2 * QEMU Block driver for RADOS (Ceph)
4 * Copyright (C) 2010-2011 Christian Brunner <chb@muc.de>,
5 * Josh Durgin <josh.durgin@dreamhost.com>
7 * This work is licensed under the terms of the GNU GPL, version 2. See
8 * the COPYING file in the top-level directory.
10 * Contributions after 2012-01-13 are licensed under the terms of the
11 * GNU GPL, version 2 or (at your option) any later version.
14 #include "qemu/osdep.h"
16 #include <rbd/librbd.h>
17 #include "qapi/error.h"
18 #include "qemu/error-report.h"
19 #include "qemu/option.h"
20 #include "block/block_int.h"
21 #include "crypto/secret.h"
22 #include "qemu/cutils.h"
23 #include "qapi/qmp/qstring.h"
24 #include "qapi/qmp/qdict.h"
25 #include "qapi/qmp/qjson.h"
26 #include "qapi/qmp/qlist.h"
29 * When specifying the image filename use:
31 * rbd:poolname/devicename[@snapshotname][:option1=value1[:option2=value2...]]
33 * poolname must be the name of an existing rados pool.
35 * devicename is the name of the rbd image.
37 * Each option given is used to configure rados, and may be any valid
38 * Ceph option, "id", or "conf".
40 * The "id" option indicates what user we should authenticate as to
41 * the Ceph cluster. If it is excluded we will use the Ceph default
44 * The "conf" option specifies a Ceph configuration file to read. If
45 * it is not specified, we will read from the default Ceph locations
46 * (e.g., /etc/ceph/ceph.conf). To avoid reading _any_ configuration
47 * file, specify conf=/dev/null.
49 * Configuration values containing :, @, or = can be escaped with a
53 /* rbd_aio_discard added in 0.1.2 */
54 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 2)
55 #define LIBRBD_SUPPORTS_DISCARD
57 #undef LIBRBD_SUPPORTS_DISCARD
60 #define OBJ_MAX_SIZE (1UL << OBJ_DEFAULT_OBJ_ORDER)
62 #define RBD_MAX_SNAPS 100
64 /* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */
65 #ifdef LIBRBD_SUPPORTS_IOVEC
66 #define LIBRBD_USE_IOVEC 1
68 #define LIBRBD_USE_IOVEC 0
78 typedef struct RBDAIOCB
{
85 struct BDRVRBDState
*s
;
88 typedef struct RADOSCB
{
90 struct BDRVRBDState
*s
;
96 typedef struct BDRVRBDState
{
104 static char *qemu_rbd_next_tok(char *src
, char delim
, char **p
)
110 for (end
= src
; *end
; ++end
) {
114 if (*end
== '\\' && end
[1] != '\0') {
125 static void qemu_rbd_unescape(char *src
)
129 for (p
= src
; *src
; ++src
, ++p
) {
130 if (*src
== '\\' && src
[1] != '\0') {
138 static void qemu_rbd_parse_filename(const char *filename
, QDict
*options
,
143 QList
*keypairs
= NULL
;
146 if (!strstart(filename
, "rbd:", &start
)) {
147 error_setg(errp
, "File name must start with 'rbd:'");
151 buf
= g_strdup(start
);
154 found_str
= qemu_rbd_next_tok(p
, '/', &p
);
156 error_setg(errp
, "Pool name is required");
159 qemu_rbd_unescape(found_str
);
160 qdict_put_str(options
, "pool", found_str
);
162 if (strchr(p
, '@')) {
163 found_str
= qemu_rbd_next_tok(p
, '@', &p
);
164 qemu_rbd_unescape(found_str
);
165 qdict_put_str(options
, "image", found_str
);
167 found_str
= qemu_rbd_next_tok(p
, ':', &p
);
168 qemu_rbd_unescape(found_str
);
169 qdict_put_str(options
, "snapshot", found_str
);
171 found_str
= qemu_rbd_next_tok(p
, ':', &p
);
172 qemu_rbd_unescape(found_str
);
173 qdict_put_str(options
, "image", found_str
);
179 /* The following are essentially all key/value pairs, and we treat
180 * 'id' and 'conf' a bit special. Key/value pairs may be in any order. */
183 name
= qemu_rbd_next_tok(p
, '=', &p
);
185 error_setg(errp
, "conf option %s has no value", name
);
189 qemu_rbd_unescape(name
);
191 value
= qemu_rbd_next_tok(p
, ':', &p
);
192 qemu_rbd_unescape(value
);
194 if (!strcmp(name
, "conf")) {
195 qdict_put_str(options
