xref: /openbmc/linux/net/ceph/ceph_common.c (revision bbb774d9)
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 	ret = ceph_compare_crush_locs(&opt1->crush_locs, &opt2->crush_locs);
180 	if (ret)
181 		return ret;
182 
183 	/* any matching mon ip implies a match */
184 	for (i = 0; i < opt1->num_mon; i++) {
185 		if (ceph_monmap_contains(client->monc.monmap,
186 				 &opt1->mon_addr[i]))
187 			return 0;
188 	}
189 	return -1;
190 }
191 EXPORT_SYMBOL(ceph_compare_options);
192 
193 int ceph_parse_fsid(const char *str, struct ceph_fsid *fsid)
194 {
195 	int i = 0;
196 	char tmp[3];
197 	int err = -EINVAL;
198 	int d;
199 
200 	dout("%s '%s'\n", __func__, str);
201 	tmp[2] = 0;
202 	while (*str && i < 16) {
203 		if (ispunct(*str)) {
204 			str++;
205 			continue;
206 		}
207 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
208 			break;
209 		tmp[0] = str[0];
210 		tmp[1] = str[1];
211 		if (sscanf(tmp, "%x", &d) < 1)
212 			break;
213 		fsid->fsid[i] = d & 0xff;
214 		i++;
215 		str += 2;
216 	}
217 
218 	if (i == 16)
219 		err = 0;
220 	dout("%s ret %d got fsid %pU\n", __func__, err, fsid);
221 	return err;
222 }
223 EXPORT_SYMBOL(ceph_parse_fsid);
224 
225 /*
226  * ceph options
227  */
228 enum {
229 	Opt_osdkeepalivetimeout,
230 	Opt_mount_timeout,
231 	Opt_osd_idle_ttl,
232 	Opt_osd_request_timeout,
233 	/* int args above */
234 	Opt_fsid,
235 	Opt_name,
236 	Opt_secret,
237 	Opt_key,
238 	Opt_ip,
239 	Opt_crush_location,
240 	Opt_read_from_replica,
241 	Opt_ms_mode,
242 	/* string args above */
243 	Opt_share,
244 	Opt_crc,
245 	Opt_cephx_require_signatures,
246 	Opt_cephx_sign_messages,
247 	Opt_tcp_nodelay,
248 	Opt_abort_on_full,
249 	Opt_rxbounce,
250 };
251 
252 enum {
253 	Opt_read_from_replica_no,
254 	Opt_read_from_replica_balance,
255 	Opt_read_from_replica_localize,
256 };
257 
258 static const struct constant_table ceph_param_read_from_replica[] = {
259 	{"no",		Opt_read_from_replica_no},
260 	{"balance",	Opt_read_from_replica_balance},
261 	{"localize",	Opt_read_from_replica_localize},
262 	{}
263 };
264 
265 enum ceph_ms_mode {
266 	Opt_ms_mode_legacy,
267 	Opt_ms_mode_crc,
268 	Opt_ms_mode_secure,
269 	Opt_ms_mode_prefer_crc,
270 	Opt_ms_mode_prefer_secure
271 };
272 
273 static const struct constant_table ceph_param_ms_mode[] = {
274 	{"legacy",		Opt_ms_mode_legacy},
275 	{"crc",			Opt_ms_mode_crc},
276 	{"secure",		Opt_ms_mode_secure},
277 	{"prefer-crc",		Opt_ms_mode_prefer_crc},
278 	{"prefer-secure",	Opt_ms_mode_prefer_secure},
279 	{}
280 };
281 
282 static const struct fs_parameter_spec ceph_parameters[] = {
283 	fsparam_flag	("abort_on_full",		Opt_abort_on_full),
284 	__fsparam	(NULL, "cephx_require_signatures", Opt_cephx_require_signatures,
285 			 fs_param_neg_with_no|fs_param_deprecated, NULL),
286 	fsparam_flag_no ("cephx_sign_messages",		Opt_cephx_sign_messages),
287 	fsparam_flag_no ("crc",				Opt_crc),
288 	fsparam_string	("crush_location",		Opt_crush_location),
289 	fsparam_string	("fsid",			Opt_fsid),
290 	fsparam_string	("ip",				Opt_ip),
291 	fsparam_string	("key",				Opt_key),
292 	fsparam_u32	("mount_timeout",		Opt_mount_timeout),
293 	fsparam_string	("name",			Opt_name),
294 	fsparam_u32	("osd_idle_ttl",		Opt_osd_idle_ttl),
295 	fsparam_u32	("osd_request_timeout",		Opt_osd_request_timeout),
296 	fsparam_u32	("osdkeepalive",		Opt_osdkeepalivetimeout),
297 	fsparam_enum	("read_from_replica",		Opt_read_from_replica,
298 			 ceph_param_read_from_replica),
299 	fsparam_flag	("rxbounce",			Opt_rxbounce),
300 	fsparam_enum	("ms_mode",			Opt_ms_mode,
301 			 ceph_param_ms_mode),
302 	fsparam_string	("secret",			Opt_secret),
303 	fsparam_flag_no ("share",			Opt_share),
