block: Clamp BlockBackend requests
[qemu/ar7.git] / blockdev.c
blob6eedcf56f89da45a1fcb9b2f1076048bbed44dd7
1 /*
2 * QEMU host block devices
4 * Copyright (c) 2003-2008 Fabrice Bellard
6 * This work is licensed under the terms of the GNU GPL, version 2 or
7 * later. See the COPYING file in the top-level directory.
9 * This file incorporates work covered by the following copyright and
10 * permission notice:
12 * Copyright (c) 2003-2008 Fabrice Bellard
14 * Permission is hereby granted, free of charge, to any person obtaining a copy
15 * of this software and associated documentation files (the "Software"), to deal
16 * in the Software without restriction, including without limitation the rights
17 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
18 * copies of the Software, and to permit persons to whom the Software is
19 * furnished to do so, subject to the following conditions:
21 * The above copyright notice and this permission notice shall be included in
22 * all copies or substantial portions of the Software.
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
27 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
30 * THE SOFTWARE.
33 #include "sysemu/block-backend.h"
34 #include "sysemu/blockdev.h"
35 #include "hw/block/block.h"
36 #include "block/blockjob.h"
37 #include "monitor/monitor.h"
38 #include "qemu/option.h"
39 #include "qemu/config-file.h"
40 #include "qapi/qmp/types.h"
41 #include "qapi-visit.h"
42 #include "qapi/qmp-output-visitor.h"
43 #include "qapi/util.h"
44 #include "sysemu/sysemu.h"
45 #include "block/block_int.h"
46 #include "qmp-commands.h"
47 #include "trace.h"
48 #include "sysemu/arch_init.h"
50 static const char *const if_name[IF_COUNT] = {
51 [IF_NONE] = "none",
52 [IF_IDE] = "ide",
53 [IF_SCSI] = "scsi",
54 [IF_FLOPPY] = "floppy",
55 [IF_PFLASH] = "pflash",
56 [IF_MTD] = "mtd",
57 [IF_SD] = "sd",
58 [IF_VIRTIO] = "virtio",
59 [IF_XEN] = "xen",
62 static int if_max_devs[IF_COUNT] = {
64 * Do not change these numbers! They govern how drive option
65 * index maps to unit and bus. That mapping is ABI.
67 * All controllers used to imlement if=T drives need to support
68 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
69 * Otherwise, some index values map to "impossible" bus, unit
70 * values.
72 * For instance, if you change [IF_SCSI] to 255, -drive
73 * if=scsi,index=12 no longer means bus=1,unit=5, but
74 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
75 * the drive can't be set up. Regression.
77 [IF_IDE] = 2,
78 [IF_SCSI] = 7,
81 /**
82 * Boards may call this to offer board-by-board overrides
83 * of the default, global values.
85 void override_max_devs(BlockInterfaceType type, int max_devs)
87 BlockBackend *blk;
88 DriveInfo *dinfo;
90 if (max_devs <= 0) {
91 return;
94 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
95 dinfo = blk_legacy_dinfo(blk);
96 if (dinfo->type == type) {
97 fprintf(stderr, "Cannot override units-per-bus property of"
98 " the %s interface, because a drive of that type has"
99 " already been added.\n", if_name[type]);
100 g_assert_not_reached();
104 if_max_devs[type] = max_devs;
108 * We automatically delete the drive when a device using it gets
109 * unplugged. Questionable feature, but we can't just drop it.
110 * Device models call blockdev_mark_auto_del() to schedule the
111 * automatic deletion, and generic qdev code calls blockdev_auto_del()
112 * when deletion is actually safe.
114 void blockdev_mark_auto_del(BlockBackend *blk)
116 DriveInfo *dinfo = blk_legacy_dinfo(blk);
117 BlockDriverState *bs = blk_bs(blk);
118 AioContext *aio_context;
120 if (!dinfo) {
121 return;
124 aio_context = bdrv_get_aio_context(bs);
125 aio_context_acquire(aio_context);
127 if (bs->job) {
128 block_job_cancel(bs->job);
131 aio_context_release(aio_context);
133 dinfo->auto_del = 1;
136 void blockdev_auto_del(BlockBackend *blk)
138 DriveInfo *dinfo = blk_legacy_dinfo(blk);
140 if (dinfo && dinfo->auto_del) {
141 blk_unref(blk);
146 * Returns the current mapping of how many units per bus
147 * a particular interface can support.
149 * A positive integer indicates n units per bus.
150 * 0 implies the mapping has not been established.
151 * -1 indicates an invalid BlockInterfaceType was given.
153 int drive_get_max_devs(BlockInterfaceType type)
155 if (type >= IF_IDE && type < IF_COUNT) {
156 return if_max_devs[type];
159 return -1;
162 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
164 int max_devs = if_max_devs[type];
165 return max_devs ? index / max_devs : 0;
168 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
170 int max_devs = if_max_devs[type];
171 return max_devs ? index % max_devs : index;
174 QemuOpts *drive_def(const char *optstr)
176 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
179 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
180 const char *optstr)
182 QemuOpts *opts;
183 char buf[32];
185 opts = drive_def(optstr);
186 if (!opts) {
187 return NULL;
189 if (type != IF_DEFAULT) {
190 qemu_opt_set(opts, "if", if_name[type]);
192 if (index >= 0) {
193 snprintf(buf, sizeof(buf), "%d", index);
194 qemu_opt_set(opts, "index", buf);
196 if (file)
197 qemu_opt_set(opts, "file", file);
198 return opts;
201 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
203 BlockBackend *blk;
204 DriveInfo *dinfo;
206 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
207 dinfo = blk_legacy_dinfo(blk);
208 if (dinfo && dinfo->type == type
209 && dinfo->bus == bus && dinfo->unit == unit) {
210 return dinfo;
214 return NULL;
217 bool drive_check_orphaned(void)
219 BlockBackend *blk;
220 DriveInfo *dinfo;
221 bool rs = false;
223 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
224 dinfo = blk_legacy_dinfo(blk);
225 /* If dinfo->bdrv->dev is NULL, it has no device attached. */
226 /* Unless this is a default drive, this may be an oversight. */
227 if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
228 dinfo->type != IF_NONE) {
229 fprintf(stderr, "Warning: Orphaned drive without device: "
230 "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
231 blk_name(blk), blk_bs(blk)->filename, if_name[dinfo->type],
232 dinfo->bus, dinfo->unit);
233 rs = true;
237 return rs;
240 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
242 return drive_get(type,
243 drive_index_to_bus_id(type, index),
244 drive_index_to_unit_id(type, index));
247 int drive_get_max_bus(BlockInterfaceType type)
249 int max_bus;
250 BlockBackend *blk;
251 DriveInfo *dinfo;
253 max_bus = -1;
254 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
255 dinfo = blk_legacy_dinfo(blk);
256 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
257 max_bus = dinfo->bus;
260 return max_bus;
263 /* Get a block device. This should only be used for single-drive devices
264 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
265 appropriate bus. */
266 DriveInfo *drive_get_next(BlockInterfaceType type)
268 static int next_block_unit[IF_COUNT];
270 return drive_get(type, 0, next_block_unit[type]++);
273 static void bdrv_format_print(void *opaque, const char *name)
275 error_printf(" %s", name);
278 typedef struct {
279 QEMUBH *bh;
280 BlockDriverState *bs;
281 } BDRVPutRefBH;
283 static void bdrv_put_ref_bh(void *opaque)
285 BDRVPutRefBH *s = opaque;
287 bdrv_unref(s->bs);
288 qemu_bh_delete(s->bh);
289 g_free(s);
293 * Release a BDS reference in a BH
295 * It is not safe to use bdrv_unref() from a callback function when the callers
296 * still need the BlockDriverState. In such cases we schedule a BH to release
297 * the reference.
