359 translation = BIOS_ATA_TRANSLATION_AUTO;
360 else {
361 error_report("'%s' invalid translation type", buf);
362 return NULL;
366 if ((buf = qemu_opt_get(opts, "media")) != NULL) {
367 if (!strcmp(buf, "disk")) {
368 media = MEDIA_DISK;
369 } else if (!strcmp(buf, "cdrom")) {
370 if (cyls || secs || heads) {
371 error_report("CHS can't be set with media=%s", buf);
372 return NULL;
374 media = MEDIA_CDROM;
375 } else {
376 error_report("'%s' invalid media", buf);
377 return NULL;
381 bdrv_flags |= BDRV_O_CACHE_WB;
382 if ((buf = qemu_opt_get(opts, "cache")) != NULL) {
383 if (bdrv_parse_cache_flags(buf, &bdrv_flags) != 0) {
384 error_report("invalid cache option");
385 return NULL;
389 #ifdef CONFIG_LINUX_AIO
390 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
391 if (!strcmp(buf, "native")) {
392 bdrv_flags |= BDRV_O_NATIVE_AIO;
393 } else if (!strcmp(buf, "threads")) {
394 /* this is the default */
395 } else {
396 error_report("invalid aio option");
397 return NULL;
400 #endif
402 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
403 if (is_help_option(buf)) {
404 error_printf("Supported formats:");
405 bdrv_iterate_format(bdrv_format_print, NULL);
406 error_printf("\n");
407 return NULL;
409 drv = bdrv_find_whitelisted_format(buf);
410 if (!drv) {
411 error_report("'%s' invalid format", buf);
412 return NULL;
416 /* disk I/O throttling */
417 io_limits.bps[BLOCK_IO_LIMIT_TOTAL] =
418 qemu_opt_get_number(opts, "bps", 0);
419 io_limits.bps[BLOCK_IO_LIMIT_READ] =
420 qemu_opt_get_number(opts, "bps_rd", 0);
421 io_limits.bps[BLOCK_IO_LIMIT_WRITE] =
422 qemu_opt_get_number(opts, "bps_wr", 0);
423 io_limits.iops[BLOCK_IO_LIMIT_TOTAL] =
424 qemu_opt_get_number(opts, "iops", 0);
425 io_limits.iops[BLOCK_IO_LIMIT_READ] =
426 qemu_opt_get_number(opts, "iops_rd", 0);
427 io_limits.iops[BLOCK_IO_LIMIT_WRITE] =
428 qemu_opt_get_number(opts, "iops_wr", 0);
430 if (!do_check_io_limits(&io_limits)) {
431 error_report("bps(iops) and bps_rd/bps_wr(iops_rd/iops_wr) "
432 "cannot be used at the same time");
433 return NULL;
436 if (qemu_opt_get(opts, "boot") != NULL) {
437 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
438 "ignored. Future versions will reject this parameter. Please "
439 "update your scripts.\n");
442 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
443 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
444 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
445 error_report("werror is not supported by this bus type");
446 return NULL;
449 on_write_error = parse_block_error_action(buf, 0);
450 if (on_write_error < 0) {
451 return NULL;
455 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
456 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
457 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
458 error_report("rerror is not supported by this bus type");
459 return NULL;
462 on_read_error = parse_block_error_action(buf, 1);
463 if (on_read_error < 0) {
464 return NULL;
468 if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
469 if (type != IF_VIRTIO) {
470 error_report("addr is not supported by this bus type");
471 return NULL;
475 /* compute bus and unit according index */
477 if (index != -1) {
478 if (bus_id != 0 || unit_id != -1) {
479 error_report("index cannot be used with bus and unit");
480 return NULL;
482 bus_id = drive_index_to_bus_id(type, index);
483 unit_id = drive_index_to_unit_id(type, index);
486 /* if user doesn't specify a unit_id,
487 * try to find the first free
490 if (unit_id == -1) {
491 unit_id = 0;
492 while (drive_get(type, bus_id, unit_id) != NULL) {
493 unit_id++;
494 if (max_devs && unit_id >= max_devs) {
495 unit_id -= max_devs;
496 bus_id++;
501 /* check unit id */
503 if (max_devs && unit_id >= max_devs) {
504 error_report("unit %d too big (max is %d)",
505 unit_id, max_devs - 1);
506 return NULL;
510 * catch multiple definitions
