2 * QEMU Block driver for RADOS (Ceph)
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_str(options, "pool", found_str);
159 if (strchr(p, '@')) {
160 found_str = qemu_rbd_next_tok(p, '@', &p);
161 qemu_rbd_unescape(found_str);
162 qdict_put_str(options, "image", found_str);
164 found_str = qemu_rbd_next_tok(p, ':', &p);
165 qemu_rbd_unescape(found_str);
166 qdict_put_str(options, "snapshot", found_str);
168 found_str = qemu_rbd_next_tok(p, ':', &p);
169 qemu_rbd_unescape(found_str);
170 qdict_put_str(options, "image", 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_str(options, "conf", value);
193 } else if (!strcmp(name, "id")) {
194 qdict_put_str(options, "user", 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_str(keypairs, name);
208 qlist_append_str(keypairs, 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",
345 .type = QEMU_OPT_STRING,
347 { /* end of list */ }
351 static int qemu_rbd_create(const char *filename, QemuOpts *opts, Error **errp)
353 Error *local_err = NULL;
357 const char *pool, *image_name, *conf, *user, *keypairs;
358 const char *secretid;
360 rados_ioctx_t io_ctx;
361 QDict *options = NULL;
364 secretid = qemu_opt_get(opts, "password-secret");
366 /* Read out options */
367 bytes = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
369 objsize = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, 0);
371 if ((objsize - 1) & objsize) { /* not a power of 2? */
372 error_setg(errp, "obj size needs to be power of 2");
376 if (objsize < 4096) {
377 error_setg(errp, "obj size too small");
381 obj_order = ctz32(objsize);
384 options = qdict_new();
385 qemu_rbd_parse_filename(filename, options, &local_err);
388 error_propagate(errp, local_err);
393 * Caution: while qdict_get_try_str() is fine, getting non-string
394 * types would require more care. When @options come from -blockdev
395 * or blockdev_add, its members are typed according to the QAPI
396 * schema, but when they come from -drive, they're all QString.
398 pool = qdict_get_try_str(options, "pool");
399 conf = qdict_get_try_str(options, "conf");
400 user = qdict_get_try_str(options, "user");
401 image_name = qdict_get_try_str(options, "image");
402 keypairs = qdict_get_try_str(options, "=keyvalue-pairs");
404 ret = rados_create(&cluster, user);
406 error_setg_errno(errp, -ret, "error initializing");
410 /* try default location when conf=NULL, but ignore failure */
411 ret = rados_conf_read_file(cluster, conf);
412 if (conf && ret < 0) {
413 error_setg_errno(errp, -ret, "error reading conf file %s", conf);
418 ret = qemu_rbd_set_keypairs(cluster, keypairs, errp);
424 if (qemu_rbd_set_auth(cluster, secretid, errp) < 0) {
429 ret = rados_connect(cluster);
431 error_setg_errno(errp, -ret, "error connecting");
435 ret = rados_ioctx_create(cluster, pool, &io_ctx);
437 error_setg_errno(errp, -ret, "error opening pool %s", pool);
441 ret = rbd_create(io_ctx, image_name, bytes, &obj_order);
443 error_setg_errno(errp, -ret, "error rbd create");
446 rados_ioctx_destroy(io_ctx);
449 rados_shutdown(cluster);
457 * This aio completion is being called from rbd_finish_bh() and runs in qemu
460 static void qemu_rbd_complete_aio(RADOSCB *rcb)
462 RBDAIOCB *acb = rcb->acb;
467 if (acb->cmd != RBD_AIO_READ) {
471 } else if (!acb->error) {
472 acb->ret = rcb->size;
476 qemu_rbd_memset(rcb, 0);
