xref: /openbmc/linux/net/ceph/ceph_common.c (revision 1cad78932a0d139dceff78e68808e160a224d57a)
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/libceph.h>
21 #include <linux/ceph/debugfs.h>
22 #include <linux/ceph/decode.h>
23 #include <linux/ceph/mon_client.h>
24 #include <linux/ceph/auth.h>
25 #include "crypto.h"
26 
27 
28 
29 /*
30  * find filename portion of a path (/foo/bar/baz -> baz)
31  */
32 const char *ceph_file_part(const char *s, int len)
33 {
34 	const char *e = s + len;
35 
36 	while (e != s && *(e-1) != '/')
37 		e--;
38 	return e;
39 }
40 EXPORT_SYMBOL(ceph_file_part);
41 
42 const char *ceph_msg_type_name(int type)
43 {
44 	switch (type) {
45 	case CEPH_MSG_SHUTDOWN: return "shutdown";
46 	case CEPH_MSG_PING: return "ping";
47 	case CEPH_MSG_AUTH: return "auth";
48 	case CEPH_MSG_AUTH_REPLY: return "auth_reply";
49 	case CEPH_MSG_MON_MAP: return "mon_map";
50 	case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
51 	case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
52 	case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
53 	case CEPH_MSG_STATFS: return "statfs";
54 	case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
55 	case CEPH_MSG_MDS_MAP: return "mds_map";
56 	case CEPH_MSG_CLIENT_SESSION: return "client_session";
57 	case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
58 	case CEPH_MSG_CLIENT_REQUEST: return "client_request";
59 	case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
60 	case CEPH_MSG_CLIENT_REPLY: return "client_reply";
61 	case CEPH_MSG_CLIENT_CAPS: return "client_caps";
62 	case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
63 	case CEPH_MSG_CLIENT_SNAP: return "client_snap";
64 	case CEPH_MSG_CLIENT_LEASE: return "client_lease";
65 	case CEPH_MSG_OSD_MAP: return "osd_map";
66 	case CEPH_MSG_OSD_OP: return "osd_op";
67 	case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
68 	case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
69 	default: return "unknown";
70 	}
71 }
72 EXPORT_SYMBOL(ceph_msg_type_name);
73 
74 /*
75  * Initially learn our fsid, or verify an fsid matches.
76  */
77 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
78 {
79 	if (client->have_fsid) {
80 		if (ceph_fsid_compare(&client->fsid, fsid)) {
81 			pr_err("bad fsid, had %pU got %pU",
82 			       &client->fsid, fsid);
83 			return -1;
84 		}
85 	} else {
86 		pr_info("client%lld fsid %pU\n", ceph_client_id(client), fsid);
87 		memcpy(&client->fsid, fsid, sizeof(*fsid));
88 		ceph_debugfs_client_init(client);
89 		client->have_fsid = true;
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_noshare,
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_noshare, "noshare"},
223 	{Opt_nocrc, "nocrc"},
224 	{-1, NULL}
225 };
226 
227 void ceph_destroy_options(struct ceph_options *opt)
228 {
229 	dout("destroy_options %p\n", opt);
230 	kfree(opt->name);
231 	if (opt->key) {
232 		ceph_crypto_key_destroy(opt->key);
233 		kfree(opt->key);
234 	}
235 	kfree(opt->mon_addr);
236 	kfree(opt);
237 }
238 EXPORT_SYMBOL(ceph_destroy_options);
239 
240 /* get secret from key store */
241 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
242 	struct key *ukey;
243 	int key_err;
244 	int err = 0;
245 	struct ceph_crypto_key *ckey;
246 
247 	ukey = request_key(&key_type_ceph, name, NULL);
248 	if (!ukey || IS_ERR(ukey)) {
249 		/* request_key errors don't map nicely to mount(2)
250 		   errors; don't even try, but still printk */
251 		key_err = PTR_ERR(ukey);
252 		switch (key_err) {
253 		case -ENOKEY:
254 			pr_warning("ceph: Mount failed due to key not found: %s\n", name);
255 			break;
256 		case -EKEYEXPIRED:
257 			pr_warning("ceph: Mount failed due to expired key: %s\n", name);
