]> git.proxmox.com Git - mirror_qemu.git/blob - blockdev.c
qmp: add drive-mirror command
[mirror_qemu.git] / blockdev.c
1 /*
2 * QEMU host block devices
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
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 */
9
10 #include "blockdev.h"
11 #include "hw/block-common.h"
12 #include "blockjob.h"
13 #include "monitor.h"
14 #include "qerror.h"
15 #include "qemu-option.h"
16 #include "qemu-config.h"
17 #include "qemu-objects.h"
18 #include "sysemu.h"
19 #include "block_int.h"
20 #include "qmp-commands.h"
21 #include "trace.h"
22 #include "arch_init.h"
23
24 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
25
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",
36 };
37
38 static const int if_max_devs[IF_COUNT] = {
39 /*
40 * Do not change these numbers! They govern how drive option
41 * index maps to unit and bus. That mapping is ABI.
42 *
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.
47 *
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.
52 */
53 [IF_IDE] = 2,
54 [IF_SCSI] = 7,
55 };
56
57 /*
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.
63 */
64 void blockdev_mark_auto_del(BlockDriverState *bs)
65 {
66 DriveInfo *dinfo = drive_get_by_blockdev(bs);
67
68 if (bs->job) {
69 block_job_cancel(bs->job);
70 }
71 if (dinfo) {
72 dinfo->auto_del = 1;
73 }
74 }
75
76 void blockdev_auto_del(BlockDriverState *bs)
77 {
78 DriveInfo *dinfo = drive_get_by_blockdev(bs);
79
80 if (dinfo && dinfo->auto_del) {
81 drive_put_ref(dinfo);
82 }
83 }
84
85 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
86 {
87 int max_devs = if_max_devs[type];
88 return max_devs ? index / max_devs : 0;
89 }
90
91 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
92 {
93 int max_devs = if_max_devs[type];
94 return max_devs ? index % max_devs : index;
95 }
96
97 QemuOpts *drive_def(const char *optstr)
98 {
99 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
100 }
101
102 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
103 const char *optstr)
104 {
105 QemuOpts *opts;
106 char buf[32];
107
108 opts = drive_def(optstr);
109 if (!opts) {
110 return NULL;
111 }
112 if (type != IF_DEFAULT) {
113 qemu_opt_set(opts, "if", if_name[type]);
114 }
115 if (index >= 0) {
116 snprintf(buf, sizeof(buf), "%d", index);
117 qemu_opt_set(opts, "index", buf);
118 }
119 if (file)
120 qemu_opt_set(opts, "file", file);
121 return opts;
122 }
123
124 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
125 {
126 DriveInfo *dinfo;
127
128 /* seek interface, bus and unit */
129
130 QTAILQ_FOREACH(dinfo, &drives, next) {
131 if (dinfo->type == type &&
132 dinfo->bus == bus &&
133 dinfo->unit == unit)
134 return dinfo;
135 }
136
137 return NULL;
138 }
139
140 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
141 {
142 return drive_get(type,
143 drive_index_to_bus_id(type, index),
144 drive_index_to_unit_id(type, index));
145 }
146
147 int drive_get_max_bus(BlockInterfaceType type)
148 {
149 int max_bus;
150 DriveInfo *dinfo;
151
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;
157 }
158 return max_bus;
159 }
160
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)
165 {
166 static int next_block_unit[IF_COUNT];
167
168 return drive_get(type, 0, next_block_unit[type]++);
169 }
170
171 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
172 {
173 DriveInfo *dinfo;
174
175 QTAILQ_FOREACH(dinfo, &drives, next) {
176 if (dinfo->bdrv == bs) {
177 return dinfo;
178 }
179 }
180 return NULL;
181 }
182
183 static void bdrv_format_print(void *opaque, const char *name)
184 {
185 error_printf(" %s", name);
186 }
187
188 static void drive_uninit(DriveInfo *dinfo)
189 {
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);
195 }
196
197 void drive_put_ref(DriveInfo *dinfo)
198 {
199 assert(dinfo->refcount);
200 if (--dinfo->refcount == 0) {
201 drive_uninit(dinfo);
202 }
203 }
204
205 void drive_get_ref(DriveInfo *dinfo)
206 {
207 dinfo->refcount++;
208 }
209
210 typedef struct {
211 QEMUBH *bh;
212 DriveInfo *dinfo;
213 } DrivePutRefBH;
214
215 static void drive_put_ref_bh(void *opaque)
216 {
217 DrivePutRefBH *s = opaque;
218
219 drive_put_ref(s->dinfo);
220 qemu_bh_delete(s->bh);
221 g_free(s);
222 }
223
224 /*
225 * Release a drive reference in a BH
226 *
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.
