2 #include <linux/ceph/ceph_debug.h>
3 #include <linux/backing-dev.h>
4 #include <linux/ctype.h>
6 #include <linux/inet.h>
9 #include <keys/ceph-type.h>
10 #include <linux/module.h>
11 #include <linux/mount.h>
12 #include <linux/nsproxy.h>
13 #include <linux/parser.h>
14 #include <linux/sched.h>
15 #include <linux/seq_file.h>
16 #include <linux/slab.h>
17 #include <linux/statfs.h>
18 #include <linux/string.h>
19 #include <linux/vmalloc.h>
22 #include <linux/ceph/ceph_features.h>
23 #include <linux/ceph/libceph.h>
24 #include <linux/ceph/debugfs.h>
25 #include <linux/ceph/decode.h>
26 #include <linux/ceph/mon_client.h>
27 #include <linux/ceph/auth.h>
32 * Module compatibility interface. For now it doesn't do anything,
33 * but its existence signals a certain level of functionality.
35 * The data buffer is used to pass information both to and from
36 * libceph. The return value indicates whether libceph determines
37 * it is compatible with the caller (from another kernel module),
38 * given the provided data.
40 * The data pointer can be null.
42 bool libceph_compatible(void *data)
46 EXPORT_SYMBOL(libceph_compatible);
48 const char *ceph_msg_type_name(int type)
51 case CEPH_MSG_SHUTDOWN: return "shutdown";
52 case CEPH_MSG_PING: return "ping";
53 case CEPH_MSG_AUTH: return "auth";
54 case CEPH_MSG_AUTH_REPLY: return "auth_reply";
55 case CEPH_MSG_MON_MAP: return "mon_map";
56 case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
57 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
58 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
59 case CEPH_MSG_STATFS: return "statfs";
60 case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
61 case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
62 case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
63 case CEPH_MSG_MDS_MAP: return "mds_map";
64 case CEPH_MSG_CLIENT_SESSION: return "client_session";
65 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
66 case CEPH_MSG_CLIENT_REQUEST: return "client_request";
67 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
68 case CEPH_MSG_CLIENT_REPLY: return "client_reply";
69 case CEPH_MSG_CLIENT_CAPS: return "client_caps";
70 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
71 case CEPH_MSG_CLIENT_SNAP: return "client_snap";
72 case CEPH_MSG_CLIENT_LEASE: return "client_lease";
73 case CEPH_MSG_OSD_MAP: return "osd_map";
74 case CEPH_MSG_OSD_OP: return "osd_op";
75 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
76 case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
77 default: return "unknown";
80 EXPORT_SYMBOL(ceph_msg_type_name);
83 * Initially learn our fsid, or verify an fsid matches.
85 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
87 if (client->have_fsid) {
88 if (ceph_fsid_compare(&client->fsid, fsid)) {
89 pr_err("bad fsid, had %pU got %pU",
94 memcpy(&client->fsid, fsid, sizeof(*fsid));
98 EXPORT_SYMBOL(ceph_check_fsid);
100 static int strcmp_null(const char *s1, const char *s2)
108 return strcmp(s1, s2);
111 int ceph_compare_options(struct ceph_options *new_opt,
112 struct ceph_client *client)
114 struct ceph_options *opt1 = new_opt;
115 struct ceph_options *opt2 = client->options;
116 int ofs = offsetof(struct ceph_options, mon_addr);
121 * Don't bother comparing options if network namespaces don't
124 if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
127 ret = memcmp(opt1, opt2, ofs);
131 ret = strcmp_null(opt1->name, opt2->name);
135 if (opt1->key && !opt2->key)
137 if (!opt1->key && opt2->key)
139 if (opt1->key && opt2->key) {
140 if (opt1->key->type != opt2->key->type)
142 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
144 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
146 if (opt1->key->len != opt2->key->len)
148 if (opt1->key->key && !opt2->key->key)
150 if (!opt1->key->key && opt2->key->key)
