1 #include <linux/ceph/ceph_debug.h>
4 #include <linux/wait.h>
5 #include <linux/slab.h>
7 #include <linux/sched.h>
8 #include <linux/debugfs.h>
9 #include <linux/seq_file.h>
10 #include <linux/utsname.h>
11 #include <linux/ratelimit.h>
14 #include "mds_client.h"
16 #include <linux/ceph/ceph_features.h>
17 #include <linux/ceph/messenger.h>
18 #include <linux/ceph/decode.h>
19 #include <linux/ceph/pagelist.h>
20 #include <linux/ceph/auth.h>
21 #include <linux/ceph/debugfs.h>
24 * A cluster of MDS (metadata server) daemons is responsible for
25 * managing the file system namespace (the directory hierarchy and
26 * inodes) and for coordinating shared access to storage. Metadata is
27 * partitioning hierarchically across a number of servers, and that
28 * partition varies over time as the cluster adjusts the distribution
29 * in order to balance load.
31 * The MDS client is primarily responsible to managing synchronous
32 * metadata requests for operations like open, unlink, and so forth.
33 * If there is a MDS failure, we find out about it when we (possibly
34 * request and) receive a new MDS map, and can resubmit affected
37 * For the most part, though, we take advantage of a lossless
38 * communications channel to the MDS, and do not need to worry about
39 * timing out or resubmitting requests.
41 * We maintain a stateful "session" with each MDS we interact with.
42 * Within each session, we sent periodic heartbeat messages to ensure
43 * any capabilities or leases we have been issues remain valid. If
44 * the session times out and goes stale, our leases and capabilities
45 * are no longer valid.
48 struct ceph_reconnect_state
{
50 struct ceph_pagelist
*pagelist
;
54 static void __wake_requests(struct ceph_mds_client
*mdsc
,
55 struct list_head
*head
);
57 static const struct ceph_connection_operations mds_con_ops
;
65 * parse individual inode info
67 static int parse_reply_info_in(void **p
, void *end
,
68 struct ceph_mds_reply_info_in
*info
,
74 *p
+= sizeof(struct ceph_mds_reply_inode
) +
75 sizeof(*info
->in
->fragtree
.splits
) *
76 le32_to_cpu(info
->in
->fragtree
.nsplits
);
78 ceph_decode_32_safe(p
, end
, info
->symlink_len
, bad
);
79 ceph_decode_need(p
, end
, info
->symlink_len
, bad
);
81 *p
+= info
->symlink_len
;
83 if (features
& CEPH_FEATURE_DIRLAYOUTHASH
)
84 ceph_decode_copy_safe(p
, end
, &info
->dir_layout
,
85 sizeof(info
->dir_layout
), bad
);
87 memset(&info
->dir_layout
, 0, sizeof(info
->dir_layout
));
89 ceph_decode_32_safe(p
, end
, info
->xattr_len
, bad
);
90 ceph_decode_need(p
, end
, info
->xattr_len
, bad
);
91 info
->xattr_data
= *p
;
92 *p
+= info
->xattr_len
;
94 if (features
& CEPH_FEATURE_MDS_INLINE_DATA
) {
95 ceph_decode_64_safe(p
, end
, info
->inline_version
, bad
);
96 ceph_decode_32_safe(p
, end
, info
->inline_len
, bad
);
97 ceph_decode_need(p
, end
, info
->inline_len
, bad
);
98 info
->inline_data
= *p
;
99 *p
+= info
->inline_len
;
101 info
->inline_version
= CEPH_INLINE_NONE
;
103 if (features
& CEPH_FEATURE_FS_FILE_LAYOUT_V2
) {
104 ceph_decode_32_safe(p
, end
, info
->pool_ns_len
, bad
);
105 ceph_decode_need(p
, end
, info
->pool_ns_len
, bad
);
106 *p
+= info
->pool_ns_len
;
108 info
->pool_ns_len
= 0;
117 * parse a normal reply, which may contain a (dir+)dentry and/or a
120 static int parse_reply_info_trace(void **p
, void *end
,
121 struct ceph_mds_reply_info_parsed
*info
,
126 if (info
->head
->is_dentry
) {
127 err
= parse_reply_info_in(p
, end
, &info
->diri
, features
);
131 if (unlikely(*p
+ sizeof(*info
->dirfrag
) > end
))
134 *p
+= sizeof(*info
->dirfrag
) +
135 sizeof(u32
)*le32_to_cpu(info
->dirfrag
->ndist
);
136 if (unlikely(*p
> end
))
139 ceph_decode_32_safe(p
, end
, info
->dname_len
, bad
);
140 ceph_decode_need(p
, end
, info
->dname_len
, bad
);
142 *p
+= info
->dname_len
;
144 *p
+= sizeof(*info
->dlease
);
147 if (info
->head
->is_target
) {
148 err
= parse_reply_info_in(p
, end
, &info
->targeti
, features
);
153 if (unlikely(*p
!= end
))
160 pr_err("problem parsing mds trace %d\n", err
);
165 * parse readdir results
167 static int parse_reply_info_dir(void **p
, void *end
,
168 struct ceph_mds_reply_info_parsed
*info
,
175 if (*p
+ sizeof(*info
->dir_dir
) > end
)
177 *p
+= sizeof(*info
->dir_dir
) +
178 sizeof(u32
)*le32_to_cpu(info
->dir_dir
->ndist
);
182 ceph_decode_need(p
, end
, sizeof(num
) + 2, bad
);
183 num
= ceph_decode_32(p
);
185 u16 flags
= ceph_decode_16(p
);
186 info
->dir_end
= !!(flags
& CEPH_READDIR_FRAG_END
);
187 info
->dir_complete
= !!(flags
& CEPH_READDIR_FRAG_COMPLETE
);
188 info
->hash_order
= !!(flags
& CEPH_READDIR_HASH_ORDER
);
193 BUG_ON(!info
->dir_entries
);
194 if ((unsigned long)(info
->dir_entries
+ num
) >
195 (unsigned long)info
->dir_entries
+ info
->dir_buf_size
) {
196 pr_err("dir contents are larger than expected\n");
203 struct ceph_mds_reply_dir_entry
*rde
= info
->dir_entries
+ i
;
205 ceph_decode_need(p
, end
, sizeof(u32
)*2, bad
);
206 rde
->name_len
= ceph_decode_32(p
);
207 ceph_decode_need(p
, end
, rde
->name_len
, bad
);
210 dout("parsed dir dname '%.*s'\n", rde
->name_len
, rde
->name
);
212 *p
+= sizeof(struct ceph_mds_reply_lease
);
215 err
= parse_reply_info_in(p
, end
, &rde
->inode
, features
);
218 /* ceph_readdir_prepopulate() will update it */
232 pr_err("problem parsing dir contents %d\n", err
);
237 * parse fcntl F_GETLK results
239 static int parse_reply_info_filelock(void **p
, void *end
,
240 struct ceph_mds_reply_info_parsed
*info
,
243 if (*p
+ sizeof(*info
->filelock_reply
) > end
)
246 info
->filelock_reply
= *p
;
247 *p
+= sizeof(*info
->filelock_reply
);
249 if (unlikely(*p
!= end
))
258 * parse create results
260 static int parse_reply_info_create(void **p
, void *end
,
261 struct ceph_mds_reply_info_parsed
*info
,
264 if (features
& CEPH_FEATURE_REPLY_CREATE_INODE
) {
266 info
->has_create_ino
= false;
268 info
->has_create_ino
= true;
269 info
->ino
= ceph_decode_64(p
);
273 if (unlikely(*p
!= end
))
282 * parse extra results
284 static int parse_reply_info_extra(void **p
, void *end
,
285 struct ceph_mds_reply_info_parsed
*info
,
288 if (info
->head
->op
== CEPH_MDS_OP_GETFILELOCK
)
289 return parse_reply_info_filelock(p
, end
, info
, features
);
290 else if (info
->head
->op
== CEPH_MDS_OP_READDIR
||
291 info
->head
->op
== CEPH_MDS_OP_LSSNAP
)
292 return parse_reply_info_dir(p
, end
, info
, features
);
293 else if (info
->head
->op
== CEPH_MDS_OP_CREATE
)
294 return parse_reply_info_create(p
, end
, info
, features
);
300 * parse entire mds reply
302 static int parse_reply_info(struct ceph_msg
*msg
,
303 struct ceph_mds_reply_info_parsed
*info
,
310 info
->head
= msg
->front
.iov_base
;
311 p
= msg
->front
.iov_base
+ sizeof(struct ceph_mds_reply_head
);
312 end
= p
+ msg
->front
.iov_len
- sizeof(struct ceph_mds_reply_head
);
315 ceph_decode_32_safe(&p
, end
, len
, bad
);
317 ceph_decode_need(&p
, end
, len
, bad
);
318 err
= parse_reply_info_trace(&p
, p
+len
, info
, features
);
324 ceph_decode_32_safe(&p
, end
, len
, bad
);
326 ceph_decode_need(&p
, end
, len
, bad
);
327 err
= parse_reply_info_extra(&p
, p
+len
, info
, features
);
333 ceph_decode_32_safe(&p
, end
, len
, bad
);
334 info
->snapblob_len
= len
;
345 pr_err("mds parse_reply err %d\n", err
);
349 static void destroy_reply_info(struct ceph_mds_reply_info_parsed
*info
)
351 if (!info
->dir_entries
)
353 free_pages((unsigned long)info
->dir_entries
, get_order(info
->dir_buf_size
));
360 const char *ceph_session_state_name(int s
)
363 case CEPH_MDS_SESSION_NEW
: return "new";
364 case CEPH_MDS_SESSION_OPENING
: return "opening";
365 case CEPH_MDS_SESSION_OPEN
: return "open";
366 case CEPH_MDS_SESSION_HUNG
: return "hung";
367 case CEPH_MDS_SESSION_CLOSING
: return "closing";
368 case CEPH_MDS_SESSION_RESTARTING
: return "restarting";
369 case CEPH_MDS_SESSION_RECONNECTING
: return "reconnecting";
370 default: return "???";
374 static struct ceph_mds_session
*get_session(struct ceph_mds_session
*s
)
376 if (atomic_inc_not_zero(&s
->s_ref
)) {
377 dout("mdsc get_session %p %d -> %d\n", s
,
378 atomic_read(&s
->s_ref
)-1, atomic_read(&s
->s_ref
));
381 dout("mdsc get_session %p 0 -- FAIL", s
);
386 void ceph_put_mds_session(struct ceph_mds_session
*s
)
388 dout("mdsc put_session %p %d -> %d\n", s
,
389 atomic_read(&s
->s_ref
), atomic_read(&s
->s_ref
)-1);
390 if (atomic_dec_and_test(&s
->s_ref
)) {
391 if (s
->s_auth
.authorizer
)
392 ceph_auth_destroy_authorizer(s
->s_auth
.authorizer
);
398 * called under mdsc->mutex
400 struct ceph_mds_session
*__ceph_lookup_mds_session(struct ceph_mds_client
*mdsc
,
403 struct ceph_mds_session
*session
;
405 if (mds
>= mdsc
->max_sessions
|| mdsc
->sessions
[mds
] == NULL
)
407 session
= mdsc
->sessions
[mds
];
408 dout("lookup_mds_session %p %d\n", session
,
409 atomic_read(&session
->s_ref
));
410 get_session(session
);
414 static bool __have_session(struct ceph_mds_client
*mdsc
, int mds
)
416 if (mds
>= mdsc
->max_sessions
)
418 return mdsc
->sessions
[mds
];
421 static int __verify_registered_session(struct ceph_mds_client
*mdsc
,
422 struct ceph_mds_session
*s
)
424 if (s
->s_mds
>= mdsc
->max_sessions
||
425 mdsc
->sessions
[s
->s_mds
] != s
)
431 * create+register a new session for given mds.
