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 static int param_get_supported_features(char *buffer,
49 const struct kernel_param *kp)
51 return sprintf(buffer, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT);
53 static const struct kernel_param_ops param_ops_supported_features = {
54 .get = param_get_supported_features,
56 module_param_cb(supported_features, ¶m_ops_supported_features, NULL,
59 const char *ceph_msg_type_name(int type)
62 case CEPH_MSG_SHUTDOWN: return "shutdown";
63 case CEPH_MSG_PING: return "ping";
64 case CEPH_MSG_AUTH: return "auth";
65 case CEPH_MSG_AUTH_REPLY: return "auth_reply";
66 case CEPH_MSG_MON_MAP: return "mon_map";
67 case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
68 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
69 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
70 case CEPH_MSG_STATFS: return "statfs";
71 case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
72 case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
73 case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
74 case CEPH_MSG_MDS_MAP: return "mds_map";
75 case CEPH_MSG_FS_MAP_USER: return "fs_map_user";
76 case CEPH_MSG_CLIENT_SESSION: return "client_session";
77 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
78 case CEPH_MSG_CLIENT_REQUEST: return "client_request";
79 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
80 case CEPH_MSG_CLIENT_REPLY: return "client_reply";
81 case CEPH_MSG_CLIENT_CAPS: return "client_caps";
82 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
83 case CEPH_MSG_CLIENT_QUOTA: return "client_quota";
84 case CEPH_MSG_CLIENT_SNAP: return "client_snap";
85 case CEPH_MSG_CLIENT_LEASE: return "client_lease";
86 case CEPH_MSG_POOLOP_REPLY: return "poolop_reply";
87 case CEPH_MSG_POOLOP: return "poolop";
88 case CEPH_MSG_MON_COMMAND: return "mon_command";
89 case CEPH_MSG_MON_COMMAND_ACK: return "mon_command_ack";
90 case CEPH_MSG_OSD_MAP: return "osd_map";
91 case CEPH_MSG_OSD_OP: return "osd_op";
92 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
93 case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
94 case CEPH_MSG_OSD_BACKOFF: return "osd_backoff";
95 default: return "unknown";
98 EXPORT_SYMBOL(ceph_msg_type_name);
101 * Initially learn our fsid, or verify an fsid matches.
103 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
105 if (client->have_fsid) {
106 if (ceph_fsid_compare(&client->fsid, fsid)) {
107 pr_err("bad fsid, had %pU got %pU",
108 &client->fsid, fsid);
112 memcpy(&client->fsid, fsid, sizeof(*fsid));
116 EXPORT_SYMBOL(ceph_check_fsid);
118 static int strcmp_null(const char *s1, const char *s2)
126 return strcmp(s1, s2);
129 int ceph_compare_options(struct ceph_options *new_opt,
130 struct ceph_client *client)
132 struct ceph_options *opt1 = new_opt;
133 struct ceph_options *opt2 = client->options;
134 int ofs = offsetof(struct ceph_options, mon_addr);
139 * Don't bother comparing options if network namespaces don't
142 if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
145 ret = memcmp(opt1, opt2, ofs);
149 ret = strcmp_null(opt1->name, opt2->name);
153 if (opt1->key && !opt2->key)
155 if (!opt1->key && opt2->key)
157 if (opt1->key && opt2->key) {
158 if (opt1->key->type != opt2->key->type)
160 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
162 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
164 if (opt1->key->len != opt2->key->len)
166 if (opt1->key->key && !opt2->key->key)
168 if (!opt1->key->key && opt2->key->key)
170 if (opt1->key->key && opt2->key->key) {
171 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
177 /* any matching mon ip implies a match */
178 for (i = 0; i < opt1->num_mon; i++) {
179 if (ceph_monmap_contains(client->monc.monmap,
185 EXPORT_SYMBOL(ceph_compare_options);
