xref: /linux/net/ceph/ceph_common.c (revision 87c9c16317882dd6dbbc07e349bc3223e14f3244)
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 /*
194  * kvmalloc() doesn't fall back to the vmalloc allocator unless flags are
195  * compatible with (a superset of) GFP_KERNEL.  This is because while the
196  * actual pages are allocated with the specified flags, the page table pages
197  * are always allocated with GFP_KERNEL.
198  *
199  * ceph_kvmalloc() may be called with GFP_KERNEL, GFP_NOFS or GFP_NOIO.
200  */
201 void *ceph_kvmalloc(size_t size, gfp_t flags)
202 {
203 	void *p;
204 
205 	if ((flags & (__GFP_IO | __GFP_FS)) == (__GFP_IO | __GFP_FS)) {
206 		p = kvmalloc(size, flags);
207 	} else if ((flags & (__GFP_IO | __GFP_FS)) == __GFP_IO) {
208 		unsigned int nofs_flag = memalloc_nofs_save();
209 		p = kvmalloc(size, GFP_KERNEL);
210 		memalloc_nofs_restore(nofs_flag);
211 	} else {
212 		unsigned int noio_flag = memalloc_noio_save();
213 		p = kvmalloc(size, GFP_KERNEL);
214 		memalloc_noio_restore(noio_flag);
215 	}
216 
217 	return p;
218 }
219 
220 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
221 {
222 	int i = 0;
223 	char tmp[3];
224 	int err = -EINVAL;
225 	int d;
226 
227 	dout("parse_fsid '%s'\n", str);
228 	tmp[2] = 0;
229 	while (*str && i < 16) {
230 		if (ispunct(*str)) {
231 			str++;
232 			continue;
233 		}
234 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
235 			break;
236 		tmp[0] = str[0];
237 		tmp[1] = str[1];
238 		if (sscanf(tmp, "%x", &d) < 1)
239 			break;
240 		fsid->fsid[i] = d & 0xff;
241 		i++;
242 		str += 2;
243 	}
244 
245 	if (i == 16)
246 		err = 0;
247 	dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
248 	return err;
249 }
250 
251 /*
252  * ceph options
253  */
254 enum {
255 	Opt_osdkeepalivetimeout,
256 	Opt_mount_timeout,
257 	Opt_osd_idle_ttl,
258 	Opt_osd_request_timeout,
259 	/* int args above */
260 	Opt_fsid,
261 	Opt_name,
262 	Opt_secret,
263 	Opt_key,
264 	Opt_ip,
265 	Opt_crush_location,
266 	Opt_read_from_replica,
267 	Opt_ms_mode,
268 	/* string args above */
269 	Opt_share,
270 	Opt_crc,
271 	Opt_cephx_require_signatures,
272 	Opt_cephx_sign_messages,
273 	Opt_tcp_nodelay,
274 	Opt_abort_on_full,
275 };
276 
277 enum {
278 	Opt_read_from_replica_no,
279 	Opt_read_from_replica_balance,
280 	Opt_read_from_replica_localize,
281 };
282 
283 static const struct constant_table ceph_param_read_from_replica[] = {
284 	{"no",		Opt_read_from_replica_no},
285 	{"balance",	Opt_read_from_replica_balance},
286 	{"localize",	Opt_read_from_replica_localize},
287 	{}
288 };
289 
290 enum ceph_ms_mode {
291 	Opt_ms_mode_legacy,
292 	Opt_ms_mode_crc,
293 	Opt_ms_mode_secure,
294 	Opt_ms_mode_prefer_crc,
295 	Opt_ms_mode_prefer_secure
296 };
297 
298 static const struct constant_table ceph_param_ms_mode[] = {
299 	{"legacy",		Opt_ms_mode_legacy},
300 	{"crc",			Opt_ms_mode_crc},
301 	{"secure",		Opt_ms_mode_secure},
302 	{"prefer-crc",		Opt_ms_mode_prefer_crc},
303 	{"prefer-secure",	Opt_ms_mode_prefer_secure},
304 	{}
305 };
306 
307 static const struct fs_parameter_spec ceph_parameters[] = {
308 	fsparam_flag	("abort_on_full",		Opt_abort_on_full),
309 	__fsparam	(NULL, "cephx_require_signatures", Opt_cephx_require_signatures,
310 			 fs_param_neg_with_no|fs_param_deprecated, NULL),
311 	fsparam_flag_no ("cephx_sign_messages",		Opt_cephx_sign_messages),
312 	fsparam_flag_no ("crc",				Opt_crc),
313 	fsparam_string	("crush_location",		Opt_crush_location),
314 	fsparam_string	("fsid",			Opt_fsid),
315 	fsparam_string	("ip",				Opt_ip),
316 	fsparam_string	("key",				Opt_key),
317 	fsparam_u32	("mount_timeout",		Opt_mount_timeout),
318 	fsparam_string	("name",			Opt_name),
319 	fsparam_u32	("osd_idle_ttl",		Opt_osd_idle_ttl),