432 * called under mdsc->mutex.
434 static struct ceph_mds_session
*register_session(struct ceph_mds_client
*mdsc
,
437 struct ceph_mds_session
*s
;
439 if (mds
>= mdsc
->mdsmap
->m_max_mds
)
440 return ERR_PTR(-EINVAL
);
442 s
= kzalloc(sizeof(*s
), GFP_NOFS
);
444 return ERR_PTR(-ENOMEM
);
447 s
->s_state
= CEPH_MDS_SESSION_NEW
;
450 mutex_init(&s
->s_mutex
);
452 ceph_con_init(&s
->s_con
, s
, &mds_con_ops
, &mdsc
->fsc
->client
->msgr
);
454 spin_lock_init(&s
->s_gen_ttl_lock
);
456 s
->s_cap_ttl
= jiffies
- 1;
458 spin_lock_init(&s
->s_cap_lock
);
459 s
->s_renew_requested
= 0;
461 INIT_LIST_HEAD(&s
->s_caps
);
464 atomic_set(&s
->s_ref
, 1);
465 INIT_LIST_HEAD(&s
->s_waiting
);
466 INIT_LIST_HEAD(&s
->s_unsafe
);
467 s
->s_num_cap_releases
= 0;
468 s
->s_cap_reconnect
= 0;
469 s
->s_cap_iterator
= NULL
;
470 INIT_LIST_HEAD(&s
->s_cap_releases
);
471 INIT_LIST_HEAD(&s
->s_cap_flushing
);
472 INIT_LIST_HEAD(&s
->s_cap_snaps_flushing
);
474 dout("register_session mds%d\n", mds
);
475 if (mds
>= mdsc
->max_sessions
) {
476 int newmax
= 1 << get_count_order(mds
+1);
477 struct ceph_mds_session
**sa
;
479 dout("register_session realloc to %d\n", newmax
);
480 sa
= kcalloc(newmax
, sizeof(void *), GFP_NOFS
);
483 if (mdsc
->sessions
) {
484 memcpy(sa
, mdsc
->sessions
,
485 mdsc
->max_sessions
* sizeof(void *));
486 kfree(mdsc
->sessions
);
489 mdsc
->max_sessions
= newmax
;
491 mdsc
->sessions
[mds
] = s
;
492 atomic_inc(&mdsc
->num_sessions
);
493 atomic_inc(&s
->s_ref
); /* one ref to sessions[], one to caller */
495 ceph_con_open(&s
->s_con
, CEPH_ENTITY_TYPE_MDS
, mds
,
496 ceph_mdsmap_get_addr(mdsc
->mdsmap
, mds
));
502 return ERR_PTR(-ENOMEM
);
506 * called under mdsc->mutex
508 static void __unregister_session(struct ceph_mds_client
*mdsc
,
509 struct ceph_mds_session
*s
)
511 dout("__unregister_session mds%d %p\n", s
->s_mds
, s
);
512 BUG_ON(mdsc
->sessions
[s
->s_mds
] != s
);
513 mdsc
->sessions
[s
->s_mds
] = NULL
;
514 ceph_con_close(&s
->s_con
);
515 ceph_put_mds_session(s
);
516 atomic_dec(&mdsc
->num_sessions
);
520 * drop session refs in request.
522 * should be last request ref, or hold mdsc->mutex
524 static void put_request_session(struct ceph_mds_request
*req
)
526 if (req
->r_session
) {
527 ceph_put_mds_session(req
->r_session
);
528 req
->r_session
= NULL
;
532 void ceph_mdsc_release_request(struct kref
*kref
)
534 struct ceph_mds_request
*req
= container_of(kref
,
535 struct ceph_mds_request
,
537 destroy_reply_info(&req
->r_reply_info
);
539 ceph_msg_put(req
->r_request
);
541 ceph_msg_put(req
->r_reply
);
543 ceph_put_cap_refs(ceph_inode(req
->r_inode
), CEPH_CAP_PIN
);
546 if (req
->r_locked_dir
)
547 ceph_put_cap_refs(ceph_inode(req
->r_locked_dir
), CEPH_CAP_PIN
);
548 iput(req
->r_target_inode
);
551 if (req
->r_old_dentry
)
552 dput(req
->r_old_dentry
);
553 if (req
->r_old_dentry_dir
) {
555 * track (and drop pins for) r_old_dentry_dir
556 * separately, since r_old_dentry's d_parent may have
557 * changed between the dir mutex being dropped and
558 * this request being freed.
560 ceph_put_cap_refs(ceph_inode(req
->r_old_dentry_dir
),
562 iput(req
->r_old_dentry_dir
);
567 ceph_pagelist_release(req
->r_pagelist
);
568 put_request_session(req
);
569 ceph_unreserve_caps(req
->r_mdsc
, &req
->r_caps_reservation
);
573 DEFINE_RB_FUNCS(request
, struct ceph_mds_request
, r_tid
, r_node
)
576 * lookup session, bump ref if found.
578 * called under mdsc->mutex.
580 static struct ceph_mds_request
*
581 lookup_get_request(struct ceph_mds_client
*mdsc
, u64 tid
)
583 struct ceph_mds_request
*req
;
585 req
= lookup_request(&mdsc
->request_tree
, tid
);
587 ceph_mdsc_get_request(req
);
593 * Register an in-flight request, and assign a tid. Link to directory
594 * are modifying (if any).
596 * Called under mdsc->mutex.
598 static void __register_request(struct ceph_mds_client
*mdsc
,
599 struct ceph_mds_request
*req
,
602 req
->r_tid
= ++mdsc
->last_tid
;
604 ceph_reserve_caps(mdsc
, &req
->r_caps_reservation
,
606 dout("__register_request %p tid %lld\n", req
, req
->r_tid
);
607 ceph_mdsc_get_request(req
);
608 insert_request(&mdsc
->request_tree
, req
);
610 req
->r_uid
= current_fsuid();
611 req
->r_gid
= current_fsgid();
613 if (mdsc
->oldest_tid
== 0 && req
->r_op
!= CEPH_MDS_OP_SETFILELOCK
)
614 mdsc
->oldest_tid
= req
->r_tid
;
618 req
->r_unsafe_dir
= dir
;
622 static void __unregister_request(struct ceph_mds_client
*mdsc
,
623 struct ceph_mds_request
*req
)
625 dout("__unregister_request %p tid %lld\n", req
, req
->r_tid
);
627 if (req
->r_tid
== mdsc
->oldest_tid
) {
628 struct rb_node
*p
= rb_next(&req
->r_node
);
629 mdsc
->oldest_tid
= 0;
631 struct ceph_mds_request
*next_req
=
632 rb_entry(p
, struct ceph_mds_request
, r_node
);
633 if (next_req
->r_op
!= CEPH_MDS_OP_SETFILELOCK
) {
634 mdsc
->oldest_tid
= next_req
->r_tid
;
641 erase_request(&mdsc
->request_tree
, req
);
643 if (req
->r_unsafe_dir
&& req
->r_got_unsafe
) {
644 struct ceph_inode_info
*ci
= ceph_inode(req
->r_unsafe_dir
);
645 spin_lock(&ci
->i_unsafe_lock
);
646 list_del_init(&req
->r_unsafe_dir_item
);
647 spin_unlock(&ci
->i_unsafe_lock
);
649 if (req
->r_target_inode
&& req
->r_got_unsafe
) {
650 struct ceph_inode_info
*ci
= ceph_inode(req
->r_target_inode
);
651 spin_lock(&ci
->i_unsafe_lock
);
652 list_del_init(&req
->r_unsafe_target_item
);
653 spin_unlock(&ci
->i_unsafe_lock
);
656 if (req
->r_unsafe_dir
) {
657 iput(req
->r_unsafe_dir
);
658 req
->r_unsafe_dir
= NULL
;
661 complete_all(&req
->r_safe_completion
);
663 ceph_mdsc_put_request(req
);
667 * Choose mds to send request to next. If there is a hint set in the
668 * request (e.g., due to a prior forward hint from the mds), use that.
669 * Otherwise, consult frag tree and/or caps to identify the
670 * appropriate mds. If all else fails, choose randomly.
672 * Called under mdsc->mutex.
674 static struct dentry
*get_nonsnap_parent(struct dentry
*dentry
)
677 * we don't need to worry about protecting the d_parent access
678 * here because we never renaming inside the snapped namespace
679 * except to resplice to another snapdir, and either the old or new
680 * result is a valid result.
682 while (!IS_ROOT(dentry
) && ceph_snap(d_inode(dentry
)) != CEPH_NOSNAP
)
683 dentry
= dentry
->d_parent
;
687 static int __choose_mds(struct ceph_mds_client
*mdsc
,
688 struct ceph_mds_request
*req
)
691 struct ceph_inode_info
*ci
;
692 struct ceph_cap
*cap
;
693 int mode
= req
->r_direct_mode
;
695 u32 hash
= req
->r_direct_hash
;
696 bool is_hash
= req
->r_direct_is_hash
;
699 * is there a specific mds we should try? ignore hint if we have
700 * no session and the mds is not up (active or recovering).
