xref: /openbmc/linux/net/ceph/ceph_common.c (revision d0b73b48)
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 		memcpy(&client->fsid, fsid, sizeof(*fsid));
88 	}
89 	return 0;
90 }
91 EXPORT_SYMBOL(ceph_check_fsid);
92 
93 static int strcmp_null(const char *s1, const char *s2)
94 {
95 	if (!s1 && !s2)
96 		return 0;
97 	if (s1 && !s2)
98 		return -1;
99 	if (!s1 && s2)
100 		return 1;
101 	return strcmp(s1, s2);
102 }
103 
104 int ceph_compare_options(struct ceph_options *new_opt,
105 			 struct ceph_client *client)
106 {
107 	struct ceph_options *opt1 = new_opt;
108 	struct ceph_options *opt2 = client->options;
109 	int ofs = offsetof(struct ceph_options, mon_addr);
110 	int i;
111 	int ret;
112 
113 	ret = memcmp(opt1, opt2, ofs);
114 	if (ret)
115 		return ret;
116 
117 	ret = strcmp_null(opt1->name, opt2->name);
118 	if (ret)
119 		return ret;
120 
121 	if (opt1->key && !opt2->key)
122 		return -1;
123 	if (!opt1->key && opt2->key)
124 		return 1;
125 	if (opt1->key && opt2->key) {
126 		if (opt1->key->type != opt2->key->type)
127 			return -1;
128 		if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
129 			return -1;
130 		if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
131 			return -1;
132 		if (opt1->key->len != opt2->key->len)
133 			return -1;
134 		if (opt1->key->key && !opt2->key->key)
135 			return -1;
136 		if (!opt1->key->key && opt2->key->key)
137 			return 1;
138 		if (opt1->key->key && opt2->key->key) {
139 			ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
140 			if (ret)
141 				return ret;
142 		}
143 	}
144 
145 	/* any matching mon ip implies a match */
146 	for (i = 0; i < opt1->num_mon; i++) {
147 		if (ceph_monmap_contains(client->monc.monmap,
148 				 &opt1->mon_addr[i]))
149 			return 0;
150 	}
151 	return -1;
152 }
153 EXPORT_SYMBOL(ceph_compare_options);
154 
155 
156 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
157 {
158 	int i = 0;
159 	char tmp[3];
160 	int err = -EINVAL;
161 	int d;
162 
163 	dout("parse_fsid '%s'\n", str);
164 	tmp[2] = 0;
165 	while (*str && i < 16) {
166 		if (ispunct(*str)) {
167 			str++;
168 			continue;
169 		}
170 		if (!isxdigit(str[0]) || !isxdigit(str[1]))
171 			break;
172 		tmp[0] = str[0];
173 		tmp[1] = str[1];
174 		if (sscanf(tmp, "%x", &d) < 1)
175 			break;
176 		fsid->fsid[i] = d & 0xff;
177 		i++;
178 		str += 2;
179 	}
180 
181 	if (i == 16)
182 		err = 0;
183 	dout("parse_fsid ret %d got fsid %pU", err, fsid);
184 	return err;
185 }
186 
187 /*
188  * ceph options
189  */
190 enum {
191 	Opt_osdtimeout,
192 	Opt_osdkeepalivetimeout,
193 	Opt_mount_timeout,
194 	Opt_osd_idle_ttl,
195 	Opt_last_int,
196 	/* int args above */
197 	Opt_fsid,
198 	Opt_name,
199 	Opt_secret,
200 	Opt_key,
201 	Opt_ip,
202 	Opt_last_string,
203 	/* string args above */
204 	Opt_share,
205 	Opt_noshare,
206 	Opt_crc,
207 	Opt_nocrc,
208 };
209 
210 static match_table_t opt_tokens = {
211 	{Opt_osdtimeout, "osdtimeout=%d"},
212 	{Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
213 	{Opt_mount_timeout, "mount_timeout=%d"},
214 	{Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
215 	/* int args above */
216 	{Opt_fsid, "fsid=%s"},
217 	{Opt_name, "name=%s"},
218 	{Opt_secret, "secret=%s"},
219 	{Opt_key, "key=%s"},
220 	{Opt_ip, "ip=%s"},
221 	/* string args above */
222 	{Opt_share, "share"},
223 	{Opt_noshare, "noshare"},
224 	{Opt_crc, "crc"},
225 	{Opt_nocrc, "nocrc"},
226 	{-1, NULL}
227 };
228 
