kvm_stat: Add RESET support for perf event ioctl
[qemu/ar7.git] / blockdev.c
blobd59efd3f157e85f557c2043b27b943963ae855b6
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 int ret;
358 Error *error = NULL;
359 QemuOpts *opts;
360 const char *id;
361 bool has_driver_specific_opts;
362 BlockdevDetectZeroesOptions detect_zeroes;
363 BlockDriver *drv = NULL;
365 /* Check common options by copying from bs_opts to opts, all other options
366 * stay in bs_opts for processing by bdrv_open(). */
367 id = qdict_get_try_str(bs_opts, "id");
368 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
369 if (error) {
370 error_propagate(errp, error);
371 goto err_no_opts;
374 qemu_opts_absorb_qdict(opts, bs_opts, &error);
375 if (error) {
376 error_propagate(errp, error);
377 goto early_err;
380 if (id) {
381 qdict_del(bs_opts, "id");
384 has_driver_specific_opts = !!qdict_size(bs_opts);
386 /* extract parameters */
387 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
388 ro = qemu_opt_get_bool(opts, "read-only", 0);
389 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
391 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
392 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
393 error_setg(errp, "invalid discard option");
394 goto early_err;
398 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
399 bdrv_flags |= BDRV_O_CACHE_WB;
401 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
402 bdrv_flags |= BDRV_O_NOCACHE;
404 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
405 bdrv_flags |= BDRV_O_NO_FLUSH;
408 #ifdef CONFIG_LINUX_AIO
409 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
410 if (!strcmp(buf, "native")) {
411 bdrv_flags |= BDRV_O_NATIVE_AIO;
412 } else if (!strcmp(buf, "threads")) {
413 /* this is the default */
414 } else {
415 error_setg(errp, "invalid aio option");
416 goto early_err;
419 #endif
421 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
422 if (is_help_option(buf)) {
423 error_printf("Supported formats:");
424 bdrv_iterate_format(bdrv_format_print, NULL);
425 error_printf("\n");
426 goto early_err;
429 drv = bdrv_find_format(buf);
430 if (!drv) {
431 error_setg(errp, "'%s' invalid format", buf);
432 goto early_err;
436 /* disk I/O throttling */
437 memset(&cfg, 0, sizeof(cfg));
438 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
439 qemu_opt_get_number(opts, "throttling.bps-total", 0);
440 cfg.buckets[THROTTLE_BPS_READ].avg =
441 qemu_opt_get_number(opts, "throttling.bps-read", 0);
442 cfg.buckets[THROTTLE_BPS_WRITE].avg =
443 qemu_opt_get_number(opts, "throttling.bps-write", 0);
444 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
445 qemu_opt_get_number(opts, "throttling.iops-total", 0);
446 cfg.buckets[THROTTLE_OPS_READ].avg =
447 qemu_opt_get_number(opts, "throttling.iops-read", 0);
448 cfg.buckets[THROTTLE_OPS_WRITE].avg =
449 qemu_opt_get_number(opts, "throttling.iops-write", 0);
451 cfg.buckets[THROTTLE_BPS_TOTAL].max =
452 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
453 cfg.buckets[THROTTLE_BPS_READ].max =
454 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
455 cfg.buckets[THROTTLE_BPS_WRITE].max =
456 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
457 cfg.buckets[THROTTLE_OPS_TOTAL].max =
458 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
459 cfg.buckets[THROTTLE_OPS_READ].max =
460 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
461 cfg.buckets[THROTTLE_OPS_WRITE].max =
462 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
464 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
466 if (!check_throttle_config(&cfg, &error)) {
467 error_propagate(errp, error);
468 goto early_err;
471 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
472 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
473 on_write_error = parse_block_error_action(buf, 0, &error);
474 if (error) {
475 error_propagate(errp, error);
476 goto early_err;
480 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
481 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
482 on_read_error = parse_block_error_action(buf, 1, &error);
483 if (error) {
484 error_propagate(errp, error);
485 goto early_err;
489 detect_zeroes =
490 qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
491 qemu_opt_get(opts, "detect-zeroes"),
492 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
493 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
494 &error);
495 if (error) {
496 error_propagate(errp, error);
497 goto early_err;
500 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
501 !(bdrv_flags & BDRV_O_UNMAP)) {
502 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
503 "without setting discard operation to unmap");
504 goto early_err;
507 /* init */
508 blk = blk_new_with_bs(qemu_opts_id(opts), errp);
509 if (!blk) {
510 goto early_err;
512 bs = blk_bs(blk);
513 bs->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
514 bs->read_only = ro;
515 bs->detect_zeroes = detect_zeroes;
517 bdrv_set_on_error(bs, on_read_error, on_write_error);
519 /* disk I/O throttling */
520 if (throttle_enabled(&cfg)) {
521 bdrv_io_limits_enable(bs);
522 bdrv_set_io_limits(bs, &cfg);
525 if (!file || !*file) {
526 if (has_driver_specific_opts) {
527 file = NULL;
528 } else {
529 QDECREF(bs_opts);
530 qemu_opts_del(opts);
531 return blk;
534 if (snapshot) {
535 /* always use cache=unsafe with snapshot */
536 bdrv_flags &= ~BDRV_O_CACHE_MASK;
537 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
540 if (copy_on_read) {
541 bdrv_flags |= BDRV_O_COPY_ON_READ;
544 if (runstate_check(RUN_STATE_INMIGRATE)) {
545 bdrv_flags |= BDRV_O_INCOMING;
548 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
550 QINCREF(bs_opts);
551 ret = bdrv_open(&bs, file, NULL, bs_opts, bdrv_flags, drv, &error);
552 assert(bs == blk_bs(blk));
554 if (ret < 0) {
555 error_setg(errp, "could not open disk image %s: %s",
556 file ?: blk_name(blk), error_get_pretty(error));
557 error_free(error);
558 goto err;
561 if (bdrv_key_required(bs)) {
562 autostart = 0;
565 QDECREF(bs_opts);
566 qemu_opts_del(opts);
568 return blk;
570 err:
571 blk_unref(blk);
572 early_err:
573 qemu_opts_del(opts);
574 err_no_opts:
575 QDECREF(bs_opts);
576 return NULL;
579 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
580 Error **errp)
582 const char *value;
584 value = qemu_opt_get(opts, from);
585 if (value) {
586 if (qemu_opt_find(opts, to)) {
587 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
588 "same time", to, from);
589 return;
593 /* rename all items in opts */
594 while ((value = qemu_opt_get(opts, from))) {
595 qemu_opt_set(opts, to, value);
596 qemu_opt_unset(opts, from);
600 QemuOptsList qemu_legacy_drive_opts = {
601 .name = "drive",
602 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
603 .desc = {
605 .name = "bus",
606 .type = QEMU_OPT_NUMBER,
607 .help = "bus number",
609 .name = "unit",
610 .type = QEMU_OPT_NUMBER,
611 .help = "unit number (i.e. lun for scsi)",
613 .name = "index",
614 .type = QEMU_OPT_NUMBER,
615 .help = "index number",
617 .name = "media",
618 .type = QEMU_OPT_STRING,
619 .help = "media type (disk, cdrom)",
621 .name = "if",
622 .type = QEMU_OPT_STRING,
623 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
625 .name = "cyls",
626 .type = QEMU_OPT_NUMBER,
627 .help = "number of cylinders (ide disk geometry)",
629 .name = "heads",
630 .type = QEMU_OPT_NUMBER,
631 .help = "number of heads (ide disk geometry)",
633 .name = "secs",
634 .type = QEMU_OPT_NUMBER,
635 .help = "number of sectors (ide disk geometry)",
637 .name = "trans",
638 .type = QEMU_OPT_STRING,
639 .help = "chs translation (auto, lba, none)",
641 .name = "boot",
642 .type = QEMU_OPT_BOOL,
643 .help = "(deprecated, ignored)",
645 .name = "addr",
646 .type = QEMU_OPT_STRING,
647 .help = "pci address (virtio only)",
649 .name = "serial",
650 .type = QEMU_OPT_STRING,
651 .help = "disk serial number",
653 .name = "file",
654 .type = QEMU_OPT_STRING,
655 .help = "file name",
658 /* Options that are passed on, but have special semantics with -drive */
660 .name = "read-only",
661 .type = QEMU_OPT_BOOL,
662 .help = "open drive file as read-only",
664 .name = "rerror",
665 .type = QEMU_OPT_STRING,
666 .help = "read error action",
668 .name = "werror",
669 .type = QEMU_OPT_STRING,
670 .help = "write error action",
672 .name = "copy-on-read",
673 .type = QEMU_OPT_BOOL,
674 .help = "copy read data from backing file into image file",
677 { /* end of list */ }
681 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
683 const char *value;
684 BlockBackend *blk;
685 DriveInfo *dinfo = NULL;
686 QDict *bs_opts;
687 QemuOpts *legacy_opts;
688 DriveMediaType media = MEDIA_DISK;
689 BlockInterfaceType type;
690 int cyls, heads, secs, translation;
691 int max_devs, bus_id, unit_id, index;
692 const char *devaddr;
693 const char *werror, *rerror;
694 bool read_only = false;
695 bool copy_on_read;
696 const char *serial;
697 const char *filename;
698 Error *local_err = NULL;
