xref: /linux/net/ceph/ceph_common.c (revision 25aee3debe0464f6c680173041fa3de30ec9ff54)
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 
19 
20 #include <linux/ceph/ceph_features.h>
21 #include <linux/ceph/libceph.h>
22 #include <linux/ceph/debugfs.h>
23 #include <linux/ceph/decode.h>
24 #include <linux/ceph/mon_client.h>
25 #include <linux/ceph/auth.h>
26 #include "crypto.h"
27 
28 
29 
30 /*
31  * find filename portion of a path (/foo/bar/baz -> baz)
32  */
33 const char *ceph_file_part(const char *s, int len)
34 {
35 	const char *e = s + len;
36 
37 	while (e != s && *(e-1) != '/')
38 		e--;
39 	return e;
40 }
41 EXPORT_SYMBOL(ceph_file_part);
42 
43 const char *ceph_msg_type_name(int type)
44 {
45 	switch (type) {
46 	case CEPH_MSG_SHUTDOWN: return "shutdown";
47 	case CEPH_MSG_PING: return "ping";
48 	case CEPH_MSG_AUTH: return "auth";
49 	case CEPH_MSG_AUTH_REPLY: return "auth_reply";
50 	case CEPH_MSG_MON_MAP: return "mon_map";
51 	case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
52 	case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
53 	case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
54 	case CEPH_MSG_STATFS: return "statfs";
55 	case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
56 	case CEPH_MSG_MDS_MAP: return "mds_map";
57 	case CEPH_MSG_CLIENT_SESSION: return "client_session";
58 	case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
59 	case CEPH_MSG_CLIENT_REQUEST: return "client_request";
60 	case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
61 	case CEPH_MSG_CLIENT_REPLY: return "client_reply";
62 	case CEPH_MSG_CLIENT_CAPS: return "client_caps";
63 	case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
64 	case CEPH_MSG_CLIENT_SNAP: return "client_snap";
65 	case CEPH_MSG_CLIENT_LEASE: return "client_lease";
66 	case CEPH_MSG_OSD_MAP: return "osd_map";
67 	case CEPH_MSG_OSD_OP: return "osd_op";
68 	case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
69 	case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
70 	default: return "unknown";
71 	}
72 }
73 EXPORT_SYMBOL(ceph_msg_type_name);
74 
75 /*
76  * Initially learn our fsid, or verify an fsid matches.
77  */
78 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
79 {
80 	if (client->have_fsid) {
81 		if (ceph_fsid_compare(&client->fsid, fsid)) {
82 			pr_err("bad fsid, had %pU got %pU",
83 			       &client->fsid, fsid);
84 			return -1;
85 		}
86 	} else {
87 		pr_info("client%lld fsid %pU\n", ceph_client_id(client), fsid);
88 		memcpy(&client->fsid, fsid, sizeof(*fsid));
89 	}
90 	return 0;
91 }
92 EXPORT_SYMBOL(ceph_check_fsid);
93 
94 static int strcmp_null(const char *s1, const char *s2)
95 {
96 	if (!s1 && !s2)
97 		return 0;
98 	if (s1 && !s2)
99 		return -1;
100 	if (!s1 && s2)
101 		return 1;
102 	return strcmp(s1, s2);
103 }
104 
105 int ceph_compare_options(struct ceph_options *new_opt,
106 			 struct ceph_client *client)
107 {
108 	struct ceph_options *opt1 = new_opt;
109 	struct ceph_options *opt2 = client->options;
110 	int ofs = offsetof(struct ceph_options, mon_addr);
111 	int i;
112 	int ret;
113 
114 	ret = memcmp(opt1, opt2, ofs);
115 	if (ret)
116 		return ret;
117 
118 	ret = strcmp_null(opt1->name, opt2->name);
119 	if (ret)
120 		return ret;
121 
122 	if (opt1->key && !opt2->key)
123 		return -1;
124 	if (!opt1->key && opt2->key)
125 		return 1;
126 	if (opt1->key && opt2->key) {
127 		if (opt1->key->type != opt2->key->type)
128 			return -1;
129 		if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
130 			return -1;
131 		if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
132 			return -1;
133 		if (opt1->key->len != opt2->key->len)
134 			return -1;
135 		if (opt1->key->key && !opt2->key->key)
136 			return -1;
137 		if (!opt1->key->key && opt2->key->key)
138 			return 1;
139 		if (opt1->key->key && opt2->key->key) {
140 			ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
141 			if (ret)
142 				return ret;
143 		}
144 	}
145 
146 	/* any matching mon ip implies a match */
147 	for (i = 0; i < opt1->num_mon; i++) {