513 if (drive_get(type, bus_id, unit_id) != NULL) {
514 error_report("drive with bus=%d, unit=%d (index=%d) exists",
515 bus_id, unit_id, index);
516 return NULL;
519 /* init */
521 dinfo = g_malloc0(sizeof(*dinfo));
522 if ((buf = qemu_opts_id(opts)) != NULL) {
523 dinfo->id = g_strdup(buf);
524 } else {
525 /* no id supplied -> create one */
526 dinfo->id = g_malloc0(32);
527 if (type == IF_IDE || type == IF_SCSI)
528 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
529 if (max_devs)
530 snprintf(dinfo->id, 32, "%s%i%s%i",
531 if_name[type], bus_id, mediastr, unit_id);
532 else
533 snprintf(dinfo->id, 32, "%s%s%i",
534 if_name[type], mediastr, unit_id);
536 dinfo->bdrv = bdrv_new(dinfo->id);
537 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
538 dinfo->bdrv->read_only = ro;
539 dinfo->devaddr = devaddr;
540 dinfo->type = type;
541 dinfo->bus = bus_id;
542 dinfo->unit = unit_id;
543 dinfo->cyls = cyls;
544 dinfo->heads = heads;
545 dinfo->secs = secs;
546 dinfo->trans = translation;
547 dinfo->opts = opts;
548 dinfo->refcount = 1;
549 dinfo->serial = serial;
550 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
552 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
554 /* disk I/O throttling */
555 bdrv_set_io_limits(dinfo->bdrv, &io_limits);
557 switch(type) {
558 case IF_IDE:
559 case IF_SCSI:
560 case IF_XEN:
561 case IF_NONE:
562 dinfo->media_cd = media == MEDIA_CDROM;
563 break;
564 case IF_SD:
565 case IF_FLOPPY:
566 case IF_PFLASH:
567 case IF_MTD:
568 break;
569 case IF_VIRTIO:
570 /* add virtio block device */
571 opts = qemu_opts_create_nofail(qemu_find_opts("device"));
572 if (arch_type == QEMU_ARCH_S390X) {
573 qemu_opt_set(opts, "driver", "virtio-blk-s390");
574 } else {
575 qemu_opt_set(opts, "driver", "virtio-blk-pci");
577 qemu_opt_set(opts, "drive", dinfo->id);
578 if (devaddr)
579 qemu_opt_set(opts, "addr", devaddr);
580 break;
581 default:
582 abort();
584 if (!file || !*file) {
585 return dinfo;
587 if (snapshot) {
588 /* always use cache=unsafe with snapshot */
589 bdrv_flags &= ~BDRV_O_CACHE_MASK;
590 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
593 if (copy_on_read) {
594 bdrv_flags |= BDRV_O_COPY_ON_READ;
597 if (runstate_check(RUN_STATE_INMIGRATE)) {
598 bdrv_flags |= BDRV_O_INCOMING;
601 if (media == MEDIA_CDROM) {
602 /* CDROM is fine for any interface, don't check. */
603 ro = 1;
604 } else if (ro == 1) {
605 if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
606 type != IF_NONE && type != IF_PFLASH) {
607 error_report("readonly not supported by this bus type");
608 goto err;
612 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
614 if (ro && copy_on_read) {
615 error_report("warning: disabling copy_on_read on readonly drive");
618 ret = bdrv_open(dinfo->bdrv, file, bdrv_flags, drv);
619 if (ret < 0) {
620 error_report("could not open disk image %s: %s",
621 file, strerror(-ret));
622 goto err;
625 if (bdrv_key_required(dinfo->bdrv))
626 autostart = 0;
627 return dinfo;
629 err:
630 bdrv_delete(dinfo->bdrv);
631 g_free(dinfo->id);
632 QTAILQ_REMOVE(&drives, dinfo, next);
633 g_free(dinfo);
634 return NULL;
637 void do_commit(Monitor *mon, const QDict *qdict)
639 const char *device = qdict_get_str(qdict, "device");
640 BlockDriverState *bs;
641 int ret;
643 if (!strcmp(device, "all")) {
644 ret = bdrv_commit_all();
645 } else {
646 bs = bdrv_find(device);
647 if (!bs) {
648 monitor_printf(mon, "Device '%s' not found\n", device);
649 return;
651 ret = bdrv_commit(bs);
653 if (ret < 0) {
654 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
655 strerror(-ret));
659 static void blockdev_do_action(int kind, void *data, Error **errp)
661 BlockdevAction action;
662 BlockdevActionList list;
664 action.kind = kind;
665 action.data = data;
666 list.value = &action;
667 list.next = NULL;
668 qmp_transaction(&list, errp);