699 int i;
701 /* Change legacy command line options into QMP ones */
702 static const struct {
703 const char *from;
704 const char *to;
705 } opt_renames[] = {
706 { "iops", "throttling.iops-total" },
707 { "iops_rd", "throttling.iops-read" },
708 { "iops_wr", "throttling.iops-write" },
710 { "bps", "throttling.bps-total" },
711 { "bps_rd", "throttling.bps-read" },
712 { "bps_wr", "throttling.bps-write" },
714 { "iops_max", "throttling.iops-total-max" },
715 { "iops_rd_max", "throttling.iops-read-max" },
716 { "iops_wr_max", "throttling.iops-write-max" },
718 { "bps_max", "throttling.bps-total-max" },
719 { "bps_rd_max", "throttling.bps-read-max" },
720 { "bps_wr_max", "throttling.bps-write-max" },
722 { "iops_size", "throttling.iops-size" },
724 { "readonly", "read-only" },
727 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
728 qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
729 &local_err);
730 if (local_err) {
731 error_report("%s", error_get_pretty(local_err));
732 error_free(local_err);
733 return NULL;
737 value = qemu_opt_get(all_opts, "cache");
738 if (value) {
739 int flags = 0;
741 if (bdrv_parse_cache_flags(value, &flags) != 0) {
742 error_report("invalid cache option");
743 return NULL;
746 /* Specific options take precedence */
747 if (!qemu_opt_get(all_opts, "cache.writeback")) {
748 qemu_opt_set_bool(all_opts, "cache.writeback",
749 !!(flags & BDRV_O_CACHE_WB));
751 if (!qemu_opt_get(all_opts, "cache.direct")) {
752 qemu_opt_set_bool(all_opts, "cache.direct",
753 !!(flags & BDRV_O_NOCACHE));
755 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
756 qemu_opt_set_bool(all_opts, "cache.no-flush",
757 !!(flags & BDRV_O_NO_FLUSH));
759 qemu_opt_unset(all_opts, "cache");
762 /* Get a QDict for processing the options */
763 bs_opts = qdict_new();
764 qemu_opts_to_qdict(all_opts, bs_opts);
766 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
767 &error_abort);
768 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
769 if (local_err) {
770 error_report("%s", error_get_pretty(local_err));
771 error_free(local_err);
772 goto fail;
775 /* Deprecated option boot=[on|off] */
776 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
777 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
778 "ignored. Future versions will reject this parameter. Please "
779 "update your scripts.\n");
782 /* Media type */
783 value = qemu_opt_get(legacy_opts, "media");
784 if (value) {
785 if (!strcmp(value, "disk")) {
786 media = MEDIA_DISK;
787 } else if (!strcmp(value, "cdrom")) {
788 media = MEDIA_CDROM;
789 read_only = true;
790 } else {
791 error_report("'%s' invalid media", value);
792 goto fail;
796 /* copy-on-read is disabled with a warning for read-only devices */
797 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
798 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
800 if (read_only && copy_on_read) {
801 error_report("warning: disabling copy-on-read on read-only drive");
802 copy_on_read = false;
805 qdict_put(bs_opts, "read-only",
806 qstring_from_str(read_only ? "on" : "off"));
807 qdict_put(bs_opts, "copy-on-read",
808 qstring_from_str(copy_on_read ? "on" :"off"));
810 /* Controller type */
811 value = qemu_opt_get(legacy_opts, "if");
812 if (value) {
813 for (type = 0;
814 type < IF_COUNT && strcmp(value, if_name[type]);
815 type++) {
817 if (type == IF_COUNT) {
818 error_report("unsupported bus type '%s'", value);
819 goto fail;
821 } else {
822 type = block_default_type;
825 /* Geometry */
826 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
827 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
828 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
830 if (cyls || heads || secs) {
831 if (cyls < 1) {
832 error_report("invalid physical cyls number");
833 goto fail;
835 if (heads < 1) {
836 error_report("invalid physical heads number");
837 goto fail;
839 if (secs < 1) {
840 error_report("invalid physical secs number");
841 goto fail;
845 translation = BIOS_ATA_TRANSLATION_AUTO;
846 value = qemu_opt_get(legacy_opts, "trans");
847 if (value != NULL) {
848 if (!cyls) {
849 error_report("'%s' trans must be used with cyls, heads and secs",
850 value);
851 goto fail;
853 if (!strcmp(value, "none")) {
854 translation = BIOS_ATA_TRANSLATION_NONE;
855 } else if (!strcmp(value, "lba")) {
856 translation = BIOS_ATA_TRANSLATION_LBA;
857 } else if (!strcmp(value, "large")) {
858 translation = BIOS_ATA_TRANSLATION_LARGE;
859 } else if (!strcmp(value, "rechs")) {
860 translation = BIOS_ATA_TRANSLATION_RECHS;
861 } else if (!strcmp(value, "auto")) {
862 translation = BIOS_ATA_TRANSLATION_AUTO;
863 } else {
864 error_report("'%s' invalid translation type", value);
865 goto fail;
869 if (media == MEDIA_CDROM) {
870 if (cyls || secs || heads) {
871 error_report("CHS can't be set with media=cdrom");
872 goto fail;
876 /* Device address specified by bus/unit or index.
877 * If none was specified, try to find the first free one. */
878 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
879 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
880 index = qemu_opt_get_number(legacy_opts, "index", -1);
882 max_devs = if_max_devs[type];
884 if (index != -1) {
885 if (bus_id != 0 || unit_id != -1) {
886 error_report("index cannot be used with bus and unit");
887 goto fail;
889 bus_id = drive_index_to_bus_id(type, index);
890 unit_id = drive_index_to_unit_id(type, index);
893 if (unit_id == -1) {
894 unit_id = 0;
895 while (drive_get(type, bus_id, unit_id) != NULL) {
896 unit_id++;
897 if (max_devs && unit_id >= max_devs) {
898 unit_id -= max_devs;
899 bus_id++;
904 if (max_devs && unit_id >= max_devs) {
905 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
906 goto fail;
909 if (drive_get(type, bus_id, unit_id) != NULL) {
910 error_report("drive with bus=%d, unit=%d (index=%d) exists",
911 bus_id, unit_id, index);
912 goto fail;
915 /* Serial number */
916 serial = qemu_opt_get(legacy_opts, "serial");
918 /* no id supplied -> create one */
919 if (qemu_opts_id(all_opts) == NULL) {
920 char *new_id;
921 const char *mediastr = "";
922 if (type == IF_IDE || type == IF_SCSI) {
923 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
925 if (max_devs) {
926 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
927 mediastr, unit_id);
928 } else {
929 new_id = g_strdup_printf("%s%s%i", if_name[type],
930 mediastr, unit_id);
932 qdict_put(bs_opts, "id", qstring_from_str(new_id));
933 g_free(new_id);
936 /* Add virtio block device */
937 devaddr = qemu_opt_get(legacy_opts, "addr");
938 if (devaddr && type != IF_VIRTIO) {
939 error_report("addr is not supported by this bus type");
940 goto fail;
943 if (type == IF_VIRTIO) {
944 QemuOpts *devopts;
945 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
946 &error_abort);
947 if (arch_type == QEMU_ARCH_S390X) {
948 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
949 } else {
950 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
952 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
953 if (devaddr) {
954 qemu_opt_set(devopts, "addr", devaddr);
958 filename = qemu_opt_get(legacy_opts, "file");
960 /* Check werror/rerror compatibility with if=... */
961 werror = qemu_opt_get(legacy_opts, "werror");
962 if (werror != NULL) {
963 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
964 type != IF_NONE) {
965 error_report("werror is not supported by this bus type");
966 goto fail;
968 qdict_put(bs_opts, "werror", qstring_from_str(werror));
971 rerror = qemu_opt_get(legacy_opts, "rerror");
972 if (rerror != NULL) {
973 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
974 type != IF_NONE) {
975 error_report("rerror is not supported by this bus type");
976 goto fail;
978 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
981 /* Actual block device init: Functionality shared with blockdev-add */
982 blk = blockdev_init(filename, bs_opts, &local_err);
983 bs_opts = NULL;
984 if (!blk) {
985 if (local_err) {
986 error_report("%s", error_get_pretty(local_err));
987 error_free(local_err);
989 goto fail;
990 } else {
991 assert(!local_err);
994 /* Create legacy DriveInfo */
995 dinfo = g_malloc0(sizeof(*dinfo));
996 dinfo->opts = all_opts;
998 dinfo->cyls = cyls;
999 dinfo->heads = heads;
1000 dinfo->secs = secs;
1001 dinfo->trans = translation;
1003 dinfo->type = type;
1004 dinfo->bus = bus_id;
1005 dinfo->unit = unit_id;
1006 dinfo->devaddr = devaddr;
1007 dinfo->serial = g_strdup(serial);
1009 blk_set_legacy_dinfo(blk, dinfo);
1011 switch(type) {
1012 case IF_IDE:
1013 case IF_SCSI:
1014 case IF_XEN:
1015 case IF_NONE:
1016 dinfo->media_cd = media == MEDIA_CDROM;
1017 break;
1018 default:
1019 break;
1022 fail:
1023 qemu_opts_del(legacy_opts);
1024 QDECREF(bs_opts);
1025 return dinfo;
1028 void do_commit(Monitor *mon, const QDict *qdict)
1030 const char *device = qdict_get_str(qdict, "device");
1031 BlockDriverState *bs;
1032 int ret;
1034 if (!strcmp(device, "all")) {
1035 ret = bdrv_commit_all();
