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/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>
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 ret = ceph_compare_crush_locs(&opt1->crush_locs, &opt2->crush_locs);
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,
191 EXPORT_SYMBOL(ceph_compare_options);
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.
199 * ceph_kvmalloc() may be called with GFP_KERNEL, GFP_NOFS or GFP_NOIO.
201 void *ceph_kvmalloc(size_t size, gfp_t flags)
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);
212 unsigned int noio_flag = memalloc_noio_save();
213 p = kvmalloc(size, GFP_KERNEL);
214 memalloc_noio_restore(noio_flag);
220 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
227 dout("parse_fsid '%s'\n", str);
229 while (*str && i < 16) {
234 if (!isxdigit(str[0]) || !isxdigit(str[1]))
238 if (sscanf(tmp, "%x", &d) < 1)
240 fsid->fsid[i] = d & 0xff;
247 dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
256 Opt_osdkeepalivetimeout,
259 Opt_osd_request_timeout,
267 Opt_read_from_replica,
268 /* string args above */
271 Opt_cephx_require_signatures,
272 Opt_cephx_sign_messages,
278 Opt_read_from_replica_no,
279 Opt_read_from_replica_balance,
280 Opt_read_from_replica_localize,
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},
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),
314 struct ceph_options *ceph_alloc_options(void)
316 struct ceph_options *opt;
318 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
322 opt->crush_locs = RB_ROOT;
323 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
325 if (!opt->mon_addr) {
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 opt->read_from_replica = CEPH_READ_FROM_REPLICA_DEFAULT;
338 EXPORT_SYMBOL(ceph_alloc_options);
340 void ceph_destroy_options(struct ceph_options *opt)
342 dout("destroy_options %p\n", opt);
346 ceph_clear_crush_locs(&opt->crush_locs);
349 ceph_crypto_key_destroy(opt->key);
352 kfree(opt->mon_addr);
355 EXPORT_SYMBOL(ceph_destroy_options);
357 /* get secret from key store */
358 static int get_secret(struct ceph_crypto_key *dst, const char *name,
364 struct ceph_crypto_key *ckey;
366 ukey = request_key(&key_type_ceph, name, NULL);
368 /* request_key errors don't map nicely to mount(2)
369 errors; don't even try, but still printk */
370 key_err = PTR_ERR(ukey);
373 error_plog(log, "Failed due to key not found: %s",
377 error_plog(log, "Failed due to expired key: %s",
381 error_plog(log, "Failed due to revoked key: %s",
385 error_plog(log, "Failed due to key error %d: %s",
392 ckey = ukey->payload.data[0];
393 err = ceph_crypto_key_clone(dst, ckey);
396 /* pass through, err is 0 */
404 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
407 struct p_log log = {.prefix = "libceph", .log = l};
410 /* ip1[:port1][,ip2[:port2]...] */
411 ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
414 error_plog(&log, "Failed to parse monitor IPs: %d", ret);
420 EXPORT_SYMBOL(ceph_parse_mon_ips);
422 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
425 struct fs_parse_result result;
427 struct p_log log = {.prefix = "libceph", .log = l};
429 token = __fs_parse(&log, ceph_parameters, param, &result);
430 dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
436 err = ceph_parse_ips(param->string,
437 param->string + param->size,
441 error_plog(&log, "Failed to parse ip: %d", err);
444 opt->flags |= CEPH_OPT_MYIP;
448 err = parse_fsid(param->string, &opt->fsid);
450 error_plog(&log, "Failed to parse fsid: %d", err);
453 opt->flags |= CEPH_OPT_FSID;
457 opt->name = param->string;
458 param->string = NULL;
461 ceph_crypto_key_destroy(opt->key);
464 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
467 err = ceph_crypto_key_unarmor(opt->key, param->string);
469 error_plog(&log, "Failed to parse secret: %d", err);
474 ceph_crypto_key_destroy(opt->key);
477 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
480 return get_secret(opt->key, param->string, &log);
481 case Opt_crush_location:
482 ceph_clear_crush_locs(&opt->crush_locs);
483 err = ceph_parse_crush_location(param->string,