, "conf", value
);
196 } else if (!strcmp(name
, "id")) {
197 qdict_put_str(options
, "user", value
);
200 * We pass these internally to qemu_rbd_set_keypairs(), so
201 * we can get away with the simpler list of [ "key1",
202 * "value1", "key2", "value2" ] rather than a raw dict
203 * { "key1": "value1", "key2": "value2" } where we can't
204 * guarantee order, or even a more correct but complex
205 * [ { "key1": "value1" }, { "key2": "value2" } ]
208 keypairs
= qlist_new();
210 qlist_append_str(keypairs
, name
);
211 qlist_append_str(keypairs
, value
);
216 qdict_put(options
, "=keyvalue-pairs",
217 qobject_to_json(QOBJECT(keypairs
)));
227 static int qemu_rbd_set_auth(rados_t cluster
, const char *secretid
,
234 gchar
*secret
= qcrypto_secret_lookup_as_base64(secretid
,
240 rados_conf_set(cluster
, "key", secret
);
246 static int qemu_rbd_set_keypairs(rados_t cluster
, const char *keypairs_json
,
256 if (!keypairs_json
) {
259 keypairs
= qobject_to_qlist(qobject_from_json(keypairs_json
,
261 remaining
= qlist_size(keypairs
) / 2;
264 while (remaining
--) {
265 name
= qobject_to_qstring(qlist_pop(keypairs
));
266 value
= qobject_to_qstring(qlist_pop(keypairs
));
267 assert(name
&& value
);
268 key
= qstring_get_str(name
);
270 ret
= rados_conf_set(cluster
, key
, qstring_get_str(value
));
274 error_setg_errno(errp
, -ret
, "invalid conf option %s", key
);
284 static void qemu_rbd_memset(RADOSCB
*rcb
, int64_t offs
)
286 if (LIBRBD_USE_IOVEC
) {
287 RBDAIOCB
*acb
= rcb
->acb
;
288 iov_memset(acb
->qiov
->iov
, acb
->qiov
->niov
, offs
, 0,
289 acb
->qiov
->size
- offs
);
291 memset(rcb
->buf
+ offs
, 0, rcb
->size
- offs
);
295 static QemuOptsList runtime_opts
= {
297 .head
= QTAILQ_HEAD_INITIALIZER(runtime_opts
.head
),
301 .type
= QEMU_OPT_STRING
,
302 .help
= "Rados pool name",
306 .type
= QEMU_OPT_STRING
,
307 .help
= "Image name in the pool",
311 .type
= QEMU_OPT_STRING
,
312 .help
= "Rados config file location",
316 .type
= QEMU_OPT_STRING
,
317 .help
= "Ceph snapshot name",
320 /* maps to 'id' in rados_create() */
322 .type
= QEMU_OPT_STRING
,
323 .help
= "Rados id name",
326 * server.* extracted manually, see qemu_rbd_mon_host()
329 .name
= "password-secret",
330 .type
= QEMU_OPT_STRING
,
331 .help
= "ID of secret providing the password",
335 * Keys for qemu_rbd_parse_filename(), not in the QAPI schema
339 * HACK: name starts with '=' so that qemu_opts_parse()
342 .name
= "=keyvalue-pairs",
343 .type
= QEMU_OPT_STRING
,
344 .help
= "Legacy rados key/value option parameters",
348 .type
= QEMU_OPT_STRING
,
350 { /* end of list */ }
354 static int qemu_rbd_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
356 Error
*local_err
= NULL
;
360 const char *pool
, *image_name
, *conf
, *user
, *keypairs
;
361 const char *secretid
;
363 rados_ioctx_t io_ctx
;
364 QDict
*options
= NULL
;
367 secretid
= qemu_opt_get(opts
, "password-secret");
369 /* Read out options */
370 bytes
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
372 objsize
= qemu_opt_get_size_del(opts
, BLOCK_OPT_CLUSTER_SIZE
, 0);
374 if ((objsize
- 1) & objsize
) { /* not a power of 2? */
375 error_setg(errp
, "obj size needs to be power of 2");
379 if (objsize
< 4096) {
380 error_setg(errp
, "obj size too small");
384 obj_order
= ctz32(objsize
);
387 options
= qdict_new();
388 qemu_rbd_parse_filename(filename
, options
, &local_err
);
391 error_propagate(errp
, local_err
);
396 * Caution: while qdict_get_try_str() is fine, getting non-string
397 * types would require more care. When @options come from -blockdev
398 * or blockdev_add, its members are typed according to the QAPI
399 * schema, but when they come from -drive, they're all QString.
