xref: /linux/net/ceph/ceph_common.c (revision 615f2e5c531bc57d5a190f321d697988e950ae4d)
1 
2 #include <linux/ceph/ceph_debug.h>
3 #include <linux/backing-dev.h>
4 #include <linux/ctype.h>
5 #include <linux/fs.h>
6 #include <linux/inet.h>
7 #include <linux/in6.h>
8 #include <linux/key.h>
9 #include <keys/ceph-type.h>
10 #include <linux/module.h>
11 #include <linux/mount.h>
12 #include <linux/parser.h>
13 #include <linux/sched.h>
14 #include <linux/seq_file.h>
15 #include <linux/slab.h>
16 #include <linux/statfs.h>
17 #include <linux/string.h>
18 #include <linux/nsproxy.h>
19 #include <net/net_namespace.h>
20 
21 
22 #include <linux/ceph/ceph_features.h>
23 #include <linux/ceph/libceph.h>
24 #include <linux/ceph/debugfs.h>
25 #include <linux/ceph/decode.h>
26 #include <linux/ceph/mon_client.h>
27 #include <linux/ceph/auth.h>
28 #include "crypto.h"
29 
30 
31 
32 /*
33  * find filename portion of a path (/foo/bar/baz -> baz)
34  */
35 const char *ceph_file_part(const char *s, int len)
36 {
37 	const char *e = s + len;
38 
39 	while (e != s && *(e-1) != '/')
40 		e--;
41 	return e;
42 }
43 EXPORT_SYMBOL(ceph_file_part);
44 
45 const char *ceph_msg_type_name(int type)
46 {
47 	switch (type) {
48 	case CEPH_MSG_SHUTDOWN: return "shutdown";
49 	case CEPH_MSG_PING: return "ping";
50 	case CEPH_MSG_AUTH: return "auth";
51 	case CEPH_MSG_AUTH_REPLY: return "auth_reply";
52 	case CEPH_MSG_MON_MAP: return "mon_map";
53 	case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
54 	case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
55 	case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
56 	case CEPH_MSG_STATFS: return "statfs";
57 	case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
58 	case CEPH_MSG_MDS_MAP: return "mds_map";
59 	case CEPH_MSG_CLIENT_SESSION: return "client_session";
60 	case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
61 	case CEPH_MSG_CLIENT_REQUEST: return "client_request";
62 	case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
63 	case CEPH_MSG_CLIENT_REPLY: return "client_reply";
64 	case CEPH_MSG_CLIENT_CAPS: return "client_caps";
65 	case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
66 	case CEPH_MSG_CLIENT_SNAP: return "client_snap";
67 	case CEPH_MSG_CLIENT_LEASE: return "client_lease";
68 	case CEPH_MSG_OSD_MAP: return "osd_map";
69 	case CEPH_MSG_OSD_OP: return "osd_op";
70 	case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
71 	case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
72 	default: return "unknown";
73 	}
74 }
75 EXPORT_SYMBOL(ceph_msg_type_name);
76 
77 /*
78  * Initially learn our fsid, or verify an fsid matches.
