]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * QEMU Block driver for RADOS (Ceph) | |
3 | * | |
4 | * Copyright (C) 2010-2011 Christian Brunner <chb@muc.de>, | |
5 | * Josh Durgin <josh.durgin@dreamhost.com> | |
6 | * | |
7 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
8 | * the COPYING file in the top-level directory. | |
9 | * | |
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. | |
12 | */ | |
13 | ||
14 | #include "qemu/osdep.h" | |
15 | ||
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 "block/qdict.h" | |
22 | #include "crypto/secret.h" | |
23 | #include "qemu/cutils.h" | |
24 | #include "qapi/qmp/qstring.h" | |
25 | #include "qapi/qmp/qdict.h" | |
26 | #include "qapi/qmp/qjson.h" | |
27 | #include "qapi/qmp/qlist.h" | |
28 | #include "qapi/qobject-input-visitor.h" | |
29 | #include "qapi/qapi-visit-block-core.h" | |
30 | ||
31 | /* | |
32 | * When specifying the image filename use: | |
33 | * | |
34 | * rbd:poolname/devicename[@snapshotname][:option1=value1[:option2=value2...]] | |
35 | * | |
36 | * poolname must be the name of an existing rados pool. | |
37 | * | |
38 | * devicename is the name of the rbd image. | |
39 | * | |
40 | * Each option given is used to configure rados, and may be any valid | |
41 | * Ceph option, "id", or "conf". | |
42 | * | |
43 | * The "id" option indicates what user we should authenticate as to | |
44 | * the Ceph cluster. If it is excluded we will use the Ceph default | |
45 | * (normally 'admin'). | |
46 | * | |
47 | * The "conf" option specifies a Ceph configuration file to read. If | |
48 | * it is not specified, we will read from the default Ceph locations | |
49 | * (e.g., /etc/ceph/ceph.conf). To avoid reading _any_ configuration | |
50 | * file, specify conf=/dev/null. | |
51 | * | |
52 | * Configuration values containing :, @, or = can be escaped with a | |
53 | * leading "\". | |
54 | */ | |
55 | ||
56 | /* rbd_aio_discard added in 0.1.2 */ | |
57 | #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 2) | |
58 | #define LIBRBD_SUPPORTS_DISCARD | |
59 | #else | |
60 | #undef LIBRBD_SUPPORTS_DISCARD | |
61 | #endif | |
62 | ||
63 | #define OBJ_MAX_SIZE (1UL << OBJ_DEFAULT_OBJ_ORDER) | |
64 | ||
65 | #define RBD_MAX_SNAPS 100 | |
66 | ||
67 | /* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */ | |
68 | #ifdef LIBRBD_SUPPORTS_IOVEC | |
69 | #define LIBRBD_USE_IOVEC 1 | |
70 | #else | |
71 | #define LIBRBD_USE_IOVEC 0 | |
72 | #endif | |
73 | ||
74 | typedef enum { | |
75 | RBD_AIO_READ, | |
76 | RBD_AIO_WRITE, | |
77 | RBD_AIO_DISCARD, | |
78 | RBD_AIO_FLUSH | |
79 | } RBDAIOCmd; | |
80 | ||
81 | typedef struct RBDAIOCB { | |
82 | BlockAIOCB common; | |
83 | int64_t ret; | |
84 | QEMUIOVector *qiov; | |
85 | char *bounce; | |
86 | RBDAIOCmd cmd; | |
87 | int error; | |
88 | struct BDRVRBDState *s; | |
89 | } RBDAIOCB; | |
90 | ||
91 | typedef struct RADOSCB { | |
92 | RBDAIOCB *acb; | |
93 | struct BDRVRBDState *s; | |
94 | int64_t size; | |
95 | char *buf; | |
96 | int64_t ret; | |
97 | } RADOSCB; | |
98 | ||
99 | typedef struct BDRVRBDState { | |
100 | rados_t cluster; | |
101 | rados_ioctx_t io_ctx; | |
102 | rbd_image_t image; | |
103 | char *image_name; | |
104 | char *snap; | |
105 | } BDRVRBDState; | |
106 | ||
107 | static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, | |
108 | BlockdevOptionsRbd *opts, bool cache, | |
109 | const char *keypairs, const char *secretid, | |
110 | Error **errp); | |
111 | ||
112 | static char *qemu_rbd_next_tok(char *src, char delim, char **p) | |
113 | { | |
114 | char *end; | |
115 | ||
116 | *p = NULL; | |
117 | ||
118 | for (end = src; *end; ++end) { | |
119 | if (*end == delim) { | |
120 | break; | |
121 | } | |
122 | if (*end == '\\' && end[1] != '\0') { | |
123 | end++; | |
124 | } | |
125 | } | |
126 | if (*end == delim) { | |
127 | *p = end + 1; | |
128 | *end = '\0'; | |
129 | } | |
130 | return src; | |
131 | } | |
132 | ||
133 | static void qemu_rbd_unescape(char *src) | |
134 | { | |
135 | char *p; | |
136 | ||
137 | for (p = src; *src; ++src, ++p) { | |
138 | if (*src == '\\' && src[1] != '\0') { | |
139 | src++; | |
140 | } | |
141 | *p = *src; | |
142 | } | |
143 | *p = '\0'; | |
144 | } | |
145 | ||
146 | static void qemu_rbd_parse_filename(const char *filename, QDict *options, | |
147 | Error **errp) | |
148 | { | |
149 | const char *start; | |
150 | char *p, *buf; | |
151 | QList *keypairs = NULL; | |
152 | char *found_str; | |
153 | ||
154 | if (!strstart(filename, "rbd:", &start)) { | |
155 | error_setg(errp, "File name must start with 'rbd:'"); | |
156 | return; | |
157 | } | |
158 | ||
159 | buf = g_strdup(start); | |
160 | p = buf; | |
161 | ||
162 | found_str = qemu_rbd_next_tok(p, '/', &p); | |
163 | if (!p) { | |
164 | error_setg(errp, "Pool name is required"); | |
165 | goto done; | |
166 | } | |
167 | qemu_rbd_unescape(found_str); | |
168 | qdict_put_str(options, "pool", found_str); | |
169 | ||
170 | if (strchr(p, '@')) { | |
171 | found_str = qemu_rbd_next_tok(p, '@', &p); | |
172 | qemu_rbd_unescape(found_str); | |