320 	fsparam_u32	("osd_request_timeout",		Opt_osd_request_timeout),
321 	fsparam_u32	("osdkeepalive",		Opt_osdkeepalivetimeout),
322 	fsparam_enum	("read_from_replica",		Opt_read_from_replica,
323 			 ceph_param_read_from_replica),
324 	fsparam_enum	("ms_mode",			Opt_ms_mode,
325 			 ceph_param_ms_mode),
326 	fsparam_string	("secret",			Opt_secret),
327 	fsparam_flag_no ("share",			Opt_share),
328 	fsparam_flag_no ("tcp_nodelay",			Opt_tcp_nodelay),
329 	{}
330 };
331 
332 struct ceph_options *ceph_alloc_options(void)
333 {
334 	struct ceph_options *opt;
335 
336 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
337 	if (!opt)
338 		return NULL;
339 
340 	opt->crush_locs = RB_ROOT;
341 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
342 				GFP_KERNEL);
343 	if (!opt->mon_addr) {
344 		kfree(opt);
345 		return NULL;
346 	}
347 
348 	opt->flags = CEPH_OPT_DEFAULT;
349 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
350 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
351 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
352 	opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
353 	opt->read_from_replica = CEPH_READ_FROM_REPLICA_DEFAULT;
354 	opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
355 	opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
356 	return opt;
357 }
358 EXPORT_SYMBOL(ceph_alloc_options);
359 
360 void ceph_destroy_options(struct ceph_options *opt)
361 {
362 	dout("destroy_options %p\n", opt);
363 	if (!opt)
364 		return;
365 
366 	ceph_clear_crush_locs(&opt->crush_locs);
367 	kfree(opt->name);
368 	if (opt->key) {
369 		ceph_crypto_key_destroy(opt->key);
370 		kfree(opt->key);
371 	}
372 	kfree(opt->mon_addr);
373 	kfree(opt);
374 }
375 EXPORT_SYMBOL(ceph_destroy_options);
376 
377 /* get secret from key store */
378 static int get_secret(struct ceph_crypto_key *dst, const char *name,
379 		      struct p_log *log)
380 {
381 	struct key *ukey;
382 	int key_err;
383 	int err = 0;
384 	struct ceph_crypto_key *ckey;
385 
386 	ukey = request_key(&key_type_ceph, name, NULL);
387 	if (IS_ERR(ukey)) {
388 		/* request_key errors don't map nicely to mount(2)
389 		   errors; don't even try, but still printk */
390 		key_err = PTR_ERR(ukey);
391 		switch (key_err) {
392 		case -ENOKEY:
393 			error_plog(log, "Failed due to key not found: %s",
394 			       name);
395 			break;
396 		case -EKEYEXPIRED:
397 			error_plog(log, "Failed due to expired key: %s",
398 			       name);
399 			break;
400 		case -EKEYREVOKED:
401 			error_plog(log, "Failed due to revoked key: %s",
402 			       name);
403 			break;
404 		default:
405 			error_plog(log, "Failed due to key error %d: %s",
406 			       key_err, name);
407 		}
408 		err = -EPERM;
409 		goto out;
410 	}
411 
412 	ckey = ukey->payload.data[0];
413 	err = ceph_crypto_key_clone(dst, ckey);
414 	if (err)
415 		goto out_key;
416 	/* pass through, err is 0 */
417 
418 out_key:
419 	key_put(ukey);
420 out:
421 	return err;
422 }
423 
424 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
425 		       struct fc_log *l)
426 {
427 	struct p_log log = {.prefix = "libceph", .log = l};
428 	int ret;
429 
430 	/* ip1[:port1][,ip2[:port2]...] */
431 	ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
432 			     &opt->num_mon);
433 	if (ret) {
434 		error_plog(&log, "Failed to parse monitor IPs: %d", ret);
435 		return ret;
436 	}
437 
438 	return 0;
439 }
440 EXPORT_SYMBOL(ceph_parse_mon_ips);
441 
442 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
443 		     struct fc_log *l)
444 {
445 	struct fs_parse_result result;
446 	int token, err;
447 	struct p_log log = {.prefix = "libceph", .log = l};
448 
449 	token = __fs_parse(&log, ceph_parameters, param, &result);
