xref: /linux/net/ceph/ceph_common.c (revision af873fcecef567abf8a3468b06dd4e4aab46da6d)
1 // SPDX-License-Identifier: GPL-2.0-only
2 
3 #include <linux/ceph/ceph_debug.h>
4 #include <linux/backing-dev.h>
5 #include <linux/ctype.h>
6 #include <linux/fs.h>
7 #include <linux/inet.h>
8 #include <linux/in6.h>
9 #include <linux/key.h>
10 #include <keys/ceph-type.h>
11 #include <linux/module.h>
12 #include <linux/mount.h>
13 #include <linux/nsproxy.h>
14 #include <linux/parser.h>
15 #include <linux/sched.h>
16 #include <linux/seq_file.h>
17 #include <linux/slab.h>
18 #include <linux/statfs.h>
19 #include <linux/string.h>
20 #include <linux/vmalloc.h>
21 
22 
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>
29 #include "crypto.h"
30 
31 
32 /*
33  * Module compatibility interface.  For now it doesn't do anything,
34  * but its existence signals a certain level of functionality.
35  *
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.
40  *
41  * The data pointer can be null.
42  */
43 bool libceph_compatible(void *data)
44 {
45 	return true;
46 }
47 EXPORT_SYMBOL(libceph_compatible);
48 
49 static int param_get_supported_features(char *buffer,
50 					const struct kernel_param *kp)
51 {
52 	return sprintf(buffer, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT);
53 }
54 static const struct kernel_param_ops param_ops_supported_features = {
55 	.get = param_get_supported_features,
56 };
57 module_param_cb(supported_features, &param_ops_supported_features, NULL,
58 		0444);
59 
60 const char *ceph_msg_type_name(int type)
61 {
62 	switch (type) {
63 	case CEPH_MSG_SHUTDOWN: return "shutdown";
64 	case CEPH_MSG_PING: return "ping";
65 	case CEPH_MSG_AUTH: return "auth";
66 	case CEPH_MSG_AUTH_REPLY: return "auth_reply";
67 	case CEPH_MSG_MON_MAP: return "mon_map";
68 	case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
69 	case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
70 	case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
71 	case CEPH_MSG_STATFS: return "statfs";
72 	case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
73 	case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
74 	case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
75 	case CEPH_MSG_MDS_MAP: return "mds_map";
76 	case CEPH_MSG_FS_MAP_USER: return "fs_map_user";
77 	case CEPH_MSG_CLIENT_SESSION: return "client_session";
78 	case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
79 	case CEPH_MSG_CLIENT_REQUEST: return "client_request";
80 	case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
81 	case CEPH_MSG_CLIENT_REPLY: return "client_reply";
82 	case CEPH_MSG_CLIENT_CAPS: return "client_caps";
83 	case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
84 	case CEPH_MSG_CLIENT_QUOTA: return "client_quota";
85 	case CEPH_MSG_CLIENT_SNAP: return "client_snap";
86 	case CEPH_MSG_CLIENT_LEASE: return "client_lease";
87 	case CEPH_MSG_POOLOP_REPLY: return "poolop_reply";
88 	case CEPH_MSG_POOLOP: return "poolop";
89 	case CEPH_MSG_MON_COMMAND: return "mon_command";
90 	case CEPH_MSG_MON_COMMAND_ACK: return "mon_command_ack";
91 	case CEPH_MSG_OSD_MAP: return "osd_map";
92 	case CEPH_MSG_OSD_OP: return "osd_op";
93 	case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
94 	case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
95 	case CEPH_MSG_OSD_BACKOFF: return "osd_backoff";
96 	default: return "unknown";
97 	}
98 }
99 EXPORT_SYMBOL(ceph_msg_type_name);
100 
101 /*
102  * Initially learn our fsid, or verify an fsid matches.
103  */
104 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
105 {
106 	if (client->have_fsid) {
107 		if (ceph_fsid_compare(&client->fsid, fsid)) {
108 			pr_err("bad fsid, had %pU got %pU",
109 			       &client->fsid, fsid);
110 			return -1;
111 		}
112 	} else {
113 		memcpy(&client->fsid, fsid, sizeof(*fsid));
114 	}
115 	return 0;
116 }
117 EXPORT_SYMBOL(ceph_check_fsid);
118 
119 static int strcmp_null(const char *s1, const char *s2)
120 {
121 	if (!s1 && !s2)
122 		return 0;
123 	if (s1 && !s2)
124 		return -1;
125 	if (!s1 && s2)
126 		return 1;
127 	return strcmp(s1, s2);
128 }
129 
130 int ceph_compare_options(struct ceph_options *new_opt,
131 			 struct ceph_client *client)
132 {
133 	struct ceph_options *opt1 = new_opt;
134 	struct ceph_options *opt2 = client->options;
135 	int ofs = offsetof(struct ceph_options, mon_addr);
136 	int i;
137 	int ret;
138 
139 	/*
140 	 * Don't bother comparing options if network namespaces don't
141 	 * match.