401 pool
= qdict_get_try_str(options
, "pool");
402 conf
= qdict_get_try_str(options
, "conf");
403 user
= qdict_get_try_str(options
, "user");
404 image_name
= qdict_get_try_str(options
, "image");
405 keypairs
= qdict_get_try_str(options
, "=keyvalue-pairs");
407 ret
= rados_create(&cluster
, user
);
409 error_setg_errno(errp
, -ret
, "error initializing");
413 /* try default location when conf=NULL, but ignore failure */
414 ret
= rados_conf_read_file(cluster
, conf
);
415 if (conf
&& ret
< 0) {
416 error_setg_errno(errp
, -ret
, "error reading conf file %s", conf
);
421 ret
= qemu_rbd_set_keypairs(cluster
, keypairs
, errp
);
427 if (qemu_rbd_set_auth(cluster
, secretid
, errp
) < 0) {
432 ret
= rados_connect(cluster
);
434 error_setg_errno(errp
, -ret
, "error connecting");
438 ret
= rados_ioctx_create(cluster
, pool
, &io_ctx
);
440 error_setg_errno(errp
, -ret
, "error opening pool %s", pool
);
444 ret
= rbd_create(io_ctx
, image_name
, bytes
, &obj_order
);
446 error_setg_errno(errp
, -ret
, "error rbd create");
449 rados_ioctx_destroy(io_ctx
);
452 rados_shutdown(cluster
);
460 * This aio completion is being called from rbd_finish_bh() and runs in qemu
463 static void qemu_rbd_complete_aio(RADOSCB
*rcb
)
465 RBDAIOCB
*acb
= rcb
->acb
;
470 if (acb
->cmd
!= RBD_AIO_READ
) {
474 } else if (!acb
->error
) {
475 acb
->ret
= rcb
->size
;
479 qemu_rbd_memset(rcb
, 0);
482 } else if (r
< rcb
->size
) {
483 qemu_rbd_memset(rcb
, r
);
485 acb
->ret
= rcb
->size
;
487 } else if (!acb
->error
) {
494 if (!LIBRBD_USE_IOVEC
) {
495 if (acb
->cmd
== RBD_AIO_READ
) {
496 qemu_iovec_from_buf(acb
->qiov
, 0, acb
->bounce
, acb
->qiov
->size
);
498 qemu_vfree(acb
->bounce
);
501 acb
->common
.cb(acb
->common
.opaque
, (acb
->ret
> 0 ? 0 : acb
->ret
));
506 static char *qemu_rbd_mon_host(QDict
*options
, Error
**errp
)
508 const char **vals
= g_new(const char *, qdict_size(options
) + 1);
510 const char *host
, *port
;
515 sprintf(keybuf
, "server.%d.host", i
);
516 host
= qdict_get_try_str(options
, keybuf
);
517 qdict_del(options
, keybuf
);
518 sprintf(keybuf
, "server.%d.port", i
);
519 port
= qdict_get_try_str(options
, keybuf
);
520 qdict_del(options
, keybuf
);
521 if (!host
&& !port
) {
525 error_setg(errp
, "Parameter server.%d.host is missing", i
);
530 if (strchr(host
, ':')) {
531 vals
[i
] = port
? g_strdup_printf("[%s]:%s", host
, port
)
532 : g_strdup_printf("[%s]", host
);
534 vals
[i
] = port
? g_strdup_printf("%s:%s", host
, port
)
540 rados_str
= i
? g_strjoinv(";", (char **)vals
) : NULL
;
542 g_strfreev((char **)vals
);
546 static int qemu_rbd_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
549 BDRVRBDState
*s
= bs
->opaque
;
550 const char *pool
, *snap
, *conf
, *user
, *image_name
, *keypairs
;
551 const char *secretid
, *filename
;
553 Error
*local_err
= NULL
;
554 char *mon_host
= NULL
;
557 /* If we are given a filename, parse the filename, with precedence given to
558 * filename encoded options */
559 filename
= qdict_get_try_str(options
, "filename");
561 warn_report("'filename' option specified. "
562 "This is an unsupported option, and may be deprecated "
564 qemu_rbd_parse_filename(filename
, options
, &local_err
);
567 error_propagate(errp
, local_err
);
572 opts
= qemu_opts_create(&runtime_opts
, NULL
, 0, &error_abort
);
573 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
575 error_propagate(errp
, local_err
);
580 mon_host
= qemu_rbd_mon_host(options
, &local_err
);
582 error_propagate(errp
, local_err
);
587 secretid
= qemu_opt_get(opts
, "password-secret");
589 pool
= qemu_opt_get(opts
, "pool");
590 conf