450 	dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
451 	if (token < 0)
452 		return token;
453 
454 	switch (token) {
455 	case Opt_ip:
456 		err = ceph_parse_ips(param->string,
457 				     param->string + param->size,
458 				     &opt->my_addr,
459 				     1, NULL);
460 		if (err) {
461 			error_plog(&log, "Failed to parse ip: %d", err);
462 			return err;
463 		}
464 		opt->flags |= CEPH_OPT_MYIP;
465 		break;
466 
467 	case Opt_fsid:
468 		err = parse_fsid(param->string, &opt->fsid);
469 		if (err) {
470 			error_plog(&log, "Failed to parse fsid: %d", err);
471 			return err;
472 		}
473 		opt->flags |= CEPH_OPT_FSID;
474 		break;
475 	case Opt_name:
476 		kfree(opt->name);
477 		opt->name = param->string;
478 		param->string = NULL;
479 		break;
480 	case Opt_secret:
481 		ceph_crypto_key_destroy(opt->key);
482 		kfree(opt->key);
483 
484 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
485 		if (!opt->key)
486 			return -ENOMEM;
487 		err = ceph_crypto_key_unarmor(opt->key, param->string);
488 		if (err) {
489 			error_plog(&log, "Failed to parse secret: %d", err);
490 			return err;
491 		}
492 		break;
493 	case Opt_key:
494 		ceph_crypto_key_destroy(opt->key);
495 		kfree(opt->key);
496 
497 		opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
498 		if (!opt->key)
499 			return -ENOMEM;
500 		return get_secret(opt->key, param->string, &log);
501 	case Opt_crush_location:
502 		ceph_clear_crush_locs(&opt->crush_locs);
503 		err = ceph_parse_crush_location(param->string,
504 						&opt->crush_locs);
505 		if (err) {
506 			error_plog(&log, "Failed to parse CRUSH location: %d",
507 				   err);
508 			return err;
509 		}
510 		break;
511 	case Opt_read_from_replica:
512 		switch (result.uint_32) {
513 		case Opt_read_from_replica_no:
514 			opt->read_from_replica = 0;
515 			break;
516 		case Opt_read_from_replica_balance:
517 			opt->read_from_replica = CEPH_OSD_FLAG_BALANCE_READS;
518 			break;
519 		case Opt_read_from_replica_localize:
520 			opt->read_from_replica = CEPH_OSD_FLAG_LOCALIZE_READS;
521 			break;
522 		default:
523 			BUG();
524 		}
525 		break;
526 	case Opt_ms_mode:
527 		switch (result.uint_32) {
528 		case Opt_ms_mode_legacy:
529 			opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
530 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
531 			break;
532 		case Opt_ms_mode_crc:
533 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
534 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
535 			break;
536 		case Opt_ms_mode_secure:
537 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
538 			opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
539 			break;
540 		case Opt_ms_mode_prefer_crc:
541 			opt->con_modes[0] = CEPH_CON_MODE_CRC;
542 			opt->con_modes[1] = CEPH_CON_MODE_SECURE;
543 			break;
544 		case Opt_ms_mode_prefer_secure:
545 			opt->con_modes[0] = CEPH_CON_MODE_SECURE;
546 			opt->con_modes[1] = CEPH_CON_MODE_CRC;
547 			break;
548 		default:
549 			BUG();
550 		}
551 		break;
552 
553 	case Opt_osdkeepalivetimeout:
554 		/* 0 isn't well defined right now, reject it */
555 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
556 			goto out_of_range;
557 		opt->osd_keepalive_timeout =
558 		    msecs_to_jiffies(result.uint_32 * 1000);
559 		break;
560 	case Opt_osd_idle_ttl:
561 		/* 0 isn't well defined right now, reject it */
562 		if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
563 			goto out_of_range;
564 		opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
565 		break;
566 	case Opt_mount_timeout:
567 		/* 0 is "wait forever" (i.e. infinite timeout) */
568 		if (result.uint_32 > INT_MAX / 1000)
569 			goto out_of_range;