142 	 */
143 	if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
144 		return -1;
145 
146 	ret = memcmp(opt1, opt2, ofs);
147 	if (ret)
148 		return ret;
149 
150 	ret = strcmp_null(opt1->name, opt2->name);
151 	if (ret)
152 		return ret;
153 
154 	if (opt1->key && !opt2->key)
155 		return -1;
156 	if (!opt1->key && opt2->key)
157 		return 1;
158 	if (opt1->key && opt2->key) {
159 		if (opt1->key->type != opt2->key->type)
160 			return -1;
161 		if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
162 			return -1;
163 		if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
164 			return -1;
165 		if (opt1->key->len != opt2->key->len)
166 			return -1;
167 		if (opt1->key->key && !opt2->key->key)
168 			return -1;
169 		if (!opt1->key->key && opt2->key->key)
170 			return 1;
171 		if (opt1->key->key && opt2->key->key) {
172 			ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
173 			if (ret)
174 				return ret;
175 		}
176 	}
177 
178 	/* any matching mon ip implies a match */
179 	for (i = 0; i < opt1->num_mon; i++) {
180 		if (ceph_monmap_contains(client->monc.monmap,
181 				 &opt1->mon_addr[i]))
182 			return 0;
183 	}
184 	return -1;
185 }
186 EXPORT_SYMBOL(ceph_compare_options);
187 
188 void *ceph_kvmalloc(size_t size, gfp_t flags)
189 {
190 	if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
191 		void *ptr = kmalloc(size, flags | __GFP_NOWARN);
192 		if (ptr)
193 			return ptr;
194 	}
195 
196 	return __vmalloc(size, flags, PAGE_KERNEL);
197 }
198 
199 
200 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
201 {
202 	int i = 0;
203 	char tmp[3];
204 	int err = -EINVAL;
205 	int d;
206 
207 	dout("parse_fsid '%s'\n", str);
208 	tmp[2] = 0;
209 	while (*str && i < 16) {
210 		if (ispunct(*str)) {
211 			str++;
212 			continue;
213 		}
214 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
215 			break;
216 		tmp[0] = str[0];
217 		tmp[1] = str[1];
218 		if (sscanf(tmp, "%x", &d) < 1)
219 			break;
220 		fsid->fsid[i] = d & 0xff;
221 		i++;
222 		str += 2;
223 	}
224 
225 	if (i == 16)
226 		err = 0;
227 	dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
228 	return err;
229 }
230 
231 /*
232  * ceph options
233  */
234 enum {
235 	Opt_osdtimeout,
236 	Opt_osdkeepalivetimeout,
237 	Opt_mount_timeout,
238 	Opt_osd_idle_ttl,
239 	Opt_osd_request_timeout,
240 	Opt_last_int,
241 	/* int args above */
242 	Opt_fsid,
243 	Opt_name,
244 	Opt_secret,
245 	Opt_key,
246 	Opt_ip,
247 	Opt_last_string,
248 	/* string args above */
249 	Opt_share,
250 	Opt_noshare,
251 	Opt_crc,
252 	Opt_nocrc,
253 	Opt_cephx_require_signatures,
254 	Opt_nocephx_require_signatures,
255 	Opt_cephx_sign_messages,
256 	Opt_nocephx_sign_messages,
257 	Opt_tcp_nodelay,
258 	Opt_notcp_nodelay,
259 	Opt_abort_on_full,
260 };
261 
262 static match_table_t opt_tokens = {
263 	{Opt_osdtimeout, "osdtimeout=%d"},
264 	{Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
265 	{Opt_mount_timeout, "mount_timeout=%d"},
266 	{Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
267 	{Opt_osd_request_timeout, "osd_request_timeout=%d"},
268 	/* int args above */
269 	{Opt_fsid, "fsid=%s"},
270 	{Opt_name, "name=%s"},
271 	{Opt_secret, "secret=%s"},
272 	{Opt_key, "key=%s"},
273 	{Opt_ip, "ip=%s"},
274 	/* string args above */
275 	{Opt_share, "share"},
276 	{Opt_noshare, "noshare"},
277 	{Opt_crc, "crc"},
278 	{Opt_nocrc, "nocrc"},
279 	{Opt_cephx_require_signatures, "cephx_require_signatures"},
280 	{Opt_nocephx_require_signatures, "nocephx_require_signatures"},
281 	{Opt_cephx_sign_messages, "cephx_sign_messages"},
282 	{Opt_nocephx_sign_messages, "nocephx_sign_messages"},
283 	{Opt_tcp_nodelay, "tcp_nodelay"},
284 	{Opt_notcp_nodelay, "notcp_nodelay"},
285 	{Opt_abort_on_full, "abort_on_full"},
286 	{-1, NULL}
287 };
288 
289 void ceph_destroy_options(struct ceph_options *opt)
290 {
291 	dout("destroy_options %p\n", opt);
292 	kfree(opt->name);
