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 "block/block_int.h"
20 #include "crypto/secret.h"
21 #include "qemu/cutils.h"
22 #include "qapi/qmp/qstring.h"
23 #include "qapi/qmp/qjson.h"
26 * When specifying the image filename use:
28 * rbd:poolname/devicename[@snapshotname][:option1=value1[:option2=value2...]]
30 * poolname must be the name of an existing rados pool.
32 * devicename is the name of the rbd image.
34 * Each option given is used to configure rados, and may be any valid
35 * Ceph option, "id", or "conf".
37 * The "id" option indicates what user we should authenticate as to
38 * the Ceph cluster. If it is excluded we will use the Ceph default
41 * The "conf" option specifies a Ceph configuration file to read. If
42 * it is not specified, we will read from the default Ceph locations
43 * (e.g., /etc/ceph/ceph.conf). To avoid reading _any_ configuration
44 * file, specify conf=/dev/null.
46 * Configuration values containing :, @, or = can be escaped with a
50 /* rbd_aio_discard added in 0.1.2 */
51 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 2)
52 #define LIBRBD_SUPPORTS_DISCARD
54 #undef LIBRBD_SUPPORTS_DISCARD
57 #define OBJ_MAX_SIZE (1UL << OBJ_DEFAULT_OBJ_ORDER)
59 #define RBD_MAX_SNAPS 100
61 /* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */
62 #ifdef LIBRBD_SUPPORTS_IOVEC
63 #define LIBRBD_USE_IOVEC 1
65 #define LIBRBD_USE_IOVEC 0
75 typedef struct RBDAIOCB
{
82 struct BDRVRBDState
*s
;
85 typedef struct RADOSCB
{
87 struct BDRVRBDState
*s
;
93 typedef struct BDRVRBDState
{
101 static char *qemu_rbd_next_tok(char *src
, char delim
, char **p
)
107 for (end
= src
; *end
; ++end
) {
111 if (*end
== '\\' && end
[1] != '\0') {
122 static void qemu_rbd_unescape(char *src
)
126 for (p
= src
; *src
; ++src
, ++p
) {
127 if (*src
== '\\' && src
[1] != '\0') {
135 static void qemu_rbd_parse_filename(const char *filename
, QDict
*options
,
140 QList
*keypairs
= NULL
;
143 if (!strstart(filename
, "rbd:", &start
)) {
144 error_setg(errp
, "File name must start with 'rbd:'");
148 buf
= g_strdup(start
);
151 found_str
= qemu_rbd_next_tok(p
, '/', &p
);
153 error_setg(errp
, "Pool name is required");
156 qemu_rbd_unescape(found_str
);
157 qdict_put(options
, "pool", qstring_from_str(found_str
));
159 if (strchr(p
, '@')) {
160 found_str
= qemu_rbd_next_tok(p
, '@', &p
);
161 qemu_rbd_unescape(found_str
);
162 qdict_put(options
, "image", qstring_from_str(found_str
));
164 found_str
= qemu_rbd_next_tok(p
, ':', &p
);
165 qemu_rbd_unescape(found_str
);
166 qdict_put(options
, "snapshot", qstring_from_str(found_str
));
168 found_str
= qemu_rbd_next_tok(p
, ':', &p
);
169 qemu_rbd_unescape(found_str
);
170 qdict_put(options
, "image", qstring_from_str(found_str
));
176 /* The following are essentially all key/value pairs, and we treat
177 * 'id' and 'conf' a bit special. Key/value pairs may be in any order. */
180 name
= qemu_rbd_next_tok(p
, '=', &p
);
182 error_setg(errp
, "conf option %s has no value", name
);
186 qemu_rbd_unescape(name
);
188 value
= qemu_rbd_next_tok(p
, ':', &p
);
189 qemu_rbd_unescape(value
);
191 if (!strcmp(name
, "conf")) {
192 qdict_put(options
, "conf", qstring_from_str(value
));
193 } else if (!strcmp(name
, "id")) {
194 qdict_put(options
, "user" , qstring_from_str(value
));
197 * We pass these internally to qemu_rbd_set_keypairs(), so
198 * we can get away with the simpler list of [ "key1",
