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/parser.h>
13 #include <linux/sched.h>
14 #include <linux/seq_file.h>
15 #include <linux/slab.h>
16 #include <linux/statfs.h>
17 #include <linux/string.h>
18 #include <linux/vmalloc.h>
19 #include <linux/nsproxy.h>
20 #include <net/net_namespace.h>
23 #include <linux/ceph/ceph_features.h>
24 #include <linux/ceph/libceph.h>
25 #include <linux/ceph/debugfs.h>
26 #include <linux/ceph/decode.h>
27 #include <linux/ceph/mon_client.h>
28 #include <linux/ceph/auth.h>
33 * Module compatibility interface. For now it doesn't do anything,
34 * but its existence signals a certain level of functionality.
36 * The data buffer is used to pass information both to and from
37 * libceph. The return value indicates whether libceph determines
38 * it is compatible with the caller (from another kernel module),
39 * given the provided data.
41 * The data pointer can be null.
43 bool libceph_compatible(void *data)
47 EXPORT_SYMBOL(libceph_compatible);
50 * find filename portion of a path (/foo/bar/baz -> baz)
52 const char *ceph_file_part(const char *s, int len)
54 const char *e = s + len;
56 while (e != s && *(e-1) != '/')
60 EXPORT_SYMBOL(ceph_file_part);
62 const char *ceph_msg_type_name(int type)
65 case CEPH_MSG_SHUTDOWN: return "shutdown";
66 case CEPH_MSG_PING: return "ping";
67 case CEPH_MSG_AUTH: return "auth";
68 case CEPH_MSG_AUTH_REPLY: return "auth_reply";
69 case CEPH_MSG_MON_MAP: return "mon_map";
70 case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
71 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
72 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
73 case CEPH_MSG_STATFS: return "statfs";
74 case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
75 case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
76 case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
77 case CEPH_MSG_MDS_MAP: return "mds_map";
78 case CEPH_MSG_CLIENT_SESSION: return "client_session";
79 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
80 case CEPH_MSG_CLIENT_REQUEST: return "client_request";
81 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
82 case CEPH_MSG_CLIENT_REPLY: return "client_reply";
83 case CEPH_MSG_CLIENT_CAPS: return "client_caps";
84 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
85 case CEPH_MSG_CLIENT_SNAP: return "client_snap";
86 case CEPH_MSG_CLIENT_LEASE: return "client_lease";
87 case CEPH_MSG_OSD_MAP: return "osd_map";
88 case CEPH_MSG_OSD_OP: return "osd_op";
89 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
90 case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
91 default: return "unknown";
94 EXPORT_SYMBOL(ceph_msg_type_name);
97 * Initially learn our fsid, or verify an fsid matches.
99 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
101 if (client->have_fsid) {
102 if (ceph_fsid_compare(&client->fsid, fsid)) {
103 pr_err("bad fsid, had %pU got %pU",
104 &client->fsid, fsid);
108 memcpy(&client->fsid, fsid, sizeof(*fsid));
112 EXPORT_SYMBOL(ceph_check_fsid);
114 static int strcmp_null(const char *s1, const char *s2)
122 return strcmp(s1, s2);
125 int ceph_compare_options(struct ceph_options *new_opt,
126 struct ceph_client *client)
128 struct ceph_options *opt1 = new_opt;
129 struct ceph_options *opt2 = client->options;
130 int ofs = offsetof(struct ceph_options, mon_addr);
134 ret = memcmp(opt1, opt2, ofs);
138 ret = strcmp_null(opt1->name, opt2->name);
142 if (opt1->key && !opt2->key)
144 if (!opt1->key && opt2->key)
146 if (opt1->key && opt2->key) {
147 if (opt1->key->type != opt2->key->type)
149 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
151 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
153 if (opt1->key->len != opt2->key->len)
155 if (opt1->key->key && !opt2->key->key)
157 if (!opt1->key->key && opt2->key->key)
159 if (opt1->key->key && opt2->key->key) {
160 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
166 /* any matching mon ip implies a match */
167 for (i = 0; i < opt1->num_mon; i++) {
168 if (ceph_monmap_contains(client->monc.monmap,
174 EXPORT_SYMBOL(ceph_compare_options);
176 void *ceph_kvmalloc(size_t size, gfp_t flags)
178 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
179 void *ptr = kmalloc(size, flags | __GFP_NOWARN);
184 return __vmalloc(size, flags | __GFP_HIGHMEM, PAGE_KERNEL);
188 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
195 dout("parse_fsid '%s'\n", str);
197 while (*str && i < 16) {
202 if (!isxdigit(str[0]) || !isxdigit(str[1]))
206 if (sscanf(tmp, "%x", &d) < 1)
208 fsid->fsid[i] = d & 0xff;
215 dout("parse_fsid ret %d got fsid %pU", err, fsid);