79  */
80 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
81 {
82 	if (client->have_fsid) {
83 		if (ceph_fsid_compare(&client->fsid, fsid)) {
84 			pr_err("bad fsid, had %pU got %pU",
85 			       &client->fsid, fsid);
86 			return -1;
87 		}
88 	} else {
89 		memcpy(&client->fsid, fsid, sizeof(*fsid));
90 	}
91 	return 0;
92 }
93 EXPORT_SYMBOL(ceph_check_fsid);
94 
95 static int strcmp_null(const char *s1, const char *s2)
96 {
97 	if (!s1 && !s2)
98 		return 0;
99 	if (s1 && !s2)
100 		return -1;
101 	if (!s1 && s2)
102 		return 1;
103 	return strcmp(s1, s2);
104 }
105 
106 int ceph_compare_options(struct ceph_options *new_opt,
107 			 struct ceph_client *client)
108 {
109 	struct ceph_options *opt1 = new_opt;
110 	struct ceph_options *opt2 = client->options;
111 	int ofs = offsetof(struct ceph_options, mon_addr);
112 	int i;
113 	int ret;
114 
115 	ret = memcmp(opt1, opt2, ofs);
116 	if (ret)
117 		return ret;
118 
119 	ret = strcmp_null(opt1->name, opt2->name);
120 	if (ret)
121 		return ret;
122 
123 	if (opt1->key && !opt2->key)
124 		return -1;
125 	if (!opt1->key && opt2->key)
126 		return 1;
127 	if (opt1->key && opt2->key) {
128 		if (opt1->key->type != opt2->key->type)
129 			return -1;
130 		if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
131 			return -1;
132 		if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
133 			return -1;
134 		if (opt1->key->len != opt2->key->len)
135 			return -1;
136 		if (opt1->key->key && !opt2->key->key)
137 			return -1;
138 		if (!opt1->key->key && opt2->key->key)
139 			return 1;
140 		if (opt1->key->key && opt2->key->key) {
141 			ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
142 			if (ret)
143 				return ret;
144 		}
145 	}
146 
147 	/* any matching mon ip implies a match */
148 	for (i = 0; i < opt1->num_mon; i++) {
149 		if (ceph_monmap_contains(client->monc.monmap,
150 				 &opt1->mon_addr[i]))
151 			return 0;
152 	}
153 	return -1;
154 }
155 EXPORT_SYMBOL(ceph_compare_options);
156 
157 
158 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
159 {
160 	int i = 0;
161 	char tmp[3];
162 	int err = -EINVAL;
163 	int d;
164 
165 	dout("parse_fsid '%s'\n", str);
166 	tmp[2] = 0;
167 	while (*str && i < 16) {
168 		if (ispunct(*str)) {
169 			str++;
170 			continue;
171 		}
172 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
173 			break;
174 		tmp[0] = str[0];
175 		tmp[1] = str[1];
176 		if (sscanf(tmp, "%x", &d) < 1)
177 			break;
178 		fsid->fsid[i] = d & 0xff;
179 		i++;
180 		str += 2;
181 	}
182 
183 	if (i == 16)
184 		err = 0;
185 	dout("parse_fsid ret %d got fsid %pU", err, fsid);
186 	return err;
187 }
188 
189 /*
190  * ceph options
191  */
192 enum {
193 	Opt_osdtimeout,
194 	Opt_osdkeepalivetimeout,
195 	Opt_mount_timeout,
196 	Opt_osd_idle_ttl,
197 	Opt_last_int,
198 	/* int args above */
199 	Opt_fsid,
200 	Opt_name,
201 	Opt_secret,
202 	Opt_key,
203 	Opt_ip,
204 	Opt_last_string,
205 	/* string args above */
206 	Opt_share,
207 	Opt_noshare,
208 	Opt_crc,
209 	Opt_nocrc,
210 };
211 
212 static match_table_t opt_tokens = {
213 	{Opt_osdtimeout, "osdtimeout=%d"},
214 	{Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
215 	{Opt_mount_timeout, "mount_timeout=%d"},
216 	{Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
217 	/* int args above */
218 	{Opt_fsid, "fsid=%s"},
219 	{Opt_name, "name=%s"},
220 	{Opt_secret, "secret=%s"},
221 	{Opt_key, "key=%s"},
222 	{Opt_ip, "ip=%s"},
223 	/* string args above */
224 	{Opt_share, "share"},
225 	{Opt_noshare, "noshare"},
226 	{Opt_crc, "crc"},
227 	{Opt_nocrc, "nocrc"},
228 	{-1, NULL}
229 };
230 
231 void ceph_destroy_options(struct ceph_options *opt)
232 {
233 	dout("destroy_options %p\n", opt);
234 	kfree(opt->name);
235 	if (opt->key) {
236 		ceph_crypto_key_destroy(opt->key);
237 		kfree(opt->key);
238 	}
239 	kfree(opt->mon_addr);
240 	kfree(opt);
241 }
