Merge remote-tracking branch 'mst/tags/for_anthony' into staging
[qemu/rayw.git] / blockdev.c
blob9126587c459e5e2e5efea640994b337f8d991de0
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 if (ret == -EBUSY) {
646 qerror_report(QERR_DEVICE_IN_USE, device);
647 return;
649 } else {
650 bs = bdrv_find(device);
651 if (!bs) {
652 qerror_report(QERR_DEVICE_NOT_FOUND, device);
653 return;
655 ret = bdrv_commit(bs);
656 if (ret == -EBUSY) {
657 qerror_report(QERR_DEVICE_IN_USE, device);
658 return;
663 static void blockdev_do_action(int kind, void *data, Error **errp)
665 BlockdevAction action;
666 BlockdevActionList list;
668 action.kind = kind;
669 action.data = data;
670 list.value = &action;
671 list.next = NULL;
672 qmp_transaction(&list, errp);
675 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
676 bool has_format, const char *format,
677 bool has_mode, enum NewImageMode mode,
678 Error **errp)
680 BlockdevSnapshot snapshot = {
681 .device = (char *) device,
682 .snapshot_file = (char *) snapshot_file,
683 .has_format = has_format,
684 .format = (char *) format,
685 .has_mode = has_mode,
686 .mode = mode,
688 blockdev_do_action(BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC, &snapshot,
689 errp);
693 /* New and old BlockDriverState structs for group snapshots */
694 typedef struct BlkTransactionStates {
695 BlockDriverState *old_bs;
696 BlockDriverState *new_bs;
697 QSIMPLEQ_ENTRY(BlkTransactionStates) entry;
698 } BlkTransactionStates;
701 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
702 * then we do not pivot any of the devices in the group, and abandon the
703 * snapshots
705 void qmp_transaction(BlockdevActionList *dev_list, Error **errp)
707 int ret = 0;
708 BlockdevActionList *dev_entry = dev_list;
709 BlkTransactionStates *states, *next;
710 Error *local_err = NULL;
712 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionStates) snap_bdrv_states;
713 QSIMPLEQ_INIT(&snap_bdrv_states);
715 /* drain all i/o before any snapshots */
716 bdrv_drain_all();
718 /* We don't do anything in this loop that commits us to the snapshot */
719 while (NULL != dev_entry) {
720 BlockdevAction *dev_info = NULL;
721 BlockDriver *proto_drv;
722 BlockDriver *drv;
723 int flags;
724 enum NewImageMode mode;
725 const char *new_image_file;
726 const char *device;
727 const char *format = "qcow2";
729 dev_info = dev_entry->value;
730 dev_entry = dev_entry->next;
732 states = g_malloc0(sizeof(BlkTransactionStates));
733 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, states, entry);
735 switch (dev_info->kind) {
736 case BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
737 device = dev_info->blockdev_snapshot_sync->device;
738 if (!dev_info->blockdev_snapshot_sync->has_mode) {
739 dev_info->blockdev_snapshot_sync->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
741 new_image_file = dev_info->blockdev_snapshot_sync->snapshot_file;
742 if (dev_info->blockdev_snapshot_sync->has_format) {
743 format = dev_info->blockdev_snapshot_sync->format;
745 mode = dev_info->blockdev_snapshot_sync->mode;
746 break;
747 default:
748 abort();
751 drv = bdrv_find_format(format);
752 if (!drv) {
753 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
754 goto delete_and_fail;
757 states->old_bs = bdrv_find(device);
758 if (!states->old_bs) {
759 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
760 goto delete_and_fail;
763 if (!bdrv_is_inserted(states->old_bs)) {
764 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
765 goto delete_and_fail;
768 if (bdrv_in_use(states->old_bs)) {
769 error_set(errp, QERR_DEVICE_IN_USE, device);
770 goto delete_and_fail;
773 if (!bdrv_is_read_only(states->old_bs)) {
774 if (bdrv_flush(states->old_bs)) {
775 error_set(errp, QERR_IO_ERROR);
776 goto delete_and_fail;
780 flags = states->old_bs->open_flags;
782 proto_drv = bdrv_find_protocol(new_image_file);
783 if (!proto_drv) {
784 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
785 goto delete_and_fail;
788 /* create new image w/backing file */
789 if (mode != NEW_IMAGE_MODE_EXISTING) {
790 bdrv_img_create(new_image_file, format,
791 states->old_bs->filename,
792 states->old_bs->drv->format_name,
793 NULL, -1, flags, &local_err);
794 if (error_is_set(&local_err)) {
795 error_propagate(errp, local_err);
796 goto delete_and_fail;
800 /* We will manually add the backing_hd field to the bs later */
801 states->new_bs = bdrv_new("");
802 ret = bdrv_open(states->new_bs, new_image_file,
803 flags | BDRV_O_NO_BACKING, drv);
804 if (ret != 0) {
805 error_set(errp, QERR_OPEN_FILE_FAILED, new_image_file);
806 goto delete_and_fail;
811 /* Now we are going to do the actual pivot. Everything up to this point
812 * is reversible, but we are committed at this point */
813 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
814 /* This removes our old bs from the bdrv_states, and adds the new bs */
815 bdrv_append(states->new_bs, states->old_bs);
816 /* We don't need (or want) to use the transactional
