xref: /linux/net/ceph/ceph_common.c (revision 69050f8d6d075dc01af7a5f2f550a8067510366f)
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_obj(*opt, GFP_KERNEL);
313 	if (!opt)
314 		return NULL;
315 
316 	opt->crush_locs = RB_ROOT;
317 	opt->mon_addr = kzalloc_objs(*opt->mon_addr, CEPH_MAX_MON, GFP_KERNEL);
318 	if (!opt->mon_addr) {
319 		kfree(opt);
320 		return NULL;
321 	}
322 
323 	opt->flags = CEPH_OPT_DEFAULT;
324 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
325 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
326 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
327 	opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
328 	opt->read_from_replica = CEPH_READ_FROM_REPLICA_DEFAULT;
329 	opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
330 	opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
331 	return opt;
332 }
333 EXPORT_SYMBOL(ceph_alloc_options);
334 
335 void ceph_destroy_options(struct ceph_options *opt)
336 {
337 	dout("destroy_options %p\n", opt);
338 	if (!opt)
339 		return;
340 
341 	ceph_clear_crush_locs(&opt->crush_locs);
342 	kfree(opt->name);
343 	if (opt->key) {
344 		ceph_crypto_key_destroy(opt->key);
345 		kfree(opt->key);
346 	}
347 	kfree(opt->mon_addr);
348 	kfree(opt);
349 }
350 EXPORT_SYMBOL(ceph_destroy_options);
351 
352 /* get secret from key store */
353 static int get_secret(struct ceph_crypto_key *dst, const char *name,
354 		      struct p_log *log)
355 {
356 	struct key *ukey;
357 	int key_err;
358 	int err = 0;
359 	struct ceph_crypto_key *ckey;
360 
361 	ukey = request_key(&key_type_ceph, name, NULL);
362 	if (IS_ERR(ukey)) {
363 		/* request_key errors don't map nicely to mount(2)
364 		   errors; don't even try, but still printk */
365 		key_err = PTR_ERR(ukey);
366 		switch (key_err) {
367 		case -ENOKEY:
368 			error_plog(log, "Failed due to key not found: %s",
369 			       name);
370 			break;
371 		case -EKEYEXPIRED:
372 			error_plog(log, "Failed due to expired key: %s",
373 			       name);
374 			break;
375 		case -EKEYREVOKED:
376 			error_plog(log, "Failed due to revoked key: %s",
377 			       name);
378 			break;
379 		default:
380 			error_plog(log, "Failed due to key error %d: %s",
381 			       key_err, name);
382 		}
383 		err = -EPERM;
384 		goto out;
385 	}
386 
387 	ckey = ukey->payload.data[0];
388 	err = ceph_crypto_key_clone(dst, ckey);
389 	if (err)
390 		goto out_key;
391 	/* pass through, err is 0 */
392 
393 out_key:
394 	key_put(ukey);
395 out:
396 	return err;
397 }
398 
399 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
400 		       struct fc_log *l, char delim)
401 {
402 	struct p_log log = {.prefix = "libceph", .log = l};
403 	int ret;
404 
405 	/* ip1[:port1][<delim>ip2[:port2]...] */
406 	ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
407 			     &opt->num_mon, delim);
408 	if (ret) {
409 		error_plog(&log, "Failed to parse monitor IPs: %d", ret);
410 		return ret;
411 	}
412 
413 	return 0;
414 }
415 EXPORT_SYMBOL(ceph_parse_mon_ips);
416 
417 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
418 		     struct fc_log *l)
419 {
420 	struct fs_parse_result result;
421 	int token, err;
422 	struct p_log log = {.prefix = "libceph", .log = l};
423 
424 	token = __fs_parse(&log, ceph_parameters, param, &result);
425 	dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
426 	if (token < 0)
427 		return token;
428 
429 	switch (token) {
430 	case Opt_ip:
431 		err = ceph_parse_ips(param->string,
432 				     param->string + param->size,
433 				     &opt->my_addr, 1, NULL, ',');
434 		if (err) {
