xref: /openbmc/linux/net/ceph/ceph_common.c (revision 367e5927)
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 	Opt_abort_on_full,
259 };
260 
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"},
267 	/* int args above */
268 	{Opt_fsid, "fsid=%s"},
269 	{Opt_name, "name=%s"},
270 	{Opt_secret, "secret=%s"},
271 	{Opt_key, "key=%s"},
272 	{Opt_ip, "ip=%s"},
273 	/* string args above */
274 	{Opt_share, "share"},
275 	{Opt_noshare, "noshare"},
276 	{Opt_crc, "crc"},
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"},
285 	{-1, NULL}
286 };
287 
288 void ceph_destroy_options(struct ceph_options *opt)
289 {
290 	dout("destroy_options %p\n", opt);
291 	kfree(opt->name);
292 	if (opt->key) {
293 		ceph_crypto_key_destroy(opt->key);
294 		kfree(opt->key);
295 	}
296 	kfree(opt->mon_addr);
297 	kfree(opt);
298 }
299 EXPORT_SYMBOL(ceph_destroy_options);
300 
301 /* get secret from key store */
302 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
303 	struct key *ukey;
304 	int key_err;
305 	int err = 0;
306 	struct ceph_crypto_key *ckey;
307 
308 	ukey = request_key(&key_type_ceph, name, NULL);
309 	if (IS_ERR(ukey)) {
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);
313 		switch (key_err) {
314 		case -ENOKEY:
315 			pr_warn("ceph: Mount failed due to key not found: %s\n",
316 				name);
317 			break;
318 		case -EKEYEXPIRED:
319 			pr_warn("ceph: Mount failed due to expired key: %s\n",
320 				name);
321 			break;
322 		case -EKEYREVOKED:
323 			pr_warn("ceph: Mount failed due to revoked key: %s\n",
324 				name);
325 			break;
326 		default:
327 			pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
328 				key_err, name);
329 		}
330 		err = -EPERM;
331 		goto out;
332 	}
333 
334 	ckey = ukey->payload.data[0];
335 	err = ceph_crypto_key_clone(dst, ckey);
336 	if (err)
337 		goto out_key;
338 	/* pass through, err is 0 */
339 
340 out_key:
341 	key_put(ukey);
342 out:
343 	return err;
344 }
345 
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),
350 			void *private)
351 {
352 	struct ceph_options *opt;
353 	const char *c;
354 	int err = -ENOMEM;
355 	substring_t argstr[MAX_OPT_ARGS];
356 
357 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
358 	if (!opt)
359 		return ERR_PTR(-ENOMEM);
360 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
361 				GFP_KERNEL);
362 	if (!opt->mon_addr)
363 		goto out;
364 
365 	dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
366 	     dev_name);
367 
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;
374 
375 	/* get mon ip(s) */
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);
379 	if (err < 0)
380 		goto out;
381 
382 	/* parse mount options */
383 	while ((c = strsep(&options, ",")) != NULL) {
384 		int token, intval;
385 		if (!*c)
386 			continue;
387 		err = -EINVAL;
388 		token = match_token((char *)c, opt_tokens, argstr);
389 		if (token < 0 && parse_extra_token) {
390 			/* extra? */
391 			err = parse_extra_token((char *)c, private);
392 			if (err < 0) {
393 				pr_err("bad option at '%s'\n", c);
394 				goto out;
395 			}
396 			continue;
397 		}
398 		if (token < Opt_last_int) {
399 			err = match_int(&argstr[0], &intval);
400 			if (err < 0) {
401 				pr_err("bad option arg (not int) at '%s'\n", c);
402 				goto out;
403 			}
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,
407 			     argstr[0].from);
408 		} else {
409 			dout("got token %d\n", token);
410 		}
411 		switch (token) {
412 		case Opt_ip:
413 			err = ceph_parse_ips(argstr[0].from,
414 					     argstr[0].to,
415 					     &opt->my_addr,
416 					     1, NULL);
417 			if (err < 0)
418 				goto out;
419 			opt->flags |= CEPH_OPT_MYIP;
420 			break;
421 
422 		case Opt_fsid:
423 			err = parse_fsid(argstr[0].from, &opt->fsid);
424 			if (err == 0)
425 				opt->flags |= CEPH_OPT_FSID;
426 			break;
427 		case Opt_name:
428 			kfree(opt->name);
429 			opt->name = kstrndup(argstr[0].from,
430 					      argstr[0].to-argstr[0].from,
431 					      GFP_KERNEL);
432 			if (!opt->name) {
433 				err = -ENOMEM;
434 				goto out;
435 			}
436 			break;
437 		case Opt_secret:
438 			ceph_crypto_key_destroy(opt->key);
439 			kfree(opt->key);
440 
441 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
442 			if (!opt->key) {
