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 */
ceph_file_part(const char * s,int len)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
ceph_msg_type_name(int type)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 */
ceph_check_fsid(struct ceph_client * client,struct ceph_fsid * fsid)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 }
89 return 0;
90 }
91 EXPORT_SYMBOL(ceph_check_fsid);
92
strcmp_null(const char * s1,const char * s2)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
ceph_compare_options(struct ceph_options * new_opt,struct ceph_client * client)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
parse_fsid(const char * str,struct ceph_fsid * fsid)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_noshare,
205 Opt_nocrc,
206 };
207
208 static match_table_t opt_tokens = {
209 {Opt_osdtimeout, "osdtimeout=%d"},
210 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
211 {Opt_mount_timeout, "mount_timeout=%d"},
212 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
213 /* int args above */
214 {Opt_fsid, "fsid=%s"},
215 {Opt_name, "name=%s"},
216 {Opt_secret, "secret=%s"},
217 {Opt_key, "key=%s"},
218 {Opt_ip, "ip=%s"},
219 /* string args above */
220 {Opt_noshare, "noshare"},
221 {Opt_nocrc, "nocrc"},
222 {-1, NULL}
223 };
224
ceph_destroy_options(struct ceph_options * opt)225 void ceph_destroy_options(struct ceph_options *opt)
226 {
227 dout("destroy_options %p\n", opt);
228 kfree(opt->name);
229 if (opt->key) {
230 ceph_crypto_key_destroy(opt->key);
231 kfree(opt->key);
232 }
233 kfree(opt->mon_addr);
234 kfree(opt);
235 }
236 EXPORT_SYMBOL(ceph_destroy_options);
237
238 /* get secret from key store */
get_secret(struct ceph_crypto_key * dst,const char * name)239 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
240 struct key *ukey;
241 int key_err;
242 int err = 0;
243 struct ceph_crypto_key *ckey;
244
245 ukey = request_key(&key_type_ceph, name, NULL);
246 if (!ukey || IS_ERR(ukey)) {
247 /* request_key errors don't map nicely to mount(2)
248 errors; don't even try, but still printk */
249 key_err = PTR_ERR(ukey);
250 switch (key_err) {
251 case -ENOKEY:
252 pr_warning("ceph: Mount failed due to key not found: %s\n", name);
253 break;
254 case -EKEYEXPIRED:
255 pr_warning("ceph: Mount failed due to expired key: %s\n", name);
256 break;
257 case -EKEYREVOKED:
258 pr_warning("ceph: Mount failed due to revoked key: %s\n", name);
259 break;
260 default:
261 pr_warning("ceph: Mount failed due to unknown key error"
262 " %d: %s\n", key_err, name);
263 }
264 err = -EPERM;
265 goto out;
266 }
267
268 ckey = ukey->payload.data;
269 err = ceph_crypto_key_clone(dst, ckey);
270 if (err)
271 goto out_key;
272 /* pass through, err is 0 */
273
274 out_key:
275 key_put(ukey);
276 out:
277 return err;
278 }
279
ceph_parse_options(struct ceph_options ** popt,char * options,const char * dev_name,const char * dev_name_end,int (* parse_extra_token)(char * c,void * private),void * private)280 int ceph_parse_options(struct ceph_options **popt, char *options,
281 const char *dev_name, const char *dev_name_end,
282 int (*parse_extra_token)(char *c, void *private),
283 void *private)
284 {
285 struct ceph_options *opt;
286 const char *c;
287 int err = -ENOMEM;
288 substring_t argstr[MAX_OPT_ARGS];
289
290 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
291 if (!opt)
292 return err;
293 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
294 GFP_KERNEL);
295 if (!opt->mon_addr)
296 goto out;
297
298 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
299 dev_name);
300
301 /* start with defaults */
302 opt->flags = CEPH_OPT_DEFAULT;
303 opt->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
304 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
305 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
306 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* seconds */
307
308 /* get mon ip(s) */
309 /* ip1[:port1][,ip2[:port2]...] */
310 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
311 CEPH_MAX_MON, &opt->num_mon);
312 if (err < 0)
313 goto out;
314
315 /* parse mount options */
316 while ((c = strsep(&options, ",")) != NULL) {
317 int token, intval, ret;
318 if (!*c)
319 continue;
320 err = -EINVAL;
321 token = match_token((char *)c, opt_tokens, argstr);
322 if (token < 0 && parse_extra_token) {
323 /* extra? */
