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);
49 * find filename portion of a path (/foo/bar/baz -> baz)
51 const char *ceph_file_part(const char *s, int len)
53 const char *e = s + len;
55 while (e != s && *(e-1) != '/')
59 EXPORT_SYMBOL(ceph_file_part);
61 const char *ceph_msg_type_name(int type)
64 case CEPH_MSG_SHUTDOWN: return "shutdown";
65 case CEPH_MSG_PING: return "ping";
66 case CEPH_MSG_AUTH: return "auth";
67 case CEPH_MSG_AUTH_REPLY: return "auth_reply";
68 case CEPH_MSG_MON_MAP: return "mon_map";
69 case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
70 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
71 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
72 case CEPH_MSG_STATFS: return "statfs";
73 case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
74 case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
75 case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
76 case CEPH_MSG_MDS_MAP: return "mds_map";
77 case CEPH_MSG_CLIENT_SESSION: return "client_session";
78 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
79 case CEPH_MSG_CLIENT_REQUEST: return "client_request";
80 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
81 case CEPH_MSG_CLIENT_REPLY: return "client_reply";
82 case CEPH_MSG_CLIENT_CAPS: return "client_caps";
83 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
84 case CEPH_MSG_CLIENT_SNAP: return "client_snap";
85 case CEPH_MSG_CLIENT_LEASE: return "client_lease";
86 case CEPH_MSG_OSD_MAP: return "osd_map";
87 case CEPH_MSG_OSD_OP: return "osd_op";
88 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
89 case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
90 default: return "unknown";
93 EXPORT_SYMBOL(ceph_msg_type_name);
96 * Initially learn our fsid, or verify an fsid matches.
98 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
100 if (client->have_fsid) {
101 if (ceph_fsid_compare(&client->fsid, fsid)) {
102 pr_err("bad fsid, had %pU got %pU",
103 &client->fsid, fsid);
107 memcpy(&client->fsid, fsid, sizeof(*fsid));
111 EXPORT_SYMBOL(ceph_check_fsid);
113 static int strcmp_null(const char *s1, const char *s2)
121 return strcmp(s1, s2);
124 int ceph_compare_options(struct ceph_options *new_opt,
125 struct ceph_client *client)
127 struct ceph_options *opt1 = new_opt;
128 struct ceph_options *opt2 = client->options;
129 int ofs = offsetof(struct ceph_options, mon_addr);
134 * Don't bother comparing options if network namespaces don't
137 if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
140 ret = memcmp(opt1, opt2, ofs);
144 ret = strcmp_null(opt1->name, opt2->name);
148 if (opt1->key && !opt2->key)
150 if (!opt1->key && opt2->key)
152 if (opt1->key && opt2->key) {
153 if (opt1->key->type != opt2->key->type)
155 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
157 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
159 if (opt1->key->len != opt2->key->len)
161 if (opt1->key->key && !opt2->key->key)
163 if (!opt1->key->key && opt2->key->key)
165 if (opt1->key->key && opt2->key->key) {
166 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
172 /* any matching mon ip implies a match */
173 for (i = 0; i < opt1->num_mon; i++) {
174 if (ceph_monmap_contains(client->monc.monmap,
180 EXPORT_SYMBOL(ceph_compare_options);
182 void *ceph_kvmalloc(size_t size, gfp_t flags)
184 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
185 void *ptr = kmalloc(size, flags | __GFP_NOWARN);
190 return __vmalloc(size, flags, PAGE_KERNEL);
194 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
201 dout("parse_fsid '%s'\n", str);
203 while (*str && i < 16) {
208 if (!isxdigit(str[0]) || !isxdigit(str[1]))