702 if (req
->r_resend_mds
>= 0 &&
703 (__have_session(mdsc
, req
->r_resend_mds
) ||
704 ceph_mdsmap_get_state(mdsc
->mdsmap
, req
->r_resend_mds
) > 0)) {
705 dout("choose_mds using resend_mds mds%d\n",
707 return req
->r_resend_mds
;
710 if (mode
== USE_RANDOM_MDS
)
715 inode
= req
->r_inode
;
716 } else if (req
->r_dentry
) {
717 /* ignore race with rename; old or new d_parent is okay */
718 struct dentry
*parent
= req
->r_dentry
->d_parent
;
719 struct inode
*dir
= d_inode(parent
);
721 if (dir
->i_sb
!= mdsc
->fsc
->sb
) {
723 inode
= d_inode(req
->r_dentry
);
724 } else if (ceph_snap(dir
) != CEPH_NOSNAP
) {
725 /* direct snapped/virtual snapdir requests
726 * based on parent dir inode */
727 struct dentry
*dn
= get_nonsnap_parent(parent
);
729 dout("__choose_mds using nonsnap parent %p\n", inode
);
732 inode
= d_inode(req
->r_dentry
);
733 if (!inode
|| mode
== USE_AUTH_MDS
) {
736 hash
= ceph_dentry_hash(dir
, req
->r_dentry
);
742 dout("__choose_mds %p is_hash=%d (%d) mode %d\n", inode
, (int)is_hash
,
746 ci
= ceph_inode(inode
);
748 if (is_hash
&& S_ISDIR(inode
->i_mode
)) {
749 struct ceph_inode_frag frag
;
752 ceph_choose_frag(ci
, hash
, &frag
, &found
);
754 if (mode
== USE_ANY_MDS
&& frag
.ndist
> 0) {
757 /* choose a random replica */
758 get_random_bytes(&r
, 1);
761 dout("choose_mds %p %llx.%llx "
762 "frag %u mds%d (%d/%d)\n",
763 inode
, ceph_vinop(inode
),
766 if (ceph_mdsmap_get_state(mdsc
->mdsmap
, mds
) >=
767 CEPH_MDS_STATE_ACTIVE
)
771 /* since this file/dir wasn't known to be
772 * replicated, then we want to look for the
773 * authoritative mds. */
776 /* choose auth mds */
778 dout("choose_mds %p %llx.%llx "
779 "frag %u mds%d (auth)\n",
780 inode
, ceph_vinop(inode
), frag
.frag
, mds
);
781 if (ceph_mdsmap_get_state(mdsc
->mdsmap
, mds
) >=
782 CEPH_MDS_STATE_ACTIVE
)
788 spin_lock(&ci
->i_ceph_lock
);
790 if (mode
== USE_AUTH_MDS
)
791 cap
= ci
->i_auth_cap
;
792 if (!cap
&& !RB_EMPTY_ROOT(&ci
->i_caps
))
793 cap
= rb_entry(rb_first(&ci
->i_caps
), struct ceph_cap
, ci_node
);
795 spin_unlock(&ci
->i_ceph_lock
);
798 mds
= cap
->session
->s_mds
;
799 dout("choose_mds %p %llx.%llx mds%d (%scap %p)\n",
800 inode
, ceph_vinop(inode
), mds
,
801 cap
== ci
->i_auth_cap
? "auth " : "", cap
);
802 spin_unlock(&ci
->i_ceph_lock
);
806 mds
= ceph_mdsmap_get_random_mds(mdsc
->mdsmap
);
807 dout("choose_mds chose random mds%d\n", mds
);
815 static struct ceph_msg
*create_session_msg(u32 op
, u64 seq
)
817 struct ceph_msg
*msg
;
818 struct ceph_mds_session_head
*h
;
820 msg
= ceph_msg_new(CEPH_MSG_CLIENT_SESSION
, sizeof(*h
), GFP_NOFS
,
823 pr_err("create_session_msg ENOMEM creating msg\n");
826 h
= msg
->front
.iov_base
;
827 h
->op
= cpu_to_le32(op
);
828 h
->seq
= cpu_to_le64(seq
);
834 * session message, specialization for CEPH_SESSION_REQUEST_OPEN
835 * to include additional client metadata fields.
837 static struct ceph_msg
*create_session_open_msg(struct ceph_mds_client
*mdsc
, u64 seq
)
839 struct ceph_msg
*msg
;
840 struct ceph_mds_session_head
*h
;
842 int metadata_bytes
= 0;
843 int metadata_key_count
= 0;
844 struct ceph_options
*opt
= mdsc
->fsc
->client
->options
;
845 struct ceph_mount_options
*fsopt
= mdsc
->fsc
->mount_options
;
848 const char* metadata
[][2] = {
849 {"hostname", utsname()->nodename
},
850 {"kernel_version", utsname()->release
},
851 {"entity_id", opt
->name
? : ""},
852 {"root", fsopt
->server_path
? : "/"},
856 /* Calculate serialized length of metadata */
857 metadata_bytes
= 4; /* map length */
858 for (i
= 0; metadata
[i
][0] != NULL
; ++i
) {
859 metadata_bytes
+= 8 + strlen(metadata
[i
][0]) +
860 strlen(metadata
[i
][1]);
861 metadata_key_count
++;
864 /* Allocate the message */
865 msg
= ceph_msg_new(CEPH_MSG_CLIENT_SESSION
, sizeof(*h
) + metadata_bytes
,
868 pr_err("create_session_msg ENOMEM creating msg\n");
871 h
= msg
->front
.iov_base
;
872 h
->op
= cpu_to_le32(CEPH_SESSION_REQUEST_OPEN
);
873 h
->seq
= cpu_to_le64(seq
);
876 * Serialize client metadata into waiting buffer space, using
877 * the format that userspace expects for map<string, string>
879 * ClientSession messages with metadata are v2
881 msg
->hdr
.version
= cpu_to_le16(2);
882 msg
->hdr
.compat_version
= cpu_to_le16(1);
884 /* The write pointer, following the session_head structure */
885 p
= msg
->front
.iov_base
+ sizeof(*h
);
887 /* Number of entries in the map */
888 ceph_encode_32(&p
, metadata_key_count
);
890 /* Two length-prefixed strings for each entry in the map */
891 for (i
= 0; metadata
[i
][0] != NULL
; ++i
) {
892 size_t const key_len
= strlen(metadata
[i
][0]);
893 size_t const val_len
= strlen(metadata
[i
][1]);
895 ceph_encode_32(&p
, key_len
);
896 memcpy(p
, metadata
[i
][0], key_len
);
898 ceph_encode_32(&p
, val_len
);
899 memcpy(p
, metadata
[i
][1], val_len
);
907 * send session open request.
909 * called under mdsc->mutex
911 static int __open_session(struct ceph_mds_client
*mdsc
,
912 struct ceph_mds_session
*session
)
914 struct ceph_msg
*msg
;
916 int mds
= session
->s_mds
;
918 /* wait for mds to go active? */
919 mstate
= ceph_mdsmap_get_state(mdsc
->mdsmap
, mds
);
920 dout("open_session to mds%d (%s)\n", mds
,
921 ceph_mds_state_name(mstate
));
922 session
->s_state
= CEPH_MDS_SESSION_OPENING
;
923 session
->s_renew_requested
= jiffies
;
925 /* send connect message */
926 msg
= create_session_open_msg(mdsc
, session
->s_seq
);
929 ceph_con_send(&session
->s_con
, msg
);
934 * open sessions for any export targets for the given mds
936 * called under mdsc->mutex
938 static struct ceph_mds_session
*
939 __open_export_target_session(struct ceph_mds_client
*mdsc
, int target
)
941 struct ceph_mds_session
*session
;
943 session
= __ceph_lookup_mds_session(mdsc
, target
);
945 session
= register_session(mdsc
, target
);
949 if (session
->s_state
== CEPH_MDS_SESSION_NEW
||
950 session
->s_state
== CEPH_MDS_SESSION_CLOSING
)
951 __open_session(mdsc
, session
);
956 struct ceph_mds_session
*
957 ceph_mdsc_open_export_target_session(struct ceph_mds_client
*mdsc
, int target
)
959 struct ceph_mds_session
*session
;
961 dout("open_export_target_session to mds%d\n", target
);
963 mutex_lock(&mdsc
->mutex
);
964 session
= __open_export_target_session(mdsc
, target
);
965 mutex_unlock(&mdsc
->mutex
);
970 static void __open_export_target_sessions(struct ceph_mds_client
*mdsc
,
971 struct ceph_mds_session
*session
)
973 struct ceph_mds_info
*mi
;
974 struct ceph_mds_session
*ts
;
975 int i
, mds
= session
->s_mds
;
977 if (mds
>= mdsc
->mdsmap
->m_max_mds
)
980 mi
= &mdsc
->mdsmap
->m_info
[mds
];
981 dout("open_export_target_sessions for mds%d (%d targets)\n",
982 session
->s_mds
, mi
->num_export_targets
);
984 for (i
= 0; i
< mi
->num_export_targets
; i
++) {
985 ts
= __open_export_target_session(mdsc
, mi
->export_targets
[i
]);
987 ceph_put_mds_session(ts
);
991 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client
*mdsc
,
992 struct ceph_mds_session
*session
)
994 mutex_lock(&mdsc
->mutex
);
995 __open_export_target_sessions(mdsc
, session
);
996 mutex_unlock(&mdsc
->mutex
);
1003 /* caller holds s_cap_lock, we drop it */
1004 static void cleanup_cap_releases(struct ceph_mds_client
*mdsc
,
1005 struct ceph_mds_session
*session
)
1006 __releases(session
->s_cap_lock
)
1008 LIST_HEAD(tmp_list
);
1009 list_splice_init(&session
->s_cap_releases
, &tmp_list
);
1010 session
->s_num_cap_releases
= 0;
1011 spin_unlock(&session
->s_cap_lock
);
1013 dout("cleanup_cap_releases mds%d\n", session
->s_mds
);
1014 while (!list_empty(&tmp_list
)) {
1015 struct ceph_cap
*cap
;
1016 /* zero out the in-progress message */
1017 cap
= list_first_entry(&tmp_list
,
1018 struct ceph_cap
, session_caps
);
1019 list_del(&cap
->session_caps
);
1020 ceph_put_cap(mdsc
, cap
);
1024 static void cleanup_session_requests(struct ceph_mds_client
*mdsc
,
1025 struct ceph_mds_session
*session
)
1027 struct ceph_mds_request
*req
;
1030 dout("cleanup_session_requests mds%d\n", session
->s_mds
);
1031 mutex_lock(&mdsc
->mutex
);
1032 while (!list_empty(&session
->s_unsafe
)) {
1033 req
= list_first_entry(&session
->s_unsafe
,
1034 struct ceph_mds_request
, r_unsafe_item
);
1035 list_del_init(&req
->r_unsafe_item
);
1036 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1038 __unregister_request(mdsc
, req
);
1040 /* zero r_attempts, so kick_requests() will re-send requests */
1041 p
= rb_first(&mdsc
->request_tree
);
1043 req
= rb_entry(p
, struct ceph_mds_request
, r_node
);
1045 if (req
->r_session
&&
1046 req
->r_session
->s_mds
== session
->s_mds
)
1047 req
->r_attempts
= 0;
1049 mutex_unlock(&mdsc
->mutex
);
1053 * Helper to safely iterate over all caps associated with a session, with
1054 * special care taken to handle a racing __ceph_remove_cap().
