xref: /openbmc/linux/net/ceph/ceph_common.c (revision 6e10e219)
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 static int 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("parse_fsid '%s'\n", 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("parse_fsid ret %d got fsid %pU\n", err, fsid);
221 	return err;
222 }
223 
224 /*
225  * ceph options
226  */
227 enum {
228 	Opt_osdkeepalivetimeout,
229 	Opt_mount_timeout,
230 	Opt_osd_idle_ttl,
231 	Opt_osd_request_timeout,
232 	/* int args above */
233 	Opt_fsid,
234 	Opt_name,
235 	Opt_secret,
236 	Opt_key,
237 	Opt_ip,
238 	Opt_crush_location,
239 	Opt_read_from_replica,
240 	Opt_ms_mode,
241 	/* string args above */
242 	Opt_share,
243 	Opt_crc,
244 	Opt_cephx_require_signatures,
245 	Opt_cephx_sign_messages,
246 	Opt_tcp_nodelay,
247 	Opt_abort_on_full,
248 };
249 
250 enum {
251 	Opt_read_from_replica_no,
252 	Opt_read_from_replica_balance,
253 	Opt_read_from_replica_localize,
254 };
255 
256 static const struct constant_table ceph_param_read_from_replica[] = {
257 	{"no",		Opt_read_from_replica_no},
258 	{"balance",	Opt_read_from_replica_balance},
259 	{"localize",	Opt_read_from_replica_localize},
260 	{}
261 };
262 
263 enum ceph_ms_mode {
264 	Opt_ms_mode_legacy,
265 	Opt_ms_mode_crc,
266 	Opt_ms_mode_secure,
267 	Opt_ms_mode_prefer_crc,
268 	Opt_ms_mode_prefer_secure
269 };
270 
271 static const struct constant_table ceph_param_ms_mode[] = {
272 	{"legacy",		Opt_ms_mode_legacy},
273 	{"crc",			Opt_ms_mode_crc},
274 	{"secure",		Opt_ms_mode_secure},
275 	{"prefer-crc",		Opt_ms_mode_prefer_crc},
276 	{"prefer-secure",	Opt_ms_mode_prefer_secure},
277 	{}
278 };
279 
280 static const struct fs_parameter_spec ceph_parameters[] = {
281 	fsparam_flag	("abort_on_full",		Opt_abort_on_full),
282 	__fsparam	(NULL, "cephx_require_signatures", Opt_cephx_require_signatures,
283 			 fs_param_neg_with_no|fs_param_deprecated, NULL),
284 	fsparam_flag_no ("cephx_sign_messages",		Opt_cephx_sign_messages),
285 	fsparam_flag_no ("crc",				Opt_crc),
286 	fsparam_string	("crush_location",		Opt_crush_location),
287 	fsparam_string	("fsid",			Opt_fsid),
288 	fsparam_string	("ip",				Opt_ip),
289 	fsparam_string	("key",				Opt_key),
290 	fsparam_u32	("mount_timeout",		Opt_mount_timeout),
291 	fsparam_string	("name",			Opt_name),
292 	fsparam_u32	("osd_idle_ttl",		Opt_osd_idle_ttl),
293 	fsparam_u32	("osd_request_timeout",		Opt_osd_request_timeout),
294 	fsparam_u32	("osdkeepalive",		Opt_osdkeepalivetimeout),
295 	fsparam_enum	("read_from_replica",		Opt_read_from_replica,
296 			 ceph_param_read_from_replica),
297 	fsparam_enum	("ms_mode",			Opt_ms_mode,
298 			 ceph_param_ms_mode),
299 	fsparam_string	("secret",			Opt_secret),
300 	fsparam_flag_no ("share",			Opt_share),
301 	fsparam_flag_no ("tcp_nodelay",			Opt_tcp_nodelay),
302 	{}
303 };
304 
305 struct ceph_options *ceph_alloc_options(void)
306 {
307 	struct ceph_options *opt;
308 
309 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
310 	if (!opt)
311 		return NULL;
312 
313 	opt->crush_locs = RB_ROOT;
314 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
315 				GFP_KERNEL);
316 	if (!opt->mon_addr) {
317 		kfree(opt);
318 		return NULL;
319 	}
320 
321 	opt->flags = CEPH_OPT_DEFAULT;
322 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
323 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
324 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
325 	opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