230 */
231 static void drive_put_ref_bh_schedule(DriveInfo *dinfo)
232 {
233 DrivePutRefBH *s;
234
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);
239 }
240
241 static int parse_block_error_action(const char *buf, bool is_read)
242 {
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;
255 }
256 }
257
258 static bool do_check_io_limits(BlockIOLimit *io_limits)
259 {
260 bool bps_flag;
261 bool iops_flag;
262
263 assert(io_limits);
264
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;
273 }
274
275 return true;
276 }
277
278 DriveInfo *drive_init(QemuOpts *opts, int default_to_scsi)
279 {
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;
300
301 translation = BIOS_ATA_TRANSLATION_AUTO;
302 media = MEDIA_DISK;
303
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);
308
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);
312
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);
316
317 file = qemu_opt_get(opts, "file");
318 serial = qemu_opt_get(opts, "serial");
319
320 if ((buf = qemu_opt_get(opts, "if")) != NULL) {
321 for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
322 ;
323 if (type == IF_COUNT) {
324 error_report("unsupported bus type '%s'", buf);
325 return NULL;
326 }
327 } else {
328 type = default_to_scsi ? IF_SCSI : IF_IDE;
329 }
330
331 max_devs = if_max_devs[type];
332
333 if (cyls || heads || secs) {
334 if (cyls < 1) {
335 error_report("invalid physical cyls number");
336 return NULL;
337 }
338 if (heads < 1) {
339 error_report("invalid physical heads number");
340 return NULL;
341 }
342 if (secs < 1) {
343 error_report("invalid physical secs number");
344 return NULL;
345 }
346 }
347
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;
353 }
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;
363 }
364 }
365
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;
373 }
374 media = MEDIA_CDROM;
375 } else {
376 error_report("'%s' invalid media", buf);
377 return NULL;
378 }
379 }
380
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;
386 }
387 }
388
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;
398 }
399 }
400 #endif
401
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;
408 }
409 drv = bdrv_find_whitelisted_format(buf);
410 if (!drv) {
411 error_report("'%s' invalid format", buf);
412 return NULL;
413 }
414 }
415
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);
429
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;
434 }
435
436 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
437 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
438 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
439 error_report("werror is not supported by this bus type");
440 return NULL;
441 }
442
443 on_write_error = parse_block_error_action(buf, 0);
444 if (on_write_error < 0) {
445 return NULL;
446 }
447 }
448
449 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
450 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
451 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
452 error_report("rerror is not supported by this bus type");
453 return NULL;
454 }
455
456 on_read_error = parse_block_error_action(buf, 1);
457 if (on_read_error < 0) {
458 return NULL;
459 }
460 }
461
462 if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
463 if (type != IF_VIRTIO) {
464 error_report("addr is not supported by this bus type");
465 return NULL;
466 }
467 }
468
469 /* compute bus and unit according index */
470
471 if (index != -1) {
472 if (bus_id != 0 || unit_id != -1) {
473 error_report("index cannot be used with bus and unit");
474 return NULL;
475 }
476 bus_id = drive_index_to_bus_id(type, index);
477 unit_id = drive_index_to_unit_id(type, index);
478 }
479
480 /* if user doesn't specify a unit_id,
481 * try to find the first free
482 */
483
484 if (unit_id == -1) {
485 unit_id = 0;
486 while (drive_get(type, bus_id, unit_id) != NULL) {
487 unit_id++;
488 if (max_devs && unit_id >= max_devs) {
489 unit_id -= max_devs;
490 bus_id++;
491 }
492 }
493 }
494
495 /* check unit id */
496
497 if (max_devs && unit_id >= max_devs) {
498 error_report("unit %d too big (max is %d)",
499 unit_id, max_devs - 1);