299 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
301 BDRVPutRefBH *s;
303 s = g_new(BDRVPutRefBH, 1);
304 s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
305 s->bs = bs;
306 qemu_bh_schedule(s->bh);
309 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
311 if (!strcmp(buf, "ignore")) {
312 return BLOCKDEV_ON_ERROR_IGNORE;
313 } else if (!is_read && !strcmp(buf, "enospc")) {
314 return BLOCKDEV_ON_ERROR_ENOSPC;
315 } else if (!strcmp(buf, "stop")) {
316 return BLOCKDEV_ON_ERROR_STOP;
317 } else if (!strcmp(buf, "report")) {
318 return BLOCKDEV_ON_ERROR_REPORT;
319 } else {
320 error_setg(errp, "'%s' invalid %s error action",
321 buf, is_read ? "read" : "write");
322 return -1;
326 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
328 if (throttle_conflicting(cfg)) {
329 error_setg(errp, "bps/iops/max total values and read/write values"
330 " cannot be used at the same time");
331 return false;
334 if (!throttle_is_valid(cfg)) {
335 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
336 return false;
339 return true;
342 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
344 /* Takes the ownership of bs_opts */
345 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
346 Error **errp)
348 const char *buf;
349 int ro = 0;
350 int bdrv_flags = 0;
351 int on_read_error, on_write_error;
352 BlockBackend *blk;
353 BlockDriverState *bs;
354 ThrottleConfig cfg;
355 int snapshot = 0;
356 bool copy_on_read;
357 Error *error = NULL;
358 QemuOpts *opts;
359 const char *id;
360 bool has_driver_specific_opts;
361 BlockdevDetectZeroesOptions detect_zeroes;
363 /* Check common options by copying from bs_opts to opts, all other options
364 * stay in bs_opts for processing by bdrv_open(). */
365 id = qdict_get_try_str(bs_opts, "id");
366 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
367 if (error) {
368 error_propagate(errp, error);
369 goto err_no_opts;
372 qemu_opts_absorb_qdict(opts, bs_opts, &error);
373 if (error) {
374 error_propagate(errp, error);
375 goto early_err;
378 if (id) {
379 qdict_del(bs_opts, "id");
382 has_driver_specific_opts = !!qdict_size(bs_opts);
384 /* extract parameters */
385 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
386 ro = qemu_opt_get_bool(opts, "read-only", 0);
387 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
389 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
390 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
391 error_setg(errp, "invalid discard option");
392 goto early_err;
396 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
397 bdrv_flags |= BDRV_O_CACHE_WB;
399 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
400 bdrv_flags |= BDRV_O_NOCACHE;
402 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
403 bdrv_flags |= BDRV_O_NO_FLUSH;
406 #ifdef CONFIG_LINUX_AIO
407 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
408 if (!strcmp(buf, "native")) {
409 bdrv_flags |= BDRV_O_NATIVE_AIO;
410 } else if (!strcmp(buf, "threads")) {
411 /* this is the default */
412 } else {
413 error_setg(errp, "invalid aio option");
414 goto early_err;
417 #endif
419 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
420 if (is_help_option(buf)) {
421 error_printf("Supported formats:");
422 bdrv_iterate_format(bdrv_format_print, NULL);
423 error_printf("\n");
424 goto early_err;
427 if (qdict_haskey(bs_opts, "driver")) {
428 error_setg(errp, "Cannot specify both 'driver' and 'format'");
429 goto early_err;
431 qdict_put(bs_opts, "driver", qstring_from_str(buf));
434 /* disk I/O throttling */
435 memset(&cfg, 0, sizeof(cfg));
436 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
437 qemu_opt_get_number(opts, "throttling.bps-total", 0);
438 cfg.buckets[THROTTLE_BPS_READ].avg =
439 qemu_opt_get_number(opts, "throttling.bps-read", 0);
440 cfg.buckets[THROTTLE_BPS_WRITE].avg =
441 qemu_opt_get_number(opts, "throttling.bps-write", 0);
442 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
443 qemu_opt_get_number(opts, "throttling.iops-total", 0);
444 cfg.buckets[THROTTLE_OPS_READ].avg =
445 qemu_opt_get_number(opts, "throttling.iops-read", 0);
446 cfg.buckets[THROTTLE_OPS_WRITE].avg =
447 qemu_opt_get_number(opts, "throttling.iops-write", 0);
449 cfg.buckets[THROTTLE_BPS_TOTAL].max =
450 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
451 cfg.buckets[THROTTLE_BPS_READ].max =
452 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
453 cfg.buckets[THROTTLE_BPS_WRITE].max =
454 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
455 cfg.buckets[THROTTLE_OPS_TOTAL].max =
456 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
457 cfg.buckets[THROTTLE_OPS_READ].max =
458 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
459 cfg.buckets[THROTTLE_OPS_WRITE].max =
460 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
462 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
464 if (!check_throttle_config(&cfg, &error)) {
465 error_propagate(errp, error);
466 goto early_err;
469 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
470 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
471 on_write_error = parse_block_error_action(buf, 0, &error);
472 if (error) {
473 error_propagate(errp, error);
474 goto early_err;
478 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
479 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
480 on_read_error = parse_block_error_action(buf, 1, &error);
481 if (error) {
482 error_propagate(errp, error);
483 goto early_err;
487 detect_zeroes =
488 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
489 qemu_opt_get(opts, "detect-zeroes"),
490 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
491 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
492 &error);
493 if (error) {
494 error_propagate(errp, error);
495 goto early_err;
498 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
499 !(bdrv_flags & BDRV_O_UNMAP)) {
500 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
501 "without setting discard operation to unmap");
502 goto early_err;
505 /* init */
506 if ((!file || !*file) && !has_driver_specific_opts) {
507 blk = blk_new_with_bs(qemu_opts_id(opts), errp);
508 if (!blk) {
509 goto early_err;
512 bs = blk_bs(blk);
513 bs->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
514 bs->read_only = ro;
516 QDECREF(bs_opts);
517 } else {
518 if (file && !*file) {
519 file = NULL;
522 if (snapshot) {
523 /* always use cache=unsafe with snapshot */
524 bdrv_flags &= ~BDRV_O_CACHE_MASK;
525 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
528 if (copy_on_read) {
529 bdrv_flags |= BDRV_O_COPY_ON_READ;
532 if (runstate_check(RUN_STATE_INMIGRATE)) {
533 bdrv_flags |= BDRV_O_INCOMING;
536 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
538 blk = blk_new_open(qemu_opts_id(opts), file, NULL, bs_opts, bdrv_flags,
539 errp);
540 if (!blk) {
541 goto err_no_bs_opts;
543 bs = blk_bs(blk);
546 bs->detect_zeroes = detect_zeroes;
548 bdrv_set_on_error(bs, on_read_error, on_write_error);
550 /* disk I/O throttling */
551 if (throttle_enabled(&cfg)) {
552 bdrv_io_limits_enable(bs);
553 bdrv_set_io_limits(bs, &cfg);
556 if (bdrv_key_required(bs)) {
557 autostart = 0;
560 err_no_bs_opts:
561 qemu_opts_del(opts);
562 return blk;
564 early_err:
565 qemu_opts_del(opts);
566 err_no_opts:
567 QDECREF(bs_opts);
568 return NULL;
571 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
572 Error **errp)
574 const char *value;
576 value = qemu_opt_get(opts, from);
577 if (value) {
578 if (qemu_opt_find(opts, to)) {
579 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
580 "same time", to, from);
581 return;
585 /* rename all items in opts */
586 while ((value = qemu_opt_get(opts, from))) {
587 qemu_opt_set(opts, to, value);
588 qemu_opt_unset(opts, from);
592 QemuOptsList qemu_legacy_drive_opts = {
593 .name = "drive",
594 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
595 .desc = {
597 .name = "bus",
598 .type = QEMU_OPT_NUMBER,
599 .help = "bus number",
601 .name = "unit",
602 .type = QEMU_OPT_NUMBER,
603 .help = "unit number (i.e. lun for scsi)",
605 .name = "index",
606 .type = QEMU_OPT_NUMBER,
607 .help = "index number",
609 .name = "media",
610 .type = QEMU_OPT_STRING,
611 .help = "media type (disk, cdrom)",
613 .name = "if",
614 .type = QEMU_OPT_STRING,
615 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
617 .name = "cyls",
618 .type = QEMU_OPT_NUMBER,
619 .help = "number of cylinders (ide disk geometry)",
621 .name = "heads",
622 .type = QEMU_OPT_NUMBER,
623 .help = "number of heads (ide disk geometry)",
625 .name = "secs",
626 .type = QEMU_OPT_NUMBER,
627 .help = "number of sectors (ide disk geometry)",
629 .name = "trans",
630 .type = QEMU_OPT_STRING,
631 .help = "chs translation (auto, lba, none)",
633 .name = "boot",
634 .type = QEMU_OPT_BOOL,
635 .help = "(deprecated, ignored)",
637 .name = "addr",
638 .type = QEMU_OPT_STRING,
639 .help = "pci address (virtio only)",
641 .name = "serial",
642 .type = QEMU_OPT_STRING,
643 .help = "disk serial number",
645 .name = "file",
646 .type = QEMU_OPT_STRING,
647 .help = "file name",
650 /* Options that are passed on, but have special semantics with -drive */
652 .name = "read-only",
653 .type = QEMU_OPT_BOOL,
654 .help = "open drive file as read-only",
656 .name = "rerror",
657 .type = QEMU_OPT_STRING,
658 .help = "read error action",
660 .name = "werror",
661 .type = QEMU_OPT_STRING,
662 .help = "write error action",
664 .name = "copy-on-read",
665 .type = QEMU_OPT_BOOL,
666 .help = "copy read data from backing file into image file",
669 { /* end of list */ }
673 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
675 const char *value;
676 BlockBackend *blk;
677 DriveInfo *dinfo = NULL;
678 QDict *bs_opts;
679 QemuOpts *legacy_opts;
680 DriveMediaType media = MEDIA_DISK;
681 BlockInterfaceType type;
682 int cyls, heads, secs, translation;
683 int max_devs, bus_id, unit_id, index;
684 const char *devaddr;
685 const char *werror, *rerror;
686 bool read_only = false;
687 bool copy_on_read;
688 const char *serial;
689 const char *filename;
690 Error *local_err = NULL;
691 int i;
693 /* Change legacy command line options into QMP ones */
694 static const struct {
695 const char *from;
696 const char *to;
697 } opt_renames[] = {
698 { "iops", "throttling.iops-total" },
699 { "iops_rd", "throttling.iops-read" },
700 { "iops_wr", "throttling.iops-write" },
702 { "bps", "throttling.bps-total" },
703 { "bps_rd", "throttling.bps-read" },
704 { "bps_wr", "throttling.bps-write" },
706 { "iops_max", "throttling.iops-total-max" },
707 { "iops_rd_max", "throttling.iops-read-max" },
708 { "iops_wr_max", "throttling.iops-write-max" },
710 { "bps_max", "throttling.bps-total-max" },
711 { "bps_rd_max", "throttling.bps-read-max" },
712 { "bps_wr_max", "throttling.bps-write-max" },
714 { "iops_size", "throttling.iops-size" },
716 { "readonly", "read-only" },
719 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
720 qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
721 &local_err);
722 if (local_err) {
723 error_report("%s", error_get_pretty(local_err));
724 error_free(local_err);
725 return NULL;
729 value = qemu_opt_get(all_opts, "cache");
730 if (value) {
731 int flags = 0;
733 if (bdrv_parse_cache_flags(value, &flags) != 0) {
734 error_report("invalid cache option");
735 return NULL;
738 /* Specific options take precedence */
739 if (!qemu_opt_get(all_opts, "cache.writeback")) {
740 qemu_opt_set_bool(all_opts, "cache.writeback",
741 !!(flags & BDRV_O_CACHE_WB));
743 if (!qemu_opt_get(all_opts, "cache.direct")) {
744 qemu_opt_set_bool(all_opts, "cache.direct",
745 !!(flags & BDRV_O_NOCACHE));
747 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
