2 #include <linux/ceph/ceph_debug.h>
3 #include <linux/backing-dev.h>
4 #include <linux/ctype.h>
6 #include <linux/inet.h>
9 #include <keys/ceph-type.h>
10 #include <linux/module.h>
11 #include <linux/mount.h>
12 #include <linux/nsproxy.h>
13 #include <linux/parser.h>
14 #include <linux/sched.h>
15 #include <linux/seq_file.h>
16 #include <linux/slab.h>
17 #include <linux/statfs.h>
18 #include <linux/string.h>
19 #include <linux/vmalloc.h>
22 #include <linux/ceph/ceph_features.h>
23 #include <linux/ceph/libceph.h>
24 #include <linux/ceph/debugfs.h>
25 #include <linux/ceph/decode.h>
26 #include <linux/ceph/mon_client.h>
27 #include <linux/ceph/auth.h>
32 * Module compatibility interface. For now it doesn't do anything,
33 * but its existence signals a certain level of functionality.
35 * The data buffer is used to pass information both to and from
36 * libceph. The return value indicates whether libceph determines
37 * it is compatible with the caller (from another kernel module),
38 * given the provided data.
40 * The data pointer can be null.
42 bool libceph_compatible(void *data
)
46 EXPORT_SYMBOL(libceph_compatible
);
48 static int param_get_supported_features(char *buffer
,
49 const struct kernel_param
*kp
)
51 return sprintf(buffer
, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT
);
53 static const struct kernel_param_ops param_ops_supported_features
= {
54 .get
= param_get_supported_features
,
56 module_param_cb(supported_features
, ¶m_ops_supported_features
, NULL
,
59 const char *ceph_msg_type_name(int type
)
62 case CEPH_MSG_SHUTDOWN
: return "shutdown";
63 case CEPH_MSG_PING
: return "ping";
64 case CEPH_MSG_AUTH
: return "auth";
65 case CEPH_MSG_AUTH_REPLY
: return "auth_reply";
66 case CEPH_MSG_MON_MAP
: return "mon_map";
67 case CEPH_MSG_MON_GET_MAP
: return "mon_get_map";
68 case CEPH_MSG_MON_SUBSCRIBE
: return "mon_subscribe";
69 case CEPH_MSG_MON_SUBSCRIBE_ACK
: return "mon_subscribe_ack";
70 case CEPH_MSG_STATFS
: return "statfs";
71 case CEPH_MSG_STATFS_REPLY
: return "statfs_reply";
72 case CEPH_MSG_MON_GET_VERSION
: return "mon_get_version";
73 case CEPH_MSG_MON_GET_VERSION_REPLY
: return "mon_get_version_reply";
74 case CEPH_MSG_MDS_MAP
: return "mds_map";
75 case CEPH_MSG_CLIENT_SESSION
: return "client_session";
76 case CEPH_MSG_CLIENT_RECONNECT
: return "client_reconnect";
77 case CEPH_MSG_CLIENT_REQUEST
: return "client_request";
78 case CEPH_MSG_CLIENT_REQUEST_FORWARD
: return "client_request_forward";
79 case CEPH_MSG_CLIENT_REPLY
: return "client_reply";
80 case CEPH_MSG_CLIENT_CAPS
: return "client_caps";
81 case CEPH_MSG_CLIENT_CAPRELEASE
: return "client_cap_release";
82 case CEPH_MSG_CLIENT_SNAP
: return "client_snap";
83 case CEPH_MSG_CLIENT_LEASE
: return "client_lease";
84 case CEPH_MSG_OSD_MAP
: return "osd_map";
85 case CEPH_MSG_OSD_OP
: return "osd_op";
86 case CEPH_MSG_OSD_OPREPLY
: return "osd_opreply";
87 case CEPH_MSG_WATCH_NOTIFY
: return "watch_notify";
88 case CEPH_MSG_OSD_BACKOFF
: return "osd_backoff";
89 default: return "unknown";
92 EXPORT_SYMBOL(ceph_msg_type_name
);
95 * Initially learn our fsid, or verify an fsid matches.