258 			break;
259 		case -EKEYREVOKED:
260 			pr_warning("ceph: Mount failed due to revoked key: %s\n", name);
261 			break;
262 		default:
263 			pr_warning("ceph: Mount failed due to unknown key error"
264 			       " %d: %s\n", key_err, name);
265 		}
266 		err = -EPERM;
267 		goto out;
268 	}
269 
270 	ckey = ukey->payload.data;
271 	err = ceph_crypto_key_clone(dst, ckey);
272 	if (err)
273 		goto out_key;
274 	/* pass through, err is 0 */
275 
276 out_key:
277 	key_put(ukey);
278 out:
279 	return err;
280 }
281 
282 int ceph_parse_options(struct ceph_options **popt, char *options,
283 		       const char *dev_name, const char *dev_name_end,
284 		       int (*parse_extra_token)(char *c, void *private),
285 		       void *private)
286 {
287 	struct ceph_options *opt;
288 	const char *c;
289 	int err = -ENOMEM;
290 	substring_t argstr[MAX_OPT_ARGS];
291 
292 	opt = kzalloc(sizeof(*opt), GFP_KERNEL);
293 	if (!opt)
294 		return err;
295 	opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
296 				GFP_KERNEL);
297 	if (!opt->mon_addr)
298 		goto out;
299 
300 	dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
301 	     dev_name);
302 
303 	/* start with defaults */
304 	opt->flags = CEPH_OPT_DEFAULT;
305 	opt->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
306 	opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
307 	opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
308 	opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
309 
310 	/* get mon ip(s) */
311 	/* ip1[:port1][,ip2[:port2]...] */
312 	err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
313 			     CEPH_MAX_MON, &opt->num_mon);
314 	if (err < 0)
315 		goto out;
316 
317 	/* parse mount options */
318 	while ((c = strsep(&options, ",")) != NULL) {
319 		int token, intval, ret;
320 		if (!*c)
321 			continue;
322 		err = -EINVAL;
323 		token = match_token((char *)c, opt_tokens, argstr);
324 		if (token < 0 && parse_extra_token) {
325 			/* extra? */
326 			err = parse_extra_token((char *)c, private);
327 			if (err < 0) {
328 				pr_err("bad option at '%s'\n", c);
329 				goto out;
330 			}
331 			continue;
332 		}
333 		if (token < Opt_last_int) {
334 			ret = match_int(&argstr[0], &intval);
335 			if (ret < 0) {
336 				pr_err("bad mount option arg (not int) "
337 				       "at '%s'\n", c);
338 				continue;
339 			}
340 			dout("got int token %d val %d\n", token, intval);
341 		} else if (token > Opt_last_int && token < Opt_last_string) {
342 			dout("got string token %d val %s\n", token,
343 			     argstr[0].from);
344 		} else {
345 			dout("got token %d\n", token);
346 		}
347 		switch (token) {
348 		case Opt_ip:
349 			err = ceph_parse_ips(argstr[0].from,
350 					     argstr[0].to,
351 					     &opt->my_addr,
352 					     1, NULL);
353 			if (err < 0)
354 				goto out;
355 			opt->flags |= CEPH_OPT_MYIP;
356 			break;
357 
358 		case Opt_fsid:
359 			err = parse_fsid(argstr[0].from, &opt->fsid);
360 			if (err == 0)
361 				opt->flags |= CEPH_OPT_FSID;
362 			break;
363 		case Opt_name:
364 			opt->name = kstrndup(argstr[0].from,
365 					      argstr[0].to-argstr[0].from,
366 					      GFP_KERNEL);
367 			break;
368 		case Opt_secret:
369 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
370 			if (!opt->key) {
371 				err = -ENOMEM;
372 				goto out;
373 			}
374 			err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
375 			if (err < 0)
376 				goto out;
377 			break;
378 		case Opt_key:
379 		        opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
380 			if (!opt->key) {
381 				err = -ENOMEM;
382 				goto out;
383 			}
384 			err = get_secret(opt->key, argstr[0].from);
385 			if (err < 0)
386 				goto out;
387 			break;
388 