671 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
672 bool has_format, const char *format,
673 bool has_mode, enum NewImageMode mode,
674 Error **errp)
676 BlockdevSnapshot snapshot = {
677 .device = (char *) device,
678 .snapshot_file = (char *) snapshot_file,
679 .has_format = has_format,
680 .format = (char *) format,
681 .has_mode = has_mode,
682 .mode = mode,
684 blockdev_do_action(BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC, &snapshot,
685 errp);
689 /* New and old BlockDriverState structs for group snapshots */
690 typedef struct BlkTransactionStates {
691 BlockDriverState *old_bs;
692 BlockDriverState *new_bs;
693 QSIMPLEQ_ENTRY(BlkTransactionStates) entry;
694 } BlkTransactionStates;
697 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
698 * then we do not pivot any of the devices in the group, and abandon the
699 * snapshots
701 void qmp_transaction(BlockdevActionList *dev_list, Error **errp)
703 int ret = 0;
704 BlockdevActionList *dev_entry = dev_list;
705 BlkTransactionStates *states, *next;
706 Error *local_err = NULL;
708 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionStates) snap_bdrv_states;
709 QSIMPLEQ_INIT(&snap_bdrv_states);
711 /* drain all i/o before any snapshots */
712 bdrv_drain_all();
714 /* We don't do anything in this loop that commits us to the snapshot */
715 while (NULL != dev_entry) {
716 BlockdevAction *dev_info = NULL;
717 BlockDriver *proto_drv;
718 BlockDriver *drv;
719 int flags;
720 enum NewImageMode mode;
721 const char *new_image_file;
722 const char *device;
723 const char *format = "qcow2";
725 dev_info = dev_entry->value;
726 dev_entry = dev_entry->next;
728 states = g_malloc0(sizeof(BlkTransactionStates));
729 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, states, entry);
731 switch (dev_info->kind) {
732 case BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
733 device = dev_info->blockdev_snapshot_sync->device;
734 if (!dev_info->blockdev_snapshot_sync->has_mode) {
735 dev_info->blockdev_snapshot_sync->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
737 new_image_file = dev_info->blockdev_snapshot_sync->snapshot_file;
738 if (dev_info->blockdev_snapshot_sync->has_format) {
739 format = dev_info->blockdev_snapshot_sync->format;
741 mode = dev_info->blockdev_snapshot_sync->mode;
742 break;
743 default:
744 abort();
747 drv = bdrv_find_format(format);
748 if (!drv) {
749 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
750 goto delete_and_fail;
753 states->old_bs = bdrv_find(device);
754 if (!states->old_bs) {
755 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
756 goto delete_and_fail;
759 if (!bdrv_is_inserted(states->old_bs)) {
760 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
761 goto delete_and_fail;
764 if (bdrv_in_use(states->old_bs)) {
765 error_set(errp, QERR_DEVICE_IN_USE, device);
766 goto delete_and_fail;
769 if (!bdrv_is_read_only(states->old_bs)) {
770 if (bdrv_flush(states->old_bs)) {
771 error_set(errp, QERR_IO_ERROR);
772 goto delete_and_fail;
776 flags = states->old_bs->open_flags;
778 proto_drv = bdrv_find_protocol(new_image_file);
779 if (!proto_drv) {
780 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
781 goto delete_and_fail;
784 /* create new image w/backing file */
785 if (mode != NEW_IMAGE_MODE_EXISTING) {
786 bdrv_img_create(new_image_file, format,
787 states->old_bs->filename,
788 states->old_bs->drv->format_name,
789 NULL, -1, flags, &local_err);
790 if (error_is_set(&local_err)) {
791 error_propagate(errp, local_err);
792 goto delete_and_fail;
796 /* We will manually add the backing_hd field to the bs later */
797 states->new_bs = bdrv_new("");
798 ret = bdrv_open(states->new_bs, new_image_file,
799 flags | BDRV_O_NO_BACKING, drv);
800 if (ret != 0) {
801 error_set(errp, QERR_OPEN_FILE_FAILED, new_image_file);
802 goto delete_and_fail;
807 /* Now we are going to do the actual pivot. Everything up to this point