148 		if (ceph_monmap_contains(client->monc.monmap,
149 				 &opt1->mon_addr[i]))
150 			return 0;
151 	}
152 	return -1;
153 }
154 EXPORT_SYMBOL(ceph_compare_options);
155 
156 
157 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
158 {
159 	int i = 0;
160 	char tmp[3];
161 	int err = -EINVAL;
162 	int d;
163 
164 	dout("parse_fsid '%s'\n", str);
165 	tmp[2] = 0;
166 	while (*str && i < 16) {
167 		if (ispunct(*str)) {
168 			str++;
169 			continue;
170 		}
171 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
172 			break;
173 		tmp[0] = str[0];
174 		tmp[1] = str[1];
175 		if (sscanf(tmp, "%x", &d) < 1)
176 			break;
177 		fsid->fsid[i] = d & 0xff;
178 		i++;
179 		str += 2;
180 	}
181 
182 	if (i == 16)
183 		err = 0;
184 	dout("parse_fsid ret %d got fsid %pU", err, fsid);
185 	return err;
186 }
187 
188 /*
189  * ceph options
190  */
191 enum {
192 	Opt_osdtimeout,
193 	Opt_osdkeepalivetimeout,
194 	Opt_mount_timeout,
195 	Opt_osd_idle_ttl,
196 	Opt_last_int,
197 	/* int args above */
198 	Opt_fsid,
199 	Opt_name,
200 	Opt_secret,
201 	Opt_key,
202 	Opt_ip,
203 	Opt_last_string,
204 	/* string args above */
205 	Opt_share,
206 	Opt_noshare,
207 	Opt_crc,
208 	Opt_nocrc,
209 };
210 
211 static match_table_t opt_tokens = {
212 	{Opt_osdtimeout, "osdtimeout=%d"},
213 	{Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
214 	{Opt_mount_timeout, "mount_timeout=%d"},
215 	{Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
216 	/* int args above */
217 	{Opt_fsid, "fsid=%s"},
218 	{Opt_name, "name=%s"},
219 	{Opt_secret, "secret=%s"},
220 	{Opt_key, "key=%s"},
221 	{Opt_ip, "ip=%s"},
222 	/* string args above */
223 	{Opt_share, "share"},
224 	{Opt_noshare, "noshare"},
225 	{Opt_crc, "crc"},
226 	{Opt_nocrc, "nocrc"},
227 	{-1, NULL}
228 };
229 
230 void ceph_destroy_options(struct ceph_options *opt)
231 {
232 	dout("destroy_options %p\n", opt);
233 	kfree(opt->name);
234 	if (opt->key) {
235 		ceph_crypto_key_destroy(opt->key);
236 		kfree(opt->key);
237 	}
238 	kfree(opt->mon_addr);
239 	kfree(opt);
240 }
241 EXPORT_SYMBOL(ceph_destroy_options);
242 
243 /* get secret from key store */
244 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
245 	struct key *ukey;
246 	int key_err;
247 	int err = 0;
248 	struct ceph_crypto_key *ckey;
249 
250 	ukey = request_key(&key_type_ceph, name, NULL);
251 	if (!ukey || IS_ERR(ukey)) {
252 		/* request_key errors don't map nicely to mount(2)
253 		   errors; don't even try, but still printk */
254 		key_err = PTR_ERR(ukey);
255 		switch (key_err) {
256 		case -ENOKEY:
257 			pr_warning("ceph: Mount failed due to key not found: %s\n", name);
258 			break;
259 		case -EKEYEXPIRED:
260 			pr_warning("ceph: Mount failed due to expired key: %s\n", name);
261 			break;
262 		case -EKEYREVOKED:
263 			pr_warning("ceph: Mount failed due to revoked key: %s\n", name);
264 			break;
265 		default:
266 			pr_warning("ceph: Mount failed due to unknown key error"
267 			       " %d: %s\n", key_err, name);
268 		}
269 		err = -EPERM;
270 		goto out;
271 	}
272 
273 	ckey = ukey->payload.data;
274 	err = ceph_crypto_key_clone(dst, ckey);
275 	if (err)
276 		goto out_key;
277 	/* pass through, err is 0 */
278 
279 out_key:
280 	key_put(ukey);
281 out:
282 	return err;
283 }
284 
285 struct ceph_options *
286 ceph_parse_options(char *options, const char *dev_name,
287 			const char *dev_name_end,
288 			int (*parse_extra_token)(char *c, void *private),
289 			void *private)
290 {
291 	struct ceph_options *opt;
292 	const char *c;
293 	int err = -ENOMEM;
294 	substring_t argstr[MAX_OPT_ARGS];
295 
296 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
297 	if (!opt)
298 		return ERR_PTR(-ENOMEM);
299 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
300 				GFP_KERNEL);
301 	if (!opt->mon_addr)
302 		goto out;
303 
304 	dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
305 	     dev_name);
306 
307 	/* start with defaults */
308 	opt->flags = CEPH_OPT_DEFAULT;
309 	opt->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
310 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