229 void ceph_destroy_options(struct ceph_options *opt)
230 {
231 	dout("destroy_options %p\n", opt);
232 	kfree(opt->name);
233 	if (opt->key) {
234 		ceph_crypto_key_destroy(opt->key);
235 		kfree(opt->key);
236 	}
237 	kfree(opt->mon_addr);
238 	kfree(opt);
239 }
240 EXPORT_SYMBOL(ceph_destroy_options);
241 
242 /* get secret from key store */
243 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
244 	struct key *ukey;
245 	int key_err;
246 	int err = 0;
247 	struct ceph_crypto_key *ckey;
248 
249 	ukey = request_key(&key_type_ceph, name, NULL);
250 	if (!ukey || IS_ERR(ukey)) {
251 		/* request_key errors don't map nicely to mount(2)
252 		   errors; don't even try, but still printk */
253 		key_err = PTR_ERR(ukey);
254 		switch (key_err) {
255 		case -ENOKEY:
256 			pr_warning("ceph: Mount failed due to key not found: %s\n", name);
257 			break;
258 		case -EKEYEXPIRED:
259 			pr_warning("ceph: Mount failed due to expired key: %s\n", name);
260 			break;
261 		case -EKEYREVOKED:
262 			pr_warning("ceph: Mount failed due to revoked key: %s\n", name);
263 			break;
264 		default:
265 			pr_warning("ceph: Mount failed due to unknown key error"
266 			       " %d: %s\n", key_err, name);
267 		}
268 		err = -EPERM;
269 		goto out;
270 	}
271 
272 	ckey = ukey->payload.data;
273 	err = ceph_crypto_key_clone(dst, ckey);
274 	if (err)
275 		goto out_key;
276 	/* pass through, err is 0 */
277 
278 out_key:
279 	key_put(ukey);
280 out:
281 	return err;
282 }
283 
284 struct ceph_options *
285 ceph_parse_options(char *options, const char *dev_name,
286 			const char *dev_name_end,
287 			int (*parse_extra_token)(char *c, void *private),
288 			void *private)
289 {
290 	struct ceph_options *opt;
291 	const char *c;
292 	int err = -ENOMEM;
293 	substring_t argstr[MAX_OPT_ARGS];
294 
295 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
296 	if (!opt)
297 		return ERR_PTR(-ENOMEM);
298 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
299 				GFP_KERNEL);
300 	if (!opt->mon_addr)
301 		goto out;
302 
303 	dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
304 	     dev_name);
305 
306 	/* start with defaults */
307 	opt->flags = CEPH_OPT_DEFAULT;
308 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
309 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
310 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
311 
312 	/* get mon ip(s) */
313 	/* ip1[:port1][,ip2[:port2]...] */
314 	err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
315 			     CEPH_MAX_MON, &opt->num_mon);
316 	if (err < 0)
317 		goto out;
318 
319 	/* parse mount options */
320 	while ((c = strsep(&options, ",")) != NULL) {
321 		int token, intval, ret;
322 		if (!*c)
323 			continue;
324 		err = -EINVAL;
325 		token = match_token((char *)c, opt_tokens, argstr);
326 		if (token < 0 && parse_extra_token) {
327 			/* extra? */
328 			err = parse_extra_token((char *)c, private);
329 			if (err < 0) {
330 				pr_err("bad option at '%s'\n", c);
331 				goto out;
332 			}
333 			continue;
334 		}
335 		if (token < Opt_last_int) {
336 			ret = match_int(&argstr[0], &intval);
337 			if (ret < 0) {
338 				pr_err("bad mount option arg (not int) "
339 				       "at '%s'\n", c);
340 				continue;
341 			}
342 			dout("got int token %d val %d\n", token, intval);
343 		} else if (token > Opt_last_int && token < Opt_last_string) {
344 			dout("got string token %d val %s\n", token,
345 			     argstr[0].from);
346 		} else {
347 			dout("got token %d\n", token);