500 return NULL;
501 }
502
503 /*
504 * catch multiple definitions
505 */
506
507 if (drive_get(type, bus_id, unit_id) != NULL) {
508 error_report("drive with bus=%d, unit=%d (index=%d) exists",
509 bus_id, unit_id, index);
510 return NULL;
511 }
512
513 /* init */
514
515 dinfo = g_malloc0(sizeof(*dinfo));
516 if ((buf = qemu_opts_id(opts)) != NULL) {
517 dinfo->id = g_strdup(buf);
518 } else {
519 /* no id supplied -> create one */
520 dinfo->id = g_malloc0(32);
521 if (type == IF_IDE || type == IF_SCSI)
522 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
523 if (max_devs)
524 snprintf(dinfo->id, 32, "%s%i%s%i",
525 if_name[type], bus_id, mediastr, unit_id);
526 else
527 snprintf(dinfo->id, 32, "%s%s%i",
528 if_name[type], mediastr, unit_id);
529 }
530 dinfo->bdrv = bdrv_new(dinfo->id);
531 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
532 dinfo->bdrv->read_only = ro;
533 dinfo->devaddr = devaddr;
534 dinfo->type = type;
535 dinfo->bus = bus_id;
536 dinfo->unit = unit_id;
537 dinfo->cyls = cyls;
538 dinfo->heads = heads;
539 dinfo->secs = secs;
540 dinfo->trans = translation;
541 dinfo->opts = opts;
542 dinfo->refcount = 1;
543 dinfo->serial = serial;
544 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
545
546 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
547
548 /* disk I/O throttling */
549 bdrv_set_io_limits(dinfo->bdrv, &io_limits);
550
551 switch(type) {
552 case IF_IDE:
553 case IF_SCSI:
554 case IF_XEN:
555 case IF_NONE:
556 dinfo->media_cd = media == MEDIA_CDROM;
557 break;
558 case IF_SD:
559 case IF_FLOPPY:
560 case IF_PFLASH:
561 case IF_MTD:
562 break;
563 case IF_VIRTIO:
564 /* add virtio block device */
565 opts = qemu_opts_create(qemu_find_opts("device"), NULL, 0, NULL);
566 if (arch_type == QEMU_ARCH_S390X) {
567 qemu_opt_set(opts, "driver", "virtio-blk-s390");
568 } else {
569 qemu_opt_set(opts, "driver", "virtio-blk-pci");
570 }
571 qemu_opt_set(opts, "drive", dinfo->id);
572 if (devaddr)
573 qemu_opt_set(opts, "addr", devaddr);
574 break;
575 default:
576 abort();
577 }
578 if (!file || !*file) {
579 return dinfo;
580 }
581 if (snapshot) {
582 /* always use cache=unsafe with snapshot */
583 bdrv_flags &= ~BDRV_O_CACHE_MASK;
584 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
585 }
586
587 if (copy_on_read) {
588 bdrv_flags |= BDRV_O_COPY_ON_READ;
589 }
590
591 if (runstate_check(RUN_STATE_INMIGRATE)) {
592 bdrv_flags |= BDRV_O_INCOMING;
593 }
594
595 if (media == MEDIA_CDROM) {
596 /* CDROM is fine for any interface, don't check. */
597 ro = 1;
598 } else if (ro == 1) {
599 if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
600 type != IF_NONE && type != IF_PFLASH) {
601 error_report("readonly not supported by this bus type");
602 goto err;
603 }
604 }
605
606 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
607
608 if (ro && copy_on_read) {
609 error_report("warning: disabling copy_on_read on readonly drive");
610 }
611
612 ret = bdrv_open(dinfo->bdrv, file, bdrv_flags, drv);
613 if (ret < 0) {
614 error_report("could not open disk image %s: %s",
615 file, strerror(-ret));
616 goto err;
617 }
618
619 if (bdrv_key_required(dinfo->bdrv))
620 autostart = 0;
621 return dinfo;
622
623 err:
624 bdrv_delete(dinfo->bdrv);
625 g_free(dinfo->id);
626 QTAILQ_REMOVE(&drives, dinfo, next);
627 g_free(dinfo);
628 return NULL;
629 }
630
631 void do_commit(Monitor *mon, const QDict *qdict)
632 {
633 const char *device = qdict_get_str(qdict, "device");
634 BlockDriverState *bs;
635 int ret;
636
637 if (!strcmp(device, "all")) {
638 ret = bdrv_commit_all();
639 if (ret == -EBUSY) {
640 qerror_report(QERR_DEVICE_IN_USE, device);
641 return;
642 }
643 } else {
644 bs = bdrv_find(device);
645 if (!bs) {
646 qerror_report(QERR_DEVICE_NOT_FOUND, device);
647 return;
648 }
649 ret = bdrv_commit(bs);
650 if (ret == -EBUSY) {
651 qerror_report(QERR_DEVICE_IN_USE, device);
652 return;
653 }
654 }
655 }
656
657 static void blockdev_do_action(int kind, void *data, Error **errp)
658 {
659 BlockdevAction action;
660 BlockdevActionList list;
661
662 action.kind = kind;
663 action.data = data;
664 list.value = &action;
665 list.next = NULL;