1036 } else {
1037 bs = bdrv_find(device);
1038 if (!bs) {
1039 monitor_printf(mon, "Device '%s' not found\n", device);
1040 return;
1042 ret = bdrv_commit(bs);
1044 if (ret < 0) {
1045 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1046 strerror(-ret));
1050 static void blockdev_do_action(int kind, void *data, Error **errp)
1052 TransactionAction action;
1053 TransactionActionList list;
1055 action.kind = kind;
1056 action.data = data;
1057 list.value = &action;
1058 list.next = NULL;
1059 qmp_transaction(&list, errp);
1062 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1063 bool has_node_name, const char *node_name,
1064 const char *snapshot_file,
1065 bool has_snapshot_node_name,
1066 const char *snapshot_node_name,
1067 bool has_format, const char *format,
1068 bool has_mode, NewImageMode mode, Error **errp)
1070 BlockdevSnapshot snapshot = {
1071 .has_device = has_device,
1072 .device = (char *) device,
1073 .has_node_name = has_node_name,
1074 .node_name = (char *) node_name,
1075 .snapshot_file = (char *) snapshot_file,
1076 .has_snapshot_node_name = has_snapshot_node_name,
1077 .snapshot_node_name = (char *) snapshot_node_name,
1078 .has_format = has_format,
1079 .format = (char *) format,
1080 .has_mode = has_mode,
1081 .mode = mode,
1083 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1084 &snapshot, errp);
1087 void qmp_blockdev_snapshot_internal_sync(const char *device,
1088 const char *name,
1089 Error **errp)
1091 BlockdevSnapshotInternal snapshot = {
1092 .device = (char *) device,
1093 .name = (char *) name
1096 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1097 &snapshot, errp);
1100 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1101 bool has_id,
1102 const char *id,
1103 bool has_name,
1104 const char *name,
1105 Error **errp)
1107 BlockDriverState *bs = bdrv_find(device);
1108 AioContext *aio_context;
1109 QEMUSnapshotInfo sn;
1110 Error *local_err = NULL;
1111 SnapshotInfo *info = NULL;
1112 int ret;
1114 if (!bs) {
1115 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1116 return NULL;
1119 if (!has_id) {
1120 id = NULL;
1123 if (!has_name) {
1124 name = NULL;
1127 if (!id && !name) {
1128 error_setg(errp, "Name or id must be provided");
1129 return NULL;
1132 aio_context = bdrv_get_aio_context(bs);
1133 aio_context_acquire(aio_context);
1135 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1136 goto out_aio_context;
1139 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1140 if (local_err) {
1141 error_propagate(errp, local_err);
1142 goto out_aio_context;
1144 if (!ret) {
1145 error_setg(errp,
1146 "Snapshot with id '%s' and name '%s' does not exist on "
1147 "device '%s'",
1148 STR_OR_NULL(id), STR_OR_NULL(name), device);
1149 goto out_aio_context;
1152 bdrv_snapshot_delete(bs, id, name, &local_err);
1153 if (local_err) {
1154 error_propagate(errp, local_err);
1155 goto out_aio_context;
1158 aio_context_release(aio_context);
1160 info = g_new0(SnapshotInfo, 1);
1161 info->id = g_strdup(sn.id_str);
1162 info->name = g_strdup(sn.name);
1163 info->date_nsec = sn.date_nsec;
1164 info->date_sec = sn.date_sec;
1165 info->vm_state_size = sn.vm_state_size;
1166 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1167 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1169 return info;
1171 out_aio_context:
1172 aio_context_release(aio_context);
1173 return NULL;
1176 /* New and old BlockDriverState structs for atomic group operations */
1178 typedef struct BlkTransactionState BlkTransactionState;
1180 /* Only prepare() may fail. In a single transaction, only one of commit() or
1181 abort() will be called, clean() will always be called if it present. */
1182 typedef struct BdrvActionOps {
1183 /* Size of state struct, in bytes. */
1184 size_t instance_size;
1185 /* Prepare the work, must NOT be NULL. */
1186 void (*prepare)(BlkTransactionState *common, Error **errp);
1187 /* Commit the changes, can be NULL. */
1188 void (*commit)(BlkTransactionState *common);
1189 /* Abort the changes on fail, can be NULL. */
1190 void (*abort)(BlkTransactionState *common);
1191 /* Clean up resource in the end, can be NULL. */
1192 void (*clean)(BlkTransactionState *common);
1193 } BdrvActionOps;
1196 * This structure must be arranged as first member in child type, assuming
1197 * that compiler will also arrange it to the same address with parent instance.
1198 * Later it will be used in free().
1200 struct BlkTransactionState {
1201 TransactionAction *action;
1202 const BdrvActionOps *ops;
1203 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1206 /* internal snapshot private data */
1207 typedef struct InternalSnapshotState {
1208 BlkTransactionState common;
1209 BlockDriverState *bs;
1210 AioContext *aio_context;
1211 QEMUSnapshotInfo sn;
1212 } InternalSnapshotState;
1214 static void internal_snapshot_prepare(BlkTransactionState *common,
1215 Error **errp)
1217 Error *local_err = NULL;
1218 const char *device;
1219 const char *name;
1220 BlockDriverState *bs;
1221 QEMUSnapshotInfo old_sn, *sn;
1222 bool ret;
1223 qemu_timeval tv;
1224 BlockdevSnapshotInternal *internal;
1225 InternalSnapshotState *state;
1226 int ret1;
1228 g_assert(common->action->kind ==
1229 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1230 internal = common->action->blockdev_snapshot_internal_sync;
1231 state = DO_UPCAST(InternalSnapshotState, common, common);
1233 /* 1. parse input */
1234 device = internal->device;
1235 name = internal->name;
1237 /* 2. check for validation */
1238 bs = bdrv_find(device);
1239 if (!bs) {
1240 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1241 return;
1244 /* AioContext is released in .clean() */
1245 state->aio_context = bdrv_get_aio_context(bs);
1246 aio_context_acquire(state->aio_context);
1248 if (!bdrv_is_inserted(bs)) {
1249 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1250 return;
1253 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1254 return;
1257 if (bdrv_is_read_only(bs)) {
1258 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1259 return;
1262 if (!bdrv_can_snapshot(bs)) {
1263 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1264 bs->drv->format_name, device, "internal snapshot");
1265 return;
1268 if (!strlen(name)) {
1269 error_setg(errp, "Name is empty");
1270 return;
1273 /* check whether a snapshot with name exist */
1274 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1275 &local_err);
1276 if (local_err) {
1277 error_propagate(errp, local_err);
1278 return;
1279 } else if (ret) {
1280 error_setg(errp,
1281 "Snapshot with name '%s' already exists on device '%s'",
1282 name, device);
1283 return;
1286 /* 3. take the snapshot */
1287 sn = &state->sn;
1288 pstrcpy(sn->name, sizeof(sn->name), name);
1289 qemu_gettimeofday(&tv);
1290 sn->date_sec = tv.tv_sec;
1291 sn->date_nsec = tv.tv_usec * 1000;
1292 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1294 ret1 = bdrv_snapshot_create(bs, sn);
1295 if (ret1 < 0) {
1296 error_setg_errno(errp, -ret1,
1297 "Failed to create snapshot '%s' on device '%s'",
1298 name, device);
1299 return;
1302 /* 4. succeed, mark a snapshot is created */
1303 state->bs = bs;
1306 static void internal_snapshot_abort(BlkTransactionState *common)
1308 InternalSnapshotState *state =
1309 DO_UPCAST(InternalSnapshotState, common, common);
1310 BlockDriverState *bs = state->bs;
1311 QEMUSnapshotInfo *sn = &state->sn;
1312 Error *local_error = NULL;
1314 if (!bs) {
1315 return;
1318 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1319 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1320 "device '%s' in abort: %s",
1321 sn->id_str,
1322 sn->name,
1323 bdrv_get_device_name(bs),
1324 error_get_pretty(local_error));
1325 error_free(local_error);
1329 static void internal_snapshot_clean(BlkTransactionState *common)
1331 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1332 common, common);
1334 if (state->aio_context) {
1335 aio_context_release(state->aio_context);
1339 /* external snapshot private data */
1340 typedef struct ExternalSnapshotState {
1341 BlkTransactionState common;
1342 BlockDriverState *old_bs;
1343 BlockDriverState *new_bs;
1344 AioContext *aio_context;
1345 } ExternalSnapshotState;
1347 static void external_snapshot_prepare(BlkTransactionState *common,
1348 Error **errp)
1350 BlockDriver *drv;
1351 int flags, ret;
1352 QDict *options = NULL;
1353 Error *local_err = NULL;
1354 bool has_device = false;
1355 const char *device;
1356 bool has_node_name = false;
1357 const char *node_name;
1358 bool has_snapshot_node_name = false;
1359 const char *snapshot_node_name;
1360 const char *new_image_file;
1361 const char *format = "qcow2";
1362 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1363 ExternalSnapshotState *state =
1364 DO_UPCAST(ExternalSnapshotState, common, common);
1365 TransactionAction *action = common->action;
1367 /* get parameters */
1368 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1370 has_device = action->blockdev_snapshot_sync->has_device;