348 		}
349 		switch (token) {
350 		case Opt_ip:
351 			err = ceph_parse_ips(argstr[0].from,
352 					     argstr[0].to,
353 					     &opt->my_addr,
354 					     1, NULL);
355 			if (err < 0)
356 				goto out;
357 			opt->flags |= CEPH_OPT_MYIP;
358 			break;
359 
360 		case Opt_fsid:
361 			err = parse_fsid(argstr[0].from, &opt->fsid);
362 			if (err == 0)
363 				opt->flags |= CEPH_OPT_FSID;
364 			break;
365 		case Opt_name:
366 			opt->name = kstrndup(argstr[0].from,
367 					      argstr[0].to-argstr[0].from,
368 					      GFP_KERNEL);
369 			break;
370 		case Opt_secret:
371 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
372 			if (!opt->key) {
373 				err = -ENOMEM;
374 				goto out;
375 			}
376 			err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
377 			if (err < 0)
378 				goto out;
379 			break;
380 		case Opt_key:
381 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
382 			if (!opt->key) {
383 				err = -ENOMEM;
384 				goto out;
385 			}
386 			err = get_secret(opt->key, argstr[0].from);
387 			if (err < 0)
388 				goto out;
389 			break;
390 
391 			/* misc */
392 		case Opt_osdtimeout:
393 			pr_warning("ignoring deprecated osdtimeout option\n");
394 			break;
395 		case Opt_osdkeepalivetimeout:
396 			opt->osd_keepalive_timeout = intval;
397 			break;
398 		case Opt_osd_idle_ttl:
399 			opt->osd_idle_ttl = intval;
400 			break;
401 		case Opt_mount_timeout:
402 			opt->mount_timeout = intval;
403 			break;
404 
405 		case Opt_share:
406 			opt->flags &= ~CEPH_OPT_NOSHARE;
407 			break;
408 		case Opt_noshare:
409 			opt->flags |= CEPH_OPT_NOSHARE;
410 			break;
411 
412 		case Opt_crc:
413 			opt->flags &= ~CEPH_OPT_NOCRC;
414 			break;
415 		case Opt_nocrc:
416 			opt->flags |= CEPH_OPT_NOCRC;
417 			break;
418 
419 		default:
420 			BUG_ON(token);
421 		}
422 	}
423 
424 	/* success */
425 	return opt;
426 
427 out:
428 	ceph_destroy_options(opt);
429 	return ERR_PTR(err);
430 }
431 EXPORT_SYMBOL(ceph_parse_options);
432 
433 u64 ceph_client_id(struct ceph_client *client)
434 {
435 	return client->monc.auth->global_id;
436 }
437 EXPORT_SYMBOL(ceph_client_id);
438 
439 /*
440  * create a fresh client instance
441  */
442 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
443 				       unsigned int supported_features,
444 				       unsigned int required_features)
445 {
446 	struct ceph_client *client;
447 	struct ceph_entity_addr *myaddr = NULL;
448 	int err = -ENOMEM;
449 
450 	client = kzalloc(sizeof(*client), GFP_KERNEL);
451 	if (client == NULL)
452 		return ERR_PTR(-ENOMEM);
453 
454 	client->private = private;
455 	client->options = opt;
456 
457 	mutex_init(&client->mount_mutex);
458 	init_waitqueue_head(&client->auth_wq);
459 	client->auth_err = 0;
460 
461 	client->extra_mon_dispatch = NULL;
462 	client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
463 		supported_features;
464 	client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
465 		required_features;
466 
467 	/* msgr */
468 	if (ceph_test_opt(client, MYIP))
469 		myaddr = &client->options->my_addr;
470 	ceph_messenger_init(&client->msgr, myaddr,
471 		client->supported_features,
472 		client->required_features,
473 		ceph_test_opt(client, NOCRC));
474 
475 	/* subsystems */
476 	err = ceph_monc_init(&client->monc, client);
477 	if (err < 0)
478 		goto fail;
479 	err = ceph_osdc_init(&client->osdc, client);
480 	if (err < 0)
481 		goto fail_monc;
482 
483 	return client;
