xref: /openbmc/linux/net/ceph/ceph_common.c (revision 4cff79e9)
1 
2 #include <linux/ceph/ceph_debug.h>
3 #include <linux/backing-dev.h>
4 #include <linux/ctype.h>
5 #include <linux/fs.h>
6 #include <linux/inet.h>
7 #include <linux/in6.h>
8 #include <linux/key.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>
20 
21 
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>
28 #include "crypto.h"
29 
30 
31 /*
32  * Module compatibility interface.  For now it doesn't do anything,
33  * but its existence signals a certain level of functionality.
34  *
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.
39  *
40  * The data pointer can be null.
41  */
42 bool libceph_compatible(void *data)
43 {
44 	return true;
45 }
46 EXPORT_SYMBOL(libceph_compatible);
47 
48 static int param_get_supported_features(char *buffer,
49 					const struct kernel_param *kp)
50 {
51 	return sprintf(buffer, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT);
52 }
53 static const struct kernel_param_ops param_ops_supported_features = {
54 	.get = param_get_supported_features,
55 };
56 module_param_cb(supported_features, &param_ops_supported_features, NULL,
57 		0444);
58 
59 const char *ceph_msg_type_name(int type)
60 {
61 	switch (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";
96 	}
97 }
98 EXPORT_SYMBOL(ceph_msg_type_name);
99 
100 /*
101  * Initially learn our fsid, or verify an fsid matches.
102  */
103 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
104 {
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);
109 			return -1;
110 		}
111 	} else {
112 		memcpy(&client->fsid, fsid, sizeof(*fsid));
113 	}
114 	return 0;
115 }
116 EXPORT_SYMBOL(ceph_check_fsid);
117 
118 static int strcmp_null(const char *s1, const char *s2)
119 {
120 	if (!s1 && !s2)
121 		return 0;
122 	if (s1 && !s2)
123 		return -1;
124 	if (!s1 && s2)
125 		return 1;
126 	return strcmp(s1, s2);
127 }
128 
129 int ceph_compare_options(struct ceph_options *new_opt,
130 			 struct ceph_client *client)
131 {
132 	struct ceph_options *opt1 = new_opt;
133 	struct ceph_options *opt2 = client->options;
134 	int ofs = offsetof(struct ceph_options, mon_addr);
135 	int i;
136 	int ret;
137 
138 	/*
139 	 * Don't bother comparing options if network namespaces don't
140 	 * match.
141 	 */
142 	if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
143 		return -1;
144 
145 	ret = memcmp(opt1, opt2, ofs);
146 	if (ret)
147 		return ret;
148 
149 	ret = strcmp_null(opt1->name, opt2->name);
150 	if (ret)
151 		return ret;
152 
153 	if (opt1->key && !opt2->key)
154 		return -1;
155 	if (!opt1->key && opt2->key)
156 		return 1;
157 	if (opt1->key && opt2->key) {
158 		if (opt1->key->type != opt2->key->type)
159 			return -1;
160 		if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
161 			return -1;
162 		if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
163 			return -1;
164 		if (opt1->key->len != opt2->key->len)
165 			return -1;
166 		if (opt1->key->key && !opt2->key->key)
167 			return -1;
168 		if (!opt1->key->key && opt2->key->key)
169 			return 1;
170 		if (opt1->key->key && opt2->key->key) {
171 			ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
172 			if (ret)
173 				return ret;
174 		}
175 	}
176 
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,
180 				 &opt1->mon_addr[i]))
181 			return 0;
182 	}
183 	return -1;
184 }
185 EXPORT_SYMBOL(ceph_compare_options);
186 
187 void *ceph_kvmalloc(size_t size, gfp_t flags)
188 {
189 	if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
190 		void *ptr = kmalloc(size, flags | __GFP_NOWARN);
191 		if (ptr)
192 			return ptr;
193 	}
194 
195 	return __vmalloc(size, flags, PAGE_KERNEL);
196 }
197 
198 
199 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
200 {
201 	int i = 0;
202 	char tmp[3];
203 	int err = -EINVAL;