97 int ceph_check_fsid(struct ceph_client
*client
, struct ceph_fsid
*fsid
)
99 if (client
->have_fsid
) {
100 if (ceph_fsid_compare(&client
->fsid
, fsid
)) {
101 pr_err("bad fsid, had %pU got %pU",
102 &client
->fsid
, fsid
);
106 memcpy(&client
->fsid
, fsid
, sizeof(*fsid
));
110 EXPORT_SYMBOL(ceph_check_fsid
);
112 static int strcmp_null(const char *s1
, const char *s2
)
120 return strcmp(s1
, s2
);
123 int ceph_compare_options(struct ceph_options
*new_opt
,
124 struct ceph_client
*client
)
126 struct ceph_options
*opt1
= new_opt
;
127 struct ceph_options
*opt2
= client
->options
;
128 int ofs
= offsetof(struct ceph_options
, mon_addr
);
133 * Don't bother comparing options if network namespaces don't
136 if (!net_eq(current
->nsproxy
->net_ns
, read_pnet(&client
->msgr
.net
)))
139 ret
= memcmp(opt1
, opt2
, ofs
);
143 ret
= strcmp_null(opt1
->name
, opt2
->name
);
147 if (opt1
->key
&& !opt2
->key
)
149 if (!opt1
->key
&& opt2
->key
)
151 if (opt1
->key
&& opt2
->key
) {
152 if (opt1
->key
->type
!= opt2
->key
->type
)
154 if (opt1
->key
->created
.tv_sec
!= opt2
->key
->created
.tv_sec
)
156 if (opt1
->key
->created
.tv_nsec
!= opt2
->key
->created
.tv_nsec
)
158 if (opt1
->key
->len
!= opt2
->key
->len
)
160 if (opt1
->key
->key
&& !opt2
->key
->key
)
162 if (!opt1
->key
->key
&& opt2
->key
->key
)
164 if (opt1
->key
->key
&& opt2
->key
->key
) {
165 ret
= memcmp(opt1
->key
->key
, opt2
->key
->key
, opt1
->key
->len
);
171 /* any matching mon ip implies a match */
172 for (i
= 0; i
< opt1
->num_mon
; i
++) {
173 if (ceph_monmap_contains(client
->monc
.monmap
,
179 EXPORT_SYMBOL(ceph_compare_options
);
181 void *ceph_kvmalloc(size_t size
, gfp_t flags
)
183 if (size
<= (PAGE_SIZE
<< PAGE_ALLOC_COSTLY_ORDER
)) {
184 void *ptr
= kmalloc(size
, flags
| __GFP_NOWARN
);
189 return __vmalloc(size
, flags
, PAGE_KERNEL
);
193 static int parse_fsid(const char *str
, struct ceph_fsid
*fsid
)
200 dout("parse_fsid '%s'\n", str
);
202 while (*str
&& i
< 16) {
207 if (!isxdigit(str
[0]) || !isxdigit(str
[1]))
211 if (sscanf(tmp
, "%x", &d
) < 1)
213 fsid
->fsid
[i
] = d
& 0xff;
220 dout("parse_fsid ret %d got fsid %pU", err
, fsid
);
229 Opt_osdkeepalivetimeout
,
232 Opt_osd_request_timeout
,
241 /* string args above */
246 Opt_cephx_require_signatures
,
247 Opt_nocephx_require_signatures
,
248 Opt_cephx_sign_messages
,
249 Opt_nocephx_sign_messages
,
254 static match_table_t opt_tokens
= {
255 {Opt_osdtimeout
, "osdtimeout=%d"},
256 {Opt_osdkeepalivetimeout
, "osdkeepalive=%d"},
257 {Opt_mount_timeout
, "mount_timeout=%d"},
258 {Opt_osd_idle_ttl
, "osd_idle_ttl=%d"},
259 {Opt_osd_request_timeout
, "osd_request_timeout=%d"},
261 {Opt_fsid
, "fsid=%s"},
262 {Opt_name
, "name=%s"},
263 {Opt_secret
, "secret=%s"},
266 /* string args above */
267 {Opt_share
, "share"},
268 {Opt_noshare
, "noshare"},
270 {Opt_nocrc
, "nocrc"},
271 {Opt_cephx_require_signatures
, "cephx_require_signatures"},
272 {Opt_nocephx_require_signatures
, "nocephx_require_signatures"},
273 {Opt_cephx_sign_messages
, "cephx_sign_messages"},
274 {Opt_nocephx_sign_messages
, "nocephx_sign_messages"},
275 {Opt_tcp_nodelay
, "tcp_nodelay"},
276 {Opt_notcp_nodelay
, "notcp_nodelay"},
280 void ceph_destroy_options(struct ceph_options
*opt
)
282 dout("destroy_options %p\n", opt
);
285 ceph_crypto_key_destroy(opt
->key
);
288 kfree(opt