484 
485 fail_monc:
486 	ceph_monc_stop(&client->monc);
487 fail:
488 	kfree(client);
489 	return ERR_PTR(err);
490 }
491 EXPORT_SYMBOL(ceph_create_client);
492 
493 void ceph_destroy_client(struct ceph_client *client)
494 {
495 	dout("destroy_client %p\n", client);
496 
497 	atomic_set(&client->msgr.stopping, 1);
498 
499 	/* unmount */
500 	ceph_osdc_stop(&client->osdc);
501 
502 	ceph_monc_stop(&client->monc);
503 
504 	ceph_debugfs_client_cleanup(client);
505 
506 	ceph_destroy_options(client->options);
507 
508 	kfree(client);
509 	dout("destroy_client %p done\n", client);
510 }
511 EXPORT_SYMBOL(ceph_destroy_client);
512 
513 /*
514  * true if we have the mon map (and have thus joined the cluster)
515  */
516 static int have_mon_and_osd_map(struct ceph_client *client)
517 {
518 	return client->monc.monmap && client->monc.monmap->epoch &&
519 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
520 }
521 
522 /*
523  * mount: join the ceph cluster, and open root directory.
524  */
525 int __ceph_open_session(struct ceph_client *client, unsigned long started)
526 {
527 	int err;
528 	unsigned long timeout = client->options->mount_timeout * HZ;
529 
530 	/* open session, and wait for mon and osd maps */
531 	err = ceph_monc_open_session(&client->monc);
532 	if (err < 0)
533 		return err;
534 
535 	while (!have_mon_and_osd_map(client)) {
536 		err = -EIO;
537 		if (timeout && time_after_eq(jiffies, started + timeout))
538 			return err;
539 
540 		/* wait */
541 		dout("mount waiting for mon_map\n");
542 		err = wait_event_interruptible_timeout(client->auth_wq,
543 			have_mon_and_osd_map(client) || (client->auth_err < 0),
544 			timeout);
545 		if (err == -EINTR || err == -ERESTARTSYS)
546 			return err;
547 		if (client->auth_err < 0)
548 			return client->auth_err;
549 	}
550 
551 	return 0;
552 }
553 EXPORT_SYMBOL(__ceph_open_session);
554 
555 
556 int ceph_open_session(struct ceph_client *client)
557 {
558 	int ret;
559 	unsigned long started = jiffies;  /* note the start time */
560 
561 	dout("open_session start\n");
562 	mutex_lock(&client->mount_mutex);
563 
564 	ret = __ceph_open_session(client, started);
565 
566 	mutex_unlock(&client->mount_mutex);
567 	return ret;
568 }
569 EXPORT_SYMBOL(ceph_open_session);
570 
571 
572 static int __init init_ceph_lib(void)
573 {
574 	int ret = 0;
575 
576 	ret = ceph_debugfs_init();
577 	if (ret < 0)
578 		goto out;
579 
580 	ret = ceph_crypto_init();
581 	if (ret < 0)
582 		goto out_debugfs;
583 
584 	ret = ceph_msgr_init();
585 	if (ret < 0)
586 		goto out_crypto;
587 
588 	pr_info("loaded (mon/osd proto %d/%d, osdmap %d/%d %d/%d)\n",
589 		CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL,
590 		CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
591 		CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
592 
593 	return 0;
594 
595 out_crypto:
596 	ceph_crypto_shutdown();
597 out_debugfs:
598 	ceph_debugfs_cleanup();
599 out:
600 	return ret;
601 }
602 
603 static void __exit exit_ceph_lib(void)
604 {
605 	dout("exit_ceph_lib\n");
606 	ceph_msgr_exit();
607 	ceph_crypto_shutdown();
608 	ceph_debugfs_cleanup();
609 }
610 
611 module_init(init_ceph_lib);
612 module_exit(exit_ceph_lib);
613 
614 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
615 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
616 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
617 MODULE_DESCRIPTION("Ceph filesystem for Linux");
618 MODULE_LICENSE("GPL");
619