204 	int d;
205 
206 	dout("parse_fsid '%s'\n", str);
207 	tmp[2] = 0;
208 	while (*str && i < 16) {
209 		if (ispunct(*str)) {
210 			str++;
211 			continue;
212 		}
213 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
214 			break;
215 		tmp[0] = str[0];
216 		tmp[1] = str[1];
217 		if (sscanf(tmp, "%x", &d) < 1)
218 			break;
219 		fsid->fsid[i] = d & 0xff;
220 		i++;
221 		str += 2;
222 	}
223 
224 	if (i == 16)
225 		err = 0;
226 	dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
227 	return err;
228 }
229 
230 /*
231  * ceph options
232  */
233 enum {
234 	Opt_osdtimeout,
235 	Opt_osdkeepalivetimeout,
236 	Opt_mount_timeout,
237 	Opt_osd_idle_ttl,
238 	Opt_osd_request_timeout,
239 	Opt_last_int,
240 	/* int args above */
241 	Opt_fsid,
242 	Opt_name,
243 	Opt_secret,
244 	Opt_key,
245 	Opt_ip,
246 	Opt_last_string,
247 	/* string args above */
248 	Opt_share,
249 	Opt_noshare,
250 	Opt_crc,
251 	Opt_nocrc,
252 	Opt_cephx_require_signatures,
253 	Opt_nocephx_require_signatures,
254 	Opt_cephx_sign_messages,
255 	Opt_nocephx_sign_messages,
256 	Opt_tcp_nodelay,
257 	Opt_notcp_nodelay,
258 };
259 
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"},
266 	/* int args above */
267 	{Opt_fsid, "fsid=%s"},
268 	{Opt_name, "name=%s"},
269 	{Opt_secret, "secret=%s"},
270 	{Opt_key, "key=%s"},
271 	{Opt_ip, "ip=%s"},
272 	/* string args above */
273 	{Opt_share, "share"},
274 	{Opt_noshare, "noshare"},
275 	{Opt_crc, "crc"},
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"},
283 	{-1, NULL}
284 };
285 
286 void ceph_destroy_options(struct ceph_options *opt)
287 {
288 	dout("destroy_options %p\n", opt);
289 	kfree(opt->name);
290 	if (opt->key) {
291 		ceph_crypto_key_destroy(opt->key);
292 		kfree(opt->key);
293 	}
294 	kfree(opt->mon_addr);
295 	kfree(opt);
296 }
297 EXPORT_SYMBOL(ceph_destroy_options);
298 
299 /* get secret from key store */
300 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
301 	struct key *ukey;
302 	int key_err;
303 	int err = 0;
304 	struct ceph_crypto_key *ckey;
305 
306 	ukey = request_key(&key_type_ceph, name, NULL);
307 	if (!ukey || IS_ERR(ukey)) {
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);
311 		switch (key_err) {
312 		case -ENOKEY:
313 			pr_warn("ceph: Mount failed due to key not found: %s\n",
314 				name);
315 			break;
316 		case -EKEYEXPIRED:
317 			pr_warn("ceph: Mount failed due to expired key: %s\n",
318 				name);
319 			break;
320 		case -EKEYREVOKED:
321 			pr_warn("ceph: Mount failed due to revoked key: %s\n",
322 				name);
323 			break;
324 		default:
325 			pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
326 				key_err, name);
327 		}
328 		err = -EPERM;
329 		goto out;
330 	}
331 
332 	ckey = ukey->payload.data[0];
333 	err = ceph_crypto_key_clone(dst, ckey);
334 	if (err)
335 		goto out_key;
336 	/* pass through, err is 0 */
337 
338 out_key:
339 	key_put(ukey);
340 out:
341 	return err;
342 }
343 
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),
348 			void *private)
349 {
350 	struct ceph_options *opt;
351 	const char *c;
352 	int err = -ENOMEM;
353 	substring_t argstr[MAX_OPT_ARGS];
354 
355 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
356 	if (!opt)
357 		return ERR_PTR(-ENOMEM);
358 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
359 				GFP_KERNEL);
360 	if (!opt->mon_addr)
361 		goto out;
362 
363 	dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
364 	     dev_name);
365 
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;
372 
373 	/* get mon ip(s) */
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);
377 	if (err < 0)
378 		goto out;
379 
380 	/* parse mount options */
381 	while ((c = strsep(&options, ",")) != NULL) {
382 		int token, intval, ret;
383 		if (!*c)
384 			continue;
385 		err = -EINVAL;