666 qmp_transaction(&list, errp);
667 }
668
669 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
670 bool has_format, const char *format,
671 bool has_mode, enum NewImageMode mode,
672 Error **errp)
673 {
674 BlockdevSnapshot snapshot = {
675 .device = (char *) device,
676 .snapshot_file = (char *) snapshot_file,
677 .has_format = has_format,
678 .format = (char *) format,
679 .has_mode = has_mode,
680 .mode = mode,
681 };
682 blockdev_do_action(BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC, &snapshot,
683 errp);
684 }
685
686
687 /* New and old BlockDriverState structs for group snapshots */
688 typedef struct BlkTransactionStates {
689 BlockDriverState *old_bs;
690 BlockDriverState *new_bs;
691 QSIMPLEQ_ENTRY(BlkTransactionStates) entry;
692 } BlkTransactionStates;
693
694 /*
695 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
696 * then we do not pivot any of the devices in the group, and abandon the
697 * snapshots
698 */
699 void qmp_transaction(BlockdevActionList *dev_list, Error **errp)
700 {
701 int ret = 0;
702 BlockdevActionList *dev_entry = dev_list;
703 BlkTransactionStates *states, *next;
704
705 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionStates) snap_bdrv_states;
706 QSIMPLEQ_INIT(&snap_bdrv_states);
707
708 /* drain all i/o before any snapshots */
709 bdrv_drain_all();
710
711 /* We don't do anything in this loop that commits us to the snapshot */
712 while (NULL != dev_entry) {
713 BlockdevAction *dev_info = NULL;
714 BlockDriver *proto_drv;
715 BlockDriver *drv;
716 int flags;
717 enum NewImageMode mode;
718 const char *new_image_file;
719 const char *device;
720 const char *format = "qcow2";
721
722 dev_info = dev_entry->value;
723 dev_entry = dev_entry->next;
724
725 states = g_malloc0(sizeof(BlkTransactionStates));
726 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, states, entry);
727
728 switch (dev_info->kind) {
729 case BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
730 device = dev_info->blockdev_snapshot_sync->device;
731 if (!dev_info->blockdev_snapshot_sync->has_mode) {
732 dev_info->blockdev_snapshot_sync->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
733 }
734 new_image_file = dev_info->blockdev_snapshot_sync->snapshot_file;
735 if (dev_info->blockdev_snapshot_sync->has_format) {
736 format = dev_info->blockdev_snapshot_sync->format;
737 }
738 mode = dev_info->blockdev_snapshot_sync->mode;
739 break;
740 default:
741 abort();
742 }
743
744 drv = bdrv_find_format(format);
745 if (!drv) {
746 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
747 goto delete_and_fail;
748 }
749
750 states->old_bs = bdrv_find(device);
751 if (!states->old_bs) {
752 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
753 goto delete_and_fail;
754 }
755
756 if (!bdrv_is_inserted(states->old_bs)) {
757 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
758 goto delete_and_fail;
759 }
760
761 if (bdrv_in_use(states->old_bs)) {
762 error_set(errp, QERR_DEVICE_IN_USE, device);
763 goto delete_and_fail;
764 }
765
766 if (!bdrv_is_read_only(states->old_bs)) {
767 if (bdrv_flush(states->old_bs)) {
768 error_set(errp, QERR_IO_ERROR);
769 goto delete_and_fail;
770 }
771 }
772
773 flags = states->old_bs->open_flags;
774
775 proto_drv = bdrv_find_protocol(new_image_file);
776 if (!proto_drv) {
777 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
778 goto delete_and_fail;
779 }
780
781 /* create new image w/backing file */
782 if (mode != NEW_IMAGE_MODE_EXISTING) {
783 ret = bdrv_img_create(new_image_file, format,
784 states->old_bs->filename,
785 states->old_bs->drv->format_name,
786 NULL, -1, flags);
787 if (ret) {
788 error_set(errp, QERR_OPEN_FILE_FAILED, new_image_file);
789 goto delete_and_fail;
790 }
791 }
792
793 /* We will manually add the backing_hd field to the bs later */
794 states->new_bs = bdrv_new("");
795 ret = bdrv_open(states->new_bs, new_image_file,
796 flags | BDRV_O_NO_BACKING, drv);
797 if (ret != 0) {
798 error_set(errp, QERR_OPEN_FILE_FAILED, new_image_file);
799 goto delete_and_fail;
800 }
801 }
802
803
804 /* Now we are going to do the actual pivot. Everything up to this point