808 * is reversible, but we are committed at this point */
809 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
810 /* This removes our old bs from the bdrv_states, and adds the new bs */
811 bdrv_append(states->new_bs, states->old_bs);
812 /* We don't need (or want) to use the transactional
813 * bdrv_reopen_multiple() across all the entries at once, because we
814 * don't want to abort all of them if one of them fails the reopen */
815 bdrv_reopen(states->new_bs, states->new_bs->open_flags & ~BDRV_O_RDWR,
816 NULL);
819 /* success */
820 goto exit;
822 delete_and_fail:
824 * failure, and it is all-or-none; abandon each new bs, and keep using
825 * the original bs for all images
827 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
828 if (states->new_bs) {
829 bdrv_delete(states->new_bs);
832 exit:
833 QSIMPLEQ_FOREACH_SAFE(states, &snap_bdrv_states, entry, next) {
834 g_free(states);
839 static void eject_device(BlockDriverState *bs, int force, Error **errp)
841 if (bdrv_in_use(bs)) {
842 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
843 return;
845 if (!bdrv_dev_has_removable_media(bs)) {
846 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
847 return;
850 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
851 bdrv_dev_eject_request(bs, force);
852 if (!force) {
853 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
854 return;
858 bdrv_close(bs);
861 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
863 BlockDriverState *bs;
865 bs = bdrv_find(device);
866 if (!bs) {
867 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
868 return;
871 eject_device(bs, force, errp);
874 void qmp_block_passwd(const char *device, const char *password, Error **errp)
876 BlockDriverState *bs;
877 int err;
879 bs = bdrv_find(device);
880 if (!bs) {
881 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
882 return;
885 err = bdrv_set_key(bs, password);
886 if (err == -EINVAL) {
887 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
888 return;
889 } else if (err < 0) {
890 error_set(errp, QERR_INVALID_PASSWORD);
891 return;
895 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
896 int bdrv_flags, BlockDriver *drv,
897 const char *password, Error **errp)
899 if (bdrv_open(bs, filename, bdrv_flags, drv) < 0) {
900 error_set(errp, QERR_OPEN_FILE_FAILED, filename);
901 return;
904 if (bdrv_key_required(bs)) {
905 if (password) {
906 if (bdrv_set_key(bs, password) < 0) {
907 error_set(errp, QERR_INVALID_PASSWORD);
909 } else {
910 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
911 bdrv_get_encrypted_filename(bs));
913 } else if (password) {
914 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
918 void qmp_change_blockdev(const char *device, const char *filename,
919 bool has_format, const char *format, Error **errp)
921 BlockDriverState *bs;
922 BlockDriver *drv = NULL;
923 int bdrv_flags;
924 Error *err = NULL;
926 bs = bdrv_find(device);
927 if (!bs) {
928 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
929 return;
932 if (format) {
933 drv = bdrv_find_whitelisted_format(format);
934 if (!drv) {
935 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
936 return;
940 eject_device(bs, 0, &err);
941 if (error_is_set(&err)) {
942 error_propagate(errp, err);
943 return;
946 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
947 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
949 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
952 /* throttling disk I/O limits */
953 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
954 int64_t bps_wr, int64_t iops, int64_t iops_rd,
955 int64_t iops_wr, Error **errp)
957 BlockIOLimit io_limits;
958 BlockDriverState *bs;
960 bs = bdrv_find(device);
961 if (!bs) {
962 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
963 return;
966 io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
