Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
[linux-2.6.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/module.h>
9 #include <linux/mount.h>
10 #include <linux/parser.h>
11 #include <linux/sched.h>
12 #include <linux/seq_file.h>
13 #include <linux/slab.h>
14 #include <linux/statfs.h>
15 #include <linux/string.h>
16
17
18 #include <linux/ceph/libceph.h>
19 #include <linux/ceph/debugfs.h>
20 #include <linux/ceph/decode.h>
21 #include <linux/ceph/mon_client.h>
22 #include <linux/ceph/auth.h>
23
24
25
26 /*
27  * find filename portion of a path (/foo/bar/baz -> baz)
28  */
29 const char *ceph_file_part(const char *s, int len)
30 {
31         const char *e = s + len;
32
33         while (e != s && *(e-1) != '/')
34                 e--;
35         return e;
36 }
37 EXPORT_SYMBOL(ceph_file_part);
38
39 const char *ceph_msg_type_name(int type)
40 {
41         switch (type) {
42         case CEPH_MSG_SHUTDOWN: return "shutdown";
43         case CEPH_MSG_PING: return "ping";
44         case CEPH_MSG_AUTH: return "auth";
45         case CEPH_MSG_AUTH_REPLY: return "auth_reply";
46         case CEPH_MSG_MON_MAP: return "mon_map";
47         case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
48         case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
49         case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
50         case CEPH_MSG_STATFS: return "statfs";
51         case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
52         case CEPH_MSG_MDS_MAP: return "mds_map";
53         case CEPH_MSG_CLIENT_SESSION: return "client_session";
54         case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
55         case CEPH_MSG_CLIENT_REQUEST: return "client_request";
56         case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
57         case CEPH_MSG_CLIENT_REPLY: return "client_reply";
58         case CEPH_MSG_CLIENT_CAPS: return "client_caps";
59         case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
60         case CEPH_MSG_CLIENT_SNAP: return "client_snap";
61         case CEPH_MSG_CLIENT_LEASE: return "client_lease";
62         case CEPH_MSG_OSD_MAP: return "osd_map";
63         case CEPH_MSG_OSD_OP: return "osd_op";
64         case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
65         default: return "unknown";
66         }
67 }
68 EXPORT_SYMBOL(ceph_msg_type_name);
69
70 /*
71  * Initially learn our fsid, or verify an fsid matches.
72  */
73 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
74 {
75         if (client->have_fsid) {
76                 if (ceph_fsid_compare(&client->fsid, fsid)) {
77                         pr_err("bad fsid, had %pU got %pU",
78                                &client->fsid, fsid);
79                         return -1;
80                 }
81         } else {
82                 pr_info("client%lld fsid %pU\n", ceph_client_id(client), fsid);
83                 memcpy(&client->fsid, fsid, sizeof(*fsid));
84                 ceph_debugfs_client_init(client);
85                 client->have_fsid = true;
86         }
87         return 0;
88 }
89 EXPORT_SYMBOL(ceph_check_fsid);
90
91 static int strcmp_null(const char *s1, const char *s2)
92 {
93         if (!s1 && !s2)
94                 return 0;
95         if (s1 && !s2)
96                 return -1;
97         if (!s1 && s2)
98                 return 1;
99         return strcmp(s1, s2);
100 }
101
102 int ceph_compare_options(struct ceph_options *new_opt,
103                          struct ceph_client *client)
104 {
105         struct ceph_options *opt1 = new_opt;
106         struct ceph_options *opt2 = client->options;
107         int ofs = offsetof(struct ceph_options, mon_addr);
108         int i;
109         int ret;
110
111         ret = memcmp(opt1, opt2, ofs);
112         if (ret)
113                 return ret;
114
115         ret = strcmp_null(opt1->name, opt2->name);
116         if (ret)
117                 return ret;
118
119         ret = strcmp_null(opt1->secret, opt2->secret);
120         if (ret)
121                 return ret;
122
123         /* any matching mon ip implies a match */
124         for (i = 0; i < opt1->num_mon; i++) {
125                 if (ceph_monmap_contains(client->monc.monmap,
126                                  &opt1->mon_addr[i]))
127                         return 0;
128         }
129         return -1;
130 }
131 EXPORT_SYMBOL(ceph_compare_options);
132
133
134 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
135 {
136         int i = 0;
137         char tmp[3];
138         int err = -EINVAL;
139         int d;
140
141         dout("parse_fsid '%s'\n", str);