242 EXPORT_SYMBOL(ceph_destroy_options);
243 
244 /* get secret from key store */
245 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
246 	struct key *ukey;
247 	int key_err;
248 	int err = 0;
249 	struct ceph_crypto_key *ckey;
250 
251 	ukey = request_key(&key_type_ceph, name, NULL);
252 	if (!ukey || IS_ERR(ukey)) {
253 		/* request_key errors don't map nicely to mount(2)
254 		   errors; don't even try, but still printk */
255 		key_err = PTR_ERR(ukey);
256 		switch (key_err) {
257 		case -ENOKEY:
258 			pr_warning("ceph: Mount failed due to key not found: %s\n", name);
259 			break;
260 		case -EKEYEXPIRED:
261 			pr_warning("ceph: Mount failed due to expired key: %s\n", name);
262 			break;
263 		case -EKEYREVOKED:
264 			pr_warning("ceph: Mount failed due to revoked key: %s\n", name);
265 			break;
266 		default:
267 			pr_warning("ceph: Mount failed due to unknown key error"
268 			       " %d: %s\n", key_err, name);
269 		}
270 		err = -EPERM;
271 		goto out;
272 	}
273 
274 	ckey = ukey->payload.data;
275 	err = ceph_crypto_key_clone(dst, ckey);
276 	if (err)
277 		goto out_key;
278 	/* pass through, err is 0 */
279 
280 out_key:
281 	key_put(ukey);
282 out:
283 	return err;
284 }
285 
286 struct ceph_options *
287 ceph_parse_options(char *options, const char *dev_name,
288 			const char *dev_name_end,
289 			int (*parse_extra_token)(char *c, void *private),
290 			void *private)
291 {
292 	struct ceph_options *opt;
293 	const char *c;
294 	int err = -ENOMEM;
295 	substring_t argstr[MAX_OPT_ARGS];
296 
297 	if (current->nsproxy->net_ns != &init_net)
298 		return ERR_PTR(-EINVAL);
299 
300 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
301 	if (!opt)
302 		return ERR_PTR(-ENOMEM);
303 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
304 				GFP_KERNEL);
305 	if (!opt->mon_addr)
306 		goto out;
307 
308 	dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
309 	     dev_name);
310 
311 	/* start with defaults */
312 	opt->flags = CEPH_OPT_DEFAULT;
313 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
314 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
315 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
316 
317 	/* get mon ip(s) */
318 	/* ip1[:port1][,ip2[:port2]...] */
319 	err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
320 			     CEPH_MAX_MON, &opt->num_mon);
321 	if (err < 0)
322 		goto out;
323 
324 	/* parse mount options */
325 	while ((c = strsep(&options, ",")) != NULL) {
326 		int token, intval, ret;
327 		if (!*c)
328 			continue;
329 		err = -EINVAL;
330 		token = match_token((char *)c, opt_tokens, argstr);
331 		if (token < 0 && parse_extra_token) {
332 			/* extra? */
333 			err = parse_extra_token((char *)c, private);
334 			if (err < 0) {
335 				pr_err("bad option at '%s'\n", c);
336 				goto out;
337 			}
338 			continue;
339 		}
340 		if (token < Opt_last_int) {
341 			ret = match_int(&argstr[0], &intval);
342 			if (ret < 0) {
343 				pr_err("bad mount option arg (not int) "
344 				       "at '%s'\n", c);
345 				continue;
346 			}
347 			dout("got int token %d val %d\n", token, intval);
348 		} else if (token > Opt_last_int && token < Opt_last_string) {
349 			dout("got string token %d val %s\n", token,
350 			     argstr[0].from);
351 		} else {
352 			dout("got token %d\n", token);
353 		}
354 		switch (token) {
355 		case Opt_ip:
356 			err = ceph_parse_ips(argstr[0].from,
357 					     argstr[0].to,
358 					     &opt->my_addr,
359 					     1, NULL);
360 			if (err < 0)
361 				goto out;
362 			opt->flags |= CEPH_OPT_MYIP;
363 			break;
364 
365 		case Opt_fsid:
366 			err = parse_fsid(argstr[0].from, &opt->fsid);
367 			if (err == 0)
368 				opt->flags |= CEPH_OPT_FSID;
369 			break;
370 		case Opt_name:
371 			opt->name = kstrndup(argstr[0].from,
372 					      argstr[0].to-argstr[0].from,
373 					      GFP_KERNEL);
