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,
60 * find filename portion of a path (/foo/bar/baz -> baz)
62 const char *ceph_file_part(const char *s, int len)
64 const char *e = s + len;
66 while (e != s && *(e-1) != '/')
70 EXPORT_SYMBOL(ceph_file_part);
72 const char *ceph_msg_type_name(int type)
75 case CEPH_MSG_SHUTDOWN: return "shutdown";
76 case CEPH_MSG_PING: return "ping";
77 case CEPH_MSG_AUTH: return "auth";
78 case CEPH_MSG_AUTH_REPLY: return "auth_reply";
79 case CEPH_MSG_MON_MAP: return "mon_map";
80 case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
81 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
82 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
83 case CEPH_MSG_STATFS: return "statfs";
84 case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
85 case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
86 case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
87 case CEPH_MSG_MDS_MAP: return "mds_map";
88 case CEPH_MSG_CLIENT_SESSION: return "client_session";
89 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
90 case CEPH_MSG_CLIENT_REQUEST: return "client_request";
91 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
92 case CEPH_MSG_CLIENT_REPLY: return "client_reply";
93 case CEPH_MSG_CLIENT_CAPS: return "client_caps";
94 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
95 case CEPH_MSG_CLIENT_SNAP: return "client_snap";
96 case CEPH_MSG_CLIENT_LEASE: return "client_lease";
97 case CEPH_MSG_OSD_MAP: return "osd_map";
98 case CEPH_MSG_OSD_OP: return "osd_op";
99 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
100 case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
101 default: return "unknown";
104 EXPORT_SYMBOL(ceph_msg_type_name);
107 * Initially learn our fsid, or verify an fsid matches.
109 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
111 if (client->have_fsid) {
112 if (ceph_fsid_compare(&client->fsid, fsid)) {
113 pr_err("bad fsid, had %pU got %pU",
114 &client->fsid, fsid);
118 memcpy(&client->fsid, fsid, sizeof(*fsid));
122 EXPORT_SYMBOL(ceph_check_fsid);
124 static int strcmp_null(const char *s1, const char *s2)
132 return strcmp(s1, s2);
135 int ceph_compare_options(struct ceph_options *new_opt,
136 struct ceph_client *client)
138 struct ceph_options *opt1 = new_opt;
139 struct ceph_options *opt2 = client->options;
140 int ofs = offsetof(struct ceph_options, mon_addr);
145 * Don't bother comparing options if network namespaces don't
148 if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
151 ret = memcmp(opt1, opt2, ofs);
155 ret = strcmp_null(opt1->name, opt2->name);
159 if (opt1->key && !opt2->key)
161 if (!opt1->key && opt2->key)
163 if (opt1->key && opt2->key) {
164 if (opt1->key->type != opt2->key->type)
166 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
168 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
170 if (opt1->key->len != opt2->key->len)
172 if (opt1->key->key && !opt2->key->key)
174 if (!opt1->key->key && opt2->key->key)
176 if (opt1->key->key && opt2->key->key) {
177 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
183 /* any matching mon ip implies a match */
184 for (i = 0; i < opt1->num_mon; i++) {
185 if (ceph_monmap_contains(client->monc.monmap,
191 EXPORT_SYMBOL(ceph_compare_options);
193 void *ceph_kvmalloc(size_t size, gfp_t flags)
195 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
196 void *ptr = kmalloc(size, flags | __GFP_NOWARN);
201 return __vmalloc(size, flags, PAGE_KERNEL);
205 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
212 dout("parse_fsid '%s'\n", str);
214 while (*str && i < 16) {
219 if (!isxdigit(str[0]) || !isxdigit(str[1]))