142         tmp[2] = 0;
143         while (*str && i < 16) {
144                 if (ispunct(*str)) {
145                         str++;
146                         continue;
147                 }
148                 if (!isxdigit(str[0]) || !isxdigit(str[1]))
149                         break;
150                 tmp[0] = str[0];
151                 tmp[1] = str[1];
152                 if (sscanf(tmp, "%x", &d) < 1)
153                         break;
154                 fsid->fsid[i] = d & 0xff;
155                 i++;
156                 str += 2;
157         }
158
159         if (i == 16)
160                 err = 0;
161         dout("parse_fsid ret %d got fsid %pU", err, fsid);
162         return err;
163 }
164
165 /*
166  * ceph options
167  */
168 enum {
169         Opt_osdtimeout,
170         Opt_osdkeepalivetimeout,
171         Opt_mount_timeout,
172         Opt_osd_idle_ttl,
173         Opt_last_int,
174         /* int args above */
175         Opt_fsid,
176         Opt_name,
177         Opt_secret,
178         Opt_ip,
179         Opt_last_string,
180         /* string args above */
181         Opt_noshare,
182         Opt_nocrc,
183 };
184
185 static match_table_t opt_tokens = {
186         {Opt_osdtimeout, "osdtimeout=%d"},
187         {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
188         {Opt_mount_timeout, "mount_timeout=%d"},
189         {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
190         /* int args above */
191         {Opt_fsid, "fsid=%s"},
192         {Opt_name, "name=%s"},
193         {Opt_secret, "secret=%s"},
194         {Opt_ip, "ip=%s"},
195         /* string args above */
196         {Opt_noshare, "noshare"},
197         {Opt_nocrc, "nocrc"},
198         {-1, NULL}
199 };
200
201 void ceph_destroy_options(struct ceph_options *opt)
202 {
203         dout("destroy_options %p\n", opt);
204         kfree(opt->name);
205         kfree(opt->secret);
206         kfree(opt);
207 }
208 EXPORT_SYMBOL(ceph_destroy_options);
209
210 int ceph_parse_options(struct ceph_options **popt, char *options,
211                        const char *dev_name, const char *dev_name_end,
212                        int (*parse_extra_token)(char *c, void *private),
213                        void *private)
214 {
215         struct ceph_options *opt;
216         const char *c;
217         int err = -ENOMEM;
218         substring_t argstr[MAX_OPT_ARGS];
219
220         opt = kzalloc(sizeof(*opt), GFP_KERNEL);
221         if (!opt)
222                 return err;
223         opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
224                                 GFP_KERNEL);
225         if (!opt->mon_addr)
226                 goto out;
227
228         dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
229              dev_name);
230
231         /* start with defaults */
232         opt->flags = CEPH_OPT_DEFAULT;
233         opt->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
234         opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
235         opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
236         opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
237
238         /* get mon ip(s) */
239         /* ip1[:port1][,ip2[:port2]...] */
240         err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
241                              CEPH_MAX_MON, &opt->num_mon);
242         if (err < 0)
243                 goto out;
244
245         /* parse mount options */
246         while ((c = strsep(&options, ",")) != NULL) {
247                 int token, intval, ret;
248                 if (!*c)
249                         continue;
250                 err = -EINVAL;
251                 token = match_token((char *)c, opt_tokens, argstr);
252                 if (token < 0 && parse_extra_token) {
253                         /* extra? */
254                         err = parse_extra_token((char *)c, private);
255                         if (err < 0) {
256                                 pr_err("bad option at '%s'\n", c);
257                                 goto out;
258                         }
259                         continue;
260                 }
261                 if (token < Opt_last_int) {
262                         ret = match_int(&argstr[0], &intval);
263                         if (ret < 0) {
264                                 pr_err("bad mount option arg (not int) "
265                                        "at '%s'\n", c);
266                                 continue;
267                         }
268                         dout("got int token %d val %d\n", token, intval);