748 qemu_opt_set_bool(all_opts, "cache.no-flush",
749 !!(flags & BDRV_O_NO_FLUSH));
751 qemu_opt_unset(all_opts, "cache");
754 /* Get a QDict for processing the options */
755 bs_opts = qdict_new();
756 qemu_opts_to_qdict(all_opts, bs_opts);
758 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
759 &error_abort);
760 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
761 if (local_err) {
762 error_report("%s", error_get_pretty(local_err));
763 error_free(local_err);
764 goto fail;
767 /* Deprecated option boot=[on|off] */
768 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
769 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
770 "ignored. Future versions will reject this parameter. Please "
771 "update your scripts.\n");
774 /* Media type */
775 value = qemu_opt_get(legacy_opts, "media");
776 if (value) {
777 if (!strcmp(value, "disk")) {
778 media = MEDIA_DISK;
779 } else if (!strcmp(value, "cdrom")) {
780 media = MEDIA_CDROM;
781 read_only = true;
782 } else {
783 error_report("'%s' invalid media", value);
784 goto fail;
788 /* copy-on-read is disabled with a warning for read-only devices */
789 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
790 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
792 if (read_only && copy_on_read) {
793 error_report("warning: disabling copy-on-read on read-only drive");
794 copy_on_read = false;
797 qdict_put(bs_opts, "read-only",
798 qstring_from_str(read_only ? "on" : "off"));
799 qdict_put(bs_opts, "copy-on-read",
800 qstring_from_str(copy_on_read ? "on" :"off"));
802 /* Controller type */
803 value = qemu_opt_get(legacy_opts, "if");
804 if (value) {
805 for (type = 0;
806 type < IF_COUNT && strcmp(value, if_name[type]);
807 type++) {
809 if (type == IF_COUNT) {
810 error_report("unsupported bus type '%s'", value);
811 goto fail;
813 } else {
814 type = block_default_type;
817 /* Geometry */
818 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
819 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
820 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
822 if (cyls || heads || secs) {
823 if (cyls < 1) {
824 error_report("invalid physical cyls number");
825 goto fail;
827 if (heads < 1) {
828 error_report("invalid physical heads number");
829 goto fail;
831 if (secs < 1) {
832 error_report("invalid physical secs number");
833 goto fail;
837 translation = BIOS_ATA_TRANSLATION_AUTO;
838 value = qemu_opt_get(legacy_opts, "trans");
839 if (value != NULL) {
840 if (!cyls) {
841 error_report("'%s' trans must be used with cyls, heads and secs",
842 value);
843 goto fail;
845 if (!strcmp(value, "none")) {
846 translation = BIOS_ATA_TRANSLATION_NONE;
847 } else if (!strcmp(value, "lba")) {
848 translation = BIOS_ATA_TRANSLATION_LBA;
849 } else if (!strcmp(value, "large")) {
850 translation = BIOS_ATA_TRANSLATION_LARGE;
851 } else if (!strcmp(value, "rechs")) {
852 translation = BIOS_ATA_TRANSLATION_RECHS;
853 } else if (!strcmp(value, "auto")) {
854 translation = BIOS_ATA_TRANSLATION_AUTO;
855 } else {
856 error_report("'%s' invalid translation type", value);
857 goto fail;
861 if (media == MEDIA_CDROM) {
862 if (cyls || secs || heads) {
863 error_report("CHS can't be set with media=cdrom");
864 goto fail;
868 /* Device address specified by bus/unit or index.
869 * If none was specified, try to find the first free one. */
870 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
871 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
872 index = qemu_opt_get_number(legacy_opts, "index", -1);
874 max_devs = if_max_devs[type];
876 if (index != -1) {
877 if (bus_id != 0 || unit_id != -1) {
878 error_report("index cannot be used with bus and unit");
879 goto fail;
881 bus_id = drive_index_to_bus_id(type, index);
882 unit_id = drive_index_to_unit_id(type, index);
885 if (unit_id == -1) {
886 unit_id = 0;
887 while (drive_get(type, bus_id, unit_id) != NULL) {
888 unit_id++;
889 if (max_devs && unit_id >= max_devs) {
890 unit_id -= max_devs;
891 bus_id++;
896 if (max_devs && unit_id >= max_devs) {
897 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
898 goto fail;
901 if (drive_get(type, bus_id, unit_id) != NULL) {
902 error_report("drive with bus=%d, unit=%d (index=%d) exists",
903 bus_id, unit_id, index);
904 goto fail;
907 /* Serial number */
908 serial = qemu_opt_get(legacy_opts, "serial");
910 /* no id supplied -> create one */
911 if (qemu_opts_id(all_opts) == NULL) {
912 char *new_id;
913 const char *mediastr = "";
914 if (type == IF_IDE || type == IF_SCSI) {
915 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
917 if (max_devs) {
918 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
919 mediastr, unit_id);
920 } else {
921 new_id = g_strdup_printf("%s%s%i", if_name[type],
922 mediastr, unit_id);
924 qdict_put(bs_opts, "id", qstring_from_str(new_id));
925 g_free(new_id);
928 /* Add virtio block device */
929 devaddr = qemu_opt_get(legacy_opts, "addr");
930 if (devaddr && type != IF_VIRTIO) {
931 error_report("addr is not supported by this bus type");
932 goto fail;
935 if (type == IF_VIRTIO) {
936 QemuOpts *devopts;
937 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
938 &error_abort);
939 if (arch_type == QEMU_ARCH_S390X) {
940 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
941 } else {
942 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
944 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
945 if (devaddr) {
946 qemu_opt_set(devopts, "addr", devaddr);
950 filename = qemu_opt_get(legacy_opts, "file");
952 /* Check werror/rerror compatibility with if=... */
953 werror = qemu_opt_get(legacy_opts, "werror");
954 if (werror != NULL) {
955 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
956 type != IF_NONE) {
957 error_report("werror is not supported by this bus type");
958 goto fail;
960 qdict_put(bs_opts, "werror", qstring_from_str(werror));
963 rerror = qemu_opt_get(legacy_opts, "rerror");
964 if (rerror != NULL) {
965 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
966 type != IF_NONE) {
967 error_report("rerror is not supported by this bus type");
968 goto fail;
970 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
973 /* Actual block device init: Functionality shared with blockdev-add */
974 blk = blockdev_init(filename, bs_opts, &local_err);
975 bs_opts = NULL;
976 if (!blk) {
977 if (local_err) {
978 error_report("%s", error_get_pretty(local_err));
979 error_free(local_err);
981 goto fail;
982 } else {
983 assert(!local_err);
986 /* Create legacy DriveInfo */
987 dinfo = g_malloc0(sizeof(*dinfo));
988 dinfo->opts = all_opts;
990 dinfo->cyls = cyls;
991 dinfo->heads = heads;
992 dinfo->secs = secs;
993 dinfo->trans = translation;
995 dinfo->type = type;
996 dinfo->bus = bus_id;
997 dinfo->unit = unit_id;
998 dinfo->devaddr = devaddr;
999 dinfo->serial = g_strdup(serial);
1001 blk_set_legacy_dinfo(blk, dinfo);
1003 switch(type) {
1004 case IF_IDE:
1005 case IF_SCSI:
1006 case IF_XEN:
1007 case IF_NONE:
1008 dinfo->media_cd = media == MEDIA_CDROM;
1009 break;
1010 default:
1011 break;
1014 fail:
1015 qemu_opts_del(legacy_opts);
1016 QDECREF(bs_opts);
1017 return dinfo;
1020 void do_commit(Monitor *mon, const QDict *qdict)
1022 const char *device = qdict_get_str(qdict, "device");
1023 BlockDriverState *bs;
1024 int ret;
1026 if (!strcmp(device, "all")) {
1027 ret = bdrv_commit_all();
1028 } else {
1029 bs = bdrv_find(device);
1030 if (!bs) {
1031 monitor_printf(mon, "Device '%s' not found\n", device);
1032 return;
1034 ret = bdrv_commit(bs);
1036 if (ret < 0) {
1037 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1038 strerror(-ret));
1042 static void blockdev_do_action(int kind, void *data, Error **errp)
1044 TransactionAction action;
1045 TransactionActionList list;
1047 action.kind = kind;
1048 action.data = data;
1049 list.value = &action;
1050 list.next = NULL;
1051 qmp_transaction(&list, errp);
1054 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1055 bool has_node_name, const char *node_name,
1056 const char *snapshot_file,
1057 bool has_snapshot_node_name,
1058 const char *snapshot_node_name,
1059 bool has_format, const char *format,
1060 bool has_mode, NewImageMode mode, Error **errp)
1062 BlockdevSnapshot snapshot = {
1063 .has_device = has_device,
1064 .device = (char *) device,
1065 .has_node_name = has_node_name,
1066 .node_name = (char *) node_name,
1067 .snapshot_file = (char *) snapshot_file,
1068 .has_snapshot_node_name = has_snapshot_node_name,
1069 .snapshot_node_name = (char *) snapshot_node_name,
1070 .has_format = has_format,
1071 .format = (char *) format,
1072 .has_mode = has_mode,
1073 .mode = mode,
1075 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1076 &snapshot, errp);
1079 void qmp_blockdev_snapshot_internal_sync(const char *device,
1080 const char *name,
1081 Error **errp)
1083 BlockdevSnapshotInternal snapshot = {
1084 .device = (char *) device,
1085 .name = (char *) name
1088 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1089 &snapshot, errp);
1092 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1093 bool has_id,
1094 const char *id,
1095 bool has_name,
1096 const char *name,
1097 Error **errp)
1099 BlockDriverState *bs = bdrv_find(device);
1100 AioContext *aio_context;
1101 QEMUSnapshotInfo sn;
1102 Error *local_err = NULL;
1103 SnapshotInfo *info = NULL;
1104 int ret;
1106 if (!bs) {
1107 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1108 return NULL;
1111 if (!has_id) {
1112 id = NULL;
1115 if (!has_name) {
1116 name = NULL;
1119 if (!id && !name) {
1120 error_setg(errp, "Name or id must be provided");
1121 return NULL;
1124 aio_context = bdrv_get_aio_context(bs);
1125 aio_context_acquire(aio_context);
1127 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1128 goto out_aio_context;
1131 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1132 if (local_err) {
1133 error_propagate(errp, local_err);
1134 goto out_aio_context;
1136 if (!ret) {
1137 error_setg(errp,
1138 "Snapshot with id '%s' and name '%s' does not exist on "
1139 "device '%s'",
1140 STR_OR_NULL(id), STR_OR_NULL(name), device);
1141 goto out_aio_context;
1144 bdrv_snapshot_delete(bs, id, name, &local_err);
1145 if (local_err) {
1146 error_propagate(errp, local_err);
1147 goto out_aio_context;
1150 aio_context_release(aio_context);
1152 info = g_new0(SnapshotInfo, 1);
1153 info->id = g_strdup(sn.id_str);
1154 info->name = g_strdup(sn.name);
1155 info->date_nsec = sn.date_nsec;
1156 info->date_sec = sn.date_sec;
1157 info->vm_state_size = sn.vm_state_size;
1158 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1159 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1161 return info;
1163 out_aio_context:
1164 aio_context_release(aio_context);
1165 return NULL;
1168 /* New and old BlockDriverState structs for atomic group operations */
1170 typedef struct BlkTransactionState BlkTransactionState;
1172 /* Only prepare() may fail. In a single transaction, only one of commit() or
1173 abort() will be called, clean() will always be called if it present. */
1174 typedef struct BdrvActionOps {
1175 /* Size of state struct, in bytes. */
1176 size_t instance_size;
1177 /* Prepare the work, must NOT be NULL. */
1178 void (*prepare)(BlkTransactionState *common, Error **errp);
1179 /* Commit the changes, can be NULL. */
1180 void (*commit)(BlkTransactionState *common);
1181 /* Abort the changes on fail, can be NULL. */
1182 void (*abort)(BlkTransactionState *common);
1183 /* Clean up resource in the end, can be NULL. */
1184 void (*clean)(BlkTransactionState *common);
1185 } BdrvActionOps;
1188 * This structure must be arranged as first member in child type, assuming
1189 * that compiler will also arrange it to the same address with parent instance.