570 		opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
571 		break;
572 	case Opt_osd_request_timeout:
573 		/* 0 is "wait forever" (i.e. infinite timeout) */
574 		if (result.uint_32 > INT_MAX / 1000)
575 			goto out_of_range;
576 		opt->osd_request_timeout =
577 		    msecs_to_jiffies(result.uint_32 * 1000);
578 		break;
579 
580 	case Opt_share:
581 		if (!result.negated)
582 			opt->flags &= ~CEPH_OPT_NOSHARE;
583 		else
584 			opt->flags |= CEPH_OPT_NOSHARE;
585 		break;
586 	case Opt_crc:
587 		if (!result.negated)
588 			opt->flags &= ~CEPH_OPT_NOCRC;
589 		else
590 			opt->flags |= CEPH_OPT_NOCRC;
591 		break;
592 	case Opt_cephx_require_signatures:
593 		if (!result.negated)
594 			warn_plog(&log, "Ignoring cephx_require_signatures");
595 		else
596 			warn_plog(&log, "Ignoring nocephx_require_signatures, use nocephx_sign_messages");
597 		break;
598 	case Opt_cephx_sign_messages:
599 		if (!result.negated)
600 			opt->flags &= ~CEPH_OPT_NOMSGSIGN;
601 		else
602 			opt->flags |= CEPH_OPT_NOMSGSIGN;
603 		break;
604 	case Opt_tcp_nodelay:
605 		if (!result.negated)
606 			opt->flags |= CEPH_OPT_TCP_NODELAY;
607 		else
608 			opt->flags &= ~CEPH_OPT_TCP_NODELAY;
609 		break;
610 
611 	case Opt_abort_on_full:
612 		opt->flags |= CEPH_OPT_ABORT_ON_FULL;
613 		break;
614 
615 	default:
616 		BUG();
617 	}
618 
619 	return 0;
620 
621 out_of_range:
622 	return inval_plog(&log, "%s out of range", param->key);
623 }
624 EXPORT_SYMBOL(ceph_parse_param);
625 
626 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
627 			      bool show_all)
628 {
629 	struct ceph_options *opt = client->options;
630 	size_t pos = m->count;
631 	struct rb_node *n;
632 
633 	if (opt->name) {
634 		seq_puts(m, "name=");
635 		seq_escape(m, opt->name, ", \t\n\\");
636 		seq_putc(m, ',');
637 	}
638 	if (opt->key)
639 		seq_puts(m, "secret=<hidden>,");
640 
641 	if (!RB_EMPTY_ROOT(&opt->crush_locs)) {
642 		seq_puts(m, "crush_location=");
643 		for (n = rb_first(&opt->crush_locs); ; ) {
644 			struct crush_loc_node *loc =
645 			    rb_entry(n, struct crush_loc_node, cl_node);
646 
647 			seq_printf(m, "%s:%s", loc->cl_loc.cl_type_name,
648 				   loc->cl_loc.cl_name);
649 			n = rb_next(n);
650 			if (!n)
651 				break;
652 
653 			seq_putc(m, '|');
654 		}
655 		seq_putc(m, ',');
656 	}
657 	if (opt->read_from_replica == CEPH_OSD_FLAG_BALANCE_READS) {
658 		seq_puts(m, "read_from_replica=balance,");
659 	} else if (opt->read_from_replica == CEPH_OSD_FLAG_LOCALIZE_READS) {
660 		seq_puts(m, "read_from_replica=localize,");
661 	}
662 	if (opt->con_modes[0] != CEPH_CON_MODE_UNKNOWN) {
663 		if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
664 		    opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
665 			seq_puts(m, "ms_mode=crc,");
666 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
667 			   opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
668 			seq_puts(m, "ms_mode=secure,");
669 		} else if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
670 			   opt->con_modes[1] == CEPH_CON_MODE_SECURE) {
671 			seq_puts(m, "ms_mode=prefer-crc,");
672 		} else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
673 			   opt->con_modes[1] == CEPH_CON_MODE_CRC) {
674 			seq_puts(m, "ms_mode=prefer-secure,");
675 		}
676 	}
677 
678 	if (opt->flags & CEPH_OPT_FSID)
679 		seq_printf(m, "fsid=%pU,", &opt->fsid);
680 	if (opt->flags & CEPH_OPT_NOSHARE)
681 		seq_puts(m, "noshare,");
682 	if (opt->flags & CEPH_OPT_NOCRC)
683 		seq_puts(m, "nocrc,");
684 	if (opt->flags & CEPH_OPT_NOMSGSIGN)
685 		seq_puts(m, "nocephx_sign_messages,");
686 	if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
687 		seq_puts(m, "notcp_nodelay,");
688 	if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