967 io_limits.bps[BLOCK_IO_LIMIT_READ] = bps_rd;
968 io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
969 io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
970 io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
971 io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
973 if (!do_check_io_limits(&io_limits)) {
974 error_set(errp, QERR_INVALID_PARAMETER_COMBINATION);
975 return;
978 bs->io_limits = io_limits;
979 bs->slice_time = BLOCK_IO_SLICE_TIME;
981 if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
982 bdrv_io_limits_enable(bs);
983 } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
984 bdrv_io_limits_disable(bs);
985 } else {
986 if (bs->block_timer) {
987 qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
992 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
994 const char *id = qdict_get_str(qdict, "id");
995 BlockDriverState *bs;
997 bs = bdrv_find(id);
998 if (!bs) {
999 qerror_report(QERR_DEVICE_NOT_FOUND, id);
1000 return -1;
1002 if (bdrv_in_use(bs)) {
1003 qerror_report(QERR_DEVICE_IN_USE, id);
1004 return -1;
1007 /* quiesce block driver; prevent further io */
1008 bdrv_drain_all();
1009 bdrv_flush(bs);
1010 bdrv_close(bs);
1012 /* if we have a device attached to this BlockDriverState
1013 * then we need to make the drive anonymous until the device
1014 * can be removed. If this is a drive with no device backing
1015 * then we can just get rid of the block driver state right here.
1017 if (bdrv_get_attached_dev(bs)) {
1018 bdrv_make_anon(bs);
1019 } else {
1020 drive_uninit(drive_get_by_blockdev(bs));
1023 return 0;
1026 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1028 BlockDriverState *bs;
1030 bs = bdrv_find(device);
1031 if (!bs) {
1032 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1033 return;
1036 if (size < 0) {
1037 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1038 return;
1041 switch (bdrv_truncate(bs, size)) {
1042 case 0:
1043 break;
1044 case -ENOMEDIUM:
1045 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1046 break;
1047 case -ENOTSUP:
1048 error_set(errp, QERR_UNSUPPORTED);
1049 break;
1050 case -EACCES:
1051 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1052 break;
1053 case -EBUSY:
1054 error_set(errp, QERR_DEVICE_IN_USE, device);
1055 break;
1056 default:
1057 error_set(errp, QERR_UNDEFINED_ERROR);
1058 break;
1062 static void block_job_cb(void *opaque, int ret)
1064 BlockDriverState *bs = opaque;
1065 QObject *obj;
1067 trace_block_job_cb(bs, bs->job, ret);
1069 assert(bs->job);
1070 obj = qobject_from_block_job(bs->job);
1071 if (ret < 0) {
1072 QDict *dict = qobject_to_qdict(obj);
1073 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1076 if (block_job_is_cancelled(bs->job)) {
1077 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1078 } else {
1079 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1081 qobject_decref(obj);
1083 drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1086 void qmp_block_stream(const char *device, bool has_base,
1087 const char *base, bool has_speed, int64_t speed,
1088 bool has_on_error, BlockdevOnError on_error,
1089 Error **errp)
1091 BlockDriverState *bs;
1092 BlockDriverState *base_bs = NULL;
1093 Error *local_err = NULL;
1095 if (!has_on_error) {
1096 on_error = BLOCKDEV_ON_ERROR_REPORT;
1099 bs = bdrv_find(device);
1100 if (!bs) {
1101 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1102 return;
1105 if (base) {
1106 base_bs = bdrv_find_backing_image(bs, base);
1107 if (base_bs == NULL) {
1108 error_set(errp, QERR_BASE_NOT_FOUND, base);
1109 return;
1113 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1114 on_error, block_job_cb, bs, &local_err);
1115 if (error_is_set(&local_err)) {
1116 error_propagate(errp, local_err);
1117 return;
1120 /* Grab a reference so hotplug does not delete the BlockDriverState from
1121 * underneath us.