->mon_addr
);
291 EXPORT_SYMBOL(ceph_destroy_options
);
293 /* get secret from key store */
294 static int get_secret(struct ceph_crypto_key
*dst
, const char *name
) {
298 struct ceph_crypto_key
*ckey
;
300 ukey
= request_key(&key_type_ceph
, name
, NULL
);
301 if (!ukey
|| IS_ERR(ukey
)) {
302 /* request_key errors don't map nicely to mount(2)
303 errors; don't even try, but still printk */
304 key_err
= PTR_ERR(ukey
);
307 pr_warn("ceph: Mount failed due to key not found: %s\n",
311 pr_warn("ceph: Mount failed due to expired key: %s\n",
315 pr_warn("ceph: Mount failed due to revoked key: %s\n",
319 pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
326 ckey
= ukey
->payload
.data
[0];
327 err
= ceph_crypto_key_clone(dst
, ckey
);
330 /* pass through, err is 0 */
338 struct ceph_options
*
339 ceph_parse_options(char *options
, const char *dev_name
,
340 const char *dev_name_end
,
341 int (*parse_extra_token
)(char *c
, void *private),
344 struct ceph_options
*opt
;
347 substring_t argstr
[MAX_OPT_ARGS
];
349 opt
= kzalloc(sizeof(*opt
), GFP_KERNEL
);
351 return ERR_PTR(-ENOMEM
);
352 opt
->mon_addr
= kcalloc(CEPH_MAX_MON
, sizeof(*opt
->mon_addr
),
357 dout("parse_options %p options '%s' dev_name '%s'\n", opt
, options
,
360 /* start with defaults */
361 opt
->flags
= CEPH_OPT_DEFAULT
;
362 opt
->osd_keepalive_timeout
= CEPH_OSD_KEEPALIVE_DEFAULT
;
363 opt
->mount_timeout
= CEPH_MOUNT_TIMEOUT_DEFAULT
;
364 opt
->osd_idle_ttl
= CEPH_OSD_IDLE_TTL_DEFAULT
;
365 opt
->osd_request_timeout
= CEPH_OSD_REQUEST_TIMEOUT_DEFAULT
;
368 /* ip1[:port1][,ip2[:port2]...] */
369 err
= ceph_parse_ips(dev_name
, dev_name_end
, opt
->mon_addr
,
370 CEPH_MAX_MON
, &opt
->num_mon
);
374 /* parse mount options */
375 while ((c
= strsep(&options
, ",")) != NULL
) {
376 int token
, intval
, ret
;
380 token
= match_token((char *)c
, opt_tokens
, argstr
);
381 if (token
< 0 && parse_extra_token
) {
383 err
= parse_extra_token((char *)c
, private);
385 pr_err("bad option at '%s'\n", c
);
390 if (token
< Opt_last_int
) {
391 ret
= match_int(&argstr
[0], &intval
);
393 pr_err("bad mount option arg (not int) "
397 dout("got int token %d val %d\n", token
, intval
);
398 } else if (token
> Opt_last_int
&& token
< Opt_last_string
) {
399 dout("got string token %d val %s\n", token
,
402 dout("got token %d\n", token
);
406 err
= ceph_parse_ips(argstr
[0].from
,
412 opt
->flags
|= CEPH_OPT_MYIP
;
416 err
= parse_fsid(argstr
[0].from
, &opt
->fsid
);
418 opt
->flags
|= CEPH_OPT_FSID
;
421 opt
->name
= kstrndup(argstr
[0].from
,
422 argstr
[0].to
-argstr
[0].from
,
426 opt
->key
= kzalloc(sizeof(*opt
->key
), GFP_KERNEL
);
431 err
= ceph_crypto_key_unarmor(opt
->key
, argstr
[0].from
);
436 opt
->key
= kzalloc(sizeof(*opt
->key
), GFP_KERNEL
);
441 err
= get_secret(opt
->key
, argstr
[0].from
);
448 pr_warn("ignoring deprecated osdtimeout option\n");
450 case Opt_osdkeepalivetimeout
:
451 /* 0 isn't well defined right now, reject it */
452 if (intval
< 1 || intval
> INT_MAX
/ 1000) {
453 pr_err("osdkeepalive out of range\n");
457 opt
->osd_keepalive_timeout
=
458 msecs_to_jiffies(intval
* 1000);
460 case Opt_osd_idle_ttl
:
461 /* 0 isn't well defined right now, reject it */
462 if (intval
< 1 || intval
> INT_MAX
/ 1000) {
463 pr_err("osd_idle_ttl out of range\n");
467 opt
->osd_idle_ttl
= msecs_to_jiffies(intval
* 1000);