212 if (sscanf(tmp, "%x", &d) < 1)
214 fsid->fsid[i] = d & 0xff;
221 dout("parse_fsid ret %d got fsid %pU", err, fsid);
230 Opt_osdkeepalivetimeout,
233 Opt_osd_request_timeout,
242 /* string args above */
247 Opt_cephx_require_signatures,
248 Opt_nocephx_require_signatures,
249 Opt_cephx_sign_messages,
250 Opt_nocephx_sign_messages,
255 static match_table_t opt_tokens = {
256 {Opt_osdtimeout, "osdtimeout=%d"},
257 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
258 {Opt_mount_timeout, "mount_timeout=%d"},
259 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
260 {Opt_osd_request_timeout, "osd_request_timeout=%d"},
262 {Opt_fsid, "fsid=%s"},
263 {Opt_name, "name=%s"},
264 {Opt_secret, "secret=%s"},
267 /* string args above */
268 {Opt_share, "share"},
269 {Opt_noshare, "noshare"},
271 {Opt_nocrc, "nocrc"},
272 {Opt_cephx_require_signatures, "cephx_require_signatures"},
273 {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
274 {Opt_cephx_sign_messages, "cephx_sign_messages"},
275 {Opt_nocephx_sign_messages, "nocephx_sign_messages"},
276 {Opt_tcp_nodelay, "tcp_nodelay"},
277 {Opt_notcp_nodelay, "notcp_nodelay"},
281 void ceph_destroy_options(struct ceph_options *opt)
283 dout("destroy_options %p\n", opt);
286 ceph_crypto_key_destroy(opt->key);
289 kfree(opt->mon_addr);
292 EXPORT_SYMBOL(ceph_destroy_options);
294 /* get secret from key store */
295 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
299 struct ceph_crypto_key *ckey;
301 ukey = request_key(&key_type_ceph, name, NULL);
302 if (!ukey || IS_ERR(ukey)) {
303 /* request_key errors don't map nicely to mount(2)
304 errors; don't even try, but still printk */
305 key_err = PTR_ERR(ukey);
308 pr_warn("ceph: Mount failed due to key not found: %s\n",
312 pr_warn("ceph: Mount failed due to expired key: %s\n",
316 pr_warn("ceph: Mount failed due to revoked key: %s\n",
320 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
327 ckey = ukey->payload.data[0];
328 err = ceph_crypto_key_clone(dst, ckey);
331 /* pass through, err is 0 */
339 struct ceph_options *
340 ceph_parse_options(char *options, const char *dev_name,
341 const char *dev_name_end,
342 int (*parse_extra_token)(char *c, void *private),
345 struct ceph_options *opt;
348 substring_t argstr[MAX_OPT_ARGS];
350 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
352 return ERR_PTR(-ENOMEM);
353 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
358 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
361 /* start with defaults */
362 opt->flags = CEPH_OPT_DEFAULT;
363 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
364 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
365 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
366 opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
369 /* ip1[:port1][,ip2[:port2]...] */
370 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
371 CEPH_MAX_MON, &opt->num_mon);
375 /* parse mount options */
376 while ((c = strsep(&options, ",")) != NULL) {
377 int token, intval, ret;
381 token = match_token((char *)c, opt_tokens, argstr);
382 if (token < 0 && parse_extra_token) {
384 err = parse_extra_token((char *)c, private);
386 pr_err("bad option at '%s'\n", c);
391 if (token < Opt_last_int) {
392 ret = match_int(&argstr[0], &intval);
394 pr_err("bad mount option arg (not int) "
398 dout("got int token %d val %d\n", token, intval);
399 } else if (token > Opt_last_int && token < Opt_last_string) {
400 dout("got string token %d val %s\n", token,
403 dout("got token %d\n", token);
407 err = ceph_parse_ips(argstr[0].from,
413 opt->flags |= CEPH_OPT_MYIP;
417 err = parse_fsid(argstr[0].from, &opt->fsid);
419 opt->flags |= CEPH_OPT_FSID;
