Merge tag 'ceph-for-5.8-rc1' of git://github.com/ceph/ceph-client
[platform/kernel/linux-rpi.git] / net / ceph / ceph_common.c
1 // SPDX-License-Identifier: GPL-2.0-only
2
3 #include <linux/ceph/ceph_debug.h>
4 #include <linux/backing-dev.h>
5 #include <linux/ctype.h>
6 #include <linux/fs.h>
7 #include <linux/inet.h>
8 #include <linux/in6.h>
9 #include <linux/key.h>
10 #include <keys/ceph-type.h>
11 #include <linux/module.h>
12 #include <linux/mount.h>
13 #include <linux/nsproxy.h>
14 #include <linux/fs_parser.h>
15 #include <linux/sched.h>
16 #include <linux/sched/mm.h>
17 #include <linux/seq_file.h>
18 #include <linux/slab.h>
19 #include <linux/statfs.h>
20 #include <linux/string.h>
21 #include <linux/vmalloc.h>
22
23
24 #include <linux/ceph/ceph_features.h>
25 #include <linux/ceph/libceph.h>
26 #include <linux/ceph/debugfs.h>
27 #include <linux/ceph/decode.h>
28 #include <linux/ceph/mon_client.h>
29 #include <linux/ceph/auth.h>
30 #include "crypto.h"
31
32
33 /*
34  * Module compatibility interface.  For now it doesn't do anything,
35  * but its existence signals a certain level of functionality.
36  *
37  * The data buffer is used to pass information both to and from
38  * libceph.  The return value indicates whether libceph determines
39  * it is compatible with the caller (from another kernel module),
40  * given the provided data.
41  *
42  * The data pointer can be null.
43  */
44 bool libceph_compatible(void *data)
45 {
46         return true;
47 }
48 EXPORT_SYMBOL(libceph_compatible);
49
50 static int param_get_supported_features(char *buffer,
51                                         const struct kernel_param *kp)
52 {
53         return sprintf(buffer, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT);
54 }
55 static const struct kernel_param_ops param_ops_supported_features = {
56         .get = param_get_supported_features,
57 };
58 module_param_cb(supported_features, &param_ops_supported_features, NULL,
59                 0444);
60
61 const char *ceph_msg_type_name(int type)
62 {
63         switch (type) {
64         case CEPH_MSG_SHUTDOWN: return "shutdown";
65         case CEPH_MSG_PING: return "ping";
66         case CEPH_MSG_AUTH: return "auth";
67         case CEPH_MSG_AUTH_REPLY: return "auth_reply";
68         case CEPH_MSG_MON_MAP: return "mon_map";
69         case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
70         case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
71         case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
72         case CEPH_MSG_STATFS: return "statfs";
73         case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
74         case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
75         case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
76         case CEPH_MSG_MDS_MAP: return "mds_map";
77         case CEPH_MSG_FS_MAP_USER: return "fs_map_user";
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_QUOTA: return "client_quota";
86         case CEPH_MSG_CLIENT_SNAP: return "client_snap";
87         case CEPH_MSG_CLIENT_LEASE: return "client_lease";
88         case CEPH_MSG_POOLOP_REPLY: return "poolop_reply";
89         case CEPH_MSG_POOLOP: return "poolop";
90         case CEPH_MSG_MON_COMMAND: return "mon_command";
91         case CEPH_MSG_MON_COMMAND_ACK: return "mon_command_ack";
92         case CEPH_MSG_OSD_MAP: return "osd_map";
93         case CEPH_MSG_OSD_OP: return "osd_op";
94         case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
95         case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
96         case CEPH_MSG_OSD_BACKOFF: return "osd_backoff";
97         default: return "unknown";
98         }
99 }
100 EXPORT_SYMBOL(ceph_msg_type_name);
101
102 /*
103  * Initially learn our fsid, or verify an fsid matches.