805 * is reversible, but we are committed at this point */
806 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
807 /* This removes our old bs from the bdrv_states, and adds the new bs */
808 bdrv_append(states->new_bs, states->old_bs);
809 /* We don't need (or want) to use the transactional
810 * bdrv_reopen_multiple() across all the entries at once, because we
811 * don't want to abort all of them if one of them fails the reopen */
812 bdrv_reopen(states->new_bs, states->new_bs->open_flags & ~BDRV_O_RDWR,
813 NULL);
814 }
815
816 /* success */
817 goto exit;
818
819 delete_and_fail:
820 /*
821 * failure, and it is all-or-none; abandon each new bs, and keep using
822 * the original bs for all images
823 */
824 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
825 if (states->new_bs) {
826 bdrv_delete(states->new_bs);
827 }
828 }
829 exit:
830 QSIMPLEQ_FOREACH_SAFE(states, &snap_bdrv_states, entry, next) {
831 g_free(states);
832 }
833 }
834
835
836 static void eject_device(BlockDriverState *bs, int force, Error **errp)
837 {
838 if (bdrv_in_use(bs)) {
839 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
840 return;
841 }
842 if (!bdrv_dev_has_removable_media(bs)) {
843 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
844 return;
845 }
846
847 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
848 bdrv_dev_eject_request(bs, force);
849 if (!force) {
850 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
851 return;
852 }
853 }
854
855 bdrv_close(bs);
856 }
857
858 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
859 {
860 BlockDriverState *bs;
861
862 bs = bdrv_find(device);
863 if (!bs) {
864 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
865 return;
866 }
867
868 eject_device(bs, force, errp);
869 }
870
871 void qmp_block_passwd(const char *device, const char *password, Error **errp)
872 {
873 BlockDriverState *bs;
874 int err;
875
876 bs = bdrv_find(device);
877 if (!bs) {
878 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
879 return;
880 }
881
882 err = bdrv_set_key(bs, password);
883 if (err == -EINVAL) {
884 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
885 return;
886 } else if (err < 0) {
887 error_set(errp, QERR_INVALID_PASSWORD);
888 return;
889 }
890 }
891
892 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
893 int bdrv_flags, BlockDriver *drv,
894 const char *password, Error **errp)
895 {
896 if (bdrv_open(bs, filename, bdrv_flags, drv) < 0) {
897 error_set(errp, QERR_OPEN_FILE_FAILED, filename);
898 return;
899 }
900
901 if (bdrv_key_required(bs)) {
902 if (password) {
903 if (bdrv_set_key(bs, password) < 0) {
904 error_set(errp, QERR_INVALID_PASSWORD);
905 }
906 } else {
907 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
908 bdrv_get_encrypted_filename(bs));
909 }
910 } else if (password) {
911 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
912 }
913 }
914
915 void qmp_change_blockdev(const char *device, const char *filename,
916 bool has_format, const char *format, Error **errp)
917 {
918 BlockDriverState *bs;
919 BlockDriver *drv = NULL;
920 int bdrv_flags;
921 Error *err = NULL;
922
923 bs = bdrv_find(device);
924 if (!bs) {
925 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
926 return;
927 }
928
929 if (format) {
930 drv = bdrv_find_whitelisted_format(format);
931 if (!drv) {
932 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
933 return;
934 }
935 }
936
937 eject_device(bs, 0, &err);
938 if (error_is_set(&err)) {
939 error_propagate(errp, err);
940 return;
941 }
942
943 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
944 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
945
946 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
947 }
948
949 /* throttling disk I/O limits */
950 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
951 int64_t bps_wr, int64_t iops, int64_t iops_rd,
952 int64_t iops_wr, Error **errp)
953 {
954 BlockIOLimit io_limits;
955 BlockDriverState *bs;
956
957 bs = bdrv_find(device);
958 if (!bs) {
959 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
960 return;
961 }
962
963 io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
