xref: /openbmc/linux/net/ceph/ceph_common.c (revision 3ce311af)
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/fs_parser.h>
15 #include <linux/sched.h>
16 #include <linux/sched/mm.h>
17 #include <linux/seq_file.h>
18 #include <linux/slab.h>
19 #include <linux/statfs.h>
20 #include <linux/string.h>
21 #include <linux/vmalloc.h>
22 
23 
24 #include <linux/ceph/ceph_features.h>
25 #include <linux/ceph/libceph.h>
26 #include <linux/ceph/debugfs.h>
27 #include <linux/ceph/decode.h>
28 #include <linux/ceph/mon_client.h>
29 #include <linux/ceph/auth.h>
30 #include "crypto.h"
31 
32 
33 /*
34  * Module compatibility interface.  For now it doesn't do anything,
35  * but its existence signals a certain level of functionality.
36  *
37  * The data buffer is used to pass information both to and from
38  * libceph.  The return value indicates whether libceph determines
39  * it is compatible with the caller (from another kernel module),
40  * given the provided data.
41  *
42  * The data pointer can be null.
43  */
44 bool libceph_compatible(void *data)
45 {
46 	return true;
47 }
48 EXPORT_SYMBOL(libceph_compatible);
49 
50 static int param_get_supported_features(char *buffer,
51 					const struct kernel_param *kp)
52 {
53 	return sprintf(buffer, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT);
54 }
55 static const struct kernel_param_ops param_ops_supported_features = {
56 	.get = param_get_supported_features,
57 };
58 module_param_cb(supported_features, &param_ops_supported_features, NULL,
59 		0444);
60 
61 const char *ceph_msg_type_name(int type)
62 {
63 	switch (type) {
64 	case CEPH_MSG_SHUTDOWN: return "shutdown";
65 	case CEPH_MSG_PING: return "ping";
66 	case CEPH_MSG_AUTH: return "auth";
67 	case CEPH_MSG_AUTH_REPLY: return "auth_reply";
68 	case CEPH_MSG_MON_MAP: return "mon_map";
69 	case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
70 	case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
71 	case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
72 	case CEPH_MSG_STATFS: return "statfs";
73 	case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
74 	case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
75 	case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
76 	case CEPH_MSG_MDS_MAP: return "mds_map";
77 	case CEPH_MSG_FS_MAP_USER: return "fs_map_user";
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_QUOTA: return "client_quota";
86 	case CEPH_MSG_CLIENT_SNAP: return "client_snap";
87 	case CEPH_MSG_CLIENT_LEASE: return "client_lease";
88 	case CEPH_MSG_POOLOP_REPLY: return "poolop_reply";
89 	case CEPH_MSG_POOLOP: return "poolop";
90 	case CEPH_MSG_MON_COMMAND: return "mon_command";
91 	case CEPH_MSG_MON_COMMAND_ACK: return "mon_command_ack";
92 	case CEPH_MSG_OSD_MAP: return "osd_map";
93 	case CEPH_MSG_OSD_OP: return "osd_op";
94 	case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
95 	case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
96 	case CEPH_MSG_OSD_BACKOFF: return "osd_backoff";
97 	default: return "unknown";
98 	}
99 }
100 EXPORT_SYMBOL(ceph_msg_type_name);
101 
102 /*
103  * Initially learn our fsid, or verify an fsid matches.
104  */
105 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
106 {
107 	if (client->have_fsid) {
108 		if (ceph_fsid_compare(&client->fsid, fsid)) {
109 			pr_err("bad fsid, had %pU got %pU",
110 			       &client->fsid, fsid);
111 			return -1;
112 		}
113 	} else {
114 		memcpy(&client->fsid, fsid, sizeof(*fsid));
115 	}
116 	return 0;
117 }
118 EXPORT_SYMBOL(ceph_check_fsid);
119 
120 static int strcmp_null(const char *s1, const char *s2)
121 {
122 	if (!s1 && !s2)
123 		return 0;
124 	if (s1 && !s2)
125 		return -1;
126 	if (!s1 && s2)
127 		return 1;
128 	return strcmp(s1, s2);
129 }
130 
131 int ceph_compare_options(struct ceph_options *new_opt,
132 			 struct ceph_client *client)
133 {
134 	struct ceph_options *opt1 = new_opt;
135 	struct ceph_options *opt2 = client->options;
136 	int ofs = offsetof(struct ceph_options, mon_addr);
137 	int i;
138 	int ret;
139 
140 	/*
141 	 * Don't bother comparing options if network namespaces don't
142 	 * match.