199 * "value1", "key2", "value2" ] rather than a raw dict
200 * { "key1": "value1", "key2": "value2" } where we can't
201 * guarantee order, or even a more correct but complex
202 * [ { "key1": "value1" }, { "key2": "value2" } ]
205 keypairs
= qlist_new();
207 qlist_append(keypairs
, qstring_from_str(name
));
208 qlist_append(keypairs
, qstring_from_str(value
));
213 qdict_put(options
, "=keyvalue-pairs",
214 qobject_to_json(QOBJECT(keypairs
)));
224 static int qemu_rbd_set_auth(rados_t cluster
, const char *secretid
,
231 gchar
*secret
= qcrypto_secret_lookup_as_base64(secretid
,
237 rados_conf_set(cluster
, "key", secret
);
243 static int qemu_rbd_set_keypairs(rados_t cluster
, const char *keypairs_json
,
253 if (!keypairs_json
) {
256 keypairs
= qobject_to_qlist(qobject_from_json(keypairs_json
,
258 remaining
= qlist_size(keypairs
) / 2;
261 while (remaining
--) {
262 name
= qobject_to_qstring(qlist_pop(keypairs
));
263 value
= qobject_to_qstring(qlist_pop(keypairs
));
264 assert(name
&& value
);
265 key
= qstring_get_str(name
);
267 ret
= rados_conf_set(cluster
, key
, qstring_get_str(value
));
271 error_setg_errno(errp
, -ret
, "invalid conf option %s", key
);
281 static void qemu_rbd_memset(RADOSCB
*rcb
, int64_t offs
)
283 if (LIBRBD_USE_IOVEC
) {
284 RBDAIOCB
*acb
= rcb
->acb
;
285 iov_memset(acb
->qiov
->iov
, acb
->qiov
->niov
, offs
, 0,
286 acb
->qiov
->size
- offs
);
288 memset(rcb
->buf
+ offs
, 0, rcb
->size
- offs
);
292 static QemuOptsList runtime_opts
= {
294 .head
= QTAILQ_HEAD_INITIALIZER(runtime_opts
.head
),
298 .type
= QEMU_OPT_STRING
,
299 .help
= "Rados pool name",
303 .type
= QEMU_OPT_STRING
,
304 .help
= "Image name in the pool",
308 .type
= QEMU_OPT_STRING
,
309 .help
= "Rados config file location",
313 .type
= QEMU_OPT_STRING
,
314 .help
= "Ceph snapshot name",
317 /* maps to 'id' in rados_create() */
319 .type
= QEMU_OPT_STRING
,
320 .help
= "Rados id name",
323 * server.* extracted manually, see qemu_rbd_mon_host()
326 .name
= "password-secret",
327 .type
= QEMU_OPT_STRING
,
328 .help
= "ID of secret providing the password",
332 * Keys for qemu_rbd_parse_filename(), not in the QAPI schema
336 * HACK: name starts with '=' so that qemu_opts_parse()
339 .name
= "=keyvalue-pairs",
340 .type
= QEMU_OPT_STRING
,
341 .help
= "Legacy rados key/value option parameters",
343 { /* end of list */ }
347 static int qemu_rbd_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
349 Error
*local_err
= NULL
;
353 const char *pool
, *name
, *conf
, *clientname
, *keypairs
;
354 const char *secretid
;
356 rados_ioctx_t io_ctx
;
357 QDict
*options
= NULL
;
360 secretid
= qemu_opt_get(opts
, "password-secret");
362 /* Read out options */
363 bytes
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
365 objsize
= qemu_opt_get_size_del(opts
, BLOCK_OPT_CLUSTER_SIZE
, 0);
367 if ((objsize
- 1) & objsize
) { /* not a power of 2? */
368 error_setg(errp
, "obj size needs to be power of 2");
372 if (objsize
< 4096) {
373 error_setg(errp
, "obj size too small");
377 obj_order
= ctz32(objsize
);
380 options
= qdict_new();
381 qemu_rbd_parse_filename(filename
, options
, &local_err
);
384 error_propagate(errp
, local_err
);
389 * Caution: while qdict_get_try_str() is fine, getting non-string
390 * types would require more care. When @options come from -blockdev
391 * or blockdev_add, its members are typed according to the QAPI
392 * schema, but when they come from -drive, they're all QString.