479 } else if (r < rcb->size) {
480 qemu_rbd_memset(rcb, r);
482 acb->ret = rcb->size;
484 } else if (!acb->error) {
491 if (!LIBRBD_USE_IOVEC) {
492 if (acb->cmd == RBD_AIO_READ) {
493 qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
495 qemu_vfree(acb->bounce);
498 acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
503 static char *qemu_rbd_mon_host(QDict *options, Error **errp)
505 const char **vals = g_new(const char *, qdict_size(options) + 1);
507 const char *host, *port;
512 sprintf(keybuf, "server.%d.host", i);
513 host = qdict_get_try_str(options, keybuf);
514 qdict_del(options, keybuf);
515 sprintf(keybuf, "server.%d.port", i);
516 port = qdict_get_try_str(options, keybuf);
517 qdict_del(options, keybuf);
518 if (!host && !port) {
522 error_setg(errp, "Parameter server.%d.host is missing", i);
527 if (strchr(host, ':')) {
528 vals[i] = port ? g_strdup_printf("[%s]:%s", host, port)
529 : g_strdup_printf("[%s]", host);
531 vals[i] = port ? g_strdup_printf("%s:%s", host, port)
537 rados_str = i ? g_strjoinv(";", (char **)vals) : NULL;
539 g_strfreev((char **)vals);
543 static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
546 BDRVRBDState *s = bs->opaque;
547 const char *pool, *snap, *conf, *user, *image_name, *keypairs;
548 const char *secretid, *filename;
550 Error *local_err = NULL;
551 char *mon_host = NULL;
554 /* If we are given a filename, parse the filename, with precedence given to
555 * filename encoded options */
556 filename = qdict_get_try_str(options, "filename");
558 warn_report("'filename' option specified. "
559 "This is an unsupported option, and may be deprecated "
561 qemu_rbd_parse_filename(filename, options, &local_err);
564 error_propagate(errp, local_err);
569 opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
570 qemu_opts_absorb_qdict(opts, options, &local_err);
572 error_propagate(errp, local_err);
577 mon_host = qemu_rbd_mon_host(options, &local_err);
579 error_propagate(errp, local_err);
584 secretid = qemu_opt_get(opts, "password-secret");
586 pool = qemu_opt_get(opts, "pool");
587 conf = qemu_opt_get(opts, "conf");
588 snap = qemu_opt_get(opts, "snapshot");
589 user = qemu_opt_get(opts, "user");
590 image_name = qemu_opt_get(opts, "image");
591 keypairs = qemu_opt_get(opts, "=keyvalue-pairs");
593 if (!pool || !image_name) {
594 error_setg(errp, "Parameters 'pool' and 'image' are required");
599 r = rados_create(&s->cluster, user);
601 error_setg_errno(errp, -r, "error initializing");
605 s->snap = g_strdup(snap);
606 s->image_name = g_strdup(image_name);
608 /* try default location when conf=NULL, but ignore failure */
609 r = rados_conf_read_file(s->cluster, conf);
611 error_setg_errno(errp, -r, "error reading conf file %s", conf);
612 goto failed_shutdown;
615 r = qemu_rbd_set_keypairs(s->cluster, keypairs, errp);
617 goto failed_shutdown;
621 r = rados_conf_set(s->cluster, "mon_host", mon_host);
623 goto failed_shutdown;
627 if (qemu_rbd_set_auth(s->cluster, secretid, errp) < 0) {
629 goto failed_shutdown;
633 * Fallback to more conservative semantics if setting cache
634 * options fails. Ignore errors from setting rbd_cache because the
635 * only possible error is that the option does not exist, and
636 * librbd defaults to no caching. If write through caching cannot
637 * be set up, fall back to no caching.
