1 // SPDX-License-Identifier: GPL-2.0-only
3 #include <linux/ceph/ceph_debug.h>
4 #include <linux/backing-dev.h>
5 #include <linux/ctype.h>
7 #include <linux/inet.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/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>
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>
34 * Module compatibility interface. For now it doesn't do anything,
35 * but its existence signals a certain level of functionality.
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.
42 * The data pointer can be null.
44 bool libceph_compatible(void *data)
48 EXPORT_SYMBOL(libceph_compatible);
50 static int param_get_supported_features(char *buffer,
51 const struct kernel_param *kp)
53 return sprintf(buffer, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT);
55 static const struct kernel_param_ops param_ops_supported_features = {
56 .get = param_get_supported_features,
58 module_param_cb(supported_features, ¶m_ops_supported_features, NULL,
61 const char *ceph_msg_type_name(int 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";
100 EXPORT_SYMBOL(ceph_msg_type_name);
103 * Initially learn our fsid, or verify an fsid matches.
105 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
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);
114 memcpy(&client->fsid, fsid, sizeof(*fsid));
118 EXPORT_SYMBOL(ceph_check_fsid);
120 static int strcmp_null(const char *s1, const char *s2)
128 return strcmp(s1, s2);
131 int ceph_compare_options(struct ceph_options *new_opt,
132 struct ceph_client *client)
134 struct ceph_options *opt1 = new_opt;
135 struct ceph_options *opt2 = client->options;
136 int ofs = offsetof(struct ceph_options, mon_addr);
141 * Don't bother comparing options if network namespaces don't
144 if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
147 ret = memcmp(opt1, opt2, ofs);
151 ret = strcmp_null(opt1->name, opt2->name);
155 if (opt1->key && !opt2->key)
157 if (!opt1->key && opt2->key)
159 if (opt1->key && opt2->key) {
160 if (opt1->key->type != opt2->key->type)
162 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
164 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
166 if (opt1->key->len != opt2->key->len)
168 if (opt1->key->key && !opt2->key->key)
170 if (!opt1->key->key && opt2->key->key)
172 if (opt1->key->key && opt2->key->key) {
173 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
179 /* any matching mon ip implies a match */
180 for (i = 0; i < opt1->num_mon; i++) {
181 if (ceph_monmap_contains(client->monc.monmap,
187 EXPORT_SYMBOL(ceph_compare_options);
190 * kvmalloc() doesn't fall back to the vmalloc allocator unless flags are
191 * compatible with (a superset of) GFP_KERNEL. This is because while the
192 * actual pages are allocated with the specified flags, the page table pages
193 * are always allocated with GFP_KERNEL. map_vm_area() doesn't even take
194 * flags because GFP_KERNEL is hard-coded in {p4d,pud,pmd,pte}_alloc().
196 * ceph_kvmalloc() may be called with GFP_KERNEL, GFP_NOFS or GFP_NOIO.
198 void *ceph_kvmalloc(size_t size, gfp_t flags)
202 if ((flags & (__GFP_IO | __GFP_FS)) == (__GFP_IO | __GFP_FS)) {
203 p = kvmalloc(size, flags);
204 } else if ((flags & (__GFP_IO | __GFP_FS)) == __GFP_IO) {
205 unsigned int nofs_flag = memalloc_nofs_save();
206 p = kvmalloc(size, GFP_KERNEL);
207 memalloc_nofs_restore(nofs_flag);
209 unsigned int noio_flag = memalloc_noio_save();
210 p = kvmalloc(size, GFP_KERNEL);
211 memalloc_noio_restore(noio_flag);
217 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
224 dout("parse_fsid '%s'\n", str);
226 while (*str && i < 16) {
231 if (!isxdigit(str[0]) || !isxdigit(str[1]))
235 if (sscanf(tmp, "%x", &d) < 1)
237 fsid->fsid[i] = d & 0xff;
244 dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
253 Opt_osdkeepalivetimeout,
256 Opt_osd_request_timeout,
265 /* string args above */