389 			/* misc */
390 		case Opt_osdtimeout:
391 			opt->osd_timeout = intval;
392 			break;
393 		case Opt_osdkeepalivetimeout:
394 			opt->osd_keepalive_timeout = intval;
395 			break;
396 		case Opt_osd_idle_ttl:
397 			opt->osd_idle_ttl = intval;
398 			break;
399 		case Opt_mount_timeout:
400 			opt->mount_timeout = intval;
401 			break;
402 
403 		case Opt_noshare:
404 			opt->flags |= CEPH_OPT_NOSHARE;
405 			break;
406 
407 		case Opt_nocrc:
408 			opt->flags |= CEPH_OPT_NOCRC;
409 			break;
410 
411 		default:
412 			BUG_ON(token);
413 		}
414 	}
415 
416 	/* success */
417 	*popt = opt;
418 	return 0;
419 
420 out:
421 	ceph_destroy_options(opt);
422 	return err;
423 }
424 EXPORT_SYMBOL(ceph_parse_options);
425 
426 u64 ceph_client_id(struct ceph_client *client)
427 {
428 	return client->monc.auth->global_id;
429 }
430 EXPORT_SYMBOL(ceph_client_id);
431 
432 /*
433  * create a fresh client instance
434  */
435 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
436 {
437 	struct ceph_client *client;
438 	int err = -ENOMEM;
439 
440 	client = kzalloc(sizeof(*client), GFP_KERNEL);
441 	if (client == NULL)
442 		return ERR_PTR(-ENOMEM);
443 
444 	client->private = private;
445 	client->options = opt;
446 
447 	mutex_init(&client->mount_mutex);
448 	init_waitqueue_head(&client->auth_wq);
449 	client->auth_err = 0;
450 
451 	client->extra_mon_dispatch = NULL;
452 	client->supported_features = CEPH_FEATURE_SUPPORTED_DEFAULT;
453 	client->required_features = CEPH_FEATURE_REQUIRED_DEFAULT;
454 
455 	client->msgr = NULL;
456 
457 	/* subsystems */
458 	err = ceph_monc_init(&client->monc, client);
459 	if (err < 0)
460 		goto fail;
461 	err = ceph_osdc_init(&client->osdc, client);
462 	if (err < 0)
463 		goto fail_monc;
464 
465 	return client;
466 
467 fail_monc:
468 	ceph_monc_stop(&client->monc);
469 fail:
470 	kfree(client);
471 	return ERR_PTR(err);
472 }
473 EXPORT_SYMBOL(ceph_create_client);
474 
475 void ceph_destroy_client(struct ceph_client *client)
476 {
477 	dout("destroy_client %p\n", client);
478 
479 	/* unmount */
480 	ceph_osdc_stop(&client->osdc);
481 
482 	/*
483 	 * make sure osd connections close out before destroying the
484 	 * auth module, which is needed to free those connections'
485 	 * ceph_authorizers.
486 	 */
487 	ceph_msgr_flush();
488 
489 	ceph_monc_stop(&client->monc);
490 
491 	ceph_debugfs_client_cleanup(client);
492 
493 	if (client->msgr)
494 		ceph_messenger_destroy(client->msgr);
495 
496 	ceph_destroy_options(client->options);
497 
498 	kfree(client);
499 	dout("destroy_client %p done\n", client);
500 }
501 EXPORT_SYMBOL(ceph_destroy_client);
502 
503 /*
504  * true if we have the mon map (and have thus joined the cluster)
505  */
506 static int have_mon_and_osd_map(struct ceph_client *client)
507 {
508 	return client->monc.monmap && client->monc.monmap->epoch &&
509 	       client->osdc.osdmap && client->osdc.osdmap->epoch;
510 }
511 
512 /*
513  * mount: join the ceph cluster, and open root directory.
514  */
515 int __ceph_open_session(struct ceph_client *client, unsigned long started)
516 {
517 	struct ceph_entity_addr *myaddr = NULL;
518 	int err;
519 	unsigned long timeout = client->options->mount_timeout * HZ;
520 
521 	/* initialize the messenger */
522 	if (client->msgr == NULL) {
523 		if (ceph_test_opt(client, MYIP))
524 			myaddr = &client->options->my_addr;
525 		client->msgr = ceph_messenger_create(myaddr,
526 					client->supported_features,
527 					client->required_features);
528 		if (IS_ERR(client->msgr)) {
529 			client->msgr = NULL;
530 			return PTR_ERR(client->msgr);
531 		}
532 		client->msgr->nocrc = ceph_test_opt(client, NOCRC);
533 	}
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