173 | qdict_put_str(options, "image", found_str); | |
174 | ||
175 | found_str = qemu_rbd_next_tok(p, ':', &p); | |
176 | qemu_rbd_unescape(found_str); | |
177 | qdict_put_str(options, "snapshot", found_str); | |
178 | } else { | |
179 | found_str = qemu_rbd_next_tok(p, ':', &p); | |
180 | qemu_rbd_unescape(found_str); | |
181 | qdict_put_str(options, "image", found_str); | |
182 | } | |
183 | if (!p) { | |
184 | goto done; | |
185 | } | |
186 | ||
187 | /* The following are essentially all key/value pairs, and we treat | |
188 | * 'id' and 'conf' a bit special. Key/value pairs may be in any order. */ | |
189 | while (p) { | |
190 | char *name, *value; | |
191 | name = qemu_rbd_next_tok(p, '=', &p); | |
192 | if (!p) { | |
193 | error_setg(errp, "conf option %s has no value", name); | |
194 | break; | |
195 | } | |
196 | ||
197 | qemu_rbd_unescape(name); | |
198 | ||
199 | value = qemu_rbd_next_tok(p, ':', &p); | |
200 | qemu_rbd_unescape(value); | |
201 | ||
202 | if (!strcmp(name, "conf")) { | |
203 | qdict_put_str(options, "conf", value); | |
204 | } else if (!strcmp(name, "id")) { | |
205 | qdict_put_str(options, "user", value); | |
206 | } else { | |
207 | /* | |
208 | * We pass these internally to qemu_rbd_set_keypairs(), so | |
209 | * we can get away with the simpler list of [ "key1", | |
210 | * "value1", "key2", "value2" ] rather than a raw dict | |
211 | * { "key1": "value1", "key2": "value2" } where we can't | |
212 | * guarantee order, or even a more correct but complex | |
213 | * [ { "key1": "value1" }, { "key2": "value2" } ] | |
214 | */ | |
215 | if (!keypairs) { | |
216 | keypairs = qlist_new(); | |
217 | } | |
218 | qlist_append_str(keypairs, name); | |
219 | qlist_append_str(keypairs, value); | |
220 | } | |
221 | } | |
222 | ||
223 | if (keypairs) { | |
224 | qdict_put(options, "=keyvalue-pairs", | |
225 | qobject_to_json(QOBJECT(keypairs))); | |
226 | } | |
227 | ||
228 | done: | |
229 | g_free(buf); | |
230 | qobject_unref(keypairs); | |
231 | return; | |
232 | } | |
233 | ||
234 | ||
235 | static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp) | |
236 | { | |
237 | /* XXX Does RBD support AIO on less than 512-byte alignment? */ | |
238 | bs->bl.request_alignment = 512; | |
239 | } | |
240 | ||
241 | ||
242 | static int qemu_rbd_set_auth(rados_t cluster, BlockdevOptionsRbd *opts, | |
243 | Error **errp) | |
244 | { | |
245 | char *key, *acr; | |
246 | int r; | |
247 | GString *accu; | |
248 | RbdAuthModeList *auth; | |
249 | ||
250 | if (opts->key_secret) { | |
251 | key = qcrypto_secret_lookup_as_base64(opts->key_secret, errp); | |
252 | if (!key) { | |
253 | return -EIO; | |
254 | } | |
255 | r = rados_conf_set(cluster, "key", key); | |
256 | g_free(key); | |
257 | if (r < 0) { | |
258 | error_setg_errno(errp, -r, "Could not set 'key'"); | |
259 | return r; | |
260 | } | |
261 | } | |
262 | ||
263 | if (opts->has_auth_client_required) { | |
264 | accu = g_string_new(""); | |
265 | for (auth = opts->auth_client_required; auth; auth = auth->next) { | |
266 | if (accu->str[0]) { | |
267 | g_string_append_c(accu, ';'); | |
268 | } | |
269 | g_string_append(accu, RbdAuthMode_str(auth->value)); | |
270 | } | |
271 | acr = g_string_free(accu, FALSE); | |
272 | r = rados_conf_set(cluster, "auth_client_required", acr); | |
273 | g_free(acr); | |
274 | if (r < 0) { | |
275 | error_setg_errno(errp, -r, | |
276 | "Could not set 'auth_client_required'"); | |
277 | return r; | |
278 | } | |
279 | } | |
280 | ||
281 | return 0; | |
282 | } | |
283 | ||
284 | static int qemu_rbd_set_keypairs(rados_t cluster, const char *keypairs_json, | |
285 | Error **errp) | |
286 | { | |
287 | QList *keypairs; | |
288 | QString *name; | |
289 | QString *value; | |
290 | const char *key; | |
291 | size_t remaining; | |
292 | int ret = 0; | |
293 | ||
294 | if (!keypairs_json) { | |
295 | return ret; | |
296 | } | |
297 | keypairs = qobject_to(QList, | |
298 | qobject_from_json(keypairs_json, &error_abort)); | |
299 | remaining = qlist_size(keypairs) / 2; | |
300 | assert(remaining); | |
301 | ||
302 | while (remaining--) { | |
303 | name = qobject_to(QString, qlist_pop(keypairs)); | |
304 | value = qobject_to(QString, qlist_pop(keypairs)); | |
305 | assert(name && value); | |
306 | key = qstring_get_str(name); | |
307 | ||
308 | ret = rados_conf_set(cluster, key, qstring_get_str(value)); | |
309 | qobject_unref(value); | |
310 | if (ret < 0) { | |
311 | error_setg_errno(errp, -ret, "invalid conf option %s", key); | |
312 | qobject_unref(name); | |
313 | ret = -EINVAL; | |
314 | break; | |
315 | } | |
316 | qobject_unref(name); | |
317 | } | |
318 | ||
319 | qobject_unref(keypairs); | |
320 | return ret; | |
321 | } | |
322 | ||
323 | static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs) | |
324 | { | |
325 | if (LIBRBD_USE_IOVEC) { | |
326 | RBDAIOCB *acb = rcb->acb; | |
327 | iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0, | |
328 | acb->qiov->size - offs); | |
329 | } else { | |
330 | memset(rcb->buf + offs, 0, rcb->size - offs); | |
331 | } | |
332 | } | |
333 | ||
334 | static QemuOptsList runtime_opts = { | |