394 pool
= qdict_get_try_str(options
, "pool");
395 conf
= qdict_get_try_str(options
, "conf");
396 clientname
= qdict_get_try_str(options
, "user");
397 name
= qdict_get_try_str(options
, "image");
398 keypairs
= qdict_get_try_str(options
, "=keyvalue-pairs");
400 ret
= rados_create(&cluster
, clientname
);
402 error_setg_errno(errp
, -ret
, "error initializing");
406 /* try default location when conf=NULL, but ignore failure */
407 ret
= rados_conf_read_file(cluster
, conf
);
408 if (conf
&& ret
< 0) {
409 error_setg_errno(errp
, -ret
, "error reading conf file %s", conf
);
414 ret
= qemu_rbd_set_keypairs(cluster
, keypairs
, errp
);
420 if (qemu_rbd_set_auth(cluster
, secretid
, errp
) < 0) {
425 ret
= rados_connect(cluster
);
427 error_setg_errno(errp
, -ret
, "error connecting");
431 ret
= rados_ioctx_create(cluster
, pool
, &io_ctx
);
433 error_setg_errno(errp
, -ret
, "error opening pool %s", pool
);
437 ret
= rbd_create(io_ctx
, name
, bytes
, &obj_order
);
439 error_setg_errno(errp
, -ret
, "error rbd create");
442 rados_ioctx_destroy(io_ctx
);
445 rados_shutdown(cluster
);
453 * This aio completion is being called from rbd_finish_bh() and runs in qemu
456 static void qemu_rbd_complete_aio(RADOSCB
*rcb
)
458 RBDAIOCB
*acb
= rcb
->acb
;
463 if (acb
->cmd
!= RBD_AIO_READ
) {
467 } else if (!acb
->error
) {
468 acb
->ret
= rcb
->size
;
472 qemu_rbd_memset(rcb
, 0);
475 } else if (r
< rcb
->size
) {
476 qemu_rbd_memset(rcb
, r
);
478 acb
->ret
= rcb
->size
;
480 } else if (!acb
->error
) {
487 if (!LIBRBD_USE_IOVEC
) {
488 if (acb
->cmd
== RBD_AIO_READ
) {
489 qemu_iovec_from_buf(acb
->qiov
, 0, acb
->bounce
, acb
->qiov
->size
);
491 qemu_vfree(acb
->bounce
);
494 acb
->common
.cb(acb
->common
.opaque
, (acb
->ret
> 0 ? 0 : acb
->ret
));
499 static char *qemu_rbd_mon_host(QDict
*options
, Error
**errp
)
501 const char **vals
= g_new(const char *, qdict_size(options
) + 1);
503 const char *host
, *port
;
508 sprintf(keybuf
, "server.%d.host", i
);
509 host
= qdict_get_try_str(options
, keybuf
);
510 qdict_del(options
, keybuf
);
511 sprintf(keybuf
, "server.%d.port", i
);
512 port
= qdict_get_try_str(options
, keybuf
);
513 qdict_del(options
, keybuf
);
514 if (!host
&& !port
) {
518 error_setg(errp
, "Parameter server.%d.host is missing", i
);
523 if (strchr(host
, ':')) {
524 vals
[i
] = port
? g_strdup_printf("[%s]:%s", host
, port
)
525 : g_strdup_printf("[%s]", host
);
527 vals
[i
] = port
? g_strdup_printf("%s:%s", host
, port
)
533 rados_str
= i
? g_strjoinv(";", (char **)vals
) : NULL
;
535 g_strfreev((char **)vals
);
539 static int qemu_rbd_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
542 BDRVRBDState
*s
= bs
->opaque
;
543 const char *pool
, *snap
, *conf
, *clientname
, *name
, *keypairs
;
544 const char *secretid
;
546 Error
*local_err
= NULL
;
547 char *mon_host
= NULL
;
550 opts
= qemu_opts_create(&runtime_opts
, NULL
, 0, &error_abort
);
