]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - net/ceph/ceph_common.c
libceph: nuke ceph_kvfree()
[mirror_ubuntu-zesty-kernel.git] / net / ceph / ceph_common.c
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 #include <linux/vmalloc.h>
19 #include <linux/nsproxy.h>
20 #include <net/net_namespace.h>
21
22
23 #include <linux/ceph/ceph_features.h>
24 #include <linux/ceph/libceph.h>
25 #include <linux/ceph/debugfs.h>
26 #include <linux/ceph/decode.h>
27 #include <linux/ceph/mon_client.h>
28 #include <linux/ceph/auth.h>
29 #include "crypto.h"
30
31
32 /*
33 * Module compatibility interface. For now it doesn't do anything,
34 * but its existence signals a certain level of functionality.
35 *
36 * The data buffer is used to pass information both to and from
37 * libceph. The return value indicates whether libceph determines
38 * it is compatible with the caller (from another kernel module),
39 * given the provided data.
40 *
41 * The data pointer can be null.
42 */
43 bool libceph_compatible(void *data)
44 {
45 return true;
46 }
47 EXPORT_SYMBOL(libceph_compatible);
48
49 /*
50 * find filename portion of a path (/foo/bar/baz -> baz)
51 */
52 const char *ceph_file_part(const char *s, int len)
53 {
54 const char *e = s + len;
55
56 while (e != s && *(e-1) != '/')
57 e--;
58 return e;
59 }
60 EXPORT_SYMBOL(ceph_file_part);
61
62 const char *ceph_msg_type_name(int type)
63 {
64 switch (type) {
65 case CEPH_MSG_SHUTDOWN: return "shutdown";
66 case CEPH_MSG_PING: return "ping";
67 case CEPH_MSG_AUTH: return "auth";
68 case CEPH_MSG_AUTH_REPLY: return "auth_reply";
69 case CEPH_MSG_MON_MAP: return "mon_map";
70 case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
71 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
72 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
73 case CEPH_MSG_STATFS: return "statfs";
74 case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
75 case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
76 case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
77 case CEPH_MSG_MDS_MAP: return "mds_map";
78 case CEPH_MSG_CLIENT_SESSION: return "client_session";
79 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
80 case CEPH_MSG_CLIENT_REQUEST: return "client_request";
81 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
82 case CEPH_MSG_CLIENT_REPLY: return "client_reply";
83 case CEPH_MSG_CLIENT_CAPS: return "client_caps";
84 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
85 case CEPH_MSG_CLIENT_SNAP: return "client_snap";
86 case CEPH_MSG_CLIENT_LEASE: return "client_lease";
87 case CEPH_MSG_OSD_MAP: return "osd_map";
88 case CEPH_MSG_OSD_OP: return "osd_op";
89 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
90 case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
91 default: return "unknown";
92 }
93 }
94 EXPORT_SYMBOL(ceph_msg_type_name);
95
96 /*
97 * Initially learn our fsid, or verify an fsid matches.
98 */
99 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
100 {
101 if (client->have_fsid) {
102 if (ceph_fsid_compare(&client->fsid, fsid)) {
103 pr_err("bad fsid, had %pU got %pU",
104 &client->fsid, fsid);
105 return -1;
106 }
107 } else {
108 memcpy(&client->fsid, fsid, sizeof(*fsid));
109 }
110 return 0;
111 }
112 EXPORT_SYMBOL(ceph_check_fsid);
113
114 static int strcmp_null(const char *s1, const char *s2)
115 {
116 if (!s1 && !s2)
117 return 0;
118 if (s1 && !s2)
119 return -1;
120 if (!s1 && s2)
121 return 1;
122 return strcmp(s1, s2);
123 }
124
125 int ceph_compare_options(struct ceph_options *new_opt,
126 struct ceph_client *client)
127 {
128 struct ceph_options *opt1 = new_opt;
129 struct ceph_options *opt2 = client->options;
130 int ofs = offsetof(struct ceph_options, mon_addr);
131 int i;
132 int ret;
133
134 ret = memcmp(opt1, opt2, ofs);
135 if (ret)
136 return ret;
137
138 ret = strcmp_null(opt1->name, opt2->name);
139 if (ret)
140 return ret;
141
142 if (opt1->key && !opt2->key)
143 return -1;
144 if (!opt1->key && opt2->key)
145 return 1;
146 if (opt1->key && opt2->key) {
147 if (opt1->key->type != opt2->key->type)
148 return -1;