964 io_limits.bps[BLOCK_IO_LIMIT_READ] = bps_rd;
965 io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
966 io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
967 io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
968 io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
969
970 if (!do_check_io_limits(&io_limits)) {
971 error_set(errp, QERR_INVALID_PARAMETER_COMBINATION);
972 return;
973 }
974
975 bs->io_limits = io_limits;
976 bs->slice_time = BLOCK_IO_SLICE_TIME;
977
978 if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
979 bdrv_io_limits_enable(bs);
980 } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
981 bdrv_io_limits_disable(bs);
982 } else {
983 if (bs->block_timer) {
984 qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
985 }
986 }
987 }
988
989 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
990 {
991 const char *id = qdict_get_str(qdict, "id");
992 BlockDriverState *bs;
993
994 bs = bdrv_find(id);
995 if (!bs) {
996 qerror_report(QERR_DEVICE_NOT_FOUND, id);
997 return -1;
998 }
999 if (bdrv_in_use(bs)) {
1000 qerror_report(QERR_DEVICE_IN_USE, id);
1001 return -1;
1002 }
1003
1004 /* quiesce block driver; prevent further io */
1005 bdrv_drain_all();
1006 bdrv_flush(bs);
1007 bdrv_close(bs);
1008
1009 /* if we have a device attached to this BlockDriverState
1010 * then we need to make the drive anonymous until the device
1011 * can be removed. If this is a drive with no device backing
1012 * then we can just get rid of the block driver state right here.
1013 */
1014 if (bdrv_get_attached_dev(bs)) {
1015 bdrv_make_anon(bs);
1016 } else {
1017 drive_uninit(drive_get_by_blockdev(bs));
1018 }
1019
1020 return 0;
1021 }
1022
1023 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1024 {
1025 BlockDriverState *bs;
1026
1027 bs = bdrv_find(device);
1028 if (!bs) {
1029 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1030 return;
1031 }
1032
1033 if (size < 0) {
1034 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1035 return;
1036 }
1037
1038 switch (bdrv_truncate(bs, size)) {
1039 case 0:
1040 break;
1041 case -ENOMEDIUM:
1042 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1043 break;
1044 case -ENOTSUP:
1045 error_set(errp, QERR_UNSUPPORTED);
1046 break;
1047 case -EACCES:
1048 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1049 break;
1050 case -EBUSY:
1051 error_set(errp, QERR_DEVICE_IN_USE, device);
1052 break;
1053 default:
1054 error_set(errp, QERR_UNDEFINED_ERROR);
1055 break;
1056 }
1057 }
1058
1059 static void block_job_cb(void *opaque, int ret)
1060 {
1061 BlockDriverState *bs = opaque;
1062 QObject *obj;
1063
1064 trace_block_job_cb(bs, bs->job, ret);
1065
1066 assert(bs->job);
1067 obj = qobject_from_block_job(bs->job);
1068 if (ret < 0) {
1069 QDict *dict = qobject_to_qdict(obj);
1070 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1071 }
1072
1073 if (block_job_is_cancelled(bs->job)) {
1074 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1075 } else {
1076 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1077 }
1078 qobject_decref(obj);
1079
1080 drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1081 }
1082
1083 void qmp_block_stream(const char *device, bool has_base,
1084 const char *base, bool has_speed, int64_t speed,
1085 bool has_on_error, BlockdevOnError on_error,
1086 Error **errp)
1087 {
1088 BlockDriverState *bs;
1089 BlockDriverState *base_bs = NULL;
1090 Error *local_err = NULL;
1091
1092 if (!has_on_error) {
1093 on_error = BLOCKDEV_ON_ERROR_REPORT;
1094 }
1095
1096 bs = bdrv_find(device);
1097 if (!bs) {
1098 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1099 return;
1100 }
1101
1102 if (base) {
1103 base_bs = bdrv_find_backing_image(bs, base);
1104 if (base_bs == NULL) {
1105 error_set(errp, QERR_BASE_NOT_FOUND, base);
1106 return;
1107 }
1108 }
1109
1110 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1111 on_error, block_job_cb, bs, &local_err);
1112 if (error_is_set(&local_err)) {
1113 error_propagate(errp, local_err);
1114 return;
1115 }
1116
1117 /* Grab a reference so hotplug does not delete the BlockDriverState from
1118 * underneath us.