304 	fsparam_flag_no ("tcp_nodelay",			Opt_tcp_nodelay),
305 	{}
306 };
307 
308 struct ceph_options *ceph_alloc_options(void)
309 {
310 	struct ceph_options *opt;
311 
312 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
313 	if (!opt)
314 		return NULL;
315 
316 	opt->crush_locs = RB_ROOT;
317 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
318 				GFP_KERNEL);
319 	if (!opt->mon_addr) {
320 		kfree(opt);
321 		return NULL;
322 	}
323 
324 	opt->flags = CEPH_OPT_DEFAULT;
325 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
326 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
327 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
328 	opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
329 	opt->read_from_replica = CEPH_READ_FROM_REPLICA_DEFAULT;
330 	opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
331 	opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
332 	return opt;
333 }
334 EXPORT_SYMBOL(ceph_alloc_options);
335 
336 void ceph_destroy_options(struct ceph_options *opt)
337 {
338 	dout("destroy_options %p\n", opt);
339 	if (!opt)
340 		return;
341 
342 	ceph_clear_crush_locs(&opt->crush_locs);
343 	kfree(opt->name);
344 	if (opt->key) {
345 		ceph_crypto_key_destroy(opt->key);
346 		kfree(opt->key);
347 	}
348 	kfree(opt->mon_addr);
349 	kfree(opt);
350 }
351 EXPORT_SYMBOL(ceph_destroy_options);
352 
353 /* get secret from key store */
354 static int get_secret(struct ceph_crypto_key *dst, const char *name,
355 		      struct p_log *log)
356 {
357 	struct key *ukey;
358 	int key_err;
359 	int err = 0;
360 	struct ceph_crypto_key *ckey;
361 
362 	ukey = request_key(&key_type_ceph, name, NULL);
363 	if (IS_ERR(ukey)) {
364 		/* request_key errors don't map nicely to mount(2)
365 		   errors; don't even try, but still printk */
366 		key_err = PTR_ERR(ukey);
367 		switch (key_err) {
368 		case -ENOKEY:
369 			error_plog(log, "Failed due to key not found: %s",
370 			       name);
371 			break;
372 		case -EKEYEXPIRED:
373 			error_plog(log, "Failed due to expired key: %s",
374 			       name);
375 			break;
376 		case -EKEYREVOKED:
377 			error_plog(log, "Failed due to revoked key: %s",
378 			       name);
379 			break;
380 		default:
381 			error_plog(log, "Failed due to key error %d: %s",
382 			       key_err, name);
383 		}
384 		err = -EPERM;
385 		goto out;
386 	}
387 
388 	ckey = ukey->payload.data[0];
389 	err = ceph_crypto_key_clone(dst, ckey);
390 	if (err)
391 		goto out_key;
392 	/* pass through, err is 0 */
393 
394 out_key:
395 	key_put(ukey);
396 out:
397 	return err;
398 }
399 
400 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
401 		       struct fc_log *l, char delim)
402 {
403 	struct p_log log = {.prefix = "libceph", .log = l};
404 	int ret;
405 
406 	/* ip1[:port1][<delim>ip2[:port2]...] */
407 	ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
408 			     &opt->num_mon, delim);
409 	if (ret) {
410 		error_plog(&log, "Failed to parse monitor IPs: %d", ret);
411 		return ret;
412 	}
413 
414 	return 0;
415 }
416 EXPORT_SYMBOL(ceph_parse_mon_ips);
417 
418 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
419 		     struct fc_log *l)
420 {
421 	struct fs_parse_result result;
422 	int token, err;
423 	struct p_log log = {.prefix = "libceph", .log = l};
424 
425 	token = __fs_parse(&log, ceph_parameters, param, &result);
426 	dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
427 	if (token < 0)
428 		return token;
429 
430 	switch (token) {
431 	case Opt_ip:
432 		err = ceph_parse_ips(param->string,
433 				     param->string + param->size,
434 				     &opt->my_addr, 1, NULL, ',');
435 		if (err) {
436 			error_plog(&log, "Failed to parse ip: %d", err);