149 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
150 return -1;
151 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
152 return -1;
153 if (opt1->key->len != opt2->key->len)
154 return -1;
155 if (opt1->key->key && !opt2->key->key)
156 return -1;
157 if (!opt1->key->key && opt2->key->key)
158 return 1;
159 if (opt1->key->key && opt2->key->key) {
160 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
161 if (ret)
162 return ret;
163 }
164 }
165
166 /* any matching mon ip implies a match */
167 for (i = 0; i < opt1->num_mon; i++) {
168 if (ceph_monmap_contains(client->monc.monmap,
169 &opt1->mon_addr[i]))
170 return 0;
171 }
172 return -1;
173 }
174 EXPORT_SYMBOL(ceph_compare_options);
175
176 void *ceph_kvmalloc(size_t size, gfp_t flags)
177 {
178 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
179 void *ptr = kmalloc(size, flags | __GFP_NOWARN);
180 if (ptr)
181 return ptr;
182 }
183
184 return __vmalloc(size, flags | __GFP_HIGHMEM, PAGE_KERNEL);
185 }
186
187
188 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
189 {
190 int i = 0;
191 char tmp[3];
192 int err = -EINVAL;
193 int d;
194
195 dout("parse_fsid '%s'\n", str);
196 tmp[2] = 0;
197 while (*str && i < 16) {
198 if (ispunct(*str)) {
199 str++;
200 continue;
201 }
202 if (!isxdigit(str[0]) || !isxdigit(str[1]))
203 break;
204 tmp[0] = str[0];
205 tmp[1] = str[1];
206 if (sscanf(tmp, "%x", &d) < 1)
207 break;
208 fsid->fsid[i] = d & 0xff;
209 i++;
210 str += 2;
211 }
212
213 if (i == 16)
214 err = 0;
215 dout("parse_fsid ret %d got fsid %pU", err, fsid);
216 return err;
217 }
218
219 /*
220 * ceph options
221 */
222 enum {
223 Opt_osdtimeout,
224 Opt_osdkeepalivetimeout,
225 Opt_mount_timeout,
226 Opt_osd_idle_ttl,
227 Opt_last_int,
228 /* int args above */
229 Opt_fsid,
230 Opt_name,
231 Opt_secret,
232 Opt_key,
233 Opt_ip,
234 Opt_last_string,
235 /* string args above */
236 Opt_share,
237 Opt_noshare,
238 Opt_crc,
239 Opt_nocrc,
240 };
241
242 static match_table_t opt_tokens = {
243 {Opt_osdtimeout, "osdtimeout=%d"},
244 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
245 {Opt_mount_timeout, "mount_timeout=%d"},
246 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
247 /* int args above */
248 {Opt_fsid, "fsid=%s"},
249 {Opt_name, "name=%s"},
250 {Opt_secret, "secret=%s"},
251 {Opt_key, "key=%s"},
252 {Opt_ip, "ip=%s"},
253 /* string args above */
254 {Opt_share, "share"},
255 {Opt_noshare, "noshare"},
256 {Opt_crc, "crc"},
257 {Opt_nocrc, "nocrc"},
258 {-1, NULL}
259 };
260
261 void ceph_destroy_options(struct ceph_options *opt)
262 {
263 dout("destroy_options %p\n", opt);
264 kfree(opt->name);
265 if (opt->key) {
266 ceph_crypto_key_destroy(opt->key);
267 kfree(opt->key);
268 }
269 kfree(opt->mon_addr);
270 kfree(opt);
271 }
272 EXPORT_SYMBOL(ceph_destroy_options);
273
274 /* get secret from key store */
275 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
276 struct key *ukey;
277 int key_err;
278 int err = 0;
279 struct ceph_crypto_key *ckey;
280
281 ukey = request_key(&key_type_ceph, name, NULL);
282 if (!ukey || IS_ERR(ukey)) {
283 /* request_key errors don't map nicely to mount(2)
284 errors; don't even try, but still printk */
285 key_err = PTR_ERR(ukey);
286 switch (key_err) {
287 case -ENOKEY:
288 pr_warn("ceph: Mount failed due to key not found: %s\n",
289 name);
290 break;
291 case -EKEYEXPIRED:
292 pr_warn("ceph: Mount failed due to expired key: %s\n",
293 name);
294 break;
295 case -EKEYREVOKED:
296 pr_warn("ceph: Mount failed due to revoked key: %s\n",
297 name);
298 break;
299 default:
300 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
301 key_err, name);
302 }
303 err = -EPERM;
304 goto out;
305 }
306
307 ckey = ukey->payload.data;
308 err = ceph_crypto_key_clone(dst, ckey);
309 if (err)
310 goto out_key;
311 /* pass through, err is 0 */
312
313 out_key:
314 key_put(ukey);
315 out:
316 return err;
317 }
318
319 struct ceph_options *
320 ceph_parse_options(char *options, const char *dev_name,
321 const char *dev_name_end,
322 int (*parse_extra_token)(char *c, void *private),
323 void *private)
324 {