224 Opt_osdkeepalivetimeout,
235 /* string args above */
240 Opt_cephx_require_signatures,
241 Opt_nocephx_require_signatures,
246 static match_table_t opt_tokens = {
247 {Opt_osdtimeout, "osdtimeout=%d"},
248 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
249 {Opt_mount_timeout, "mount_timeout=%d"},
250 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
252 {Opt_fsid, "fsid=%s"},
253 {Opt_name, "name=%s"},
254 {Opt_secret, "secret=%s"},
257 /* string args above */
258 {Opt_share, "share"},
259 {Opt_noshare, "noshare"},
261 {Opt_nocrc, "nocrc"},
262 {Opt_cephx_require_signatures, "cephx_require_signatures"},
263 {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
264 {Opt_tcp_nodelay, "tcp_nodelay"},
265 {Opt_notcp_nodelay, "notcp_nodelay"},
269 void ceph_destroy_options(struct ceph_options *opt)
271 dout("destroy_options %p\n", opt);
274 ceph_crypto_key_destroy(opt->key);
277 kfree(opt->mon_addr);
280 EXPORT_SYMBOL(ceph_destroy_options);
282 /* get secret from key store */
283 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
287 struct ceph_crypto_key *ckey;
289 ukey = request_key(&key_type_ceph, name, NULL);
290 if (!ukey || IS_ERR(ukey)) {
291 /* request_key errors don't map nicely to mount(2)
292 errors; don't even try, but still printk */
293 key_err = PTR_ERR(ukey);
296 pr_warn("ceph: Mount failed due to key not found: %s\n",
300 pr_warn("ceph: Mount failed due to expired key: %s\n",
304 pr_warn("ceph: Mount failed due to revoked key: %s\n",
308 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
315 ckey = ukey->payload.data;
316 err = ceph_crypto_key_clone(dst, ckey);
319 /* pass through, err is 0 */
327 struct ceph_options *
328 ceph_parse_options(char *options, const char *dev_name,
329 const char *dev_name_end,
330 int (*parse_extra_token)(char *c, void *private),
333 struct ceph_options *opt;
336 substring_t argstr[MAX_OPT_ARGS];
338 if (current->nsproxy->net_ns != &init_net)
339 return ERR_PTR(-EINVAL);
341 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
343 return ERR_PTR(-ENOMEM);
344 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
349 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
352 /* start with defaults */
353 opt->flags = CEPH_OPT_DEFAULT;
354 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
355 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
356 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* seconds */
359 /* ip1[:port1][,ip2[:port2]...] */
360 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
361 CEPH_MAX_MON, &opt->num_mon);
365 /* parse mount options */
366 while ((c = strsep(&options, ",")) != NULL) {
367 int token, intval, ret;
371 token = match_token((char *)c, opt_tokens, argstr);
372 if (token < 0 && parse_extra_token) {
374 err = parse_extra_token((char *)c, private);
376 pr_err("bad option at '%s'\n", c);
381 if (token < Opt_last_int) {
382 ret = match_int(&argstr[0], &intval);
384 pr_err("bad mount option arg (not int) "
388 dout("got int token %d val %d\n", token, intval);
389 } else if (token > Opt_last_int && token < Opt_last_string) {
390 dout("got string token %d val %s\n", token,
393 dout("got token %d\n", token);
397 err = ceph_parse_ips(argstr[0].from,
403 opt->flags |= CEPH_OPT_MYIP;
407 err = parse_fsid(argstr[0].from, &opt->fsid);
409 opt->flags |= CEPH_OPT_FSID;
412 opt->name = kstrndup(argstr[0].from,
413 argstr[0].to-argstr[0].from,
417 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
422 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
427 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
432 err = get_secret(opt->key, argstr[0].from);
439 pr_warn("ignoring deprecated osdtimeout option\n");
441 case Opt_osdkeepalivetimeout:
442 opt->osd_keepalive_timeout = intval;
444 case Opt_osd_idle_ttl:
445 opt->osd_idle_ttl = intval;
447 case Opt_mount_timeout:
448 opt->mount_timeout = intval;
452 opt->flags &= ~CEPH_OPT_NOSHARE;
455 opt->flags |= CEPH_OPT_NOSHARE;
459 opt->flags &= ~CEPH_OPT_NOCRC;
462 opt->flags |= CEPH_OPT_NOCRC;
465 case Opt_cephx_require_signatures:
466 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
468 case Opt_nocephx_require_signatures:
469 opt->flags |= CEPH_OPT_NOMSGAUTH;
472 case Opt_tcp_nodelay:
473 opt->flags |= CEPH_OPT_TCP_NODELAY;
475 case Opt_notcp_nodelay:
476 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
488 ceph_destroy_options(opt);
491 EXPORT_SYMBOL(ceph_parse_options);
493 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client)
495 struct ceph_options *opt = client->options;
496 size_t pos = m->count;
499 seq_printf(m, "name=%s,", opt->name);