293 	if (opt->key) {
294 		ceph_crypto_key_destroy(opt->key);
295 		kfree(opt->key);
296 	}
297 	kfree(opt->mon_addr);
298 	kfree(opt);
299 }
300 EXPORT_SYMBOL(ceph_destroy_options);
301 
302 /* get secret from key store */
303 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
304 	struct key *ukey;
305 	int key_err;
306 	int err = 0;
307 	struct ceph_crypto_key *ckey;
308 
309 	ukey = request_key(&key_type_ceph, name, NULL);
310 	if (IS_ERR(ukey)) {
311 		/* request_key errors don't map nicely to mount(2)
312 		   errors; don't even try, but still printk */
313 		key_err = PTR_ERR(ukey);
314 		switch (key_err) {
315 		case -ENOKEY:
316 			pr_warn("ceph: Mount failed due to key not found: %s\n",
317 				name);
318 			break;
319 		case -EKEYEXPIRED:
320 			pr_warn("ceph: Mount failed due to expired key: %s\n",
321 				name);
322 			break;
323 		case -EKEYREVOKED:
324 			pr_warn("ceph: Mount failed due to revoked key: %s\n",
325 				name);
326 			break;
327 		default:
328 			pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
329 				key_err, name);
330 		}
331 		err = -EPERM;
332 		goto out;
333 	}
334 
335 	ckey = ukey->payload.data[0];
336 	err = ceph_crypto_key_clone(dst, ckey);
337 	if (err)
338 		goto out_key;
339 	/* pass through, err is 0 */
340 
341 out_key:
342 	key_put(ukey);
343 out:
344 	return err;
345 }
346 
347 struct ceph_options *
348 ceph_parse_options(char *options, const char *dev_name,
349 			const char *dev_name_end,
350 			int (*parse_extra_token)(char *c, void *private),
351 			void *private)
352 {
353 	struct ceph_options *opt;
354 	const char *c;
355 	int err = -ENOMEM;
356 	substring_t argstr[MAX_OPT_ARGS];
357 
358 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
359 	if (!opt)
360 		return ERR_PTR(-ENOMEM);
361 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
362 				GFP_KERNEL);
363 	if (!opt->mon_addr)
364 		goto out;
365 
366 	dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
367 	     dev_name);
368 
369 	/* start with defaults */
370 	opt->flags = CEPH_OPT_DEFAULT;
371 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
372 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
373 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
374 	opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
375 
376 	/* get mon ip(s) */
377 	/* ip1[:port1][,ip2[:port2]...] */
378 	err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
379 			     CEPH_MAX_MON, &opt->num_mon);
380 	if (err < 0)
381 		goto out;
382 
383 	/* parse mount options */
384 	while ((c = strsep(&options, ",")) != NULL) {
385 		int token, intval;
386 		if (!*c)
387 			continue;
388 		err = -EINVAL;
389 		token = match_token((char *)c, opt_tokens, argstr);
390 		if (token < 0 && parse_extra_token) {
391 			/* extra? */
392 			err = parse_extra_token((char *)c, private);
393 			if (err < 0) {
394 				pr_err("bad option at '%s'\n", c);
395 				goto out;
396 			}
397 			continue;
398 		}
399 		if (token < Opt_last_int) {
400 			err = match_int(&argstr[0], &intval);
401 			if (err < 0) {
402 				pr_err("bad option arg (not int) at '%s'\n", c);
403 				goto out;
404 			}
405 			dout("got int token %d val %d\n", token, intval);
406 		} else if (token > Opt_last_int && token < Opt_last_string) {
407 			dout("got string token %d val %s\n", token,
408 			     argstr[0].from);
409 		} else {
410 			dout("got token %d\n", token);
411 		}
412 		switch (token) {
413 		case Opt_ip:
414 			err = ceph_parse_ips(argstr[0].from,
415 					     argstr[0].to,
416 					     &opt->my_addr,
417 					     1, NULL);
418 			if (err < 0)
419 				goto out;
420 			opt->flags |= CEPH_OPT_MYIP;
421 			break;
422 
423 		case Opt_fsid:
424 			err = parse_fsid(argstr[0].from, &opt->fsid);
425 			if (err == 0)
426 				opt->flags |= CEPH_OPT_FSID;
427 			break;
428 		case Opt_name:
429 			kfree(opt->name);
430 			opt->name = kstrndup(argstr[0].from,