143 	 */
144 	if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
145 		return -1;
146 
147 	ret = memcmp(opt1, opt2, ofs);
148 	if (ret)
149 		return ret;
150 
151 	ret = strcmp_null(opt1->name, opt2->name);
152 	if (ret)
153 		return ret;
154 
155 	if (opt1->key && !opt2->key)
156 		return -1;
157 	if (!opt1->key && opt2->key)
158 		return 1;
159 	if (opt1->key && opt2->key) {
160 		if (opt1->key->type != opt2->key->type)
161 			return -1;
162 		if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
163 			return -1;
164 		if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
165 			return -1;
166 		if (opt1->key->len != opt2->key->len)
167 			return -1;
168 		if (opt1->key->key && !opt2->key->key)
169 			return -1;
170 		if (!opt1->key->key && opt2->key->key)
171 			return 1;
172 		if (opt1->key->key && opt2->key->key) {
173 			ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
174 			if (ret)
175 				return ret;
176 		}
177 	}
178 
179 	/* any matching mon ip implies a match */
180 	for (i = 0; i < opt1->num_mon; i++) {
181 		if (ceph_monmap_contains(client->monc.monmap,
182 				 &opt1->mon_addr[i]))
183 			return 0;
184 	}
185 	return -1;
186 }
187 EXPORT_SYMBOL(ceph_compare_options);
188 
189 /*
190  * kvmalloc() doesn't fall back to the vmalloc allocator unless flags are
191  * compatible with (a superset of) GFP_KERNEL.  This is because while the
192  * actual pages are allocated with the specified flags, the page table pages
193  * are always allocated with GFP_KERNEL.  map_vm_area() doesn't even take
194  * flags because GFP_KERNEL is hard-coded in {p4d,pud,pmd,pte}_alloc().
195  *
196  * ceph_kvmalloc() may be called with GFP_KERNEL, GFP_NOFS or GFP_NOIO.
197  */
198 void *ceph_kvmalloc(size_t size, gfp_t flags)
199 {
200 	void *p;
201 
202 	if ((flags & (__GFP_IO | __GFP_FS)) == (__GFP_IO | __GFP_FS)) {
203 		p = kvmalloc(size, flags);
204 	} else if ((flags & (__GFP_IO | __GFP_FS)) == __GFP_IO) {
205 		unsigned int nofs_flag = memalloc_nofs_save();
206 		p = kvmalloc(size, GFP_KERNEL);
207 		memalloc_nofs_restore(nofs_flag);
208 	} else {
209 		unsigned int noio_flag = memalloc_noio_save();
210 		p = kvmalloc(size, GFP_KERNEL);
211 		memalloc_noio_restore(noio_flag);
212 	}
213 
214 	return p;
215 }
216 
217 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
218 {
219 	int i = 0;
220 	char tmp[3];
221 	int err = -EINVAL;
222 	int d;
223 
224 	dout("parse_fsid '%s'\n", str);
225 	tmp[2] = 0;
226 	while (*str && i < 16) {
227 		if (ispunct(*str)) {
228 			str++;
229 			continue;
230 		}
231 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
232 			break;
233 		tmp[0] = str[0];
234 		tmp[1] = str[1];
235 		if (sscanf(tmp, "%x", &d) < 1)
236 			break;
237 		fsid->fsid[i] = d & 0xff;
238 		i++;
239 		str += 2;
240 	}
241 
242 	if (i == 16)
243 		err = 0;
244 	dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
245 	return err;
246 }
247 
248 /*
249  * ceph options
250  */
251 enum {
252 	Opt_osdtimeout,
253 	Opt_osdkeepalivetimeout,
254 	Opt_mount_timeout,
255 	Opt_osd_idle_ttl,
256 	Opt_osd_request_timeout,
257 	/* int args above */
258 	Opt_fsid,
259 	Opt_name,
260 	Opt_secret,
261 	Opt_key,
262 	Opt_ip,
263 	/* string args above */
264 	Opt_share,
265 	Opt_crc,
266 	Opt_cephx_require_signatures,
267 	Opt_cephx_sign_messages,
268 	Opt_tcp_nodelay,
269 	Opt_abort_on_full,
270 };
271 
272 static const struct fs_parameter_spec ceph_param_specs[] = {
273 	fsparam_flag	("abort_on_full",		Opt_abort_on_full),
274 	fsparam_flag_no ("cephx_require_signatures",	Opt_cephx_require_signatures),
275 	fsparam_flag_no ("cephx_sign_messages",		Opt_cephx_sign_messages),
276 	fsparam_flag_no ("crc",				Opt_crc),
277 	fsparam_string	("fsid",			Opt_fsid),
278 	fsparam_string	("ip",				Opt_ip),
279 	fsparam_string	("key",				Opt_key),
280 	fsparam_u32	("mount_timeout",		Opt_mount_timeout),
281 	fsparam_string	("name",			Opt_name),
