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