326 	opt->read_from_replica = CEPH_READ_FROM_REPLICA_DEFAULT;
327 	opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
328 	opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
329 	return opt;
330 }
331 EXPORT_SYMBOL(ceph_alloc_options);
332 
333 void ceph_destroy_options(struct ceph_options *opt)
334 {
335 	dout("destroy_options %p\n", opt);
336 	if (!opt)
337 		return;
338 
339 	ceph_clear_crush_locs(&opt->crush_locs);
340 	kfree(opt->name);
341 	if (opt->key) {
342 		ceph_crypto_key_destroy(opt->key);
343 		kfree(opt->key);
344 	}
345 	kfree(opt->mon_addr);
346 	kfree(opt);
347 }
348 EXPORT_SYMBOL(ceph_destroy_options);
349 
350 /* get secret from key store */
351 static int get_secret(struct ceph_crypto_key *dst, const char *name,
352 		      struct p_log *log)
353 {
354 	struct key *ukey;
355 	int key_err;
356 	int err = 0;
357 	struct ceph_crypto_key *ckey;
358 
359 	ukey = request_key(&key_type_ceph, name, NULL);
360 	if (IS_ERR(ukey)) {
361 		/* request_key errors don't map nicely to mount(2)
362 		   errors; don't even try, but still printk */
363 		key_err = PTR_ERR(ukey);
364 		switch (key_err) {
365 		case -ENOKEY:
366 			error_plog(log, "Failed due to key not found: %s",
367 			       name);
368 			break;
369 		case -EKEYEXPIRED:
370 			error_plog(log, "Failed due to expired key: %s",
371 			       name);
372 			break;
373 		case -EKEYREVOKED:
374 			error_plog(log, "Failed due to revoked key: %s",
375 			       name);
376 			break;
377 		default:
378 			error_plog(log, "Failed due to key error %d: %s",
379 			       key_err, name);
380 		}
381 		err = -EPERM;
382 		goto out;
383 	}
384 
385 	ckey = ukey->payload.data[0];
386 	err = ceph_crypto_key_clone(dst, ckey);
387 	if (err)
388 		goto out_key;
389 	/* pass through, err is 0 */
390 
391 out_key:
392 	key_put(ukey);
393 out:
394 	return err;
395 }
396 
397 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
398 		       struct fc_log *l)
399 {
400 	struct p_log log = {.prefix = "libceph", .log = l};
401 	int ret;
402 
403 	/* ip1[:port1][,ip2[:port2]...] */
404 	ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
405 			     &opt->num_mon);
406 	if (ret) {
407 		error_plog(&log, "Failed to parse monitor IPs: %d", ret);
408 		return ret;
409 	}
410 
411 	return 0;
412 }
413 EXPORT_SYMBOL(ceph_parse_mon_ips);
414 
415 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
416 		     struct fc_log *l)
417 {
418 	struct fs_parse_result result;
419 	int token, err;
420 	struct p_log log = {.prefix = "libceph", .log = l};
421 
422 	token = __fs_parse(&log, ceph_parameters, param, &result);
423 	dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
424 	if (token < 0)
425 		return token;
426 
427 	switch (token) {
428 	case Opt_ip:
429 		err = ceph_parse_ips(param->string,
430 				     param->string + param->size,
431 				     &opt->my_addr,
432 				     1, NULL);
433 		if (err) {
434 			error_plog(&log, "Failed to parse ip: %d", err);
435 			return err;
436 		}
437 		opt->flags |= CEPH_OPT_MYIP;
438 		break;
439 
440 	case Opt_fsid:
441 		err = parse_fsid(param->string, &opt->fsid);
442 		if (err) {
443 			error_plog(&log, "Failed to parse fsid: %d", err);
444 			return err;
445 		}
446 		opt->flags |= CEPH_OPT_FSID;
447 		break;
448 	case Opt_name:
449 		kfree(opt->name);
450 		opt->name = param->string;
451 		param->string = NULL;
452 		break;
453 	case Opt_secret:
454 		ceph_crypto_key_destroy(opt->key);