311 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
312 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
313 
314 	/* get mon ip(s) */
315 	/* ip1[:port1][,ip2[:port2]...] */
316 	err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
317 			     CEPH_MAX_MON, &opt->num_mon);
318 	if (err < 0)
319 		goto out;
320 
321 	/* parse mount options */
322 	while ((c = strsep(&options, ",")) != NULL) {
323 		int token, intval, ret;
324 		if (!*c)
325 			continue;
326 		err = -EINVAL;
327 		token = match_token((char *)c, opt_tokens, argstr);
328 		if (token < 0 && parse_extra_token) {
329 			/* extra? */
330 			err = parse_extra_token((char *)c, private);
331 			if (err < 0) {
332 				pr_err("bad option at '%s'\n", c);
333 				goto out;
334 			}
335 			continue;
336 		}
337 		if (token < Opt_last_int) {
338 			ret = match_int(&argstr[0], &intval);
339 			if (ret < 0) {
340 				pr_err("bad mount option arg (not int) "
341 				       "at '%s'\n", c);
342 				continue;
343 			}
344 			dout("got int token %d val %d\n", token, intval);
345 		} else if (token > Opt_last_int && token < Opt_last_string) {
346 			dout("got string token %d val %s\n", token,
347 			     argstr[0].from);
348 		} else {
349 			dout("got token %d\n", token);
350 		}
351 		switch (token) {
352 		case Opt_ip:
353 			err = ceph_parse_ips(argstr[0].from,
354 					     argstr[0].to,
355 					     &opt->my_addr,
356 					     1, NULL);
357 			if (err < 0)
358 				goto out;
359 			opt->flags |= CEPH_OPT_MYIP;
360 			break;
361 
362 		case Opt_fsid:
363 			err = parse_fsid(argstr[0].from, &opt->fsid);
364 			if (err == 0)
365 				opt->flags |= CEPH_OPT_FSID;
366 			break;
367 		case Opt_name:
368 			opt->name = kstrndup(argstr[0].from,
369 					      argstr[0].to-argstr[0].from,
370 					      GFP_KERNEL);
371 			break;
372 		case Opt_secret:
373 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
374 			if (!opt->key) {
375 				err = -ENOMEM;
376 				goto out;
377 			}
378 			err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
379 			if (err < 0)
380 				goto out;
381 			break;
382 		case Opt_key:
383 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
384 			if (!opt->key) {
385 				err = -ENOMEM;
386 				goto out;
387 			}
388 			err = get_secret(opt->key, argstr[0].from);
389 			if (err < 0)
390 				goto out;
391 			break;
392 
393 			/* misc */
394 		case Opt_osdtimeout:
395 			opt->osd_timeout = intval;
396 			break;
397 		case Opt_osdkeepalivetimeout:
398 			opt->osd_keepalive_timeout = intval;
399 			break;
400 		case Opt_osd_idle_ttl:
401 			opt->osd_idle_ttl = intval;
402 			break;
403 		case Opt_mount_timeout:
404 			opt->mount_timeout = intval;
405 			break;
406 
407 		case Opt_share:
408 			opt->flags &= ~CEPH_OPT_NOSHARE;
409 			break;
410 		case Opt_noshare:
411 			opt->flags |= CEPH_OPT_NOSHARE;
412 			break;
413 
414 		case Opt_crc:
415 			opt->flags &= ~CEPH_OPT_NOCRC;
416 			break;
417 		case Opt_nocrc:
418 			opt->flags |= CEPH_OPT_NOCRC;
419 			break;
420 
421 		default:
422 			BUG_ON(token);
423 		}
424 	}
425 
426 	/* success */
427 	return opt;
428 
429 out:
430 	ceph_destroy_options(opt);
431 	return ERR_PTR(err);
432 }
433 EXPORT_SYMBOL(ceph_parse_options);
434 
435 u64 ceph_client_id(struct ceph_client *client)
436 {
437 	return client->monc.auth->global_id;
438 }
439 EXPORT_SYMBOL(ceph_client_id);
440 
441 /*
442  * create a fresh client instance
443  */
444 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
445 				       unsigned int supported_features,
446 				       unsigned int required_features)
447 {
448 	struct ceph_client *client;
449 	struct ceph_entity_addr *myaddr = NULL;
450 	int err = -ENOMEM;
451 
452 	client = kzalloc(sizeof(*client), GFP_KERNEL);
453 	if (client == NULL)
454 		return ERR_PTR(-ENOMEM);
455 
456 	client->private = private;
457 	client->options = opt;
458 
459 	mutex_init(&client->mount_mutex);
460 	init_waitqueue_head(&client->auth_wq);
461 	client->auth_err = 0;
462 
463 	client->extra_mon_dispatch = NULL;
464 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
