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,
261 static match_table_t opt_tokens = {
262 {Opt_osdtimeout, "osdtimeout=%d"},
263 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
264 {Opt_mount_timeout, "mount_timeout=%d"},
265 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
266 {Opt_osd_request_timeout, "osd_request_timeout=%d"},
268 {Opt_fsid, "fsid=%s"},
269 {Opt_name, "name=%s"},
270 {Opt_secret, "secret=%s"},
273 /* string args above */
274 {Opt_share, "share"},
275 {Opt_noshare, "noshare"},
277 {Opt_nocrc, "nocrc"},
278 {Opt_cephx_require_signatures, "cephx_require_signatures"},
279 {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
280 {Opt_cephx_sign_messages, "cephx_sign_messages"},
281 {Opt_nocephx_sign_messages, "nocephx_sign_messages"},
282 {Opt_tcp_nodelay, "tcp_nodelay"},
283 {Opt_notcp_nodelay, "notcp_nodelay"},
284 {Opt_abort_on_full, "abort_on_full"},
288 void ceph_destroy_options(struct ceph_options *opt)
290 dout("destroy_options %p\n", opt);
293 ceph_crypto_key_destroy(opt->key);
296 kfree(opt->mon_addr);
299 EXPORT_SYMBOL(ceph_destroy_options);
301 /* get secret from key store */
302 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
306 struct ceph_crypto_key *ckey;
308 ukey = request_key(&key_type_ceph, name, NULL);
310 /* request_key errors don't map nicely to mount(2)
311 errors; don't even try, but still printk */
312 key_err = PTR_ERR(ukey);
315 pr_warn("ceph: Mount failed due to key not found: %s\n",
319 pr_warn("ceph: Mount failed due to expired key: %s\n",
323 pr_warn("ceph: Mount failed due to revoked key: %s\n",
327 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
334 ckey = ukey->payload.data[0];
335 err = ceph_crypto_key_clone(dst, ckey);
338 /* pass through, err is 0 */
346 struct ceph_options *
347 ceph_parse_options(char *options, const char *dev_name,
348 const char *dev_name_end,
349 int (*parse_extra_token)(char *c, void *private),
352 struct ceph_options *opt;
355 substring_t argstr[MAX_OPT_ARGS];
357 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
359 return ERR_PTR(-ENOMEM);
360 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
365 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
368 /* start with defaults */
369 opt->flags = CEPH_OPT_DEFAULT;
370 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
371 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
372 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
373 opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
376 /* ip1[:port1][,ip2[:port2]...] */
377 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
378 CEPH_MAX_MON, &opt->num_mon);
382 /* parse mount options */
383 while ((c = strsep(&options, ",")) != NULL) {
388 token = match_token((char *)c, opt_tokens, argstr);
389 if (token < 0 && parse_extra_token) {
391 err = parse_extra_token((char *)c, private);
393 pr_err("bad option at '%s'\n", c);
398 if (token < Opt_last_int) {
399 err = match_int(&argstr[0], &intval);
401 pr_err("bad option arg (not int) at '%s'\n", c);
404 dout("got int token %d val %d\n", token, intval);
405 } else if (token > Opt_last_int && token < Opt_last_string) {
406 dout("got string token %d val %s\n", token,
409 dout("got token %d\n", token);
413 err = ceph_parse_ips(argstr[0].from,
419 opt->flags |= CEPH_OPT_MYIP;
423 err = parse_fsid(argstr[0].from, &opt->fsid);
425 opt->flags |= CEPH_OPT_FSID;
429 opt->name = kstrndup(argstr[0].from,
430 argstr[0].to-argstr[0].from,
438 ceph_crypto_key_destroy(opt->key);
441 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
446 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
451 ceph_crypto_key_destroy(opt->key);
454 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
459 err = get_secret(opt->key, argstr[0].from);