1123 drive_get_ref(drive_get_by_blockdev(bs));
1125 trace_qmp_block_stream(bs, bs->job);
1128 void qmp_block_commit(const char *device,
1129 bool has_base, const char *base, const char *top,
1130 bool has_speed, int64_t speed,
1131 Error **errp)
1133 BlockDriverState *bs;
1134 BlockDriverState *base_bs, *top_bs;
1135 Error *local_err = NULL;
1136 /* This will be part of the QMP command, if/when the
1137 * BlockdevOnError change for blkmirror makes it in
1139 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1141 /* drain all i/o before commits */
1142 bdrv_drain_all();
1144 bs = bdrv_find(device);
1145 if (!bs) {
1146 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1147 return;
1150 /* default top_bs is the active layer */
1151 top_bs = bs;
1153 if (top) {
1154 if (strcmp(bs->filename, top) != 0) {
1155 top_bs = bdrv_find_backing_image(bs, top);
1159 if (top_bs == NULL) {
1160 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1161 return;
1164 if (has_base && base) {
1165 base_bs = bdrv_find_backing_image(top_bs, base);
1166 } else {
1167 base_bs = bdrv_find_base(top_bs);
1170 if (base_bs == NULL) {
1171 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1172 return;
1175 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1176 &local_err);
1177 if (local_err != NULL) {
1178 error_propagate(errp, local_err);
1179 return;
1181 /* Grab a reference so hotplug does not delete the BlockDriverState from
1182 * underneath us.
1184 drive_get_ref(drive_get_by_blockdev(bs));
1187 void qmp_drive_mirror(const char *device, const char *target,
1188 bool has_format, const char *format,
1189 enum MirrorSyncMode sync,
1190 bool has_mode, enum NewImageMode mode,
1191 bool has_speed, int64_t speed,
1192 bool has_on_source_error, BlockdevOnError on_source_error,
1193 bool has_on_target_error, BlockdevOnError on_target_error,
1194 Error **errp)
1196 BlockDriverInfo bdi;
1197 BlockDriverState *bs;
1198 BlockDriverState *source, *target_bs;
1199 BlockDriver *proto_drv;
1200 BlockDriver *drv = NULL;
1201 Error *local_err = NULL;
1202 int flags;
1203 uint64_t size;
1204 int ret;
1206 if (!has_speed) {
1207 speed = 0;
1209 if (!has_on_source_error) {
1210 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1212 if (!has_on_target_error) {
1213 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1215 if (!has_mode) {
1216 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1219 bs = bdrv_find(device);
1220 if (!bs) {
1221 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1222 return;
1225 if (!bdrv_is_inserted(bs)) {
1226 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1227 return;
1230 if (!has_format) {
1231 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1233 if (format) {
1234 drv = bdrv_find_format(format);
1235 if (!drv) {
1236 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1237 return;
1241 if (bdrv_in_use(bs)) {
1242 error_set(errp, QERR_DEVICE_IN_USE, device);
1243 return;
1246 flags = bs->open_flags | BDRV_O_RDWR;
1247 source = bs->backing_hd;
1248 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1249 sync = MIRROR_SYNC_MODE_FULL;
1252 proto_drv = bdrv_find_protocol(target);
1253 if (!proto_drv) {
1254 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1255 return;
1258 if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1259 /* create new image w/o backing file */