455 		kfree(opt->key);
456 
457 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
458 		if (!opt->key)
459 			return -ENOMEM;
460 		err = ceph_crypto_key_unarmor(opt->key, param->string);
461 		if (err) {
462 			error_plog(&log, "Failed to parse secret: %d", err);
463 			return err;
464 		}
465 		break;
466 	case Opt_key:
467 		ceph_crypto_key_destroy(opt->key);
468 		kfree(opt->key);
469 
470 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
471 		if (!opt->key)
472 			return -ENOMEM;
473 		return get_secret(opt->key, param->string, &log);
474 	case Opt_crush_location:
475 		ceph_clear_crush_locs(&opt->crush_locs);
476 		err = ceph_parse_crush_location(param->string,
477 						&opt->crush_locs);
478 		if (err) {
479 			error_plog(&log, "Failed to parse CRUSH location: %d",
480 				   err);
481 			return err;
482 		}
483 		break;
484 	case Opt_read_from_replica:
485 		switch (result.uint_32) {
486 		case Opt_read_from_replica_no:
487 			opt->read_from_replica = 0;
488 			break;
489 		case Opt_read_from_replica_balance:
490 			opt->read_from_replica = CEPH_OSD_FLAG_BALANCE_READS;
491 			break;
492 		case Opt_read_from_replica_localize:
493 			opt->read_from_replica = CEPH_OSD_FLAG_LOCALIZE_READS;
494 			break;
495 		default:
496 			BUG();
497 		}
498 		break;
499 	case Opt_ms_mode:
500 		switch (result.uint_32) {
501 		case Opt_ms_mode_legacy:
502 			opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
503 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
504 			break;
505 		case Opt_ms_mode_crc:
506 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
507 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
508 			break;
509 		case Opt_ms_mode_secure:
510 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
511 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
512 			break;
513 		case Opt_ms_mode_prefer_crc:
514 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
515 			opt->con_modes[1] = CEPH_CON_MODE_SECURE;
516 			break;
517 		case Opt_ms_mode_prefer_secure:
518 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
519 			opt->con_modes[1] = CEPH_CON_MODE_CRC;
520 			break;
521 		default:
522 			BUG();
523 		}
524 		break;
525 
526 	case Opt_osdkeepalivetimeout:
527 		/* 0 isn't well defined right now, reject it */
528 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
529 			goto out_of_range;
530 		opt->osd_keepalive_timeout =
531 		    msecs_to_jiffies(result.uint_32 * 1000);
532 		break;
533 	case Opt_osd_idle_ttl:
534 		/* 0 isn't well defined right now, reject it */
535 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
536 			goto out_of_range;
537 		opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
538 		break;
539 	case Opt_mount_timeout:
540 		/* 0 is "wait forever" (i.e. infinite timeout) */
541 		if (result.uint_32 > INT_MAX / 1000)
542 			goto out_of_range;
543 		opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
544 		break;
545 	case Opt_osd_request_timeout:
546 		/* 0 is "wait forever" (i.e. infinite timeout) */
547 		if (result.uint_32 > INT_MAX / 1000)
548 			goto out_of_range;
549 		opt->osd_request_timeout =
550 		    msecs_to_jiffies(result.uint_32 * 1000);
551 		break;
552 
553 	case Opt_share:
554 		if (!result.negated)
555 			opt->flags &= ~CEPH_OPT_NOSHARE;
556 		else
557 			opt->flags |= CEPH_OPT_NOSHARE;
558 		break;
559 	case Opt_crc:
560 		if (!result.negated)
561 			opt->flags &= ~CEPH_OPT_NOCRC;
562 		else