374 			break;
375 		case Opt_secret:
376 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
377 			if (!opt->key) {
378 				err = -ENOMEM;
379 				goto out;
380 			}
381 			err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
382 			if (err < 0)
383 				goto out;
384 			break;
385 		case Opt_key:
386 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
387 			if (!opt->key) {
388 				err = -ENOMEM;
389 				goto out;
390 			}
391 			err = get_secret(opt->key, argstr[0].from);
392 			if (err < 0)
393 				goto out;
394 			break;
395 
396 			/* misc */
397 		case Opt_osdtimeout:
398 			pr_warning("ignoring deprecated osdtimeout option\n");
399 			break;
400 		case Opt_osdkeepalivetimeout:
401 			opt->osd_keepalive_timeout = intval;
402 			break;
403 		case Opt_osd_idle_ttl:
404 			opt->osd_idle_ttl = intval;
405 			break;
406 		case Opt_mount_timeout:
407 			opt->mount_timeout = intval;
408 			break;
409 
410 		case Opt_share:
411 			opt->flags &= ~CEPH_OPT_NOSHARE;
412 			break;
413 		case Opt_noshare:
414 			opt->flags |= CEPH_OPT_NOSHARE;
415 			break;
416 
417 		case Opt_crc:
418 			opt->flags &= ~CEPH_OPT_NOCRC;
419 			break;
420 		case Opt_nocrc:
421 			opt->flags |= CEPH_OPT_NOCRC;
422 			break;
423 
424 		default:
425 			BUG_ON(token);
426 		}
427 	}
428 
429 	/* success */
430 	return opt;
431 
432 out:
433 	ceph_destroy_options(opt);
434 	return ERR_PTR(err);
435 }
436 EXPORT_SYMBOL(ceph_parse_options);
437 
438 u64 ceph_client_id(struct ceph_client *client)
439 {
440 	return client->monc.auth->global_id;
441 }
442 EXPORT_SYMBOL(ceph_client_id);
443 
444 /*
445  * create a fresh client instance
446  */
447 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
448 				       unsigned int supported_features,
449 				       unsigned int required_features)
450 {
451 	struct ceph_client *client;
452 	struct ceph_entity_addr *myaddr = NULL;
453 	int err = -ENOMEM;
454 
455 	client = kzalloc(sizeof(*client), GFP_KERNEL);
456 	if (client == NULL)
457 		return ERR_PTR(-ENOMEM);
458 
459 	client->private = private;
460 	client->options = opt;
461 
462 	mutex_init(&client->mount_mutex);
463 	init_waitqueue_head(&client->auth_wq);
464 	client->auth_err = 0;
465 
466 	client->extra_mon_dispatch = NULL;
467 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
468 		supported_features;
469 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
470 		required_features;
471 
472 	/* msgr */
473 	if (ceph_test_opt(client, MYIP))
474 		myaddr = &client->options->my_addr;
475 	ceph_messenger_init(&client->msgr, myaddr,
476 		client->supported_features,
477 		client->required_features,
478 		ceph_test_opt(client, NOCRC));
479 
480 	/* subsystems */
481 	err = ceph_monc_init(&client->monc, client);
482 	if (err < 0)
483 		goto fail;
484 	err = ceph_osdc_init(&client->osdc, client);
485 	if (err < 0)
486 		goto fail_monc;
487 
488 	return client;
489 
490 fail_monc:
491 	ceph_monc_stop(&client->monc);
492 fail:
493 	kfree(client);
494 	return ERR_PTR(err);
495 }
496 EXPORT_SYMBOL(ceph_create_client);
497 
498 void ceph_destroy_client(struct ceph_client *client)
499 {
500 	dout("destroy_client %p\n", client);
501 
502 	atomic_set(&client->msgr.stopping, 1);
503 
504 	/* unmount */
505 	ceph_osdc_stop(&client->osdc);
506 
507 	ceph_monc_stop(&client->monc);
508 
509 	ceph_debugfs_client_cleanup(client);
510 
511 	ceph_destroy_options(client->options);
512 
513 	kfree(client);
514 	dout("destroy_client %p done\n", client);
515 }
516 EXPORT_SYMBOL(ceph_destroy_client);
517 
518 /*
519  * true if we have the mon map (and have thus joined the cluster)
520  */
521 static int have_mon_and_osd_map(struct ceph_client *client)
522 {
523 	return client->monc.monmap && client->monc.monmap->epoch &&
524 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
525 }
526 
527 /*
528  * mount: join the ceph cluster, and open root directory.