104  */
105 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
106 {
107         if (client->have_fsid) {
108                 if (ceph_fsid_compare(&client->fsid, fsid)) {
109                         pr_err("bad fsid, had %pU got %pU",
110                                &client->fsid, fsid);
111                         return -1;
112                 }
113         } else {
114                 memcpy(&client->fsid, fsid, sizeof(*fsid));
115         }
116         return 0;
117 }
118 EXPORT_SYMBOL(ceph_check_fsid);
119
120 static int strcmp_null(const char *s1, const char *s2)
121 {
122         if (!s1 && !s2)
123                 return 0;
124         if (s1 && !s2)
125                 return -1;
126         if (!s1 && s2)
127                 return 1;
128         return strcmp(s1, s2);
129 }
130
131 int ceph_compare_options(struct ceph_options *new_opt,
132                          struct ceph_client *client)
133 {
134         struct ceph_options *opt1 = new_opt;
135         struct ceph_options *opt2 = client->options;
136         int ofs = offsetof(struct ceph_options, mon_addr);
137         int i;
138         int ret;
139
140         /*
141          * Don't bother comparing options if network namespaces don't
142          * match.
143          */
144         if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
145                 return -1;
146
147         ret = memcmp(opt1, opt2, ofs);
148         if (ret)
149                 return ret;
150
151         ret = strcmp_null(opt1->name, opt2->name);
152         if (ret)
153                 return ret;
154
155         if (opt1->key && !opt2->key)
156                 return -1;
157         if (!opt1->key && opt2->key)
158                 return 1;
159         if (opt1->key && opt2->key) {
160                 if (opt1->key->type != opt2->key->type)
161                         return -1;
162                 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
163                         return -1;
164                 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
165                         return -1;
166                 if (opt1->key->len != opt2->key->len)
167                         return -1;
168                 if (opt1->key->key && !opt2->key->key)
169                         return -1;
170                 if (!opt1->key->key && opt2->key->key)
171                         return 1;
172                 if (opt1->key->key && opt2->key->key) {
173                         ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
174                         if (ret)
175                                 return ret;
176                 }
177         }
178
179         ret = ceph_compare_crush_locs(&opt1->crush_locs, &opt2->crush_locs);
180         if (ret)
181                 return ret;
182
183         /* any matching mon ip implies a match */
184         for (i = 0; i < opt1->num_mon; i++) {
185                 if (ceph_monmap_contains(client->monc.monmap,
186                                  &opt1->mon_addr[i]))
187                         return 0;
188         }
189         return -1;
190 }
191 EXPORT_SYMBOL(ceph_compare_options);
192
193 /*
194  * kvmalloc() doesn't fall back to the vmalloc allocator unless flags are
195  * compatible with (a superset of) GFP_KERNEL.  This is because while the
196  * actual pages are allocated with the specified flags, the page table pages
197  * are always allocated with GFP_KERNEL.
198  *
199  * ceph_kvmalloc() may be called with GFP_KERNEL, GFP_NOFS or GFP_NOIO.
200  */
201 void *ceph_kvmalloc(size_t size, gfp_t flags)
202 {
203         void *p;
204
205         if ((flags & (__GFP_IO | __GFP_FS)) == (__GFP_IO | __GFP_FS)) {
206                 p = kvmalloc(size, flags);
207         } else if ((flags & (__GFP_IO | __GFP_FS)) == __GFP_IO) {
208                 unsigned int nofs_flag = memalloc_nofs_save();
209                 p = kvmalloc(size, GFP_KERNEL);
210                 memalloc_nofs_restore(nofs_flag);
211         } else {
212                 unsigned int noio_flag = memalloc_noio_save();
213                 p = kvmalloc(size, GFP_KERNEL);
214                 memalloc_noio_restore(noio_flag);
215         }
216
217         return p;
218 }
219
220 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
221 {
222         int i = 0;
223         char tmp[3];
224         int err = -EINVAL;
225         int d;
226
227         dout("parse_fsid '%s'\n", str);
228         tmp[2] = 0;
229         while (*str && i < 16) {
230                 if (ispunct(*str)) {
231                         str++;
232                         continue;
233                 }
234                 if (!isxdigit(str[0]) || !isxdigit(str[1]))
235                         break;
236                 tmp[0] = str[0];
237                 tmp[1] = str[1];
238                 if (sscanf(tmp, "%x", &d) < 1)
239                         break;
240                 fsid->fsid[i] = d & 0xff;
241                 i++;
242                 str += 2;
243         }