563 			opt->flags |= CEPH_OPT_NOCRC;
564 		break;
565 	case Opt_cephx_require_signatures:
566 		if (!result.negated)
567 			warn_plog(&log, "Ignoring cephx_require_signatures");
568 		else
569 			warn_plog(&log, "Ignoring nocephx_require_signatures, use nocephx_sign_messages");
570 		break;
571 	case Opt_cephx_sign_messages:
572 		if (!result.negated)
573 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
574 		else
575 			opt->flags |= CEPH_OPT_NOMSGSIGN;
576 		break;
577 	case Opt_tcp_nodelay:
578 		if (!result.negated)
579 			opt->flags |= CEPH_OPT_TCP_NODELAY;
580 		else
581 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
582 		break;
583 
584 	case Opt_abort_on_full:
585 		opt->flags |= CEPH_OPT_ABORT_ON_FULL;
586 		break;
587 
588 	default:
589 		BUG();
590 	}
591 
592 	return 0;
593 
594 out_of_range:
595 	return inval_plog(&log, "%s out of range", param->key);
596 }
597 EXPORT_SYMBOL(ceph_parse_param);
598 
599 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
600 			      bool show_all)
601 {
602 	struct ceph_options *opt = client->options;
603 	size_t pos = m->count;
604 	struct rb_node *n;
605 
606 	if (opt->name) {
607 		seq_puts(m, "name=");
608 		seq_escape(m, opt->name, ", \t\n\\");
609 		seq_putc(m, ',');
610 	}
611 	if (opt->key)
612 		seq_puts(m, "secret=<hidden>,");
613 
614 	if (!RB_EMPTY_ROOT(&opt->crush_locs)) {
615 		seq_puts(m, "crush_location=");
616 		for (n = rb_first(&opt->crush_locs); ; ) {
617 			struct crush_loc_node *loc =
618 			    rb_entry(n, struct crush_loc_node, cl_node);
619 
620 			seq_printf(m, "%s:%s", loc->cl_loc.cl_type_name,
621 				   loc->cl_loc.cl_name);
622 			n = rb_next(n);
623 			if (!n)
624 				break;
625 
626 			seq_putc(m, '|');
627 		}
628 		seq_putc(m, ',');
629 	}
630 	if (opt->read_from_replica == CEPH_OSD_FLAG_BALANCE_READS) {
631 		seq_puts(m, "read_from_replica=balance,");
632 	} else if (opt->read_from_replica == CEPH_OSD_FLAG_LOCALIZE_READS) {
633 		seq_puts(m, "read_from_replica=localize,");
634 	}
635 	if (opt->con_modes[0] != CEPH_CON_MODE_UNKNOWN) {
636 		if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
637 		    opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
638 			seq_puts(m, "ms_mode=crc,");
639 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
640 			   opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
641 			seq_puts(m, "ms_mode=secure,");
642 		} else if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
643 			   opt->con_modes[1] == CEPH_CON_MODE_SECURE) {
644 			seq_puts(m, "ms_mode=prefer-crc,");
645 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
646 			   opt->con_modes[1] == CEPH_CON_MODE_CRC) {
647 			seq_puts(m, "ms_mode=prefer-secure,");
648 		}
649 	}
650 
651 	if (opt->flags & CEPH_OPT_FSID)
652 		seq_printf(m, "fsid=%pU,", &opt->fsid);
653 	if (opt->flags & CEPH_OPT_NOSHARE)
654 		seq_puts(m, "noshare,");
655 	if (opt->flags & CEPH_OPT_NOCRC)
656 		seq_puts(m, "nocrc,");
657 	if (opt->flags & CEPH_OPT_NOMSGSIGN)
658 		seq_puts(m, "nocephx_sign_messages,");
659 	if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
660 		seq_puts(m, "notcp_nodelay,");
661 	if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
662 		seq_puts(m, "abort_on_full,");
663 
664 	if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
665 		seq_printf(m, "mount_timeout=%d,",
666 			   jiffies_to_msecs(opt->mount_timeout) / 1000);
667 	if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