1371 device = action->blockdev_snapshot_sync->device;
1372 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1373 node_name = action->blockdev_snapshot_sync->node_name;
1374 has_snapshot_node_name =
1375 action->blockdev_snapshot_sync->has_snapshot_node_name;
1376 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1378 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1379 if (action->blockdev_snapshot_sync->has_format) {
1380 format = action->blockdev_snapshot_sync->format;
1382 if (action->blockdev_snapshot_sync->has_mode) {
1383 mode = action->blockdev_snapshot_sync->mode;
1386 /* start processing */
1387 drv = bdrv_find_format(format);
1388 if (!drv) {
1389 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1390 return;
1393 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1394 has_node_name ? node_name : NULL,
1395 &local_err);
1396 if (local_err) {
1397 error_propagate(errp, local_err);
1398 return;
1401 if (has_node_name && !has_snapshot_node_name) {
1402 error_setg(errp, "New snapshot node name missing");
1403 return;
1406 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1407 error_setg(errp, "New snapshot node name already existing");
1408 return;
1411 /* Acquire AioContext now so any threads operating on old_bs stop */
1412 state->aio_context = bdrv_get_aio_context(state->old_bs);
1413 aio_context_acquire(state->aio_context);
1415 if (!bdrv_is_inserted(state->old_bs)) {
1416 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1417 return;
1420 if (bdrv_op_is_blocked(state->old_bs,
1421 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1422 return;
1425 if (!bdrv_is_read_only(state->old_bs)) {
1426 if (bdrv_flush(state->old_bs)) {
1427 error_set(errp, QERR_IO_ERROR);
1428 return;
1432 if (!bdrv_is_first_non_filter(state->old_bs)) {
1433 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1434 return;
1437 flags = state->old_bs->open_flags;
1439 /* create new image w/backing file */
1440 if (mode != NEW_IMAGE_MODE_EXISTING) {
1441 bdrv_img_create(new_image_file, format,
1442 state->old_bs->filename,
1443 state->old_bs->drv->format_name,
1444 NULL, -1, flags, &local_err, false);
1445 if (local_err) {
1446 error_propagate(errp, local_err);
1447 return;
1451 if (has_snapshot_node_name) {
1452 options = qdict_new();
1453 qdict_put(options, "node-name",
1454 qstring_from_str(snapshot_node_name));
1457 /* TODO Inherit bs->options or only take explicit options with an
1458 * extended QMP command? */
1459 assert(state->new_bs == NULL);
1460 ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1461 flags | BDRV_O_NO_BACKING, drv, &local_err);
1462 /* We will manually add the backing_hd field to the bs later */
1463 if (ret != 0) {
1464 error_propagate(errp, local_err);
1468 static void external_snapshot_commit(BlkTransactionState *common)
1470 ExternalSnapshotState *state =
1471 DO_UPCAST(ExternalSnapshotState, common, common);
1473 bdrv_set_aio_context(state->new_bs, state->aio_context);
1475 /* This removes our old bs and adds the new bs */
1476 bdrv_append(state->new_bs, state->old_bs);
1477 /* We don't need (or want) to use the transactional
1478 * bdrv_reopen_multiple() across all the entries at once, because we
1479 * don't want to abort all of them if one of them fails the reopen */
1480 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1481 NULL);
1483 aio_context_release(state->aio_context);
1486 static void external_snapshot_abort(BlkTransactionState *common)
1488 ExternalSnapshotState *state =
1489 DO_UPCAST(ExternalSnapshotState, common, common);
1490 if (state->new_bs) {
1491 bdrv_unref(state->new_bs);
1493 if (state->aio_context) {
1494 aio_context_release(state->aio_context);
1498 typedef struct DriveBackupState {
1499 BlkTransactionState common;
1500 BlockDriverState *bs;
1501 AioContext *aio_context;
1502 BlockJob *job;
1503 } DriveBackupState;
1505 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1507 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1508 BlockDriverState *bs;
1509 DriveBackup *backup;
1510 Error *local_err = NULL;
1512 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1513 backup = common->action->drive_backup;
1515 bs = bdrv_find(backup->device);
1516 if (!bs) {
1517 error_set(errp, QERR_DEVICE_NOT_FOUND, backup->device);
1518 return;
1521 /* AioContext is released in .clean() */
1522 state->aio_context = bdrv_get_aio_context(bs);
1523 aio_context_acquire(state->aio_context);
1525 qmp_drive_backup(backup->device, backup->target,
1526 backup->has_format, backup->format,
1527 backup->sync,
1528 backup->has_mode, backup->mode,
1529 backup->has_speed, backup->speed,
1530 backup->has_on_source_error, backup->on_source_error,
1531 backup->has_on_target_error, backup->on_target_error,
1532 &local_err);
1533 if (local_err) {
1534 error_propagate(errp, local_err);
1535 return;
1538 state->bs = bs;
1539 state->job = state->bs->job;
1542 static void drive_backup_abort(BlkTransactionState *common)
1544 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1545 BlockDriverState *bs = state->bs;
1547 /* Only cancel if it's the job we started */
1548 if (bs && bs->job && bs->job == state->job) {
1549 block_job_cancel_sync(bs->job);
1553 static void drive_backup_clean(BlkTransactionState *common)
1555 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1557 if (state->aio_context) {
1558 aio_context_release(state->aio_context);
1562 typedef struct BlockdevBackupState {
1563 BlkTransactionState common;
1564 BlockDriverState *bs;
1565 BlockJob *job;
1566 AioContext *aio_context;
1567 } BlockdevBackupState;
1569 static void blockdev_backup_prepare(BlkTransactionState *common, Error **errp)
1571 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1572 BlockdevBackup *backup;
1573 BlockDriverState *bs, *target;
1574 Error *local_err = NULL;
1576 assert(common->action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1577 backup = common->action->blockdev_backup;
1579 bs = bdrv_find(backup->device);
1580 if (!bs) {
1581 error_set(errp, QERR_DEVICE_NOT_FOUND, backup->device);
1582 return;
1585 target = bdrv_find(backup->target);
1586 if (!target) {
1587 error_set(errp, QERR_DEVICE_NOT_FOUND, backup->target);
1588 return;
1591 /* AioContext is released in .clean() */
1592 state->aio_context = bdrv_get_aio_context(bs);
1593 if (state->aio_context != bdrv_get_aio_context(target)) {
1594 state->aio_context = NULL;
1595 error_setg(errp, "Backup between two IO threads is not implemented");
1596 return;
1598 aio_context_acquire(state->aio_context);
1600 qmp_blockdev_backup(backup->device, backup->target,
1601 backup->sync,
1602 backup->has_speed, backup->speed,
1603 backup->has_on_source_error, backup->on_source_error,
1604 backup->has_on_target_error, backup->on_target_error,
1605 &local_err);
1606 if (local_err) {
1607 error_propagate(errp, local_err);
1608 return;
1611 state->bs = bs;
1612 state->job = state->bs->job;
1615 static void blockdev_backup_abort(BlkTransactionState *common)
1617 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1618 BlockDriverState *bs = state->bs;
1620 /* Only cancel if it's the job we started */
1621 if (bs && bs->job && bs->job == state->job) {
1622 block_job_cancel_sync(bs->job);
1626 static void blockdev_backup_clean(BlkTransactionState *common)
1628 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1630 if (state->aio_context) {
1631 aio_context_release(state->aio_context);
1635 static void abort_prepare(BlkTransactionState *common, Error **errp)
1637 error_setg(errp, "Transaction aborted using Abort action");
1640 static void abort_commit(BlkTransactionState *common)
1642 g_assert_not_reached(); /* this action never succeeds */
1645 static const BdrvActionOps actions[] = {
1646 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1647 .instance_size = sizeof(ExternalSnapshotState),
1648 .prepare = external_snapshot_prepare,
1649 .commit = external_snapshot_commit,
1650 .abort = external_snapshot_abort,
1652 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1653 .instance_size = sizeof(DriveBackupState),
1654 .prepare = drive_backup_prepare,
1655 .abort = drive_backup_abort,
1656 .clean = drive_backup_clean,
1658 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
1659 .instance_size = sizeof(BlockdevBackupState),
1660 .prepare = blockdev_backup_prepare,
1661 .abort = blockdev_backup_abort,
1662 .clean = blockdev_backup_clean,
1664 [TRANSACTION_ACTION_KIND_ABORT] = {
1665 .instance_size = sizeof(BlkTransactionState),
1666 .prepare = abort_prepare,
1667 .commit = abort_commit,
1669 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1670 .instance_size = sizeof(InternalSnapshotState),
1671 .prepare = internal_snapshot_prepare,
1672 .abort = internal_snapshot_abort,
1673 .clean = internal_snapshot_clean,
1678 * 'Atomic' group operations. The operations are performed as a set, and if
1679 * any fail then we roll back all operations in the group.