465 		supported_features;
466 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
467 		required_features;
468 
469 	/* msgr */
470 	if (ceph_test_opt(client, MYIP))
471 		myaddr = &client->options->my_addr;
472 	ceph_messenger_init(&client->msgr, myaddr,
473 		client->supported_features,
474 		client->required_features,
475 		ceph_test_opt(client, NOCRC));
476 
477 	/* subsystems */
478 	err = ceph_monc_init(&client->monc, client);
479 	if (err < 0)
480 		goto fail;
481 	err = ceph_osdc_init(&client->osdc, client);
482 	if (err < 0)
483 		goto fail_monc;
484 
485 	return client;
486 
487 fail_monc:
488 	ceph_monc_stop(&client->monc);
489 fail:
490 	kfree(client);
491 	return ERR_PTR(err);
492 }
493 EXPORT_SYMBOL(ceph_create_client);
494 
495 void ceph_destroy_client(struct ceph_client *client)
496 {
497 	dout("destroy_client %p\n", client);
498 
499 	atomic_set(&client->msgr.stopping, 1);
500 
501 	/* unmount */
502 	ceph_osdc_stop(&client->osdc);
503 
504 	ceph_monc_stop(&client->monc);
505 
506 	ceph_debugfs_client_cleanup(client);
507 
508 	ceph_destroy_options(client->options);
509 
510 	kfree(client);
511 	dout("destroy_client %p done\n", client);
512 }
513 EXPORT_SYMBOL(ceph_destroy_client);
514 
515 /*
516  * true if we have the mon map (and have thus joined the cluster)
517  */
518 static int have_mon_and_osd_map(struct ceph_client *client)
519 {
520 	return client->monc.monmap && client->monc.monmap->epoch &&
521 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
522 }
523 
524 /*
525  * mount: join the ceph cluster, and open root directory.
526  */
527 int __ceph_open_session(struct ceph_client *client, unsigned long started)
528 {
529 	int err;
530 	unsigned long timeout = client->options->mount_timeout * HZ;
531 
532 	/* open session, and wait for mon and osd maps */
533 	err = ceph_monc_open_session(&client->monc);
534 	if (err < 0)
535 		return err;
536 
537 	while (!have_mon_and_osd_map(client)) {
538 		err = -EIO;
539 		if (timeout && time_after_eq(jiffies, started + timeout))
540 			return err;
541 
542 		/* wait */
543 		dout("mount waiting for mon_map\n");
544 		err = wait_event_interruptible_timeout(client->auth_wq,
545 			have_mon_and_osd_map(client) || (client->auth_err < 0),
546 			timeout);
547 		if (err == -EINTR || err == -ERESTARTSYS)
548 			return err;
549 		if (client->auth_err < 0)
550 			return client->auth_err;
551 	}
552 
553 	return 0;
554 }
555 EXPORT_SYMBOL(__ceph_open_session);
556 
557 
558 int ceph_open_session(struct ceph_client *client)
559 {
560 	int ret;
561 	unsigned long started = jiffies;  /* note the start time */
562 
563 	dout("open_session start\n");
564 	mutex_lock(&client->mount_mutex);
565 
566 	ret = __ceph_open_session(client, started);
567 
568 	mutex_unlock(&client->mount_mutex);
569 	return ret;
570 }
571 EXPORT_SYMBOL(ceph_open_session);
572 
573 
574 static int __init init_ceph_lib(void)
575 {
576 	int ret = 0;
577 
578 	ret = ceph_debugfs_init();
579 	if (ret < 0)
580 		goto out;
581 
582 	ret = ceph_crypto_init();
583 	if (ret < 0)
584 		goto out_debugfs;
585 
586 	ret = ceph_msgr_init();
587 	if (ret < 0)
588 		goto out_crypto;
589 
590 	pr_info("loaded (mon/osd proto %d/%d, osdmap %d/%d %d/%d)\n",
591 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL,
592 		CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
593 		CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
594 
595 	return 0;
596 
597 out_crypto:
598 	ceph_crypto_shutdown();
599 out_debugfs:
600 	ceph_debugfs_cleanup();
601 out:
602 	return ret;
603 }
604 
605 static void __exit exit_ceph_lib(void)
606 {
607 	dout("exit_ceph_lib\n");
608 	ceph_msgr_exit();
609 	ceph_crypto_shutdown();
610 	ceph_debugfs_cleanup();
611 }
612 
613 module_init(init_ceph_lib);
614 module_exit(exit_ceph_lib);
615 
616 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
617 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
618 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
619 MODULE_DESCRIPTION("Ceph filesystem for Linux");
620 MODULE_LICENSE("GPL");
621