335 | .name = "rbd", | |
336 | .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head), | |
337 | .desc = { | |
338 | { | |
339 | .name = "pool", | |
340 | .type = QEMU_OPT_STRING, | |
341 | .help = "Rados pool name", | |
342 | }, | |
343 | { | |
344 | .name = "image", | |
345 | .type = QEMU_OPT_STRING, | |
346 | .help = "Image name in the pool", | |
347 | }, | |
348 | { | |
349 | .name = "conf", | |
350 | .type = QEMU_OPT_STRING, | |
351 | .help = "Rados config file location", | |
352 | }, | |
353 | { | |
354 | .name = "snapshot", | |
355 | .type = QEMU_OPT_STRING, | |
356 | .help = "Ceph snapshot name", | |
357 | }, | |
358 | { | |
359 | /* maps to 'id' in rados_create() */ | |
360 | .name = "user", | |
361 | .type = QEMU_OPT_STRING, | |
362 | .help = "Rados id name", | |
363 | }, | |
364 | /* | |
365 | * server.* extracted manually, see qemu_rbd_mon_host() | |
366 | */ | |
367 | { /* end of list */ } | |
368 | }, | |
369 | }; | |
370 | ||
371 | /* FIXME Deprecate and remove keypairs or make it available in QMP. */ | |
372 | static int qemu_rbd_do_create(BlockdevCreateOptions *options, | |
373 | const char *keypairs, const char *password_secret, | |
374 | Error **errp) | |
375 | { | |
376 | BlockdevCreateOptionsRbd *opts = &options->u.rbd; | |
377 | rados_t cluster; | |
378 | rados_ioctx_t io_ctx; | |
379 | int obj_order = 0; | |
380 | int ret; | |
381 | ||
382 | assert(options->driver == BLOCKDEV_DRIVER_RBD); | |
383 | if (opts->location->has_snapshot) { | |
384 | error_setg(errp, "Can't use snapshot name for image creation"); | |
385 | return -EINVAL; | |
386 | } | |
387 | ||
388 | if (opts->has_cluster_size) { | |
389 | int64_t objsize = opts->cluster_size; | |
390 | if ((objsize - 1) & objsize) { /* not a power of 2? */ | |
391 | error_setg(errp, "obj size needs to be power of 2"); | |
392 | return -EINVAL; | |
393 | } | |
394 | if (objsize < 4096) { | |
395 | error_setg(errp, "obj size too small"); | |
396 | return -EINVAL; | |
397 | } | |
398 | obj_order = ctz32(objsize); | |
399 | } | |
400 | ||
401 | ret = qemu_rbd_connect(&cluster, &io_ctx, opts->location, false, keypairs, | |
402 | password_secret, errp); | |
403 | if (ret < 0) { | |
404 | return ret; | |
405 | } | |
406 | ||
407 | ret = rbd_create(io_ctx, opts->location->image, opts->size, &obj_order); | |
408 | if (ret < 0) { | |
409 | error_setg_errno(errp, -ret, "error rbd create"); | |
410 | goto out; | |
411 | } | |
412 | ||
413 | ret = 0; | |
414 | out: | |
415 | rados_ioctx_destroy(io_ctx); | |
416 | rados_shutdown(cluster); | |
417 | return ret; | |
418 | } | |
419 | ||
420 | static int qemu_rbd_co_create(BlockdevCreateOptions *options, Error **errp) | |
421 | { | |
422 | return qemu_rbd_do_create(options, NULL, NULL, errp); | |
423 | } | |
424 | ||
425 | static int coroutine_fn qemu_rbd_co_create_opts(const char *filename, | |
426 | QemuOpts *opts, | |
427 | Error **errp) | |
428 | { | |
429 | BlockdevCreateOptions *create_options; | |
430 | BlockdevCreateOptionsRbd *rbd_opts; | |
431 | BlockdevOptionsRbd *loc; | |
432 | Error *local_err = NULL; | |
433 | const char *keypairs, *password_secret; | |
434 | QDict *options = NULL; | |
435 | int ret = 0; | |
436 | ||
437 | create_options = g_new0(BlockdevCreateOptions, 1); | |
438 | create_options->driver = BLOCKDEV_DRIVER_RBD; | |
439 | rbd_opts = &create_options->u.rbd; | |
440 | ||
441 | rbd_opts->location = g_new0(BlockdevOptionsRbd, 1); | |
442 | ||
443 | password_secret = qemu_opt_get(opts, "password-secret"); | |
444 | ||
445 | /* Read out options */ | |
446 | rbd_opts->size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), | |
447 | BDRV_SECTOR_SIZE); | |
448 | rbd_opts->cluster_size = qemu_opt_get_size_del(opts, | |
449 | BLOCK_OPT_CLUSTER_SIZE, 0); | |
450 | rbd_opts->has_cluster_size = (rbd_opts->cluster_size != 0); | |
451 | ||
452 | options = qdict_new(); | |
453 | qemu_rbd_parse_filename(filename, options, &local_err); | |
454 | if (local_err) { | |
455 | ret = -EINVAL; | |
456 | error_propagate(errp, local_err); | |
457 | goto exit; | |
458 | } | |
459 | ||
460 | /* | |
461 | * Caution: while qdict_get_try_str() is fine, getting non-string | |
462 | * types would require more care. When @options come from -blockdev | |
463 | * or blockdev_add, its members are typed according to the QAPI | |
464 | * schema, but when they come from -drive, they're all QString. | |
465 | */ | |
466 | loc = rbd_opts->location; | |
467 | loc->pool = g_strdup(qdict_get_try_str(options, "pool")); | |
468 | loc->conf = g_strdup(qdict_get_try_str(options, "conf")); | |
469 | loc->has_conf = !!loc->conf; | |
470 | loc->user = g_strdup(qdict_get_try_str(options, "user")); | |
471 | loc->has_user = !!loc->user; | |
472 | loc->image = g_strdup(qdict_get_try_str(options, "image")); | |
473 | keypairs = qdict_get_try_str(options, "=keyvalue-pairs"); | |
474 | ||
475 | ret = qemu_rbd_do_create(create_options, keypairs, password_secret, errp); | |
476 | if (ret < 0) { | |
477 | goto exit; | |
478 | } | |
479 | ||
480 | exit: | |
481 | qobject_unref(options); | |