422 opt->name = kstrndup(argstr[0].from,
423 argstr[0].to-argstr[0].from,
427 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
432 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
437 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
442 err = get_secret(opt->key, argstr[0].from);
449 pr_warn("ignoring deprecated osdtimeout option\n");
451 case Opt_osdkeepalivetimeout:
452 /* 0 isn't well defined right now, reject it */
453 if (intval < 1 || intval > INT_MAX / 1000) {
454 pr_err("osdkeepalive out of range\n");
458 opt->osd_keepalive_timeout =
459 msecs_to_jiffies(intval * 1000);
461 case Opt_osd_idle_ttl:
462 /* 0 isn't well defined right now, reject it */
463 if (intval < 1 || intval > INT_MAX / 1000) {
464 pr_err("osd_idle_ttl out of range\n");
468 opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
470 case Opt_mount_timeout:
471 /* 0 is "wait forever" (i.e. infinite timeout) */
472 if (intval < 0 || intval > INT_MAX / 1000) {
473 pr_err("mount_timeout out of range\n");
477 opt->mount_timeout = msecs_to_jiffies(intval * 1000);
479 case Opt_osd_request_timeout:
480 /* 0 is "wait forever" (i.e. infinite timeout) */
481 if (intval < 0 || intval > INT_MAX / 1000) {
482 pr_err("osd_request_timeout out of range\n");
486 opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
490 opt->flags &= ~CEPH_OPT_NOSHARE;
493 opt->flags |= CEPH_OPT_NOSHARE;
497 opt->flags &= ~CEPH_OPT_NOCRC;
500 opt->flags |= CEPH_OPT_NOCRC;
503 case Opt_cephx_require_signatures:
504 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
506 case Opt_nocephx_require_signatures:
507 opt->flags |= CEPH_OPT_NOMSGAUTH;
509 case Opt_cephx_sign_messages:
510 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
512 case Opt_nocephx_sign_messages:
513 opt->flags |= CEPH_OPT_NOMSGSIGN;
516 case Opt_tcp_nodelay:
517 opt->flags |= CEPH_OPT_TCP_NODELAY;
519 case Opt_notcp_nodelay:
520 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
532 ceph_destroy_options(opt);
535 EXPORT_SYMBOL(ceph_parse_options);
537 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client)
539 struct ceph_options *opt = client->options;
540 size_t pos = m->count;
543 seq_puts(m, "name=");
544 seq_escape(m, opt->name, ", \t\n\\");
548 seq_puts(m, "secret=<hidden>,");
550 if (opt->flags & CEPH_OPT_FSID)
551 seq_printf(m, "fsid=%pU,", &opt->fsid);
552 if (opt->flags & CEPH_OPT_NOSHARE)
553 seq_puts(m, "noshare,");
554 if (opt->flags & CEPH_OPT_NOCRC)
555 seq_puts(m, "nocrc,");
556 if (opt->flags & CEPH_OPT_NOMSGAUTH)
557 seq_puts(m, "nocephx_require_signatures,");
558 if (opt->flags & CEPH_OPT_NOMSGSIGN)
559 seq_puts(m, "nocephx_sign_messages,");
560 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
561 seq_puts(m, "notcp_nodelay,");
563 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
564 seq_printf(m, "mount_timeout=%d,",
565 jiffies_to_msecs(opt->mount_timeout) / 1000);
566 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
567 seq_printf(m, "osd_idle_ttl=%d,",
568 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
569 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
570 seq_printf(m, "osdkeepalivetimeout=%d,",
571 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
572 if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
573 seq_printf(m, "osd_request_timeout=%d,",
574 jiffies_to_msecs(opt->osd_request_timeout) / 1000);
576 /* drop redundant comma */
582 EXPORT_SYMBOL(ceph_print_client_options);
584 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
586 return &client->msgr.inst.addr;
588 EXPORT_SYMBOL(ceph_client_addr);
590 u64 ceph_client_gid(struct ceph_client *client)
592 return client->monc.auth->global_id;