668 		seq_printf(m, "osd_idle_ttl=%d,",
669 			   jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
670 	if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
671 		seq_printf(m, "osdkeepalivetimeout=%d,",
672 		    jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
673 	if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
674 		seq_printf(m, "osd_request_timeout=%d,",
675 			   jiffies_to_msecs(opt->osd_request_timeout) / 1000);
676 
677 	/* drop redundant comma */
678 	if (m->count != pos)
679 		m->count--;
680 
681 	return 0;
682 }
683 EXPORT_SYMBOL(ceph_print_client_options);
684 
685 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
686 {
687 	return &client->msgr.inst.addr;
688 }
689 EXPORT_SYMBOL(ceph_client_addr);
690 
691 u64 ceph_client_gid(struct ceph_client *client)
692 {
693 	return client->monc.auth->global_id;
694 }
695 EXPORT_SYMBOL(ceph_client_gid);
696 
697 /*
698  * create a fresh client instance
699  */
700 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
701 {
702 	struct ceph_client *client;
703 	struct ceph_entity_addr *myaddr = NULL;
704 	int err;
705 
706 	err = wait_for_random_bytes();
707 	if (err < 0)
708 		return ERR_PTR(err);
709 
710 	client = kzalloc(sizeof(*client), GFP_KERNEL);
711 	if (client == NULL)
712 		return ERR_PTR(-ENOMEM);
713 
714 	client->private = private;
715 	client->options = opt;
716 
717 	mutex_init(&client->mount_mutex);
718 	init_waitqueue_head(&client->auth_wq);
719 	client->auth_err = 0;
720 
721 	client->extra_mon_dispatch = NULL;
722 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
723 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
724 
725 	if (!ceph_test_opt(client, NOMSGSIGN))
726 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
727 
728 	/* msgr */
729 	if (ceph_test_opt(client, MYIP))
730 		myaddr = &client->options->my_addr;
731 
732 	ceph_messenger_init(&client->msgr, myaddr);
733 
734 	/* subsystems */
735 	err = ceph_monc_init(&client->monc, client);
736 	if (err < 0)
737 		goto fail;
738 	err = ceph_osdc_init(&client->osdc, client);
739 	if (err < 0)
740 		goto fail_monc;
741 
742 	return client;
743 
744 fail_monc:
745 	ceph_monc_stop(&client->monc);
746 fail:
747 	ceph_messenger_fini(&client->msgr);
748 	kfree(client);
749 	return ERR_PTR(err);
750 }
751 EXPORT_SYMBOL(ceph_create_client);
752 
753 void ceph_destroy_client(struct ceph_client *client)
754 {
755 	dout("destroy_client %p\n", client);
756 
757 	atomic_set(&client->msgr.stopping, 1);
758 
759 	/* unmount */
760 	ceph_osdc_stop(&client->osdc);
761 	ceph_monc_stop(&client->monc);
762 	ceph_messenger_fini(&client->msgr);
763 
764 	ceph_debugfs_client_cleanup(client);
765 
766 	ceph_destroy_options(client->options);
767 
768 	kfree(client);
769 	dout("destroy_client %p done\n", client);
770 }
771 EXPORT_SYMBOL(ceph_destroy_client);
772 
773 void ceph_reset_client_addr(struct ceph_client *client)
774 {
775 	ceph_messenger_reset_nonce(&client->msgr);
776 	ceph_monc_reopen_session(&client->monc);
777 	ceph_osdc_reopen_osds(&client->osdc);
778 }
779 EXPORT_SYMBOL(ceph_reset_client_addr);
780 
781 /*
782  * true if we have the mon map (and have thus joined the cluster)
783  */
784 static bool have_mon_and_osd_map(struct ceph_client *client)
785 {
786 	return client->monc.monmap && client->monc.monmap->epoch &&
787 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
788 }
789 
790 /*
791  * mount: join the ceph cluster, and open root directory.