482 | qapi_free_BlockdevCreateOptions(create_options); | |
483 | return ret; | |
484 | } | |
485 | ||
486 | /* | |
487 | * This aio completion is being called from rbd_finish_bh() and runs in qemu | |
488 | * BH context. | |
489 | */ | |
490 | static void qemu_rbd_complete_aio(RADOSCB *rcb) | |
491 | { | |
492 | RBDAIOCB *acb = rcb->acb; | |
493 | int64_t r; | |
494 | ||
495 | r = rcb->ret; | |
496 | ||
497 | if (acb->cmd != RBD_AIO_READ) { | |
498 | if (r < 0) { | |
499 | acb->ret = r; | |
500 | acb->error = 1; | |
501 | } else if (!acb->error) { | |
502 | acb->ret = rcb->size; | |
503 | } | |
504 | } else { | |
505 | if (r < 0) { | |
506 | qemu_rbd_memset(rcb, 0); | |
507 | acb->ret = r; | |
508 | acb->error = 1; | |
509 | } else if (r < rcb->size) { | |
510 | qemu_rbd_memset(rcb, r); | |
511 | if (!acb->error) { | |
512 | acb->ret = rcb->size; | |
513 | } | |
514 | } else if (!acb->error) { | |
515 | acb->ret = r; | |
516 | } | |
517 | } | |
518 | ||
519 | g_free(rcb); | |
520 | ||
521 | if (!LIBRBD_USE_IOVEC) { | |
522 | if (acb->cmd == RBD_AIO_READ) { | |
523 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); | |
524 | } | |
525 | qemu_vfree(acb->bounce); | |
526 | } | |
527 | ||
528 | acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret)); | |
529 | ||
530 | qemu_aio_unref(acb); | |
531 | } | |
532 | ||
533 | static char *qemu_rbd_mon_host(BlockdevOptionsRbd *opts, Error **errp) | |
534 | { | |
535 | const char **vals; | |
536 | const char *host, *port; | |
537 | char *rados_str; | |
538 | InetSocketAddressBaseList *p; | |
539 | int i, cnt; | |
540 | ||
541 | if (!opts->has_server) { | |
542 | return NULL; | |
543 | } | |
544 | ||
545 | for (cnt = 0, p = opts->server; p; p = p->next) { | |
546 | cnt++; | |
547 | } | |
548 | ||
549 | vals = g_new(const char *, cnt + 1); | |
550 | ||
551 | for (i = 0, p = opts->server; p; p = p->next, i++) { | |
552 | host = p->value->host; | |
553 | port = p->value->port; | |
554 | ||
555 | if (strchr(host, ':')) { | |
556 | vals[i] = g_strdup_printf("[%s]:%s", host, port); | |
557 | } else { | |
558 | vals[i] = g_strdup_printf("%s:%s", host, port); | |
559 | } | |
560 | } | |
561 | vals[i] = NULL; | |
562 | ||
563 | rados_str = i ? g_strjoinv(";", (char **)vals) : NULL; | |
564 | g_strfreev((char **)vals); | |
565 | return rados_str; | |
566 | } | |
567 | ||
568 | static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, | |
569 | BlockdevOptionsRbd *opts, bool cache, | |
570 | const char *keypairs, const char *secretid, | |
571 | Error **errp) | |
572 | { | |
573 | char *mon_host = NULL; | |
574 | Error *local_err = NULL; | |
575 | int r; | |
576 | ||
577 | if (secretid) { | |
578 | if (opts->key_secret) { | |
579 | error_setg(errp, | |
580 | "Legacy 'password-secret' clashes with 'key-secret'"); | |
581 | return -EINVAL; | |
582 | } | |
583 | opts->key_secret = g_strdup(secretid); | |
584 | opts->has_key_secret = true; | |
585 | } | |
586 | ||
587 | mon_host = qemu_rbd_mon_host(opts, &local_err); | |
588 | if (local_err) { | |
589 | error_propagate(errp, local_err); | |
590 | r = -EINVAL; | |
591 | goto failed_opts; | |
592 | } | |
593 | ||
594 | r = rados_create(cluster, opts->user); | |
595 | if (r < 0) { | |
596 | error_setg_errno(errp, -r, "error initializing"); | |
597 | goto failed_opts; | |
598 | } | |
599 | ||
600 | /* try default location when conf=NULL, but ignore failure */ | |
601 | r = rados_conf_read_file(*cluster, opts->conf); | |
602 | if (opts->has_conf && r < 0) { | |
603 | error_setg_errno(errp, -r, "error reading conf file %s", opts->conf); | |
604 | goto failed_shutdown; | |
605 | } | |
606 | ||
607 | r = qemu_rbd_set_keypairs(*cluster, keypairs, errp); | |
608 | if (r < 0) { | |
609 | goto failed_shutdown; | |
610 | } | |
611 | ||
612 | if (mon_host) { | |
613 | r = rados_conf_set(*cluster, "mon_host", mon_host); | |
614 | if (r < 0) { | |
615 | goto failed_shutdown; | |
616 | } | |
617 | } | |
618 | ||
619 | r = qemu_rbd_set_auth(*cluster, opts, errp); | |
620 | if (r < 0) { | |
621 | goto failed_shutdown; | |
622 | } | |
623 | ||
624 | /* | |
625 | * Fallback to more conservative semantics if setting cache | |
626 | * options fails. Ignore errors from setting rbd_cache because the | |
627 | * only possible error is that the option does not exist, and | |
628 | * librbd defaults to no caching. If write through caching cannot | |
629 | * be set up, fall back to no caching. | |
630 | */ | |
631 | if (cache) { | |
632 | rados_conf_set(*cluster, "rbd_cache", "true"); | |
633 | } else { | |
634 | rados_conf_set(*cluster, "rbd_cache", "false"); | |
635 | } | |
636 | ||
637 | r = rados_connect(*cluster); | |
638 | if (r < 0) { | |
639 | error_setg_errno(errp, -r, "error connecting"); | |
640 | goto failed_shutdown; | |
641 | } | |
642 | ||
643 | r = rados_ioctx_create(*cluster, opts->pool, io_ctx); | |
644 | if (r < 0) { | |
645 | error_setg_errno(errp, -r, "error opening pool %s", opts->pool); | |
646 | goto failed_shutdown; | |
647 | } | |
648 | ||
649 | return 0; | |
650 | ||
651 | failed_shutdown: | |
652 | rados_shutdown(*cluster); | |
