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 /* 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,
263 /* string args above */
266 Opt_cephx_require_signatures,
267 Opt_cephx_sign_messages,
272 static const struct fs_parameter_spec ceph_parameters[] = {
273 fsparam_flag ("abort_on_full", Opt_abort_on_full),
274 fsparam_flag_no ("cephx_require_signatures", Opt_cephx_require_signatures),
275 fsparam_flag_no ("cephx_sign_messages", Opt_cephx_sign_messages),
276 fsparam_flag_no ("crc", Opt_crc),
277 fsparam_string ("fsid", Opt_fsid),
278 fsparam_string ("ip", Opt_ip),
279 fsparam_string ("key", Opt_key),
280 fsparam_u32 ("mount_timeout", Opt_mount_timeout),
281 fsparam_string ("name", Opt_name),
282 fsparam_u32 ("osd_idle_ttl", Opt_osd_idle_ttl),
283 fsparam_u32 ("osd_request_timeout", Opt_osd_request_timeout),
284 fsparam_u32 ("osdkeepalive", Opt_osdkeepalivetimeout),
285 __fsparam (fs_param_is_s32, "osdtimeout", Opt_osdtimeout,
286 fs_param_deprecated, NULL),
287 fsparam_string ("secret", Opt_secret),
288 fsparam_flag_no ("share", Opt_share),
289 fsparam_flag_no ("tcp_nodelay", Opt_tcp_nodelay),
293 struct ceph_options *ceph_alloc_options(void)
295 struct ceph_options *opt;
297 opt = kzalloc(sizeof(*opt), GFP_KERNEL);
301 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
303 if (!opt->mon_addr) {
308 opt->flags = CEPH_OPT_DEFAULT;
309 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
310 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
311 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
312 opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
315 EXPORT_SYMBOL(ceph_alloc_options);
317 void ceph_destroy_options(struct ceph_options *opt)
319 dout("destroy_options %p\n", opt);
325 ceph_crypto_key_destroy(opt->key);
328 kfree(opt->mon_addr);
331 EXPORT_SYMBOL(ceph_destroy_options);
333 /* get secret from key store */
334 static int get_secret(struct ceph_crypto_key *dst, const char *name,
340 struct ceph_crypto_key *ckey;
342 ukey = request_key(&key_type_ceph, name, NULL);
344 /* request_key errors don't map nicely to mount(2)
345 errors; don't even try, but still printk */
346 key_err = PTR_ERR(ukey);
349 error_plog(log, "Failed due to key not found: %s",
353 error_plog(log, "Failed due to expired key: %s",
357 error_plog(log, "Failed due to revoked key: %s",
361 error_plog(log, "Failed due to key error %d: %s",
368 ckey = ukey->payload.data[0];
369 err = ceph_crypto_key_clone(dst, ckey);
372 /* pass through, err is 0 */
380 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
383 struct p_log log = {.prefix = "libceph", .log = l};
386 /* ip1[:port1][,ip2[:port2]...] */
387 ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
390 error_plog(&log, "Failed to parse monitor IPs: %d", ret);
396 EXPORT_SYMBOL(ceph_parse_mon_ips);
398 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
401 struct fs_parse_result result;
403 struct p_log log = {.prefix = "libceph", .log = l};
405 token = __fs_parse(&log, ceph_parameters, param, &result);
406 dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
412 err = ceph_parse_ips(param->string,
413 param->string + param->size,
417 error_plog(&log, "Failed to parse ip: %d", err);
420 opt->flags |= CEPH_OPT_MYIP;
424 err = parse_fsid(param->string, &opt->fsid);
426 error_plog(&log, "Failed to parse fsid: %d", err);
429 opt->flags |= CEPH_OPT_FSID;
433 opt->name = param->string;
434 param->string = NULL;
437 ceph_crypto_key_destroy(opt->key);
440 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
443 err = ceph_crypto_key_unarmor(opt->key, param->string);
445 error_plog(&log, "Failed to parse secret: %d", err);
450 ceph_crypto_key_destroy(opt->key);
453 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