1190 * Later it will be used in free().
1192 struct BlkTransactionState {
1193 TransactionAction *action;
1194 const BdrvActionOps *ops;
1195 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1198 /* internal snapshot private data */
1199 typedef struct InternalSnapshotState {
1200 BlkTransactionState common;
1201 BlockDriverState *bs;
1202 AioContext *aio_context;
1203 QEMUSnapshotInfo sn;
1204 } InternalSnapshotState;
1206 static void internal_snapshot_prepare(BlkTransactionState *common,
1207 Error **errp)
1209 Error *local_err = NULL;
1210 const char *device;
1211 const char *name;
1212 BlockDriverState *bs;
1213 QEMUSnapshotInfo old_sn, *sn;
1214 bool ret;
1215 qemu_timeval tv;
1216 BlockdevSnapshotInternal *internal;
1217 InternalSnapshotState *state;
1218 int ret1;
1220 g_assert(common->action->kind ==
1221 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1222 internal = common->action->blockdev_snapshot_internal_sync;
1223 state = DO_UPCAST(InternalSnapshotState, common, common);
1225 /* 1. parse input */
1226 device = internal->device;
1227 name = internal->name;
1229 /* 2. check for validation */
1230 bs = bdrv_find(device);
1231 if (!bs) {
1232 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1233 return;
1236 /* AioContext is released in .clean() */
1237 state->aio_context = bdrv_get_aio_context(bs);
1238 aio_context_acquire(state->aio_context);
1240 if (!bdrv_is_inserted(bs)) {
1241 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1242 return;
1245 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1246 return;
1249 if (bdrv_is_read_only(bs)) {
1250 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1251 return;
1254 if (!bdrv_can_snapshot(bs)) {
1255 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1256 bs->drv->format_name, device, "internal snapshot");
1257 return;
1260 if (!strlen(name)) {
1261 error_setg(errp, "Name is empty");
1262 return;
1265 /* check whether a snapshot with name exist */
1266 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1267 &local_err);
1268 if (local_err) {
1269 error_propagate(errp, local_err);
1270 return;
1271 } else if (ret) {
1272 error_setg(errp,
1273 "Snapshot with name '%s' already exists on device '%s'",
1274 name, device);
1275 return;
1278 /* 3. take the snapshot */
1279 sn = &state->sn;
1280 pstrcpy(sn->name, sizeof(sn->name), name);
1281 qemu_gettimeofday(&tv);
1282 sn->date_sec = tv.tv_sec;
1283 sn->date_nsec = tv.tv_usec * 1000;
1284 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1286 ret1 = bdrv_snapshot_create(bs, sn);
1287 if (ret1 < 0) {
1288 error_setg_errno(errp, -ret1,
1289 "Failed to create snapshot '%s' on device '%s'",
1290 name, device);
1291 return;
1294 /* 4. succeed, mark a snapshot is created */
1295 state->bs = bs;
1298 static void internal_snapshot_abort(BlkTransactionState *common)
1300 InternalSnapshotState *state =
1301 DO_UPCAST(InternalSnapshotState, common, common);
1302 BlockDriverState *bs = state->bs;
1303 QEMUSnapshotInfo *sn = &state->sn;
1304 Error *local_error = NULL;
1306 if (!bs) {
1307 return;
1310 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1311 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1312 "device '%s' in abort: %s",
1313 sn->id_str,
1314 sn->name,
1315 bdrv_get_device_name(bs),
1316 error_get_pretty(local_error));
1317 error_free(local_error);
1321 static void internal_snapshot_clean(BlkTransactionState *common)
1323 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1324 common, common);
1326 if (state->aio_context) {
1327 aio_context_release(state->aio_context);
1331 /* external snapshot private data */
1332 typedef struct ExternalSnapshotState {
1333 BlkTransactionState common;
1334 BlockDriverState *old_bs;
1335 BlockDriverState *new_bs;
1336 AioContext *aio_context;
1337 } ExternalSnapshotState;
1339 static void external_snapshot_prepare(BlkTransactionState *common,
1340 Error **errp)
1342 BlockDriver *drv;
1343 int flags, ret;
1344 QDict *options = NULL;
1345 Error *local_err = NULL;
1346 bool has_device = false;
1347 const char *device;
1348 bool has_node_name = false;
1349 const char *node_name;
1350 bool has_snapshot_node_name = false;
1351 const char *snapshot_node_name;
1352 const char *new_image_file;
1353 const char *format = "qcow2";
1354 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1355 ExternalSnapshotState *state =
1356 DO_UPCAST(ExternalSnapshotState, common, common);
1357 TransactionAction *action = common->action;
1359 /* get parameters */
1360 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1362 has_device = action->blockdev_snapshot_sync->has_device;
1363 device = action->blockdev_snapshot_sync->device;
1364 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1365 node_name = action->blockdev_snapshot_sync->node_name;
1366 has_snapshot_node_name =
1367 action->blockdev_snapshot_sync->has_snapshot_node_name;
1368 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1370 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1371 if (action->blockdev_snapshot_sync->has_format) {
1372 format = action->blockdev_snapshot_sync->format;
1374 if (action->blockdev_snapshot_sync->has_mode) {
1375 mode = action->blockdev_snapshot_sync->mode;
1378 /* start processing */
1379 drv = bdrv_find_format(format);
1380 if (!drv) {
1381 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1382 return;
1385 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1386 has_node_name ? node_name : NULL,
1387 &local_err);
1388 if (local_err) {
1389 error_propagate(errp, local_err);
1390 return;
1393 if (has_node_name && !has_snapshot_node_name) {
1394 error_setg(errp, "New snapshot node name missing");
1395 return;
1398 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1399 error_setg(errp, "New snapshot node name already existing");
1400 return;
1403 /* Acquire AioContext now so any threads operating on old_bs stop */
1404 state->aio_context = bdrv_get_aio_context(state->old_bs);
1405 aio_context_acquire(state->aio_context);
1407 if (!bdrv_is_inserted(state->old_bs)) {
1408 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1409 return;
1412 if (bdrv_op_is_blocked(state->old_bs,
1413 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1414 return;
1417 if (!bdrv_is_read_only(state->old_bs)) {
1418 if (bdrv_flush(state->old_bs)) {
1419 error_set(errp, QERR_IO_ERROR);
1420 return;
1424 if (!bdrv_is_first_non_filter(state->old_bs)) {
1425 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1426 return;
1429 flags = state->old_bs->open_flags;
1431 /* create new image w/backing file */
1432 if (mode != NEW_IMAGE_MODE_EXISTING) {
1433 bdrv_img_create(new_image_file, format,
1434 state->old_bs->filename,
1435 state->old_bs->drv->format_name,
1436 NULL, -1, flags, &local_err, false);
1437 if (local_err) {
1438 error_propagate(errp, local_err);
1439 return;
1443 if (has_snapshot_node_name) {
1444 options = qdict_new();
1445 qdict_put(options, "node-name",
1446 qstring_from_str(snapshot_node_name));
1449 /* TODO Inherit bs->options or only take explicit options with an
1450 * extended QMP command? */
1451 assert(state->new_bs == NULL);
1452 ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1453 flags | BDRV_O_NO_BACKING, drv, &local_err);
1454 /* We will manually add the backing_hd field to the bs later */
1455 if (ret != 0) {
1456 error_propagate(errp, local_err);
1460 static void external_snapshot_commit(BlkTransactionState *common)
1462 ExternalSnapshotState *state =
1463 DO_UPCAST(ExternalSnapshotState, common, common);
1465 bdrv_set_aio_context(state->new_bs, state->aio_context);
1467 /* This removes our old bs and adds the new bs */
1468 bdrv_append(state->new_bs, state->old_bs);
1469 /* We don't need (or want) to use the transactional
1470 * bdrv_reopen_multiple() across all the entries at once, because we
1471 * don't want to abort all of them if one of them fails the reopen */
1472 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1473 NULL);
1475 aio_context_release(state->aio_context);
1478 static void external_snapshot_abort(BlkTransactionState *common)
1480 ExternalSnapshotState *state =
1481 DO_UPCAST(ExternalSnapshotState, common, common);
1482 if (state->new_bs) {
1483 bdrv_unref(state->new_bs);
1485 if (state->aio_context) {
1486 aio_context_release(state->aio_context);
1490 typedef struct DriveBackupState {
1491 BlkTransactionState common;
1492 BlockDriverState *bs;
1493 AioContext *aio_context;
1494 BlockJob *job;
1495 } DriveBackupState;
1497 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1499 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1500 BlockDriverState *bs;
1501 DriveBackup *backup;
1502 Error *local_err = NULL;
1504 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1505 backup = common->action->drive_backup;
1507 bs = bdrv_find(backup->device);
1508 if (!bs) {
1509 error_set(errp, QERR_DEVICE_NOT_FOUND, backup->device);
1510 return;
1513 /* AioContext is released in .clean() */
1514 state->aio_context = bdrv_get_aio_context(bs);
1515 aio_context_acquire(state->aio_context);
1517 qmp_drive_backup(backup->device, backup->target,
1518 backup->has_format, backup->format,