529  */
530 int __ceph_open_session(struct ceph_client *client, unsigned long started)
531 {
532 	int err;
533 	unsigned long timeout = client->options->mount_timeout * HZ;
534 
535 	/* open session, and wait for mon and osd maps */
536 	err = ceph_monc_open_session(&client->monc);
537 	if (err < 0)
538 		return err;
539 
540 	while (!have_mon_and_osd_map(client)) {
541 		err = -EIO;
542 		if (timeout && time_after_eq(jiffies, started + timeout))
543 			return err;
544 
545 		/* wait */
546 		dout("mount waiting for mon_map\n");
547 		err = wait_event_interruptible_timeout(client->auth_wq,
548 			have_mon_and_osd_map(client) || (client->auth_err < 0),
549 			timeout);
550 		if (err == -EINTR || err == -ERESTARTSYS)
551 			return err;
552 		if (client->auth_err < 0)
553 			return client->auth_err;
554 	}
555 
556 	return 0;
557 }
558 EXPORT_SYMBOL(__ceph_open_session);
559 
560 
561 int ceph_open_session(struct ceph_client *client)
562 {
563 	int ret;
564 	unsigned long started = jiffies;  /* note the start time */
565 
566 	dout("open_session start\n");
567 	mutex_lock(&client->mount_mutex);
568 
569 	ret = __ceph_open_session(client, started);
570 
571 	mutex_unlock(&client->mount_mutex);
572 	return ret;
573 }
574 EXPORT_SYMBOL(ceph_open_session);
575 
576 
577 static int __init init_ceph_lib(void)
578 {
579 	int ret = 0;
580 
581 	ret = ceph_debugfs_init();
582 	if (ret < 0)
583 		goto out;
584 
585 	ret = ceph_crypto_init();
586 	if (ret < 0)
587 		goto out_debugfs;
588 
589 	ret = ceph_msgr_init();
590 	if (ret < 0)
591 		goto out_crypto;
592 
593 	pr_info("loaded (mon/osd proto %d/%d, osdmap %d/%d %d/%d)\n",
594 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL,
595 		CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
596 		CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
597 
598 	return 0;
599 
600 out_crypto:
601 	ceph_crypto_shutdown();
602 out_debugfs:
603 	ceph_debugfs_cleanup();
604 out:
605 	return ret;
606 }
607 
608 static void __exit exit_ceph_lib(void)
609 {
610 	dout("exit_ceph_lib\n");
611 	ceph_msgr_exit();
612 	ceph_crypto_shutdown();
613 	ceph_debugfs_cleanup();
614 }
615 
616 module_init(init_ceph_lib);
617 module_exit(exit_ceph_lib);
618 
619 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
620 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
621 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
622 MODULE_DESCRIPTION("Ceph filesystem for Linux");
623 MODULE_LICENSE("GPL");
624