1056 * Caller must hold session s_mutex.
1058 static int iterate_session_caps(struct ceph_mds_session
*session
,
1059 int (*cb
)(struct inode
*, struct ceph_cap
*,
1062 struct list_head
*p
;
1063 struct ceph_cap
*cap
;
1064 struct inode
*inode
, *last_inode
= NULL
;
1065 struct ceph_cap
*old_cap
= NULL
;
1068 dout("iterate_session_caps %p mds%d\n", session
, session
->s_mds
);
1069 spin_lock(&session
->s_cap_lock
);
1070 p
= session
->s_caps
.next
;
1071 while (p
!= &session
->s_caps
) {
1072 cap
= list_entry(p
, struct ceph_cap
, session_caps
);
1073 inode
= igrab(&cap
->ci
->vfs_inode
);
1078 session
->s_cap_iterator
= cap
;
1079 spin_unlock(&session
->s_cap_lock
);
1086 ceph_put_cap(session
->s_mdsc
, old_cap
);
1090 ret
= cb(inode
, cap
, arg
);
1093 spin_lock(&session
->s_cap_lock
);
1095 if (cap
->ci
== NULL
) {
1096 dout("iterate_session_caps finishing cap %p removal\n",
1098 BUG_ON(cap
->session
!= session
);
1099 cap
->session
= NULL
;
1100 list_del_init(&cap
->session_caps
);
1101 session
->s_nr_caps
--;
1102 if (cap
->queue_release
) {
1103 list_add_tail(&cap
->session_caps
,
1104 &session
->s_cap_releases
);
1105 session
->s_num_cap_releases
++;
1107 old_cap
= cap
; /* put_cap it w/o locks held */
1115 session
->s_cap_iterator
= NULL
;
1116 spin_unlock(&session
->s_cap_lock
);
1120 ceph_put_cap(session
->s_mdsc
, old_cap
);
1125 static int remove_session_caps_cb(struct inode
*inode
, struct ceph_cap
*cap
,
1128 struct ceph_fs_client
*fsc
= (struct ceph_fs_client
*)arg
;
1129 struct ceph_inode_info
*ci
= ceph_inode(inode
);
1130 LIST_HEAD(to_remove
);
1132 bool invalidate
= false;
1134 dout("removing cap %p, ci is %p, inode is %p\n",
1135 cap
, ci
, &ci
->vfs_inode
);
1136 spin_lock(&ci
->i_ceph_lock
);
1137 __ceph_remove_cap(cap
, false);
1138 if (!ci
->i_auth_cap
) {
1139 struct ceph_cap_flush
*cf
;
1140 struct ceph_mds_client
*mdsc
= fsc
->mdsc
;
1142 ci
->i_ceph_flags
|= CEPH_I_CAP_DROPPED
;
1144 if (ci
->i_wrbuffer_ref
> 0 &&
1145 ACCESS_ONCE(fsc
->mount_state
) == CEPH_MOUNT_SHUTDOWN
)
1149 struct rb_node
*n
= rb_first(&ci
->i_cap_flush_tree
);
1152 cf
= rb_entry(n
, struct ceph_cap_flush
, i_node
);
1153 rb_erase(&cf
->i_node
, &ci
->i_cap_flush_tree
);
1154 list_add(&cf
->list
, &to_remove
);
1157 spin_lock(&mdsc
->cap_dirty_lock
);
1159 list_for_each_entry(cf
, &to_remove
, list
)
1160 rb_erase(&cf
->g_node
, &mdsc
->cap_flush_tree
);
1162 if (!list_empty(&ci
->i_dirty_item
)) {
1163 pr_warn_ratelimited(
1164 " dropping dirty %s state for %p %lld\n",
1165 ceph_cap_string(ci
->i_dirty_caps
),
1166 inode
, ceph_ino(inode
));
1167 ci
->i_dirty_caps
= 0;
1168 list_del_init(&ci
->i_dirty_item
);
1171 if (!list_empty(&ci
->i_flushing_item
)) {
1172 pr_warn_ratelimited(
1173 " dropping dirty+flushing %s state for %p %lld\n",
1174 ceph_cap_string(ci
->i_flushing_caps
),
1175 inode
, ceph_ino(inode
));
1176 ci
->i_flushing_caps
= 0;
1177 list_del_init(&ci
->i_flushing_item
);
1178 mdsc
->num_cap_flushing
--;
1181 spin_unlock(&mdsc
->cap_dirty_lock
);
1183 if (!ci
->i_dirty_caps
&& ci
->i_prealloc_cap_flush
) {
1184 list_add(&ci
->i_prealloc_cap_flush
->list
, &to_remove
);
1185 ci
->i_prealloc_cap_flush
= NULL
;
1188 spin_unlock(&ci
->i_ceph_lock
);
1189 while (!list_empty(&to_remove
)) {
1190 struct ceph_cap_flush
*cf
;
1191 cf
= list_first_entry(&to_remove
,
1192 struct ceph_cap_flush
, list
);
1193 list_del(&cf
->list
);
1194 ceph_free_cap_flush(cf
);
1197 wake_up_all(&ci
->i_cap_wq
);
1199 ceph_queue_invalidate(inode
);
1206 * caller must hold session s_mutex
1208 static void remove_session_caps(struct ceph_mds_session
*session
)
1210 struct ceph_fs_client
*fsc
= session
->s_mdsc
->fsc
;
1211 struct super_block
*sb
= fsc
->sb
;
1212 dout("remove_session_caps on %p\n", session
);
1213 iterate_session_caps(session
, remove_session_caps_cb
, fsc
);
1215 spin_lock(&session
->s_cap_lock
);
1216 if (session
->s_nr_caps
> 0) {
1217 struct inode
*inode
;
1218 struct ceph_cap
*cap
, *prev
= NULL
;
1219 struct ceph_vino vino
;
1221 * iterate_session_caps() skips inodes that are being
1222 * deleted, we need to wait until deletions are complete.
1223 * __wait_on_freeing_inode() is designed for the job,
1224 * but it is not exported, so use lookup inode function
1227 while (!list_empty(&session
->s_caps
)) {
1228 cap
= list_entry(session
->s_caps
.next
,
1229 struct ceph_cap
, session_caps
);
1233 vino
= cap
->ci
->i_vino
;
1234 spin_unlock(&session
->s_cap_lock
);
1236 inode
= ceph_find_inode(sb
, vino
);
1239 spin_lock(&session
->s_cap_lock
);
1243 // drop cap expires and unlock s_cap_lock
1244 cleanup_cap_releases(session
->s_mdsc
, session
);
1246 BUG_ON(session
->s_nr_caps
> 0);
1247 BUG_ON(!list_empty(&session
->s_cap_flushing
));
1251 * wake up any threads waiting on this session's caps. if the cap is
1252 * old (didn't get renewed on the client reconnect), remove it now.
1254 * caller must hold s_mutex.
1256 static int wake_up_session_cb(struct inode
*inode
, struct ceph_cap
*cap
,
1259 struct ceph_inode_info
*ci
= ceph_inode(inode
);
1262 spin_lock(&ci
->i_ceph_lock
);
1263 ci
->i_wanted_max_size
= 0;
1264 ci
->i_requested_max_size
= 0;
1265 spin_unlock(&ci
->i_ceph_lock
);
1267 wake_up_all(&ci
->i_cap_wq
);
1271 static void wake_up_session_caps(struct ceph_mds_session
*session
,
1274 dout("wake_up_session_caps %p mds%d\n", session
, session
->s_mds
);
1275 iterate_session_caps(session
, wake_up_session_cb
,
1276 (void *)(unsigned long)reconnect
);
1280 * Send periodic message to MDS renewing all currently held caps. The
1281 * ack will reset the expiration for all caps from this session.