= qemu_opt_get(opts
, "conf");
591 snap
= qemu_opt_get(opts
, "snapshot");
592 user
= qemu_opt_get(opts
, "user");
593 image_name
= qemu_opt_get(opts
, "image");
594 keypairs
= qemu_opt_get(opts
, "=keyvalue-pairs");
596 if (!pool
|| !image_name
) {
597 error_setg(errp
, "Parameters 'pool' and 'image' are required");
602 r
= rados_create(&s
->cluster
, user
);
604 error_setg_errno(errp
, -r
, "error initializing");
608 s
->snap
= g_strdup(snap
);
609 s
->image_name
= g_strdup(image_name
);
611 /* try default location when conf=NULL, but ignore failure */
612 r
= rados_conf_read_file(s
->cluster
, conf
);
614 error_setg_errno(errp
, -r
, "error reading conf file %s", conf
);
615 goto failed_shutdown
;
618 r
= qemu_rbd_set_keypairs(s
->cluster
, keypairs
, errp
);
620 goto failed_shutdown
;
624 r
= rados_conf_set(s
->cluster
, "mon_host", mon_host
);
626 goto failed_shutdown
;
630 if (qemu_rbd_set_auth(s
->cluster
, secretid
, errp
) < 0) {
632 goto failed_shutdown
;
636 * Fallback to more conservative semantics if setting cache
637 * options fails. Ignore errors from setting rbd_cache because the
638 * only possible error is that the option does not exist, and
639 * librbd defaults to no caching. If write through caching cannot
640 * be set up, fall back to no caching.
642 if (flags
& BDRV_O_NOCACHE
) {
643 rados_conf_set(s
->cluster
, "rbd_cache", "false");
645 rados_conf_set(s
->cluster
, "rbd_cache", "true");
648 r
= rados_connect(s
->cluster
);
650 error_setg_errno(errp
, -r
, "error connecting");
651 goto failed_shutdown
;
654 r
= rados_ioctx_create(s
->cluster
, pool
, &s
->io_ctx
);
656 error_setg_errno(errp
, -r
, "error opening pool %s", pool
);
657 goto failed_shutdown
;
660 /* rbd_open is always r/w */
661 r
= rbd_open(s
->io_ctx
, s
->image_name
, &s
->image
, s
->snap
);
663 error_setg_errno(errp
, -r
, "error reading header from %s",
668 /* If we are using an rbd snapshot, we must be r/o, otherwise
670 if (s
->snap
!= NULL
) {
671 if (!bdrv_is_read_only(bs
)) {
672 error_report("Opening rbd snapshots without an explicit "
673 "read-only=on option is deprecated. Future versions "
674 "will refuse to open the image instead of "
675 "automatically marking the image read-only.");
676 r
= bdrv_set_read_only(bs
, true, &local_err
);
678 error_propagate(errp
, local_err
);
688 rados_ioctx_destroy(s
->io_ctx
);
690 rados_shutdown(s
->cluster
);
692 g_free(s
->image_name
);
701 /* Since RBD is currently always opened R/W via the API,
702 * we just need to check if we are using a snapshot or not, in
703 * order to determine if we will allow it to be R/W */
704 static int qemu_rbd_reopen_prepare(BDRVReopenState
*state
,
705 BlockReopenQueue
*queue
, Error
**errp
)
707 BDRVRBDState
*s
= state
->bs
->opaque
;
710 if (s
->snap
&& state
->flags
& BDRV_O_RDWR
) {
712 "Cannot change node '%s' to r/w when using RBD snapshot",
713 bdrv_get_device_or_node_name(state
->bs
));
720 static void qemu_rbd_close(BlockDriverState
*bs
)
722 BDRVRBDState
*s
= bs
->opaque
;
725 rados_ioctx_destroy(s
->io_ctx
);
727 g_free(s
->image_name
);
728 rados_shutdown(s
->cluster
);
731 static const AIOCBInfo rbd_aiocb_info
= {
732 .aiocb_size
= sizeof(RBDAIOCB
),
735 static void rbd_finish_bh(void *opaque
)
737 RADOSCB
*rcb
= opaque
;
738 qemu_rbd_complete_aio(rcb
);
742 * This is the callback function for rbd_aio_read and _write
744 * Note: this function is being called from a non qemu thread so
745 * we need to be careful about what we do here. Generally we only
746 * schedule a BH, and do the rest of the io completion handling
747 * from rbd_finish_bh() which runs in a qemu context.