152 if (opt1->key->key && opt2->key->key) {
153 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
159 /* any matching mon ip implies a match */
160 for (i = 0; i < opt1->num_mon; i++) {
161 if (ceph_monmap_contains(client->monc.monmap,
167 EXPORT_SYMBOL(ceph_compare_options);
169 void *ceph_kvmalloc(size_t size, gfp_t flags)
171 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
172 void *ptr = kmalloc(size, flags | __GFP_NOWARN);
177 return __vmalloc(size, flags | __GFP_HIGHMEM, PAGE_KERNEL);
181 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
188 dout("parse_fsid '%s'\n", str);
190 while (*str && i < 16) {
195 if (!isxdigit(str[0]) || !isxdigit(str[1]))
199 if (sscanf(tmp, "%x", &d) < 1)
201 fsid->fsid[i] = d & 0xff;
208 dout("parse_fsid ret %d got fsid %pU", err, fsid);
217 Opt_osdkeepalivetimeout,
228 /* string args above */
233 Opt_cephx_require_signatures,
234 Opt_nocephx_require_signatures,
235 Opt_cephx_sign_messages,
236 Opt_nocephx_sign_messages,
241 static match_table_t opt_tokens = {
242 {Opt_osdtimeout, "osdtimeout=%d"},
243 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
244 {Opt_mount_timeout, "mount_timeout=%d"},
245 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
247 {Opt_fsid, "fsid=%s"},
248 {Opt_name, "name=%s"},
249 {Opt_secret, "secret=%s"},
252 /* string args above */
253 {Opt_share, "share"},
254 {Opt_noshare, "noshare"},
256 {Opt_nocrc, "nocrc"},
257 {Opt_cephx_require_signatures, "cephx_require_signatures"},
258 {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
259 {Opt_cephx_sign_messages, "cephx_sign_messages"},
260 {Opt_nocephx_sign_messages, "nocephx_sign_messages"},
261 {Opt_tcp_nodelay, "tcp_nodelay"},
262 {Opt_notcp_nodelay, "notcp_nodelay"},
266 void ceph_destroy_options(struct ceph_options *opt)
268 dout("destroy_options %p\n", opt);
271 ceph_crypto_key_destroy(opt->key);
274 kfree(opt->mon_addr);
277 EXPORT_SYMBOL(ceph_destroy_options);
279 /* get secret from key store */
280 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
284 struct ceph_crypto_key *ckey;
286 ukey = request_key(&key_type_ceph, name, NULL);
287 if (!ukey || IS_ERR(ukey)) {
288 /* request_key errors don't map nicely to mount(2)
289 errors; don't even try, but still printk */
290 key_err = PTR_ERR(ukey);
293 pr_warn("ceph: Mount failed due to key not found: %s\n",
297 pr_warn("ceph: Mount failed due to expired key: %s\n",
301 pr_warn("ceph: Mount failed due to revoked key: %s\n",
305 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
312 ckey = ukey->payload.data[0];
313 err = ceph_crypto_key_clone(dst, ckey);
316 /* pass through, err is 0 */
324 struct ceph_options *
325 ceph_parse_options(char *options, const char *dev_name,
326 const char *dev_name_end,
327 int (*parse_extra_token)(char *c, void *private),
330 struct ceph_options *opt;
333 substring_t argstr[MAX_OPT_ARGS];
335 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
337 return ERR_PTR(-ENOMEM);
338 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
343 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
346 /* start with defaults */
347 opt->flags = CEPH_OPT_DEFAULT;
348 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
349 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
350 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
353 /* ip1[:port1][,ip2[:port2]...] */
354 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
355 CEPH_MAX_MON, &opt->num_mon);
359 /* parse mount options */
360 while ((c = strsep(&options, ",")) != NULL) {
361 int token, intval, ret;
365 token = match_token((char *)c, opt_tokens, argstr);
366 if (token < 0 && parse_extra_token) {
368 err = parse_extra_token((char *)c, private);
370 pr_err("bad option at '%s'\n", c);
375 if (token < Opt_last_int) {
376 ret = match_int(&argstr[0], &intval);
378 pr_err("bad mount option arg (not int) "