1283 * caller holds s_mutex
1285 static int send_renew_caps(struct ceph_mds_client
*mdsc
,
1286 struct ceph_mds_session
*session
)
1288 struct ceph_msg
*msg
;
1291 if (time_after_eq(jiffies
, session
->s_cap_ttl
) &&
1292 time_after_eq(session
->s_cap_ttl
, session
->s_renew_requested
))
1293 pr_info("mds%d caps stale\n", session
->s_mds
);
1294 session
->s_renew_requested
= jiffies
;
1296 /* do not try to renew caps until a recovering mds has reconnected
1297 * with its clients. */
1298 state
= ceph_mdsmap_get_state(mdsc
->mdsmap
, session
->s_mds
);
1299 if (state
< CEPH_MDS_STATE_RECONNECT
) {
1300 dout("send_renew_caps ignoring mds%d (%s)\n",
1301 session
->s_mds
, ceph_mds_state_name(state
));
1305 dout("send_renew_caps to mds%d (%s)\n", session
->s_mds
,
1306 ceph_mds_state_name(state
));
1307 msg
= create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS
,
1308 ++session
->s_renew_seq
);
1311 ceph_con_send(&session
->s_con
, msg
);
1315 static int send_flushmsg_ack(struct ceph_mds_client
*mdsc
,
1316 struct ceph_mds_session
*session
, u64 seq
)
1318 struct ceph_msg
*msg
;
1320 dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1321 session
->s_mds
, ceph_session_state_name(session
->s_state
), seq
);
1322 msg
= create_session_msg(CEPH_SESSION_FLUSHMSG_ACK
, seq
);
1325 ceph_con_send(&session
->s_con
, msg
);
1331 * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1333 * Called under session->s_mutex
1335 static void renewed_caps(struct ceph_mds_client
*mdsc
,
1336 struct ceph_mds_session
*session
, int is_renew
)
1341 spin_lock(&session
->s_cap_lock
);
1342 was_stale
= is_renew
&& time_after_eq(jiffies
, session
->s_cap_ttl
);
1344 session
->s_cap_ttl
= session
->s_renew_requested
+
1345 mdsc
->mdsmap
->m_session_timeout
*HZ
;
1348 if (time_before(jiffies
, session
->s_cap_ttl
)) {
1349 pr_info("mds%d caps renewed\n", session
->s_mds
);
1352 pr_info("mds%d caps still stale\n", session
->s_mds
);
1355 dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1356 session
->s_mds
, session
->s_cap_ttl
, was_stale
? "stale" : "fresh",
1357 time_before(jiffies
, session
->s_cap_ttl
) ? "stale" : "fresh");
1358 spin_unlock(&session
->s_cap_lock
);
1361 wake_up_session_caps(session
, 0);
1365 * send a session close request
1367 static int request_close_session(struct ceph_mds_client
*mdsc
,
1368 struct ceph_mds_session
*session
)
1370 struct ceph_msg
*msg
;
1372 dout("request_close_session mds%d state %s seq %lld\n",
1373 session
->s_mds
, ceph_session_state_name(session
->s_state
),
1375 msg
= create_session_msg(CEPH_SESSION_REQUEST_CLOSE
, session
->s_seq
);
1378 ceph_con_send(&session
->s_con
, msg
);
1383 * Called with s_mutex held.
1385 static int __close_session(struct ceph_mds_client
*mdsc
,
1386 struct ceph_mds_session
*session
)
1388 if (session
->s_state
>= CEPH_MDS_SESSION_CLOSING
)
1390 session
->s_state
= CEPH_MDS_SESSION_CLOSING
;
1391 return request_close_session(mdsc
, session
);
1395 * Trim old(er) caps.
1397 * Because we can't cache an inode without one or more caps, we do
1398 * this indirectly: if a cap is unused, we prune its aliases, at which
1399 * point the inode will hopefully get dropped to.
1401 * Yes, this is a bit sloppy. Our only real goal here is to respond to
1402 * memory pressure from the MDS, though, so it needn't be perfect.
1404 static int trim_caps_cb(struct inode
*inode
, struct ceph_cap
*cap
, void *arg
)
1406 struct ceph_mds_session
*session
= arg
;
1407 struct ceph_inode_info
*ci
= ceph_inode(inode
);
1408 int used
, wanted
, oissued
, mine
;
1410 if (session
->s_trim_caps
<= 0)
1413 spin_lock(&ci
->i_ceph_lock
);
1414 mine
= cap
->issued
| cap
->implemented
;
1415 used
= __ceph_caps_used(ci
);
1416 wanted
= __ceph_caps_file_wanted(ci
);
1417 oissued
= __ceph_caps_issued_other(ci
, cap
);
1419 dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1420 inode
, cap
, ceph_cap_string(mine
), ceph_cap_string(oissued
),
1421 ceph_cap_string(used
), ceph_cap_string(wanted
));
1422 if (cap
== ci
->i_auth_cap
) {
1423 if (ci
->i_dirty_caps
|| ci
->i_flushing_caps
||
1424 !list_empty(&ci
->i_cap_snaps
))
1426 if ((used
| wanted
) & CEPH_CAP_ANY_WR
)
1429 /* The inode has cached pages, but it's no longer used.
1430 * we can safely drop it */
1431 if (wanted
== 0 && used
== CEPH_CAP_FILE_CACHE
&&
1432 !(oissued
& CEPH_CAP_FILE_CACHE
)) {
1436 if ((used
| wanted
) & ~oissued
& mine
)
1437 goto out
; /* we need these caps */
1439 session
->s_trim_caps
--;
1441 /* we aren't the only cap.. just remove us */
1442 __ceph_remove_cap(cap
, true);
1444 /* try dropping referring dentries */
1445 spin_unlock(&ci
->i_ceph_lock
);
1446 d_prune_aliases(inode
);
1447 dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1448 inode
, cap
, atomic_read(&inode
->i_count
));
1453 spin_unlock(&ci
->i_ceph_lock
);
1458 * Trim session cap count down to some max number.
1460 static int trim_caps(struct ceph_mds_client
*mdsc
,
1461 struct ceph_mds_session
*session
,
1464 int trim_caps
= session
->s_nr_caps
- max_caps
;
1466 dout("trim_caps mds%d start: %d / %d, trim %d\n",
1467 session
->s_mds
, session
->s_nr_caps
, max_caps
, trim_caps
);
1468 if (trim_caps
> 0) {
1469 session
->s_trim_caps
= trim_caps
;
1470 iterate_session_caps(session
, trim_caps_cb
, session
);
1471 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
1472 session
->s_mds
, session
->s_nr_caps
, max_caps
,
1473 trim_caps
- session
->s_trim_caps
);
1474 session
->s_trim_caps
= 0;
1477 ceph_send_cap_releases(mdsc
, session
);
1481 static int check_capsnap_flush(struct ceph_inode_info
*ci
,
1485 spin_lock(&ci
->i_ceph_lock
);
1486 if (want_snap_seq
> 0 && !list_empty(&ci
->i_cap_snaps
)) {
1487 struct ceph_cap_snap
*capsnap
=
1488 list_first_entry(&ci
->i_cap_snaps
,
1489 struct ceph_cap_snap
, ci_item
);
1490 ret
= capsnap
->follows
>= want_snap_seq
;
1492 spin_unlock(&ci
->i_ceph_lock
);
1496 static int check_caps_flush(struct ceph_mds_client
*mdsc
,
1500 struct ceph_cap_flush
*cf
;
1503 spin_lock(&mdsc
->cap_dirty_lock
);
1504 n
= rb_first(&mdsc
->cap_flush_tree
);
1505 cf
= n
? rb_entry(n
, struct ceph_cap_flush
, g_node
) : NULL
;
1506 if (cf
&& cf
->tid
<= want_flush_tid
) {
1507 dout("check_caps_flush still flushing tid %llu <= %llu\n",
1508 cf
->tid
, want_flush_tid
);
1511 spin_unlock(&mdsc
->cap_dirty_lock
);
1516 * flush all dirty inode data to disk.
1518 * returns true if we've flushed through want_flush_tid
1520 static void wait_caps_flush(struct ceph_mds_client
*mdsc
,
1521 u64 want_flush_tid
, u64 want_snap_seq
)
1525 dout("check_caps_flush want %llu snap want %llu\n",
1526 want_flush_tid
, want_snap_seq
);
1527 mutex_lock(&mdsc
->mutex
);
1528 for (mds
= 0; mds
< mdsc
->max_sessions
; ) {
1529 struct ceph_mds_session
*session
= mdsc
->sessions
[mds
];
1530 struct inode
*inode
= NULL
;
1536 get_session(session
);
1537 mutex_unlock(&mdsc
->mutex
);
1539 mutex_lock(&session
->s_mutex
);
1540 if (!list_empty(&session
->s_cap_snaps_flushing
)) {
1541 struct ceph_cap_snap
*capsnap
=
1542 list_first_entry(&session
->s_cap_snaps_flushing
,
1543 struct ceph_cap_snap
,
1545 struct ceph_inode_info
*ci
= capsnap
->ci
;
1546 if (!check_capsnap_flush(ci
, want_snap_seq
)) {
1547 dout("check_cap_flush still flushing snap %p "
1548 "follows %lld <= %lld to mds%d\n",
1549 &ci
->vfs_inode
, capsnap
->follows
,
1550 want_snap_seq
, mds
);
1551 inode
= igrab(&ci
->vfs_inode
);
1554 mutex_unlock(&session
->s_mutex
);
1555 ceph_put_mds_session(session
);
1558 wait_event(mdsc
->cap_flushing_wq
,
1559 check_capsnap_flush(ceph_inode(inode
),
1566 mutex_lock(&mdsc
->mutex
);
1568 mutex_unlock(&mdsc
->mutex
);
1570 wait_event(mdsc
->cap_flushing_wq
,
1571 check_caps_flush(mdsc
, want_flush_tid
));
1573 dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid
);
1577 * called under s_mutex
1579 void ceph_send_cap_releases(struct ceph_mds_client
*mdsc
,
1580 struct ceph_mds_session
*session
)
1582 struct ceph_msg
*msg
= NULL
;
1583 struct ceph_mds_cap_release
*head
;
1584 struct ceph_mds_cap_item
*item
;
1585 struct ceph_cap
*cap
;
1586 LIST_HEAD(tmp_list
);
1587 int num_cap_releases
;
1589 spin_lock(&session
->s_cap_lock
);
1591 list_splice_init(&session
->s_cap_releases
, &tmp_list
);
1592 num_cap_releases
= session
->s_num_cap_releases
;
1593 session
->s_num_cap_releases
= 0;
1594 spin_unlock(&session
->s_cap_lock
);
1596 while (!list_empty(&tmp_list
)) {
1598 msg
= ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE
,
1599 PAGE_SIZE
, GFP_NOFS
, false);
1602 head
= msg
->front
.iov_base
;
1603 head
->num
= cpu_to_le32(0);
1604 msg
->front
.iov_len
= sizeof(*head
);
1606 cap
= list_first_entry(&tmp_list
, struct ceph_cap
,
1608 list_del(&cap
->session_caps
);
1611 head
= msg
->front
.iov_base
;
1612 le32_add_cpu(&head
->num
, 1);
1613 item
= msg
->front
.iov_base
+ msg
->front
.iov_len
;
1614 item
->ino
= cpu_to_le64(cap
->cap_ino
);
1615 item
->cap_id
= cpu_to_le64(cap
->cap_id
);
1616 item
->migrate_seq
= cpu_to_le32(cap
->mseq
);
1617 item
->seq
= cpu_to_le32(cap
->issue_seq
);
1618 msg
->front
.iov_len
+= sizeof(*item
);
1620 ceph_put_cap(mdsc
, cap
);
1622 if (le32_to_cpu(head
->num
) == CEPH_CAPS_PER_RELEASE
) {
1623 msg
->hdr
.front_len
= cpu_to_le32(msg
->front
.iov_len
);
1624 dout("send_cap_releases mds%d %p\n", session
->s_mds
, msg
);
1625 ceph_con_send(&session
->s_con
, msg
);
1630 BUG_ON(num_cap_releases
!= 0);
1632 spin_lock(&session
->s_cap_lock
);
1633 if (!list_empty(&session
->s_cap_releases
))
1635 spin_unlock(&session
->s_cap_lock
);
1638 msg
->hdr
.front_len
= cpu_to_le32(msg
->front
.iov_len
);
1639 dout("send_cap_releases mds%d %p\n", session
->s_mds
, msg
);
1640 ceph_con_send(&session
->s_con
, msg
);
1644 pr_err("send_cap_releases mds%d, failed to allocate message\n",
1646 spin_lock(&session
->s_cap_lock
);
1647 list_splice(&tmp_list
, &session
->s_cap_releases
);
1648 session
->s_num_cap_releases
+= num_cap_releases
;
1649 spin_unlock(&session
->s_cap_lock
);
1656 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request
*req
,
1659 struct ceph_inode_info
*ci
= ceph_inode(dir
);
1660 struct ceph_mds_reply_info_parsed
*rinfo
= &req
->r_reply_info
;
1661 struct ceph_mount_options
*opt
= req
->r_mdsc
->fsc
->mount_options
;
1662 size_t size
= sizeof(struct ceph_mds_reply_dir_entry
);
1663 int order
, num_entries
;
1665 spin_lock(&ci
->i_ceph_lock
);
1666 num_entries
= ci
->i_files
+ ci
->i_subdirs
;
1667 spin_unlock(&ci
->i_ceph_lock
);
1668 num_entries
= max(num_entries
, 1);
1669 num_entries
= min(num_entries
, opt
->max_readdir
);
1671 order
= get_order(size
* num_entries
);
1672 while (order
>= 0) {
1673 rinfo
->dir_entries
= (void*)__get_free_pages(GFP_KERNEL
|
1676 if (rinfo
->dir_entries
)
1680 if (!rinfo
->dir_entries
)
1683 num_entries
= (PAGE_SIZE
<< order
) / size
;
1684 num_entries
= min(num_entries
, opt
->max_readdir
);
1686 rinfo
->dir_buf_size
= PAGE_SIZE
<< order
;
1687 req
->r_num_caps
= num_entries
+ 1;
1688 req
->r_args
.readdir
.max_entries
= cpu_to_le32(num_entries
);
1689 req
->r_args
.readdir
.max_bytes
= cpu_to_le32(opt
->max_readdir_bytes
);
1694 * Create an mds request.