270 Opt_cephx_require_signatures,
271 Opt_nocephx_require_signatures,
272 Opt_cephx_sign_messages,
273 Opt_nocephx_sign_messages,
279 static match_table_t opt_tokens = {
280 {Opt_osdtimeout, "osdtimeout=%d"},
281 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
282 {Opt_mount_timeout, "mount_timeout=%d"},
283 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
284 {Opt_osd_request_timeout, "osd_request_timeout=%d"},
286 {Opt_fsid, "fsid=%s"},
287 {Opt_name, "name=%s"},
288 {Opt_secret, "secret=%s"},
291 /* string args above */
292 {Opt_share, "share"},
293 {Opt_noshare, "noshare"},
295 {Opt_nocrc, "nocrc"},
296 {Opt_cephx_require_signatures, "cephx_require_signatures"},
297 {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
298 {Opt_cephx_sign_messages, "cephx_sign_messages"},
299 {Opt_nocephx_sign_messages, "nocephx_sign_messages"},
300 {Opt_tcp_nodelay, "tcp_nodelay"},
301 {Opt_notcp_nodelay, "notcp_nodelay"},
302 {Opt_abort_on_full, "abort_on_full"},
306 void ceph_destroy_options(struct ceph_options *opt)
308 dout("destroy_options %p\n", opt);
311 ceph_crypto_key_destroy(opt->key);
314 kfree(opt->mon_addr);
317 EXPORT_SYMBOL(ceph_destroy_options);
319 /* get secret from key store */
320 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
324 struct ceph_crypto_key *ckey;
326 ukey = request_key(&key_type_ceph, name, NULL);
328 /* request_key errors don't map nicely to mount(2)
329 errors; don't even try, but still printk */
330 key_err = PTR_ERR(ukey);
333 pr_warn("ceph: Mount failed due to key not found: %s\n",
337 pr_warn("ceph: Mount failed due to expired key: %s\n",
341 pr_warn("ceph: Mount failed due to revoked key: %s\n",
345 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
352 ckey = ukey->payload.data[0];
353 err = ceph_crypto_key_clone(dst, ckey);
356 /* pass through, err is 0 */
364 struct ceph_options *
365 ceph_parse_options(char *options, const char *dev_name,
366 const char *dev_name_end,
367 int (*parse_extra_token)(char *c, void *private),
370 struct ceph_options *opt;
373 substring_t argstr[MAX_OPT_ARGS];
375 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
377 return ERR_PTR(-ENOMEM);
378 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
383 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
386 /* start with defaults */
387 opt->flags = CEPH_OPT_DEFAULT;
388 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
389 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
390 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
391 opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
394 /* ip1[:port1][,ip2[:port2]...] */
395 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
396 CEPH_MAX_MON, &opt->num_mon);
400 /* parse mount options */
401 while ((c = strsep(&options, ",")) != NULL) {
406 token = match_token((char *)c, opt_tokens, argstr);
407 if (token < 0 && parse_extra_token) {
409 err = parse_extra_token((char *)c, private);
411 pr_err("bad option at '%s'\n", c);
416 if (token < Opt_last_int) {
417 err = match_int(&argstr[0], &intval);
419 pr_err("bad option arg (not int) at '%s'\n", c);
422 dout("got int token %d val %d\n", token, intval);
423 } else if (token > Opt_last_int && token < Opt_last_string) {
424 dout("got string token %d val %s\n", token,
427 dout("got token %d\n", token);
431 err = ceph_parse_ips(argstr[0].from,
437 opt->flags |= CEPH_OPT_MYIP;
441 err = parse_fsid(argstr[0].from, &opt->fsid);
443 opt->flags |= CEPH_OPT_FSID;
447 opt->name = kstrndup(argstr[0].from,
448 argstr[0].to-argstr[0].from,
456 ceph_crypto_key_destroy(opt->key);
459 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
464 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
469 ceph_crypto_key_destroy(opt->key);
472 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
477 err = get_secret(opt->key, argstr[0].from);