469 case Opt_mount_timeout
:
470 /* 0 is "wait forever" (i.e. infinite timeout) */
471 if (intval
< 0 || intval
> INT_MAX
/ 1000) {
472 pr_err("mount_timeout out of range\n");
476 opt
->mount_timeout
= msecs_to_jiffies(intval
* 1000);
478 case Opt_osd_request_timeout
:
479 /* 0 is "wait forever" (i.e. infinite timeout) */
480 if (intval
< 0 || intval
> INT_MAX
/ 1000) {
481 pr_err("osd_request_timeout out of range\n");
485 opt
->osd_request_timeout
= msecs_to_jiffies(intval
* 1000);
489 opt
->flags
&= ~CEPH_OPT_NOSHARE
;
492 opt
->flags
|= CEPH_OPT_NOSHARE
;
496 opt
->flags
&= ~CEPH_OPT_NOCRC
;
499 opt
->flags
|= CEPH_OPT_NOCRC
;
502 case Opt_cephx_require_signatures
:
503 opt
->flags
&= ~CEPH_OPT_NOMSGAUTH
;
505 case Opt_nocephx_require_signatures
:
506 opt
->flags
|= CEPH_OPT_NOMSGAUTH
;
508 case Opt_cephx_sign_messages
:
509 opt
->flags
&= ~CEPH_OPT_NOMSGSIGN
;
511 case Opt_nocephx_sign_messages
:
512 opt
->flags
|= CEPH_OPT_NOMSGSIGN
;
515 case Opt_tcp_nodelay
:
516 opt
->flags
|= CEPH_OPT_TCP_NODELAY
;
518 case Opt_notcp_nodelay
:
519 opt
->flags
&= ~CEPH_OPT_TCP_NODELAY
;
531 ceph_destroy_options(opt
);
534 EXPORT_SYMBOL(ceph_parse_options
);
536 int ceph_print_client_options(struct seq_file
*m
, struct ceph_client
*client
)
538 struct ceph_options
*opt
= client
->options
;
539 size_t pos
= m
->count
;
542 seq_puts(m
, "name=");
543 seq_escape(m
, opt
->name
, ", \t\n\\");
547 seq_puts(m
, "secret=<hidden>,");
549 if (opt
->flags
& CEPH_OPT_FSID
)
550 seq_printf(m
, "fsid=%pU,", &opt
->fsid
);
551 if (opt
->flags
& CEPH_OPT_NOSHARE
)
552 seq_puts(m
, "noshare,");
553 if (opt
->flags
& CEPH_OPT_NOCRC
)
554 seq_puts(m
, "nocrc,");
555 if (opt
->flags
& CEPH_OPT_NOMSGAUTH
)
556 seq_puts(m
, "nocephx_require_signatures,");
557 if (opt
->flags
& CEPH_OPT_NOMSGSIGN
)
558 seq_puts(m
, "nocephx_sign_messages,");
559 if ((opt
->flags
& CEPH_OPT_TCP_NODELAY
) == 0)
560 seq_puts(m
, "notcp_nodelay,");
562 if (opt
->mount_timeout
!= CEPH_MOUNT_TIMEOUT_DEFAULT
)
563 seq_printf(m
, "mount_timeout=%d,",
564 jiffies_to_msecs(opt
->mount_timeout
) / 1000);
565 if (opt
->osd_idle_ttl
!= CEPH_OSD_IDLE_TTL_DEFAULT
)
566 seq_printf(m
, "osd_idle_ttl=%d,",
567 jiffies_to_msecs(opt
->osd_idle_ttl
) / 1000);
568 if (opt
->osd_keepalive_timeout
!= CEPH_OSD_KEEPALIVE_DEFAULT
)
569 seq_printf(m
, "osdkeepalivetimeout=%d,",
570 jiffies_to_msecs(opt
->osd_keepalive_timeout
) / 1000);
571 if (opt
->osd_request_timeout
!= CEPH_OSD_REQUEST_TIMEOUT_DEFAULT
)
572 seq_printf(m
, "osd_request_timeout=%d,",
573 jiffies_to_msecs(opt
->osd_request_timeout
) / 1000);
575 /* drop redundant comma */
581 EXPORT_SYMBOL(ceph_print_client_options
);
583 struct ceph_entity_addr
*ceph_client_addr(struct ceph_client
*client
)
585 return &client
->msgr
.inst
.addr
;
587 EXPORT_SYMBOL(ceph_client_addr
);
589 u64
ceph_client_gid(struct ceph_client
*client
)
591 return client
->monc
.auth
->global_id
;
593 EXPORT_SYMBOL(ceph_client_gid
);
596 * create a fresh client instance
598 struct ceph_client
*ceph_create_client(struct ceph_options
*opt
, void *private)
600 struct ceph_client
*client
;
601 struct ceph_entity_addr
*myaddr
= NULL
;
604 err
= wait_for_random_bytes();
608 client
= kzalloc(sizeof(*client
), GFP_KERNEL
);
610 return ERR_PTR(-ENOMEM
);
612 client
->private = private;
613 client
->options
= opt
;
615 mutex_init(&client