1519 backup->sync,
1520 backup->has_mode, backup->mode,
1521 backup->has_speed, backup->speed,
1522 backup->has_on_source_error, backup->on_source_error,
1523 backup->has_on_target_error, backup->on_target_error,
1524 &local_err);
1525 if (local_err) {
1526 error_propagate(errp, local_err);
1527 return;
1530 state->bs = bs;
1531 state->job = state->bs->job;
1534 static void drive_backup_abort(BlkTransactionState *common)
1536 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1537 BlockDriverState *bs = state->bs;
1539 /* Only cancel if it's the job we started */
1540 if (bs && bs->job && bs->job == state->job) {
1541 block_job_cancel_sync(bs->job);
1545 static void drive_backup_clean(BlkTransactionState *common)
1547 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1549 if (state->aio_context) {
1550 aio_context_release(state->aio_context);
1554 typedef struct BlockdevBackupState {
1555 BlkTransactionState common;
1556 BlockDriverState *bs;
1557 BlockJob *job;
1558 AioContext *aio_context;
1559 } BlockdevBackupState;
1561 static void blockdev_backup_prepare(BlkTransactionState *common, Error **errp)
1563 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1564 BlockdevBackup *backup;
1565 BlockDriverState *bs, *target;
1566 Error *local_err = NULL;
1568 assert(common->action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1569 backup = common->action->blockdev_backup;
1571 bs = bdrv_find(backup->device);
1572 if (!bs) {
1573 error_set(errp, QERR_DEVICE_NOT_FOUND, backup->device);
1574 return;
1577 target = bdrv_find(backup->target);
1578 if (!target) {
1579 error_set(errp, QERR_DEVICE_NOT_FOUND, backup->target);
1580 return;
1583 /* AioContext is released in .clean() */
1584 state->aio_context = bdrv_get_aio_context(bs);
1585 if (state->aio_context != bdrv_get_aio_context(target)) {
1586 state->aio_context = NULL;
1587 error_setg(errp, "Backup between two IO threads is not implemented");
1588 return;
1590 aio_context_acquire(state->aio_context);
1592 qmp_blockdev_backup(backup->device, backup->target,
1593 backup->sync,
1594 backup->has_speed, backup->speed,
1595 backup->has_on_source_error, backup->on_source_error,
1596 backup->has_on_target_error, backup->on_target_error,
1597 &local_err);
1598 if (local_err) {
1599 error_propagate(errp, local_err);
1600 return;
1603 state->bs = bs;
1604 state->job = state->bs->job;
1607 static void blockdev_backup_abort(BlkTransactionState *common)
1609 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1610 BlockDriverState *bs = state->bs;
1612 /* Only cancel if it's the job we started */
1613 if (bs && bs->job && bs->job == state->job) {
1614 block_job_cancel_sync(bs->job);
1618 static void blockdev_backup_clean(BlkTransactionState *common)
1620 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1622 if (state->aio_context) {
1623 aio_context_release(state->aio_context);
1627 static void abort_prepare(BlkTransactionState *common, Error **errp)
1629 error_setg(errp, "Transaction aborted using Abort action");
1632 static void abort_commit(BlkTransactionState *common)
1634 g_assert_not_reached(); /* this action never succeeds */
1637 static const BdrvActionOps actions[] = {
1638 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1639 .instance_size = sizeof(ExternalSnapshotState),
1640 .prepare = external_snapshot_prepare,
1641 .commit = external_snapshot_commit,
1642 .abort = external_snapshot_abort,
1644 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1645 .instance_size = sizeof(DriveBackupState),
1646 .prepare = drive_backup_prepare,
1647 .abort = drive_backup_abort,
1648 .clean = drive_backup_clean,
1650 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
1651 .instance_size = sizeof(BlockdevBackupState),
1652 .prepare = blockdev_backup_prepare,
1653 .abort = blockdev_backup_abort,
1654 .clean = blockdev_backup_clean,
1656 [TRANSACTION_ACTION_KIND_ABORT] = {
1657 .instance_size = sizeof(BlkTransactionState),
1658 .prepare = abort_prepare,
1659 .commit = abort_commit,
1661 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1662 .instance_size = sizeof(InternalSnapshotState),
1663 .prepare = internal_snapshot_prepare,
1664 .abort = internal_snapshot_abort,
1665 .clean = internal_snapshot_clean,
1670 * 'Atomic' group operations. The operations are performed as a set, and if
1671 * any fail then we roll back all operations in the group.
1673 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1675 TransactionActionList *dev_entry = dev_list;
1676 BlkTransactionState *state, *next;
1677 Error *local_err = NULL;
1679 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1680 QSIMPLEQ_INIT(&snap_bdrv_states);
1682 /* drain all i/o before any operations */
1683 bdrv_drain_all();
1685 /* We don't do anything in this loop that commits us to the operations */
1686 while (NULL != dev_entry) {
1687 TransactionAction *dev_info = NULL;
1688 const BdrvActionOps *ops;
1690 dev_info = dev_entry->value;
1691 dev_entry = dev_entry->next;
1693 assert(dev_info->kind < ARRAY_SIZE(actions));
1695 ops = &actions[dev_info->kind];
1696 assert(ops->instance_size > 0);
1698 state = g_malloc0(ops->instance_size);
1699 state->ops = ops;
1700 state->action = dev_info;
1701 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1703 state->ops->prepare(state, &local_err);
1704 if (local_err) {
1705 error_propagate(errp, local_err);
1706 goto delete_and_fail;
1710 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1711 if (state->ops->commit) {
1712 state->ops->commit(state);
1716 /* success */
1717 goto exit;
1719 delete_and_fail:
1720 /* failure, and it is all-or-none; roll back all operations */
1721 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1722 if (state->ops->abort) {
1723 state->ops->abort(state);
1726 exit:
1727 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1728 if (state->ops->clean) {
1729 state->ops->clean(state);
1731 g_free(state);
1736 static void eject_device(BlockBackend *blk, int force, Error **errp)
1738 BlockDriverState *bs = blk_bs(blk);
1739 AioContext *aio_context;
1741 aio_context = bdrv_get_aio_context(bs);
1742 aio_context_acquire(aio_context);
1744 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1745 goto out;
1747 if (!blk_dev_has_removable_media(blk)) {
1748 error_setg(errp, "Device '%s' is not removable",
1749 bdrv_get_device_name(bs));
1750 goto out;
1753 if (blk_dev_is_medium_locked(blk) && !blk_dev_is_tray_open(blk)) {
1754 blk_dev_eject_request(blk, force);
1755 if (!force) {
1756 error_setg(errp, "Device '%s' is locked",
1757 bdrv_get_device_name(bs));
1758 goto out;
1762 bdrv_close(bs);
1764 out:
1765 aio_context_release(aio_context);
1768 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1770 BlockBackend *blk;
1772 blk = blk_by_name(device);
1773 if (!blk) {
1774 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1775 return;
1778 eject_device(blk, force, errp);
1781 void qmp_block_passwd(bool has_device, const char *device,
1782 bool has_node_name, const char *node_name,
1783 const char *password, Error **errp)
1785 Error *local_err = NULL;
1786 BlockDriverState *bs;
1787 AioContext *aio_context;
1789 bs = bdrv_lookup_bs(has_device ? device : NULL,
1790 has_node_name ? node_name : NULL,
1791 &local_err);
1792 if (local_err) {
1793 error_propagate(errp, local_err);
1794 return;
1797 aio_context = bdrv_get_aio_context(bs);
1798 aio_context_acquire(aio_context);
1800 bdrv_add_key(bs, password, errp);
1802 aio_context_release(aio_context);
1805 /* Assumes AioContext is held */
1806 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1807 int bdrv_flags, BlockDriver *drv,
1808 const char *password, Error **errp)
1810 Error *local_err = NULL;
1811 int ret;
1813 ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1814 if (ret < 0) {
1815 error_propagate(errp, local_err);
1816 return;
1819 bdrv_add_key(bs, password, errp);
1822 void qmp_change_blockdev(const char *device, const char *filename,
1823 const char *format, Error **errp)
1825 BlockBackend *blk;
1826 BlockDriverState *bs;
1827 AioContext *aio_context;
1828 BlockDriver *drv = NULL;
1829 int bdrv_flags;
1830 Error *err = NULL;
1832 blk = blk_by_name(device);
1833 if (!blk) {
1834 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1835 return;
1837 bs = blk_bs(blk);
1839 aio_context = bdrv_get_aio_context(bs);
1840 aio_context_acquire(aio_context);
1842 if (format) {
1843 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1844 if (!drv) {
1845 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1846 goto out;
1850 eject_device(blk, 0, &err);
1851 if (err) {
1852 error_propagate(errp, err);
1853 goto out;
1856 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1857 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1859 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1861 out:
1862 aio_context_release(aio_context);
1865 /* throttling disk I/O limits */
1866 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1867 int64_t bps_wr,
1868 int64_t iops,
1869 int64_t iops_rd,
1870 int64_t iops_wr,
1871 bool has_bps_max,
1872 int64_t bps_max,
1873 bool has_bps_rd_max,
1874 int64_t bps_rd_max,