594 EXPORT_SYMBOL(ceph_client_gid);
597 * create a fresh client instance
599 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
600 u64 supported_features,
601 u64 required_features)
603 struct ceph_client *client;
604 struct ceph_entity_addr *myaddr = NULL;
607 client = kzalloc(sizeof(*client), GFP_KERNEL);
609 return ERR_PTR(-ENOMEM);
611 client->private = private;
612 client->options = opt;
614 mutex_init(&client->mount_mutex);
615 init_waitqueue_head(&client->auth_wq);
616 client->auth_err = 0;
618 if (!ceph_test_opt(client, NOMSGAUTH))
619 required_features |= CEPH_FEATURE_MSG_AUTH;
621 client->extra_mon_dispatch = NULL;
622 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
624 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
628 if (ceph_test_opt(client, MYIP))
629 myaddr = &client->options->my_addr;
631 ceph_messenger_init(&client->msgr, myaddr);
634 err = ceph_monc_init(&client->monc, client);
637 err = ceph_osdc_init(&client->osdc, client);
644 ceph_monc_stop(&client->monc);
646 ceph_messenger_fini(&client->msgr);
650 EXPORT_SYMBOL(ceph_create_client);
652 void ceph_destroy_client(struct ceph_client *client)
654 dout("destroy_client %p\n", client);
656 atomic_set(&client->msgr.stopping, 1);
659 ceph_osdc_stop(&client->osdc);
660 ceph_monc_stop(&client->monc);
661 ceph_messenger_fini(&client->msgr);
663 ceph_debugfs_client_cleanup(client);
665 ceph_destroy_options(client->options);
668 dout("destroy_client %p done\n", client);
670 EXPORT_SYMBOL(ceph_destroy_client);
673 * true if we have the mon map (and have thus joined the cluster)
675 static bool have_mon_and_osd_map(struct ceph_client *client)
677 return client->monc.monmap && client->monc.monmap->epoch &&
678 client->osdc.osdmap && client->osdc.osdmap->epoch;
682 * mount: join the ceph cluster, and open root directory.
684 int __ceph_open_session(struct ceph_client *client, unsigned long started)
686 unsigned long timeout = client->options->mount_timeout;
689 /* open session, and wait for mon and osd maps */
690 err = ceph_monc_open_session(&client->monc);
694 while (!have_mon_and_osd_map(client)) {
695 if (timeout && time_after_eq(jiffies, started + timeout))
699 dout("mount waiting for mon_map\n");
700 err = wait_event_interruptible_timeout(client->auth_wq,
701 have_mon_and_osd_map(client) || (client->auth_err < 0),
702 ceph_timeout_jiffies(timeout));
705 if (client->auth_err < 0)
706 return client->auth_err;
709 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
711 ceph_debugfs_client_init(client);
715 EXPORT_SYMBOL(__ceph_open_session);
718 int ceph_open_session(struct ceph_client *client)
721 unsigned long started = jiffies; /* note the start time */
723 dout("open_session start\n");
724 mutex_lock(&client->mount_mutex);
726 ret = __ceph_open_session(client, started);
728 mutex_unlock(&client->mount_mutex);
731 EXPORT_SYMBOL(ceph_open_session);
734 static int __init init_ceph_lib(void)
738 ret = ceph_debugfs_init();
742 ret = ceph_crypto_init();
746 ret = ceph_msgr_init();
750 ret = ceph_osdc_setup();
754 pr_info("loaded (mon/osd proto %d/%d)\n",
755 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
762 ceph_crypto_shutdown();
764 ceph_debugfs_cleanup();
769 static void __exit exit_ceph_lib(void)
771 dout("exit_ceph_lib\n");
772 WARN_ON(!ceph_strings_empty());
776 ceph_crypto_shutdown();
777 ceph_debugfs_cleanup();
780 module_init(init_ceph_lib);
781 module_exit(exit_ceph_lib);
783 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
784 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
785 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
786 MODULE_DESCRIPTION("Ceph core library");
787 MODULE_LICENSE("GPL");