435 			error_plog(&log, "Failed to parse ip: %d", err);
436 			return err;
437 		}
438 		opt->flags |= CEPH_OPT_MYIP;
439 		break;
440 
441 	case Opt_fsid:
442 		err = ceph_parse_fsid(param->string, &opt->fsid);
443 		if (err) {
444 			error_plog(&log, "Failed to parse fsid: %d", err);
445 			return err;
446 		}
447 		opt->flags |= CEPH_OPT_FSID;
448 		break;
449 	case Opt_name:
450 		kfree(opt->name);
451 		opt->name = param->string;
452 		param->string = NULL;
453 		break;
454 	case Opt_secret:
455 		ceph_crypto_key_destroy(opt->key);
456 		kfree(opt->key);
457 
458 		opt->key = kzalloc_obj(*opt->key, GFP_KERNEL);
459 		if (!opt->key)
460 			return -ENOMEM;
461 		err = ceph_crypto_key_unarmor(opt->key, param->string);
462 		if (err) {
463 			error_plog(&log, "Failed to parse secret: %d", err);
464 			return err;
465 		}
466 		break;
467 	case Opt_key:
468 		ceph_crypto_key_destroy(opt->key);
469 		kfree(opt->key);
470 
471 		opt->key = kzalloc_obj(*opt->key, GFP_KERNEL);
472 		if (!opt->key)
473 			return -ENOMEM;
474 		return get_secret(opt->key, param->string, &log);
475 	case Opt_crush_location:
476 		ceph_clear_crush_locs(&opt->crush_locs);
477 		err = ceph_parse_crush_location(param->string,
478 						&opt->crush_locs);
479 		if (err) {
480 			error_plog(&log, "Failed to parse CRUSH location: %d",
481 				   err);
482 			return err;
483 		}
484 		break;
485 	case Opt_read_from_replica:
486 		switch (result.uint_32) {
487 		case Opt_read_from_replica_no:
488 			opt->read_from_replica = 0;
489 			break;
490 		case Opt_read_from_replica_balance:
491 			opt->read_from_replica = CEPH_OSD_FLAG_BALANCE_READS;
492 			break;
493 		case Opt_read_from_replica_localize:
494 			opt->read_from_replica = CEPH_OSD_FLAG_LOCALIZE_READS;
495 			break;
496 		default:
497 			BUG();
498 		}
499 		break;
500 	case Opt_ms_mode:
501 		switch (result.uint_32) {
502 		case Opt_ms_mode_legacy:
503 			opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
504 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
505 			break;
506 		case Opt_ms_mode_crc:
507 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
508 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
509 			break;
510 		case Opt_ms_mode_secure:
511 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
512 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
513 			break;
514 		case Opt_ms_mode_prefer_crc:
515 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
516 			opt->con_modes[1] = CEPH_CON_MODE_SECURE;
517 			break;
518 		case Opt_ms_mode_prefer_secure:
519 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
520 			opt->con_modes[1] = CEPH_CON_MODE_CRC;
521 			break;
522 		default:
523 			BUG();
524 		}
525 		break;
526 
527 	case Opt_osdkeepalivetimeout:
528 		/* 0 isn't well defined right now, reject it */
529 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
530 			goto out_of_range;
531 		opt->osd_keepalive_timeout =
532 		    msecs_to_jiffies(result.uint_32 * 1000);
533 		break;
534 	case Opt_osd_idle_ttl:
535 		/* 0 isn't well defined right now, reject it */
536 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
537 			goto out_of_range;
538 		opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
539 		break;
540 	case Opt_mount_timeout:
541 		/* 0 is "wait forever" (i.e. infinite timeout) */
542 		if (result.uint_32 > INT_MAX / 1000)
543 			goto out_of_range;
544 		opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
545 		break;
546 	case Opt_osd_request_timeout:
547 		/* 0 is "wait forever" (i.e. infinite timeout) */
548 		if (result.uint_32 > INT_MAX / 1000)