551 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
553 error_propagate(errp
, local_err
);
558 mon_host
= qemu_rbd_mon_host(options
, &local_err
);
560 error_propagate(errp
, local_err
);
565 secretid
= qemu_opt_get(opts
, "password-secret");
567 pool
= qemu_opt_get(opts
, "pool");
568 conf
= qemu_opt_get(opts
, "conf");
569 snap
= qemu_opt_get(opts
, "snapshot");
570 clientname
= qemu_opt_get(opts
, "user");
571 name
= qemu_opt_get(opts
, "image");
572 keypairs
= qemu_opt_get(opts
, "=keyvalue-pairs");
574 if (!pool
|| !name
) {
575 error_setg(errp
, "Parameters 'pool' and 'image' are required");
580 r
= rados_create(&s
->cluster
, clientname
);
582 error_setg_errno(errp
, -r
, "error initializing");
586 s
->snap
= g_strdup(snap
);
587 s
->name
= g_strdup(name
);
589 /* try default location when conf=NULL, but ignore failure */
590 r
= rados_conf_read_file(s
->cluster
, conf
);
592 error_setg_errno(errp
, -r
, "error reading conf file %s", conf
);
593 goto failed_shutdown
;
596 r
= qemu_rbd_set_keypairs(s
->cluster
, keypairs
, errp
);
598 goto failed_shutdown
;
602 r
= rados_conf_set(s
->cluster
, "mon_host", mon_host
);
604 goto failed_shutdown
;
608 if (qemu_rbd_set_auth(s
->cluster
, secretid
, errp
) < 0) {
610 goto failed_shutdown
;
614 * Fallback to more conservative semantics if setting cache
615 * options fails. Ignore errors from setting rbd_cache because the
616 * only possible error is that the option does not exist, and
617 * librbd defaults to no caching. If write through caching cannot
618 * be set up, fall back to no caching.
620 if (flags
& BDRV_O_NOCACHE
) {
621 rados_conf_set(s
->cluster
, "rbd_cache", "false");
623 rados_conf_set(s
->cluster
, "rbd_cache", "true");
626 r
= rados_connect(s
->cluster
);
628 error_setg_errno(errp
, -r
, "error connecting");
629 goto failed_shutdown
;
632 r
= rados_ioctx_create(s
->cluster
, pool
, &s
->io_ctx
);
634 error_setg_errno(errp
, -r
, "error opening pool %s", pool
);
635 goto failed_shutdown
;
638 r
= rbd_open(s
->io_ctx
, s
->name
, &s
->image
, s
->snap
);
640 error_setg_errno(errp
, -r
, "error reading header from %s", s
->name
);
644 bs
->read_only
= (s
->snap
!= NULL
);
650 rados_ioctx_destroy(s
->io_ctx
);
652 rados_shutdown(s
->cluster
);
661 static void qemu_rbd_close(BlockDriverState
*bs
)
663 BDRVRBDState
*s
= bs
->opaque
;
666 rados_ioctx_destroy(s
->io_ctx
);
669 rados_shutdown(s
->cluster
);
672 static const AIOCBInfo rbd_aiocb_info
= {
673 .aiocb_size
= sizeof(RBDAIOCB
),
676 static void rbd_finish_bh(void *opaque
)
678 RADOSCB
*rcb
= opaque
;
679 qemu_rbd_complete_aio(rcb
);
683 * This is the callback function for rbd_aio_read and _write
685 * Note: this function is being called from a non qemu thread so
686 * we need to be careful about what we do here. Generally we only
687 * schedule a BH, and do the rest of the io completion handling
688 * from rbd_finish_bh() which runs in a qemu context.