244
245         if (i == 16)
246                 err = 0;
247         dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
248         return err;
249 }
250
251 /*
252  * ceph options
253  */
254 enum {
255         Opt_osdtimeout,
256         Opt_osdkeepalivetimeout,
257         Opt_mount_timeout,
258         Opt_osd_idle_ttl,
259         Opt_osd_request_timeout,
260         /* int args above */
261         Opt_fsid,
262         Opt_name,
263         Opt_secret,
264         Opt_key,
265         Opt_ip,
266         Opt_crush_location,
267         Opt_read_from_replica,
268         /* string args above */
269         Opt_share,
270         Opt_crc,
271         Opt_cephx_require_signatures,
272         Opt_cephx_sign_messages,
273         Opt_tcp_nodelay,
274         Opt_abort_on_full,
275 };
276
277 enum {
278         Opt_read_from_replica_no,
279         Opt_read_from_replica_balance,
280         Opt_read_from_replica_localize,
281 };
282
283 static const struct constant_table ceph_param_read_from_replica[] = {
284         {"no",          Opt_read_from_replica_no},
285         {"balance",     Opt_read_from_replica_balance},
286         {"localize",    Opt_read_from_replica_localize},
287         {}
288 };
289
290 static const struct fs_parameter_spec ceph_parameters[] = {
291         fsparam_flag    ("abort_on_full",               Opt_abort_on_full),
292         fsparam_flag_no ("cephx_require_signatures",    Opt_cephx_require_signatures),
293         fsparam_flag_no ("cephx_sign_messages",         Opt_cephx_sign_messages),
294         fsparam_flag_no ("crc",                         Opt_crc),
295         fsparam_string  ("crush_location",              Opt_crush_location),
296         fsparam_string  ("fsid",                        Opt_fsid),
297         fsparam_string  ("ip",                          Opt_ip),
298         fsparam_string  ("key",                         Opt_key),
299         fsparam_u32     ("mount_timeout",               Opt_mount_timeout),
300         fsparam_string  ("name",                        Opt_name),
301         fsparam_u32     ("osd_idle_ttl",                Opt_osd_idle_ttl),
302         fsparam_u32     ("osd_request_timeout",         Opt_osd_request_timeout),
303         fsparam_u32     ("osdkeepalive",                Opt_osdkeepalivetimeout),
304         __fsparam       (fs_param_is_s32, "osdtimeout", Opt_osdtimeout,
305                          fs_param_deprecated, NULL),
306         fsparam_enum    ("read_from_replica",           Opt_read_from_replica,
307                          ceph_param_read_from_replica),
308         fsparam_string  ("secret",                      Opt_secret),
309         fsparam_flag_no ("share",                       Opt_share),
310         fsparam_flag_no ("tcp_nodelay",                 Opt_tcp_nodelay),
311         {}
312 };
313
314 struct ceph_options *ceph_alloc_options(void)
315 {
316         struct ceph_options *opt;
317
318         opt = kzalloc(sizeof(*opt), GFP_KERNEL);
319         if (!opt)
320                 return NULL;
321
322         opt->crush_locs = RB_ROOT;
323         opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
324                                 GFP_KERNEL);
325         if (!opt->mon_addr) {
326                 kfree(opt);
327                 return NULL;
328         }
329
330         opt->flags = CEPH_OPT_DEFAULT;
331         opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
332         opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
333         opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
334         opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
335         return opt;
336 }
337 EXPORT_SYMBOL(ceph_alloc_options);
338
339 void ceph_destroy_options(struct ceph_options *opt)
340 {
341         dout("destroy_options %p\n", opt);
342         if (!opt)
343                 return;
344
345         ceph_clear_crush_locs(&opt->crush_locs);
346         kfree(opt->name);
347         if (opt->key) {
348                 ceph_crypto_key_destroy(opt->key);
349                 kfree(opt->key);
350         }
351         kfree(opt->mon_addr);
352         kfree(opt);
353 }
354 EXPORT_SYMBOL(ceph_destroy_options);
355
356 /* get secret from key store */
357 static int get_secret(struct ceph_crypto_key *dst, const char *name,
358                       struct p_log *log)
359 {
360         struct key *ukey;
361         int key_err;
362         int err = 0;
363         struct ceph_crypto_key *ckey;
364
365         ukey = request_key(&key_type_ceph, name, NULL);
366         if (IS_ERR(ukey)) {
367                 /* request_key errors don't map nicely to mount(2)
368                    errors; don't even try, but still printk */
369                 key_err = PTR_ERR(ukey);
370                 switch (key_err) {
371                 case -ENOKEY:
372                         error_plog(log, "Failed due to key not found: %s",
373                                name);
374                         break;