749 static void rbd_finish_aiocb(rbd_completion_t c
, RADOSCB
*rcb
)
751 RBDAIOCB
*acb
= rcb
->acb
;
753 rcb
->ret
= rbd_aio_get_return_value(c
);
756 aio_bh_schedule_oneshot(bdrv_get_aio_context(acb
->common
.bs
),
760 static int rbd_aio_discard_wrapper(rbd_image_t image
,
763 rbd_completion_t comp
)
765 #ifdef LIBRBD_SUPPORTS_DISCARD
766 return rbd_aio_discard(image
, off
, len
, comp
);
772 static int rbd_aio_flush_wrapper(rbd_image_t image
,
773 rbd_completion_t comp
)
775 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
776 return rbd_aio_flush(image
, comp
);
782 static BlockAIOCB
*rbd_start_aio(BlockDriverState
*bs
,
786 BlockCompletionFunc
*cb
,
795 BDRVRBDState
*s
= bs
->opaque
;
797 acb
= qemu_aio_get(&rbd_aiocb_info
, bs
, cb
, opaque
);
800 assert(!qiov
|| qiov
->size
== size
);
802 rcb
= g_new(RADOSCB
, 1);
804 if (!LIBRBD_USE_IOVEC
) {
805 if (cmd
== RBD_AIO_DISCARD
|| cmd
== RBD_AIO_FLUSH
) {
808 acb
->bounce
= qemu_try_blockalign(bs
, qiov
->size
);
809 if (acb
->bounce
== NULL
) {
813 if (cmd
== RBD_AIO_WRITE
) {
814 qemu_iovec_to_buf(acb
->qiov
, 0, acb
->bounce
, qiov
->size
);
816 rcb
->buf
= acb
->bounce
;
826 r
= rbd_aio_create_completion(rcb
, (rbd_callback_t
) rbd_finish_aiocb
, &c
);
833 #ifdef LIBRBD_SUPPORTS_IOVEC
834 r
= rbd_aio_writev(s
->image
, qiov
->iov
, qiov
->niov
, off
, c
);
836 r
= rbd_aio_write(s
->image
, off
, size
, rcb
->buf
, c
);
840 #ifdef LIBRBD_SUPPORTS_IOVEC
841 r
= rbd_aio_readv(s
->image
, qiov
->iov
, qiov
->niov
, off
, c
);
843 r
= rbd_aio_read(s
->image
, off
, size
, rcb
->buf
, c
);
846 case RBD_AIO_DISCARD
:
847 r
= rbd_aio_discard_wrapper(s
->image
, off
, size
, c
);
850 r
= rbd_aio_flush_wrapper(s
->image
, c
);
857 goto failed_completion
;
865 if (!LIBRBD_USE_IOVEC
) {
866 qemu_vfree(acb
->bounce
);
873 static BlockAIOCB
*qemu_rbd_aio_readv(BlockDriverState
*bs
,
877 BlockCompletionFunc
*cb
,
880 return rbd_start_aio(bs
, sector_num
<< BDRV_SECTOR_BITS
, qiov
,
881 (int64_t) nb_sectors
<< BDRV_SECTOR_BITS
, cb
, opaque
,
885 static BlockAIOCB
*qemu_rbd_aio_writev(BlockDriverState
*bs
,
889 BlockCompletionFunc
*cb
,
892 return rbd_start_aio(bs
, sector_num
<< BDRV_SECTOR_BITS
, qiov
,
893 (int64_t) nb_sectors
<< BDRV_SECTOR_BITS
, cb
, opaque
,
897 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
898 static BlockAIOCB
*qemu_rbd_aio_flush(BlockDriverState
*bs
,
899 BlockCompletionFunc
*cb
,
902 return rbd_start_aio(bs
, 0, NULL
, 0, cb
, opaque
, RBD_AIO_FLUSH
);
907 static int qemu_rbd_co_flush(BlockDriverState
*bs
)
909 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 1)
910 /* rbd_flush added in 0.1.1 */
911 BDRVRBDState
*s
= bs
->opaque
;
912 return rbd_flush(s
->image
);