282 	fsparam_u32	("osd_idle_ttl",		Opt_osd_idle_ttl),
283 	fsparam_u32	("osd_request_timeout",		Opt_osd_request_timeout),
284 	fsparam_u32	("osdkeepalive",		Opt_osdkeepalivetimeout),
285 	__fsparam	(fs_param_is_s32, "osdtimeout", Opt_osdtimeout,
286 			 fs_param_deprecated),
287 	fsparam_string	("secret",			Opt_secret),
288 	fsparam_flag_no ("share",			Opt_share),
289 	fsparam_flag_no ("tcp_nodelay",			Opt_tcp_nodelay),
290 	{}
291 };
292 
293 static const struct fs_parameter_description ceph_parameters = {
294         .name           = "libceph",
295         .specs          = ceph_param_specs,
296 };
297 
298 struct ceph_options *ceph_alloc_options(void)
299 {
300 	struct ceph_options *opt;
301 
302 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
303 	if (!opt)
304 		return NULL;
305 
306 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
307 				GFP_KERNEL);
308 	if (!opt->mon_addr) {
309 		kfree(opt);
310 		return NULL;
311 	}
312 
313 	opt->flags = CEPH_OPT_DEFAULT;
314 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
315 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
316 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
317 	opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
318 	return opt;
319 }
320 EXPORT_SYMBOL(ceph_alloc_options);
321 
322 void ceph_destroy_options(struct ceph_options *opt)
323 {
324 	dout("destroy_options %p\n", opt);
325 	if (!opt)
326 		return;
327 
328 	kfree(opt->name);
329 	if (opt->key) {
330 		ceph_crypto_key_destroy(opt->key);
331 		kfree(opt->key);
332 	}
333 	kfree(opt->mon_addr);
334 	kfree(opt);
335 }
336 EXPORT_SYMBOL(ceph_destroy_options);
337 
338 /* get secret from key store */
339 static int get_secret(struct ceph_crypto_key *dst, const char *name,
340 		      struct fs_context *fc)
341 {
342 	struct key *ukey;
343 	int key_err;
344 	int err = 0;
345 	struct ceph_crypto_key *ckey;
346 
347 	ukey = request_key(&key_type_ceph, name, NULL);
348 	if (IS_ERR(ukey)) {
349 		/* request_key errors don't map nicely to mount(2)
350 		   errors; don't even try, but still printk */
351 		key_err = PTR_ERR(ukey);
352 		switch (key_err) {
353 		case -ENOKEY:
354 			errorf(fc, "libceph: Failed due to key not found: %s",
355 			       name);
356 			break;
357 		case -EKEYEXPIRED:
358 			errorf(fc, "libceph: Failed due to expired key: %s",
359 			       name);
360 			break;
361 		case -EKEYREVOKED:
362 			errorf(fc, "libceph: Failed due to revoked key: %s",
363 			       name);
364 			break;
365 		default:
366 			errorf(fc, "libceph: Failed due to key error %d: %s",
367 			       key_err, name);
368 		}
369 		err = -EPERM;
370 		goto out;
371 	}
372 
373 	ckey = ukey->payload.data[0];
374 	err = ceph_crypto_key_clone(dst, ckey);
375 	if (err)
376 		goto out_key;
377 	/* pass through, err is 0 */
378 
379 out_key:
380 	key_put(ukey);
381 out:
382 	return err;
383 }
384 
385 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
386 		       struct fs_context *fc)
387 {
388 	int ret;
389 
390 	/* ip1[:port1][,ip2[:port2]...] */
391 	ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
392 			     &opt->num_mon);
393 	if (ret) {
394 		errorf(fc, "libceph: Failed to parse monitor IPs: %d", ret);
395 		return ret;
396 	}
397 
398 	return 0;
399 }
400 EXPORT_SYMBOL(ceph_parse_mon_ips);
401 
402 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
403 		     struct fs_context *fc)
404 {
405 	struct fs_parse_result result;
406 	int token, err;
407 
408 	token = fs_parse(fc, &ceph_parameters, param, &result);
409 	dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
410 	if (token < 0)
411 		return token;
412 
413 	switch (token) {
414 	case Opt_ip:
415 		err = ceph_parse_ips(param->string,
416 				     param->string + param->size,
417 				     &opt->my_addr,
418 				     1, NULL);
419 		if (err) {
420 			errorf(fc, "libceph: Failed to parse ip: %d", err);
421 			return err;
422 		}
423 		opt->flags |= CEPH_OPT_MYIP;