1696 struct ceph_mds_request
*
1697 ceph_mdsc_create_request(struct ceph_mds_client
*mdsc
, int op
, int mode
)
1699 struct ceph_mds_request
*req
= kzalloc(sizeof(*req
), GFP_NOFS
);
1702 return ERR_PTR(-ENOMEM
);
1704 mutex_init(&req
->r_fill_mutex
);
1706 req
->r_started
= jiffies
;
1707 req
->r_resend_mds
= -1;
1708 INIT_LIST_HEAD(&req
->r_unsafe_dir_item
);
1709 INIT_LIST_HEAD(&req
->r_unsafe_target_item
);
1711 kref_init(&req
->r_kref
);
1712 RB_CLEAR_NODE(&req
->r_node
);
1713 INIT_LIST_HEAD(&req
->r_wait
);
1714 init_completion(&req
->r_completion
);
1715 init_completion(&req
->r_safe_completion
);
1716 INIT_LIST_HEAD(&req
->r_unsafe_item
);
1718 req
->r_stamp
= current_fs_time(mdsc
->fsc
->sb
);
1721 req
->r_direct_mode
= mode
;
1726 * return oldest (lowest) request, tid in request tree, 0 if none.
1728 * called under mdsc->mutex.
1730 static struct ceph_mds_request
*__get_oldest_req(struct ceph_mds_client
*mdsc
)
1732 if (RB_EMPTY_ROOT(&mdsc
->request_tree
))
1734 return rb_entry(rb_first(&mdsc
->request_tree
),
1735 struct ceph_mds_request
, r_node
);
1738 static inline u64
__get_oldest_tid(struct ceph_mds_client
*mdsc
)
1740 return mdsc
->oldest_tid
;
1744 * Build a dentry's path. Allocate on heap; caller must kfree. Based
1745 * on build_path_from_dentry in fs/cifs/dir.c.
1747 * If @stop_on_nosnap, generate path relative to the first non-snapped
1750 * Encode hidden .snap dirs as a double /, i.e.
1751 * foo/.snap/bar -> foo//bar
1753 char *ceph_mdsc_build_path(struct dentry
*dentry
, int *plen
, u64
*base
,
1756 struct dentry
*temp
;
1762 return ERR_PTR(-EINVAL
);
1766 seq
= read_seqbegin(&rename_lock
);
1768 for (temp
= dentry
; !IS_ROOT(temp
);) {
1769 struct inode
*inode
= d_inode(temp
);
1770 if (inode
&& ceph_snap(inode
) == CEPH_SNAPDIR
)
1771 len
++; /* slash only */
1772 else if (stop_on_nosnap
&& inode
&&
1773 ceph_snap(inode
) == CEPH_NOSNAP
)
1776 len
+= 1 + temp
->d_name
.len
;
1777 temp
= temp
->d_parent
;
1781 len
--; /* no leading '/' */
1783 path
= kmalloc(len
+1, GFP_NOFS
);
1785 return ERR_PTR(-ENOMEM
);
1787 path
[pos
] = 0; /* trailing null */
1789 for (temp
= dentry
; !IS_ROOT(temp
) && pos
!= 0; ) {
1790 struct inode
*inode
;
1792 spin_lock(&temp
->d_lock
);
1793 inode
= d_inode(temp
);
1794 if (inode
&& ceph_snap(inode
) == CEPH_SNAPDIR
) {
1795 dout("build_path path+%d: %p SNAPDIR\n",
1797 } else if (stop_on_nosnap
&& inode
&&
1798 ceph_snap(inode
) == CEPH_NOSNAP
) {
1799 spin_unlock(&temp
->d_lock
);
1802 pos
-= temp
->d_name
.len
;
1804 spin_unlock(&temp
->d_lock
);
1807 strncpy(path
+ pos
, temp
->d_name
.name
,
1810 spin_unlock(&temp
->d_lock
);
1813 temp
= temp
->d_parent
;
1816 if (pos
!= 0 || read_seqretry(&rename_lock
, seq
)) {
1817 pr_err("build_path did not end path lookup where "
1818 "expected, namelen is %d, pos is %d\n", len
, pos
);
1819 /* presumably this is only possible if racing with a
1820 rename of one of the parent directories (we can not
1821 lock the dentries above us to prevent this, but
1822 retrying should be harmless) */
1827 *base
= ceph_ino(d_inode(temp
));
1829 dout("build_path on %p %d built %llx '%.*s'\n",
1830 dentry
, d_count(dentry
), *base
, len
, path
);
1834 static int build_dentry_path(struct dentry
*dentry
,
1835 const char **ppath
, int *ppathlen
, u64
*pino
,
1840 if (ceph_snap(d_inode(dentry
->d_parent
)) == CEPH_NOSNAP
) {
1841 *pino
= ceph_ino(d_inode(dentry
->d_parent
));
1842 *ppath
= dentry
->d_name
.name
;
1843 *ppathlen
= dentry
->d_name
.len
;
1846 path
= ceph_mdsc_build_path(dentry
, ppathlen
, pino
, 1);
1848 return PTR_ERR(path
);
1854 static int build_inode_path(struct inode
*inode
,
1855 const char **ppath
, int *ppathlen
, u64
*pino
,
1858 struct dentry
*dentry
;
1861 if (ceph_snap(inode
) == CEPH_NOSNAP
) {
1862 *pino
= ceph_ino(inode
);
1866 dentry
= d_find_alias(inode
);
1867 path
= ceph_mdsc_build_path(dentry
, ppathlen
, pino
, 1);
1870 return PTR_ERR(path
);
1877 * request arguments may be specified via an inode *, a dentry *, or
1878 * an explicit ino+path.