919 static int qemu_rbd_getinfo(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
921 BDRVRBDState
*s
= bs
->opaque
;
922 rbd_image_info_t info
;
925 r
= rbd_stat(s
->image
, &info
, sizeof(info
));
930 bdi
->cluster_size
= info
.obj_size
;
934 static int64_t qemu_rbd_getlength(BlockDriverState
*bs
)
936 BDRVRBDState
*s
= bs
->opaque
;
937 rbd_image_info_t info
;
940 r
= rbd_stat(s
->image
, &info
, sizeof(info
));
948 static int qemu_rbd_truncate(BlockDriverState
*bs
, int64_t offset
,
949 PreallocMode prealloc
, Error
**errp
)
951 BDRVRBDState
*s
= bs
->opaque
;
954 if (prealloc
!= PREALLOC_MODE_OFF
) {
955 error_setg(errp
, "Unsupported preallocation mode '%s'",
956 PreallocMode_str(prealloc
));
960 r
= rbd_resize(s
->image
, offset
);
962 error_setg_errno(errp
, -r
, "Failed to resize file");
969 static int qemu_rbd_snap_create(BlockDriverState
*bs
,
970 QEMUSnapshotInfo
*sn_info
)
972 BDRVRBDState
*s
= bs
->opaque
;
975 if (sn_info
->name
[0] == '\0') {
976 return -EINVAL
; /* we need a name for rbd snapshots */
980 * rbd snapshots are using the name as the user controlled unique identifier
981 * we can't use the rbd snapid for that purpose, as it can't be set
983 if (sn_info
->id_str
[0] != '\0' &&
984 strcmp(sn_info
->id_str
, sn_info
->name
) != 0) {
988 if (strlen(sn_info
->name
) >= sizeof(sn_info
->id_str
)) {
992 r
= rbd_snap_create(s
->image
, sn_info
->name
);
994 error_report("failed to create snap: %s", strerror(-r
));
1001 static int qemu_rbd_snap_remove(BlockDriverState
*bs
,
1002 const char *snapshot_id
,
1003 const char *snapshot_name
,
1006 BDRVRBDState
*s
= bs
->opaque
;
1009 if (!snapshot_name
) {
1010 error_setg(errp
, "rbd need a valid snapshot name");
1014 /* If snapshot_id is specified, it must be equal to name, see
1015 qemu_rbd_snap_list() */
1016 if (snapshot_id
&& strcmp(snapshot_id
, snapshot_name
)) {
1018 "rbd do not support snapshot id, it should be NULL or "
1019 "equal to snapshot name");
1023 r
= rbd_snap_remove(s
->image
, snapshot_name
);
1025 error_setg_errno(errp
, -r
, "Failed to remove the snapshot");
1030 static int qemu_rbd_snap_rollback(BlockDriverState
*bs
,
1031 const char *snapshot_name
)
1033 BDRVRBDState
*s
= bs
->opaque
;
1035 return rbd_snap_rollback(s
->image
, snapshot_name
);
1038 static int qemu_rbd_snap_list(BlockDriverState
*bs
,
1039 QEMUSnapshotInfo
**psn_tab
)
1041 BDRVRBDState
*s
= bs
->opaque
;
1042 QEMUSnapshotInfo
*sn_info
, *sn_tab
= NULL
;
1044 rbd_snap_info_t
*snaps
;
1045 int max_snaps
= RBD_MAX_SNAPS
;
1048 snaps
= g_new(rbd_snap_info_t
, max_snaps
);
1049 snap_count
= rbd_snap_list(s
->image
, snaps
, &max_snaps
);
1050 if (snap_count
<= 0) {
1053 } while (snap_count
== -ERANGE
);
1055 if (snap_count
<= 0) {
1059 sn_tab