501 seq_puts(m, "secret=<hidden>,");
503 if (opt->flags & CEPH_OPT_FSID)
504 seq_printf(m, "fsid=%pU,", &opt->fsid);
505 if (opt->flags & CEPH_OPT_NOSHARE)
506 seq_puts(m, "noshare,");
507 if (opt->flags & CEPH_OPT_NOCRC)
508 seq_puts(m, "nocrc,");
509 if (opt->flags & CEPH_OPT_NOMSGAUTH)
510 seq_puts(m, "nocephx_require_signatures,");
511 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
512 seq_puts(m, "notcp_nodelay,");
514 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
515 seq_printf(m, "mount_timeout=%d,", opt->mount_timeout);
516 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
517 seq_printf(m, "osd_idle_ttl=%d,", opt->osd_idle_ttl);
518 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
519 seq_printf(m, "osdkeepalivetimeout=%d,",
520 opt->osd_keepalive_timeout);
522 /* drop redundant comma */
528 EXPORT_SYMBOL(ceph_print_client_options);
530 u64 ceph_client_id(struct ceph_client *client)
532 return client->monc.auth->global_id;
534 EXPORT_SYMBOL(ceph_client_id);
537 * create a fresh client instance
539 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
540 u64 supported_features,
541 u64 required_features)
543 struct ceph_client *client;
544 struct ceph_entity_addr *myaddr = NULL;
547 client = kzalloc(sizeof(*client), GFP_KERNEL);
549 return ERR_PTR(-ENOMEM);
551 client->private = private;
552 client->options = opt;
554 mutex_init(&client->mount_mutex);
555 init_waitqueue_head(&client->auth_wq);
556 client->auth_err = 0;
558 if (!ceph_test_opt(client, NOMSGAUTH))
559 required_features |= CEPH_FEATURE_MSG_AUTH;
561 client->extra_mon_dispatch = NULL;
562 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
564 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
568 if (ceph_test_opt(client, MYIP))
569 myaddr = &client->options->my_addr;
571 ceph_messenger_init(&client->msgr, myaddr,
572 client->supported_features,
573 client->required_features,
574 ceph_test_opt(client, NOCRC),
575 ceph_test_opt(client, TCP_NODELAY));
578 err = ceph_monc_init(&client->monc, client);
581 err = ceph_osdc_init(&client->osdc, client);
588 ceph_monc_stop(&client->monc);
593 EXPORT_SYMBOL(ceph_create_client);
595 void ceph_destroy_client(struct ceph_client *client)
597 dout("destroy_client %p\n", client);
599 atomic_set(&client->msgr.stopping, 1);
602 ceph_osdc_stop(&client->osdc);
604 ceph_monc_stop(&client->monc);
606 ceph_debugfs_client_cleanup(client);
608 ceph_destroy_options(client->options);
611 dout("destroy_client %p done\n", client);
613 EXPORT_SYMBOL(ceph_destroy_client);
616 * true if we have the mon map (and have thus joined the cluster)
618 static int have_mon_and_osd_map(struct ceph_client *client)
620 return client->monc.monmap && client->monc.monmap->epoch &&
621 client->osdc.osdmap && client->osdc.osdmap->epoch;
625 * mount: join the ceph cluster, and open root directory.
627 int __ceph_open_session(struct ceph_client *client, unsigned long started)
630 unsigned long timeout = client->options->mount_timeout * HZ;
632 /* open session, and wait for mon and osd maps */
633 err = ceph_monc_open_session(&client->monc);
637 while (!have_mon_and_osd_map(client)) {
639 if (timeout && time_after_eq(jiffies, started + timeout))
643 dout("mount waiting for mon_map\n");
644 err = wait_event_interruptible_timeout(client->auth_wq,
645 have_mon_and_osd_map(client) || (client->auth_err < 0),
647 if (err == -EINTR || err == -ERESTARTSYS)
649 if (client->auth_err < 0)
650 return client->auth_err;
655 EXPORT_SYMBOL(__ceph_open_session);
658 int ceph_open_session(struct ceph_client *client)
661 unsigned long started = jiffies; /* note the start time */
663 dout("open_session start\n");
664 mutex_lock(&client->mount_mutex);
666 ret = __ceph_open_session(client, started);
668 mutex_unlock(&client->mount_mutex);
671 EXPORT_SYMBOL(ceph_open_session);
674 static int __init init_ceph_lib(void)
678 ret = ceph_debugfs_init();
682 ret = ceph_crypto_init();
686 ret = ceph_msgr_init();
690 ret = ceph_osdc_setup();
694 pr_info("loaded (mon/osd proto %d/%d)\n",
695 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
702 ceph_crypto_shutdown();
704 ceph_debugfs_cleanup();
709 static void __exit exit_ceph_lib(void)
711 dout("exit_ceph_lib\n");
714 ceph_crypto_shutdown();
715 ceph_debugfs_cleanup();
718 module_init(init_ceph_lib);
719 module_exit(exit_ceph_lib);
721 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
722 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
723 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
724 MODULE_DESCRIPTION("Ceph filesystem for Linux");
725 MODULE_LICENSE("GPL");