437 			return err;
438 		}
439 		opt->flags |= CEPH_OPT_MYIP;
440 		break;
441 
442 	case Opt_fsid:
443 		err = ceph_parse_fsid(param->string, &opt->fsid);
444 		if (err) {
445 			error_plog(&log, "Failed to parse fsid: %d", err);
446 			return err;
447 		}
448 		opt->flags |= CEPH_OPT_FSID;
449 		break;
450 	case Opt_name:
451 		kfree(opt->name);
452 		opt->name = param->string;
453 		param->string = NULL;
454 		break;
455 	case Opt_secret:
456 		ceph_crypto_key_destroy(opt->key);
457 		kfree(opt->key);
458 
459 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
460 		if (!opt->key)
461 			return -ENOMEM;
462 		err = ceph_crypto_key_unarmor(opt->key, param->string);
463 		if (err) {
464 			error_plog(&log, "Failed to parse secret: %d", err);
465 			return err;
466 		}
467 		break;
468 	case Opt_key:
469 		ceph_crypto_key_destroy(opt->key);
470 		kfree(opt->key);
471 
472 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
473 		if (!opt->key)
474 			return -ENOMEM;
475 		return get_secret(opt->key, param->string, &log);
476 	case Opt_crush_location:
477 		ceph_clear_crush_locs(&opt->crush_locs);
478 		err = ceph_parse_crush_location(param->string,
479 						&opt->crush_locs);
480 		if (err) {
481 			error_plog(&log, "Failed to parse CRUSH location: %d",
482 				   err);
483 			return err;
484 		}
485 		break;
486 	case Opt_read_from_replica:
487 		switch (result.uint_32) {
488 		case Opt_read_from_replica_no:
489 			opt->read_from_replica = 0;
490 			break;
491 		case Opt_read_from_replica_balance:
492 			opt->read_from_replica = CEPH_OSD_FLAG_BALANCE_READS;
493 			break;
494 		case Opt_read_from_replica_localize:
495 			opt->read_from_replica = CEPH_OSD_FLAG_LOCALIZE_READS;
496 			break;
497 		default:
498 			BUG();
499 		}
500 		break;
501 	case Opt_ms_mode:
502 		switch (result.uint_32) {
503 		case Opt_ms_mode_legacy:
504 			opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
505 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
506 			break;
507 		case Opt_ms_mode_crc:
508 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
509 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
510 			break;
511 		case Opt_ms_mode_secure:
512 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
513 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
514 			break;
515 		case Opt_ms_mode_prefer_crc:
516 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
517 			opt->con_modes[1] = CEPH_CON_MODE_SECURE;
518 			break;
519 		case Opt_ms_mode_prefer_secure:
520 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
521 			opt->con_modes[1] = CEPH_CON_MODE_CRC;
522 			break;
523 		default:
524 			BUG();
525 		}
526 		break;
527 
528 	case Opt_osdkeepalivetimeout:
529 		/* 0 isn't well defined right now, reject it */
530 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
531 			goto out_of_range;
532 		opt->osd_keepalive_timeout =
533 		    msecs_to_jiffies(result.uint_32 * 1000);
534 		break;
535 	case Opt_osd_idle_ttl:
536 		/* 0 isn't well defined right now, reject it */
537 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
538 			goto out_of_range;
539 		opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
540 		break;
541 	case Opt_mount_timeout:
542 		/* 0 is "wait forever" (i.e. infinite timeout) */
543 		if (result.uint_32 > INT_MAX / 1000)
544 			goto out_of_range;
545 		opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
546 		break;
547 	case Opt_osd_request_timeout:
548 		/* 0 is "wait forever" (i.e. infinite timeout) */
549 		if (result.uint_32 > INT_MAX / 1000)
550 			goto out_of_range;
551 		opt->osd_request_timeout =
552 		    msecs_to_jiffies(result.uint_32 * 1000);