443 				err = -ENOMEM;
444 				goto out;
445 			}
446 			err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
447 			if (err < 0)
448 				goto out;
449 			break;
450 		case Opt_key:
451 			ceph_crypto_key_destroy(opt->key);
452 			kfree(opt->key);
453 
454 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
455 			if (!opt->key) {
456 				err = -ENOMEM;
457 				goto out;
458 			}
459 			err = get_secret(opt->key, argstr[0].from);
460 			if (err < 0)
461 				goto out;
462 			break;
463 
464 			/* misc */
465 		case Opt_osdtimeout:
466 			pr_warn("ignoring deprecated osdtimeout option\n");
467 			break;
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");
472 				err = -EINVAL;
473 				goto out;
474 			}
475 			opt->osd_keepalive_timeout =
476 					msecs_to_jiffies(intval * 1000);
477 			break;
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");
482 				err = -EINVAL;
483 				goto out;
484 			}
485 			opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
486 			break;
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");
491 				err = -EINVAL;
492 				goto out;
493 			}
494 			opt->mount_timeout = msecs_to_jiffies(intval * 1000);
495 			break;
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");
500 				err = -EINVAL;
501 				goto out;
502 			}
503 			opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
504 			break;
505 
506 		case Opt_share:
507 			opt->flags &= ~CEPH_OPT_NOSHARE;
508 			break;
509 		case Opt_noshare:
510 			opt->flags |= CEPH_OPT_NOSHARE;
511 			break;
512 
513 		case Opt_crc:
514 			opt->flags &= ~CEPH_OPT_NOCRC;
515 			break;
516 		case Opt_nocrc:
517 			opt->flags |= CEPH_OPT_NOCRC;
518 			break;
519 
520 		case Opt_cephx_require_signatures:
521 			opt->flags &= ~CEPH_OPT_NOMSGAUTH;
522 			break;
523 		case Opt_nocephx_require_signatures:
524 			opt->flags |= CEPH_OPT_NOMSGAUTH;
525 			break;
526 		case Opt_cephx_sign_messages:
527 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
528 			break;
529 		case Opt_nocephx_sign_messages:
530 			opt->flags |= CEPH_OPT_NOMSGSIGN;
531 			break;
532 
533 		case Opt_tcp_nodelay:
534 			opt->flags |= CEPH_OPT_TCP_NODELAY;
535 			break;
536 		case Opt_notcp_nodelay:
537 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
538 			break;
539 
540 		case Opt_abort_on_full:
541 			opt->flags |= CEPH_OPT_ABORT_ON_FULL;
542 			break;
543 
544 		default:
545 			BUG_ON(token);
546 		}
547 	}
548 
549 	/* success */
550 	return opt;
551 
552 out:
553 	ceph_destroy_options(opt);
554 	return ERR_PTR(err);
555 }
556 EXPORT_SYMBOL(ceph_parse_options);
557 
558 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
559 			      bool show_all)
560 {
561 	struct ceph_options *opt = client->options;
562 	size_t pos = m->count;
563 
564 	if (opt->name) {
565 		seq_puts(m, "name=");
566 		seq_escape(m, opt->name, ", \t\n\\");
567 		seq_putc(m, ',');
568 	}
569 	if (opt->key)
570 		seq_puts(m, "secret=<hidden>,");
571 
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,");
586 
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);
599 
600 	/* drop redundant comma */
601 	if (m->count != pos)
602 		m->count--;
603 
604 	return 0;
605 }
606 EXPORT_SYMBOL(ceph_print_client_options);
607 
608 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
609 {
610 	return &client->msgr.inst.addr;
611 }
612 EXPORT_SYMBOL(ceph_client_addr);
613 
614 u64 ceph_client_gid(struct ceph_client *client)
615 {
616 	return client->monc.auth->global_id;
617 }
618 EXPORT_SYMBOL(ceph_client_gid);
619 
620 /*
621  * create a fresh client instance
622  */
623 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
624 {
625 	struct ceph_client *client;
626 	struct ceph_entity_addr *myaddr = NULL;
627 	int err;
628 
629 	err = wait_for_random_bytes();
630 	if (err < 0)
631 		return ERR_PTR(err);
632 
633 	client = kzalloc(sizeof(*client), GFP_KERNEL);
634 	if (client == NULL)
635 		return ERR_PTR(-ENOMEM);
636 
637 	client->private = private;
638 	client->options = opt;
639 
640 	mutex_init(&client->mount_mutex);
641 	init_waitqueue_head(&client->auth_wq);
642 	client->auth_err = 0;
643 
644 	client->extra_mon_dispatch = NULL;
645 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
646 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
647 