424 		break;
425 
426 	case Opt_fsid:
427 		err = parse_fsid(param->string, &opt->fsid);
428 		if (err) {
429 			errorf(fc, "libceph: Failed to parse fsid: %d", err);
430 			return err;
431 		}
432 		opt->flags |= CEPH_OPT_FSID;
433 		break;
434 	case Opt_name:
435 		kfree(opt->name);
436 		opt->name = param->string;
437 		param->string = NULL;
438 		break;
439 	case Opt_secret:
440 		ceph_crypto_key_destroy(opt->key);
441 		kfree(opt->key);
442 
443 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
444 		if (!opt->key)
445 			return -ENOMEM;
446 		err = ceph_crypto_key_unarmor(opt->key, param->string);
447 		if (err) {
448 			errorf(fc, "libceph: Failed to parse secret: %d", err);
449 			return err;
450 		}
451 		break;
452 	case Opt_key:
453 		ceph_crypto_key_destroy(opt->key);
454 		kfree(opt->key);
455 
456 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
457 		if (!opt->key)
458 			return -ENOMEM;
459 		return get_secret(opt->key, param->string, fc);
460 
461 	case Opt_osdtimeout:
462 		warnf(fc, "libceph: Ignoring osdtimeout");
463 		break;
464 	case Opt_osdkeepalivetimeout:
465 		/* 0 isn't well defined right now, reject it */
466 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
467 			goto out_of_range;
468 		opt->osd_keepalive_timeout =
469 		    msecs_to_jiffies(result.uint_32 * 1000);
470 		break;
471 	case Opt_osd_idle_ttl:
472 		/* 0 isn't well defined right now, reject it */
473 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
474 			goto out_of_range;
475 		opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
476 		break;
477 	case Opt_mount_timeout:
478 		/* 0 is "wait forever" (i.e. infinite timeout) */
479 		if (result.uint_32 > INT_MAX / 1000)
480 			goto out_of_range;
481 		opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
482 		break;
483 	case Opt_osd_request_timeout:
484 		/* 0 is "wait forever" (i.e. infinite timeout) */
485 		if (result.uint_32 > INT_MAX / 1000)
486 			goto out_of_range;
487 		opt->osd_request_timeout =
488 		    msecs_to_jiffies(result.uint_32 * 1000);
489 		break;
490 
491 	case Opt_share:
492 		if (!result.negated)
493 			opt->flags &= ~CEPH_OPT_NOSHARE;
494 		else
495 			opt->flags |= CEPH_OPT_NOSHARE;
496 		break;
497 	case Opt_crc:
498 		if (!result.negated)
499 			opt->flags &= ~CEPH_OPT_NOCRC;
500 		else
501 			opt->flags |= CEPH_OPT_NOCRC;
502 		break;
503 	case Opt_cephx_require_signatures:
504 		if (!result.negated)
505 			opt->flags &= ~CEPH_OPT_NOMSGAUTH;
506 		else
507 			opt->flags |= CEPH_OPT_NOMSGAUTH;
508 		break;
509 	case Opt_cephx_sign_messages:
510 		if (!result.negated)
511 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
512 		else
513 			opt->flags |= CEPH_OPT_NOMSGSIGN;
514 		break;
515 	case Opt_tcp_nodelay:
516 		if (!result.negated)
517 			opt->flags |= CEPH_OPT_TCP_NODELAY;
518 		else
519 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
520 		break;
521 
522 	case Opt_abort_on_full:
523 		opt->flags |= CEPH_OPT_ABORT_ON_FULL;
524 		break;
525 
526 	default:
527 		BUG();
528 	}
529 
530 	return 0;
531 
532 out_of_range:
533 	return invalf(fc, "libceph: %s out of range", param->key);
534 }
535 EXPORT_SYMBOL(ceph_parse_param);
536 
537 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
538 			      bool show_all)
539 {
540 	struct ceph_options *opt = client->options;
541 	size_t pos = m->count;
542 
543 	if (opt->name) {
544 		seq_puts(m, "name=");
545 		seq_escape(m, opt->name, ", \t\n\\");
546 		seq_putc(m, ',');
547 	}
548 	if (opt->key)
549 		seq_puts(m, "secret=<hidden>,");
550 
551 	if (opt->flags & CEPH_OPT_FSID)
552 		seq_printf(m, "fsid=%pU,", &opt->fsid);
553 	if (opt->flags & CEPH_OPT_NOSHARE)
554 		seq_puts(m, "noshare,");
555 	if (opt->flags & CEPH_OPT_NOCRC)
556 		seq_puts(m, "nocrc,");
557 	if (opt->flags & CEPH_OPT_NOMSGAUTH)
558 		seq_puts(m, "nocephx_require_signatures,");
559 	if (opt->flags & CEPH_OPT_NOMSGSIGN)