653 | failed_opts: | |
654 | g_free(mon_host); | |
655 | return r; | |
656 | } | |
657 | ||
658 | static int qemu_rbd_convert_options(QDict *options, BlockdevOptionsRbd **opts, | |
659 | Error **errp) | |
660 | { | |
661 | Visitor *v; | |
662 | Error *local_err = NULL; | |
663 | ||
664 | /* Convert the remaining options into a QAPI object */ | |
665 | v = qobject_input_visitor_new_flat_confused(options, errp); | |
666 | if (!v) { | |
667 | return -EINVAL; | |
668 | } | |
669 | ||
670 | visit_type_BlockdevOptionsRbd(v, NULL, opts, &local_err); | |
671 | visit_free(v); | |
672 | ||
673 | if (local_err) { | |
674 | error_propagate(errp, local_err); | |
675 | return -EINVAL; | |
676 | } | |
677 | ||
678 | return 0; | |
679 | } | |
680 | ||
681 | static int qemu_rbd_attempt_legacy_options(QDict *options, | |
682 | BlockdevOptionsRbd **opts, | |
683 | char **keypairs) | |
684 | { | |
685 | char *filename; | |
686 | int r; | |
687 | ||
688 | filename = g_strdup(qdict_get_try_str(options, "filename")); | |
689 | if (!filename) { | |
690 | return -EINVAL; | |
691 | } | |
692 | qdict_del(options, "filename"); | |
693 | ||
694 | qemu_rbd_parse_filename(filename, options, NULL); | |
695 | ||
696 | /* keypairs freed by caller */ | |
697 | *keypairs = g_strdup(qdict_get_try_str(options, "=keyvalue-pairs")); | |
698 | if (*keypairs) { | |
699 | qdict_del(options, "=keyvalue-pairs"); | |
700 | } | |
701 | ||
702 | r = qemu_rbd_convert_options(options, opts, NULL); | |
703 | ||
704 | g_free(filename); | |
705 | return r; | |
706 | } | |
707 | ||
708 | static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags, | |
709 | Error **errp) | |
710 | { | |
711 | BDRVRBDState *s = bs->opaque; | |
712 | BlockdevOptionsRbd *opts = NULL; | |
713 | const QDictEntry *e; | |
714 | Error *local_err = NULL; | |
715 | char *keypairs, *secretid; | |
716 | int r; | |
717 | ||
718 | keypairs = g_strdup(qdict_get_try_str(options, "=keyvalue-pairs")); | |
719 | if (keypairs) { | |
720 | qdict_del(options, "=keyvalue-pairs"); | |
721 | } | |
722 | ||
723 | secretid = g_strdup(qdict_get_try_str(options, "password-secret")); | |
724 | if (secretid) { | |
725 | qdict_del(options, "password-secret"); | |
726 | } | |
727 | ||
728 | r = qemu_rbd_convert_options(options, &opts, &local_err); | |
729 | if (local_err) { | |
730 | /* If keypairs are present, that means some options are present in | |
731 | * the modern option format. Don't attempt to parse legacy option | |
732 | * formats, as we won't support mixed usage. */ | |
733 | if (keypairs) { | |
734 | error_propagate(errp, local_err); | |
735 | goto out; | |
736 | } | |
737 | ||
738 | /* If the initial attempt to convert and process the options failed, | |
739 | * we may be attempting to open an image file that has the rbd options | |
740 | * specified in the older format consisting of all key/value pairs | |
741 | * encoded in the filename. Go ahead and attempt to parse the | |
742 | * filename, and see if we can pull out the required options. */ | |
743 | r = qemu_rbd_attempt_legacy_options(options, &opts, &keypairs); | |
744 | if (r < 0) { | |
745 | /* Propagate the original error, not the legacy parsing fallback | |
746 | * error, as the latter was just a best-effort attempt. */ | |
747 | error_propagate(errp, local_err); | |
748 | goto out; | |
749 | } | |
750 | /* Take care whenever deciding to actually deprecate; once this ability | |
751 | * is removed, we will not be able to open any images with legacy-styled | |
752 | * backing image strings. */ | |
753 | warn_report("RBD options encoded in the filename as keyvalue pairs " | |
754 | "is deprecated"); | |
755 | } | |
756 | ||
757 | /* Remove the processed options from the QDict (the visitor processes | |
758 | * _all_ options in the QDict) */ | |
759 | while ((e = qdict_first(options))) { | |
760 | qdict_del(options, e->key); | |
761 | } | |
762 | ||
763 | r = qemu_rbd_connect(&s->cluster, &s->io_ctx, opts, | |
764 | !(flags & BDRV_O_NOCACHE), keypairs, secretid, errp); | |
765 | if (r < 0) { | |
766 | goto out; | |
767 | } | |
768 | ||
769 | s->snap = g_strdup(opts->snapshot); | |
770 | s->image_name = g_strdup(opts->image); | |
771 | ||
772 | /* rbd_open is always r/w */ | |
773 | r = rbd_open(s->io_ctx, s->image_name, &s->image, s->snap); | |
774 | if (r < 0) { | |
775 | error_setg_errno(errp, -r, "error reading header from %s", | |
776 | s->image_name); | |
777 | goto failed_open; | |
778 | } | |
779 | ||
780 | /* If we are using an rbd snapshot, we must be r/o, otherwise | |
781 | * leave as-is */ | |
782 | if (s->snap != NULL) { | |
783 | r = bdrv_apply_auto_read_only(bs, "rbd snapshots are read-only", errp); | |
784 | if (r < 0) { | |
785 | rbd_close(s->image); | |
786 | goto failed_open; | |
787 | } | |
788 | } | |
789 | ||
790 | r = 0; | |
791 | goto out; | |
792 | ||
793 | failed_open: | |
794 | rados_ioctx_destroy(s->io_ctx); | |
795 | g_free(s->snap); | |
796 | g_free(s->image_name); | |
797 | rados_shutdown(s->cluster); | |
798 | out: | |
799 | qapi_free_BlockdevOptionsRbd(opts); | |
800 | g_free(keypairs); | |