382 dout("got int token %d val %d\n", token, intval);
383 } else if (token > Opt_last_int && token < Opt_last_string) {
384 dout("got string token %d val %s\n", token,
387 dout("got token %d\n", token);
391 err = ceph_parse_ips(argstr[0].from,
397 opt->flags |= CEPH_OPT_MYIP;
401 err = parse_fsid(argstr[0].from, &opt->fsid);
403 opt->flags |= CEPH_OPT_FSID;
406 opt->name = kstrndup(argstr[0].from,
407 argstr[0].to-argstr[0].from,
411 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
416 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
421 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
426 err = get_secret(opt->key, argstr[0].from);
433 pr_warn("ignoring deprecated osdtimeout option\n");
435 case Opt_osdkeepalivetimeout:
436 /* 0 isn't well defined right now, reject it */
437 if (intval < 1 || intval > INT_MAX / 1000) {
438 pr_err("osdkeepalive out of range\n");
442 opt->osd_keepalive_timeout =
443 msecs_to_jiffies(intval * 1000);
445 case Opt_osd_idle_ttl:
446 /* 0 isn't well defined right now, reject it */
447 if (intval < 1 || intval > INT_MAX / 1000) {
448 pr_err("osd_idle_ttl out of range\n");
452 opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
454 case Opt_mount_timeout:
455 /* 0 is "wait forever" (i.e. infinite timeout) */
456 if (intval < 0 || intval > INT_MAX / 1000) {
457 pr_err("mount_timeout out of range\n");
461 opt->mount_timeout = msecs_to_jiffies(intval * 1000);
465 opt->flags &= ~CEPH_OPT_NOSHARE;
468 opt->flags |= CEPH_OPT_NOSHARE;
472 opt->flags &= ~CEPH_OPT_NOCRC;
475 opt->flags |= CEPH_OPT_NOCRC;
478 case Opt_cephx_require_signatures:
479 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
481 case Opt_nocephx_require_signatures:
482 opt->flags |= CEPH_OPT_NOMSGAUTH;
484 case Opt_cephx_sign_messages:
485 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
487 case Opt_nocephx_sign_messages:
488 opt->flags |= CEPH_OPT_NOMSGSIGN;
491 case Opt_tcp_nodelay:
492 opt->flags |= CEPH_OPT_TCP_NODELAY;
494 case Opt_notcp_nodelay:
495 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
507 ceph_destroy_options(opt);
510 EXPORT_SYMBOL(ceph_parse_options);
512 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client)
514 struct ceph_options *opt = client->options;
515 size_t pos = m->count;
518 seq_puts(m, "name=");
519 seq_escape(m, opt->name, ", \t\n\\");
523 seq_puts(m, "secret=<hidden>,");
525 if (opt->flags & CEPH_OPT_FSID)
526 seq_printf(m, "fsid=%pU,", &opt->fsid);
527 if (opt->flags & CEPH_OPT_NOSHARE)
528 seq_puts(m, "noshare,");
529 if (opt->flags & CEPH_OPT_NOCRC)
530 seq_puts(m, "nocrc,");
531 if (opt->flags & CEPH_OPT_NOMSGAUTH)
532 seq_puts(m, "nocephx_require_signatures,");
533 if (opt->flags & CEPH_OPT_NOMSGSIGN)
534 seq_puts(m, "nocephx_sign_messages,");
535 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
536 seq_puts(m, "notcp_nodelay,");
538 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
539 seq_printf(m, "mount_timeout=%d,",
540 jiffies_to_msecs(opt->mount_timeout) / 1000);
541 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
542 seq_printf(m, "osd_idle_ttl=%d,",
543 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
544 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
545 seq_printf(m, "osdkeepalivetimeout=%d,",
546 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
548 /* drop redundant comma */
554 EXPORT_SYMBOL(ceph_print_client_options);
556 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
558 return &client->msgr.inst.addr;
560 EXPORT_SYMBOL(ceph_client_addr);
562 u64 ceph_client_gid(struct ceph_client *client)
564 return client->monc.auth->global_id;
566 EXPORT_SYMBOL(ceph_client_gid);
569 * create a fresh client instance
571 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
572 u64 supported_features,
573 u64 required_features)