553 		break;
554 
555 	case Opt_share:
556 		if (!result.negated)
557 			opt->flags &= ~CEPH_OPT_NOSHARE;
558 		else
559 			opt->flags |= CEPH_OPT_NOSHARE;
560 		break;
561 	case Opt_crc:
562 		if (!result.negated)
563 			opt->flags &= ~CEPH_OPT_NOCRC;
564 		else
565 			opt->flags |= CEPH_OPT_NOCRC;
566 		break;
567 	case Opt_cephx_require_signatures:
568 		if (!result.negated)
569 			warn_plog(&log, "Ignoring cephx_require_signatures");
570 		else
571 			warn_plog(&log, "Ignoring nocephx_require_signatures, use nocephx_sign_messages");
572 		break;
573 	case Opt_cephx_sign_messages:
574 		if (!result.negated)
575 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
576 		else
577 			opt->flags |= CEPH_OPT_NOMSGSIGN;
578 		break;
579 	case Opt_tcp_nodelay:
580 		if (!result.negated)
581 			opt->flags |= CEPH_OPT_TCP_NODELAY;
582 		else
583 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
584 		break;
585 
586 	case Opt_abort_on_full:
587 		opt->flags |= CEPH_OPT_ABORT_ON_FULL;
588 		break;
589 	case Opt_rxbounce:
590 		opt->flags |= CEPH_OPT_RXBOUNCE;
591 		break;
592 
593 	default:
594 		BUG();
595 	}
596 
597 	return 0;
598 
599 out_of_range:
600 	return inval_plog(&log, "%s out of range", param->key);
601 }
602 EXPORT_SYMBOL(ceph_parse_param);
603 
604 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
605 			      bool show_all)
606 {
607 	struct ceph_options *opt = client->options;
608 	size_t pos = m->count;
609 	struct rb_node *n;
610 
611 	if (opt->name) {
612 		seq_puts(m, "name=");
613 		seq_escape(m, opt->name, ", \t\n\\");
614 		seq_putc(m, ',');
615 	}
616 	if (opt->key)
617 		seq_puts(m, "secret=<hidden>,");
618 
619 	if (!RB_EMPTY_ROOT(&opt->crush_locs)) {
620 		seq_puts(m, "crush_location=");
621 		for (n = rb_first(&opt->crush_locs); ; ) {
622 			struct crush_loc_node *loc =
623 			    rb_entry(n, struct crush_loc_node, cl_node);
624 
625 			seq_printf(m, "%s:%s", loc->cl_loc.cl_type_name,
626 				   loc->cl_loc.cl_name);
627 			n = rb_next(n);
628 			if (!n)
629 				break;
630 
631 			seq_putc(m, '|');
632 		}
633 		seq_putc(m, ',');
634 	}
635 	if (opt->read_from_replica == CEPH_OSD_FLAG_BALANCE_READS) {
636 		seq_puts(m, "read_from_replica=balance,");
637 	} else if (opt->read_from_replica == CEPH_OSD_FLAG_LOCALIZE_READS) {
638 		seq_puts(m, "read_from_replica=localize,");
639 	}
640 	if (opt->con_modes[0] != CEPH_CON_MODE_UNKNOWN) {
641 		if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
642 		    opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
643 			seq_puts(m, "ms_mode=crc,");
644 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
645 			   opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
646 			seq_puts(m, "ms_mode=secure,");
647 		} else if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
648 			   opt->con_modes[1] == CEPH_CON_MODE_SECURE) {
649 			seq_puts(m, "ms_mode=prefer-crc,");
650 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
651 			   opt->con_modes[1] == CEPH_CON_MODE_CRC) {
652 			seq_puts(m, "ms_mode=prefer-secure,");
653 		}
654 	}
655 
656 	if (opt->flags & CEPH_OPT_FSID)
657 		seq_printf(m, "fsid=%pU,", &opt->fsid);
658 	if (opt->flags & CEPH_OPT_NOSHARE)
659 		seq_puts(m, "noshare,");
660 	if (opt->flags & CEPH_OPT_NOCRC)
661 		seq_puts(m, "nocrc,");
662 	if (opt->flags & CEPH_OPT_NOMSGSIGN)
663 		seq_puts(m, "nocephx_sign_messages,");
664 	if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
665 		seq_puts(m, "notcp_nodelay,");
666 	if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
667 		seq_puts(m, "abort_on_full,");
668 	if (opt->flags & CEPH_OPT_RXBOUNCE)
669 		seq_puts(m, "rxbounce,");
670 