431 					      argstr[0].to-argstr[0].from,
432 					      GFP_KERNEL);
433 			if (!opt->name) {
434 				err = -ENOMEM;
435 				goto out;
436 			}
437 			break;
438 		case Opt_secret:
439 			ceph_crypto_key_destroy(opt->key);
440 			kfree(opt->key);
441 
442 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
443 			if (!opt->key) {
444 				err = -ENOMEM;
445 				goto out;
446 			}
447 			err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
448 			if (err < 0)
449 				goto out;
450 			break;
451 		case Opt_key:
452 			ceph_crypto_key_destroy(opt->key);
453 			kfree(opt->key);
454 
455 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
456 			if (!opt->key) {
457 				err = -ENOMEM;
458 				goto out;
459 			}
460 			err = get_secret(opt->key, argstr[0].from);
461 			if (err < 0)
462 				goto out;
463 			break;
464 
465 			/* misc */
466 		case Opt_osdtimeout:
467 			pr_warn("ignoring deprecated osdtimeout option\n");
468 			break;
469 		case Opt_osdkeepalivetimeout:
470 			/* 0 isn't well defined right now, reject it */
471 			if (intval < 1 || intval > INT_MAX / 1000) {
472 				pr_err("osdkeepalive out of range\n");
473 				err = -EINVAL;
474 				goto out;
475 			}
476 			opt->osd_keepalive_timeout =
477 					msecs_to_jiffies(intval * 1000);
478 			break;
479 		case Opt_osd_idle_ttl:
480 			/* 0 isn't well defined right now, reject it */
481 			if (intval < 1 || intval > INT_MAX / 1000) {
482 				pr_err("osd_idle_ttl out of range\n");
483 				err = -EINVAL;
484 				goto out;
485 			}
486 			opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
487 			break;
488 		case Opt_mount_timeout:
489 			/* 0 is "wait forever" (i.e. infinite timeout) */
490 			if (intval < 0 || intval > INT_MAX / 1000) {
491 				pr_err("mount_timeout out of range\n");
492 				err = -EINVAL;
493 				goto out;
494 			}
495 			opt->mount_timeout = msecs_to_jiffies(intval * 1000);
496 			break;
497 		case Opt_osd_request_timeout:
498 			/* 0 is "wait forever" (i.e. infinite timeout) */
499 			if (intval < 0 || intval > INT_MAX / 1000) {
500 				pr_err("osd_request_timeout out of range\n");
501 				err = -EINVAL;
502 				goto out;
503 			}
504 			opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
505 			break;
506 
507 		case Opt_share:
508 			opt->flags &= ~CEPH_OPT_NOSHARE;
509 			break;
510 		case Opt_noshare:
511 			opt->flags |= CEPH_OPT_NOSHARE;
512 			break;
513 
514 		case Opt_crc:
515 			opt->flags &= ~CEPH_OPT_NOCRC;
516 			break;
517 		case Opt_nocrc:
518 			opt->flags |= CEPH_OPT_NOCRC;
519 			break;
520 
521 		case Opt_cephx_require_signatures:
522 			opt->flags &= ~CEPH_OPT_NOMSGAUTH;
523 			break;
524 		case Opt_nocephx_require_signatures:
525 			opt->flags |= CEPH_OPT_NOMSGAUTH;
526 			break;
527 		case Opt_cephx_sign_messages:
528 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
529 			break;
530 		case Opt_nocephx_sign_messages:
531 			opt->flags |= CEPH_OPT_NOMSGSIGN;
532 			break;
533 
534 		case Opt_tcp_nodelay:
535 			opt->flags |= CEPH_OPT_TCP_NODELAY;
536 			break;
537 		case Opt_notcp_nodelay:
538 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
539 			break;
540 
541 		case Opt_abort_on_full:
542 			opt->flags |= CEPH_OPT_ABORT_ON_FULL;
543 			break;
544 
545 		default:
546 			BUG_ON(token);
547 		}
548 	}
549 
550 	/* success */
551 	return opt;
552 
553 out:
554 	ceph_destroy_options(opt);
555 	return ERR_PTR(err);
556 }
557 EXPORT_SYMBOL(ceph_parse_options);
558 
559 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
560 			      bool show_all)
561 {
562 	struct ceph_options *opt = client->options;
563 	size_t pos = m->count;
564 
565 	if (opt->name) {
566 		seq_puts(m, "name=");
567 		seq_escape(m, opt->name, ", \t\n\\");
568 		seq_putc(m, ',');
569 	}
570 	if (opt->key)
571 		seq_puts(m, "secret=<hidden>,");
572 
573 	if (opt->flags & CEPH_OPT_FSID)
574 		seq_printf(m, "fsid=%pU,", &opt->fsid);
575 	if (opt->flags & CEPH_OPT_NOSHARE)