466 pr_warn("ignoring deprecated osdtimeout option\n");
468 case Opt_osdkeepalivetimeout:
469 /* 0 isn't well defined right now, reject it */
470 if (intval < 1 || intval > INT_MAX / 1000) {
471 pr_err("osdkeepalive out of range\n");
475 opt->osd_keepalive_timeout =
476 msecs_to_jiffies(intval * 1000);
478 case Opt_osd_idle_ttl:
479 /* 0 isn't well defined right now, reject it */
480 if (intval < 1 || intval > INT_MAX / 1000) {
481 pr_err("osd_idle_ttl out of range\n");
485 opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
487 case Opt_mount_timeout:
488 /* 0 is "wait forever" (i.e. infinite timeout) */
489 if (intval < 0 || intval > INT_MAX / 1000) {
490 pr_err("mount_timeout out of range\n");
494 opt->mount_timeout = msecs_to_jiffies(intval * 1000);
496 case Opt_osd_request_timeout:
497 /* 0 is "wait forever" (i.e. infinite timeout) */
498 if (intval < 0 || intval > INT_MAX / 1000) {
499 pr_err("osd_request_timeout out of range\n");
503 opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
507 opt->flags &= ~CEPH_OPT_NOSHARE;
510 opt->flags |= CEPH_OPT_NOSHARE;
514 opt->flags &= ~CEPH_OPT_NOCRC;
517 opt->flags |= CEPH_OPT_NOCRC;
520 case Opt_cephx_require_signatures:
521 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
523 case Opt_nocephx_require_signatures:
524 opt->flags |= CEPH_OPT_NOMSGAUTH;
526 case Opt_cephx_sign_messages:
527 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
529 case Opt_nocephx_sign_messages:
530 opt->flags |= CEPH_OPT_NOMSGSIGN;
533 case Opt_tcp_nodelay:
534 opt->flags |= CEPH_OPT_TCP_NODELAY;
536 case Opt_notcp_nodelay:
537 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
540 case Opt_abort_on_full:
541 opt->flags |= CEPH_OPT_ABORT_ON_FULL;
553 ceph_destroy_options(opt);
556 EXPORT_SYMBOL(ceph_parse_options);
558 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
561 struct ceph_options *opt = client->options;
562 size_t pos = m->count;
565 seq_puts(m, "name=");
566 seq_escape(m, opt->name, ", \t\n\\");
570 seq_puts(m, "secret=<hidden>,");
572 if (opt->flags & CEPH_OPT_FSID)
573 seq_printf(m, "fsid=%pU,", &opt->fsid);
574 if (opt->flags & CEPH_OPT_NOSHARE)
575 seq_puts(m, "noshare,");
576 if (opt->flags & CEPH_OPT_NOCRC)
577 seq_puts(m, "nocrc,");
578 if (opt->flags & CEPH_OPT_NOMSGAUTH)
579 seq_puts(m, "nocephx_require_signatures,");
580 if (opt->flags & CEPH_OPT_NOMSGSIGN)
581 seq_puts(m, "nocephx_sign_messages,");
582 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
583 seq_puts(m, "notcp_nodelay,");
584 if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
585 seq_puts(m, "abort_on_full,");
587 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
588 seq_printf(m, "mount_timeout=%d,",
589 jiffies_to_msecs(opt->mount_timeout) / 1000);
590 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
591 seq_printf(m, "osd_idle_ttl=%d,",
592 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
593 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
594 seq_printf(m, "osdkeepalivetimeout=%d,",
595 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
596 if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
597 seq_printf(m, "osd_request_timeout=%d,",
598 jiffies_to_msecs(opt->osd_request_timeout) / 1000);
600 /* drop redundant comma */
606 EXPORT_SYMBOL(ceph_print_client_options);
608 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
610 return &client->msgr.inst.addr;
612 EXPORT_SYMBOL(ceph_client_addr);
614 u64 ceph_client_gid(struct ceph_client *client)
616 return client->monc.auth->global_id;
618 EXPORT_SYMBOL(ceph_client_gid);
621 * create a fresh client instance
623 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
625 struct ceph_client *client;
626 struct ceph_entity_addr *myaddr = NULL;
629 err = wait_for_random_bytes();