375                 case -EKEYEXPIRED:
376                         error_plog(log, "Failed due to expired key: %s",
377                                name);
378                         break;
379                 case -EKEYREVOKED:
380                         error_plog(log, "Failed due to revoked key: %s",
381                                name);
382                         break;
383                 default:
384                         error_plog(log, "Failed due to key error %d: %s",
385                                key_err, name);
386                 }
387                 err = -EPERM;
388                 goto out;
389         }
390
391         ckey = ukey->payload.data[0];
392         err = ceph_crypto_key_clone(dst, ckey);
393         if (err)
394                 goto out_key;
395         /* pass through, err is 0 */
396
397 out_key:
398         key_put(ukey);
399 out:
400         return err;
401 }
402
403 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
404                        struct fc_log *l)
405 {
406         struct p_log log = {.prefix = "libceph", .log = l};
407         int ret;
408
409         /* ip1[:port1][,ip2[:port2]...] */
410         ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
411                              &opt->num_mon);
412         if (ret) {
413                 error_plog(&log, "Failed to parse monitor IPs: %d", ret);
414                 return ret;
415         }
416
417         return 0;
418 }
419 EXPORT_SYMBOL(ceph_parse_mon_ips);
420
421 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
422                      struct fc_log *l)
423 {
424         struct fs_parse_result result;
425         int token, err;
426         struct p_log log = {.prefix = "libceph", .log = l};
427
428         token = __fs_parse(&log, ceph_parameters, param, &result);
429         dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
430         if (token < 0)
431                 return token;
432
433         switch (token) {
434         case Opt_ip:
435                 err = ceph_parse_ips(param->string,
436                                      param->string + param->size,
437                                      &opt->my_addr,
438                                      1, NULL);
439                 if (err) {
440                         error_plog(&log, "Failed to parse ip: %d", err);
441                         return err;
442                 }
443                 opt->flags |= CEPH_OPT_MYIP;
444                 break;
445
446         case Opt_fsid:
447                 err = parse_fsid(param->string, &opt->fsid);
448                 if (err) {
449                         error_plog(&log, "Failed to parse fsid: %d", err);
450                         return err;
451                 }
452                 opt->flags |= CEPH_OPT_FSID;
453                 break;
454         case Opt_name:
455                 kfree(opt->name);
456                 opt->name = param->string;
457                 param->string = NULL;
458                 break;
459         case Opt_secret:
460                 ceph_crypto_key_destroy(opt->key);
461                 kfree(opt->key);
462
463                 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
464                 if (!opt->key)
465                         return -ENOMEM;
466                 err = ceph_crypto_key_unarmor(opt->key, param->string);
467                 if (err) {
468                         error_plog(&log, "Failed to parse secret: %d", err);
469                         return err;
470                 }
471                 break;
472         case Opt_key:
473                 ceph_crypto_key_destroy(opt->key);
474                 kfree(opt->key);
475
476                 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
477                 if (!opt->key)
478                         return -ENOMEM;
479                 return get_secret(opt->key, param->string, &log);
480         case Opt_crush_location:
481                 ceph_clear_crush_locs(&opt->crush_locs);
482                 err = ceph_parse_crush_location(param->string,
483                                                 &opt->crush_locs);
484                 if (err) {
485                         error_plog(&log, "Failed to parse CRUSH location: %d",
486                                    err);
487                         return err;
488                 }
489                 break;
490         case Opt_read_from_replica:
491                 switch (result.uint_32) {
492                 case Opt_read_from_replica_no:
493                         opt->osd_req_flags &= ~(CEPH_OSD_FLAG_BALANCE_READS |
494                                                 CEPH_OSD_FLAG_LOCALIZE_READS);
495                         break;
496                 case Opt_read_from_replica_balance:
497                         opt->osd_req_flags |= CEPH_OSD_FLAG_BALANCE_READS;
498                         opt->osd_req_flags &= ~CEPH_OSD_FLAG_LOCALIZE_READS;
499                         break;