690 static void rbd_finish_aiocb(rbd_completion_t c
, RADOSCB
*rcb
)
692 RBDAIOCB
*acb
= rcb
->acb
;
694 rcb
->ret
= rbd_aio_get_return_value(c
);
697 aio_bh_schedule_oneshot(bdrv_get_aio_context(acb
->common
.bs
),
701 static int rbd_aio_discard_wrapper(rbd_image_t image
,
704 rbd_completion_t comp
)
706 #ifdef LIBRBD_SUPPORTS_DISCARD
707 return rbd_aio_discard(image
, off
, len
, comp
);
713 static int rbd_aio_flush_wrapper(rbd_image_t image
,
714 rbd_completion_t comp
)
716 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
717 return rbd_aio_flush(image
, comp
);
723 static BlockAIOCB
*rbd_start_aio(BlockDriverState
*bs
,
727 BlockCompletionFunc
*cb
,
736 BDRVRBDState
*s
= bs
->opaque
;
738 acb
= qemu_aio_get(&rbd_aiocb_info
, bs
, cb
, opaque
);
741 assert(!qiov
|| qiov
->size
== size
);
743 rcb
= g_new(RADOSCB
, 1);
745 if (!LIBRBD_USE_IOVEC
) {
746 if (cmd
== RBD_AIO_DISCARD
|| cmd
== RBD_AIO_FLUSH
) {
749 acb
->bounce
= qemu_try_blockalign(bs
, qiov
->size
);
750 if (acb
->bounce
== NULL
) {
754 if (cmd
== RBD_AIO_WRITE
) {
755 qemu_iovec_to_buf(acb
->qiov
, 0, acb
->bounce
, qiov
->size
);
757 rcb
->buf
= acb
->bounce
;
767 r
= rbd_aio_create_completion(rcb
, (rbd_callback_t
) rbd_finish_aiocb
, &c
);
774 #ifdef LIBRBD_SUPPORTS_IOVEC
775 r
= rbd_aio_writev(s
->image
, qiov
->iov
, qiov
->niov
, off
, c
);
777 r
= rbd_aio_write(s
->image
, off
, size
, rcb
->buf
, c
);
781 #ifdef LIBRBD_SUPPORTS_IOVEC
782 r
= rbd_aio_readv(s
->image
, qiov
->iov
, qiov
->niov
, off
, c
);
784 r
= rbd_aio_read(s
->image
, off
, size
, rcb
->buf
, c
);
787 case RBD_AIO_DISCARD
:
788 r
= rbd_aio_discard_wrapper(s
->image
, off
, size
, c
);
791 r
= rbd_aio_flush_wrapper(s
->image
, c
);
798 goto failed_completion
;
806 if (!LIBRBD_USE_IOVEC
) {
807 qemu_vfree(acb
->bounce
);
814 static BlockAIOCB
*qemu_rbd_aio_readv(BlockDriverState
*bs
,
818 BlockCompletionFunc
*cb
,
821 return rbd_start_aio(bs
, sector_num
<< BDRV_SECTOR_BITS
, qiov
,
822 (int64_t) nb_sectors
<< BDRV_SECTOR_BITS
, cb
, opaque
,
826 static BlockAIOCB
*qemu_rbd_aio_writev(BlockDriverState
*bs
,
830 BlockCompletionFunc
*cb
,
833 return rbd_start_aio(bs
, sector_num
<< BDRV_SECTOR_BITS
, qiov
,
834 (int64_t) nb_sectors
<< BDRV_SECTOR_BITS
, cb
, opaque
,
838 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
839 static BlockAIOCB
*qemu_rbd_aio_flush(BlockDriverState
*bs
,
840 BlockCompletionFunc
*cb
,
843 return rbd_start_aio(bs
, 0, NULL
, 0, cb
, opaque
, RBD_AIO_FLUSH
);
848 static int qemu_rbd_co_flush(BlockDriverState
*bs
)
850 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 1)
851 /* rbd_flush added in 0.1.1 */