223 if (sscanf(tmp, "%x", &d) < 1)
225 fsid->fsid[i] = d & 0xff;
232 dout("parse_fsid ret %d got fsid %pU", err, fsid);
241 Opt_osdkeepalivetimeout,
244 Opt_osd_request_timeout,
253 /* string args above */
258 Opt_cephx_require_signatures,
259 Opt_nocephx_require_signatures,
260 Opt_cephx_sign_messages,
261 Opt_nocephx_sign_messages,
266 static match_table_t opt_tokens = {
267 {Opt_osdtimeout, "osdtimeout=%d"},
268 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
269 {Opt_mount_timeout, "mount_timeout=%d"},
270 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
271 {Opt_osd_request_timeout, "osd_request_timeout=%d"},
273 {Opt_fsid, "fsid=%s"},
274 {Opt_name, "name=%s"},
275 {Opt_secret, "secret=%s"},
278 /* string args above */
279 {Opt_share, "share"},
280 {Opt_noshare, "noshare"},
282 {Opt_nocrc, "nocrc"},
283 {Opt_cephx_require_signatures, "cephx_require_signatures"},
284 {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
285 {Opt_cephx_sign_messages, "cephx_sign_messages"},
286 {Opt_nocephx_sign_messages, "nocephx_sign_messages"},
287 {Opt_tcp_nodelay, "tcp_nodelay"},
288 {Opt_notcp_nodelay, "notcp_nodelay"},
292 void ceph_destroy_options(struct ceph_options *opt)
294 dout("destroy_options %p\n", opt);
297 ceph_crypto_key_destroy(opt->key);
300 kfree(opt->mon_addr);
303 EXPORT_SYMBOL(ceph_destroy_options);
305 /* get secret from key store */
306 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
310 struct ceph_crypto_key *ckey;
312 ukey = request_key(&key_type_ceph, name, NULL);
313 if (!ukey || IS_ERR(ukey)) {
314 /* request_key errors don't map nicely to mount(2)
315 errors; don't even try, but still printk */
316 key_err = PTR_ERR(ukey);
319 pr_warn("ceph: Mount failed due to key not found: %s\n",
323 pr_warn("ceph: Mount failed due to expired key: %s\n",
327 pr_warn("ceph: Mount failed due to revoked key: %s\n",
331 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
338 ckey = ukey->payload.data[0];
339 err = ceph_crypto_key_clone(dst, ckey);
342 /* pass through, err is 0 */
350 struct ceph_options *
351 ceph_parse_options(char *options, const char *dev_name,
352 const char *dev_name_end,
353 int (*parse_extra_token)(char *c, void *private),
356 struct ceph_options *opt;
359 substring_t argstr[MAX_OPT_ARGS];
361 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
363 return ERR_PTR(-ENOMEM);
364 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
369 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
372 /* start with defaults */
373 opt->flags = CEPH_OPT_DEFAULT;
374 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
375 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
376 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
377 opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
380 /* ip1[:port1][,ip2[:port2]...] */
381 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
382 CEPH_MAX_MON, &opt->num_mon);
386 /* parse mount options */
387 while ((c = strsep(&options, ",")) != NULL) {
388 int token, intval, ret;
392 token = match_token((char *)c, opt_tokens, argstr);
393 if (token < 0 && parse_extra_token) {
395 err = parse_extra_token((char *)c, private);
397 pr_err("bad option at '%s'\n", c);
402 if (token < Opt_last_int) {
403 ret = match_int(&argstr[0], &intval);
405 pr_err("bad mount option arg (not int) "
409 dout("got int token %d val %d\n", token, intval);
410 } else if (token > Opt_last_int && token < Opt_last_string) {
411 dout("got string token %d val %s\n", token,
414 dout("got token %d\n", token);
418 err = ceph_parse_ips(argstr[0].from,
424 opt->flags |= CEPH_OPT_MYIP;
428 err = parse_fsid(argstr[0].from, &opt->fsid);