325 struct ceph_options *opt;
326 const char *c;
327 int err = -ENOMEM;
328 substring_t argstr[MAX_OPT_ARGS];
329
330 if (current->nsproxy->net_ns != &init_net)
331 return ERR_PTR(-EINVAL);
332
333 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
334 if (!opt)
335 return ERR_PTR(-ENOMEM);
336 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
337 GFP_KERNEL);
338 if (!opt->mon_addr)
339 goto out;
340
341 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
342 dev_name);
343
344 /* start with defaults */
345 opt->flags = CEPH_OPT_DEFAULT;
346 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
347 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
348 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* seconds */
349
350 /* get mon ip(s) */
351 /* ip1[:port1][,ip2[:port2]...] */
352 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
353 CEPH_MAX_MON, &opt->num_mon);
354 if (err < 0)
355 goto out;
356
357 /* parse mount options */
358 while ((c = strsep(&options, ",")) != NULL) {
359 int token, intval, ret;
360 if (!*c)
361 continue;
362 err = -EINVAL;
363 token = match_token((char *)c, opt_tokens, argstr);
364 if (token < 0 && parse_extra_token) {
365 /* extra? */
366 err = parse_extra_token((char *)c, private);
367 if (err < 0) {
368 pr_err("bad option at '%s'\n", c);
369 goto out;
370 }
371 continue;
372 }
373 if (token < Opt_last_int) {
374 ret = match_int(&argstr[0], &intval);
375 if (ret < 0) {
376 pr_err("bad mount option arg (not int) "
377 "at '%s'\n", c);
378 continue;
379 }
380 dout("got int token %d val %d\n", token, intval);
381 } else if (token > Opt_last_int && token < Opt_last_string) {
382 dout("got string token %d val %s\n", token,
383 argstr[0].from);
384 } else {
385 dout("got token %d\n", token);
386 }
387 switch (token) {
388 case Opt_ip:
389 err = ceph_parse_ips(argstr[0].from,
390 argstr[0].to,
391 &opt->my_addr,
392 1, NULL);
393 if (err < 0)
394 goto out;
395 opt->flags |= CEPH_OPT_MYIP;
396 break;
397
398 case Opt_fsid:
399 err = parse_fsid(argstr[0].from, &opt->fsid);
400 if (err == 0)
401 opt->flags |= CEPH_OPT_FSID;
402 break;
403 case Opt_name:
404 opt->name = kstrndup(argstr[0].from,
405 argstr[0].to-argstr[0].from,
406 GFP_KERNEL);
407 break;
408 case Opt_secret:
409 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
410 if (!opt->key) {
411 err = -ENOMEM;
412 goto out;
413 }
414 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
415 if (err < 0)
416 goto out;
417 break;
418 case Opt_key:
419 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
420 if (!opt->key) {
421 err = -ENOMEM;
422 goto out;
423 }
424 err = get_secret(opt->key, argstr[0].from);
425 if (err < 0)
426 goto out;
427 break;
428
429 /* misc */
430 case Opt_osdtimeout:
431 pr_warn("ignoring deprecated osdtimeout option\n");
432 break;
433 case Opt_osdkeepalivetimeout:
434 opt->osd_keepalive_timeout = intval;
435 break;
436 case Opt_osd_idle_ttl:
437 opt->osd_idle_ttl = intval;
438 break;
439 case Opt_mount_timeout:
440 opt->mount_timeout = intval;
441 break;
442
443 case Opt_share:
444 opt->flags &= ~CEPH_OPT_NOSHARE;
445 break;
446 case Opt_noshare:
447 opt->flags |= CEPH_OPT_NOSHARE;
448 break;
449
450 case Opt_crc:
451 opt->flags &= ~CEPH_OPT_NOCRC;
452 break;
453 case Opt_nocrc:
454 opt->flags |= CEPH_OPT_NOCRC;
455 break;
456
457 default:
458 BUG_ON(token);
459 }
460 }
461
462 /* success */
463 return opt;
464
465 out:
466 ceph_destroy_options(opt);
467 return ERR_PTR(err);
468 }
469 EXPORT_SYMBOL(ceph_parse_options);
470
471 u64 ceph_client_id(struct ceph_client *client)
472 {
473 return client->monc.auth->global_id;
474 }
475 EXPORT_SYMBOL(ceph_client_id);
476
477 /*
478 * create a fresh client instance
479 */
480 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
481 u64 supported_features,
482 u64 required_features)
483 {
484 struct ceph_client *client;
485 struct ceph_entity_addr *myaddr = NULL;
486 int err = -ENOMEM;
487
488 client = kzalloc(sizeof(*client), GFP_KERNEL);
489 if (client == NULL)
490 return ERR_PTR(-ENOMEM);
491
492 client->private = private;
493 client->options = opt;
494