1880 static int set_request_path_attr(struct inode
*rinode
, struct dentry
*rdentry
,
1881 const char *rpath
, u64 rino
,
1882 const char **ppath
, int *pathlen
,
1883 u64
*ino
, int *freepath
)
1888 r
= build_inode_path(rinode
, ppath
, pathlen
, ino
, freepath
);
1889 dout(" inode %p %llx.%llx\n", rinode
, ceph_ino(rinode
),
1891 } else if (rdentry
) {
1892 r
= build_dentry_path(rdentry
, ppath
, pathlen
, ino
, freepath
);
1893 dout(" dentry %p %llx/%.*s\n", rdentry
, *ino
, *pathlen
,
1895 } else if (rpath
|| rino
) {
1898 *pathlen
= rpath
? strlen(rpath
) : 0;
1899 dout(" path %.*s\n", *pathlen
, rpath
);
1906 * called under mdsc->mutex
1908 static struct ceph_msg
*create_request_message(struct ceph_mds_client
*mdsc
,
1909 struct ceph_mds_request
*req
,
1910 int mds
, bool drop_cap_releases
)
1912 struct ceph_msg
*msg
;
1913 struct ceph_mds_request_head
*head
;
1914 const char *path1
= NULL
;
1915 const char *path2
= NULL
;
1916 u64 ino1
= 0, ino2
= 0;
1917 int pathlen1
= 0, pathlen2
= 0;
1918 int freepath1
= 0, freepath2
= 0;
1924 ret
= set_request_path_attr(req
->r_inode
, req
->r_dentry
,
1925 req
->r_path1
, req
->r_ino1
.ino
,
1926 &path1
, &pathlen1
, &ino1
, &freepath1
);
1932 ret
= set_request_path_attr(NULL
, req
->r_old_dentry
,
1933 req
->r_path2
, req
->r_ino2
.ino
,
1934 &path2
, &pathlen2
, &ino2
, &freepath2
);
1940 len
= sizeof(*head
) +
1941 pathlen1
+ pathlen2
+ 2*(1 + sizeof(u32
) + sizeof(u64
)) +
1942 sizeof(struct ceph_timespec
);
1944 /* calculate (max) length for cap releases */
1945 len
+= sizeof(struct ceph_mds_request_release
) *
1946 (!!req
->r_inode_drop
+ !!req
->r_dentry_drop
+
1947 !!req
->r_old_inode_drop
+ !!req
->r_old_dentry_drop
);
1948 if (req
->r_dentry_drop
)
1949 len
+= req
->r_dentry
->d_name
.len
;
1950 if (req
->r_old_dentry_drop
)
1951 len
+= req
->r_old_dentry
->d_name
.len
;
1953 msg
= ceph_msg_new(CEPH_MSG_CLIENT_REQUEST
, len
, GFP_NOFS
, false);
1955 msg
= ERR_PTR(-ENOMEM
);
1959 msg
->hdr
.version
= cpu_to_le16(2);
1960 msg
->hdr
.tid
= cpu_to_le64(req
->r_tid
);
1962 head
= msg
->front
.iov_base
;
1963 p
= msg
->front
.iov_base
+ sizeof(*head
);
1964 end
= msg
->front
.iov_base
+ msg
->front
.iov_len
;
1966 head
->mdsmap_epoch
= cpu_to_le32(mdsc
->mdsmap
->m_epoch
);
1967 head
->op
= cpu_to_le32(req
->r_op
);
1968 head
->caller_uid
= cpu_to_le32(from_kuid(&init_user_ns
, req
->r_uid
));
1969 head
->caller_gid
= cpu_to_le32(from_kgid(&init_user_ns
, req
->r_gid
));
1970 head
->args
= req
->r_args
;
1972 ceph_encode_filepath(&p
, end
, ino1
, path1
);
1973 ceph_encode_filepath(&p
, end
, ino2
, path2
);
1975 /* make note of release offset, in case we need to replay */
1976 req
->r_request_release_offset
= p
- msg
->front
.iov_base
;
1980 if (req
->r_inode_drop
)
1981 releases
+= ceph_encode_inode_release(&p
,
1982 req
->r_inode
? req
->r_inode
: d_inode(req
->r_dentry
),
1983 mds
, req
->r_inode_drop
, req
->r_inode_unless
, 0);
1984 if (req
->r_dentry_drop
)
1985 releases
+= ceph_encode_dentry_release(&p
, req
->r_dentry
,
1986 mds
, req
->r_dentry_drop
, req
->r_dentry_unless
);
1987 if (req
->r_old_dentry_drop
)
1988 releases
+= ceph_encode_dentry_release(&p
, req
->r_old_dentry
,
1989 mds
, req
->r_old_dentry_drop
, req
->r_old_dentry_unless
);
1990 if (req
->r_old_inode_drop
)
1991 releases
+= ceph_encode_inode_release(&p
,
1992 d_inode(req
->r_old_dentry
),
1993 mds
, req
->r_old_inode_drop
, req
->r_old_inode_unless
, 0);
1995 if (drop_cap_releases
) {
1997 p
= msg
->front
.iov_base
+ req
->r_request_release_offset
;
2000 head
->num_releases
= cpu_to_le16(releases
);
2004 struct ceph_timespec ts
;
2005 ceph_encode_timespec(&ts
, &req
->r_stamp
);
2006 ceph_encode_copy(&p
, &ts
, sizeof(ts
));
2010 msg
->front
.iov_len
= p
- msg
->front
.iov_base
;
2011 msg
->hdr
.front_len
= cpu_to_le32(msg
->front
.iov_len
);
2013 if (req
->r_pagelist
) {
2014 struct ceph_pagelist
*pagelist
= req
->r_pagelist
;
2015 atomic_inc(&pagelist
->refcnt
);
2016 ceph_msg_data_add_pagelist(msg
, pagelist
);
2017 msg
->hdr
.data_len
= cpu_to_le32(pagelist
->length
);
2019 msg
->hdr
.data_len
= 0;
2022 msg
->hdr
.data_off
= cpu_to_le16(0);
2026 kfree((char *)path2
);
2029 kfree((char *)path1
);
2035 * called under mdsc->mutex if error, under no mutex if
2038 static void complete_request(struct ceph_mds_client
*mdsc
,
2039 struct ceph_mds_request
*req
)
2041 if (req
->r_callback
)
2042 req
->r_callback(mdsc
, req
);
2044 complete_all(&req
->r_completion
);
2048 * called under mdsc->mutex
2050 static int __prepare_send_request(struct ceph_mds_client
*mdsc
,
2051 struct ceph_mds_request
*req
,
2052 int mds
, bool drop_cap_releases
)
2054 struct ceph_mds_request_head
*rhead
;
2055 struct ceph_msg
*msg
;
2060 struct ceph_cap
*cap
=
2061 ceph_get_cap_for_mds(ceph_inode(req
->r_inode
), mds
);
2064 req
->r_sent_on_mseq
= cap
->mseq
;
2066 req
->r_sent_on_mseq
= -1;
2068 dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req
,
2069 req
->r_tid
, ceph_mds_op_name(req
->r_op
), req
->r_attempts
);
2071 if (req
->r_got_unsafe
) {
2074 * Replay. Do not regenerate message (and rebuild
2075 * paths, etc.); just use the original message.
2076 * Rebuilding paths will break for renames because
2077 * d_move mangles the src name.
2079 msg
= req
->r_request
;
2080 rhead
= msg
->front
.iov_base
;
2082 flags
= le32_to_cpu(rhead
->flags
);
2083 flags
|= CEPH_MDS_FLAG_REPLAY
;
2084 rhead
->flags
= cpu_to_le32(flags
);
2086 if (req
->r_target_inode
)
2087 rhead
->ino
= cpu_to_le64(ceph_ino(req
->r_target_inode
));
2089 rhead
->num_retry
= req
->r_attempts
- 1;
2091 /* remove cap/dentry releases from message */
2092 rhead
->num_releases
= 0;
2095 p
= msg
->front
.iov_base
+ req
->r_request_release_offset
;
2097 struct ceph_timespec ts
;
2098 ceph_encode_timespec(&ts
, &req
->r_stamp
);
2099 ceph_encode_copy(&p
, &ts
, sizeof(ts
));
2102 msg
->front
.iov_len
= p
- msg
->front
.iov_base
;
2103 msg
->hdr
.front_len
= cpu_to_le32(msg
->front
.iov_len
);
2107 if (req
->r_request
) {
2108 ceph_msg_put(req
->r_request
);
2109 req
->r_request
= NULL
;
2111 msg
= create_request_message(mdsc
, req
, mds
, drop_cap_releases
);
2113 req
->r_err
= PTR_ERR(msg
);
2114 return PTR_ERR(msg
);
2116 req
->r_request
= msg
;
2118 rhead
= msg
->front
.iov_base
;
2119 rhead
->oldest_client_tid
= cpu_to_le64(__get_oldest_tid(mdsc
));
2120 if (req
->r_got_unsafe
)
2121 flags
|= CEPH_MDS_FLAG_REPLAY
;
2122 if (req
->r_locked_dir
)
2123 flags
|= CEPH_MDS_FLAG_WANT_DENTRY
;
2124 rhead
->flags
= cpu_to_le32(flags
);
2125 rhead
->num_fwd
= req
->r_num_fwd
;
2126 rhead
->num_retry
= req
->r_attempts
- 1;
2129 dout(" r_locked_dir = %p\n", req
->r_locked_dir
);
2134 * send request, or put it on the appropriate wait list.
2136 static int __do_request(struct ceph_mds_client
*mdsc
,
2137 struct ceph_mds_request
*req
)
2139 struct ceph_mds_session
*session
= NULL
;
2143 if (req
->r_err
|| req
->r_got_result
) {
2145 __unregister_request(mdsc
, req
);
2149 if (req
->r_timeout
&&
2150 time_after_eq(jiffies
, req
->r_started
+ req
->r_timeout
)) {
2151 dout("do_request timed out\n");
2155 if (ACCESS_ONCE(mdsc
->fsc
->mount_state
) == CEPH_MOUNT_SHUTDOWN
) {
2156 dout("do_request forced umount\n");
2161 put_request_session(req
);
2163 mds
= __choose_mds(mdsc
, req
);
2165 ceph_mdsmap_get_state(mdsc
->mdsmap
, mds
) < CEPH_MDS_STATE_ACTIVE
) {
2166 dout("do_request no mds or not active, waiting for map\n");
2167 list_add(&req
->r_wait
, &mdsc
->waiting_for_map
);
2171 /* get, open session */
2172 session
= __ceph_lookup_mds_session(mdsc
, mds
);
2174 session
= register_session(mdsc
, mds
);
2175 if (IS_ERR(session
)) {
2176 err
= PTR_ERR(session
);
2180 req
->r_session
= get_session(session
);
2182 dout("do_request mds%d session %p state %s\n", mds
, session
,
2183 ceph_session_state_name(session
->s_state
));
2184 if (session
->s_state
!= CEPH_MDS_SESSION_OPEN
&&
2185 session
->s_state
!= CEPH_MDS_SESSION_HUNG
) {
2186 if (session
->s_state
== CEPH_MDS_SESSION_NEW
||
2187 session
->s_state
== CEPH_MDS_SESSION_CLOSING
)
2188 __open_session(mdsc
, session
);
2189 list_add(&req
->r_wait
, &session
->s_waiting
);
2194 req
->r_resend_mds
= -1; /* forget any previous mds hint */
2196 if (req
->r_request_started
== 0) /* note request start time */
2197 req
->r_request_started
= jiffies
;
2199 err
= __prepare_send_request(mdsc
, req
, mds
, false);
2201 ceph_msg_get(req
->r_request
);
2202 ceph_con_send(&session
->s_con
, req
->r_request
);
2206 ceph_put_mds_session(session
);
2209 dout("__do_request early error %d\n", err
);
2211 complete_request(mdsc
, req
);
2212 __unregister_request(mdsc
, req
);
2219 * called under mdsc->mutex
2221 static void __wake_requests(struct ceph_mds_client
*mdsc
,
2222 struct list_head
*head
)
2224 struct ceph_mds_request
*req
;
2225 LIST_HEAD(tmp_list
);
2227 list_splice_init(head
, &tmp_list
);
2229 while (!list_empty(&tmp_list
)) {
2230 req
= list_entry(tmp_list
.next
,
2231 struct ceph_mds_request
, r_wait
);
2232 list_del_init(&req
->r_wait
);
2233 dout(" wake request %p tid %llu\n", req
, req
->r_tid
);
2234 __do_request(mdsc
, req
);
2239 * Wake up threads with requests pending for @mds, so that they can
2240 * resubmit their requests to a possibly different mds.