1119 */
1120 drive_get_ref(drive_get_by_blockdev(bs));
1121
1122 trace_qmp_block_stream(bs, bs->job);
1123 }
1124
1125 void qmp_block_commit(const char *device,
1126 bool has_base, const char *base, const char *top,
1127 bool has_speed, int64_t speed,
1128 Error **errp)
1129 {
1130 BlockDriverState *bs;
1131 BlockDriverState *base_bs, *top_bs;
1132 Error *local_err = NULL;
1133 /* This will be part of the QMP command, if/when the
1134 * BlockdevOnError change for blkmirror makes it in
1135 */
1136 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1137
1138 /* drain all i/o before commits */
1139 bdrv_drain_all();
1140
1141 bs = bdrv_find(device);
1142 if (!bs) {
1143 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1144 return;
1145 }
1146
1147 /* default top_bs is the active layer */
1148 top_bs = bs;
1149
1150 if (top) {
1151 if (strcmp(bs->filename, top) != 0) {
1152 top_bs = bdrv_find_backing_image(bs, top);
1153 }
1154 }
1155
1156 if (top_bs == NULL) {
1157 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1158 return;
1159 }
1160
1161 if (has_base && base) {
1162 base_bs = bdrv_find_backing_image(top_bs, base);
1163 } else {
1164 base_bs = bdrv_find_base(top_bs);
1165 }
1166
1167 if (base_bs == NULL) {
1168 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1169 return;
1170 }
1171
1172 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1173 &local_err);
1174 if (local_err != NULL) {
1175 error_propagate(errp, local_err);
1176 return;
1177 }
1178 /* Grab a reference so hotplug does not delete the BlockDriverState from
1179 * underneath us.
1180 */
1181 drive_get_ref(drive_get_by_blockdev(bs));
1182 }
1183
1184 void qmp_drive_mirror(const char *device, const char *target,
1185 bool has_format, const char *format,
1186 enum MirrorSyncMode sync,
1187 bool has_mode, enum NewImageMode mode,
1188 bool has_speed, int64_t speed, Error **errp)
1189 {
1190 BlockDriverInfo bdi;
1191 BlockDriverState *bs;
1192 BlockDriverState *source, *target_bs;
1193 BlockDriver *proto_drv;
1194 BlockDriver *drv = NULL;
1195 Error *local_err = NULL;
1196 int flags;
1197 uint64_t size;
1198 int ret;
1199
1200 if (!has_speed) {
1201 speed = 0;
1202 }
1203 if (!has_mode) {
1204 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1205 }
1206
1207 bs = bdrv_find(device);
1208 if (!bs) {
1209 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1210 return;
1211 }
1212
1213 if (!bdrv_is_inserted(bs)) {
1214 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1215 return;
1216 }
1217
1218 if (!has_format) {
1219 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1220 }
1221 if (format) {
1222 drv = bdrv_find_format(format);
1223 if (!drv) {
1224 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1225 return;
1226 }
1227 }
1228
1229 if (bdrv_in_use(bs)) {
1230 error_set(errp, QERR_DEVICE_IN_USE, device);
1231 return;
1232 }
1233
1234 flags = bs->open_flags | BDRV_O_RDWR;
1235 source = bs->backing_hd;
1236 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1237 sync = MIRROR_SYNC_MODE_FULL;
1238 }
1239
1240 proto_drv = bdrv_find_protocol(target);
1241 if (!proto_drv) {
1242 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1243 return;
1244 }
1245
1246 if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1247 /* create new image w/o backing file */
1248 assert(format && drv);
1249 bdrv_get_geometry(bs, &size);
1250 size *= 512;
1251 ret = bdrv_img_create(target, format,
1252 NULL, NULL, NULL, size, flags);
1253 } else {
1254 switch (mode) {
1255 case NEW_IMAGE_MODE_EXISTING:
1256 ret = 0;
1257 break;
1258 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1259 /* create new image with backing file */
1260 ret = bdrv_img_create(target, format,
1261 source->filename,