386 		token = match_token((char *)c, opt_tokens, argstr);
387 		if (token < 0 && parse_extra_token) {
388 			/* extra? */
389 			err = parse_extra_token((char *)c, private);
390 			if (err < 0) {
391 				pr_err("bad option at '%s'\n", c);
392 				goto out;
393 			}
394 			continue;
395 		}
396 		if (token < Opt_last_int) {
397 			ret = match_int(&argstr[0], &intval);
398 			if (ret < 0) {
399 				pr_err("bad mount option arg (not int) "
400 				       "at '%s'\n", c);
401 				continue;
402 			}
403 			dout("got int token %d val %d\n", token, intval);
404 		} else if (token > Opt_last_int && token < Opt_last_string) {
405 			dout("got string token %d val %s\n", token,
406 			     argstr[0].from);
407 		} else {
408 			dout("got token %d\n", token);
409 		}
410 		switch (token) {
411 		case Opt_ip:
412 			err = ceph_parse_ips(argstr[0].from,
413 					     argstr[0].to,
414 					     &opt->my_addr,
415 					     1, NULL);
416 			if (err < 0)
417 				goto out;
418 			opt->flags |= CEPH_OPT_MYIP;
419 			break;
420 
421 		case Opt_fsid:
422 			err = parse_fsid(argstr[0].from, &opt->fsid);
423 			if (err == 0)
424 				opt->flags |= CEPH_OPT_FSID;
425 			break;
426 		case Opt_name:
427 			kfree(opt->name);
428 			opt->name = kstrndup(argstr[0].from,
429 					      argstr[0].to-argstr[0].from,
430 					      GFP_KERNEL);
431 			if (!opt->name) {
432 				err = -ENOMEM;
433 				goto out;
434 			}
435 			break;
436 		case Opt_secret:
437 			ceph_crypto_key_destroy(opt->key);
438 			kfree(opt->key);
439 
440 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
441 			if (!opt->key) {
442 				err = -ENOMEM;
443 				goto out;
444 			}
445 			err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
446 			if (err < 0)
447 				goto out;
448 			break;
449 		case Opt_key:
450 			ceph_crypto_key_destroy(opt->key);
451 			kfree(opt->key);
452 
453 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
454 			if (!opt->key) {
455 				err = -ENOMEM;
456 				goto out;
457 			}
458 			err = get_secret(opt->key, argstr[0].from);
459 			if (err < 0)
460 				goto out;
461 			break;
462 
463 			/* misc */
464 		case Opt_osdtimeout:
465 			pr_warn("ignoring deprecated osdtimeout option\n");
466 			break;
467 		case Opt_osdkeepalivetimeout:
468 			/* 0 isn't well defined right now, reject it */
469 			if (intval < 1 || intval > INT_MAX / 1000) {
470 				pr_err("osdkeepalive out of range\n");
471 				err = -EINVAL;
472 				goto out;
473 			}
474 			opt->osd_keepalive_timeout =
475 					msecs_to_jiffies(intval * 1000);
476 			break;
477 		case Opt_osd_idle_ttl:
478 			/* 0 isn't well defined right now, reject it */
479 			if (intval < 1 || intval > INT_MAX / 1000) {
480 				pr_err("osd_idle_ttl out of range\n");
481 				err = -EINVAL;
482 				goto out;
483 			}
484 			opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
485 			break;
486 		case Opt_mount_timeout:
487 			/* 0 is "wait forever" (i.e. infinite timeout) */
488 			if (intval < 0 || intval > INT_MAX / 1000) {
489 				pr_err("mount_timeout out of range\n");
490 				err = -EINVAL;
491 				goto out;
492 			}
493 			opt->mount_timeout = msecs_to_jiffies(intval * 1000);
494 			break;
495 		case Opt_osd_request_timeout:
496 			/* 0 is "wait forever" (i.e. infinite timeout) */
497 			if (intval < 0 || intval > INT_MAX / 1000) {
498 				pr_err("osd_request_timeout out of range\n");
499 				err = -EINVAL;
500 				goto out;
501 			}
502 			opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
503 			break;
504 
505 		case Opt_share:
506 			opt->flags &= ~CEPH_OPT_NOSHARE;
507 			break;
508 		case Opt_noshare:
509 			opt->flags |= CEPH_OPT_NOSHARE;
510 			break;
511 
512 		case Opt_crc:
513 			opt->flags &= ~CEPH_OPT_NOCRC;
514 			break;
515 		case Opt_nocrc:
516 			opt->flags |= CEPH_OPT_NOCRC;
517 			break;
518 
519 		case Opt_cephx_require_signatures:
520 			opt->flags &= ~CEPH_OPT_NOMSGAUTH;
521 			break;
522 		case Opt_nocephx_require_signatures:
523 			opt->flags |= CEPH_OPT_NOMSGAUTH;
524 			break;
525 		case Opt_cephx_sign_messages:
526 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
527 			break;
528 		case Opt_nocephx_sign_messages:
529 			opt->flags |= CEPH_OPT_NOMSGSIGN;
530 			break;
531 
532 		case Opt_tcp_nodelay:
533 			opt->flags |= CEPH_OPT_TCP_NODELAY;
534 			break;
535 		case Opt_notcp_nodelay:
536 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
537 			break;
538 
539 		default:
540 			BUG_ON(token);
541 		}
542 	}
543 
544 	/* success */
545 	return opt;
546 
547 out:
548 	ceph_destroy_options(opt);
549 	return ERR_PTR(err);
550 }
551 EXPORT_SYMBOL(ceph_parse_options);
552 
553 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client)
554 {
555 	struct ceph_options *opt = client->options;
556 	size_t pos = m->count;
557 
558 	if (opt->name) {
559 		seq_puts(m, "name=");
560 		seq_escape(m, opt->name, ", \t\n\\");
561 		seq_putc(m, ',');
562 	}
563 	if (opt->key)
564 		seq_puts(m, "secret=<hidden>,");
565 
566 	if (opt->flags & CEPH_OPT_FSID)
567 		seq_printf(m, "fsid=%pU,", &opt->fsid);
568 	if (opt->flags & CEPH_OPT_NOSHARE)
569 		seq_puts(m, "noshare,");
570 	if (opt->flags & CEPH_OPT_NOCRC)
571 		seq_puts(m, "nocrc,");
572 	if (opt->flags & CEPH_OPT_NOMSGAUTH)
573 		seq_puts(m, "nocephx_require_signatures,");
574 	if (opt->flags & CEPH_OPT_NOMSGSIGN)
575 		seq_puts(m, "nocephx_sign_messages,");
576 	if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
577 		seq_puts(m, "notcp_nodelay,");
578 
579 	if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
580 		seq_printf(m, "mount_timeout=%d,",
581 			   jiffies_to_msecs(opt->mount_timeout) / 1000);
582 	if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
583 		seq_printf(m, "osd_idle_ttl=%d,",
584 			   jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
585 	if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
586 		seq_printf(m, "osdkeepalivetimeout=%d,",
587 		    jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
588 	if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
589 		seq_printf(m, "osd_request_timeout=%d,",
590 			   jiffies_to_msecs(opt->osd_request_timeout) / 1000);
591 
592 	/* drop redundant comma */
593 	if (m->count != pos)
594 		m->count--;
595 
596 	return 0;
597 }
598 EXPORT_SYMBOL(ceph_print_client_options);
599 
600 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
601 {
602 	return &client->msgr.inst.addr;
603 }
604 EXPORT_SYMBOL(ceph_client_addr);
605 
606 u64 ceph_client_gid(struct ceph_client *client)
607 {
608 	return client->monc.auth->global_id;
609 }
610 EXPORT_SYMBOL(ceph_client_gid);
611 
612 /*
613  * create a fresh client instance
614  */
615 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
616 {
617 	struct ceph_client *client;
618 	struct ceph_entity_addr *myaddr = NULL;
619 	int err;
620 
621 	err = wait_for_random_bytes();
622 	if (err < 0)
623 		return ERR_PTR(err);
624 
625 	client = kzalloc(sizeof(*client), GFP_KERNEL);
626 	if (client == NULL)
627 		return ERR_PTR(-ENOMEM);
628 
629 	client->private = private;
630 	client->options = opt;
631 
632 	mutex_init(&client->mount_mutex);
633 	init_waitqueue_head(&client->auth_wq);
634 	client->auth_err = 0;
635 
636 	client->extra_mon_dispatch = NULL;
637 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
638 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
639 
640 	if (!ceph_test_opt(client, NOMSGAUTH))
641 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
642 
643 	/* msgr */
644 	if (ceph_test_opt(client, MYIP))
645 		myaddr = &client->options->my_addr;
646 
647 	ceph_messenger_init(&client->msgr, myaddr);
648 
649 	/* subsystems */
650 	err = ceph_monc_init(&client->monc, client);