801 | g_free(secretid); | |
802 | return r; | |
803 | } | |
804 | ||
805 | ||
806 | /* Since RBD is currently always opened R/W via the API, | |
807 | * we just need to check if we are using a snapshot or not, in | |
808 | * order to determine if we will allow it to be R/W */ | |
809 | static int qemu_rbd_reopen_prepare(BDRVReopenState *state, | |
810 | BlockReopenQueue *queue, Error **errp) | |
811 | { | |
812 | BDRVRBDState *s = state->bs->opaque; | |
813 | int ret = 0; | |
814 | ||
815 | if (s->snap && state->flags & BDRV_O_RDWR) { | |
816 | error_setg(errp, | |
817 | "Cannot change node '%s' to r/w when using RBD snapshot", | |
818 | bdrv_get_device_or_node_name(state->bs)); | |
819 | ret = -EINVAL; | |
820 | } | |
821 | ||
822 | return ret; | |
823 | } | |
824 | ||
825 | static void qemu_rbd_close(BlockDriverState *bs) | |
826 | { | |
827 | BDRVRBDState *s = bs->opaque; | |
828 | ||
829 | rbd_close(s->image); | |
830 | rados_ioctx_destroy(s->io_ctx); | |
831 | g_free(s->snap); | |
832 | g_free(s->image_name); | |
833 | rados_shutdown(s->cluster); | |
834 | } | |
835 | ||
836 | static const AIOCBInfo rbd_aiocb_info = { | |
837 | .aiocb_size = sizeof(RBDAIOCB), | |
838 | }; | |
839 | ||
840 | static void rbd_finish_bh(void *opaque) | |
841 | { | |
842 | RADOSCB *rcb = opaque; | |
843 | qemu_rbd_complete_aio(rcb); | |
844 | } | |
845 | ||
846 | /* | |
847 | * This is the callback function for rbd_aio_read and _write | |
848 | * | |
849 | * Note: this function is being called from a non qemu thread so | |
850 | * we need to be careful about what we do here. Generally we only | |
851 | * schedule a BH, and do the rest of the io completion handling | |
852 | * from rbd_finish_bh() which runs in a qemu context. | |
853 | */ | |
854 | static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb) | |
855 | { | |
856 | RBDAIOCB *acb = rcb->acb; | |
857 | ||
858 | rcb->ret = rbd_aio_get_return_value(c); | |
859 | rbd_aio_release(c); | |
860 | ||
861 | aio_bh_schedule_oneshot(bdrv_get_aio_context(acb->common.bs), | |
862 | rbd_finish_bh, rcb); | |
863 | } | |
864 | ||
865 | static int rbd_aio_discard_wrapper(rbd_image_t image, | |
866 | uint64_t off, | |
867 | uint64_t len, | |
868 | rbd_completion_t comp) | |
869 | { | |
870 | #ifdef LIBRBD_SUPPORTS_DISCARD | |
871 | return rbd_aio_discard(image, off, len, comp); | |
872 | #else | |
873 | return -ENOTSUP; | |
874 | #endif | |
875 | } | |
876 | ||
877 | static int rbd_aio_flush_wrapper(rbd_image_t image, | |
878 | rbd_completion_t comp) | |
879 | { | |
880 | #ifdef LIBRBD_SUPPORTS_AIO_FLUSH | |
881 | return rbd_aio_flush(image, comp); | |
882 | #else | |
883 | return -ENOTSUP; | |
884 | #endif | |
885 | } | |
886 | ||
887 | static BlockAIOCB *rbd_start_aio(BlockDriverState *bs, | |
888 | int64_t off, | |
889 | QEMUIOVector *qiov, | |
890 | int64_t size, | |
891 | BlockCompletionFunc *cb, | |
892 | void *opaque, | |
893 | RBDAIOCmd cmd) | |
894 | { | |
895 | RBDAIOCB *acb; | |
896 | RADOSCB *rcb = NULL; | |
897 | rbd_completion_t c; | |
898 | int r; | |
899 | ||
900 | BDRVRBDState *s = bs->opaque; | |
901 | ||
902 | acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque); | |
903 | acb->cmd = cmd; | |
904 | acb->qiov = qiov; | |
905 | assert(!qiov || qiov->size == size); | |
906 | ||
907 | rcb = g_new(RADOSCB, 1); | |
908 | ||
909 | if (!LIBRBD_USE_IOVEC) { | |
910 | if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) { | |
911 | acb->bounce = NULL; | |
912 | } else { | |
913 | acb->bounce = qemu_try_blockalign(bs, qiov->size); | |
914 | if (acb->bounce == NULL) { | |
915 | goto failed; | |
916 | } | |
917 | } | |
918 | if (cmd == RBD_AIO_WRITE) { | |
919 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); | |
920 | } | |
921 | rcb->buf = acb->bounce; | |
922 | } | |
923 | ||
924 | acb->ret = 0; | |
925 | acb->error = 0; | |
926 | acb->s = s; | |
927 | ||
928 | rcb->acb = acb; | |
929 | rcb->s = acb->s; | |
930 | rcb->size = size; | |
931 | r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c); | |
932 | if (r < 0) { | |
933 | goto failed; | |
934 | } | |
935 | ||
936 | switch (cmd) { | |
937 | case RBD_AIO_WRITE: | |
938 | #ifdef LIBRBD_SUPPORTS_IOVEC | |
939 | r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c); | |
940 | #else | |
941 | r = rbd_aio_write(s->image, off, size, rcb->buf, c); | |
942 | #endif | |
943 | break; | |
944 | case RBD_AIO_READ: | |
945 | #ifdef LIBRBD_SUPPORTS_IOVEC | |
946 | r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c); | |
947 | #else | |
948 | r = rbd_aio_read(s->image, off, size, rcb->buf, c); | |
949 | #endif | |
950 | break; | |
951 | case RBD_AIO_DISCARD: | |
952 | r = rbd_aio_discard_wrapper(s->image, off, size, c); | |
953 | break; | |
954 | case RBD_AIO_FLUSH: | |
955 | r = rbd_aio_flush_wrapper(s->image, c); | |
956 | break; | |
957 | default: | |
958 | r = -EINVAL; | |
959 | } | |
960 | ||
961 | if (r < 0) { | |
962 | goto failed_completion; | |
963 | } | |
964 | return &acb->common; | |
965 | ||
966 | failed_completion: | |
967 | rbd_aio_release(c); | |
968 | failed: | |
969 | g_free(rcb); | |