430 opt->flags |= CEPH_OPT_FSID;
433 opt->name = kstrndup(argstr[0].from,
434 argstr[0].to-argstr[0].from,
438 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
443 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
448 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
453 err = get_secret(opt->key, argstr[0].from);
460 pr_warn("ignoring deprecated osdtimeout option\n");
462 case Opt_osdkeepalivetimeout:
463 /* 0 isn't well defined right now, reject it */
464 if (intval < 1 || intval > INT_MAX / 1000) {
465 pr_err("osdkeepalive out of range\n");
469 opt->osd_keepalive_timeout =
470 msecs_to_jiffies(intval * 1000);
472 case Opt_osd_idle_ttl:
473 /* 0 isn't well defined right now, reject it */
474 if (intval < 1 || intval > INT_MAX / 1000) {
475 pr_err("osd_idle_ttl out of range\n");
479 opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
481 case Opt_mount_timeout:
482 /* 0 is "wait forever" (i.e. infinite timeout) */
483 if (intval < 0 || intval > INT_MAX / 1000) {
484 pr_err("mount_timeout out of range\n");
488 opt->mount_timeout = msecs_to_jiffies(intval * 1000);
490 case Opt_osd_request_timeout:
491 /* 0 is "wait forever" (i.e. infinite timeout) */
492 if (intval < 0 || intval > INT_MAX / 1000) {
493 pr_err("osd_request_timeout out of range\n");
497 opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
501 opt->flags &= ~CEPH_OPT_NOSHARE;
504 opt->flags |= CEPH_OPT_NOSHARE;
508 opt->flags &= ~CEPH_OPT_NOCRC;
511 opt->flags |= CEPH_OPT_NOCRC;
514 case Opt_cephx_require_signatures:
515 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
517 case Opt_nocephx_require_signatures:
518 opt->flags |= CEPH_OPT_NOMSGAUTH;
520 case Opt_cephx_sign_messages:
521 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
523 case Opt_nocephx_sign_messages:
524 opt->flags |= CEPH_OPT_NOMSGSIGN;
527 case Opt_tcp_nodelay:
528 opt->flags |= CEPH_OPT_TCP_NODELAY;
530 case Opt_notcp_nodelay:
531 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
543 ceph_destroy_options(opt);
546 EXPORT_SYMBOL(ceph_parse_options);
548 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client)
550 struct ceph_options *opt = client->options;
551 size_t pos = m->count;
554 seq_puts(m, "name=");
555 seq_escape(m, opt->name, ", \t\n\\");
559 seq_puts(m, "secret=<hidden>,");
561 if (opt->flags & CEPH_OPT_FSID)
562 seq_printf(m, "fsid=%pU,", &opt->fsid);
563 if (opt->flags & CEPH_OPT_NOSHARE)
564 seq_puts(m, "noshare,");
565 if (opt->flags & CEPH_OPT_NOCRC)
566 seq_puts(m, "nocrc,");
567 if (opt->flags & CEPH_OPT_NOMSGAUTH)
568 seq_puts(m, "nocephx_require_signatures,");
569 if (opt->flags & CEPH_OPT_NOMSGSIGN)
570 seq_puts(m, "nocephx_sign_messages,");
571 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
572 seq_puts(m, "notcp_nodelay,");
574 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
575 seq_printf(m, "mount_timeout=%d,",
576 jiffies_to_msecs(opt->mount_timeout) / 1000);
577 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
578 seq_printf(m, "osd_idle_ttl=%d,",
579 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
580 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
581 seq_printf(m, "osdkeepalivetimeout=%d,",
582 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
583 if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
584 seq_printf(m, "osd_request_timeout=%d,",
585 jiffies_to_msecs(opt->osd_request_timeout) / 1000);
587 /* drop redundant comma */
593 EXPORT_SYMBOL(ceph_print_client_options);
595 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
597 return &client->msgr.inst.addr;
599 EXPORT_SYMBOL(ceph_client_addr);
601 u64 ceph_client_gid(struct ceph_client *client)