484 pr_warn("ignoring deprecated osdtimeout option\n");
486 case Opt_osdkeepalivetimeout:
487 /* 0 isn't well defined right now, reject it */
488 if (intval < 1 || intval > INT_MAX / 1000) {
489 pr_err("osdkeepalive out of range\n");
493 opt->osd_keepalive_timeout =
494 msecs_to_jiffies(intval * 1000);
496 case Opt_osd_idle_ttl:
497 /* 0 isn't well defined right now, reject it */
498 if (intval < 1 || intval > INT_MAX / 1000) {
499 pr_err("osd_idle_ttl out of range\n");
503 opt->osd_idle_ttl = msecs_to_jiffies(intval * 1000);
505 case Opt_mount_timeout:
506 /* 0 is "wait forever" (i.e. infinite timeout) */
507 if (intval < 0 || intval > INT_MAX / 1000) {
508 pr_err("mount_timeout out of range\n");
512 opt->mount_timeout = msecs_to_jiffies(intval * 1000);
514 case Opt_osd_request_timeout:
515 /* 0 is "wait forever" (i.e. infinite timeout) */
516 if (intval < 0 || intval > INT_MAX / 1000) {
517 pr_err("osd_request_timeout out of range\n");
521 opt->osd_request_timeout = msecs_to_jiffies(intval * 1000);
525 opt->flags &= ~CEPH_OPT_NOSHARE;
528 opt->flags |= CEPH_OPT_NOSHARE;
532 opt->flags &= ~CEPH_OPT_NOCRC;
535 opt->flags |= CEPH_OPT_NOCRC;
538 case Opt_cephx_require_signatures:
539 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
541 case Opt_nocephx_require_signatures:
542 opt->flags |= CEPH_OPT_NOMSGAUTH;
544 case Opt_cephx_sign_messages:
545 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
547 case Opt_nocephx_sign_messages:
548 opt->flags |= CEPH_OPT_NOMSGSIGN;
551 case Opt_tcp_nodelay:
552 opt->flags |= CEPH_OPT_TCP_NODELAY;
554 case Opt_notcp_nodelay:
555 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
558 case Opt_abort_on_full:
559 opt->flags |= CEPH_OPT_ABORT_ON_FULL;
571 ceph_destroy_options(opt);
574 EXPORT_SYMBOL(ceph_parse_options);
576 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
579 struct ceph_options *opt = client->options;
580 size_t pos = m->count;
583 seq_puts(m, "name=");
584 seq_escape(m, opt->name, ", \t\n\\");
588 seq_puts(m, "secret=<hidden>,");
590 if (opt->flags & CEPH_OPT_FSID)
591 seq_printf(m, "fsid=%pU,", &opt->fsid);
592 if (opt->flags & CEPH_OPT_NOSHARE)
593 seq_puts(m, "noshare,");
594 if (opt->flags & CEPH_OPT_NOCRC)
595 seq_puts(m, "nocrc,");
596 if (opt->flags & CEPH_OPT_NOMSGAUTH)
597 seq_puts(m, "nocephx_require_signatures,");
598 if (opt->flags & CEPH_OPT_NOMSGSIGN)
599 seq_puts(m, "nocephx_sign_messages,");
600 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
601 seq_puts(m, "notcp_nodelay,");
602 if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
603 seq_puts(m, "abort_on_full,");
605 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
606 seq_printf(m, "mount_timeout=%d,",
607 jiffies_to_msecs(opt->mount_timeout) / 1000);
608 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
609 seq_printf(m, "osd_idle_ttl=%d,",
610 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
611 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
612 seq_printf(m, "osdkeepalivetimeout=%d,",
613 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
614 if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
615 seq_printf(m, "osd_request_timeout=%d,",
616 jiffies_to_msecs(opt->osd_request_timeout) / 1000);
618 /* drop redundant comma */
624 EXPORT_SYMBOL(ceph_print_client_options);
626 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
628 return &client->msgr.inst.addr;
630 EXPORT_SYMBOL(ceph_client_addr);
632 u64 ceph_client_gid(struct ceph_client *client)
634 return client->monc.auth->global_id;
636 EXPORT_SYMBOL(ceph_client_gid);
639 * create a fresh client instance
641 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
643 struct ceph_client *client;
644 struct ceph_entity_addr *myaddr = NULL;
647 err = wait_for_random_bytes();
651 client = kzalloc(sizeof(*client), GFP_KERNEL);
653 return ERR_PTR(-ENOMEM);
655 client->private = private;