187 void *ceph_kvmalloc(size_t size, gfp_t flags)
189 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
190 void *ptr = kmalloc(size, flags | __GFP_NOWARN);
195 return __vmalloc(size, flags, PAGE_KERNEL);
199 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
206 dout("parse_fsid '%s'\n", str);
208 while (*str && i < 16) {
213 if (!isxdigit(str[0]) || !isxdigit(str[1]))
217 if (sscanf(tmp, "%x", &d) < 1)
219 fsid->fsid[i] = d & 0xff;
226 dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
235 Opt_osdkeepalivetimeout,
238 Opt_osd_request_timeout,
247 /* string args above */
252 Opt_cephx_require_signatures,
253 Opt_nocephx_require_signatures,
254 Opt_cephx_sign_messages,
255 Opt_nocephx_sign_messages,
260 static match_table_t opt_tokens = {
261 {Opt_osdtimeout, "osdtimeout=%d"},
262 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
263 {Opt_mount_timeout, "mount_timeout=%d"},
264 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
265 {Opt_osd_request_timeout, "osd_request_timeout=%d"},
267 {Opt_fsid, "fsid=%s"},
268 {Opt_name, "name=%s"},
269 {Opt_secret, "secret=%s"},
272 /* string args above */
273 {Opt_share, "share"},
274 {Opt_noshare, "noshare"},
276 {Opt_nocrc, "nocrc"},
277 {Opt_cephx_require_signatures, "cephx_require_signatures"},
278 {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
279 {Opt_cephx_sign_messages, "cephx_sign_messages"},
280 {Opt_nocephx_sign_messages, "nocephx_sign_messages"},
281 {Opt_tcp_nodelay, "tcp_nodelay"},
282 {Opt_notcp_nodelay, "notcp_nodelay"},
286 void ceph_destroy_options(struct ceph_options *opt)
288 dout("destroy_options %p\n", opt);
291 ceph_crypto_key_destroy(opt->key);
294 kfree(opt->mon_addr);
297 EXPORT_SYMBOL(ceph_destroy_options);
299 /* get secret from key store */
300 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
304 struct ceph_crypto_key *ckey;
306 ukey = request_key(&key_type_ceph, name, NULL);
308 /* request_key errors don't map nicely to mount(2)
309 errors; don't even try, but still printk */
310 key_err = PTR_ERR(ukey);
313 pr_warn("ceph: Mount failed due to key not found: %s\n",
317 pr_warn("ceph: Mount failed due to expired key: %s\n",
321 pr_warn("ceph: Mount failed due to revoked key: %s\n",
325 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
332 ckey = ukey->payload.data[0];
333 err = ceph_crypto_key_clone(dst, ckey);
336 /* pass through, err is 0 */
344 struct ceph_options *
345 ceph_parse_options(char *options, const char *dev_name,
346 const char *dev_name_end,
347 int (*parse_extra_token)(char *c, void *private),
350 struct ceph_options *opt;
353 substring_t argstr[MAX_OPT_ARGS];
355 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
357 return ERR_PTR(-ENOMEM);
358 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
363 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
366 /* start with defaults */
367 opt->flags = CEPH_OPT_DEFAULT;
368 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
369 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
370 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
371 opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
374 /* ip1[:port1][,ip2[:port2]...] */
375 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
376 CEPH_MAX_MON, &opt->num_mon);
380 /* parse mount options */
381 while ((c = strsep(&options, ",")) != NULL) {
386 token = match_token((char *)c, opt_tokens, argstr);
387 if (token < 0 && parse_extra_token) {
389 err = parse_extra_token((char *)c, private);
391 pr_err("bad option at '%s'\n", c);
396 if (token < Opt_last_int) {
397 err = match_int(&argstr[0], &intval);
399 pr_err("bad option arg (not int) at '%s'\n", c);
402 dout("got int token %d val %d\n", token, intval);
403 } else if (token > Opt_last_int && token < Opt_last_string) {
404 dout("got string token %d val %s\n", token,