1875 bool has_bps_wr_max,
1876 int64_t bps_wr_max,
1877 bool has_iops_max,
1878 int64_t iops_max,
1879 bool has_iops_rd_max,
1880 int64_t iops_rd_max,
1881 bool has_iops_wr_max,
1882 int64_t iops_wr_max,
1883 bool has_iops_size,
1884 int64_t iops_size, Error **errp)
1886 ThrottleConfig cfg;
1887 BlockDriverState *bs;
1888 AioContext *aio_context;
1890 bs = bdrv_find(device);
1891 if (!bs) {
1892 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1893 return;
1896 memset(&cfg, 0, sizeof(cfg));
1897 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1898 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1899 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1901 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1902 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1903 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1905 if (has_bps_max) {
1906 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1908 if (has_bps_rd_max) {
1909 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1911 if (has_bps_wr_max) {
1912 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1914 if (has_iops_max) {
1915 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1917 if (has_iops_rd_max) {
1918 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1920 if (has_iops_wr_max) {
1921 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1924 if (has_iops_size) {
1925 cfg.op_size = iops_size;
1928 if (!check_throttle_config(&cfg, errp)) {
1929 return;
1932 aio_context = bdrv_get_aio_context(bs);
1933 aio_context_acquire(aio_context);
1935 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1936 bdrv_io_limits_enable(bs);
1937 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1938 bdrv_io_limits_disable(bs);
1941 if (bs->io_limits_enabled) {
1942 bdrv_set_io_limits(bs, &cfg);
1945 aio_context_release(aio_context);
1948 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1950 const char *id = qdict_get_str(qdict, "id");
1951 BlockBackend *blk;
1952 BlockDriverState *bs;
1953 AioContext *aio_context;
1954 Error *local_err = NULL;
1956 blk = blk_by_name(id);
1957 if (!blk) {
1958 error_report("Device '%s' not found", id);
1959 return -1;
1961 bs = blk_bs(blk);
1963 if (!blk_legacy_dinfo(blk)) {
1964 error_report("Deleting device added with blockdev-add"
1965 " is not supported");
1966 return -1;
1969 aio_context = bdrv_get_aio_context(bs);
1970 aio_context_acquire(aio_context);
1972 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1973 error_report("%s", error_get_pretty(local_err));
1974 error_free(local_err);
1975 aio_context_release(aio_context);
1976 return -1;
1979 /* quiesce block driver; prevent further io */
1980 bdrv_drain_all();
1981 bdrv_flush(bs);
1982 bdrv_close(bs);
1984 /* if we have a device attached to this BlockDriverState
1985 * then we need to make the drive anonymous until the device
1986 * can be removed. If this is a drive with no device backing
1987 * then we can just get rid of the block driver state right here.
1989 if (blk_get_attached_dev(blk)) {
1990 blk_hide_on_behalf_of_do_drive_del(blk);
1991 /* Further I/O must not pause the guest */
1992 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1993 BLOCKDEV_ON_ERROR_REPORT);
1994 } else {
1995 blk_unref(blk);
1998 aio_context_release(aio_context);
1999 return 0;
2002 void qmp_block_resize(bool has_device, const char *device,
2003 bool has_node_name, const char *node_name,
2004 int64_t size, Error **errp)
2006 Error *local_err = NULL;
2007 BlockDriverState *bs;
2008 AioContext *aio_context;
2009 int ret;
2011 bs = bdrv_lookup_bs(has_device ? device : NULL,
2012 has_node_name ? node_name : NULL,
2013 &local_err);
2014 if (local_err) {
2015 error_propagate(errp, local_err);
2016 return;
2019 aio_context = bdrv_get_aio_context(bs);
2020 aio_context_acquire(aio_context);
2022 if (!bdrv_is_first_non_filter(bs)) {
2023 error_set(errp, QERR_FEATURE_DISABLED, "resize");
2024 goto out;
2027 if (size < 0) {
2028 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2029 goto out;
2032 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2033 error_set(errp, QERR_DEVICE_IN_USE, device);
2034 goto out;
2037 /* complete all in-flight operations before resizing the device */
2038 bdrv_drain_all();
2040 ret = bdrv_truncate(bs, size);
2041 switch (ret) {
2042 case 0:
2043 break;
2044 case -ENOMEDIUM:
2045 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2046 break;
2047 case -ENOTSUP:
2048 error_set(errp, QERR_UNSUPPORTED);
2049 break;
2050 case -EACCES:
2051 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
2052 break;
2053 case -EBUSY:
2054 error_set(errp, QERR_DEVICE_IN_USE, device);
2055 break;
2056 default:
2057 error_setg_errno(errp, -ret, "Could not resize");
2058 break;
2061 out:
2062 aio_context_release(aio_context);
2065 static void block_job_cb(void *opaque, int ret)
2067 /* Note that this function may be executed from another AioContext besides
2068 * the QEMU main loop. If you need to access anything that assumes the
2069 * QEMU global mutex, use a BH or introduce a mutex.
2072 BlockDriverState *bs = opaque;
2073 const char *msg = NULL;
2075 trace_block_job_cb(bs, bs->job, ret);
2077 assert(bs->job);
2079 if (ret < 0) {
2080 msg = strerror(-ret);
2083 if (block_job_is_cancelled(bs->job)) {
2084 block_job_event_cancelled(bs->job);
2085 } else {
2086 block_job_event_completed(bs->job, msg);
2089 bdrv_put_ref_bh_schedule(bs);
2092 void qmp_block_stream(const char *device,
2093 bool has_base, const char *base,
2094 bool has_backing_file, const char *backing_file,
2095 bool has_speed, int64_t speed,
2096 bool has_on_error, BlockdevOnError on_error,
2097 Error **errp)
2099 BlockDriverState *bs;
2100 BlockDriverState *base_bs = NULL;
2101 AioContext *aio_context;
2102 Error *local_err = NULL;
2103 const char *base_name = NULL;
2105 if (!has_on_error) {
2106 on_error = BLOCKDEV_ON_ERROR_REPORT;
2109 bs = bdrv_find(device);
2110 if (!bs) {
2111 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2112 return;
2115 aio_context = bdrv_get_aio_context(bs);
2116 aio_context_acquire(aio_context);
2118 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
2119 goto out;
2122 if (has_base) {
2123 base_bs = bdrv_find_backing_image(bs, base);
2124 if (base_bs == NULL) {
2125 error_set(errp, QERR_BASE_NOT_FOUND, base);
2126 goto out;
2128 assert(bdrv_get_aio_context(base_bs) == aio_context);
2129 base_name = base;
2132 /* if we are streaming the entire chain, the result will have no backing
2133 * file, and specifying one is therefore an error */
2134 if (base_bs == NULL && has_backing_file) {
2135 error_setg(errp, "backing file specified, but streaming the "
2136 "entire chain");
2137 goto out;
2140 /* backing_file string overrides base bs filename */
2141 base_name = has_backing_file ? backing_file : base_name;
2143 stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2144 on_error, block_job_cb, bs, &local_err);
2145 if (local_err) {
2146 error_propagate(errp, local_err);
2147 goto out;
2150 trace_qmp_block_stream(bs, bs->job);
2152 out:
2153 aio_context_release(aio_context);
2156 void qmp_block_commit(const char *device,
2157 bool has_base, const char *base,
2158 bool has_top, const char *top,
2159 bool has_backing_file, const char *backing_file,
2160 bool has_speed, int64_t speed,
2161 Error **errp)
2163 BlockDriverState *bs;
2164 BlockDriverState *base_bs, *top_bs;
2165 AioContext *aio_context;
2166 Error *local_err = NULL;
2167 /* This will be part of the QMP command, if/when the
2168 * BlockdevOnError change for blkmirror makes it in
2170 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2172 if (!has_speed) {
2173 speed = 0;
2176 /* Important Note:
2177 * libvirt relies on the DeviceNotFound error class in order to probe for
2178 * live commit feature versions; for this to work, we must make sure to
2179 * perform the device lookup before any generic errors that may occur in a
2180 * scenario in which all optional arguments are omitted. */
2181 bs = bdrv_find(device);
2182 if (!bs) {
2183 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2184 return;
2187 aio_context = bdrv_get_aio_context(bs);
2188 aio_context_acquire(aio_context);
2190 /* drain all i/o before commits */
2191 bdrv_drain_all();
2193 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2194 goto out;
2197 /* default top_bs is the active layer */
2198 top_bs = bs;
2200 if (has_top && top) {
2201 if (strcmp(bs->filename, top) != 0) {
2202 top_bs = bdrv_find_backing_image(bs, top);
2206 if (top_bs == NULL) {
2207 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2208 goto out;
2211 assert(bdrv_get_aio_context(top_bs) == aio_context);
2213 if (has_base && base) {
2214 base_bs = bdrv_find_backing_image(top_bs, base);
2215 } else {
2216 base_bs = bdrv_find_base(top_bs);
2219 if (base_bs == NULL) {
2220 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2221 goto out;
2224 assert(bdrv_get_aio_context(base_bs) == aio_context);
2226 if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2227 goto out;
2230 /* Do not allow attempts to commit an image into itself */
2231 if (top_bs == base_bs) {
2232 error_setg(errp, "cannot commit an image into itself");