656 client->options = opt;
658 mutex_init(&client->mount_mutex);
659 init_waitqueue_head(&client->auth_wq);
660 client->auth_err = 0;
662 client->extra_mon_dispatch = NULL;
663 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
664 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
666 if (!ceph_test_opt(client, NOMSGAUTH))
667 client->required_features |= CEPH_FEATURE_MSG_AUTH;
670 if (ceph_test_opt(client, MYIP))
671 myaddr = &client->options->my_addr;
673 ceph_messenger_init(&client->msgr, myaddr);
676 err = ceph_monc_init(&client->monc, client);
679 err = ceph_osdc_init(&client->osdc, client);
686 ceph_monc_stop(&client->monc);
688 ceph_messenger_fini(&client->msgr);
692 EXPORT_SYMBOL(ceph_create_client);
694 void ceph_destroy_client(struct ceph_client *client)
696 dout("destroy_client %p\n", client);
698 atomic_set(&client->msgr.stopping, 1);
701 ceph_osdc_stop(&client->osdc);
702 ceph_monc_stop(&client->monc);
703 ceph_messenger_fini(&client->msgr);
705 ceph_debugfs_client_cleanup(client);
707 ceph_destroy_options(client->options);
710 dout("destroy_client %p done\n", client);
712 EXPORT_SYMBOL(ceph_destroy_client);
714 void ceph_reset_client_addr(struct ceph_client *client)
716 ceph_messenger_reset_nonce(&client->msgr);
717 ceph_monc_reopen_session(&client->monc);
718 ceph_osdc_reopen_osds(&client->osdc);
720 EXPORT_SYMBOL(ceph_reset_client_addr);
723 * true if we have the mon map (and have thus joined the cluster)
725 static bool have_mon_and_osd_map(struct ceph_client *client)
727 return client->monc.monmap && client->monc.monmap->epoch &&
728 client->osdc.osdmap && client->osdc.osdmap->epoch;
732 * mount: join the ceph cluster, and open root directory.
734 int __ceph_open_session(struct ceph_client *client, unsigned long started)
736 unsigned long timeout = client->options->mount_timeout;
739 /* open session, and wait for mon and osd maps */
740 err = ceph_monc_open_session(&client->monc);
744 while (!have_mon_and_osd_map(client)) {
745 if (timeout && time_after_eq(jiffies, started + timeout))
749 dout("mount waiting for mon_map\n");
750 err = wait_event_interruptible_timeout(client->auth_wq,
751 have_mon_and_osd_map(client) || (client->auth_err < 0),
752 ceph_timeout_jiffies(timeout));
755 if (client->auth_err < 0)
756 return client->auth_err;
759 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
761 ceph_debugfs_client_init(client);
765 EXPORT_SYMBOL(__ceph_open_session);
767 int ceph_open_session(struct ceph_client *client)
770 unsigned long started = jiffies; /* note the start time */
772 dout("open_session start\n");
773 mutex_lock(&client->mount_mutex);
775 ret = __ceph_open_session(client, started);
777 mutex_unlock(&client->mount_mutex);
780 EXPORT_SYMBOL(ceph_open_session);
782 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
783 unsigned long timeout)
788 ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
792 if (client->osdc.osdmap->epoch >= newest_epoch)
795 ceph_osdc_maybe_request_map(&client->osdc);
796 return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
798 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
800 static int __init init_ceph_lib(void)
806 ret = ceph_crypto_init();
810 ret = ceph_msgr_init();
814 ret = ceph_osdc_setup();
818 pr_info("loaded (mon/osd proto %d/%d)\n",
819 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
826 ceph_crypto_shutdown();
828 ceph_debugfs_cleanup();
832 static void __exit exit_ceph_lib(void)
834 dout("exit_ceph_lib\n");
835 WARN_ON(!ceph_strings_empty());
839 ceph_crypto_shutdown();
840 ceph_debugfs_cleanup();
843 module_init(init_ceph_lib);
844 module_exit(exit_ceph_lib);
846 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
847 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
848 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
849 MODULE_DESCRIPTION("Ceph core library");
850 MODULE_LICENSE("GPL");