603 return client->monc.auth->global_id;
605 EXPORT_SYMBOL(ceph_client_gid);
608 * create a fresh client instance
610 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
612 struct ceph_client *client;
613 struct ceph_entity_addr *myaddr = NULL;
616 client = kzalloc(sizeof(*client), GFP_KERNEL);
618 return ERR_PTR(-ENOMEM);
620 client->private = private;
621 client->options = opt;
623 mutex_init(&client->mount_mutex);
624 init_waitqueue_head(&client->auth_wq);
625 client->auth_err = 0;
627 client->extra_mon_dispatch = NULL;
628 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
629 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
631 if (!ceph_test_opt(client, NOMSGAUTH))
632 client->required_features |= CEPH_FEATURE_MSG_AUTH;
635 if (ceph_test_opt(client, MYIP))
636 myaddr = &client->options->my_addr;
638 ceph_messenger_init(&client->msgr, myaddr);
641 err = ceph_monc_init(&client->monc, client);
644 err = ceph_osdc_init(&client->osdc, client);
651 ceph_monc_stop(&client->monc);
653 ceph_messenger_fini(&client->msgr);
657 EXPORT_SYMBOL(ceph_create_client);
659 void ceph_destroy_client(struct ceph_client *client)
661 dout("destroy_client %p\n", client);
663 atomic_set(&client->msgr.stopping, 1);
666 ceph_osdc_stop(&client->osdc);
667 ceph_monc_stop(&client->monc);
668 ceph_messenger_fini(&client->msgr);
670 ceph_debugfs_client_cleanup(client);
672 ceph_destroy_options(client->options);
675 dout("destroy_client %p done\n", client);
677 EXPORT_SYMBOL(ceph_destroy_client);
680 * true if we have the mon map (and have thus joined the cluster)
682 static bool have_mon_and_osd_map(struct ceph_client *client)
684 return client->monc.monmap && client->monc.monmap->epoch &&
685 client->osdc.osdmap && client->osdc.osdmap->epoch;
689 * mount: join the ceph cluster, and open root directory.
691 int __ceph_open_session(struct ceph_client *client, unsigned long started)
693 unsigned long timeout = client->options->mount_timeout;
696 /* open session, and wait for mon and osd maps */
697 err = ceph_monc_open_session(&client->monc);
701 while (!have_mon_and_osd_map(client)) {
702 if (timeout && time_after_eq(jiffies, started + timeout))
706 dout("mount waiting for mon_map\n");
707 err = wait_event_interruptible_timeout(client->auth_wq,
708 have_mon_and_osd_map(client) || (client->auth_err < 0),
709 ceph_timeout_jiffies(timeout));
712 if (client->auth_err < 0)
713 return client->auth_err;
716 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
718 ceph_debugfs_client_init(client);
722 EXPORT_SYMBOL(__ceph_open_session);
725 int ceph_open_session(struct ceph_client *client)
728 unsigned long started = jiffies; /* note the start time */
730 dout("open_session start\n");
731 mutex_lock(&client->mount_mutex);
733 ret = __ceph_open_session(client, started);
735 mutex_unlock(&client->mount_mutex);
738 EXPORT_SYMBOL(ceph_open_session);
741 static int __init init_ceph_lib(void)
745 ret = ceph_debugfs_init();
749 ret = ceph_crypto_init();
753 ret = ceph_msgr_init();
757 ret = ceph_osdc_setup();
761 pr_info("loaded (mon/osd proto %d/%d)\n",
762 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
769 ceph_crypto_shutdown();
771 ceph_debugfs_cleanup();
776 static void __exit exit_ceph_lib(void)
778 dout("exit_ceph_lib\n");
779 WARN_ON(!ceph_strings_empty());
783 ceph_crypto_shutdown();
784 ceph_debugfs_cleanup();
787 module_init(init_ceph_lib);
788 module_exit(exit_ceph_lib);
790 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
791 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
792 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
793 MODULE_DESCRIPTION("Ceph core library");
794 MODULE_LICENSE("GPL");