269                 } else if (token > Opt_last_int && token < Opt_last_string) {
270                         dout("got string token %d val %s\n", token,
271                              argstr[0].from);
272                 } else {
273                         dout("got token %d\n", token);
274                 }
275                 switch (token) {
276                 case Opt_ip:
277                         err = ceph_parse_ips(argstr[0].from,
278                                              argstr[0].to,
279                                              &opt->my_addr,
280                                              1, NULL);
281                         if (err < 0)
282                                 goto out;
283                         opt->flags |= CEPH_OPT_MYIP;
284                         break;
285
286                 case Opt_fsid:
287                         err = parse_fsid(argstr[0].from, &opt->fsid);
288                         if (err == 0)
289                                 opt->flags |= CEPH_OPT_FSID;
290                         break;
291                 case Opt_name:
292                         opt->name = kstrndup(argstr[0].from,
293                                               argstr[0].to-argstr[0].from,
294                                               GFP_KERNEL);
295                         break;
296                 case Opt_secret:
297                         opt->secret = kstrndup(argstr[0].from,
298                                                 argstr[0].to-argstr[0].from,
299                                                 GFP_KERNEL);
300                         break;
301
302                         /* misc */
303                 case Opt_osdtimeout:
304                         opt->osd_timeout = intval;
305                         break;
306                 case Opt_osdkeepalivetimeout:
307                         opt->osd_keepalive_timeout = intval;
308                         break;
309                 case Opt_osd_idle_ttl:
310                         opt->osd_idle_ttl = intval;
311                         break;
312                 case Opt_mount_timeout:
313                         opt->mount_timeout = intval;
314                         break;
315
316                 case Opt_noshare:
317                         opt->flags |= CEPH_OPT_NOSHARE;
318                         break;
319
320                 case Opt_nocrc:
321                         opt->flags |= CEPH_OPT_NOCRC;
322                         break;
323
324                 default:
325                         BUG_ON(token);
326                 }
327         }
328
329         /* success */
330         *popt = opt;
331         return 0;
332
333 out:
334         ceph_destroy_options(opt);
335         return err;
336 }
337 EXPORT_SYMBOL(ceph_parse_options);
338
339 u64 ceph_client_id(struct ceph_client *client)
340 {
341         return client->monc.auth->global_id;
342 }
343 EXPORT_SYMBOL(ceph_client_id);
344
345 /*
346  * create a fresh client instance
347  */
348 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
349 {
350         struct ceph_client *client;
351         int err = -ENOMEM;
352
353         client = kzalloc(sizeof(*client), GFP_KERNEL);
354         if (client == NULL)
355                 return ERR_PTR(-ENOMEM);
356
357         client->private = private;
358         client->options = opt;
359
360         mutex_init(&client->mount_mutex);
361         init_waitqueue_head(&client->auth_wq);
362         client->auth_err = 0;
363
364         client->extra_mon_dispatch = NULL;
365         client->supported_features = CEPH_FEATURE_SUPPORTED_DEFAULT;
366         client->required_features = CEPH_FEATURE_REQUIRED_DEFAULT;
367
368         client->msgr = NULL;
369
370         /* subsystems */
371         err = ceph_monc_init(&client->monc, client);
372         if (err < 0)
373                 goto fail;
374         err = ceph_osdc_init(&client->osdc, client);
375         if (err < 0)
376                 goto fail_monc;
377
378         return client;
379
380 fail_monc:
381         ceph_monc_stop(&client->monc);
382 fail:
383         kfree(client);
384         return ERR_PTR(err);
385 }
386 EXPORT_SYMBOL(ceph_create_client);
387
388 void ceph_destroy_client(struct ceph_client *client)
389 {
390         dout("destroy_client %p\n", client);
391
392         /* unmount */
393         ceph_osdc_stop(&client->osdc);
394
395         /*
396          * make sure mds and osd connections close out before destroying
397          * the auth module, which is needed to free those connections'
398          * ceph_authorizers.