817 * bdrv_reopen_multiple() across all the entries at once, because we
818 * don't want to abort all of them if one of them fails the reopen */
819 bdrv_reopen(states->new_bs, states->new_bs->open_flags & ~BDRV_O_RDWR,
820 NULL);
823 /* success */
824 goto exit;
826 delete_and_fail:
828 * failure, and it is all-or-none; abandon each new bs, and keep using
829 * the original bs for all images
831 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
832 if (states->new_bs) {
833 bdrv_delete(states->new_bs);
836 exit:
837 QSIMPLEQ_FOREACH_SAFE(states, &snap_bdrv_states, entry, next) {
838 g_free(states);
843 static void eject_device(BlockDriverState *bs, int force, Error **errp)
845 if (bdrv_in_use(bs)) {
846 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
847 return;
849 if (!bdrv_dev_has_removable_media(bs)) {
850 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
851 return;
854 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
855 bdrv_dev_eject_request(bs, force);
856 if (!force) {
857 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
858 return;
862 bdrv_close(bs);
865 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
867 BlockDriverState *bs;
869 bs = bdrv_find(device);
870 if (!bs) {
871 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
872 return;
875 eject_device(bs, force, errp);
878 void qmp_block_passwd(const char *device, const char *password, Error **errp)
880 BlockDriverState *bs;
881 int err;
883 bs = bdrv_find(device);
884 if (!bs) {
885 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
886 return;
889 err = bdrv_set_key(bs, password);
890 if (err == -EINVAL) {
891 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
892 return;
893 } else if (err < 0) {
894 error_set(errp, QERR_INVALID_PASSWORD);
895 return;
899 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
900 int bdrv_flags, BlockDriver *drv,
901 const char *password, Error **errp)
903 if (bdrv_open(bs, filename, bdrv_flags, drv) < 0) {
904 error_set(errp, QERR_OPEN_FILE_FAILED, filename);
905 return;
908 if (bdrv_key_required(bs)) {
909 if (password) {
910 if (bdrv_set_key(bs, password) < 0) {
911 error_set(errp, QERR_INVALID_PASSWORD);
913 } else {
914 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
915 bdrv_get_encrypted_filename(bs));
917 } else if (password) {
918 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
922 void qmp_change_blockdev(const char *device, const char *filename,
923 bool has_format, const char *format, Error **errp)
925 BlockDriverState *bs;
926 BlockDriver *drv = NULL;
927 int bdrv_flags;
928 Error *err = NULL;
930 bs = bdrv_find(device);
931 if (!bs) {
932 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
933 return;
936 if (format) {
937 drv = bdrv_find_whitelisted_format(format);
938 if (!drv) {
939 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
940 return;
944 eject_device(bs, 0, &err);
945 if (error_is_set(&err)) {
946 error_propagate(errp, err);
947 return;
950 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
951 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
953 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
956 /* throttling disk I/O limits */
957 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
958 int64_t bps_wr, int64_t iops, int64_t iops_rd,
959 int64_t iops_wr, Error **errp)
961 BlockIOLimit io_limits;
962 BlockDriverState *bs;
964 bs = bdrv_find(device);
965 if (!bs) {
966 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
967 return;
970 io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
971 io_limits.bps[BLOCK_IO_LIMIT_READ] = bps_rd;
972 io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
973 io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
974 io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
975 io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
977 if (!do_check_io_limits(&io_limits)) {
978 error_set(errp, QERR_INVALID_PARAMETER_COMBINATION);
979 return;
982 bs->io_limits = io_limits;
983 bs->slice_time = BLOCK_IO_SLICE_TIME;
985 if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
986 bdrv_io_limits_enable(bs);
987 } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
988 bdrv_io_limits_disable(bs);
989 } else {
990 if (bs->block_timer) {
991 qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
996 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
998 const char *id = qdict_get_str(qdict, "id");
999 BlockDriverState *bs;
1001 bs = bdrv_find(id);
1002 if (!bs) {
1003 qerror_report(QERR_DEVICE_NOT_FOUND, id);
1004 return -1;
1006 if (bdrv_in_use(bs)) {
1007 qerror_report(QERR_DEVICE_IN_USE, id);
1008 return -1;
1011 /* quiesce block driver; prevent further io */
1012 bdrv_drain_all();
1013 bdrv_flush(bs);
1014 bdrv_close(bs);
1016 /* if we have a device attached to this BlockDriverState
1017 * then we need to make the drive anonymous until the device
1018 * can be removed. If this is a drive with no device backing
1019 * then we can just get rid of the block driver state right here.