576 		seq_puts(m, "noshare,");
577 	if (opt->flags & CEPH_OPT_NOCRC)
578 		seq_puts(m, "nocrc,");
579 	if (opt->flags & CEPH_OPT_NOMSGAUTH)
580 		seq_puts(m, "nocephx_require_signatures,");
581 	if (opt->flags & CEPH_OPT_NOMSGSIGN)
582 		seq_puts(m, "nocephx_sign_messages,");
583 	if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
584 		seq_puts(m, "notcp_nodelay,");
585 	if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
586 		seq_puts(m, "abort_on_full,");
587 
588 	if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
589 		seq_printf(m, "mount_timeout=%d,",
590 			   jiffies_to_msecs(opt->mount_timeout) / 1000);
591 	if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
592 		seq_printf(m, "osd_idle_ttl=%d,",
593 			   jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
594 	if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
595 		seq_printf(m, "osdkeepalivetimeout=%d,",
596 		    jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
597 	if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
598 		seq_printf(m, "osd_request_timeout=%d,",
599 			   jiffies_to_msecs(opt->osd_request_timeout) / 1000);
600 
601 	/* drop redundant comma */
602 	if (m->count != pos)
603 		m->count--;
604 
605 	return 0;
606 }
607 EXPORT_SYMBOL(ceph_print_client_options);
608 
609 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
610 {
611 	return &client->msgr.inst.addr;
612 }
613 EXPORT_SYMBOL(ceph_client_addr);
614 
615 u64 ceph_client_gid(struct ceph_client *client)
616 {
617 	return client->monc.auth->global_id;
618 }
619 EXPORT_SYMBOL(ceph_client_gid);
620 
621 /*
622  * create a fresh client instance
623  */
624 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
625 {
626 	struct ceph_client *client;
627 	struct ceph_entity_addr *myaddr = NULL;
628 	int err;
629 
630 	err = wait_for_random_bytes();
631 	if (err < 0)
632 		return ERR_PTR(err);
633 
634 	client = kzalloc(sizeof(*client), GFP_KERNEL);
635 	if (client == NULL)
636 		return ERR_PTR(-ENOMEM);
637 
638 	client->private = private;
639 	client->options = opt;
640 
641 	mutex_init(&client->mount_mutex);
642 	init_waitqueue_head(&client->auth_wq);
643 	client->auth_err = 0;
644 
645 	client->extra_mon_dispatch = NULL;
646 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
647 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
648 
649 	if (!ceph_test_opt(client, NOMSGAUTH))
650 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
651 
652 	/* msgr */
653 	if (ceph_test_opt(client, MYIP))
654 		myaddr = &client->options->my_addr;
655 
656 	ceph_messenger_init(&client->msgr, myaddr);
657 
658 	/* subsystems */
659 	err = ceph_monc_init(&client->monc, client);
660 	if (err < 0)
661 		goto fail;
662 	err = ceph_osdc_init(&client->osdc, client);
663 	if (err < 0)
664 		goto fail_monc;
665 
666 	return client;
667 
668 fail_monc:
669 	ceph_monc_stop(&client->monc);
670 fail:
671 	ceph_messenger_fini(&client->msgr);
672 	kfree(client);
673 	return ERR_PTR(err);
674 }
675 EXPORT_SYMBOL(ceph_create_client);
676 
677 void ceph_destroy_client(struct ceph_client *client)
678 {
679 	dout("destroy_client %p\n", client);
680 
681 	atomic_set(&client->msgr.stopping, 1);
682 
683 	/* unmount */
684 	ceph_osdc_stop(&client->osdc);
685 	ceph_monc_stop(&client->monc);
686 	ceph_messenger_fini(&client->msgr);
687 
688 	ceph_debugfs_client_cleanup(client);
689 
690 	ceph_destroy_options(client->options);
691 
692 	kfree(client);
693 	dout("destroy_client %p done\n", client);
694 }
695 EXPORT_SYMBOL(ceph_destroy_client);
696 
697 /*
698  * true if we have the mon map (and have thus joined the cluster)
699  */
700 static bool have_mon_and_osd_map(struct ceph_client *client)
701 {
702 	return client->monc.monmap && client->monc.monmap->epoch &&
703 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
704 }
705 
706 /*
707  * mount: join the ceph cluster, and open root directory.