456 return get_secret(opt->key, param->string, &log);
459 warn_plog(&log, "Ignoring osdtimeout");
461 case Opt_osdkeepalivetimeout:
462 /* 0 isn't well defined right now, reject it */
463 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
465 opt->osd_keepalive_timeout =
466 msecs_to_jiffies(result.uint_32 * 1000);
468 case Opt_osd_idle_ttl:
469 /* 0 isn't well defined right now, reject it */
470 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
472 opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
474 case Opt_mount_timeout:
475 /* 0 is "wait forever" (i.e. infinite timeout) */
476 if (result.uint_32 > INT_MAX / 1000)
478 opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
480 case Opt_osd_request_timeout:
481 /* 0 is "wait forever" (i.e. infinite timeout) */
482 if (result.uint_32 > INT_MAX / 1000)
484 opt->osd_request_timeout =
485 msecs_to_jiffies(result.uint_32 * 1000);
490 opt->flags &= ~CEPH_OPT_NOSHARE;
492 opt->flags |= CEPH_OPT_NOSHARE;
496 opt->flags &= ~CEPH_OPT_NOCRC;
498 opt->flags |= CEPH_OPT_NOCRC;
500 case Opt_cephx_require_signatures:
502 opt->flags &= ~CEPH_OPT_NOMSGAUTH;
504 opt->flags |= CEPH_OPT_NOMSGAUTH;
506 case Opt_cephx_sign_messages:
508 opt->flags &= ~CEPH_OPT_NOMSGSIGN;
510 opt->flags |= CEPH_OPT_NOMSGSIGN;
512 case Opt_tcp_nodelay:
514 opt->flags |= CEPH_OPT_TCP_NODELAY;
516 opt->flags &= ~CEPH_OPT_TCP_NODELAY;
519 case Opt_abort_on_full:
520 opt->flags |= CEPH_OPT_ABORT_ON_FULL;
530 return inval_plog(&log, "%s out of range", param->key);
532 EXPORT_SYMBOL(ceph_parse_param);
534 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
537 struct ceph_options *opt = client->options;
538 size_t pos = m->count;
541 seq_puts(m, "name=");
542 seq_escape(m, opt->name, ", \t\n\\");
546 seq_puts(m, "secret=<hidden>,");
548 if (opt->flags & CEPH_OPT_FSID)
549 seq_printf(m, "fsid=%pU,", &opt->fsid);
550 if (opt->flags & CEPH_OPT_NOSHARE)
551 seq_puts(m, "noshare,");
552 if (opt->flags & CEPH_OPT_NOCRC)
553 seq_puts(m, "nocrc,");
554 if (opt->flags & CEPH_OPT_NOMSGAUTH)
555 seq_puts(m, "nocephx_require_signatures,");
556 if (opt->flags & CEPH_OPT_NOMSGSIGN)
557 seq_puts(m, "nocephx_sign_messages,");
558 if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
559 seq_puts(m, "notcp_nodelay,");
560 if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
561 seq_puts(m, "abort_on_full,");
563 if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
564 seq_printf(m, "mount_timeout=%d,",
565 jiffies_to_msecs(opt->mount_timeout) / 1000);
566 if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
567 seq_printf(m, "osd_idle_ttl=%d,",
568 jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
569 if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
570 seq_printf(m, "osdkeepalivetimeout=%d,",
571 jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
572 if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
573 seq_printf(m, "osd_request_timeout=%d,",
574 jiffies_to_msecs(opt->osd_request_timeout) / 1000);
576 /* drop redundant comma */
582 EXPORT_SYMBOL(ceph_print_client_options);
584 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
586 return &client->msgr.inst.addr;
588 EXPORT_SYMBOL(ceph_client_addr);
590 u64 ceph_client_gid(struct ceph_client *client)
592 return client->monc.auth->global_id;
594 EXPORT_SYMBOL(ceph_client_gid);
597 * create a fresh client instance
599 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
601 struct ceph_client *client;
602 struct ceph_entity_addr *myaddr = NULL;
605 err = wait_for_random_bytes();
609 client = kzalloc(sizeof(*client), GFP_KERNEL);
611 return ERR_PTR(-ENOMEM);
613 client->private = private;
614 client->options = opt;
616 mutex_init(&client->mount_mutex);
617 init_waitqueue_head(&client->auth_wq);
618 client->auth_err = 0;