1021 if (bdrv_get_attached_dev(bs)) {
1022 bdrv_make_anon(bs);
1023 } else {
1024 drive_uninit(drive_get_by_blockdev(bs));
1027 return 0;
1030 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1032 BlockDriverState *bs;
1034 bs = bdrv_find(device);
1035 if (!bs) {
1036 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1037 return;
1040 if (size < 0) {
1041 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1042 return;
1045 switch (bdrv_truncate(bs, size)) {
1046 case 0:
1047 break;
1048 case -ENOMEDIUM:
1049 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1050 break;
1051 case -ENOTSUP:
1052 error_set(errp, QERR_UNSUPPORTED);
1053 break;
1054 case -EACCES:
1055 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1056 break;
1057 case -EBUSY:
1058 error_set(errp, QERR_DEVICE_IN_USE, device);
1059 break;
1060 default:
1061 error_set(errp, QERR_UNDEFINED_ERROR);
1062 break;
1066 static void block_job_cb(void *opaque, int ret)
1068 BlockDriverState *bs = opaque;
1069 QObject *obj;
1071 trace_block_job_cb(bs, bs->job, ret);
1073 assert(bs->job);
1074 obj = qobject_from_block_job(bs->job);
1075 if (ret < 0) {
1076 QDict *dict = qobject_to_qdict(obj);
1077 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1080 if (block_job_is_cancelled(bs->job)) {
1081 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1082 } else {
1083 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1085 qobject_decref(obj);
1087 drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1090 void qmp_block_stream(const char *device, bool has_base,
1091 const char *base, bool has_speed, int64_t speed,
1092 bool has_on_error, BlockdevOnError on_error,
1093 Error **errp)
1095 BlockDriverState *bs;
1096 BlockDriverState *base_bs = NULL;
1097 Error *local_err = NULL;
1099 if (!has_on_error) {
1100 on_error = BLOCKDEV_ON_ERROR_REPORT;
1103 bs = bdrv_find(device);
1104 if (!bs) {
1105 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1106 return;
1109 if (base) {
1110 base_bs = bdrv_find_backing_image(bs, base);
1111 if (base_bs == NULL) {
1112 error_set(errp, QERR_BASE_NOT_FOUND, base);
1113 return;
1117 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1118 on_error, block_job_cb, bs, &local_err);
1119 if (error_is_set(&local_err)) {
1120 error_propagate(errp, local_err);
1121 return;
1124 /* Grab a reference so hotplug does not delete the BlockDriverState from
1125 * underneath us.
1127 drive_get_ref(drive_get_by_blockdev(bs));
1129 trace_qmp_block_stream(bs, bs->job);
1132 void qmp_block_commit(const char *device,
1133 bool has_base, const char *base, const char *top,
1134 bool has_speed, int64_t speed,
1135 Error **errp)
1137 BlockDriverState *bs;
1138 BlockDriverState *base_bs, *top_bs;
1139 Error *local_err = NULL;
1140 /* This will be part of the QMP command, if/when the
1141 * BlockdevOnError change for blkmirror makes it in
1143 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1145 /* drain all i/o before commits */
1146 bdrv_drain_all();
1148 bs = bdrv_find(device);
1149 if (!bs) {
1150 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1151 return;
1154 /* default top_bs is the active layer */
1155 top_bs = bs;
1157 if (top) {
1158 if (strcmp(bs->filename, top) != 0) {
1159 top_bs = bdrv_find_backing_image(bs, top);
1163 if (top_bs == NULL) {
1164 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1165 return;
1168 if (has_base && base) {
1169 base_bs = bdrv_find_backing_image(top_bs, base);
1170 } else {
1171 base_bs = bdrv_find_base(top_bs);
1174 if (base_bs == NULL) {
1175 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1176 return;
1179 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1180 &local_err);
1181 if (local_err != NULL) {
1182 error_propagate(errp, local_err);
1183 return;
1185 /* Grab a reference so hotplug does not delete the BlockDriverState from
1186 * underneath us.