689 		seq_puts(m, "abort_on_full,");
690 
691 	if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
692 		seq_printf(m, "mount_timeout=%d,",
693 			   jiffies_to_msecs(opt->mount_timeout) / 1000);
694 	if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
695 		seq_printf(m, "osd_idle_ttl=%d,",
696 			   jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
697 	if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
698 		seq_printf(m, "osdkeepalivetimeout=%d,",
699 		    jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
700 	if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
701 		seq_printf(m, "osd_request_timeout=%d,",
702 			   jiffies_to_msecs(opt->osd_request_timeout) / 1000);
703 
704 	/* drop redundant comma */
705 	if (m->count != pos)
706 		m->count--;
707 
708 	return 0;
709 }
710 EXPORT_SYMBOL(ceph_print_client_options);
711 
712 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
713 {
714 	return &client->msgr.inst.addr;
715 }
716 EXPORT_SYMBOL(ceph_client_addr);
717 
718 u64 ceph_client_gid(struct ceph_client *client)
719 {
720 	return client->monc.auth->global_id;
721 }
722 EXPORT_SYMBOL(ceph_client_gid);
723 
724 /*
725  * create a fresh client instance
726  */
727 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
728 {
729 	struct ceph_client *client;
730 	struct ceph_entity_addr *myaddr = NULL;
731 	int err;
732 
733 	err = wait_for_random_bytes();
734 	if (err < 0)
735 		return ERR_PTR(err);
736 
737 	client = kzalloc(sizeof(*client), GFP_KERNEL);
738 	if (client == NULL)
739 		return ERR_PTR(-ENOMEM);
740 
741 	client->private = private;
742 	client->options = opt;
743 
744 	mutex_init(&client->mount_mutex);
745 	init_waitqueue_head(&client->auth_wq);
746 	client->auth_err = 0;
747 
748 	client->extra_mon_dispatch = NULL;
749 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
750 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
751 
752 	if (!ceph_test_opt(client, NOMSGSIGN))
753 		client->required_features |= CEPH_FEATURE_MSG_AUTH;
754 
755 	/* msgr */
756 	if (ceph_test_opt(client, MYIP))
757 		myaddr = &client->options->my_addr;
758 
759 	ceph_messenger_init(&client->msgr, myaddr);
760 
761 	/* subsystems */
762 	err = ceph_monc_init(&client->monc, client);
763 	if (err < 0)
764 		goto fail;
765 	err = ceph_osdc_init(&client->osdc, client);
766 	if (err < 0)
767 		goto fail_monc;
768 
769 	return client;
770 
771 fail_monc:
772 	ceph_monc_stop(&client->monc);
773 fail:
774 	ceph_messenger_fini(&client->msgr);
775 	kfree(client);
776 	return ERR_PTR(err);
777 }
778 EXPORT_SYMBOL(ceph_create_client);
779 
780 void ceph_destroy_client(struct ceph_client *client)
781 {
782 	dout("destroy_client %p\n", client);
783 
784 	atomic_set(&client->msgr.stopping, 1);
785 
786 	/* unmount */
787 	ceph_osdc_stop(&client->osdc);
788 	ceph_monc_stop(&client->monc);
789 	ceph_messenger_fini(&client->msgr);
790 
791 	ceph_debugfs_client_cleanup(client);
792 
793 	ceph_destroy_options(client->options);
794 
795 	kfree(client);
796 	dout("destroy_client %p done\n", client);
797 }
798 EXPORT_SYMBOL(ceph_destroy_client);
799 
800 void ceph_reset_client_addr(struct ceph_client *client)
801 {
802 	ceph_messenger_reset_nonce(&client->msgr);
803 	ceph_monc_reopen_session(&client->monc);
804 	ceph_osdc_reopen_osds(&client->osdc);
805 }
806 EXPORT_SYMBOL(ceph_reset_client_addr);
807 
808 /*
809  * true if we have the mon map (and have thus joined the cluster)
810  */
811 static bool have_mon_and_osd_map(struct ceph_client *client)
812 {
813 	return client->monc.monmap && client->monc.monmap->epoch &&
814 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
815 }
816 
817 /*
818  * mount: join the ceph cluster, and open root directory.