407 dout("got token %d\n", token);
411 err = ceph_parse_ips(argstr[0].from,
417 opt->flags |= CEPH_OPT_MYIP;
421 err = parse_fsid(argstr[0].from, &opt->fsid);
423 opt->flags |= CEPH_OPT_FSID;
427 opt->name = kstrndup(argstr[0].from,
428 argstr[0].to-argstr[0].from,
436 ceph_crypto_key_destroy(opt->key);
439 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
444 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
449 ceph_crypto_key_destroy(opt->key);
452 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
457 err = get_secret(opt->key, argstr[0].from);
464 pr_warn("ignoring deprecated osdtimeout option\n");
466 case Opt_osdkeepalivetimeout:
467 /* 0 isn't well defined right now, reject it */
468 if (intval < 1 || intval > INT_MAX / 1000) {
469 pr_err("osdkeepalive out of range\n");
473 opt->osd_keepalive_timeout =
474 msecs_to_jiffies(intval * 1000);
476 case Opt_osd_idle_ttl:
477 /* 0 isn't well defined right now, reject it */
478 if (intval < 1 || intval > INT_MAX / 1000) {
479 pr_err("osd_idle_ttl out of range\n");
483 opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
485 case Opt_mount_timeout:
486 /* 0 is "wait forever" (i.e. infinite timeout) */
487 if (intval < 0 || intval > INT_MAX / 1000) {
488 pr_err("mount_timeout out of range\n");
492 opt->mount_timeout = msecs_to_jiffies(intval * 1000);
494 case Opt_osd_request_timeout:
495 /* 0 is "wait forever" (i.e. infinite timeout) */
496 if (intval < 0 || intval > INT_MAX / 1000) {
497 pr_err("osd_request_timeout out of range\n");
501 opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
505 opt->flags &= ~CEPH_OPT_NOSHARE;
508 opt->flags |= CEPH_OPT_NOSHARE;
512 opt->flags &= ~CEPH_OPT_NOCRC;
515 opt->flags |= CEPH_OPT_NOCRC;
518 case Opt_cephx_require_signatures:
519 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
521 case Opt_nocephx_require_signatures:
522 opt->flags |= CEPH_OPT_NOMSGAUTH;
524 case Opt_cephx_sign_messages:
525 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
527 case Opt_nocephx_sign_messages:
528 opt->flags |= CEPH_OPT_NOMSGSIGN;
531 case Opt_tcp_nodelay:
532 opt->flags |= CEPH_OPT_TCP_NODELAY;
534 case Opt_notcp_nodelay:
535 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
547 ceph_destroy_options(opt);
550 EXPORT_SYMBOL(ceph_parse_options);
552 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client)
554 struct ceph_options *opt = client->options;
555 size_t pos = m->count;
558 seq_puts(m, "name=");
559 seq_escape(m, opt->name, ", \t\n\\");
563 seq_puts(m, "secret=<hidden>,");
565 if (opt->flags & CEPH_OPT_FSID)
566 seq_printf(m, "fsid=%pU,", &opt->fsid);
567 if (opt->flags & CEPH_OPT_NOSHARE)
568 seq_puts(m, "noshare,");
569 if (opt->flags & CEPH_OPT_NOCRC)
570 seq_puts(m, "nocrc,");
571 if (opt->flags & CEPH_OPT_NOMSGAUTH)
572 seq_puts(m, "nocephx_require_signatures,");
573 if (opt->flags & CEPH_OPT_NOMSGSIGN)
574 seq_puts(m, "nocephx_sign_messages,");
575 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
576 seq_puts(m, "notcp_nodelay,");
578 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
579 seq_printf(m, "mount_timeout=%d,",
580 jiffies_to_msecs(opt->mount_timeout) / 1000);
581 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
582 seq_printf(m, "osd_idle_ttl=%d,",
583 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
584 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
585 seq_printf(m, "osdkeepalivetimeout=%d,",
586 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
587 if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
588 seq_printf(m, "osd_request_timeout=%d,",
589 jiffies_to_msecs(opt->osd_request_timeout) / 1000);
591 /* drop redundant comma */
597 EXPORT_SYMBOL(ceph_print_client_options);
599 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
601 return &client->msgr.inst.addr;