633 client = kzalloc(sizeof(*client), GFP_KERNEL);
635 return ERR_PTR(-ENOMEM);
637 client->private = private;
638 client->options = opt;
640 mutex_init(&client->mount_mutex);
641 init_waitqueue_head(&client->auth_wq);
642 client->auth_err = 0;
644 client->extra_mon_dispatch = NULL;
645 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
646 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
648 if (!ceph_test_opt(client, NOMSGAUTH))
649 client->required_features |= CEPH_FEATURE_MSG_AUTH;
652 if (ceph_test_opt(client, MYIP))
653 myaddr = &client->options->my_addr;
655 ceph_messenger_init(&client->msgr, myaddr);
658 err = ceph_monc_init(&client->monc, client);
661 err = ceph_osdc_init(&client->osdc, client);
668 ceph_monc_stop(&client->monc);
670 ceph_messenger_fini(&client->msgr);
674 EXPORT_SYMBOL(ceph_create_client);
676 void ceph_destroy_client(struct ceph_client *client)
678 dout("destroy_client %p\n", client);
680 atomic_set(&client->msgr.stopping, 1);
683 ceph_osdc_stop(&client->osdc);
684 ceph_monc_stop(&client->monc);
685 ceph_messenger_fini(&client->msgr);
687 ceph_debugfs_client_cleanup(client);
689 ceph_destroy_options(client->options);
692 dout("destroy_client %p done\n", client);
694 EXPORT_SYMBOL(ceph_destroy_client);
697 * true if we have the mon map (and have thus joined the cluster)
699 static bool have_mon_and_osd_map(struct ceph_client *client)
701 return client->monc.monmap && client->monc.monmap->epoch &&
702 client->osdc.osdmap && client->osdc.osdmap->epoch;
706 * mount: join the ceph cluster, and open root directory.
708 int __ceph_open_session(struct ceph_client *client, unsigned long started)
710 unsigned long timeout = client->options->mount_timeout;
713 /* open session, and wait for mon and osd maps */
714 err = ceph_monc_open_session(&client->monc);
718 while (!have_mon_and_osd_map(client)) {
719 if (timeout && time_after_eq(jiffies, started + timeout))
723 dout("mount waiting for mon_map\n");
724 err = wait_event_interruptible_timeout(client->auth_wq,
725 have_mon_and_osd_map(client) || (client->auth_err < 0),
726 ceph_timeout_jiffies(timeout));
729 if (client->auth_err < 0)
730 return client->auth_err;
733 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
735 ceph_debugfs_client_init(client);
739 EXPORT_SYMBOL(__ceph_open_session);
741 int ceph_open_session(struct ceph_client *client)
744 unsigned long started = jiffies; /* note the start time */
746 dout("open_session start\n");
747 mutex_lock(&client->mount_mutex);
749 ret = __ceph_open_session(client, started);
751 mutex_unlock(&client->mount_mutex);
754 EXPORT_SYMBOL(ceph_open_session);
756 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
757 unsigned long timeout)
762 ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
766 if (client->osdc.osdmap->epoch >= newest_epoch)
769 ceph_osdc_maybe_request_map(&client->osdc);
770 return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
772 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
774 static int __init init_ceph_lib(void)
778 ret = ceph_debugfs_init();
782 ret = ceph_crypto_init();
786 ret = ceph_msgr_init();
790 ret = ceph_osdc_setup();
794 pr_info("loaded (mon/osd proto %d/%d)\n",
795 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
802 ceph_crypto_shutdown();
804 ceph_debugfs_cleanup();
809 static void __exit exit_ceph_lib(void)
811 dout("exit_ceph_lib\n");
812 WARN_ON(!ceph_strings_empty());
816 ceph_crypto_shutdown();
817 ceph_debugfs_cleanup();
820 module_init(init_ceph_lib);
821 module_exit(exit_ceph_lib);
823 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
824 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
825 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
826 MODULE_DESCRIPTION("Ceph core library");
827 MODULE_LICENSE("GPL");