819  */
820 int __ceph_open_session(struct ceph_client *client, unsigned long started)
821 {
822 	unsigned long timeout = client->options->mount_timeout;
823 	long err;
824 
825 	/* open session, and wait for mon and osd maps */
826 	err = ceph_monc_open_session(&client->monc);
827 	if (err < 0)
828 		return err;
829 
830 	while (!have_mon_and_osd_map(client)) {
831 		if (timeout && time_after_eq(jiffies, started + timeout))
832 			return -ETIMEDOUT;
833 
834 		/* wait */
835 		dout("mount waiting for mon_map\n");
836 		err = wait_event_interruptible_timeout(client->auth_wq,
837 			have_mon_and_osd_map(client) || (client->auth_err < 0),
838 			ceph_timeout_jiffies(timeout));
839 		if (err < 0)
840 			return err;
841 		if (client->auth_err < 0)
842 			return client->auth_err;
843 	}
844 
845 	pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
846 		&client->fsid);
847 	ceph_debugfs_client_init(client);
848 
849 	return 0;
850 }
851 EXPORT_SYMBOL(__ceph_open_session);
852 
853 int ceph_open_session(struct ceph_client *client)
854 {
855 	int ret;
856 	unsigned long started = jiffies;  /* note the start time */
857 
858 	dout("open_session start\n");
859 	mutex_lock(&client->mount_mutex);
860 
861 	ret = __ceph_open_session(client, started);
862 
863 	mutex_unlock(&client->mount_mutex);
864 	return ret;
865 }
866 EXPORT_SYMBOL(ceph_open_session);
867 
868 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
869 				unsigned long timeout)
870 {
871 	u64 newest_epoch;
872 	int ret;
873 
874 	ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
875 	if (ret)
876 		return ret;
877 
878 	if (client->osdc.osdmap->epoch >= newest_epoch)
879 		return 0;
880 
881 	ceph_osdc_maybe_request_map(&client->osdc);
882 	return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
883 }
884 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
885 
886 static int __init init_ceph_lib(void)
887 {
888 	int ret = 0;
889 
890 	ceph_debugfs_init();
891 
892 	ret = ceph_crypto_init();
893 	if (ret < 0)
894 		goto out_debugfs;
895 
896 	ret = ceph_msgr_init();
897 	if (ret < 0)
898 		goto out_crypto;
899 
900 	ret = ceph_osdc_setup();
901 	if (ret < 0)
902 		goto out_msgr;
903 
904 	pr_info("loaded (mon/osd proto %d/%d)\n",
905 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
906 
907 	return 0;
908 
909 out_msgr:
910 	ceph_msgr_exit();
911 out_crypto:
912 	ceph_crypto_shutdown();
913 out_debugfs:
914 	ceph_debugfs_cleanup();
915 	return ret;
916 }
917 
918 static void __exit exit_ceph_lib(void)
919 {
920 	dout("exit_ceph_lib\n");
921 	WARN_ON(!ceph_strings_empty());
922 
923 	ceph_osdc_cleanup();
924 	ceph_msgr_exit();
925 	ceph_crypto_shutdown();
926 	ceph_debugfs_cleanup();
927 }
928 
929 module_init(init_ceph_lib);
930 module_exit(exit_ceph_lib);
931 
932 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
933 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
934 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
935 MODULE_DESCRIPTION("Ceph core library");
936 MODULE_LICENSE("GPL");
937