970 | if (!LIBRBD_USE_IOVEC) { | |
971 | qemu_vfree(acb->bounce); | |
972 | } | |
973 | ||
974 | qemu_aio_unref(acb); | |
975 | return NULL; | |
976 | } | |
977 | ||
978 | static BlockAIOCB *qemu_rbd_aio_preadv(BlockDriverState *bs, | |
979 | uint64_t offset, uint64_t bytes, | |
980 | QEMUIOVector *qiov, int flags, | |
981 | BlockCompletionFunc *cb, | |
982 | void *opaque) | |
983 | { | |
984 | return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque, | |
985 | RBD_AIO_READ); | |
986 | } | |
987 | ||
988 | static BlockAIOCB *qemu_rbd_aio_pwritev(BlockDriverState *bs, | |
989 | uint64_t offset, uint64_t bytes, | |
990 | QEMUIOVector *qiov, int flags, | |
991 | BlockCompletionFunc *cb, | |
992 | void *opaque) | |
993 | { | |
994 | return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque, | |
995 | RBD_AIO_WRITE); | |
996 | } | |
997 | ||
998 | #ifdef LIBRBD_SUPPORTS_AIO_FLUSH | |
999 | static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs, | |
1000 | BlockCompletionFunc *cb, | |
1001 | void *opaque) | |
1002 | { | |
1003 | return rbd_start_aio(bs, 0, NULL, 0, cb, opaque, RBD_AIO_FLUSH); | |
1004 | } | |
1005 | ||
1006 | #else | |
1007 | ||
1008 | static int qemu_rbd_co_flush(BlockDriverState *bs) | |
1009 | { | |
1010 | #if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 1) | |
1011 | /* rbd_flush added in 0.1.1 */ | |
1012 | BDRVRBDState *s = bs->opaque; | |
1013 | return rbd_flush(s->image); | |
1014 | #else | |
1015 | return 0; | |
1016 | #endif | |
1017 | } | |
1018 | #endif | |
1019 | ||
1020 | static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi) | |
1021 | { | |
1022 | BDRVRBDState *s = bs->opaque; | |
1023 | rbd_image_info_t info; | |
1024 | int r; | |
1025 | ||
1026 | r = rbd_stat(s->image, &info, sizeof(info)); | |
1027 | if (r < 0) { | |
1028 | return r; | |
1029 | } | |
1030 | ||
1031 | bdi->cluster_size = info.obj_size; | |
1032 | return 0; | |
1033 | } | |
1034 | ||
1035 | static int64_t qemu_rbd_getlength(BlockDriverState *bs) | |
1036 | { | |
1037 | BDRVRBDState *s = bs->opaque; | |
1038 | rbd_image_info_t info; | |
1039 | int r; | |
1040 | ||
1041 | r = rbd_stat(s->image, &info, sizeof(info)); | |
1042 | if (r < 0) { | |
1043 | return r; | |
1044 | } | |
1045 | ||
1046 | return info.size; | |
1047 | } | |
1048 | ||
1049 | static int coroutine_fn qemu_rbd_co_truncate(BlockDriverState *bs, | |
1050 | int64_t offset, | |
1051 | PreallocMode prealloc, | |
1052 | Error **errp) | |
1053 | { | |
1054 | BDRVRBDState *s = bs->opaque; | |
1055 | int r; | |
1056 | ||
1057 | if (prealloc != PREALLOC_MODE_OFF) { | |
1058 | error_setg(errp, "Unsupported preallocation mode '%s'", | |
1059 | PreallocMode_str(prealloc)); | |
1060 | return -ENOTSUP; | |
1061 | } | |
1062 | ||
1063 | r = rbd_resize(s->image, offset); | |
1064 | if (r < 0) { | |
1065 | error_setg_errno(errp, -r, "Failed to resize file"); | |
1066 | return r; | |
1067 | } | |
1068 | ||
1069 | return 0; | |
1070 | } | |
1071 | ||
1072 | static int qemu_rbd_snap_create(BlockDriverState *bs, | |
1073 | QEMUSnapshotInfo *sn_info) | |
1074 | { | |
1075 | BDRVRBDState *s = bs->opaque; | |
1076 | int r; | |
1077 | ||
1078 | if (sn_info->name[0] == '\0') { | |
1079 | return -EINVAL; /* we need a name for rbd snapshots */ | |
1080 | } | |
1081 | ||
1082 | /* | |
1083 | * rbd snapshots are using the name as the user controlled unique identifier | |
1084 | * we can't use the rbd snapid for that purpose, as it can't be set | |
1085 | */ | |
1086 | if (sn_info->id_str[0] != '\0' && | |
1087 | strcmp(sn_info->id_str, sn_info->name) != 0) { | |
1088 | return -EINVAL; | |
1089 | } | |
1090 | ||
1091 | if (strlen(sn_info->name) >= sizeof(sn_info->id_str)) { | |
1092 | return -ERANGE; | |
1093 | } | |
1094 | ||
1095 | r = rbd_snap_create(s->image, sn_info->name); | |
1096 | if (r < 0) { | |
1097 | error_report("failed to create snap: %s", strerror(-r)); | |
1098 | return r; | |
1099 | } | |
1100 | ||
1101 | return 0; | |
1102 | } | |
1103 | ||
1104 | static int qemu_rbd_snap_remove(BlockDriverState *bs, | |
1105 | const char *snapshot_id, | |
1106 | const char *snapshot_name, | |
1107 | Error **errp) | |
1108 | { | |
1109 | BDRVRBDState *s = bs->opaque; | |
1110 | int r; | |
1111 | ||
1112 | if (!snapshot_name) { | |
1113 | error_setg(errp, "rbd need a valid snapshot name"); | |
1114 | return -EINVAL; | |
1115 | } | |
1116 | ||
1117 | /* If snapshot_id is specified, it must be equal to name, see | |
1118 | qemu_rbd_snap_list() */ | |
1119 | if (snapshot_id && strcmp(snapshot_id, snapshot_name)) { | |
1120 | error_setg(errp, | |
1121 | "rbd do not support snapshot id, it should be NULL or " | |
1122 | "equal to snapshot name"); | |
1123 | return -EINVAL; | |
1124 | } | |
1125 | ||
1126 | r = rbd_snap_remove(s->image, snapshot_name); | |
1127 | if (r < 0) { | |
1128 | error_setg_errno(errp, -r, "Failed to remove the snapshot"); | |
1129 | } | |
1130 | return r; | |
1131 | } | |
1132 | ||
1133 | static int qemu_rbd_snap_rollback(BlockDriverState *bs, | |
1134 | const char *snapshot_name) | |
1135 | { | |