324 err = parse_extra_token((char *)c, private);
325 if (err < 0) {
326 pr_err("bad option at '%s'\n", c);
327 goto out;
328 }
329 continue;
330 }
331 if (token < Opt_last_int) {
332 ret = match_int(&argstr[0], &intval);
333 if (ret < 0) {
334 pr_err("bad mount option arg (not int) "
335 "at '%s'\n", c);
336 continue;
337 }
338 dout("got int token %d val %d\n", token, intval);
339 } else if (token > Opt_last_int && token < Opt_last_string) {
340 dout("got string token %d val %s\n", token,
341 argstr[0].from);
342 } else {
343 dout("got token %d\n", token);
344 }
345 switch (token) {
346 case Opt_ip:
347 err = ceph_parse_ips(argstr[0].from,
348 argstr[0].to,
349 &opt->my_addr,
350 1, NULL);
351 if (err < 0)
352 goto out;
353 opt->flags |= CEPH_OPT_MYIP;
354 break;
355
356 case Opt_fsid:
357 err = parse_fsid(argstr[0].from, &opt->fsid);
358 if (err == 0)
359 opt->flags |= CEPH_OPT_FSID;
360 break;
361 case Opt_name:
362 opt->name = kstrndup(argstr[0].from,
363 argstr[0].to-argstr[0].from,
364 GFP_KERNEL);
365 break;
366 case Opt_secret:
367 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
368 if (!opt->key) {
369 err = -ENOMEM;
370 goto out;
371 }
372 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
373 if (err < 0)
374 goto out;
375 break;
376 case Opt_key:
377 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
378 if (!opt->key) {
379 err = -ENOMEM;
380 goto out;
381 }
382 err = get_secret(opt->key, argstr[0].from);
383 if (err < 0)
384 goto out;
385 break;
386
387 /* misc */
388 case Opt_osdtimeout:
389 opt->osd_timeout = intval;
390 break;
391 case Opt_osdkeepalivetimeout:
392 opt->osd_keepalive_timeout = intval;
393 break;
394 case Opt_osd_idle_ttl:
395 opt->osd_idle_ttl = intval;
396 break;
397 case Opt_mount_timeout:
398 opt->mount_timeout = intval;
399 break;
400
401 case Opt_noshare:
402 opt->flags |= CEPH_OPT_NOSHARE;
403 break;
404
405 case Opt_nocrc:
406 opt->flags |= CEPH_OPT_NOCRC;
407 break;
408
409 default:
410 BUG_ON(token);
411 }
412 }
413
414 /* success */
415 *popt = opt;
416 return 0;
417
418 out:
419 ceph_destroy_options(opt);
420 return err;
421 }
422 EXPORT_SYMBOL(ceph_parse_options);
423
ceph_client_id(struct ceph_client * client)424 u64 ceph_client_id(struct ceph_client *client)
425 {
426 return client->monc.auth->global_id;
427 }
428 EXPORT_SYMBOL(ceph_client_id);
429
430 /*
431 * create a fresh client instance
432 */
ceph_create_client(struct ceph_options * opt,void * private,unsigned supported_features,unsigned required_features)433 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
434 unsigned supported_features,
435 unsigned required_features)
436 {
437 struct ceph_client *client;
438 struct ceph_entity_addr *myaddr = NULL;
439 int err = -ENOMEM;
440
441 client = kzalloc(sizeof(*client), GFP_KERNEL);
442 if (client == NULL)
443 return ERR_PTR(-ENOMEM);
444
445 client->private = private;
446 client->options = opt;
447
448 mutex_init(&client->mount_mutex);
449 init_waitqueue_head(&client->auth_wq);
450 client->auth_err = 0;
451
452 client->extra_mon_dispatch = NULL;
453 client->supported_features = CEPH_FEATURE_SUPPORTED_DEFAULT |
454 supported_features;
455 client->required_features = CEPH_FEATURE_REQUIRED_DEFAULT |
456 required_features;
457
458 /* msgr */
459 if (ceph_test_opt(client, MYIP))
460 myaddr = &client->options->my_addr;
461 client->msgr = ceph_messenger_create(myaddr,
462 client->supported_features,
463 client->required_features);
464 if (IS_ERR(client->msgr)) {
465 err = PTR_ERR(client->msgr);
466 goto fail;
467 }
468 client->msgr->nocrc = ceph_test_opt(client, NOCRC);
469
470 /* subsystems */
471 err = ceph_monc_init(&client->monc, client);
472 if (err < 0)
473 goto fail_msgr;
474 err = ceph_osdc_init(&client->osdc, client);
475 if (err < 0)
476 goto fail_monc;
477
478 return client;
479
480 fail_monc:
481 ceph_monc_stop(&client->monc);
482 fail_msgr:
483 ceph_messenger_destroy(client->msgr);
484 fail:
485 kfree(client);
486 return ERR_PTR(err);
487 }
488 EXPORT_SYMBOL(ceph_create_client);
489
ceph_destroy_client(struct ceph_client * client)490 void ceph_destroy_client(struct ceph_client *client)
491 {
492 dout("destroy_client %p\n", client);
493
494 /* unmount */
495 ceph_osdc_stop(&client->osdc);
496
497 /*
498 * make sure osd connections close out before destroying the
499 * auth module, which is needed to free those connections'
500 * ceph_authorizers.