639 if (flags & BDRV_O_NOCACHE) {
640 rados_conf_set(s->cluster, "rbd_cache", "false");
642 rados_conf_set(s->cluster, "rbd_cache", "true");
645 r = rados_connect(s->cluster);
647 error_setg_errno(errp, -r, "error connecting");
648 goto failed_shutdown;
651 r = rados_ioctx_create(s->cluster, pool, &s->io_ctx);
653 error_setg_errno(errp, -r, "error opening pool %s", pool);
654 goto failed_shutdown;
657 /* rbd_open is always r/w */
658 r = rbd_open(s->io_ctx, s->image_name, &s->image, s->snap);
660 error_setg_errno(errp, -r, "error reading header from %s",
665 /* If we are using an rbd snapshot, we must be r/o, otherwise
667 if (s->snap != NULL) {
668 if (!bdrv_is_read_only(bs)) {
669 error_report("Opening rbd snapshots without an explicit "
670 "read-only=on option is deprecated. Future versions "
671 "will refuse to open the image instead of "
672 "automatically marking the image read-only.");
673 r = bdrv_set_read_only(bs, true, &local_err);
675 error_propagate(errp, local_err);
685 rados_ioctx_destroy(s->io_ctx);
687 rados_shutdown(s->cluster);
689 g_free(s->image_name);
698 /* Since RBD is currently always opened R/W via the API,
699 * we just need to check if we are using a snapshot or not, in
700 * order to determine if we will allow it to be R/W */
701 static int qemu_rbd_reopen_prepare(BDRVReopenState *state,
702 BlockReopenQueue *queue, Error **errp)
704 BDRVRBDState *s = state->bs->opaque;
707 if (s->snap && state->flags & BDRV_O_RDWR) {
709 "Cannot change node '%s' to r/w when using RBD snapshot",
710 bdrv_get_device_or_node_name(state->bs));
717 static void qemu_rbd_close(BlockDriverState *bs)
719 BDRVRBDState *s = bs->opaque;
722 rados_ioctx_destroy(s->io_ctx);
724 g_free(s->image_name);
725 rados_shutdown(s->cluster);
728 static const AIOCBInfo rbd_aiocb_info = {
729 .aiocb_size = sizeof(RBDAIOCB),
732 static void rbd_finish_bh(void *opaque)
734 RADOSCB *rcb = opaque;
735 qemu_rbd_complete_aio(rcb);
739 * This is the callback function for rbd_aio_read and _write
741 * Note: this function is being called from a non qemu thread so
742 * we need to be careful about what we do here. Generally we only
743 * schedule a BH, and do the rest of the io completion handling
744 * from rbd_finish_bh() which runs in a qemu context.
746 static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
748 RBDAIOCB *acb = rcb->acb;
750 rcb->ret = rbd_aio_get_return_value(c);
753 aio_bh_schedule_oneshot(bdrv_get_aio_context(acb->common.bs),
757 static int rbd_aio_discard_wrapper(rbd_image_t image,
760 rbd_completion_t comp)
762 #ifdef LIBRBD_SUPPORTS_DISCARD
763 return rbd_aio_discard(image, off, len, comp);
769 static int rbd_aio_flush_wrapper(rbd_image_t image,
770 rbd_completion_t comp)
772 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
773 return rbd_aio_flush(image, comp);
779 static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
783 BlockCompletionFunc *cb,
792 BDRVRBDState *s = bs->opaque;
794 acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque);
797 assert(!qiov || qiov->size == size);
799 rcb = g_new(RADOSCB, 1);
801 if (!LIBRBD_USE_IOVEC) {
802 if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
805 acb->bounce = qemu_try_blockalign(bs, qiov->size);
806 if (acb->bounce == NULL) {
810 if (cmd == RBD_AIO_WRITE) {
811 qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