671 	if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
672 		seq_printf(m, "mount_timeout=%d,",
673 			   jiffies_to_msecs(opt->mount_timeout) / 1000);
674 	if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
675 		seq_printf(m, "osd_idle_ttl=%d,",
676 			   jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
677 	if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
678 		seq_printf(m, "osdkeepalivetimeout=%d,",
679 		    jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
680 	if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
681 		seq_printf(m, "osd_request_timeout=%d,",
682 			   jiffies_to_msecs(opt->osd_request_timeout) / 1000);
683 
684 	/* drop redundant comma */
685 	if (m->count != pos)
686 		m->count--;
687 
688 	return 0;
689 }
690 EXPORT_SYMBOL(ceph_print_client_options);
691 
692 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
693 {
694 	return &client->msgr.inst.addr;
695 }
696 EXPORT_SYMBOL(ceph_client_addr);
697 
698 u64 ceph_client_gid(struct ceph_client *client)
699 {
700 	return client->monc.auth->global_id;
701 }
702 EXPORT_SYMBOL(ceph_client_gid);
703 
704 /*
705  * create a fresh client instance
706  */
707 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
708 {
709 	struct ceph_client *client;
710 	struct ceph_entity_addr *myaddr = NULL;
711 	int err;
712 
713 	err = wait_for_random_bytes();
714 	if (err < 0)
715 		return ERR_PTR(err);
716 
717 	client = kzalloc(sizeof(*client), GFP_KERNEL);
718 	if (client == NULL)
719 		return ERR_PTR(-ENOMEM);
720 
721 	client->private = private;
722 	client->options = opt;
723 
724 	mutex_init(&client->mount_mutex);
725 	init_waitqueue_head(&client->auth_wq);
726 	client->auth_err = 0;
727 
728 	client->extra_mon_dispatch = NULL;
729 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
730 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
731 
732 	if (!ceph_test_opt(client, NOMSGSIGN))
733 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
734 
735 	/* msgr */
736 	if (ceph_test_opt(client, MYIP))
737 		myaddr = &client->options->my_addr;
738 
739 	ceph_messenger_init(&client->msgr, myaddr);
740 
741 	/* subsystems */
742 	err = ceph_monc_init(&client->monc, client);
743 	if (err < 0)
744 		goto fail;
745 	err = ceph_osdc_init(&client->osdc, client);
746 	if (err < 0)
747 		goto fail_monc;
748 
749 	return client;
750 
751 fail_monc:
752 	ceph_monc_stop(&client->monc);
753 fail:
754 	ceph_messenger_fini(&client->msgr);
755 	kfree(client);
756 	return ERR_PTR(err);
757 }
758 EXPORT_SYMBOL(ceph_create_client);
759 
760 void ceph_destroy_client(struct ceph_client *client)
761 {
762 	dout("destroy_client %p\n", client);
763 
764 	atomic_set(&client->msgr.stopping, 1);
765 
766 	/* unmount */
767 	ceph_osdc_stop(&client->osdc);
768 	ceph_monc_stop(&client->monc);
769 	ceph_messenger_fini(&client->msgr);
770 
771 	ceph_debugfs_client_cleanup(client);
772 
773 	ceph_destroy_options(client->options);
774 
775 	kfree(client);
776 	dout("destroy_client %p done\n", client);
777 }
778 EXPORT_SYMBOL(ceph_destroy_client);
779 
780 void ceph_reset_client_addr(struct ceph_client *client)
781 {
782 	ceph_messenger_reset_nonce(&client->msgr);
783 	ceph_monc_reopen_session(&client->monc);
784 	ceph_osdc_reopen_osds(&client->osdc);
785 }
786 EXPORT_SYMBOL(ceph_reset_client_addr);
787 
788 /*
789  * true if we have the mon map (and have thus joined the cluster)
790  */
791 static bool have_mon_and_osd_map(struct ceph_client *client)
792 {
793 	return client->monc.monmap && client->monc.monmap->epoch &&
794 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
795 }
796 
797 /*
798  * mount: join the ceph cluster, and open root directory.