549 			goto out_of_range;
550 		opt->osd_request_timeout =
551 		    msecs_to_jiffies(result.uint_32 * 1000);
552 		break;
553 
554 	case Opt_share:
555 		if (!result.negated)
556 			opt->flags &= ~CEPH_OPT_NOSHARE;
557 		else
558 			opt->flags |= CEPH_OPT_NOSHARE;
559 		break;
560 	case Opt_crc:
561 		if (!result.negated)
562 			opt->flags &= ~CEPH_OPT_NOCRC;
563 		else
564 			opt->flags |= CEPH_OPT_NOCRC;
565 		break;
566 	case Opt_cephx_require_signatures:
567 		if (!result.negated)
568 			warn_plog(&log, "Ignoring cephx_require_signatures");
569 		else
570 			warn_plog(&log, "Ignoring nocephx_require_signatures, use nocephx_sign_messages");
571 		break;
572 	case Opt_cephx_sign_messages:
573 		if (!result.negated)
574 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
575 		else
576 			opt->flags |= CEPH_OPT_NOMSGSIGN;
577 		break;
578 	case Opt_tcp_nodelay:
579 		if (!result.negated)
580 			opt->flags |= CEPH_OPT_TCP_NODELAY;
581 		else
582 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
583 		break;
584 
585 	case Opt_abort_on_full:
586 		opt->flags |= CEPH_OPT_ABORT_ON_FULL;
587 		break;
588 	case Opt_rxbounce:
589 		opt->flags |= CEPH_OPT_RXBOUNCE;
590 		break;
591 
592 	default:
593 		BUG();
594 	}
595 
596 	return 0;
597 
598 out_of_range:
599 	return inval_plog(&log, "%s out of range", param->key);
600 }
601 EXPORT_SYMBOL(ceph_parse_param);
602 
603 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
604 			      bool show_all)
605 {
606 	struct ceph_options *opt = client->options;
607 	size_t pos = m->count;
608 	struct rb_node *n;
609 
610 	if (opt->name) {
611 		seq_puts(m, "name=");
612 		seq_escape(m, opt->name, ", \t\n\\");
613 		seq_putc(m, ',');
614 	}
615 	if (opt->key)
616 		seq_puts(m, "secret=<hidden>,");
617 
618 	if (!RB_EMPTY_ROOT(&opt->crush_locs)) {
619 		seq_puts(m, "crush_location=");
620 		for (n = rb_first(&opt->crush_locs); ; ) {
621 			struct crush_loc_node *loc =
622 			    rb_entry(n, struct crush_loc_node, cl_node);
623 
624 			seq_printf(m, "%s:%s", loc->cl_loc.cl_type_name,
625 				   loc->cl_loc.cl_name);
626 			n = rb_next(n);
627 			if (!n)
628 				break;
629 
630 			seq_putc(m, '|');
631 		}
632 		seq_putc(m, ',');
633 	}
634 	if (opt->read_from_replica == CEPH_OSD_FLAG_BALANCE_READS) {
635 		seq_puts(m, "read_from_replica=balance,");
636 	} else if (opt->read_from_replica == CEPH_OSD_FLAG_LOCALIZE_READS) {
637 		seq_puts(m, "read_from_replica=localize,");
638 	}
639 	if (opt->con_modes[0] != CEPH_CON_MODE_UNKNOWN) {
640 		if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
641 		    opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
642 			seq_puts(m, "ms_mode=crc,");
643 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
644 			   opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
645 			seq_puts(m, "ms_mode=secure,");
646 		} else if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
647 			   opt->con_modes[1] == CEPH_CON_MODE_SECURE) {
648 			seq_puts(m, "ms_mode=prefer-crc,");
649 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
650 			   opt->con_modes[1] == CEPH_CON_MODE_CRC) {
651 			seq_puts(m, "ms_mode=prefer-secure,");
652 		}
653 	}
654 
655 	if (opt->flags & CEPH_OPT_FSID)
656 		seq_printf(m, "fsid=%pU,", &opt->fsid);
657 	if (opt->flags & CEPH_OPT_NOSHARE)
658 		seq_puts(m, "noshare,");
659 	if (opt->flags & CEPH_OPT_NOCRC)
660 		seq_puts(m, "nocrc,");
661 	if (opt->flags & CEPH_OPT_NOMSGSIGN)
662 		seq_puts(m, "nocephx_sign_messages,");
663 	if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
664 		seq_puts(m, "notcp_nodelay,");