1188 drive_get_ref(drive_get_by_blockdev(bs));
1191 void qmp_drive_mirror(const char *device, const char *target,
1192 bool has_format, const char *format,
1193 enum MirrorSyncMode sync,
1194 bool has_mode, enum NewImageMode mode,
1195 bool has_speed, int64_t speed,
1196 bool has_on_source_error, BlockdevOnError on_source_error,
1197 bool has_on_target_error, BlockdevOnError on_target_error,
1198 Error **errp)
1200 BlockDriverInfo bdi;
1201 BlockDriverState *bs;
1202 BlockDriverState *source, *target_bs;
1203 BlockDriver *proto_drv;
1204 BlockDriver *drv = NULL;
1205 Error *local_err = NULL;
1206 int flags;
1207 uint64_t size;
1208 int ret;
1210 if (!has_speed) {
1211 speed = 0;
1213 if (!has_on_source_error) {
1214 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1216 if (!has_on_target_error) {
1217 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1219 if (!has_mode) {
1220 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1223 bs = bdrv_find(device);
1224 if (!bs) {
1225 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1226 return;
1229 if (!bdrv_is_inserted(bs)) {
1230 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1231 return;
1234 if (!has_format) {
1235 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1237 if (format) {
1238 drv = bdrv_find_format(format);
1239 if (!drv) {
1240 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1241 return;
1245 if (bdrv_in_use(bs)) {
1246 error_set(errp, QERR_DEVICE_IN_USE, device);
1247 return;
1250 flags = bs->open_flags | BDRV_O_RDWR;
1251 source = bs->backing_hd;
1252 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1253 sync = MIRROR_SYNC_MODE_FULL;
1256 proto_drv = bdrv_find_protocol(target);
1257 if (!proto_drv) {
1258 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1259 return;
1262 if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1263 /* create new image w/o backing file */
1264 assert(format && drv);
1265 bdrv_get_geometry(bs, &size);
1266 size *= 512;
1267 bdrv_img_create(target, format,
1268 NULL, NULL, NULL, size, flags, &local_err);
1269 } else {
1270 switch (mode) {
1271 case NEW_IMAGE_MODE_EXISTING:
1272 ret = 0;
1273 break;
1274 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1275 /* create new image with backing file */
1276 bdrv_img_create(target, format,
1277 source->filename,
1278 source->drv->format_name,
1279 NULL, -1, flags, &local_err);
1280 break;
1281 default:
1282 abort();
1286 if (error_is_set(&local_err)) {
1287 error_propagate(errp, local_err);
1288 return;
1291 target_bs = bdrv_new("");
1292 ret = bdrv_open(target_bs, target, flags | BDRV_O_NO_BACKING, drv);
1294 if (ret < 0) {
1295 bdrv_delete(target_bs);
1296 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1297 return;
1300 /* We need a backing file if we will copy parts of a cluster. */
1301 if (bdrv_get_info(target_bs, &bdi) >= 0 && bdi.cluster_size != 0 &&
1302 bdi.cluster_size >= BDRV_SECTORS_PER_DIRTY_CHUNK * 512) {
1303 ret = bdrv_open_backing_file(target_bs);
1304 if (ret < 0) {
1305 bdrv_delete(target_bs);
1306 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1307 return;
1311 mirror_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1312 block_job_cb, bs, &local_err);
1313 if (local_err != NULL) {
1314 bdrv_delete(target_bs);
1315 error_propagate(errp, local_err);
1316 return;
1319 /* Grab a reference so hotplug does not delete the BlockDriverState from
1320 * underneath us.