648 	if (!ceph_test_opt(client, NOMSGAUTH))
649 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
650 
651 	/* msgr */
652 	if (ceph_test_opt(client, MYIP))
653 		myaddr = &client->options->my_addr;
654 
655 	ceph_messenger_init(&client->msgr, myaddr);
656 
657 	/* subsystems */
658 	err = ceph_monc_init(&client->monc, client);
659 	if (err < 0)
660 		goto fail;
661 	err = ceph_osdc_init(&client->osdc, client);
662 	if (err < 0)
663 		goto fail_monc;
664 
665 	return client;
666 
667 fail_monc:
668 	ceph_monc_stop(&client->monc);
669 fail:
670 	ceph_messenger_fini(&client->msgr);
671 	kfree(client);
672 	return ERR_PTR(err);
673 }
674 EXPORT_SYMBOL(ceph_create_client);
675 
676 void ceph_destroy_client(struct ceph_client *client)
677 {
678 	dout("destroy_client %p\n", client);
679 
680 	atomic_set(&client->msgr.stopping, 1);
681 
682 	/* unmount */
683 	ceph_osdc_stop(&client->osdc);
684 	ceph_monc_stop(&client->monc);
685 	ceph_messenger_fini(&client->msgr);
686 
687 	ceph_debugfs_client_cleanup(client);
688 
689 	ceph_destroy_options(client->options);
690 
691 	kfree(client);
692 	dout("destroy_client %p done\n", client);
693 }
694 EXPORT_SYMBOL(ceph_destroy_client);
695 
696 /*
697  * true if we have the mon map (and have thus joined the cluster)
698  */
699 static bool have_mon_and_osd_map(struct ceph_client *client)
700 {
701 	return client->monc.monmap && client->monc.monmap->epoch &&
702 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
703 }
704 
705 /*
706  * mount: join the ceph cluster, and open root directory.
707  */
708 int __ceph_open_session(struct ceph_client *client, unsigned long started)
709 {
710 	unsigned long timeout = client->options->mount_timeout;
711 	long err;
712 
713 	/* open session, and wait for mon and osd maps */
714 	err = ceph_monc_open_session(&client->monc);
715 	if (err < 0)
716 		return err;
717 
718 	while (!have_mon_and_osd_map(client)) {
719 		if (timeout && time_after_eq(jiffies, started + timeout))
720 			return -ETIMEDOUT;
721 
722 		/* wait */
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));
727 		if (err < 0)
728 			return err;
729 		if (client->auth_err < 0)
730 			return client->auth_err;
731 	}
732 
733 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
734 		&client->fsid);
735 	ceph_debugfs_client_init(client);
736 
737 	return 0;
738 }
739 EXPORT_SYMBOL(__ceph_open_session);
740 
741 int ceph_open_session(struct ceph_client *client)
742 {
743 	int ret;
744 	unsigned long started = jiffies;  /* note the start time */
745 
746 	dout("open_session start\n");
747 	mutex_lock(&client->mount_mutex);
748 
749 	ret = __ceph_open_session(client, started);
750 
751 	mutex_unlock(&client->mount_mutex);
752 	return ret;
753 }
754 EXPORT_SYMBOL(ceph_open_session);
755 
756 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
757 				unsigned long timeout)
758 {
759 	u64 newest_epoch;
760 	int ret;
761 
762 	ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
763 	if (ret)
764 		return ret;
765 
766 	if (client->osdc.osdmap->epoch >= newest_epoch)
767 		return 0;
768 
769 	ceph_osdc_maybe_request_map(&client->osdc);
770 	return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
771 }
772 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
773 
774 static int __init init_ceph_lib(void)
775 {
776 	int ret = 0;
777 
778 	ret = ceph_debugfs_init();
779 	if (ret < 0)
780 		goto out;
781 
782 	ret = ceph_crypto_init();
783 	if (ret < 0)
784 		goto out_debugfs;
785 
786 	ret = ceph_msgr_init();
787 	if (ret < 0)
788 		goto out_crypto;
789 
790 	ret = ceph_osdc_setup();
791 	if (ret < 0)
792 		goto out_msgr;
793 
794 	pr_info("loaded (mon/osd proto %d/%d)\n",
795 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
796 
797 	return 0;
798 
799 out_msgr:
800 	ceph_msgr_exit();
801 out_crypto:
802 	ceph_crypto_shutdown();
803 out_debugfs:
804 	ceph_debugfs_cleanup();
805 out:
806 	return ret;
807 }
808 
809 static void __exit exit_ceph_lib(void)
810 {
811 	dout("exit_ceph_lib\n");
812 	WARN_ON(!ceph_strings_empty());
813 
814 	ceph_osdc_cleanup();
815 	ceph_msgr_exit();
816 	ceph_crypto_shutdown();
817 	ceph_debugfs_cleanup();
818 }
819 
820 module_init(init_ceph_lib);
821 module_exit(exit_ceph_lib);
822 
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");
828