= g_new0(QEMUSnapshotInfo
, snap_count
);
1061 for (i
= 0; i
< snap_count
; i
++) {
1062 const char *snap_name
= snaps
[i
].name
;
1064 sn_info
= sn_tab
+ i
;
1065 pstrcpy(sn_info
->id_str
, sizeof(sn_info
->id_str
), snap_name
);
1066 pstrcpy(sn_info
->name
, sizeof(sn_info
->name
), snap_name
);
1068 sn_info
->vm_state_size
= snaps
[i
].size
;
1069 sn_info
->date_sec
= 0;
1070 sn_info
->date_nsec
= 0;
1071 sn_info
->vm_clock_nsec
= 0;
1073 rbd_snap_list_end(snaps
);
1081 #ifdef LIBRBD_SUPPORTS_DISCARD
1082 static BlockAIOCB
*qemu_rbd_aio_pdiscard(BlockDriverState
*bs
,
1085 BlockCompletionFunc
*cb
,
1088 return rbd_start_aio(bs
, offset
, NULL
, bytes
, cb
, opaque
,
1093 #ifdef LIBRBD_SUPPORTS_INVALIDATE
1094 static void qemu_rbd_invalidate_cache(BlockDriverState
*bs
,
1097 BDRVRBDState
*s
= bs
->opaque
;
1098 int r
= rbd_invalidate_cache(s
->image
);
1100 error_setg_errno(errp
, -r
, "Failed to invalidate the cache");
1105 static QemuOptsList qemu_rbd_create_opts
= {
1106 .name
= "rbd-create-opts",
1107 .head
= QTAILQ_HEAD_INITIALIZER(qemu_rbd_create_opts
.head
),
1110 .name
= BLOCK_OPT_SIZE
,
1111 .type
= QEMU_OPT_SIZE
,
1112 .help
= "Virtual disk size"
1115 .name
= BLOCK_OPT_CLUSTER_SIZE
,
1116 .type
= QEMU_OPT_SIZE
,
1117 .help
= "RBD object size"
1120 .name
= "password-secret",
1121 .type
= QEMU_OPT_STRING
,
1122 .help
= "ID of secret providing the password",
1124 { /* end of list */ }
1128 static BlockDriver bdrv_rbd
= {
1129 .format_name
= "rbd",
1130 .instance_size
= sizeof(BDRVRBDState
),
1131 .bdrv_parse_filename
= qemu_rbd_parse_filename
,
1132 .bdrv_file_open
= qemu_rbd_open
,
1133 .bdrv_close
= qemu_rbd_close
,
1134 .bdrv_reopen_prepare
= qemu_rbd_reopen_prepare
,
1135 .bdrv_create
= qemu_rbd_create
,
1136 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1137 .bdrv_get_info
= qemu_rbd_getinfo
,
1138 .create_opts
= &qemu_rbd_create_opts
,
1139 .bdrv_getlength
= qemu_rbd_getlength
,
1140 .bdrv_truncate
= qemu_rbd_truncate
,
1141 .protocol_name
= "rbd",
1143 .bdrv_aio_readv
= qemu_rbd_aio_readv
,
1144 .bdrv_aio_writev
= qemu_rbd_aio_writev
,
1146 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
1147 .bdrv_aio_flush
= qemu_rbd_aio_flush
,
1149 .bdrv_co_flush_to_disk
= qemu_rbd_co_flush
,
1152 #ifdef LIBRBD_SUPPORTS_DISCARD
1153 .bdrv_aio_pdiscard
= qemu_rbd_aio_pdiscard
,
1156 .bdrv_snapshot_create
= qemu_rbd_snap_create
,
1157 .bdrv_snapshot_delete
= qemu_rbd_snap_remove
,
1158 .bdrv_snapshot_list
= qemu_rbd_snap_list
,
1159 .bdrv_snapshot_goto
= qemu_rbd_snap_rollback
,
1160 #ifdef LIBRBD_SUPPORTS_INVALIDATE
1161 .bdrv_invalidate_cache
= qemu_rbd_invalidate_cache
,
1165 static void bdrv_rbd_init(void)
1167 bdrv_register(&bdrv_rbd
);
1170 block_init(bdrv_rbd_init
);