1681 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1683 TransactionActionList *dev_entry = dev_list;
1684 BlkTransactionState *state, *next;
1685 Error *local_err = NULL;
1687 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1688 QSIMPLEQ_INIT(&snap_bdrv_states);
1690 /* drain all i/o before any operations */
1691 bdrv_drain_all();
1693 /* We don't do anything in this loop that commits us to the operations */
1694 while (NULL != dev_entry) {
1695 TransactionAction *dev_info = NULL;
1696 const BdrvActionOps *ops;
1698 dev_info = dev_entry->value;
1699 dev_entry = dev_entry->next;
1701 assert(dev_info->kind < ARRAY_SIZE(actions));
1703 ops = &actions[dev_info->kind];
1704 assert(ops->instance_size > 0);
1706 state = g_malloc0(ops->instance_size);
1707 state->ops = ops;
1708 state->action = dev_info;
1709 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1711 state->ops->prepare(state, &local_err);
1712 if (local_err) {
1713 error_propagate(errp, local_err);
1714 goto delete_and_fail;
1718 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1719 if (state->ops->commit) {
1720 state->ops->commit(state);
1724 /* success */
1725 goto exit;
1727 delete_and_fail:
1728 /* failure, and it is all-or-none; roll back all operations */
1729 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1730 if (state->ops->abort) {
1731 state->ops->abort(state);
1734 exit:
1735 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1736 if (state->ops->clean) {
1737 state->ops->clean(state);
1739 g_free(state);
1744 static void eject_device(BlockBackend *blk, int force, Error **errp)
1746 BlockDriverState *bs = blk_bs(blk);
1747 AioContext *aio_context;
1749 aio_context = bdrv_get_aio_context(bs);
1750 aio_context_acquire(aio_context);
1752 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1753 goto out;
1755 if (!blk_dev_has_removable_media(blk)) {
1756 error_setg(errp, "Device '%s' is not removable",
1757 bdrv_get_device_name(bs));
1758 goto out;
1761 if (blk_dev_is_medium_locked(blk) && !blk_dev_is_tray_open(blk)) {
1762 blk_dev_eject_request(blk, force);
1763 if (!force) {
1764 error_setg(errp, "Device '%s' is locked",
1765 bdrv_get_device_name(bs));
1766 goto out;
1770 bdrv_close(bs);
1772 out:
1773 aio_context_release(aio_context);
1776 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1778 BlockBackend *blk;
1780 blk = blk_by_name(device);
1781 if (!blk) {
1782 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1783 return;
1786 eject_device(blk, force, errp);
1789 void qmp_block_passwd(bool has_device, const char *device,
1790 bool has_node_name, const char *node_name,
1791 const char *password, Error **errp)
1793 Error *local_err = NULL;
1794 BlockDriverState *bs;
1795 AioContext *aio_context;
1796 int err;
1798 bs = bdrv_lookup_bs(has_device ? device : NULL,
1799 has_node_name ? node_name : NULL,
1800 &local_err);
1801 if (local_err) {
1802 error_propagate(errp, local_err);
1803 return;
1806 aio_context = bdrv_get_aio_context(bs);
1807 aio_context_acquire(aio_context);
1809 err = bdrv_set_key(bs, password);
1810 if (err == -EINVAL) {
1811 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1812 goto out;
1813 } else if (err < 0) {
1814 error_set(errp, QERR_INVALID_PASSWORD);
1815 goto out;
1818 out:
1819 aio_context_release(aio_context);
1822 /* Assumes AioContext is held */
1823 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1824 int bdrv_flags, BlockDriver *drv,
1825 const char *password, Error **errp)
1827 Error *local_err = NULL;
1828 int ret;
1830 ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1831 if (ret < 0) {
1832 error_propagate(errp, local_err);
1833 return;
1836 if (bdrv_key_required(bs)) {
1837 if (password) {
1838 if (bdrv_set_key(bs, password) < 0) {
1839 error_set(errp, QERR_INVALID_PASSWORD);
1841 } else {
1842 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1843 bdrv_get_encrypted_filename(bs));
1845 } else if (password) {
1846 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1850 void qmp_change_blockdev(const char *device, const char *filename,
1851 const char *format, Error **errp)
1853 BlockBackend *blk;
1854 BlockDriverState *bs;
1855 AioContext *aio_context;
1856 BlockDriver *drv = NULL;
1857 int bdrv_flags;
1858 Error *err = NULL;
1860 blk = blk_by_name(device);
1861 if (!blk) {
1862 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1863 return;
1865 bs = blk_bs(blk);
1867 aio_context = bdrv_get_aio_context(bs);
1868 aio_context_acquire(aio_context);
1870 if (format) {
1871 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1872 if (!drv) {
1873 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1874 goto out;
1878 eject_device(blk, 0, &err);
1879 if (err) {
1880 error_propagate(errp, err);
1881 goto out;
1884 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1885 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1887 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1889 out:
1890 aio_context_release(aio_context);
1893 /* throttling disk I/O limits */
1894 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1895 int64_t bps_wr,
1896 int64_t iops,
1897 int64_t iops_rd,
1898 int64_t iops_wr,
1899 bool has_bps_max,
1900 int64_t bps_max,
1901 bool has_bps_rd_max,
1902 int64_t bps_rd_max,
1903 bool has_bps_wr_max,
1904 int64_t bps_wr_max,
1905 bool has_iops_max,
1906 int64_t iops_max,
1907 bool has_iops_rd_max,
1908 int64_t iops_rd_max,
1909 bool has_iops_wr_max,
1910 int64_t iops_wr_max,
1911 bool has_iops_size,
1912 int64_t iops_size, Error **errp)
1914 ThrottleConfig cfg;
1915 BlockDriverState *bs;
1916 AioContext *aio_context;
1918 bs = bdrv_find(device);
1919 if (!bs) {
1920 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1921 return;
1924 memset(&cfg, 0, sizeof(cfg));
1925 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1926 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1927 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1929 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1930 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1931 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1933 if (has_bps_max) {
1934 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1936 if (has_bps_rd_max) {
1937 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1939 if (has_bps_wr_max) {
1940 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1942 if (has_iops_max) {
1943 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1945 if (has_iops_rd_max) {
1946 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1948 if (has_iops_wr_max) {
1949 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1952 if (has_iops_size) {
1953 cfg.op_size = iops_size;
1956 if (!check_throttle_config(&cfg, errp)) {
1957 return;
1960 aio_context = bdrv_get_aio_context(bs);
1961 aio_context_acquire(aio_context);
1963 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1964 bdrv_io_limits_enable(bs);
1965 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1966 bdrv_io_limits_disable(bs);
1969 if (bs->io_limits_enabled) {
1970 bdrv_set_io_limits(bs, &cfg);
1973 aio_context_release(aio_context);
1976 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1978 const char *id = qdict_get_str(qdict, "id");
1979 BlockBackend *blk;
1980 BlockDriverState *bs;
1981 AioContext *aio_context;
1982 Error *local_err = NULL;
1984 blk = blk_by_name(id);
1985 if (!blk) {
1986 error_report("Device '%s' not found", id);
1987 return -1;
1989 bs = blk_bs(blk);
1991 if (!blk_legacy_dinfo(blk)) {
1992 error_report("Deleting device added with blockdev-add"
1993 " is not supported");
1994 return -1;
1997 aio_context = bdrv_get_aio_context(bs);
1998 aio_context_acquire(aio_context);
2000 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2001 error_report("%s", error_get_pretty(local_err));
2002 error_free(local_err);
2003 aio_context_release(aio_context);
2004 return -1;
2007 /* quiesce block driver; prevent further io */
2008 bdrv_drain_all();
2009 bdrv_flush(bs);
2010 bdrv_close(bs);
2012 /* if we have a device attached to this BlockDriverState
2013 * then we need to make the drive anonymous until the device
2014 * can be removed. If this is a drive with no device backing
2015 * then we can just get rid of the block driver state right here.