1260 assert(format && drv);
1261 bdrv_get_geometry(bs, &size);
1262 size *= 512;
1263 bdrv_img_create(target, format,
1264 NULL, NULL, NULL, size, flags, &local_err);
1265 } else {
1266 switch (mode) {
1267 case NEW_IMAGE_MODE_EXISTING:
1268 ret = 0;
1269 break;
1270 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1271 /* create new image with backing file */
1272 bdrv_img_create(target, format,
1273 source->filename,
1274 source->drv->format_name,
1275 NULL, -1, flags, &local_err);
1276 break;
1277 default:
1278 abort();
1282 if (error_is_set(&local_err)) {
1283 error_propagate(errp, local_err);
1284 return;
1287 target_bs = bdrv_new("");
1288 ret = bdrv_open(target_bs, target, flags | BDRV_O_NO_BACKING, drv);
1290 if (ret < 0) {
1291 bdrv_delete(target_bs);
1292 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1293 return;
1296 /* We need a backing file if we will copy parts of a cluster. */
1297 if (bdrv_get_info(target_bs, &bdi) >= 0 && bdi.cluster_size != 0 &&
1298 bdi.cluster_size >= BDRV_SECTORS_PER_DIRTY_CHUNK * 512) {
1299 ret = bdrv_open_backing_file(target_bs);
1300 if (ret < 0) {
1301 bdrv_delete(target_bs);
1302 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1303 return;
1307 mirror_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1308 block_job_cb, bs, &local_err);
1309 if (local_err != NULL) {
1310 bdrv_delete(target_bs);
1311 error_propagate(errp, local_err);
1312 return;
1315 /* Grab a reference so hotplug does not delete the BlockDriverState from
1316 * underneath us.
1318 drive_get_ref(drive_get_by_blockdev(bs));
1321 static BlockJob *find_block_job(const char *device)
1323 BlockDriverState *bs;
1325 bs = bdrv_find(device);
1326 if (!bs || !bs->job) {
1327 return NULL;
1329 return bs->job;
1332 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1334 BlockJob *job = find_block_job(device);
1336 if (!job) {
1337 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1338 return;
1341 block_job_set_speed(job, speed, errp);
1344 void qmp_block_job_cancel(const char *device,
1345 bool has_force, bool force, Error **errp)
1347 BlockJob *job = find_block_job(device);
1349 if (!has_force) {
1350 force = false;
1353 if (!job) {
1354 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1355 return;
1357 if (job->paused && !force) {
1358 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1359 return;
1362 trace_qmp_block_job_cancel(job);
1363 block_job_cancel(job);
1366 void qmp_block_job_pause(const char *device, Error **errp)
1368 BlockJob *job = find_block_job(device);
1370 if (!job) {
1371 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1372 return;
1375 trace_qmp_block_job_pause(job);
1376 block_job_pause(job);
1379 void qmp_block_job_resume(const char *device, Error **errp)
1381 BlockJob *job = find_block_job(device);
1383 if (!job) {
1384 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1385 return;
1388 trace_qmp_block_job_resume(job);
1389 block_job_resume(job);
1392 void qmp_block_job_complete(const char *device, Error **errp)
1394 BlockJob *job = find_block_job(device);
1396 if (!job) {
1397 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1398 return;
1401 trace_qmp_block_job_complete(job);
1402 block_job_complete(job, errp);
1405 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1407 BlockJobInfoList **prev = opaque;