813 rcb->buf = acb->bounce;
823 r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
830 #ifdef LIBRBD_SUPPORTS_IOVEC
831 r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
833 r = rbd_aio_write(s->image, off, size, rcb->buf, c);
837 #ifdef LIBRBD_SUPPORTS_IOVEC
838 r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
840 r = rbd_aio_read(s->image, off, size, rcb->buf, c);
843 case RBD_AIO_DISCARD:
844 r = rbd_aio_discard_wrapper(s->image, off, size, c);
847 r = rbd_aio_flush_wrapper(s->image, c);
854 goto failed_completion;
862 if (!LIBRBD_USE_IOVEC) {
863 qemu_vfree(acb->bounce);
870 static BlockAIOCB *qemu_rbd_aio_readv(BlockDriverState *bs,
874 BlockCompletionFunc *cb,
877 return rbd_start_aio(bs, sector_num << BDRV_SECTOR_BITS, qiov,
878 (int64_t) nb_sectors << BDRV_SECTOR_BITS, cb, opaque,
882 static BlockAIOCB *qemu_rbd_aio_writev(BlockDriverState *bs,
886 BlockCompletionFunc *cb,
889 return rbd_start_aio(bs, sector_num << BDRV_SECTOR_BITS, qiov,
890 (int64_t) nb_sectors << BDRV_SECTOR_BITS, cb, opaque,
894 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
895 static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs,
896 BlockCompletionFunc *cb,
899 return rbd_start_aio(bs, 0, NULL, 0, cb, opaque, RBD_AIO_FLUSH);
904 static int qemu_rbd_co_flush(BlockDriverState *bs)
906 #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 1)
907 /* rbd_flush added in 0.1.1 */
908 BDRVRBDState *s = bs->opaque;
909 return rbd_flush(s->image);
916 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
918 BDRVRBDState *s = bs->opaque;
919 rbd_image_info_t info;
922 r = rbd_stat(s->image, &info, sizeof(info));
927 bdi->cluster_size = info.obj_size;
931 static int64_t qemu_rbd_getlength(BlockDriverState *bs)
933 BDRVRBDState *s = bs->opaque;
934 rbd_image_info_t info;
937 r = rbd_stat(s->image, &info, sizeof(info));
945 static int qemu_rbd_truncate(BlockDriverState *bs, int64_t offset,
946 PreallocMode prealloc, Error **errp)
948 BDRVRBDState *s = bs->opaque;
951 if (prealloc != PREALLOC_MODE_OFF) {
952 error_setg(errp, "Unsupported preallocation mode '%s'",
953 PreallocMode_str(prealloc));
957 r = rbd_resize(s->image, offset);
959 error_setg_errno(errp, -r, "Failed to resize file");
966 static int qemu_rbd_snap_create(BlockDriverState *bs,
967 QEMUSnapshotInfo *sn_info)
969 BDRVRBDState *s = bs->opaque;
972 if (sn_info->name[0] == '\0') {
973 return -EINVAL; /* we need a name for rbd snapshots */
977 * rbd snapshots are using the name as the user controlled unique identifier
978 * we can't use the rbd snapid for that purpose, as it can't be set
980 if (sn_info->id_str[0] != '\0' &&
981 strcmp(sn_info->id_str, sn_info->name) != 0) {
985 if (strlen(sn_info->name) >= sizeof(sn_info->id_str)) {
989 r = rbd_snap_create(s->image, sn_info->name);
991 error_report("failed to create snap: %s", strerror(-r));
998 static int qemu_rbd_snap_remove(BlockDriverState *bs,
999 const char *snapshot_id,
1000 const char *snapshot_name,
1003 BDRVRBDState *s = bs->opaque;
1006 if (!snapshot_name) {
1007 error_setg(errp, "rbd need a valid snapshot name");
1011 /* If snapshot_id is specified, it must be equal to name, see
1012 qemu_rbd_snap_list() */
1013 if (snapshot_id && strcmp(snapshot_id, snapshot_name)) {
1015 "rbd do not support snapshot id, it should be NULL or "
1016 "equal to snapshot name");
1020 r = rbd_snap_remove(s->image, snapshot_name);