486 error_plog(&log, "Failed to parse CRUSH location: %d",
491 case Opt_read_from_replica:
492 switch (result.uint_32) {
493 case Opt_read_from_replica_no:
494 opt->read_from_replica = 0;
496 case Opt_read_from_replica_balance:
497 opt->read_from_replica = CEPH_OSD_FLAG_BALANCE_READS;
499 case Opt_read_from_replica_localize:
500 opt->read_from_replica = CEPH_OSD_FLAG_LOCALIZE_READS;
508 warn_plog(&log, "Ignoring osdtimeout");
510 case Opt_osdkeepalivetimeout:
511 /* 0 isn't well defined right now, reject it */
512 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
514 opt->osd_keepalive_timeout =
515 msecs_to_jiffies(result.uint_32 * 1000);
517 case Opt_osd_idle_ttl:
518 /* 0 isn't well defined right now, reject it */
519 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
521 opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
523 case Opt_mount_timeout:
524 /* 0 is "wait forever" (i.e. infinite timeout) */
525 if (result.uint_32 > INT_MAX / 1000)
527 opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
529 case Opt_osd_request_timeout:
530 /* 0 is "wait forever" (i.e. infinite timeout) */
531 if (result.uint_32 > INT_MAX / 1000)
533 opt->osd_request_timeout =
534 msecs_to_jiffies(result.uint_32 * 1000);
539 opt->flags &= ~CEPH_OPT_NOSHARE;
541 opt->flags |= CEPH_OPT_NOSHARE;
545 opt->flags &= ~CEPH_OPT_NOCRC;
547 opt->flags |= CEPH_OPT_NOCRC;
549 case Opt_cephx_require_signatures:
551 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
553 opt->flags |= CEPH_OPT_NOMSGAUTH;
555 case Opt_cephx_sign_messages:
557 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
559 opt->flags |= CEPH_OPT_NOMSGSIGN;
561 case Opt_tcp_nodelay:
563 opt->flags |= CEPH_OPT_TCP_NODELAY;
565 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
568 case Opt_abort_on_full:
569 opt->flags |= CEPH_OPT_ABORT_ON_FULL;
579 return inval_plog(&log, "%s out of range", param->key);
581 EXPORT_SYMBOL(ceph_parse_param);
583 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
586 struct ceph_options *opt = client->options;
587 size_t pos = m->count;
591 seq_puts(m, "name=");
592 seq_escape(m, opt->name, ", \t\n\\");
596 seq_puts(m, "secret=<hidden>,");
598 if (!RB_EMPTY_ROOT(&opt->crush_locs)) {
599 seq_puts(m, "crush_location=");
600 for (n = rb_first(&opt->crush_locs); ; ) {
601 struct crush_loc_node *loc =
602 rb_entry(n, struct crush_loc_node, cl_node);
604 seq_printf(m, "%s:%s", loc->cl_loc.cl_type_name,
605 loc->cl_loc.cl_name);
614 if (opt->read_from_replica == CEPH_OSD_FLAG_BALANCE_READS) {
615 seq_puts(m, "read_from_replica=balance,");
616 } else if (opt->read_from_replica == CEPH_OSD_FLAG_LOCALIZE_READS) {
617 seq_puts(m, "read_from_replica=localize,");
620 if (opt->flags & CEPH_OPT_FSID)
621 seq_printf(m, "fsid=%pU,", &opt->fsid);
622 if (opt->flags & CEPH_OPT_NOSHARE)
623 seq_puts(m, "noshare,");
624 if (opt->flags & CEPH_OPT_NOCRC)
625 seq_puts(m, "nocrc,");
626 if (opt->flags & CEPH_OPT_NOMSGAUTH)
627 seq_puts(m, "nocephx_require_signatures,");
628 if (opt->flags & CEPH_OPT_NOMSGSIGN)
629 seq_puts(m, "nocephx_sign_messages,");
630 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
631 seq_puts(m, "notcp_nodelay,");
632 if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
633 seq_puts(m, "abort_on_full,");
635 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
636 seq_printf(m, "mount_timeout=%d,",
637 jiffies_to_msecs(opt->mount_timeout) / 1000);
638 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
639 seq_printf(m, "osd_idle_ttl=%d,",
640 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
641 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
642 seq_printf(m, "osdkeepalivetimeout=%d,",
643 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
644 if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
645 seq_printf(m, "osd_request_timeout=%d,",
646 jiffies_to_msecs(opt->osd_request_timeout) / 1000);
648 /* drop redundant comma */
654 EXPORT_SYMBOL(ceph_print_client_options);
656 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
658 return &client->msgr.inst.addr;
660 EXPORT_SYMBOL(ceph_client_addr);
662 u64 ceph_client_gid(struct ceph_client *client)