560 		seq_puts(m, "nocephx_sign_messages,");
561 	if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
562 		seq_puts(m, "notcp_nodelay,");
563 	if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
564 		seq_puts(m, "abort_on_full,");
565 
566 	if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
567 		seq_printf(m, "mount_timeout=%d,",
568 			   jiffies_to_msecs(opt->mount_timeout) / 1000);
569 	if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
570 		seq_printf(m, "osd_idle_ttl=%d,",
571 			   jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
572 	if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
573 		seq_printf(m, "osdkeepalivetimeout=%d,",
574 		    jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
575 	if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
576 		seq_printf(m, "osd_request_timeout=%d,",
577 			   jiffies_to_msecs(opt->osd_request_timeout) / 1000);
578 
579 	/* drop redundant comma */
580 	if (m->count != pos)
581 		m->count--;
582 
583 	return 0;
584 }
585 EXPORT_SYMBOL(ceph_print_client_options);
586 
587 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
588 {
589 	return &client->msgr.inst.addr;
590 }
591 EXPORT_SYMBOL(ceph_client_addr);
592 
593 u64 ceph_client_gid(struct ceph_client *client)
594 {
595 	return client->monc.auth->global_id;
596 }
597 EXPORT_SYMBOL(ceph_client_gid);
598 
599 /*
600  * create a fresh client instance
601  */
602 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
603 {
604 	struct ceph_client *client;
605 	struct ceph_entity_addr *myaddr = NULL;
606 	int err;
607 
608 	err = wait_for_random_bytes();
609 	if (err < 0)
610 		return ERR_PTR(err);
611 
612 	client = kzalloc(sizeof(*client), GFP_KERNEL);
613 	if (client == NULL)
614 		return ERR_PTR(-ENOMEM);
615 
616 	client->private = private;
617 	client->options = opt;
618 
619 	mutex_init(&client->mount_mutex);
620 	init_waitqueue_head(&client->auth_wq);
621 	client->auth_err = 0;
622 
623 	client->extra_mon_dispatch = NULL;
624 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
625 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
626 
627 	if (!ceph_test_opt(client, NOMSGAUTH))
628 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
629 
630 	/* msgr */
631 	if (ceph_test_opt(client, MYIP))
632 		myaddr = &client->options->my_addr;
633 
634 	ceph_messenger_init(&client->msgr, myaddr);
635 
636 	/* subsystems */
637 	err = ceph_monc_init(&client->monc, client);
638 	if (err < 0)
639 		goto fail;
640 	err = ceph_osdc_init(&client->osdc, client);
641 	if (err < 0)
642 		goto fail_monc;
643 
644 	return client;
645 
646 fail_monc:
647 	ceph_monc_stop(&client->monc);
648 fail:
649 	ceph_messenger_fini(&client->msgr);
650 	kfree(client);
651 	return ERR_PTR(err);
652 }
653 EXPORT_SYMBOL(ceph_create_client);
654 
655 void ceph_destroy_client(struct ceph_client *client)
656 {
657 	dout("destroy_client %p\n", client);
658 
659 	atomic_set(&client->msgr.stopping, 1);
660 
661 	/* unmount */
662 	ceph_osdc_stop(&client->osdc);
663 	ceph_monc_stop(&client->monc);
664 	ceph_messenger_fini(&client->msgr);
665 
666 	ceph_debugfs_client_cleanup(client);
667 
668 	ceph_destroy_options(client->options);
669 
670 	kfree(client);
671 	dout("destroy_client %p done\n", client);
672 }
673 EXPORT_SYMBOL(ceph_destroy_client);
674 
675 void ceph_reset_client_addr(struct ceph_client *client)
676 {
677 	ceph_messenger_reset_nonce(&client->msgr);
678 	ceph_monc_reopen_session(&client->monc);
679 	ceph_osdc_reopen_osds(&client->osdc);
680 }
681 EXPORT_SYMBOL(ceph_reset_client_addr);
682 
683 /*
684  * true if we have the mon map (and have thus joined the cluster)
685  */
686 static bool have_mon_and_osd_map(struct ceph_client *client)
687 {
688 	return client->monc.monmap && client->monc.monmap->epoch &&
689 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
690 }
691 
692 /*
693  * mount: join the ceph cluster, and open root directory.