575 struct ceph_client *client;
576 struct ceph_entity_addr *myaddr = NULL;
579 client = kzalloc(sizeof(*client), GFP_KERNEL);
581 return ERR_PTR(-ENOMEM);
583 client->private = private;
584 client->options = opt;
586 mutex_init(&client->mount_mutex);
587 init_waitqueue_head(&client->auth_wq);
588 client->auth_err = 0;
590 if (!ceph_test_opt(client, NOMSGAUTH))
591 required_features |= CEPH_FEATURE_MSG_AUTH;
593 client->extra_mon_dispatch = NULL;
594 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
596 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
600 if (ceph_test_opt(client, MYIP))
601 myaddr = &client->options->my_addr;
603 ceph_messenger_init(&client->msgr, myaddr);
606 err = ceph_monc_init(&client->monc, client);
609 err = ceph_osdc_init(&client->osdc, client);
616 ceph_monc_stop(&client->monc);
618 ceph_messenger_fini(&client->msgr);
622 EXPORT_SYMBOL(ceph_create_client);
624 void ceph_destroy_client(struct ceph_client *client)
626 dout("destroy_client %p\n", client);
628 atomic_set(&client->msgr.stopping, 1);
631 ceph_osdc_stop(&client->osdc);
632 ceph_monc_stop(&client->monc);
633 ceph_messenger_fini(&client->msgr);
635 ceph_debugfs_client_cleanup(client);
637 ceph_destroy_options(client->options);
640 dout("destroy_client %p done\n", client);
642 EXPORT_SYMBOL(ceph_destroy_client);
645 * true if we have the mon map (and have thus joined the cluster)
647 static bool have_mon_and_osd_map(struct ceph_client *client)
649 return client->monc.monmap && client->monc.monmap->epoch &&
650 client->osdc.osdmap && client->osdc.osdmap->epoch;
654 * mount: join the ceph cluster, and open root directory.
656 int __ceph_open_session(struct ceph_client *client, unsigned long started)
658 unsigned long timeout = client->options->mount_timeout;
661 /* open session, and wait for mon and osd maps */
662 err = ceph_monc_open_session(&client->monc);
666 while (!have_mon_and_osd_map(client)) {
667 if (timeout && time_after_eq(jiffies, started + timeout))
671 dout("mount waiting for mon_map\n");
672 err = wait_event_interruptible_timeout(client->auth_wq,
673 have_mon_and_osd_map(client) || (client->auth_err < 0),
674 ceph_timeout_jiffies(timeout));
677 if (client->auth_err < 0)
678 return client->auth_err;
681 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
683 ceph_debugfs_client_init(client);
687 EXPORT_SYMBOL(__ceph_open_session);
689 int ceph_open_session(struct ceph_client *client)
692 unsigned long started = jiffies; /* note the start time */
694 dout("open_session start\n");
695 mutex_lock(&client->mount_mutex);
697 ret = __ceph_open_session(client, started);
699 mutex_unlock(&client->mount_mutex);
702 EXPORT_SYMBOL(ceph_open_session);
704 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
705 unsigned long timeout)
710 ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
714 if (client->osdc.osdmap->epoch >= newest_epoch)
717 ceph_osdc_maybe_request_map(&client->osdc);
718 return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
720 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
722 static int __init init_ceph_lib(void)
726 ret = ceph_debugfs_init();
730 ret = ceph_crypto_init();
734 ret = ceph_msgr_init();
738 ret = ceph_osdc_setup();
742 pr_info("loaded (mon/osd proto %d/%d)\n",
743 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
750 ceph_crypto_shutdown();
752 ceph_debugfs_cleanup();
757 static void __exit exit_ceph_lib(void)
759 dout("exit_ceph_lib\n");
760 WARN_ON(!ceph_strings_empty());
764 ceph_crypto_shutdown();
765 ceph_debugfs_cleanup();
768 module_init(init_ceph_lib);
769 module_exit(exit_ceph_lib);
771 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
772 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
773 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
774 MODULE_DESCRIPTION("Ceph core library");
775 MODULE_LICENSE("GPL");