664 return client->monc.auth->global_id;
666 EXPORT_SYMBOL(ceph_client_gid);
669 * create a fresh client instance
671 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
673 struct ceph_client *client;
674 struct ceph_entity_addr *myaddr = NULL;
677 err = wait_for_random_bytes();
681 client = kzalloc(sizeof(*client), GFP_KERNEL);
683 return ERR_PTR(-ENOMEM);
685 client->private = private;
686 client->options = opt;
688 mutex_init(&client->mount_mutex);
689 init_waitqueue_head(&client->auth_wq);
690 client->auth_err = 0;
692 client->extra_mon_dispatch = NULL;
693 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
694 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
696 if (!ceph_test_opt(client, NOMSGAUTH))
697 client->required_features |= CEPH_FEATURE_MSG_AUTH;
700 if (ceph_test_opt(client, MYIP))
701 myaddr = &client->options->my_addr;
703 ceph_messenger_init(&client->msgr, myaddr);
706 err = ceph_monc_init(&client->monc, client);
709 err = ceph_osdc_init(&client->osdc, client);
716 ceph_monc_stop(&client->monc);
718 ceph_messenger_fini(&client->msgr);
722 EXPORT_SYMBOL(ceph_create_client);
724 void ceph_destroy_client(struct ceph_client *client)
726 dout("destroy_client %p\n", client);
728 atomic_set(&client->msgr.stopping, 1);
731 ceph_osdc_stop(&client->osdc);
732 ceph_monc_stop(&client->monc);
733 ceph_messenger_fini(&client->msgr);
735 ceph_debugfs_client_cleanup(client);
737 ceph_destroy_options(client->options);
740 dout("destroy_client %p done\n", client);
742 EXPORT_SYMBOL(ceph_destroy_client);
744 void ceph_reset_client_addr(struct ceph_client *client)
746 ceph_messenger_reset_nonce(&client->msgr);
747 ceph_monc_reopen_session(&client->monc);
748 ceph_osdc_reopen_osds(&client->osdc);
750 EXPORT_SYMBOL(ceph_reset_client_addr);
753 * true if we have the mon map (and have thus joined the cluster)
755 static bool have_mon_and_osd_map(struct ceph_client *client)
757 return client->monc.monmap && client->monc.monmap->epoch &&
758 client->osdc.osdmap && client->osdc.osdmap->epoch;
762 * mount: join the ceph cluster, and open root directory.
764 int __ceph_open_session(struct ceph_client *client, unsigned long started)
766 unsigned long timeout = client->options->mount_timeout;
769 /* open session, and wait for mon and osd maps */
770 err = ceph_monc_open_session(&client->monc);
774 while (!have_mon_and_osd_map(client)) {
775 if (timeout && time_after_eq(jiffies, started + timeout))
779 dout("mount waiting for mon_map\n");
780 err = wait_event_interruptible_timeout(client->auth_wq,
781 have_mon_and_osd_map(client) || (client->auth_err < 0),
782 ceph_timeout_jiffies(timeout));
785 if (client->auth_err < 0)
786 return client->auth_err;
789 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
791 ceph_debugfs_client_init(client);
795 EXPORT_SYMBOL(__ceph_open_session);
797 int ceph_open_session(struct ceph_client *client)
800 unsigned long started = jiffies; /* note the start time */
802 dout("open_session start\n");
803 mutex_lock(&client->mount_mutex);
805 ret = __ceph_open_session(client, started);
807 mutex_unlock(&client->mount_mutex);
810 EXPORT_SYMBOL(ceph_open_session);
812 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
813 unsigned long timeout)
818 ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
822 if (client->osdc.osdmap->epoch >= newest_epoch)
825 ceph_osdc_maybe_request_map(&client->osdc);
826 return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
828 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
830 static int __init init_ceph_lib(void)
836 ret = ceph_crypto_init();
840 ret = ceph_msgr_init();
844 ret = ceph_osdc_setup();
848 pr_info("loaded (mon/osd proto %d/%d)\n",
849 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
856 ceph_crypto_shutdown();
858 ceph_debugfs_cleanup();
862 static void __exit exit_ceph_lib(void)
864 dout("exit_ceph_lib\n");
865 WARN_ON(!ceph_strings_empty());
869 ceph_crypto_shutdown();
870 ceph_debugfs_cleanup();
873 module_init(init_ceph_lib);
874 module_exit(exit_ceph_lib);
879 MODULE_DESCRIPTION("Ceph core library");
880 MODULE_LICENSE("GPL");