799  */
800 int __ceph_open_session(struct ceph_client *client, unsigned long started)
801 {
802 	unsigned long timeout = client->options->mount_timeout;
803 	long err;
804 
805 	/* open session, and wait for mon and osd maps */
806 	err = ceph_monc_open_session(&client->monc);
807 	if (err < 0)
808 		return err;
809 
810 	while (!have_mon_and_osd_map(client)) {
811 		if (timeout && time_after_eq(jiffies, started + timeout))
812 			return -ETIMEDOUT;
813 
814 		/* wait */
815 		dout("mount waiting for mon_map\n");
816 		err = wait_event_interruptible_timeout(client->auth_wq,
817 			have_mon_and_osd_map(client) || (client->auth_err < 0),
818 			ceph_timeout_jiffies(timeout));
819 		if (err < 0)
820 			return err;
821 		if (client->auth_err < 0)
822 			return client->auth_err;
823 	}
824 
825 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
826 		&client->fsid);
827 	ceph_debugfs_client_init(client);
828 
829 	return 0;
830 }
831 EXPORT_SYMBOL(__ceph_open_session);
832 
833 int ceph_open_session(struct ceph_client *client)
834 {
835 	int ret;
836 	unsigned long started = jiffies;  /* note the start time */
837 
838 	dout("open_session start\n");
839 	mutex_lock(&client->mount_mutex);
840 
841 	ret = __ceph_open_session(client, started);
842 
843 	mutex_unlock(&client->mount_mutex);
844 	return ret;
845 }
846 EXPORT_SYMBOL(ceph_open_session);
847 
848 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
849 				unsigned long timeout)
850 {
851 	u64 newest_epoch;
852 	int ret;
853 
854 	ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
855 	if (ret)
856 		return ret;
857 
858 	if (client->osdc.osdmap->epoch >= newest_epoch)
859 		return 0;
860 
861 	ceph_osdc_maybe_request_map(&client->osdc);
862 	return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
863 }
864 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
865 
866 static int __init init_ceph_lib(void)
867 {
868 	int ret = 0;
869 
870 	ceph_debugfs_init();
871 
872 	ret = ceph_crypto_init();
873 	if (ret < 0)
874 		goto out_debugfs;
875 
876 	ret = ceph_msgr_init();
877 	if (ret < 0)
878 		goto out_crypto;
879 
880 	ret = ceph_osdc_setup();
881 	if (ret < 0)
882 		goto out_msgr;
883 
884 	pr_info("loaded (mon/osd proto %d/%d)\n",
885 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
886 
887 	return 0;
888 
889 out_msgr:
890 	ceph_msgr_exit();
891 out_crypto:
892 	ceph_crypto_shutdown();
893 out_debugfs:
894 	ceph_debugfs_cleanup();
895 	return ret;
896 }
897 
898 static void __exit exit_ceph_lib(void)
899 {
900 	dout("exit_ceph_lib\n");
901 	WARN_ON(!ceph_strings_empty());
902 
903 	ceph_osdc_cleanup();
904 	ceph_msgr_exit();
905 	ceph_crypto_shutdown();
906 	ceph_debugfs_cleanup();
907 }
908 
909 module_init(init_ceph_lib);
910 module_exit(exit_ceph_lib);
911 
912 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
913 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
914 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
915 MODULE_DESCRIPTION("Ceph core library");
916 MODULE_LICENSE("GPL");
917