1136 | BDRVRBDState *s = bs->opaque; | |
1137 | ||
1138 | return rbd_snap_rollback(s->image, snapshot_name); | |
1139 | } | |
1140 | ||
1141 | static int qemu_rbd_snap_list(BlockDriverState *bs, | |
1142 | QEMUSnapshotInfo **psn_tab) | |
1143 | { | |
1144 | BDRVRBDState *s = bs->opaque; | |
1145 | QEMUSnapshotInfo *sn_info, *sn_tab = NULL; | |
1146 | int i, snap_count; | |
1147 | rbd_snap_info_t *snaps; | |
1148 | int max_snaps = RBD_MAX_SNAPS; | |
1149 | ||
1150 | do { | |
1151 | snaps = g_new(rbd_snap_info_t, max_snaps); | |
1152 | snap_count = rbd_snap_list(s->image, snaps, &max_snaps); | |
1153 | if (snap_count <= 0) { | |
1154 | g_free(snaps); | |
1155 | } | |
1156 | } while (snap_count == -ERANGE); | |
1157 | ||
1158 | if (snap_count <= 0) { | |
1159 | goto done; | |
1160 | } | |
1161 | ||
1162 | sn_tab = g_new0(QEMUSnapshotInfo, snap_count); | |
1163 | ||
1164 | for (i = 0; i < snap_count; i++) { | |
1165 | const char *snap_name = snaps[i].name; | |
1166 | ||
1167 | sn_info = sn_tab + i; | |
1168 | pstrcpy(sn_info->id_str, sizeof(sn_info->id_str), snap_name); | |
1169 | pstrcpy(sn_info->name, sizeof(sn_info->name), snap_name); | |
1170 | ||
1171 | sn_info->vm_state_size = snaps[i].size; | |
1172 | sn_info->date_sec = 0; | |
1173 | sn_info->date_nsec = 0; | |
1174 | sn_info->vm_clock_nsec = 0; | |
1175 | } | |
1176 | rbd_snap_list_end(snaps); | |
1177 | g_free(snaps); | |
1178 | ||
1179 | done: | |
1180 | *psn_tab = sn_tab; | |
1181 | return snap_count; | |
1182 | } | |
1183 | ||
1184 | #ifdef LIBRBD_SUPPORTS_DISCARD | |
1185 | static BlockAIOCB *qemu_rbd_aio_pdiscard(BlockDriverState *bs, | |
1186 | int64_t offset, | |
1187 | int bytes, | |
1188 | BlockCompletionFunc *cb, | |
1189 | void *opaque) | |
1190 | { | |
1191 | return rbd_start_aio(bs, offset, NULL, bytes, cb, opaque, | |
1192 | RBD_AIO_DISCARD); | |
1193 | } | |
1194 | #endif | |
1195 | ||
1196 | #ifdef LIBRBD_SUPPORTS_INVALIDATE | |
1197 | static void coroutine_fn qemu_rbd_co_invalidate_cache(BlockDriverState *bs, | |
1198 | Error **errp) | |
1199 | { | |
1200 | BDRVRBDState *s = bs->opaque; | |
1201 | int r = rbd_invalidate_cache(s->image); | |
1202 | if (r < 0) { | |
1203 | error_setg_errno(errp, -r, "Failed to invalidate the cache"); | |
1204 | } | |
1205 | } | |
1206 | #endif | |
1207 | ||
1208 | static QemuOptsList qemu_rbd_create_opts = { | |
1209 | .name = "rbd-create-opts", | |
1210 | .head = QTAILQ_HEAD_INITIALIZER(qemu_rbd_create_opts.head), | |
1211 | .desc = { | |
1212 | { | |
1213 | .name = BLOCK_OPT_SIZE, | |
1214 | .type = QEMU_OPT_SIZE, | |
1215 | .help = "Virtual disk size" | |
1216 | }, | |
1217 | { | |
1218 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
1219 | .type = QEMU_OPT_SIZE, | |
1220 | .help = "RBD object size" | |
1221 | }, | |
1222 | { | |
1223 | .name = "password-secret", | |
1224 | .type = QEMU_OPT_STRING, | |
1225 | .help = "ID of secret providing the password", | |
1226 | }, | |
1227 | { /* end of list */ } | |
1228 | } | |
1229 | }; | |
1230 | ||
1231 | static const char *const qemu_rbd_strong_runtime_opts[] = { | |
1232 | "pool", | |
1233 | "image", | |
1234 | "conf", | |
1235 | "snapshot", | |
1236 | "user", | |
1237 | "server.", | |
1238 | "password-secret", | |
1239 | ||
1240 | NULL | |
1241 | }; | |
1242 | ||
1243 | static BlockDriver bdrv_rbd = { | |
1244 | .format_name = "rbd", | |
1245 | .instance_size = sizeof(BDRVRBDState), | |
1246 | .bdrv_parse_filename = qemu_rbd_parse_filename, | |
1247 | .bdrv_refresh_limits = qemu_rbd_refresh_limits, | |
1248 | .bdrv_file_open = qemu_rbd_open, | |
1249 | .bdrv_close = qemu_rbd_close, | |
1250 | .bdrv_reopen_prepare = qemu_rbd_reopen_prepare, | |
1251 | .bdrv_co_create = qemu_rbd_co_create, | |
1252 | .bdrv_co_create_opts = qemu_rbd_co_create_opts, | |
1253 | .bdrv_has_zero_init = bdrv_has_zero_init_1, | |
1254 | .bdrv_get_info = qemu_rbd_getinfo, | |
1255 | .create_opts = &qemu_rbd_create_opts, | |
1256 | .bdrv_getlength = qemu_rbd_getlength, | |
1257 | .bdrv_co_truncate = qemu_rbd_co_truncate, | |
1258 | .protocol_name = "rbd", | |
1259 | ||
1260 | .bdrv_aio_preadv = qemu_rbd_aio_preadv, | |
1261 | .bdrv_aio_pwritev = qemu_rbd_aio_pwritev, | |
1262 | ||
1263 | #ifdef LIBRBD_SUPPORTS_AIO_FLUSH | |
1264 | .bdrv_aio_flush = qemu_rbd_aio_flush, | |
1265 | #else | |
1266 | .bdrv_co_flush_to_disk = qemu_rbd_co_flush, | |
1267 | #endif | |
1268 | ||
1269 | #ifdef LIBRBD_SUPPORTS_DISCARD | |
1270 | .bdrv_aio_pdiscard = qemu_rbd_aio_pdiscard, | |
1271 | #endif | |
1272 | ||
1273 | .bdrv_snapshot_create = qemu_rbd_snap_create, | |
1274 | .bdrv_snapshot_delete = qemu_rbd_snap_remove, | |
1275 | .bdrv_snapshot_list = qemu_rbd_snap_list, | |
1276 | .bdrv_snapshot_goto = qemu_rbd_snap_rollback, | |
1277 | #ifdef LIBRBD_SUPPORTS_INVALIDATE | |
1278 | .bdrv_co_invalidate_cache = qemu_rbd_co_invalidate_cache, | |
1279 | #endif | |
1280 | ||
1281 | .strong_runtime_opts = qemu_rbd_strong_runtime_opts, | |
1282 | }; | |
1283 | ||
1284 | static void bdrv_rbd_init(void) | |
1285 | { | |
1286 | bdrv_register(&bdrv_rbd); | |
1287 | } | |
1288 | ||
1289 | block_init(bdrv_rbd_init); |