708  */
709 int __ceph_open_session(struct ceph_client *client, unsigned long started)
710 {
711 	unsigned long timeout = client->options->mount_timeout;
712 	long err;
713 
714 	/* open session, and wait for mon and osd maps */
715 	err = ceph_monc_open_session(&client->monc);
716 	if (err < 0)
717 		return err;
718 
719 	while (!have_mon_and_osd_map(client)) {
720 		if (timeout && time_after_eq(jiffies, started + timeout))
721 			return -ETIMEDOUT;
722 
723 		/* wait */
724 		dout("mount waiting for mon_map\n");
725 		err = wait_event_interruptible_timeout(client->auth_wq,
726 			have_mon_and_osd_map(client) || (client->auth_err < 0),
727 			ceph_timeout_jiffies(timeout));
728 		if (err < 0)
729 			return err;
730 		if (client->auth_err < 0)
731 			return client->auth_err;
732 	}
733 
734 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
735 		&client->fsid);
736 	ceph_debugfs_client_init(client);
737 
738 	return 0;
739 }
740 EXPORT_SYMBOL(__ceph_open_session);
741 
742 int ceph_open_session(struct ceph_client *client)
743 {
744 	int ret;
745 	unsigned long started = jiffies;  /* note the start time */
746 
747 	dout("open_session start\n");
748 	mutex_lock(&client->mount_mutex);
749 
750 	ret = __ceph_open_session(client, started);
751 
752 	mutex_unlock(&client->mount_mutex);
753 	return ret;
754 }
755 EXPORT_SYMBOL(ceph_open_session);
756 
757 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
758 				unsigned long timeout)
759 {
760 	u64 newest_epoch;
761 	int ret;
762 
763 	ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
764 	if (ret)
765 		return ret;
766 
767 	if (client->osdc.osdmap->epoch >= newest_epoch)
768 		return 0;
769 
770 	ceph_osdc_maybe_request_map(&client->osdc);
771 	return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
772 }
773 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
774 
775 static int __init init_ceph_lib(void)
776 {
777 	int ret = 0;
778 
779 	ret = ceph_debugfs_init();
780 	if (ret < 0)
781 		goto out;
782 
783 	ret = ceph_crypto_init();
784 	if (ret < 0)
785 		goto out_debugfs;
786 
787 	ret = ceph_msgr_init();
788 	if (ret < 0)
789 		goto out_crypto;
790 
791 	ret = ceph_osdc_setup();
792 	if (ret < 0)
793 		goto out_msgr;
794 
795 	pr_info("loaded (mon/osd proto %d/%d)\n",
796 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
797 
798 	return 0;
799 
800 out_msgr:
801 	ceph_msgr_exit();
802 out_crypto:
803 	ceph_crypto_shutdown();
804 out_debugfs:
805 	ceph_debugfs_cleanup();
806 out:
807 	return ret;
808 }
809 
810 static void __exit exit_ceph_lib(void)
811 {
812 	dout("exit_ceph_lib\n");
813 	WARN_ON(!ceph_strings_empty());
814 
815 	ceph_osdc_cleanup();
816 	ceph_msgr_exit();
817 	ceph_crypto_shutdown();
818 	ceph_debugfs_cleanup();
819 }
820 
821 module_init(init_ceph_lib);
822 module_exit(exit_ceph_lib);
823 
824 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
825 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
826 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
827 MODULE_DESCRIPTION("Ceph core library");
828 MODULE_LICENSE("GPL");
829