694  */
695 int __ceph_open_session(struct ceph_client *client, unsigned long started)
696 {
697 	unsigned long timeout = client->options->mount_timeout;
698 	long err;
699 
700 	/* open session, and wait for mon and osd maps */
701 	err = ceph_monc_open_session(&client->monc);
702 	if (err < 0)
703 		return err;
704 
705 	while (!have_mon_and_osd_map(client)) {
706 		if (timeout && time_after_eq(jiffies, started + timeout))
707 			return -ETIMEDOUT;
708 
709 		/* wait */
710 		dout("mount waiting for mon_map\n");
711 		err = wait_event_interruptible_timeout(client->auth_wq,
712 			have_mon_and_osd_map(client) || (client->auth_err < 0),
713 			ceph_timeout_jiffies(timeout));
714 		if (err < 0)
715 			return err;
716 		if (client->auth_err < 0)
717 			return client->auth_err;
718 	}
719 
720 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
721 		&client->fsid);
722 	ceph_debugfs_client_init(client);
723 
724 	return 0;
725 }
726 EXPORT_SYMBOL(__ceph_open_session);
727 
728 int ceph_open_session(struct ceph_client *client)
729 {
730 	int ret;
731 	unsigned long started = jiffies;  /* note the start time */
732 
733 	dout("open_session start\n");
734 	mutex_lock(&client->mount_mutex);
735 
736 	ret = __ceph_open_session(client, started);
737 
738 	mutex_unlock(&client->mount_mutex);
739 	return ret;
740 }
741 EXPORT_SYMBOL(ceph_open_session);
742 
743 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
744 				unsigned long timeout)
745 {
746 	u64 newest_epoch;
747 	int ret;
748 
749 	ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
750 	if (ret)
751 		return ret;
752 
753 	if (client->osdc.osdmap->epoch >= newest_epoch)
754 		return 0;
755 
756 	ceph_osdc_maybe_request_map(&client->osdc);
757 	return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
758 }
759 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
760 
761 static int __init init_ceph_lib(void)
762 {
763 	int ret = 0;
764 
765 	ceph_debugfs_init();
766 
767 	ret = ceph_crypto_init();
768 	if (ret < 0)
769 		goto out_debugfs;
770 
771 	ret = ceph_msgr_init();
772 	if (ret < 0)
773 		goto out_crypto;
774 
775 	ret = ceph_osdc_setup();
776 	if (ret < 0)
777 		goto out_msgr;
778 
779 	pr_info("loaded (mon/osd proto %d/%d)\n",
780 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
781 
782 	return 0;
783 
784 out_msgr:
785 	ceph_msgr_exit();
786 out_crypto:
787 	ceph_crypto_shutdown();
788 out_debugfs:
789 	ceph_debugfs_cleanup();
790 	return ret;
791 }
792 
793 static void __exit exit_ceph_lib(void)
794 {
795 	dout("exit_ceph_lib\n");
796 	WARN_ON(!ceph_strings_empty());
797 
798 	ceph_osdc_cleanup();
799 	ceph_msgr_exit();
800 	ceph_crypto_shutdown();
801 	ceph_debugfs_cleanup();
802 }
803 
804 module_init(init_ceph_lib);
805 module_exit(exit_ceph_lib);
806 
807 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
808 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
809 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
810 MODULE_DESCRIPTION("Ceph core library");
811 MODULE_LICENSE("GPL");
812