495 mutex_init(&client->mount_mutex);
496 init_waitqueue_head(&client->auth_wq);
497 client->auth_err = 0;
498
499 client->extra_mon_dispatch = NULL;
500 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
501 supported_features;
502 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
503 required_features;
504
505 /* msgr */
506 if (ceph_test_opt(client, MYIP))
507 myaddr = &client->options->my_addr;
508 ceph_messenger_init(&client->msgr, myaddr,
509 client->supported_features,
510 client->required_features,
511 ceph_test_opt(client, NOCRC));
512
513 /* subsystems */
514 err = ceph_monc_init(&client->monc, client);
515 if (err < 0)
516 goto fail;
517 err = ceph_osdc_init(&client->osdc, client);
518 if (err < 0)
519 goto fail_monc;
520
521 return client;
522
523 fail_monc:
524 ceph_monc_stop(&client->monc);
525 fail:
526 kfree(client);
527 return ERR_PTR(err);
528 }
529 EXPORT_SYMBOL(ceph_create_client);
530
531 void ceph_destroy_client(struct ceph_client *client)
532 {
533 dout("destroy_client %p\n", client);
534
535 atomic_set(&client->msgr.stopping, 1);
536
537 /* unmount */
538 ceph_osdc_stop(&client->osdc);
539
540 ceph_monc_stop(&client->monc);
541
542 ceph_debugfs_client_cleanup(client);
543
544 ceph_destroy_options(client->options);
545
546 kfree(client);
547 dout("destroy_client %p done\n", client);
548 }
549 EXPORT_SYMBOL(ceph_destroy_client);
550
551 /*
552 * true if we have the mon map (and have thus joined the cluster)
553 */
554 static int have_mon_and_osd_map(struct ceph_client *client)
555 {
556 return client->monc.monmap && client->monc.monmap->epoch &&
557 client->osdc.osdmap && client->osdc.osdmap->epoch;
558 }
559
560 /*
561 * mount: join the ceph cluster, and open root directory.
562 */
563 int __ceph_open_session(struct ceph_client *client, unsigned long started)
564 {
565 int err;
566 unsigned long timeout = client->options->mount_timeout * HZ;
567
568 /* open session, and wait for mon and osd maps */
569 err = ceph_monc_open_session(&client->monc);
570 if (err < 0)
571 return err;
572
573 while (!have_mon_and_osd_map(client)) {
574 err = -EIO;
575 if (timeout && time_after_eq(jiffies, started + timeout))
576 return err;
577
578 /* wait */
579 dout("mount waiting for mon_map\n");
580 err = wait_event_interruptible_timeout(client->auth_wq,
581 have_mon_and_osd_map(client) || (client->auth_err < 0),
582 timeout);
583 if (err == -EINTR || err == -ERESTARTSYS)
584 return err;
585 if (client->auth_err < 0)
586 return client->auth_err;
587 }
588
589 return 0;
590 }
591 EXPORT_SYMBOL(__ceph_open_session);
592
593
594 int ceph_open_session(struct ceph_client *client)
595 {
596 int ret;
597 unsigned long started = jiffies; /* note the start time */
598
599 dout("open_session start\n");
600 mutex_lock(&client->mount_mutex);
601
602 ret = __ceph_open_session(client, started);
603
604 mutex_unlock(&client->mount_mutex);
605 return ret;
606 }
607 EXPORT_SYMBOL(ceph_open_session);
608
609
610 static int __init init_ceph_lib(void)
611 {
612 int ret = 0;
613
614 ret = ceph_debugfs_init();
615 if (ret < 0)
616 goto out;
617
618 ret = ceph_crypto_init();
619 if (ret < 0)
620 goto out_debugfs;
621
622 ret = ceph_msgr_init();
623 if (ret < 0)
624 goto out_crypto;
625
626 ret = ceph_osdc_setup();
627 if (ret < 0)
628 goto out_msgr;
629
630 pr_info("loaded (mon/osd proto %d/%d)\n",
631 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
632
633 return 0;
634
635 out_msgr:
636 ceph_msgr_exit();
637 out_crypto:
638 ceph_crypto_shutdown();
639 out_debugfs:
640 ceph_debugfs_cleanup();
641 out:
642 return ret;
643 }
644
645 static void __exit exit_ceph_lib(void)
646 {
647 dout("exit_ceph_lib\n");
648 ceph_osdc_cleanup();
649 ceph_msgr_exit();
650 ceph_crypto_shutdown();
651 ceph_debugfs_cleanup();
652 }
653
654 module_init(init_ceph_lib);
655 module_exit(exit_ceph_lib);
656
657 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
658 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
659 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
660 MODULE_DESCRIPTION("Ceph filesystem for Linux");
661 MODULE_LICENSE("GPL");