665 	if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
666 		seq_puts(m, "abort_on_full,");
667 	if (opt->flags & CEPH_OPT_RXBOUNCE)
668 		seq_puts(m, "rxbounce,");
669 
670 	if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
671 		seq_printf(m, "mount_timeout=%d,",
672 			   jiffies_to_msecs(opt->mount_timeout) / 1000);
673 	if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
674 		seq_printf(m, "osd_idle_ttl=%d,",
675 			   jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
676 	if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
677 		seq_printf(m, "osdkeepalivetimeout=%d,",
678 		    jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
679 	if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
680 		seq_printf(m, "osd_request_timeout=%d,",
681 			   jiffies_to_msecs(opt->osd_request_timeout) / 1000);
682 
683 	/* drop redundant comma */
684 	if (m->count != pos)
685 		m->count--;
686 
687 	return 0;
688 }
689 EXPORT_SYMBOL(ceph_print_client_options);
690 
691 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
692 {
693 	return &client->msgr.inst.addr;
694 }
695 EXPORT_SYMBOL(ceph_client_addr);
696 
697 u64 ceph_client_gid(struct ceph_client *client)
698 {
699 	return client->monc.auth->global_id;
700 }
701 EXPORT_SYMBOL(ceph_client_gid);
702 
703 /*
704  * create a fresh client instance
705  */
706 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
707 {
708 	struct ceph_client *client;
709 	struct ceph_entity_addr *myaddr = NULL;
710 	int err;
711 
712 	err = wait_for_random_bytes();
713 	if (err < 0)
714 		return ERR_PTR(err);
715 
716 	client = kzalloc_obj(*client, GFP_KERNEL);
717 	if (client == NULL)
718 		return ERR_PTR(-ENOMEM);
719 
720 	client->private = private;
721 	client->options = opt;
722 
723 	mutex_init(&client->mount_mutex);
724 	init_waitqueue_head(&client->auth_wq);
725 	client->auth_err = 0;
726 
727 	client->extra_mon_dispatch = NULL;
728 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
729 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
730 
731 	if (!ceph_test_opt(client, NOMSGSIGN))
732 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
733 
734 	/* msgr */
735 	if (ceph_test_opt(client, MYIP))
736 		myaddr = &client->options->my_addr;
737 
738 	ceph_messenger_init(&client->msgr, myaddr);
739 
740 	/* subsystems */
741 	err = ceph_monc_init(&client->monc, client);
742 	if (err < 0)
743 		goto fail;
744 	err = ceph_osdc_init(&client->osdc, client);
745 	if (err < 0)
746 		goto fail_monc;
747 
748 	return client;
749 
750 fail_monc:
751 	ceph_monc_stop(&client->monc);
752 fail:
753 	ceph_messenger_fini(&client->msgr);
754 	kfree(client);
755 	return ERR_PTR(err);
756 }
757 EXPORT_SYMBOL(ceph_create_client);
758 
759 void ceph_destroy_client(struct ceph_client *client)
760 {
761 	dout("destroy_client %p\n", client);
762 
763 	atomic_set(&client->msgr.stopping, 1);
764 
765 	/* unmount */
766 	ceph_osdc_stop(&client->osdc);
767 	ceph_monc_stop(&client->monc);
768 	ceph_messenger_fini(&client->msgr);
769 
770 	ceph_debugfs_client_cleanup(client);
771 
772 	ceph_destroy_options(client->options);
773 
774 	kfree(client);
775 	dout("destroy_client %p done\n", client);
776 }
777 EXPORT_SYMBOL(ceph_destroy_client);
778 
779 void ceph_reset_client_addr(struct ceph_client *client)
780 {
781 	ceph_messenger_reset_nonce(&client->msgr);
782 	ceph_monc_reopen_session(&client->monc);
783 	ceph_osdc_reopen_osds(&client->osdc);
784 }
785 EXPORT_SYMBOL(ceph_reset_client_addr);
786 
787 /*
788  * mount: join the ceph cluster, and open root directory.