2017 if (blk_get_attached_dev(blk)) {
2018 blk_hide_on_behalf_of_do_drive_del(blk);
2019 /* Further I/O must not pause the guest */
2020 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
2021 BLOCKDEV_ON_ERROR_REPORT);
2022 } else {
2023 blk_unref(blk);
2026 aio_context_release(aio_context);
2027 return 0;
2030 void qmp_block_resize(bool has_device, const char *device,
2031 bool has_node_name, const char *node_name,
2032 int64_t size, Error **errp)
2034 Error *local_err = NULL;
2035 BlockDriverState *bs;
2036 AioContext *aio_context;
2037 int ret;
2039 bs = bdrv_lookup_bs(has_device ? device : NULL,
2040 has_node_name ? node_name : NULL,
2041 &local_err);
2042 if (local_err) {
2043 error_propagate(errp, local_err);
2044 return;
2047 aio_context = bdrv_get_aio_context(bs);
2048 aio_context_acquire(aio_context);
2050 if (!bdrv_is_first_non_filter(bs)) {
2051 error_set(errp, QERR_FEATURE_DISABLED, "resize");
2052 goto out;
2055 if (size < 0) {
2056 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2057 goto out;
2060 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2061 error_set(errp, QERR_DEVICE_IN_USE, device);
2062 goto out;
2065 /* complete all in-flight operations before resizing the device */
2066 bdrv_drain_all();
2068 ret = bdrv_truncate(bs, size);
2069 switch (ret) {
2070 case 0:
2071 break;
2072 case -ENOMEDIUM:
2073 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2074 break;
2075 case -ENOTSUP:
2076 error_set(errp, QERR_UNSUPPORTED);
2077 break;
2078 case -EACCES:
2079 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
2080 break;
2081 case -EBUSY:
2082 error_set(errp, QERR_DEVICE_IN_USE, device);
2083 break;
2084 default:
2085 error_setg_errno(errp, -ret, "Could not resize");
2086 break;
2089 out:
2090 aio_context_release(aio_context);
2093 static void block_job_cb(void *opaque, int ret)
2095 /* Note that this function may be executed from another AioContext besides
2096 * the QEMU main loop. If you need to access anything that assumes the
2097 * QEMU global mutex, use a BH or introduce a mutex.
2100 BlockDriverState *bs = opaque;
2101 const char *msg = NULL;
2103 trace_block_job_cb(bs, bs->job, ret);
2105 assert(bs->job);
2107 if (ret < 0) {
2108 msg = strerror(-ret);
2111 if (block_job_is_cancelled(bs->job)) {
2112 block_job_event_cancelled(bs->job);
2113 } else {
2114 block_job_event_completed(bs->job, msg);
2117 bdrv_put_ref_bh_schedule(bs);
2120 void qmp_block_stream(const char *device,
2121 bool has_base, const char *base,
2122 bool has_backing_file, const char *backing_file,
2123 bool has_speed, int64_t speed,
2124 bool has_on_error, BlockdevOnError on_error,
2125 Error **errp)
2127 BlockDriverState *bs;
2128 BlockDriverState *base_bs = NULL;
2129 AioContext *aio_context;
2130 Error *local_err = NULL;
2131 const char *base_name = NULL;
2133 if (!has_on_error) {
2134 on_error = BLOCKDEV_ON_ERROR_REPORT;
2137 bs = bdrv_find(device);
2138 if (!bs) {
2139 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2140 return;
2143 aio_context = bdrv_get_aio_context(bs);
2144 aio_context_acquire(aio_context);
2146 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
2147 goto out;
2150 if (has_base) {
2151 base_bs = bdrv_find_backing_image(bs, base);
2152 if (base_bs == NULL) {
2153 error_set(errp, QERR_BASE_NOT_FOUND, base);
2154 goto out;
2156 assert(bdrv_get_aio_context(base_bs) == aio_context);
2157 base_name = base;
2160 /* if we are streaming the entire chain, the result will have no backing
2161 * file, and specifying one is therefore an error */
2162 if (base_bs == NULL && has_backing_file) {
2163 error_setg(errp, "backing file specified, but streaming the "
2164 "entire chain");
2165 goto out;
2168 /* backing_file string overrides base bs filename */
2169 base_name = has_backing_file ? backing_file : base_name;
2171 stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2172 on_error, block_job_cb, bs, &local_err);
2173 if (local_err) {
2174 error_propagate(errp, local_err);
2175 goto out;
2178 trace_qmp_block_stream(bs, bs->job);
2180 out:
2181 aio_context_release(aio_context);
2184 void qmp_block_commit(const char *device,
2185 bool has_base, const char *base,
2186 bool has_top, const char *top,
2187 bool has_backing_file, const char *backing_file,
2188 bool has_speed, int64_t speed,
2189 Error **errp)
2191 BlockDriverState *bs;
2192 BlockDriverState *base_bs, *top_bs;
2193 AioContext *aio_context;
2194 Error *local_err = NULL;
2195 /* This will be part of the QMP command, if/when the
2196 * BlockdevOnError change for blkmirror makes it in
2198 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2200 if (!has_speed) {
2201 speed = 0;
2204 /* Important Note:
2205 * libvirt relies on the DeviceNotFound error class in order to probe for
2206 * live commit feature versions; for this to work, we must make sure to
2207 * perform the device lookup before any generic errors that may occur in a
2208 * scenario in which all optional arguments are omitted. */
2209 bs = bdrv_find(device);
2210 if (!bs) {
2211 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2212 return;
2215 aio_context = bdrv_get_aio_context(bs);
2216 aio_context_acquire(aio_context);
2218 /* drain all i/o before commits */
2219 bdrv_drain_all();
2221 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2222 goto out;
2225 /* default top_bs is the active layer */
2226 top_bs = bs;
2228 if (has_top && top) {
2229 if (strcmp(bs->filename, top) != 0) {
2230 top_bs = bdrv_find_backing_image(bs, top);
2234 if (top_bs == NULL) {
2235 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2236 goto out;
2239 assert(bdrv_get_aio_context(top_bs) == aio_context);
2241 if (has_base && base) {
2242 base_bs = bdrv_find_backing_image(top_bs, base);
2243 } else {
2244 base_bs = bdrv_find_base(top_bs);
2247 if (base_bs == NULL) {
2248 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2249 goto out;
2252 assert(bdrv_get_aio_context(base_bs) == aio_context);
2254 if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2255 goto out;
2258 /* Do not allow attempts to commit an image into itself */
2259 if (top_bs == base_bs) {
2260 error_setg(errp, "cannot commit an image into itself");
2261 goto out;
2264 if (top_bs == bs) {
2265 if (has_backing_file) {
2266 error_setg(errp, "'backing-file' specified,"
2267 " but 'top' is the active layer");
2268 goto out;
2270 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2271 bs, &local_err);
2272 } else {
2273 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2274 has_backing_file ? backing_file : NULL, &local_err);
2276 if (local_err != NULL) {
2277 error_propagate(errp, local_err);
2278 goto out;
2281 out:
2282 aio_context_release(aio_context);
2285 void qmp_drive_backup(const char *device, const char *target,
2286 bool has_format, const char *format,
2287 enum MirrorSyncMode sync,
2288 bool has_mode, enum NewImageMode mode,
2289 bool has_speed, int64_t speed,
2290 bool has_on_source_error, BlockdevOnError on_source_error,
2291 bool has_on_target_error, BlockdevOnError on_target_error,
2292 Error **errp)
2294 BlockDriverState *bs;
2295 BlockDriverState *target_bs;
2296 BlockDriverState *source = NULL;
2297 AioContext *aio_context;
2298 BlockDriver *drv = NULL;
2299 Error *local_err = NULL;
2300 int flags;
2301 int64_t size;
2302 int ret;
2304 if (!has_speed) {
2305 speed = 0;
2307 if (!has_on_source_error) {
2308 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2310 if (!has_on_target_error) {
2311 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2313 if (!has_mode) {
2314 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2317 bs = bdrv_find(device);
2318 if (!bs) {