852 BDRVRBDState
*s
= bs
->opaque
;
853 return rbd_flush(s
->image
);
860 static int qemu_rbd_getinfo(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
862 BDRVRBDState
*s
= bs
->opaque
;
863 rbd_image_info_t info
;
866 r
= rbd_stat(s
->image
, &info
, sizeof(info
));
871 bdi
->cluster_size
= info
.obj_size
;
875 static int64_t qemu_rbd_getlength(BlockDriverState
*bs
)
877 BDRVRBDState
*s
= bs
->opaque
;
878 rbd_image_info_t info
;
881 r
= rbd_stat(s
->image
, &info
, sizeof(info
));
889 static int qemu_rbd_truncate(BlockDriverState
*bs
, int64_t offset
)
891 BDRVRBDState
*s
= bs
->opaque
;
894 r
= rbd_resize(s
->image
, offset
);
902 static int qemu_rbd_snap_create(BlockDriverState
*bs
,
903 QEMUSnapshotInfo
*sn_info
)
905 BDRVRBDState
*s
= bs
->opaque
;
908 if (sn_info
->name
[0] == '\0') {
909 return -EINVAL
; /* we need a name for rbd snapshots */
913 * rbd snapshots are using the name as the user controlled unique identifier
914 * we can't use the rbd snapid for that purpose, as it can't be set
916 if (sn_info
->id_str
[0] != '\0' &&
917 strcmp(sn_info
->id_str
, sn_info
->name
) != 0) {
921 if (strlen(sn_info
->name
) >= sizeof(sn_info
->id_str
)) {
925 r
= rbd_snap_create(s
->image
, sn_info
->name
);
927 error_report("failed to create snap: %s", strerror(-r
));
934 static int qemu_rbd_snap_remove(BlockDriverState
*bs
,
935 const char *snapshot_id
,
936 const char *snapshot_name
,
939 BDRVRBDState
*s
= bs
->opaque
;
942 if (!snapshot_name
) {
943 error_setg(errp
, "rbd need a valid snapshot name");
947 /* If snapshot_id is specified, it must be equal to name, see
948 qemu_rbd_snap_list() */
949 if (snapshot_id
&& strcmp(snapshot_id
, snapshot_name
)) {
951 "rbd do not support snapshot id, it should be NULL or "
952 "equal to snapshot name");
956 r
= rbd_snap_remove(s
->image
, snapshot_name
);
958 error_setg_errno(errp
, -r
, "Failed to remove the snapshot");
963 static int qemu_rbd_snap_rollback(BlockDriverState
*bs
,
964 const char *snapshot_name
)
966 BDRVRBDState
*s
= bs
->opaque
;
968 return rbd_snap_rollback(s
->image
, snapshot_name
);
971 static int qemu_rbd_snap_list(BlockDriverState
*bs
,
972 QEMUSnapshotInfo
**psn_tab
)
974 BDRVRBDState
*s
= bs
->opaque
;
975 QEMUSnapshotInfo
*sn_info
, *sn_tab
= NULL
;
977 rbd_snap_info_t
*snaps
;
978 int max_snaps
= RBD_MAX_SNAPS
;
981 snaps
= g_new(rbd_snap_info_t
, max_snaps
);
982 snap_count
= rbd_snap_list(s
->image
, snaps
, &max_snaps
);
983 if (snap_count
<= 0) {
986 } while (snap_count
== -ERANGE
);
988 if (snap_count
<= 0) {
992 sn_tab
= g_new0(QEMUSnapshotInfo
, snap_count
);
994 for (i
= 0; i