501 */
502 ceph_msgr_flush();
503
504 ceph_monc_stop(&client->monc);
505
506 ceph_debugfs_client_cleanup(client);
507
508 ceph_messenger_destroy(client->msgr);
509
510 ceph_destroy_options(client->options);
511
512 kfree(client);
513 dout("destroy_client %p done\n", client);
514 }
515 EXPORT_SYMBOL(ceph_destroy_client);
516
517 /*
518 * true if we have the mon map (and have thus joined the cluster)
519 */
have_mon_and_osd_map(struct ceph_client * client)520 static int have_mon_and_osd_map(struct ceph_client *client)
521 {
522 return client->monc.monmap && client->monc.monmap->epoch &&
523 client->osdc.osdmap && client->osdc.osdmap->epoch;
524 }
525
526 /*
527 * mount: join the ceph cluster, and open root directory.
528 */
__ceph_open_session(struct ceph_client * client,unsigned long started)529 int __ceph_open_session(struct ceph_client *client, unsigned long started)
530 {
531 int err;
532 unsigned long timeout = client->options->mount_timeout * HZ;
533
534 /* open session, and wait for mon and osd maps */
535 err = ceph_monc_open_session(&client->monc);
536 if (err < 0)
537 return err;
538
539 while (!have_mon_and_osd_map(client)) {
540 err = -EIO;
541 if (timeout && time_after_eq(jiffies, started + timeout))
542 return err;
543
544 /* wait */
545 dout("mount waiting for mon_map\n");
546 err = wait_event_interruptible_timeout(client->auth_wq,
547 have_mon_and_osd_map(client) || (client->auth_err < 0),
548 timeout);
549 if (err == -EINTR || err == -ERESTARTSYS)
550 return err;
551 if (client->auth_err < 0)
552 return client->auth_err;
553 }
554
555 return 0;
556 }
557 EXPORT_SYMBOL(__ceph_open_session);
558
559
ceph_open_session(struct ceph_client * client)560 int ceph_open_session(struct ceph_client *client)
561 {
562 int ret;
563 unsigned long started = jiffies; /* note the start time */
564
565 dout("open_session start\n");
566 mutex_lock(&client->mount_mutex);
567
568 ret = __ceph_open_session(client, started);
569
570 mutex_unlock(&client->mount_mutex);
571 return ret;
572 }
573 EXPORT_SYMBOL(ceph_open_session);
574
575
init_ceph_lib(void)576 static int __init init_ceph_lib(void)
577 {
578 int ret = 0;
579
580 ret = ceph_debugfs_init();
581 if (ret < 0)
582 goto out;
583
584 ret = ceph_crypto_init();
585 if (ret < 0)
586 goto out_debugfs;
587
588 ret = ceph_msgr_init();
589 if (ret < 0)
590 goto out_crypto;
591
592 pr_info("loaded (mon/osd proto %d/%d, osdmap %d/%d %d/%d)\n",
593 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL,
594 CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
595 CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
596
597 return 0;
598
599 out_crypto:
600 ceph_crypto_shutdown();
601 out_debugfs:
602 ceph_debugfs_cleanup();
603 out:
604 return ret;
605 }
606
exit_ceph_lib(void)607 static void __exit exit_ceph_lib(void)
608 {
609 dout("exit_ceph_lib\n");
610 ceph_msgr_exit();
611 ceph_crypto_shutdown();
612 ceph_debugfs_cleanup();
613 }
614
615 module_init(init_ceph_lib);
616 module_exit(exit_ceph_lib);
617
618 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
619 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
620 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
621 MODULE_DESCRIPTION("Ceph filesystem for Linux");
622 MODULE_LICENSE("GPL");
623