500                 case Opt_read_from_replica_localize:
501                         opt->osd_req_flags |= CEPH_OSD_FLAG_LOCALIZE_READS;
502                         opt->osd_req_flags &= ~CEPH_OSD_FLAG_BALANCE_READS;
503                         break;
504                 default:
505                         BUG();
506                 }
507                 break;
508
509         case Opt_osdtimeout:
510                 warn_plog(&log, "Ignoring osdtimeout");
511                 break;
512         case Opt_osdkeepalivetimeout:
513                 /* 0 isn't well defined right now, reject it */
514                 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
515                         goto out_of_range;
516                 opt->osd_keepalive_timeout =
517                     msecs_to_jiffies(result.uint_32 * 1000);
518                 break;
519         case Opt_osd_idle_ttl:
520                 /* 0 isn't well defined right now, reject it */
521                 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
522                         goto out_of_range;
523                 opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
524                 break;
525         case Opt_mount_timeout:
526                 /* 0 is "wait forever" (i.e. infinite timeout) */
527                 if (result.uint_32 > INT_MAX / 1000)
528                         goto out_of_range;
529                 opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
530                 break;
531         case Opt_osd_request_timeout:
532                 /* 0 is "wait forever" (i.e. infinite timeout) */
533                 if (result.uint_32 > INT_MAX / 1000)
534                         goto out_of_range;
535                 opt->osd_request_timeout =
536                     msecs_to_jiffies(result.uint_32 * 1000);
537                 break;
538
539         case Opt_share:
540                 if (!result.negated)
541                         opt->flags &= ~CEPH_OPT_NOSHARE;
542                 else
543                         opt->flags |= CEPH_OPT_NOSHARE;
544                 break;
545         case Opt_crc:
546                 if (!result.negated)
547                         opt->flags &= ~CEPH_OPT_NOCRC;
548                 else
549                         opt->flags |= CEPH_OPT_NOCRC;
550                 break;
551         case Opt_cephx_require_signatures:
552                 if (!result.negated)
553                         opt->flags &= ~CEPH_OPT_NOMSGAUTH;
554                 else
555                         opt->flags |= CEPH_OPT_NOMSGAUTH;
556                 break;
557         case Opt_cephx_sign_messages:
558                 if (!result.negated)
559                         opt->flags &= ~CEPH_OPT_NOMSGSIGN;
560                 else
561                         opt->flags |= CEPH_OPT_NOMSGSIGN;
562                 break;
563         case Opt_tcp_nodelay:
564                 if (!result.negated)
565                         opt->flags |= CEPH_OPT_TCP_NODELAY;
566                 else
567                         opt->flags &= ~CEPH_OPT_TCP_NODELAY;
568                 break;
569
570         case Opt_abort_on_full:
571                 opt->flags |= CEPH_OPT_ABORT_ON_FULL;
572                 break;
573
574         default:
575                 BUG();
576         }
577
578         return 0;
579
580 out_of_range:
581         return inval_plog(&log, "%s out of range", param->key);
582 }
583 EXPORT_SYMBOL(ceph_parse_param);
584
585 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
586                               bool show_all)
587 {
588         struct ceph_options *opt = client->options;
589         size_t pos = m->count;
590         struct rb_node *n;
591
592         if (opt->name) {
593                 seq_puts(m, "name=");
594                 seq_escape(m, opt->name, ", \t\n\\");
595                 seq_putc(m, ',');
596         }
597         if (opt->key)
598                 seq_puts(m, "secret=<hidden>,");
599
600         if (!RB_EMPTY_ROOT(&opt->crush_locs)) {
601                 seq_puts(m, "crush_location=");
602                 for (n = rb_first(&opt->crush_locs); ; ) {
603                         struct crush_loc_node *loc =
604                             rb_entry(n, struct crush_loc_node, cl_node);
605
606                         seq_printf(m, "%s:%s", loc->cl_loc.cl_type_name,
607                                    loc->cl_loc.cl_name);
608                         n = rb_next(n);
609                         if (!n)
610                                 break;
611
612                         seq_putc(m, '|');
613                 }
614                 seq_putc(m, ',');
615         }
616         if (opt->osd_req_flags & CEPH_OSD_FLAG_BALANCE_READS) {
617                 seq_puts(m, "read_from_replica=balance,");
618         } else if (opt->osd_req_flags & CEPH_OSD_FLAG_LOCALIZE_READS) {
619                 seq_puts(m, "read_from_replica=localize,");
620         }
621
622         if (opt->flags & CEPH_OPT_FSID)