792  */
793 int __ceph_open_session(struct ceph_client *client, unsigned long started)
794 {
795 	unsigned long timeout = client->options->mount_timeout;
796 	long err;
797 
798 	/* open session, and wait for mon and osd maps */
799 	err = ceph_monc_open_session(&client->monc);
800 	if (err < 0)
801 		return err;
802 
803 	while (!have_mon_and_osd_map(client)) {
804 		if (timeout && time_after_eq(jiffies, started + timeout))
805 			return -ETIMEDOUT;
806 
807 		/* wait */
808 		dout("mount waiting for mon_map\n");
809 		err = wait_event_interruptible_timeout(client->auth_wq,
810 			have_mon_and_osd_map(client) || (client->auth_err < 0),
811 			ceph_timeout_jiffies(timeout));
812 		if (err < 0)
813 			return err;
814 		if (client->auth_err < 0)
815 			return client->auth_err;
816 	}
817 
818 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
819 		&client->fsid);
820 	ceph_debugfs_client_init(client);
821 
822 	return 0;
823 }
824 EXPORT_SYMBOL(__ceph_open_session);
825 
826 int ceph_open_session(struct ceph_client *client)
827 {
828 	int ret;
829 	unsigned long started = jiffies;  /* note the start time */
830 
831 	dout("open_session start\n");
832 	mutex_lock(&client->mount_mutex);
833 
834 	ret = __ceph_open_session(client, started);
835 
836 	mutex_unlock(&client->mount_mutex);
837 	return ret;
838 }
839 EXPORT_SYMBOL(ceph_open_session);
840 
841 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
842 				unsigned long timeout)
843 {
844 	u64 newest_epoch;
845 	int ret;
846 
847 	ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
848 	if (ret)
849 		return ret;
850 
851 	if (client->osdc.osdmap->epoch >= newest_epoch)
852 		return 0;
853 
854 	ceph_osdc_maybe_request_map(&client->osdc);
855 	return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
856 }
857 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
858 
859 static int __init init_ceph_lib(void)
860 {
861 	int ret = 0;
862 
863 	ceph_debugfs_init();
864 
865 	ret = ceph_crypto_init();
866 	if (ret < 0)
867 		goto out_debugfs;
868 
869 	ret = ceph_msgr_init();
870 	if (ret < 0)
871 		goto out_crypto;
872 
873 	ret = ceph_osdc_setup();
874 	if (ret < 0)
875 		goto out_msgr;
876 
877 	pr_info("loaded (mon/osd proto %d/%d)\n",
878 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
879 
880 	return 0;
881 
882 out_msgr:
883 	ceph_msgr_exit();
884 out_crypto:
885 	ceph_crypto_shutdown();
886 out_debugfs:
887 	ceph_debugfs_cleanup();
888 	return ret;
889 }
890 
891 static void __exit exit_ceph_lib(void)
892 {
893 	dout("exit_ceph_lib\n");
894 	WARN_ON(!ceph_strings_empty());
895 
896 	ceph_osdc_cleanup();
897 	ceph_msgr_exit();
898 	ceph_crypto_shutdown();
899 	ceph_debugfs_cleanup();
900 }
901 
902 module_init(init_ceph_lib);
903 module_exit(exit_ceph_lib);
904 
905 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
906 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
907 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
908 MODULE_DESCRIPTION("Ceph core library");
909 MODULE_LICENSE("GPL");
910