2319 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2320 return;
2323 aio_context = bdrv_get_aio_context(bs);
2324 aio_context_acquire(aio_context);
2326 /* Although backup_run has this check too, we need to use bs->drv below, so
2327 * do an early check redundantly. */
2328 if (!bdrv_is_inserted(bs)) {
2329 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2330 goto out;
2333 if (!has_format) {
2334 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2336 if (format) {
2337 drv = bdrv_find_format(format);
2338 if (!drv) {
2339 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2340 goto out;
2344 /* Early check to avoid creating target */
2345 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2346 goto out;
2349 flags = bs->open_flags | BDRV_O_RDWR;
2351 /* See if we have a backing HD we can use to create our new image
2352 * on top of. */
2353 if (sync == MIRROR_SYNC_MODE_TOP) {
2354 source = bs->backing_hd;
2355 if (!source) {
2356 sync = MIRROR_SYNC_MODE_FULL;
2359 if (sync == MIRROR_SYNC_MODE_NONE) {
2360 source = bs;
2363 size = bdrv_getlength(bs);
2364 if (size < 0) {
2365 error_setg_errno(errp, -size, "bdrv_getlength failed");
2366 goto out;
2369 if (mode != NEW_IMAGE_MODE_EXISTING) {
2370 assert(format && drv);
2371 if (source) {
2372 bdrv_img_create(target, format, source->filename,
2373 source->drv->format_name, NULL,
2374 size, flags, &local_err, false);
2375 } else {
2376 bdrv_img_create(target, format, NULL, NULL, NULL,
2377 size, flags, &local_err, false);
2381 if (local_err) {
2382 error_propagate(errp, local_err);
2383 goto out;
2386 target_bs = NULL;
2387 ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2388 if (ret < 0) {
2389 error_propagate(errp, local_err);
2390 goto out;
2393 bdrv_set_aio_context(target_bs, aio_context);
2395 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2396 block_job_cb, bs, &local_err);
2397 if (local_err != NULL) {
2398 bdrv_unref(target_bs);
2399 error_propagate(errp, local_err);
2400 goto out;
2403 out:
2404 aio_context_release(aio_context);
2407 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2409 return bdrv_named_nodes_list();
2412 void qmp_blockdev_backup(const char *device, const char *target,
2413 enum MirrorSyncMode sync,
2414 bool has_speed, int64_t speed,
2415 bool has_on_source_error,
2416 BlockdevOnError on_source_error,
2417 bool has_on_target_error,
2418 BlockdevOnError on_target_error,
2419 Error **errp)
2421 BlockDriverState *bs;
2422 BlockDriverState *target_bs;
2423 Error *local_err = NULL;
2424 AioContext *aio_context;
2426 if (!has_speed) {
2427 speed = 0;
2429 if (!has_on_source_error) {
2430 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2432 if (!has_on_target_error) {
2433 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2436 bs = bdrv_find(device);
2437 if (!bs) {
2438 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2439 return;
2442 aio_context = bdrv_get_aio_context(bs);
2443 aio_context_acquire(aio_context);
2445 target_bs = bdrv_find(target);
2446 if (!target_bs) {
2447 error_set(errp, QERR_DEVICE_NOT_FOUND, target);
2448 goto out;
2451 bdrv_ref(target_bs);
2452 bdrv_set_aio_context(target_bs, aio_context);
2453 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2454 block_job_cb, bs, &local_err);
2455 if (local_err != NULL) {
2456 bdrv_unref(target_bs);
2457 error_propagate(errp, local_err);
2459 out:
2460 aio_context_release(aio_context);
2463 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2465 void qmp_drive_mirror(const char *device, const char *target,
2466 bool has_format, const char *format,
2467 bool has_node_name, const char *node_name,
2468 bool has_replaces, const char *replaces,
2469 enum MirrorSyncMode sync,
2470 bool has_mode, enum NewImageMode mode,
2471 bool has_speed, int64_t speed,
2472 bool has_granularity, uint32_t granularity,
2473 bool has_buf_size, int64_t buf_size,
2474 bool has_on_source_error, BlockdevOnError on_source_error,
2475 bool has_on_target_error, BlockdevOnError on_target_error,
2476 Error **errp)
2478 BlockDriverState *bs;
2479 BlockDriverState *source, *target_bs;
2480 AioContext *aio_context;
2481 BlockDriver *drv = NULL;
2482 Error *local_err = NULL;
2483 QDict *options = NULL;
2484 int flags;
2485 int64_t size;
2486 int ret;
2488 if (!has_speed) {
2489 speed = 0;
2491 if (!has_on_source_error) {
2492 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2494 if (!has_on_target_error) {
2495 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2497 if (!has_mode) {
2498 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2500 if (!has_granularity) {
2501 granularity = 0;
2503 if (!has_buf_size) {
2504 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2507 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2508 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2509 "a value in range [512B, 64MB]");
2510 return;
2512 if (granularity & (granularity - 1)) {
2513 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "granularity", "power of 2");
2514 return;
2517 bs = bdrv_find(device);
2518 if (!bs) {
2519 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2520 return;
2523 aio_context = bdrv_get_aio_context(bs);
2524 aio_context_acquire(aio_context);
2526 if (!bdrv_is_inserted(bs)) {
2527 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2528 goto out;
2531 if (!has_format) {
2532 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2534 if (format) {
2535 drv = bdrv_find_format(format);
2536 if (!drv) {
2537 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2538 goto out;
2542 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2543 goto out;
2546 flags = bs->open_flags | BDRV_O_RDWR;
2547 source = bs->backing_hd;
2548 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2549 sync = MIRROR_SYNC_MODE_FULL;
2551 if (sync == MIRROR_SYNC_MODE_NONE) {
2552 source = bs;
2555 size = bdrv_getlength(bs);
2556 if (size < 0) {
2557 error_setg_errno(errp, -size, "bdrv_getlength failed");
2558 goto out;
2561 if (has_replaces) {
2562 BlockDriverState *to_replace_bs;
2563 AioContext *replace_aio_context;
2564 int64_t replace_size;
2566 if (!has_node_name) {
2567 error_setg(errp, "a node-name must be provided when replacing a"
2568 " named node of the graph");
2569 goto out;
2572 to_replace_bs = check_to_replace_node(replaces, &local_err);
2574 if (!to_replace_bs) {
2575 error_propagate(errp, local_err);
2576 goto out;
2579 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
2580 aio_context_acquire(replace_aio_context);
2581 replace_size = bdrv_getlength(to_replace_bs);
2582 aio_context_release(replace_aio_context);
2584 if (size != replace_size) {
2585 error_setg(errp, "cannot replace image with a mirror image of "
2586 "different size");
2587 goto out;
2591 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2592 && mode != NEW_IMAGE_MODE_EXISTING)
2594 /* create new image w/o backing file */
2595 assert(format && drv);
2596 bdrv_img_create(target, format,
2597 NULL, NULL, NULL, size, flags, &local_err, false);
2598 } else {
2599 switch (mode) {
2600 case NEW_IMAGE_MODE_EXISTING:
2601 break;
2602 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2603 /* create new image with backing file */
2604 bdrv_img_create(target, format,
2605 source->filename,
2606 source->drv->format_name,
2607 NULL, size, flags, &local_err, false);
2608 break;
2609 default:
2610 abort();
2614 if (local_err) {
2615 error_propagate(errp, local_err);
2616 goto out;
2619 if (has_node_name) {
2620 options = qdict_new();
2621 qdict_put(options, "node-name", qstring_from_str(node_name));
2624 /* Mirroring takes care of copy-on-write using the source's backing
2625 * file.