2233 goto out;
2236 if (top_bs == bs) {
2237 if (has_backing_file) {
2238 error_setg(errp, "'backing-file' specified,"
2239 " but 'top' is the active layer");
2240 goto out;
2242 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2243 bs, &local_err);
2244 } else {
2245 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2246 has_backing_file ? backing_file : NULL, &local_err);
2248 if (local_err != NULL) {
2249 error_propagate(errp, local_err);
2250 goto out;
2253 out:
2254 aio_context_release(aio_context);
2257 void qmp_drive_backup(const char *device, const char *target,
2258 bool has_format, const char *format,
2259 enum MirrorSyncMode sync,
2260 bool has_mode, enum NewImageMode mode,
2261 bool has_speed, int64_t speed,
2262 bool has_on_source_error, BlockdevOnError on_source_error,
2263 bool has_on_target_error, BlockdevOnError on_target_error,
2264 Error **errp)
2266 BlockDriverState *bs;
2267 BlockDriverState *target_bs;
2268 BlockDriverState *source = NULL;
2269 AioContext *aio_context;
2270 BlockDriver *drv = NULL;
2271 Error *local_err = NULL;
2272 int flags;
2273 int64_t size;
2274 int ret;
2276 if (!has_speed) {
2277 speed = 0;
2279 if (!has_on_source_error) {
2280 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2282 if (!has_on_target_error) {
2283 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2285 if (!has_mode) {
2286 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2289 bs = bdrv_find(device);
2290 if (!bs) {
2291 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2292 return;
2295 aio_context = bdrv_get_aio_context(bs);
2296 aio_context_acquire(aio_context);
2298 /* Although backup_run has this check too, we need to use bs->drv below, so
2299 * do an early check redundantly. */
2300 if (!bdrv_is_inserted(bs)) {
2301 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2302 goto out;
2305 if (!has_format) {
2306 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2308 if (format) {
2309 drv = bdrv_find_format(format);
2310 if (!drv) {
2311 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2312 goto out;
2316 /* Early check to avoid creating target */
2317 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2318 goto out;
2321 flags = bs->open_flags | BDRV_O_RDWR;
2323 /* See if we have a backing HD we can use to create our new image
2324 * on top of. */
2325 if (sync == MIRROR_SYNC_MODE_TOP) {
2326 source = bs->backing_hd;
2327 if (!source) {
2328 sync = MIRROR_SYNC_MODE_FULL;
2331 if (sync == MIRROR_SYNC_MODE_NONE) {
2332 source = bs;
2335 size = bdrv_getlength(bs);
2336 if (size < 0) {
2337 error_setg_errno(errp, -size, "bdrv_getlength failed");
2338 goto out;
2341 if (mode != NEW_IMAGE_MODE_EXISTING) {
2342 assert(format && drv);
2343 if (source) {
2344 bdrv_img_create(target, format, source->filename,
2345 source->drv->format_name, NULL,
2346 size, flags, &local_err, false);
2347 } else {
2348 bdrv_img_create(target, format, NULL, NULL, NULL,
2349 size, flags, &local_err, false);
2353 if (local_err) {
2354 error_propagate(errp, local_err);
2355 goto out;
2358 target_bs = NULL;
2359 ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2360 if (ret < 0) {
2361 error_propagate(errp, local_err);
2362 goto out;
2365 bdrv_set_aio_context(target_bs, aio_context);
2367 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2368 block_job_cb, bs, &local_err);
2369 if (local_err != NULL) {
2370 bdrv_unref(target_bs);
2371 error_propagate(errp, local_err);
2372 goto out;
2375 out:
2376 aio_context_release(aio_context);
2379 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2381 return bdrv_named_nodes_list();
2384 void qmp_blockdev_backup(const char *device, const char *target,
2385 enum MirrorSyncMode sync,
2386 bool has_speed, int64_t speed,
2387 bool has_on_source_error,
2388 BlockdevOnError on_source_error,
2389 bool has_on_target_error,
2390 BlockdevOnError on_target_error,
2391 Error **errp)
2393 BlockDriverState *bs;
2394 BlockDriverState *target_bs;
2395 Error *local_err = NULL;
2396 AioContext *aio_context;
2398 if (!has_speed) {
2399 speed = 0;
2401 if (!has_on_source_error) {
2402 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2404 if (!has_on_target_error) {
2405 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2408 bs = bdrv_find(device);
2409 if (!bs) {
2410 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2411 return;
2414 aio_context = bdrv_get_aio_context(bs);
2415 aio_context_acquire(aio_context);
2417 target_bs = bdrv_find(target);
2418 if (!target_bs) {
2419 error_set(errp, QERR_DEVICE_NOT_FOUND, target);
2420 goto out;
2423 bdrv_ref(target_bs);
2424 bdrv_set_aio_context(target_bs, aio_context);
2425 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2426 block_job_cb, bs, &local_err);
2427 if (local_err != NULL) {
2428 bdrv_unref(target_bs);
2429 error_propagate(errp, local_err);
2431 out:
2432 aio_context_release(aio_context);
2435 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2437 void qmp_drive_mirror(const char *device, const char *target,
2438 bool has_format, const char *format,
2439 bool has_node_name, const char *node_name,
2440 bool has_replaces, const char *replaces,
2441 enum MirrorSyncMode sync,
2442 bool has_mode, enum NewImageMode mode,
2443 bool has_speed, int64_t speed,
2444 bool has_granularity, uint32_t granularity,
2445 bool has_buf_size, int64_t buf_size,
2446 bool has_on_source_error, BlockdevOnError on_source_error,
2447 bool has_on_target_error, BlockdevOnError on_target_error,
2448 Error **errp)
2450 BlockDriverState *bs;
2451 BlockDriverState *source, *target_bs;
2452 AioContext *aio_context;
2453 BlockDriver *drv = NULL;
2454 Error *local_err = NULL;
2455 QDict *options = NULL;
2456 int flags;
2457 int64_t size;
2458 int ret;
2460 if (!has_speed) {
2461 speed = 0;
2463 if (!has_on_source_error) {
2464 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2466 if (!has_on_target_error) {
2467 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2469 if (!has_mode) {
2470 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2472 if (!has_granularity) {
2473 granularity = 0;
2475 if (!has_buf_size) {
2476 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2479 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2480 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2481 "a value in range [512B, 64MB]");
2482 return;
2484 if (granularity & (granularity - 1)) {
2485 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity", "power of 2");
2486 return;
2489 bs = bdrv_find(device);
2490 if (!bs) {
2491 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2492 return;
2495 aio_context = bdrv_get_aio_context(bs);
2496 aio_context_acquire(aio_context);
2498 if (!bdrv_is_inserted(bs)) {
2499 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2500 goto out;
2503 if (!has_format) {
2504 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2506 if (format) {
2507 drv = bdrv_find_format(format);
2508 if (!drv) {
2509 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2510 goto out;
2514 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2515 goto out;
2518 flags = bs->open_flags | BDRV_O_RDWR;
2519 source = bs->backing_hd;
2520 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2521 sync = MIRROR_SYNC_MODE_FULL;
2523 if (sync == MIRROR_SYNC_MODE_NONE) {
2524 source = bs;
2527 size = bdrv_getlength(bs);
2528 if (size < 0) {
2529 error_setg_errno(errp, -size, "bdrv_getlength failed");
2530 goto out;
2533 if (has_replaces) {
2534 BlockDriverState *to_replace_bs;
2535 AioContext *replace_aio_context;
2536 int64_t replace_size;
2538 if (!has_node_name) {
2539 error_setg(errp, "a node-name must be provided when replacing a"
2540 " named node of the graph");
2541 goto out;
2544 to_replace_bs = check_to_replace_node(replaces, &local_err);
2546 if (!to_replace_bs) {
2547 error_propagate(errp, local_err);
2548 goto out;
2551 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
2552 aio_context_acquire(replace_aio_context);
2553 replace_size = bdrv_getlength(to_replace_bs);
2554 aio_context_release(replace_aio_context);
2556 if (size != replace_size) {
2557 error_setg(errp, "cannot replace image with a mirror image of "
2558 "different size");
2559 goto out;
2563 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2564 && mode != NEW_IMAGE_MODE_EXISTING)
2566 /* create new image w/o backing file */
2567 assert(format && drv);
2568 bdrv_img_create(target, format,
2569 NULL, NULL, NULL, size, flags, &local_err, false);
2570 } else {
2571 switch (mode) {
2572 case NEW_IMAGE_MODE_EXISTING:
2573 break;
2574 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2575 /* create new image with backing file */
2576 bdrv_img_create(target, format,
2577 source->filename,
2578 source->drv->format_name,
2579 NULL, size, flags, &local_err, false);
2580 break;
2581 default:
2582 abort();
2586 if (local_err) {
2587 error_propagate(errp, local_err);
2588 goto out;
2591 if (has_node_name) {
2592 options = qdict_new();
2593 qdict_put(options, "node-name", qstring_from_str(node_name));
2596 /* Mirroring takes care of copy-on-write using the source's backing
2597 * file.