399          */
400         ceph_msgr_flush();
401
402         ceph_monc_stop(&client->monc);
403
404         ceph_debugfs_client_cleanup(client);
405
406         if (client->msgr)
407                 ceph_messenger_destroy(client->msgr);
408
409         ceph_destroy_options(client->options);
410
411         kfree(client);
412         dout("destroy_client %p done\n", client);
413 }
414 EXPORT_SYMBOL(ceph_destroy_client);
415
416 /*
417  * true if we have the mon map (and have thus joined the cluster)
418  */
419 static int have_mon_and_osd_map(struct ceph_client *client)
420 {
421         return client->monc.monmap && client->monc.monmap->epoch &&
422                client->osdc.osdmap && client->osdc.osdmap->epoch;
423 }
424
425 /*
426  * mount: join the ceph cluster, and open root directory.
427  */
428 int __ceph_open_session(struct ceph_client *client, unsigned long started)
429 {
430         struct ceph_entity_addr *myaddr = NULL;
431         int err;
432         unsigned long timeout = client->options->mount_timeout * HZ;
433
434         /* initialize the messenger */
435         if (client->msgr == NULL) {
436                 if (ceph_test_opt(client, MYIP))
437                         myaddr = &client->options->my_addr;
438                 client->msgr = ceph_messenger_create(myaddr,
439                                         client->supported_features,
440                                         client->required_features);
441                 if (IS_ERR(client->msgr)) {
442                         client->msgr = NULL;
443                         return PTR_ERR(client->msgr);
444                 }
445                 client->msgr->nocrc = ceph_test_opt(client, NOCRC);
446         }
447
448         /* open session, and wait for mon and osd maps */
449         err = ceph_monc_open_session(&client->monc);
450         if (err < 0)
451                 return err;
452
453         while (!have_mon_and_osd_map(client)) {
454                 err = -EIO;
455                 if (timeout && time_after_eq(jiffies, started + timeout))
456                         return err;
457
458                 /* wait */
459                 dout("mount waiting for mon_map\n");
460                 err = wait_event_interruptible_timeout(client->auth_wq,
461                         have_mon_and_osd_map(client) || (client->auth_err < 0),
462                         timeout);
463                 if (err == -EINTR || err == -ERESTARTSYS)
464                         return err;
465                 if (client->auth_err < 0)
466                         return client->auth_err;
467         }
468
469         return 0;
470 }
471 EXPORT_SYMBOL(__ceph_open_session);
472
473
474 int ceph_open_session(struct ceph_client *client)
475 {
476         int ret;
477         unsigned long started = jiffies;  /* note the start time */
478
479         dout("open_session start\n");
480         mutex_lock(&client->mount_mutex);
481
482         ret = __ceph_open_session(client, started);
483
484         mutex_unlock(&client->mount_mutex);
485         return ret;
486 }
487 EXPORT_SYMBOL(ceph_open_session);
488
489
490 static int __init init_ceph_lib(void)
491 {
492         int ret = 0;
493
494         ret = ceph_debugfs_init();
495         if (ret < 0)
496                 goto out;
497
498         ret = ceph_msgr_init();
499         if (ret < 0)
500                 goto out_debugfs;
501
502         pr_info("loaded (mon/osd proto %d/%d, osdmap %d/%d %d/%d)\n",
503                 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL,
504                 CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
505                 CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
506
507         return 0;
508
509 out_debugfs:
510         ceph_debugfs_cleanup();
511 out:
512         return ret;
513 }
514
515 static void __exit exit_ceph_lib(void)
516 {
517         dout("exit_ceph_lib\n");
518         ceph_msgr_exit();
519         ceph_debugfs_cleanup();
520 }
521
522 module_init(init_ceph_lib);
523 module_exit(exit_ceph_lib);
524
525 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
526 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
527 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
528 MODULE_DESCRIPTION("Ceph filesystem for Linux");
529 MODULE_LICENSE("GPL");