->mount_mutex
);
616 init_waitqueue_head(&client
->auth_wq
);
617 client
->auth_err
= 0;
619 client
->extra_mon_dispatch
= NULL
;
620 client
->supported_features
= CEPH_FEATURES_SUPPORTED_DEFAULT
;
621 client
->required_features
= CEPH_FEATURES_REQUIRED_DEFAULT
;
623 if (!ceph_test_opt(client
, NOMSGAUTH
))
624 client
->required_features
|= CEPH_FEATURE_MSG_AUTH
;
627 if (ceph_test_opt(client
, MYIP
))
628 myaddr
= &client
->options
->my_addr
;
630 ceph_messenger_init(&client
->msgr
, myaddr
);
633 err
= ceph_monc_init(&client
->monc
, client
);
636 err
= ceph_osdc_init(&client
->osdc
, client
);
643 ceph_monc_stop(&client
->monc
);
645 ceph_messenger_fini(&client
->msgr
);
649 EXPORT_SYMBOL(ceph_create_client
);
651 void ceph_destroy_client(struct ceph_client
*client
)
653 dout("destroy_client %p\n", client
);
655 atomic_set(&client
->msgr
.stopping
, 1);
658 ceph_osdc_stop(&client
->osdc
);
659 ceph_monc_stop(&client
->monc
);
660 ceph_messenger_fini(&client
->msgr
);
662 ceph_debugfs_client_cleanup(client
);
664 ceph_destroy_options(client
->options
);
667 dout("destroy_client %p done\n", client
);
669 EXPORT_SYMBOL(ceph_destroy_client
);
672 * true if we have the mon map (and have thus joined the cluster)
674 static bool have_mon_and_osd_map(struct ceph_client
*client
)
676 return client
->monc
.monmap
&& client
->monc
.monmap
->epoch
&&
677 client
->osdc
.osdmap
&& client
->osdc
.osdmap
->epoch
;
681 * mount: join the ceph cluster, and open root directory.
683 int __ceph_open_session(struct ceph_client
*client
, unsigned long started
)
685 unsigned long timeout
= client
->options
->mount_timeout
;
688 /* open session, and wait for mon and osd maps */
689 err
= ceph_monc_open_session(&client
->monc
);
693 while (!have_mon_and_osd_map(client
)) {
694 if (timeout
&& time_after_eq(jiffies
, started
+ timeout
))
698 dout("mount waiting for mon_map\n");
699 err
= wait_event_interruptible_timeout(client
->auth_wq
,
700 have_mon_and_osd_map(client
) || (client
->auth_err
< 0),
701 ceph_timeout_jiffies(timeout
));
704 if (client
->auth_err
< 0)
705 return client
->auth_err
;
708 pr_info("client%llu fsid %pU\n", ceph_client_gid(client
),
710 ceph_debugfs_client_init(client
);
714 EXPORT_SYMBOL(__ceph_open_session
);
717 int ceph_open_session(struct ceph_client
*client
)
720 unsigned long started
= jiffies
; /* note the start time */
722 dout("open_session start\n");
723 mutex_lock(&client
->mount_mutex
);
725 ret
= __ceph_open_session(client
, started
);
727 mutex_unlock(&client
->mount_mutex
);
730 EXPORT_SYMBOL(ceph_open_session
);
733 static int __init
init_ceph_lib(void)
737 ret
= ceph_debugfs_init();
741 ret
= ceph_crypto_init();
745 ret
= ceph_msgr_init();
749 ret
= ceph_osdc_setup();
753 pr_info("loaded (mon/osd proto %d/%d)\n",
754 CEPH_MONC_PROTOCOL
, CEPH_OSDC_PROTOCOL
);
761 ceph_crypto_shutdown();
763 ceph_debugfs_cleanup();
768 static void __exit
exit_ceph_lib(void)
770 dout("exit_ceph_lib\n");
771 WARN_ON(!ceph_strings_empty());
775 ceph_crypto_shutdown();
776 ceph_debugfs_cleanup();
779 module_init(init_ceph_lib
);
780 module_exit(exit_ceph_lib
);
782 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
783 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
784 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
785 MODULE_DESCRIPTION("Ceph core library");
786 MODULE_LICENSE("GPL");