651 	if (err < 0)
652 		goto fail;
653 	err = ceph_osdc_init(&client->osdc, client);
654 	if (err < 0)
655 		goto fail_monc;
656 
657 	return client;
658 
659 fail_monc:
660 	ceph_monc_stop(&client->monc);
661 fail:
662 	ceph_messenger_fini(&client->msgr);
663 	kfree(client);
664 	return ERR_PTR(err);
665 }
666 EXPORT_SYMBOL(ceph_create_client);
667 
668 void ceph_destroy_client(struct ceph_client *client)
669 {
670 	dout("destroy_client %p\n", client);
671 
672 	atomic_set(&client->msgr.stopping, 1);
673 
674 	/* unmount */
675 	ceph_osdc_stop(&client->osdc);
676 	ceph_monc_stop(&client->monc);
677 	ceph_messenger_fini(&client->msgr);
678 
679 	ceph_debugfs_client_cleanup(client);
680 
681 	ceph_destroy_options(client->options);
682 
683 	kfree(client);
684 	dout("destroy_client %p done\n", client);
685 }
686 EXPORT_SYMBOL(ceph_destroy_client);
687 
688 /*
689  * true if we have the mon map (and have thus joined the cluster)
690  */
691 static bool have_mon_and_osd_map(struct ceph_client *client)
692 {
693 	return client->monc.monmap && client->monc.monmap->epoch &&
694 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
695 }
696 
697 /*
698  * mount: join the ceph cluster, and open root directory.
699  */
700 int __ceph_open_session(struct ceph_client *client, unsigned long started)
701 {
702 	unsigned long timeout = client->options->mount_timeout;
703 	long err;
704 
705 	/* open session, and wait for mon and osd maps */
706 	err = ceph_monc_open_session(&client->monc);
707 	if (err < 0)
708 		return err;
709 
710 	while (!have_mon_and_osd_map(client)) {
711 		if (timeout && time_after_eq(jiffies, started + timeout))
712 			return -ETIMEDOUT;
713 
714 		/* wait */
715 		dout("mount waiting for mon_map\n");
716 		err = wait_event_interruptible_timeout(client->auth_wq,
717 			have_mon_and_osd_map(client) || (client->auth_err < 0),
718 			ceph_timeout_jiffies(timeout));
719 		if (err < 0)
720 			return err;
721 		if (client->auth_err < 0)
722 			return client->auth_err;
723 	}
724 
725 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
726 		&client->fsid);
727 	ceph_debugfs_client_init(client);
728 
729 	return 0;
730 }
731 EXPORT_SYMBOL(__ceph_open_session);
732 
733 
734 int ceph_open_session(struct ceph_client *client)
735 {
736 	int ret;
737 	unsigned long started = jiffies;  /* note the start time */
738 
739 	dout("open_session start\n");
740 	mutex_lock(&client->mount_mutex);
741 
742 	ret = __ceph_open_session(client, started);
743 
744 	mutex_unlock(&client->mount_mutex);
745 	return ret;
746 }
747 EXPORT_SYMBOL(ceph_open_session);
748 
749 
750 static int __init init_ceph_lib(void)
751 {
752 	int ret = 0;
753 
754 	ret = ceph_debugfs_init();
755 	if (ret < 0)
756 		goto out;
757 
758 	ret = ceph_crypto_init();
759 	if (ret < 0)
760 		goto out_debugfs;
761 
762 	ret = ceph_msgr_init();
763 	if (ret < 0)
764 		goto out_crypto;
765 
766 	ret = ceph_osdc_setup();
767 	if (ret < 0)
768 		goto out_msgr;
769 
770 	pr_info("loaded (mon/osd proto %d/%d)\n",
771 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
772 
773 	return 0;
774 
775 out_msgr:
776 	ceph_msgr_exit();
777 out_crypto:
778 	ceph_crypto_shutdown();
779 out_debugfs:
780 	ceph_debugfs_cleanup();
781 out:
782 	return ret;
783 }
784 
785 static void __exit exit_ceph_lib(void)
786 {
787 	dout("exit_ceph_lib\n");
788 	WARN_ON(!ceph_strings_empty());
789 
790 	ceph_osdc_cleanup();
791 	ceph_msgr_exit();
792 	ceph_crypto_shutdown();
793 	ceph_debugfs_cleanup();
794 }
795 
796 module_init(init_ceph_lib);
797 module_exit(exit_ceph_lib);
798 
799 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
800 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
801 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
802 MODULE_DESCRIPTION("Ceph core library");
803 MODULE_LICENSE("GPL");
804