789  */
790 int __ceph_open_session(struct ceph_client *client)
791 {
792 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
793 	long timeout = ceph_timeout_jiffies(client->options->mount_timeout);
794 	bool have_monmap, have_osdmap;
795 	int err;
796 
797 	/* open session, and wait for mon and osd maps */
798 	err = ceph_monc_open_session(&client->monc);
799 	if (err < 0)
800 		return err;
801 
802 	add_wait_queue(&client->auth_wq, &wait);
803 	for (;;) {
804 		mutex_lock(&client->monc.mutex);
805 		err = client->auth_err;
806 		have_monmap = client->monc.monmap && client->monc.monmap->epoch;
807 		mutex_unlock(&client->monc.mutex);
808 
809 		down_read(&client->osdc.lock);
810 		have_osdmap = client->osdc.osdmap && client->osdc.osdmap->epoch;
811 		up_read(&client->osdc.lock);
812 
813 		if (err || (have_monmap && have_osdmap))
814 			break;
815 
816 		if (signal_pending(current)) {
817 			err = -ERESTARTSYS;
818 			break;
819 		}
820 
821 		if (!timeout) {
822 			err = -ETIMEDOUT;
823 			break;
824 		}
825 
826 		/* wait */
827 		dout("mount waiting for mon_map\n");
828 		timeout = wait_woken(&wait, TASK_INTERRUPTIBLE, timeout);
829 	}
830 	remove_wait_queue(&client->auth_wq, &wait);
831 
832 	if (err)
833 		return err;
834 
835 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
836 		&client->fsid);
837 	ceph_debugfs_client_init(client);
838 
839 	return 0;
840 }
841 EXPORT_SYMBOL(__ceph_open_session);
842 
843 int ceph_open_session(struct ceph_client *client)
844 {
845 	int ret;
846 
847 	dout("open_session start\n");
848 	mutex_lock(&client->mount_mutex);
849 
850 	ret = __ceph_open_session(client);
851 
852 	mutex_unlock(&client->mount_mutex);
853 	return ret;
854 }
855 EXPORT_SYMBOL(ceph_open_session);
856 
857 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
858 				unsigned long timeout)
859 {
860 	u64 newest_epoch;
861 	int ret;
862 
863 	ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
864 	if (ret)
865 		return ret;
866 
867 	if (client->osdc.osdmap->epoch >= newest_epoch)
868 		return 0;
869 
870 	ceph_osdc_maybe_request_map(&client->osdc);
871 	return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
872 }
873 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
874 
875 static int __init init_ceph_lib(void)
876 {
877 	int ret = 0;
878 
879 	ceph_debugfs_init();
880 
881 	ret = ceph_crypto_init();
882 	if (ret < 0)
883 		goto out_debugfs;
884 
885 	ret = ceph_msgr_init();
886 	if (ret < 0)
887 		goto out_crypto;
888 
889 	ret = ceph_osdc_setup();
890 	if (ret < 0)
891 		goto out_msgr;
892 
893 	pr_info("loaded (mon/osd proto %d/%d)\n",
894 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
895 
896 	return 0;
897 
898 out_msgr:
899 	ceph_msgr_exit();
900 out_crypto:
901 	ceph_crypto_shutdown();
902 out_debugfs:
903 	ceph_debugfs_cleanup();
904 	return ret;
905 }
906 
907 static void __exit exit_ceph_lib(void)
908 {
909 	dout("exit_ceph_lib\n");
910 	WARN_ON(!ceph_strings_empty());
911 
912 	ceph_osdc_cleanup();
913 	ceph_msgr_exit();
914 	ceph_crypto_shutdown();
915 	ceph_debugfs_cleanup();
916 }
917 
918 module_init(init_ceph_lib);
919 module_exit(exit_ceph_lib);
920 
921 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
922 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
923 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
924 MODULE_DESCRIPTION("Ceph core library");
925 MODULE_LICENSE("GPL");
926