2627 target_bs = NULL;
2628 ret = bdrv_open(&target_bs, target, NULL, options,
2629 flags | BDRV_O_NO_BACKING, drv, &local_err);
2630 if (ret < 0) {
2631 error_propagate(errp, local_err);
2632 goto out;
2635 bdrv_set_aio_context(target_bs, aio_context);
2637 /* pass the node name to replace to mirror start since it's loose coupling
2638 * and will allow to check whether the node still exist at mirror completion
2640 mirror_start(bs, target_bs,
2641 has_replaces ? replaces : NULL,
2642 speed, granularity, buf_size, sync,
2643 on_source_error, on_target_error,
2644 block_job_cb, bs, &local_err);
2645 if (local_err != NULL) {
2646 bdrv_unref(target_bs);
2647 error_propagate(errp, local_err);
2648 goto out;
2651 out:
2652 aio_context_release(aio_context);
2655 /* Get the block job for a given device name and acquire its AioContext */
2656 static BlockJob *find_block_job(const char *device, AioContext **aio_context)
2658 BlockDriverState *bs;
2660 bs = bdrv_find(device);
2661 if (!bs) {
2662 goto notfound;
2665 *aio_context = bdrv_get_aio_context(bs);
2666 aio_context_acquire(*aio_context);
2668 if (!bs->job) {
2669 aio_context_release(*aio_context);
2670 goto notfound;
2673 return bs->job;
2675 notfound:
2676 *aio_context = NULL;
2677 return NULL;
2680 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2682 AioContext *aio_context;
2683 BlockJob *job = find_block_job(device, &aio_context);
2685 if (!job) {
2686 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2687 return;
2690 block_job_set_speed(job, speed, errp);
2691 aio_context_release(aio_context);
2694 void qmp_block_job_cancel(const char *device,
2695 bool has_force, bool force, Error **errp)
2697 AioContext *aio_context;
2698 BlockJob *job = find_block_job(device, &aio_context);
2700 if (!job) {
2701 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2702 return;
2705 if (!has_force) {
2706 force = false;
2709 if (job->paused && !force) {
2710 error_setg(errp, "The block job for device '%s' is currently paused",
2711 device);
2712 goto out;
2715 trace_qmp_block_job_cancel(job);
2716 block_job_cancel(job);
2717 out:
2718 aio_context_release(aio_context);
2721 void qmp_block_job_pause(const char *device, Error **errp)
2723 AioContext *aio_context;
2724 BlockJob *job = find_block_job(device, &aio_context);
2726 if (!job) {
2727 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2728 return;
2731 trace_qmp_block_job_pause(job);
2732 block_job_pause(job);
2733 aio_context_release(aio_context);
2736 void qmp_block_job_resume(const char *device, Error **errp)
2738 AioContext *aio_context;
2739 BlockJob *job = find_block_job(device, &aio_context);
2741 if (!job) {
2742 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2743 return;
2746 trace_qmp_block_job_resume(job);
2747 block_job_resume(job);
2748 aio_context_release(aio_context);
2751 void qmp_block_job_complete(const char *device, Error **errp)
2753 AioContext *aio_context;
2754 BlockJob *job = find_block_job(device, &aio_context);
2756 if (!job) {
2757 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2758 return;
2761 trace_qmp_block_job_complete(job);
2762 block_job_complete(job, errp);
2763 aio_context_release(aio_context);
2766 void qmp_change_backing_file(const char *device,
2767 const char *image_node_name,
2768 const char *backing_file,
2769 Error **errp)
2771 BlockDriverState *bs = NULL;
2772 AioContext *aio_context;
2773 BlockDriverState *image_bs = NULL;
2774 Error *local_err = NULL;
2775 bool ro;
2776 int open_flags;
2777 int ret;
2779 /* find the top layer BDS of the chain */
2780 bs = bdrv_find(device);
2781 if (!bs) {
2782 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2783 return;
2786 aio_context = bdrv_get_aio_context(bs);
2787 aio_context_acquire(aio_context);
2789 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
2790 if (local_err) {
2791 error_propagate(errp, local_err);
2792 goto out;
2795 if (!image_bs) {
2796 error_setg(errp, "image file not found");
2797 goto out;
2800 if (bdrv_find_base(image_bs) == image_bs) {
2801 error_setg(errp, "not allowing backing file change on an image "
2802 "without a backing file");
2803 goto out;
2806 /* even though we are not necessarily operating on bs, we need it to
2807 * determine if block ops are currently prohibited on the chain */
2808 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
2809 goto out;
2812 /* final sanity check */
2813 if (!bdrv_chain_contains(bs, image_bs)) {
2814 error_setg(errp, "'%s' and image file are not in the same chain",
2815 device);
2816 goto out;
2819 /* if not r/w, reopen to make r/w */
2820 open_flags = image_bs->open_flags;
2821 ro = bdrv_is_read_only(image_bs);
2823 if (ro) {
2824 bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
2825 if (local_err) {
2826 error_propagate(errp, local_err);
2827 goto out;
2831 ret = bdrv_change_backing_file(image_bs, backing_file,
2832 image_bs->drv ? image_bs->drv->format_name : "");
2834 if (ret < 0) {
2835 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
2836 backing_file);
2837 /* don't exit here, so we can try to restore open flags if
2838 * appropriate */
2841 if (ro) {
2842 bdrv_reopen(image_bs, open_flags, &local_err);
2843 if (local_err) {
2844 error_propagate(errp, local_err); /* will preserve prior errp */
2848 out:
2849 aio_context_release(aio_context);
2852 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2854 QmpOutputVisitor *ov = qmp_output_visitor_new();
2855 BlockBackend *blk;
2856 QObject *obj;
2857 QDict *qdict;
2858 Error *local_err = NULL;
2860 /* Require an ID in the top level */
2861 if (!options->has_id) {
2862 error_setg(errp, "Block device needs an ID");
2863 goto fail;
2866 /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
2867 * cache.direct=false instead of silently switching to aio=threads, except
2868 * when called from drive_new().
2870 * For now, simply forbidding the combination for all drivers will do. */
2871 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2872 bool direct = options->has_cache &&
2873 options->cache->has_direct &&
2874 options->cache->direct;
2875 if (!direct) {
2876 error_setg(errp, "aio=native requires cache.direct=true");
2877 goto fail;
2881 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2882 &options, NULL, &local_err);
2883 if (local_err) {
2884 error_propagate(errp, local_err);
2885 goto fail;
2888 obj = qmp_output_get_qobject(ov);
2889 qdict = qobject_to_qdict(obj);
2891 qdict_flatten(qdict);
2893 blk = blockdev_init(NULL, qdict, &local_err);
2894 if (local_err) {
2895 error_propagate(errp, local_err);
2896 goto fail;
2899 if (bdrv_key_required(blk_bs(blk))) {
2900 blk_unref(blk);
2901 error_setg(errp, "blockdev-add doesn't support encrypted devices");
2902 goto fail;
2905 fail:
2906 qmp_output_visitor_cleanup(ov);
2909 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2911 BlockJobInfoList *head = NULL, **p_next = &head;
2912 BlockDriverState *bs;
2914 for (bs = bdrv_next(NULL); bs; bs = bdrv_next(bs)) {
2915 AioContext *aio_context = bdrv_get_aio_context(bs);
2917 aio_context_acquire(aio_context);
2919 if (bs->job) {
2920 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2921 elem->value = block_job_query(bs->job);
2922 *p_next = elem;
2923 p_next = &elem->next;
2926 aio_context_release(aio_context);
2929 return head;
2932 QemuOptsList qemu_common_drive_opts = {
2933 .name = "drive",
2934 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2935 .desc = {
2937 .name = "snapshot",
2938 .type = QEMU_OPT_BOOL,
2939 .help = "enable/disable snapshot mode",
2941 .name = "discard",
2942 .type = QEMU_OPT_STRING,
2943 .help = "discard operation (ignore/off, unmap/on)",
2945 .name = "cache.writeback",
2946 .type = QEMU_OPT_BOOL,
2947 .help = "enables writeback mode for any caches",
2949 .name = "cache.direct",
2950 .type = QEMU_OPT_BOOL,
2951 .help = "enables use of O_DIRECT (bypass the host page cache)",
2953 .name = "cache.no-flush",
2954 .type = QEMU_OPT_BOOL,
2955 .help = "ignore any flush requests for the device",
2957 .name = "aio",
2958 .type = QEMU_OPT_STRING,
2959 .help = "host AIO implementation (threads, native)",
2961 .name = "format",
2962 .type = QEMU_OPT_STRING,
2963 .help = "disk format (raw, qcow2, ...)",
2965 .name = "rerror",
2966 .type = QEMU_OPT_STRING,
2967 .help = "read error action",
2969 .name = "werror",
2970 .type = QEMU_OPT_STRING,
2971 .help = "write error action",
2973 .name = "read-only",
2974 .type = QEMU_OPT_BOOL,
2975 .help = "open drive file as read-only",
2977 .name = "throttling.iops-total",
2978 .type = QEMU_OPT_NUMBER,
2979 .help = "limit total I/O operations per second",
2981 .name = "throttling.iops-read",
2982 .type = QEMU_OPT_NUMBER,
2983 .help = "limit read operations per second",
2985 .name = "throttling.iops-write",
2986 .type = QEMU_OPT_NUMBER,
2987 .help = "limit write operations per second",
2989 .name = "throttling.bps-total",
2990 .type = QEMU_OPT_NUMBER,
2991 .help = "limit total bytes per second",
2993 .name = "throttling.bps-read",
2994 .type = QEMU_OPT_NUMBER,
2995 .help = "limit read bytes per second",
2997 .name = "throttling.bps-write",
2998 .type = QEMU_OPT_NUMBER,
2999 .help = "limit write bytes per second",
3001 .name = "throttling.iops-total-max",
3002 .type = QEMU_OPT_NUMBER,
3003 .help = "I/O operations burst",
3005 .name = "throttling.iops-read-max",
3006 .type = QEMU_OPT_NUMBER,
3007 .help = "I/O operations read burst",
3009 .name = "throttling.iops-write-max",
3010 .type = QEMU_OPT_NUMBER,
3011 .help = "I/O operations write burst",
3013 .name = "throttling.bps-total-max",
3014 .type = QEMU_OPT_NUMBER,
3015 .help = "total bytes burst",
3017 .name = "throttling.bps-read-max",
3018 .type = QEMU_OPT_NUMBER,
3019 .help = "total bytes read burst",
3021 .name = "throttling.bps-write-max",
3022 .type = QEMU_OPT_NUMBER,
3023 .help = "total bytes write burst",
3025 .name = "throttling.iops-size",
3026 .type = QEMU_OPT_NUMBER,
3027 .help = "when limiting by iops max size of an I/O in bytes",
3029 .name = "copy-on-read",
3030 .type = QEMU_OPT_BOOL,
3031 .help = "copy read data from backing file into image file",
3033 .name = "detect-zeroes",
3034 .type = QEMU_OPT_STRING,
3035 .help = "try to optimize zero writes (off, on, unmap)",
3037 { /* end of list */ }
3041 QemuOptsList qemu_drive_opts = {
3042 .name = "drive",
3043 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3044 .desc = {
3046 * no elements => accept any params
3047 * validation will happen later
3049 { /* end of list */ }