1322 drive_get_ref(drive_get_by_blockdev(bs));
1325 static BlockJob *find_block_job(const char *device)
1327 BlockDriverState *bs;
1329 bs = bdrv_find(device);
1330 if (!bs || !bs->job) {
1331 return NULL;
1333 return bs->job;
1336 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1338 BlockJob *job = find_block_job(device);
1340 if (!job) {
1341 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1342 return;
1345 block_job_set_speed(job, speed, errp);
1348 void qmp_block_job_cancel(const char *device,
1349 bool has_force, bool force, Error **errp)
1351 BlockJob *job = find_block_job(device);
1353 if (!has_force) {
1354 force = false;
1357 if (!job) {
1358 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1359 return;
1361 if (job->paused && !force) {
1362 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1363 return;
1366 trace_qmp_block_job_cancel(job);
1367 block_job_cancel(job);
1370 void qmp_block_job_pause(const char *device, Error **errp)
1372 BlockJob *job = find_block_job(device);
1374 if (!job) {
1375 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1376 return;
1379 trace_qmp_block_job_pause(job);
1380 block_job_pause(job);
1383 void qmp_block_job_resume(const char *device, Error **errp)
1385 BlockJob *job = find_block_job(device);
1387 if (!job) {
1388 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1389 return;
1392 trace_qmp_block_job_resume(job);
1393 block_job_resume(job);
1396 void qmp_block_job_complete(const char *device, Error **errp)
1398 BlockJob *job = find_block_job(device);
1400 if (!job) {
1401 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1402 return;
1405 trace_qmp_block_job_complete(job);
1406 block_job_complete(job, errp);
1409 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1411 BlockJobInfoList **prev = opaque;
1412 BlockJob *job = bs->job;
1414 if (job) {
1415 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1416 elem->value = block_job_query(bs->job);
1417 (*prev)->next = elem;
1418 *prev = elem;
1422 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1424 /* Dummy is a fake list element for holding the head pointer */
1425 BlockJobInfoList dummy = {};
1426 BlockJobInfoList *prev = &dummy;
1427 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1428 return dummy.next;
1431 QemuOptsList qemu_drive_opts = {
1432 .name = "drive",
1433 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1434 .desc = {
1436 .name = "bus",
1437 .type = QEMU_OPT_NUMBER,
1438 .help = "bus number",
1440 .name = "unit",
1441 .type = QEMU_OPT_NUMBER,
1442 .help = "unit number (i.e. lun for scsi)",
1444 .name = "if",
1445 .type = QEMU_OPT_STRING,
1446 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1448 .name = "index",
1449 .type = QEMU_OPT_NUMBER,
1450 .help = "index number",
1452 .name = "cyls",
1453 .type = QEMU_OPT_NUMBER,
1454 .help = "number of cylinders (ide disk geometry)",
1456 .name = "heads",
1457 .type = QEMU_OPT_NUMBER,
1458 .help = "number of heads (ide disk geometry)",
1460 .name = "secs",
1461 .type = QEMU_OPT_NUMBER,
1462 .help = "number of sectors (ide disk geometry)",
1464 .name = "trans",
1465 .type = QEMU_OPT_STRING,
1466 .help = "chs translation (auto, lba. none)",
1468 .name = "media",
1469 .type = QEMU_OPT_STRING,
1470 .help = "media type (disk, cdrom)",
1472 .name = "snapshot",
1473 .type = QEMU_OPT_BOOL,
1474 .help = "enable/disable snapshot mode",
1476 .name = "file",
1477 .type = QEMU_OPT_STRING,
1478 .help = "disk image",
1480 .name = "cache",
1481 .type = QEMU_OPT_STRING,
1482 .help = "host cache usage (none, writeback, writethrough, "
1483 "directsync, unsafe)",
1485 .name = "aio",
1486 .type = QEMU_OPT_STRING,
1487 .help = "host AIO implementation (threads, native)",
1489 .name = "format",
1490 .type = QEMU_OPT_STRING,
1491 .help = "disk format (raw, qcow2, ...)",
1493 .name = "serial",
1494 .type = QEMU_OPT_STRING,
1495 .help = "disk serial number",
1497 .name = "rerror",
1498 .type = QEMU_OPT_STRING,
1499 .help = "read error action",
1501 .name = "werror",
1502 .type = QEMU_OPT_STRING,
1503 .help = "write error action",
1505 .name = "addr",
1506 .type = QEMU_OPT_STRING,
1507 .help = "pci address (virtio only)",
1509 .name = "readonly",
1510 .type = QEMU_OPT_BOOL,
1511 .help = "open drive file as read-only",
1513 .name = "iops",
1514 .type = QEMU_OPT_NUMBER,
1515 .help = "limit total I/O operations per second",
1517 .name = "iops_rd",
1518 .type = QEMU_OPT_NUMBER,
1519 .help = "limit read operations per second",
1521 .name = "iops_wr",
1522 .type = QEMU_OPT_NUMBER,
1523 .help = "limit write operations per second",
1525 .name = "bps",
1526 .type = QEMU_OPT_NUMBER,
1527 .help = "limit total bytes per second",
1529 .name = "bps_rd",
1530 .type = QEMU_OPT_NUMBER,
1531 .help = "limit read bytes per second",
1533 .name = "bps_wr",
1534 .type = QEMU_OPT_NUMBER,
1535 .help = "limit write bytes per second",
1537 .name = "copy-on-read",
1538 .type = QEMU_OPT_BOOL,
1539 .help = "copy read data from backing file into image file",
1541 .name = "boot",
1542 .type = QEMU_OPT_BOOL,
1543 .help = "(deprecated, ignored)",
1545 { /* end of list */ }