2599 target_bs = NULL;
2600 ret = bdrv_open(&target_bs, target, NULL, options,
2601 flags | BDRV_O_NO_BACKING, drv, &local_err);
2602 if (ret < 0) {
2603 error_propagate(errp, local_err);
2604 goto out;
2607 bdrv_set_aio_context(target_bs, aio_context);
2609 /* pass the node name to replace to mirror start since it's loose coupling
2610 * and will allow to check whether the node still exist at mirror completion
2612 mirror_start(bs, target_bs,
2613 has_replaces ? replaces : NULL,
2614 speed, granularity, buf_size, sync,
2615 on_source_error, on_target_error,
2616 block_job_cb, bs, &local_err);
2617 if (local_err != NULL) {
2618 bdrv_unref(target_bs);
2619 error_propagate(errp, local_err);
2620 goto out;
2623 out:
2624 aio_context_release(aio_context);
2627 /* Get the block job for a given device name and acquire its AioContext */
2628 static BlockJob *find_block_job(const char *device, AioContext **aio_context,
2629 Error **errp)
2631 BlockDriverState *bs;
2633 bs = bdrv_find(device);
2634 if (!bs) {
2635 goto notfound;
2638 *aio_context = bdrv_get_aio_context(bs);
2639 aio_context_acquire(*aio_context);
2641 if (!bs->job) {
2642 aio_context_release(*aio_context);
2643 goto notfound;
2646 return bs->job;
2648 notfound:
2649 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
2650 "No active block job on device '%s'", device);
2651 *aio_context = NULL;
2652 return NULL;
2655 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2657 AioContext *aio_context;
2658 BlockJob *job = find_block_job(device, &aio_context, errp);
2660 if (!job) {
2661 return;
2664 block_job_set_speed(job, speed, errp);
2665 aio_context_release(aio_context);
2668 void qmp_block_job_cancel(const char *device,
2669 bool has_force, bool force, Error **errp)
2671 AioContext *aio_context;
2672 BlockJob *job = find_block_job(device, &aio_context, errp);
2674 if (!job) {
2675 return;
2678 if (!has_force) {
2679 force = false;
2682 if (job->paused && !force) {
2683 error_setg(errp, "The block job for device '%s' is currently paused",
2684 device);
2685 goto out;
2688 trace_qmp_block_job_cancel(job);
2689 block_job_cancel(job);
2690 out:
2691 aio_context_release(aio_context);
2694 void qmp_block_job_pause(const char *device, Error **errp)
2696 AioContext *aio_context;
2697 BlockJob *job = find_block_job(device, &aio_context, errp);
2699 if (!job) {
2700 return;
2703 trace_qmp_block_job_pause(job);
2704 block_job_pause(job);
2705 aio_context_release(aio_context);
2708 void qmp_block_job_resume(const char *device, Error **errp)
2710 AioContext *aio_context;
2711 BlockJob *job = find_block_job(device, &aio_context, errp);
2713 if (!job) {
2714 return;
2717 trace_qmp_block_job_resume(job);
2718 block_job_resume(job);
2719 aio_context_release(aio_context);
2722 void qmp_block_job_complete(const char *device, Error **errp)
2724 AioContext *aio_context;
2725 BlockJob *job = find_block_job(device, &aio_context, errp);
2727 if (!job) {
2728 return;
2731 trace_qmp_block_job_complete(job);
2732 block_job_complete(job, errp);
2733 aio_context_release(aio_context);
2736 void qmp_change_backing_file(const char *device,
2737 const char *image_node_name,
2738 const char *backing_file,
2739 Error **errp)
2741 BlockDriverState *bs = NULL;
2742 AioContext *aio_context;
2743 BlockDriverState *image_bs = NULL;
2744 Error *local_err = NULL;
2745 bool ro;
2746 int open_flags;
2747 int ret;
2749 /* find the top layer BDS of the chain */
2750 bs = bdrv_find(device);
2751 if (!bs) {
2752 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2753 return;
2756 aio_context = bdrv_get_aio_context(bs);
2757 aio_context_acquire(aio_context);
2759 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
2760 if (local_err) {
2761 error_propagate(errp, local_err);
2762 goto out;
2765 if (!image_bs) {
2766 error_setg(errp, "image file not found");
2767 goto out;
2770 if (bdrv_find_base(image_bs) == image_bs) {
2771 error_setg(errp, "not allowing backing file change on an image "
2772 "without a backing file");
2773 goto out;
2776 /* even though we are not necessarily operating on bs, we need it to
2777 * determine if block ops are currently prohibited on the chain */
2778 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
2779 goto out;
2782 /* final sanity check */
2783 if (!bdrv_chain_contains(bs, image_bs)) {
2784 error_setg(errp, "'%s' and image file are not in the same chain",
2785 device);
2786 goto out;
2789 /* if not r/w, reopen to make r/w */
2790 open_flags = image_bs->open_flags;
2791 ro = bdrv_is_read_only(image_bs);
2793 if (ro) {
2794 bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
2795 if (local_err) {
2796 error_propagate(errp, local_err);
2797 goto out;
2801 ret = bdrv_change_backing_file(image_bs, backing_file,
2802 image_bs->drv ? image_bs->drv->format_name : "");
2804 if (ret < 0) {
2805 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
2806 backing_file);
2807 /* don't exit here, so we can try to restore open flags if
2808 * appropriate */
2811 if (ro) {
2812 bdrv_reopen(image_bs, open_flags, &local_err);
2813 if (local_err) {
2814 error_propagate(errp, local_err); /* will preserve prior errp */
2818 out:
2819 aio_context_release(aio_context);
2822 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2824 QmpOutputVisitor *ov = qmp_output_visitor_new();
2825 BlockBackend *blk;
2826 QObject *obj;
2827 QDict *qdict;
2828 Error *local_err = NULL;
2830 /* Require an ID in the top level */
2831 if (!options->has_id) {
2832 error_setg(errp, "Block device needs an ID");
2833 goto fail;
2836 /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
2837 * cache.direct=false instead of silently switching to aio=threads, except
2838 * when called from drive_new().
2840 * For now, simply forbidding the combination for all drivers will do. */
2841 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2842 bool direct = options->has_cache &&
2843 options->cache->has_direct &&
2844 options->cache->direct;
2845 if (!direct) {
2846 error_setg(errp, "aio=native requires cache.direct=true");
2847 goto fail;
2851 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2852 &options, NULL, &local_err);
2853 if (local_err) {
2854 error_propagate(errp, local_err);
2855 goto fail;
2858 obj = qmp_output_get_qobject(ov);
2859 qdict = qobject_to_qdict(obj);
2861 qdict_flatten(qdict);
2863 blk = blockdev_init(NULL, qdict, &local_err);
2864 if (local_err) {
2865 error_propagate(errp, local_err);
2866 goto fail;
2869 if (bdrv_key_required(blk_bs(blk))) {
2870 blk_unref(blk);
2871 error_setg(errp, "blockdev-add doesn't support encrypted devices");
2872 goto fail;
2875 fail:
2876 qmp_output_visitor_cleanup(ov);
2879 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2881 BlockJobInfoList *head = NULL, **p_next = &head;
2882 BlockDriverState *bs;
2884 for (bs = bdrv_next(NULL); bs; bs = bdrv_next(bs)) {
2885 AioContext *aio_context = bdrv_get_aio_context(bs);
2887 aio_context_acquire(aio_context);
2889 if (bs->job) {
2890 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2891 elem->value = block_job_query(bs->job);
2892 *p_next = elem;
2893 p_next = &elem->next;
2896 aio_context_release(aio_context);
2899 return head;
2902 QemuOptsList qemu_common_drive_opts = {
2903 .name = "drive",
2904 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2905 .desc = {
2907 .name = "snapshot",
2908 .type = QEMU_OPT_BOOL,
2909 .help = "enable/disable snapshot mode",
2911 .name = "discard",
2912 .type = QEMU_OPT_STRING,
2913 .help = "discard operation (ignore/off, unmap/on)",
2915 .name = "cache.writeback",
2916 .type = QEMU_OPT_BOOL,
2917 .help = "enables writeback mode for any caches",
2919 .name = "cache.direct",
2920 .type = QEMU_OPT_BOOL,
2921 .help = "enables use of O_DIRECT (bypass the host page cache)",
2923 .name = "cache.no-flush",
2924 .type = QEMU_OPT_BOOL,
2925 .help = "ignore any flush requests for the device",
2927 .name = "aio",
2928 .type = QEMU_OPT_STRING,
2929 .help = "host AIO implementation (threads, native)",
2931 .name = "format",
2932 .type = QEMU_OPT_STRING,
2933 .help = "disk format (raw, qcow2, ...)",
2935 .name = "rerror",
2936 .type = QEMU_OPT_STRING,
2937 .help = "read error action",
2939 .name = "werror",
2940 .type = QEMU_OPT_STRING,
2941 .help = "write error action",
2943 .name = "read-only",
2944 .type = QEMU_OPT_BOOL,
2945 .help = "open drive file as read-only",
2947 .name = "throttling.iops-total",
2948 .type = QEMU_OPT_NUMBER,
2949 .help = "limit total I/O operations per second",
2951 .name = "throttling.iops-read",
2952 .type = QEMU_OPT_NUMBER,
2953 .help = "limit read operations per second",
2955 .name = "throttling.iops-write",
2956 .type = QEMU_OPT_NUMBER,
2957 .help = "limit write operations per second",
2959 .name = "throttling.bps-total",
2960 .type = QEMU_OPT_NUMBER,
2961 .help = "limit total bytes per second",
2963 .name = "throttling.bps-read",
2964 .type = QEMU_OPT_NUMBER,
2965 .help = "limit read bytes per second",
2967 .name = "throttling.bps-write",
2968 .type = QEMU_OPT_NUMBER,
2969 .help = "limit write bytes per second",
2971 .name = "throttling.iops-total-max",
2972 .type = QEMU_OPT_NUMBER,
2973 .help = "I/O operations burst",
2975 .name = "throttling.iops-read-max",
2976 .type = QEMU_OPT_NUMBER,
2977 .help = "I/O operations read burst",
2979 .name = "throttling.iops-write-max",
2980 .type = QEMU_OPT_NUMBER,
2981 .help = "I/O operations write burst",
2983 .name = "throttling.bps-total-max",
2984 .type = QEMU_OPT_NUMBER,
2985 .help = "total bytes burst",
2987 .name = "throttling.bps-read-max",
2988 .type = QEMU_OPT_NUMBER,
2989 .help = "total bytes read burst",
2991 .name = "throttling.bps-write-max",
2992 .type = QEMU_OPT_NUMBER,
2993 .help = "total bytes write burst",
2995 .name = "throttling.iops-size",
2996 .type = QEMU_OPT_NUMBER,
2997 .help = "when limiting by iops max size of an I/O in bytes",
2999 .name = "copy-on-read",
3000 .type = QEMU_OPT_BOOL,
3001 .help = "copy read data from backing file into image file",
3003 .name = "detect-zeroes",
3004 .type = QEMU_OPT_STRING,
3005 .help = "try to optimize zero writes (off, on, unmap)",
3007 { /* end of list */ }
3011 QemuOptsList qemu_drive_opts = {
3012 .name = "drive",
3013 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3014 .desc = {
3016 * no elements => accept any params
3017 * validation will happen later
3019 { /* end of list */ }