1022 error_setg_errno(errp, -r, "Failed to remove the snapshot");
1027 static int qemu_rbd_snap_rollback(BlockDriverState *bs,
1028 const char *snapshot_name)
1030 BDRVRBDState *s = bs->opaque;
1032 return rbd_snap_rollback(s->image, snapshot_name);
1035 static int qemu_rbd_snap_list(BlockDriverState *bs,
1036 QEMUSnapshotInfo **psn_tab)
1038 BDRVRBDState *s = bs->opaque;
1039 QEMUSnapshotInfo *sn_info, *sn_tab = NULL;
1041 rbd_snap_info_t *snaps;
1042 int max_snaps = RBD_MAX_SNAPS;
1045 snaps = g_new(rbd_snap_info_t, max_snaps);
1046 snap_count = rbd_snap_list(s->image, snaps, &max_snaps);
1047 if (snap_count <= 0) {
1050 } while (snap_count == -ERANGE);
1052 if (snap_count <= 0) {
1056 sn_tab = g_new0(QEMUSnapshotInfo, snap_count);
1058 for (i = 0; i < snap_count; i++) {
1059 const char *snap_name = snaps[i].name;
1061 sn_info = sn_tab + i;
1062 pstrcpy(sn_info->id_str, sizeof(sn_info->id_str), snap_name);
1063 pstrcpy(sn_info->name, sizeof(sn_info->name), snap_name);
1065 sn_info->vm_state_size = snaps[i].size;
1066 sn_info->date_sec = 0;
1067 sn_info->date_nsec = 0;
1068 sn_info->vm_clock_nsec = 0;
1070 rbd_snap_list_end(snaps);
1078 #ifdef LIBRBD_SUPPORTS_DISCARD
1079 static BlockAIOCB *qemu_rbd_aio_pdiscard(BlockDriverState *bs,
1082 BlockCompletionFunc *cb,
1085 return rbd_start_aio(bs, offset, NULL, bytes, cb, opaque,
1090 #ifdef LIBRBD_SUPPORTS_INVALIDATE
1091 static void qemu_rbd_invalidate_cache(BlockDriverState *bs,
1094 BDRVRBDState *s = bs->opaque;
1095 int r = rbd_invalidate_cache(s->image);
1097 error_setg_errno(errp, -r, "Failed to invalidate the cache");
1102 static QemuOptsList qemu_rbd_create_opts = {
1103 .name = "rbd-create-opts",
1104 .head = QTAILQ_HEAD_INITIALIZER(qemu_rbd_create_opts.head),
1107 .name = BLOCK_OPT_SIZE,
1108 .type = QEMU_OPT_SIZE,
1109 .help = "Virtual disk size"
1112 .name = BLOCK_OPT_CLUSTER_SIZE,
1113 .type = QEMU_OPT_SIZE,
1114 .help = "RBD object size"
1117 .name = "password-secret",
1118 .type = QEMU_OPT_STRING,
1119 .help = "ID of secret providing the password",
1121 { /* end of list */ }
1125 static BlockDriver bdrv_rbd = {
1126 .format_name = "rbd",
1127 .instance_size = sizeof(BDRVRBDState),
1128 .bdrv_parse_filename = qemu_rbd_parse_filename,
1129 .bdrv_file_open = qemu_rbd_open,
1130 .bdrv_close = qemu_rbd_close,
1131 .bdrv_reopen_prepare = qemu_rbd_reopen_prepare,
1132 .bdrv_create = qemu_rbd_create,
1133 .bdrv_has_zero_init = bdrv_has_zero_init_1,
1134 .bdrv_get_info = qemu_rbd_getinfo,
1135 .create_opts = &qemu_rbd_create_opts,
1136 .bdrv_getlength = qemu_rbd_getlength,
1137 .bdrv_truncate = qemu_rbd_truncate,
1138 .protocol_name = "rbd",
1140 .bdrv_aio_readv = qemu_rbd_aio_readv,
1141 .bdrv_aio_writev = qemu_rbd_aio_writev,
1143 #ifdef LIBRBD_SUPPORTS_AIO_FLUSH
1144 .bdrv_aio_flush = qemu_rbd_aio_flush,
1146 .bdrv_co_flush_to_disk = qemu_rbd_co_flush,
1149 #ifdef LIBRBD_SUPPORTS_DISCARD
1150 .bdrv_aio_pdiscard = qemu_rbd_aio_pdiscard,
1153 .bdrv_snapshot_create = qemu_rbd_snap_create,
1154 .bdrv_snapshot_delete = qemu_rbd_snap_remove,
1155 .bdrv_snapshot_list = qemu_rbd_snap_list,
1156 .bdrv_snapshot_goto = qemu_rbd_snap_rollback,
1157 #ifdef LIBRBD_SUPPORTS_INVALIDATE
1158 .bdrv_invalidate_cache = qemu_rbd_invalidate_cache,
1162 static void bdrv_rbd_init(void)
1164 bdrv_register(&bdrv_rbd);
1167 block_init(bdrv_rbd_init);