623                 seq_printf(m, "fsid=%pU,", &opt->fsid);
624         if (opt->flags & CEPH_OPT_NOSHARE)
625                 seq_puts(m, "noshare,");
626         if (opt->flags & CEPH_OPT_NOCRC)
627                 seq_puts(m, "nocrc,");
628         if (opt->flags & CEPH_OPT_NOMSGAUTH)
629                 seq_puts(m, "nocephx_require_signatures,");
630         if (opt->flags & CEPH_OPT_NOMSGSIGN)
631                 seq_puts(m, "nocephx_sign_messages,");
632         if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
633                 seq_puts(m, "notcp_nodelay,");
634         if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
635                 seq_puts(m, "abort_on_full,");
636
637         if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
638                 seq_printf(m, "mount_timeout=%d,",
639                            jiffies_to_msecs(opt->mount_timeout) / 1000);
640         if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
641                 seq_printf(m, "osd_idle_ttl=%d,",
642                            jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
643         if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
644                 seq_printf(m, "osdkeepalivetimeout=%d,",
645                     jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
646         if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
647                 seq_printf(m, "osd_request_timeout=%d,",
648                            jiffies_to_msecs(opt->osd_request_timeout) / 1000);
649
650         /* drop redundant comma */
651         if (m->count != pos)
652                 m->count--;
653
654         return 0;
655 }
656 EXPORT_SYMBOL(ceph_print_client_options);
657
658 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
659 {
660         return &client->msgr.inst.addr;
661 }
662 EXPORT_SYMBOL(ceph_client_addr);
663
664 u64 ceph_client_gid(struct ceph_client *client)
665 {
666         return client->monc.auth->global_id;
667 }
668 EXPORT_SYMBOL(ceph_client_gid);
669
670 /*
671  * create a fresh client instance
672  */
673 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
674 {
675         struct ceph_client *client;
676         struct ceph_entity_addr *myaddr = NULL;
677         int err;
678
679         err = wait_for_random_bytes();
680         if (err < 0)
681                 return ERR_PTR(err);
682
683         client = kzalloc(sizeof(*client), GFP_KERNEL);
684         if (client == NULL)
685                 return ERR_PTR(-ENOMEM);
686
687         client->private = private;
688         client->options = opt;
689
690         mutex_init(&client->mount_mutex);
691         init_waitqueue_head(&client->auth_wq);
692         client->auth_err = 0;
693
694         client->extra_mon_dispatch = NULL;
695         client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
696         client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
697
698         if (!ceph_test_opt(client, NOMSGAUTH))
699                 client->required_features |= CEPH_FEATURE_MSG_AUTH;
700
701         /* msgr */
702         if (ceph_test_opt(client, MYIP))
703                 myaddr = &client->options->my_addr;
704
705         ceph_messenger_init(&client->msgr, myaddr);
706
707         /* subsystems */
708         err = ceph_monc_init(&client->monc, client);
709         if (err < 0)
710                 goto fail;
711         err = ceph_osdc_init(&client->osdc, client);
712         if (err < 0)
713                 goto fail_monc;
714
715         return client;
716
717 fail_monc:
718         ceph_monc_stop(&client->monc);
719 fail:
720         ceph_messenger_fini(&client->msgr);
721         kfree(client);
722         return ERR_PTR(err);
723 }
724 EXPORT_SYMBOL(ceph_create_client);
725
726 void ceph_destroy_client(struct ceph_client *client)
727 {
728         dout("destroy_client %p\n", client);
729
730         atomic_set(&client->msgr.stopping, 1);
731
732         /* unmount */
733         ceph_osdc_stop(&client->osdc);
734         ceph_monc_stop(&client->monc);
735         ceph_messenger_fini(&client->msgr);
736
737         ceph_debugfs_client_cleanup(client);
738
739         ceph_destroy_options(client->options);
740
741         kfree(client);
742         dout("destroy_client %p done\n", client);
743 }
744 EXPORT_SYMBOL(ceph_destroy_client);
745
746 void ceph_reset_client_addr(struct ceph_client *client)
747 {
748         ceph_messenger_reset_nonce(&client->msgr);
749         ceph_monc_reopen_session(&client->monc);
750         ceph_osdc_reopen_osds(&client->osdc);
751 }
752 EXPORT_SYMBOL(ceph_reset_client_addr);
753
754 /*
755  * true if we have the mon map (and have thus joined the cluster)
756  */
757 static bool have_mon_and_osd_map(struct ceph_client *client)
758 {
759         return client->monc.monmap && client->monc.monmap->epoch &&
760                client->osdc.osdmap && client->osdc.osdmap->epoch;
761 }
762
763 /*
764  * mount: join the ceph cluster, and open root directory.