< snap_count
; i
++) {
995 const char *snap_name
= snaps
[i
].name
;
997 sn_info
= sn_tab
+ i
;
998 pstrcpy(sn_info
->id_str
, sizeof(sn_info
->id_str
), snap_name
);
999 pstrcpy(sn_info
->name
, sizeof(sn_info
->name
), snap_name
);
1001 sn_info
->vm_state_size
= snaps
[i
].size
;
1002 sn_info
->date_sec
= 0;
1003 sn_info
->date_nsec
= 0;
1004 sn_info
->vm_clock_nsec
= 0;
1006 rbd_snap_list_end(snaps
);
1014 #ifdef LIBRBD_SUPPORTS_DISCARD
1015 static BlockAIOCB
*qemu_rbd_aio_pdiscard(BlockDriverState
*bs
,
1018 BlockCompletionFunc
*cb
,
1021 return rbd_start_aio(bs
, offset
, NULL
, count
, cb
, opaque
,
1026 #ifdef LIBRBD_SUPPORTS_INVALIDATE
1027 static void qemu_rbd_invalidate_cache(BlockDriverState
*bs
,
1030 BDRVRBDState
*s
= bs
->opaque
;
1031 int r
= rbd_invalidate_cache(s
->image
);
1033 error_setg_errno(errp
, -r
, "Failed to invalidate the cache");
1038 static QemuOptsList qemu_rbd_create_opts
= {
1039 .name
= "rbd-create-opts",
1040 .head
= QTAILQ_HEAD_INITIALIZER(qemu_rbd_create_opts
.head
),
1043 .name
= BLOCK_OPT_SIZE
,
1044 .type
= QEMU_OPT_SIZE
,
1045 .help
= "Virtual disk size"
1048 .name
= BLOCK_OPT_CLUSTER_SIZE
,
1049 .type
= QEMU_OPT_SIZE
,
1050 .help
= "RBD object size"
1053 .name
= "password-secret",
1054 .type
= QEMU_OPT_STRING
,
1055 .help
= "ID of secret providing the password",
1057 { /* end of list */ }
1061 static BlockDriver bdrv_rbd
= {
1062 .format_name
= "rbd",
1063 .instance_size
= sizeof(BDRVRBDState
),
1064 .bdrv_parse_filename
= qemu_rbd_parse_filename
,
1065 .bdrv_file_open
= qemu_rbd_open
,
1066 .bdrv_close
= qemu_rbd_close
,
1067 .bdrv_create
= qemu_rbd_create
,
1068 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1069 .bdrv_get_info
= qemu_rbd_getinfo
,
1070 .create_opts
= &qemu_rbd_create_opts
,
1071 .bdrv_getlength
= qemu_rbd_getlength
,
1072 .bdrv_truncate
= qemu_rbd_truncate
,
1073 .protocol_name
= "rbd",
1075 .bdrv_aio_readv
= qemu_rbd_aio_readv
,
1076 .bdrv_aio_writev
= qemu_rbd_aio_writev
,
1078 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
1079 .bdrv_aio_flush
= qemu_rbd_aio_flush
,
1081 .bdrv_co_flush_to_disk
= qemu_rbd_co_flush
,
1084 #ifdef LIBRBD_SUPPORTS_DISCARD
1085 .bdrv_aio_pdiscard
= qemu_rbd_aio_pdiscard
,
1088 .bdrv_snapshot_create
= qemu_rbd_snap_create
,
1089 .bdrv_snapshot_delete
= qemu_rbd_snap_remove
,
1090 .bdrv_snapshot_list
= qemu_rbd_snap_list
,
1091 .bdrv_snapshot_goto
= qemu_rbd_snap_rollback
,
1092 #ifdef LIBRBD_SUPPORTS_INVALIDATE
1093 .bdrv_invalidate_cache
= qemu_rbd_invalidate_cache
,
1097 static void bdrv_rbd_init(void)
1099 bdrv_register(&bdrv_rbd
);
1102 block_init(bdrv_rbd_init
);