1262 source->drv->format_name,
1263 NULL, -1, flags);
1264 break;
1265 default:
1266 abort();
1267 }
1268 }
1269
1270 if (ret) {
1271 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1272 return;
1273 }
1274
1275 target_bs = bdrv_new("");
1276 ret = bdrv_open(target_bs, target, flags | BDRV_O_NO_BACKING, drv);
1277
1278 if (ret < 0) {
1279 bdrv_delete(target_bs);
1280 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1281 return;
1282 }
1283
1284 /* We need a backing file if we will copy parts of a cluster. */
1285 if (bdrv_get_info(target_bs, &bdi) >= 0 && bdi.cluster_size != 0 &&
1286 bdi.cluster_size >= BDRV_SECTORS_PER_DIRTY_CHUNK * 512) {
1287 ret = bdrv_open_backing_file(target_bs);
1288 if (ret < 0) {
1289 bdrv_delete(target_bs);
1290 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1291 return;
1292 }
1293 }
1294
1295 mirror_start(bs, target_bs, speed, sync, block_job_cb, bs, &local_err);
1296 if (local_err != NULL) {
1297 bdrv_delete(target_bs);
1298 error_propagate(errp, local_err);
1299 return;
1300 }
1301
1302 /* Grab a reference so hotplug does not delete the BlockDriverState from
1303 * underneath us.
1304 */
1305 drive_get_ref(drive_get_by_blockdev(bs));
1306 }
1307
1308 static BlockJob *find_block_job(const char *device)
1309 {
1310 BlockDriverState *bs;
1311
1312 bs = bdrv_find(device);
1313 if (!bs || !bs->job) {
1314 return NULL;
1315 }
1316 return bs->job;
1317 }
1318
1319 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1320 {
1321 BlockJob *job = find_block_job(device);
1322
1323 if (!job) {
1324 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1325 return;
1326 }
1327
1328 block_job_set_speed(job, speed, errp);
1329 }
1330
1331 void qmp_block_job_cancel(const char *device,
1332 bool has_force, bool force, Error **errp)
1333 {
1334 BlockJob *job = find_block_job(device);
1335
1336 if (!has_force) {
1337 force = false;
1338 }
1339
1340 if (!job) {
1341 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1342 return;
1343 }
1344 if (job->paused && !force) {
1345 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1346 return;
1347 }
1348
1349 trace_qmp_block_job_cancel(job);
1350 block_job_cancel(job);
1351 }
1352
1353 void qmp_block_job_pause(const char *device, Error **errp)
1354 {
1355 BlockJob *job = find_block_job(device);
1356
1357 if (!job) {
1358 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1359 return;
1360 }
1361
1362 trace_qmp_block_job_pause(job);
1363 block_job_pause(job);
1364 }
1365
1366 void qmp_block_job_resume(const char *device, Error **errp)
1367 {
1368 BlockJob *job = find_block_job(device);
1369
1370 if (!job) {
1371 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1372 return;
1373 }
1374
1375 trace_qmp_block_job_resume(job);
1376 block_job_resume(job);
1377 }
1378
1379 void qmp_block_job_complete(const char *device, Error **errp)
1380 {
1381 BlockJob *job = find_block_job(device);
1382
1383 if (!job) {
1384 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1385 return;
1386 }
1387
1388 trace_qmp_block_job_complete(job);
1389 block_job_complete(job, errp);
1390 }
1391
1392 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1393 {
1394 BlockJobInfoList **prev = opaque;
1395 BlockJob *job = bs->job;
1396
1397 if (job) {
1398 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1399 elem->value = block_job_query(bs->job);
1400 (*prev)->next = elem;
1401 *prev = elem;
1402 }
1403 }
1404
1405 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1406 {
1407 /* Dummy is a fake list element for holding the head pointer */
1408 BlockJobInfoList dummy = {};
1409 BlockJobInfoList *prev = &dummy;
1410 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1411 return dummy.next;
1412 }