765  */
766 int __ceph_open_session(struct ceph_client *client, unsigned long started)
767 {
768         unsigned long timeout = client->options->mount_timeout;
769         long err;
770
771         /* open session, and wait for mon and osd maps */
772         err = ceph_monc_open_session(&client->monc);
773         if (err < 0)
774                 return err;
775
776         while (!have_mon_and_osd_map(client)) {
777                 if (timeout && time_after_eq(jiffies, started + timeout))
778                         return -ETIMEDOUT;
779
780                 /* wait */
781                 dout("mount waiting for mon_map\n");
782                 err = wait_event_interruptible_timeout(client->auth_wq,
783                         have_mon_and_osd_map(client) || (client->auth_err < 0),
784                         ceph_timeout_jiffies(timeout));
785                 if (err < 0)
786                         return err;
787                 if (client->auth_err < 0)
788                         return client->auth_err;
789         }
790
791         pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
792                 &client->fsid);
793         ceph_debugfs_client_init(client);
794
795         return 0;
796 }
797 EXPORT_SYMBOL(__ceph_open_session);
798
799 int ceph_open_session(struct ceph_client *client)
800 {
801         int ret;
802         unsigned long started = jiffies;  /* note the start time */
803
804         dout("open_session start\n");
805         mutex_lock(&client->mount_mutex);
806
807         ret = __ceph_open_session(client, started);
808
809         mutex_unlock(&client->mount_mutex);
810         return ret;
811 }
812 EXPORT_SYMBOL(ceph_open_session);
813
814 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
815                                 unsigned long timeout)
816 {
817         u64 newest_epoch;
818         int ret;
819
820         ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
821         if (ret)
822                 return ret;
823
824         if (client->osdc.osdmap->epoch >= newest_epoch)
825                 return 0;
826
827         ceph_osdc_maybe_request_map(&client->osdc);
828         return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
829 }
830 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
831
832 static int __init init_ceph_lib(void)
833 {
834         int ret = 0;
835
836         ceph_debugfs_init();
837
838         ret = ceph_crypto_init();
839         if (ret < 0)
840                 goto out_debugfs;
841
842         ret = ceph_msgr_init();
843         if (ret < 0)
844                 goto out_crypto;
845
846         ret = ceph_osdc_setup();
847         if (ret < 0)
848                 goto out_msgr;
849
850         pr_info("loaded (mon/osd proto %d/%d)\n",
851                 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
852
853         return 0;
854
855 out_msgr:
856         ceph_msgr_exit();
857 out_crypto:
858         ceph_crypto_shutdown();
859 out_debugfs:
860         ceph_debugfs_cleanup();
861         return ret;
862 }
863
864 static void __exit exit_ceph_lib(void)
865 {
866         dout("exit_ceph_lib\n");
867         WARN_ON(!ceph_strings_empty());
868
869         ceph_osdc_cleanup();
870         ceph_msgr_exit();
871         ceph_crypto_shutdown();
872         ceph_debugfs_cleanup();
873 }
874
875 module_init(init_ceph_lib);
876 module_exit(exit_ceph_lib);
877
878 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
879 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
880 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
881 MODULE_DESCRIPTION("Ceph core library");
882 MODULE_LICENSE("GPL");