1408 BlockJob *job = bs->job;
1410 if (job) {
1411 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1412 elem->value = block_job_query(bs->job);
1413 (*prev)->next = elem;
1414 *prev = elem;
1418 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1420 /* Dummy is a fake list element for holding the head pointer */
1421 BlockJobInfoList dummy = {};
1422 BlockJobInfoList *prev = &dummy;
1423 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1424 return dummy.next;
1427 QemuOptsList qemu_drive_opts = {
1428 .name = "drive",
1429 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1430 .desc = {
1432 .name = "bus",
1433 .type = QEMU_OPT_NUMBER,
1434 .help = "bus number",
1436 .name = "unit",
1437 .type = QEMU_OPT_NUMBER,
1438 .help = "unit number (i.e. lun for scsi)",
1440 .name = "if",
1441 .type = QEMU_OPT_STRING,
1442 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1444 .name = "index",
1445 .type = QEMU_OPT_NUMBER,
1446 .help = "index number",
1448 .name = "cyls",
1449 .type = QEMU_OPT_NUMBER,
1450 .help = "number of cylinders (ide disk geometry)",
1452 .name = "heads",
1453 .type = QEMU_OPT_NUMBER,
1454 .help = "number of heads (ide disk geometry)",
1456 .name = "secs",
1457 .type = QEMU_OPT_NUMBER,
1458 .help = "number of sectors (ide disk geometry)",
1460 .name = "trans",
1461 .type = QEMU_OPT_STRING,
1462 .help = "chs translation (auto, lba. none)",
1464 .name = "media",
1465 .type = QEMU_OPT_STRING,
1466 .help = "media type (disk, cdrom)",
1468 .name = "snapshot",
1469 .type = QEMU_OPT_BOOL,
1470 .help = "enable/disable snapshot mode",
1472 .name = "file",
1473 .type = QEMU_OPT_STRING,
1474 .help = "disk image",
1476 .name = "cache",
1477 .type = QEMU_OPT_STRING,
1478 .help = "host cache usage (none, writeback, writethrough, "
1479 "directsync, unsafe)",
1481 .name = "aio",
1482 .type = QEMU_OPT_STRING,
1483 .help = "host AIO implementation (threads, native)",
1485 .name = "format",
1486 .type = QEMU_OPT_STRING,
1487 .help = "disk format (raw, qcow2, ...)",
1489 .name = "serial",
1490 .type = QEMU_OPT_STRING,
1491 .help = "disk serial number",
1493 .name = "rerror",
1494 .type = QEMU_OPT_STRING,
1495 .help = "read error action",
1497 .name = "werror",
1498 .type = QEMU_OPT_STRING,
1499 .help = "write error action",
1501 .name = "addr",
1502 .type = QEMU_OPT_STRING,
1503 .help = "pci address (virtio only)",
1505 .name = "readonly",
1506 .type = QEMU_OPT_BOOL,
1507 .help = "open drive file as read-only",
1509 .name = "iops",
1510 .type = QEMU_OPT_NUMBER,
1511 .help = "limit total I/O operations per second",
1513 .name = "iops_rd",
1514 .type = QEMU_OPT_NUMBER,
1515 .help = "limit read operations per second",
1517 .name = "iops_wr",
1518 .type = QEMU_OPT_NUMBER,
1519 .help = "limit write operations per second",
1521 .name = "bps",
1522 .type = QEMU_OPT_NUMBER,
1523 .help = "limit total bytes per second",
1525 .name = "bps_rd",
1526 .type = QEMU_OPT_NUMBER,
1527 .help = "limit read bytes per second",
1529 .name = "bps_wr",
1530 .type = QEMU_OPT_NUMBER,
1531 .help = "limit write bytes per second",
1533 .name = "copy-on-read",
1534 .type = QEMU_OPT_BOOL,
1535 .help = "copy read data from backing file into image file",
1537 .name = "boot",
1538 .type = QEMU_OPT_BOOL,
1539 .help = "(deprecated, ignored)",
1541 { /* end of list */ }