620 client->extra_mon_dispatch = NULL;
621 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
622 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
624 if (!ceph_test_opt(client, NOMSGAUTH))
625 client->required_features |= CEPH_FEATURE_MSG_AUTH;
628 if (ceph_test_opt(client, MYIP))
629 myaddr = &client->options->my_addr;
631 ceph_messenger_init(&client->msgr, myaddr);
634 err = ceph_monc_init(&client->monc, client);
637 err = ceph_osdc_init(&client->osdc, client);
644 ceph_monc_stop(&client->monc);
646 ceph_messenger_fini(&client->msgr);
650 EXPORT_SYMBOL(ceph_create_client);
652 void ceph_destroy_client(struct ceph_client *client)
654 dout("destroy_client %p\n", client);
656 atomic_set(&client->msgr.stopping, 1);
659 ceph_osdc_stop(&client->osdc);
660 ceph_monc_stop(&client->monc);
661 ceph_messenger_fini(&client->msgr);
663 ceph_debugfs_client_cleanup(client);
665 ceph_destroy_options(client->options);
668 dout("destroy_client %p done\n", client);
670 EXPORT_SYMBOL(ceph_destroy_client);
672 void ceph_reset_client_addr(struct ceph_client *client)
674 ceph_messenger_reset_nonce(&client->msgr);
675 ceph_monc_reopen_session(&client->monc);
676 ceph_osdc_reopen_osds(&client->osdc);
678 EXPORT_SYMBOL(ceph_reset_client_addr);
681 * true if we have the mon map (and have thus joined the cluster)
683 static bool have_mon_and_osd_map(struct ceph_client *client)
685 return client->monc.monmap && client->monc.monmap->epoch &&
686 client->osdc.osdmap && client->osdc.osdmap->epoch;
690 * mount: join the ceph cluster, and open root directory.
692 int __ceph_open_session(struct ceph_client *client, unsigned long started)
694 unsigned long timeout = client->options->mount_timeout;
697 /* open session, and wait for mon and osd maps */
698 err = ceph_monc_open_session(&client->monc);
702 while (!have_mon_and_osd_map(client)) {
703 if (timeout && time_after_eq(jiffies, started + timeout))
707 dout("mount waiting for mon_map\n");
708 err = wait_event_interruptible_timeout(client->auth_wq,
709 have_mon_and_osd_map(client) || (client->auth_err < 0),
710 ceph_timeout_jiffies(timeout));
713 if (client->auth_err < 0)
714 return client->auth_err;
717 pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
719 ceph_debugfs_client_init(client);
723 EXPORT_SYMBOL(__ceph_open_session);
725 int ceph_open_session(struct ceph_client *client)
728 unsigned long started = jiffies; /* note the start time */
730 dout("open_session start\n");
731 mutex_lock(&client->mount_mutex);
733 ret = __ceph_open_session(client, started);
735 mutex_unlock(&client->mount_mutex);
738 EXPORT_SYMBOL(ceph_open_session);
740 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
741 unsigned long timeout)
746 ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
750 if (client->osdc.osdmap->epoch >= newest_epoch)
753 ceph_osdc_maybe_request_map(&client->osdc);
754 return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
756 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
758 static int __init init_ceph_lib(void)
764 ret = ceph_crypto_init();
768 ret = ceph_msgr_init();
772 ret = ceph_osdc_setup();
776 pr_info("loaded (mon/osd proto %d/%d)\n",
777 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
784 ceph_crypto_shutdown();
786 ceph_debugfs_cleanup();
790 static void __exit exit_ceph_lib(void)
792 dout("exit_ceph_lib\n");
793 WARN_ON(!ceph_strings_empty());
797 ceph_crypto_shutdown();
798 ceph_debugfs_cleanup();
801 module_init(init_ceph_lib);
802 module_exit(exit_ceph_lib);
804 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
805 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
806 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
807 MODULE_DESCRIPTION("Ceph core library");
808 MODULE_LICENSE("GPL");