603 EXPORT_SYMBOL(ceph_client_addr);
605 u64 ceph_client_gid(struct ceph_client *client)
607 return client->monc.auth->global_id;
609 EXPORT_SYMBOL(ceph_client_gid);
612 * create a fresh client instance
614 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
616 struct ceph_client *client;
617 struct ceph_entity_addr *myaddr = NULL;
620 err = wait_for_random_bytes();
624 client = kzalloc(sizeof(*client), GFP_KERNEL);
626 return ERR_PTR(-ENOMEM);
628 client->private = private;
629 client->options = opt;
631 mutex_init(&client->mount_mutex);
632 init_waitqueue_head(&client->auth_wq);
633 client->auth_err = 0;
635 client->extra_mon_dispatch = NULL;
636 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
637 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
639 if (!ceph_test_opt(client, NOMSGAUTH))
640 client->required_features |= CEPH_FEATURE_MSG_AUTH;
643 if (ceph_test_opt(client, MYIP))
644 myaddr = &client->options->my_addr;
646 ceph_messenger_init(&client->msgr, myaddr);
649 err = ceph_monc_init(&client->monc, client);
652 err = ceph_osdc_init(&client->osdc, client);
659 ceph_monc_stop(&client->monc);
661 ceph_messenger_fini(&client->msgr);
665 EXPORT_SYMBOL(ceph_create_client);
667 void ceph_destroy_client(struct ceph_client *client)
669 dout("destroy_client %p\n", client);
671 atomic_set(&client->msgr.stopping, 1);
674 ceph_osdc_stop(&client->osdc);
675 ceph_monc_stop(&client->monc);
676 ceph_messenger_fini(&client->msgr);
678 ceph_debugfs_client_cleanup(client);
680 ceph_destroy_options(client->options);
683 dout("destroy_client %p done\n", client);
685 EXPORT_SYMBOL(ceph_destroy_client);
688 * true if we have the mon map (and have thus joined the cluster)
690 static bool have_mon_and_osd_map(struct ceph_client *client)
692 return client->monc.monmap && client->monc.monmap->epoch &&
693 client->osdc.osdmap && client->osdc.osdmap->epoch;
697 * mount: join the ceph cluster, and open root directory.
699 int __ceph_open_session(struct ceph_client *client, unsigned long started)
701 unsigned long timeout = client->options->mount_timeout;
704 /* open session, and wait for mon and osd maps */
705 err = ceph_monc_open_session(&client->monc);
709 while (!have_mon_and_osd_map(client)) {
710 if (timeout && time_after_eq(jiffies, started + timeout))
714 dout("mount waiting for mon_map\n");
715 err = wait_event_interruptible_timeout(client->auth_wq,
716 have_mon_and_osd_map(client) || (client->auth_err < 0),
717 ceph_timeout_jiffies(timeout));
720 if (client->auth_err < 0)
721 return client->auth_err;
724 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
726 ceph_debugfs_client_init(client);
730 EXPORT_SYMBOL(__ceph_open_session);
733 int ceph_open_session(struct ceph_client *client)
736 unsigned long started = jiffies; /* note the start time */
738 dout("open_session start\n");
739 mutex_lock(&client->mount_mutex);
741 ret = __ceph_open_session(client, started);
743 mutex_unlock(&client->mount_mutex);
746 EXPORT_SYMBOL(ceph_open_session);
749 static int __init init_ceph_lib(void)
753 ret = ceph_debugfs_init();
757 ret = ceph_crypto_init();
761 ret = ceph_msgr_init();
765 ret = ceph_osdc_setup();
769 pr_info("loaded (mon/osd proto %d/%d)\n",
770 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
777 ceph_crypto_shutdown();
779 ceph_debugfs_cleanup();
784 static void __exit exit_ceph_lib(void)
786 dout("exit_ceph_lib\n");
787 WARN_ON(!ceph_strings_empty());
791 ceph_crypto_shutdown();
792 ceph_debugfs_cleanup();
795 module_init(init_ceph_lib);
796 module_exit(exit_ceph_lib);
798 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
799 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
800 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
801 MODULE_DESCRIPTION("Ceph core library");
802 MODULE_LICENSE("GPL");