2242 static void kick_requests(struct ceph_mds_client
*mdsc
, int mds
)
2244 struct ceph_mds_request
*req
;
2245 struct rb_node
*p
= rb_first(&mdsc
->request_tree
);
2247 dout("kick_requests mds%d\n", mds
);
2249 req
= rb_entry(p
, struct ceph_mds_request
, r_node
);
2251 if (req
->r_got_unsafe
)
2253 if (req
->r_attempts
> 0)
2254 continue; /* only new requests */
2255 if (req
->r_session
&&
2256 req
->r_session
->s_mds
== mds
) {
2257 dout(" kicking tid %llu\n", req
->r_tid
);
2258 list_del_init(&req
->r_wait
);
2259 __do_request(mdsc
, req
);
2264 void ceph_mdsc_submit_request(struct ceph_mds_client
*mdsc
,
2265 struct ceph_mds_request
*req
)
2267 dout("submit_request on %p\n", req
);
2268 mutex_lock(&mdsc
->mutex
);
2269 __register_request(mdsc
, req
, NULL
);
2270 __do_request(mdsc
, req
);
2271 mutex_unlock(&mdsc
->mutex
);
2275 * Synchrously perform an mds request. Take care of all of the
2276 * session setup, forwarding, retry details.
2278 int ceph_mdsc_do_request(struct ceph_mds_client
*mdsc
,
2280 struct ceph_mds_request
*req
)
2284 dout("do_request on %p\n", req
);
2286 /* take CAP_PIN refs for r_inode, r_locked_dir, r_old_dentry */
2288 ceph_get_cap_refs(ceph_inode(req
->r_inode
), CEPH_CAP_PIN
);
2289 if (req
->r_locked_dir
)
2290 ceph_get_cap_refs(ceph_inode(req
->r_locked_dir
), CEPH_CAP_PIN
);
2291 if (req
->r_old_dentry_dir
)
2292 ceph_get_cap_refs(ceph_inode(req
->r_old_dentry_dir
),
2295 /* deny access to directories with pool_ns layouts */
2296 if (req
->r_inode
&& S_ISDIR(req
->r_inode
->i_mode
) &&
2297 ceph_inode(req
->r_inode
)->i_pool_ns_len
)
2299 if (req
->r_locked_dir
&&
2300 ceph_inode(req
->r_locked_dir
)->i_pool_ns_len
)
2304 mutex_lock(&mdsc
->mutex
);
2305 __register_request(mdsc
, req
, dir
);
2306 __do_request(mdsc
, req
);
2314 mutex_unlock(&mdsc
->mutex
);
2315 dout("do_request waiting\n");
2316 if (!req
->r_timeout
&& req
->r_wait_for_completion
) {
2317 err
= req
->r_wait_for_completion(mdsc
, req
);
2319 long timeleft
= wait_for_completion_killable_timeout(
2321 ceph_timeout_jiffies(req
->r_timeout
));
2325 err
= -EIO
; /* timed out */
2327 err
= timeleft
; /* killed */
2329 dout("do_request waited, got %d\n", err
);
2330 mutex_lock(&mdsc
->mutex
);
2332 /* only abort if we didn't race with a real reply */
2333 if (req
->r_got_result
) {
2334 err
= le32_to_cpu(req
->r_reply_info
.head
->result
);
2335 } else if (err
< 0) {
2336 dout("aborted request %lld with %d\n", req
->r_tid
, err
);
2339 * ensure we aren't running concurrently with
2340 * ceph_fill_trace or ceph_readdir_prepopulate, which
2341 * rely on locks (dir mutex) held by our caller.
2343 mutex_lock(&req
->r_fill_mutex
);
2345 req
->r_aborted
= true;
2346 mutex_unlock(&req
->r_fill_mutex
);
2348 if (req
->r_locked_dir
&&
2349 (req
->r_op
& CEPH_MDS_OP_WRITE
))
2350 ceph_invalidate_dir_request(req
);
2356 mutex_unlock(&mdsc
->mutex
);
2357 dout("do_request %p done, result %d\n", req
, err
);
2362 * Invalidate dir's completeness, dentry lease state on an aborted MDS
2363 * namespace request.
2365 void ceph_invalidate_dir_request(struct ceph_mds_request
*req
)
2367 struct inode
*inode
= req
->r_locked_dir
;
2369 dout("invalidate_dir_request %p (complete, lease(s))\n", inode
);
2371 ceph_dir_clear_complete(inode
);
2373 ceph_invalidate_dentry_lease(req
->r_dentry
);
2374 if (req
->r_old_dentry
)
2375 ceph_invalidate_dentry_lease(req
->r_old_dentry
);
2381 * We take the session mutex and parse and process the reply immediately.
2382 * This preserves the logical ordering of replies, capabilities, etc., sent
2383 * by the MDS as they are applied to our local cache.
2385 static void handle_reply(struct ceph_mds_session
*session
, struct ceph_msg
*msg
)
2387 struct ceph_mds_client
*mdsc
= session
->s_mdsc
;
2388 struct ceph_mds_request
*req
;
2389 struct ceph_mds_reply_head
*head
= msg
->front
.iov_base
;
2390 struct ceph_mds_reply_info_parsed
*rinfo
; /* parsed reply info */
2391 struct ceph_snap_realm
*realm
;
2394 int mds
= session
->s_mds
;
2396 if (msg
->front
.iov_len
< sizeof(*head
)) {
2397 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
2402 /* get request, session */
2403 tid
= le64_to_cpu(msg
->hdr
.tid
);
2404 mutex_lock(&mdsc
->mutex
);
2405 req
= lookup_get_request(mdsc
, tid
);
2407 dout("handle_reply on unknown tid %llu\n", tid
);
2408 mutex_unlock(&mdsc
->mutex
);
2411 dout("handle_reply %p\n", req
);
2413 /* correct session? */
2414 if (req
->r_session
!= session
) {
2415 pr_err("mdsc_handle_reply got %llu on session mds%d"
2416 " not mds%d\n", tid
, session
->s_mds
,
2417 req
->r_session
? req
->r_session
->s_mds
: -1);
2418 mutex_unlock(&mdsc
->mutex
);
2423 if ((req
->r_got_unsafe
&& !head
->safe
) ||
2424 (req
->r_got_safe
&& head
->safe
)) {
2425 pr_warn("got a dup %s reply on %llu from mds%d\n",
2426 head
->safe
? "safe" : "unsafe", tid
, mds
);
2427 mutex_unlock(&mdsc
->mutex
);
2430 if (req
->r_got_safe
) {
2431 pr_warn("got unsafe after safe on %llu from mds%d\n",
2433 mutex_unlock(&mdsc
->mutex
);
2437 result
= le32_to_cpu(head
->result
);
2441 * if we're not talking to the authority, send to them
2442 * if the authority has changed while we weren't looking,
2443 * send to new authority
2444 * Otherwise we just have to return an ESTALE
2446 if (result
== -ESTALE
) {
2447 dout("got ESTALE on request %llu", req
->r_tid
);
2448 req
->r_resend_mds
= -1;
2449 if (req
->r_direct_mode
!= USE_AUTH_MDS
) {
2450 dout("not using auth, setting for that now");
2451 req
->r_direct_mode
= USE_AUTH_MDS
;
2452 __do_request(mdsc
, req
);
2453 mutex_unlock(&mdsc
->mutex
);
2456 int mds
= __choose_mds(mdsc
, req
);
2457 if (mds
>= 0 && mds
!= req
->r_session
->s_mds
) {
2458 dout("but auth changed, so resending");
2459 __do_request(mdsc
, req
);
2460 mutex_unlock(&mdsc
->mutex
);
2464 dout("have to return ESTALE on request %llu", req
->r_tid
);
2469 req
->r_got_safe
= true;
2470 __unregister_request(mdsc
, req
);
2472 if (req
->r_got_unsafe
) {
2474 * We already handled the unsafe response, now do the
2475 * cleanup. No need to examine the response; the MDS
2476 * doesn't include any result info in the safe
2477 * response. And even if it did, there is nothing
2478 * useful we could do with a revised return value.
2480 dout("got safe reply %llu, mds%d\n", tid
, mds
);
2481 list_del_init(&req
->r_unsafe_item
);
2483 /* last unsafe request during umount? */
2484 if (mdsc
->stopping
&& !__get_oldest_req(mdsc
))
2485 complete_all(&mdsc
->safe_umount_waiters
);
2486 mutex_unlock(&mdsc
->mutex
);
2490 req
->r_got_unsafe
= true;
2491 list_add_tail(&req
->r_unsafe_item
, &req
->r_session
->s_unsafe
);
2492 if (req
->r_unsafe_dir
) {
2493 struct ceph_inode_info
*ci
=
2494 ceph_inode(req
->r_unsafe_dir
);
2495 spin_lock(&ci
->i_unsafe_lock
);
2496 list_add_tail(&req
->r_unsafe_dir_item
,
2497 &ci
->i_unsafe_dirops
);
2498 spin_unlock(&ci
->i_unsafe_lock
);
2502 dout("handle_reply tid %lld result %d\n", tid
, result
);
2503 rinfo
= &req
->r_reply_info
;
2504 err
= parse_reply_info(msg
, rinfo
, session
->s_con
.peer_features
);
2505 mutex_unlock(&mdsc
->mutex
);
2507 mutex_lock(&session
->s_mutex
);
2509 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds
, tid
);
2516 if (rinfo
->snapblob_len
) {
2517 down_write(&mdsc
->snap_rwsem
);
2518 ceph_update_snap_trace(mdsc
, rinfo
->snapblob
,
2519 rinfo
->snapblob
+ rinfo
->snapblob_len
,
2520 le32_to_cpu(head
->op
) == CEPH_MDS_OP_RMSNAP
,
2522 downgrade_write(&mdsc
->snap_rwsem
);
2524 down_read(&mdsc
->snap_rwsem
);
2527 /* insert trace into our cache */
2528 mutex_lock(&req
->r_fill_mutex
);
2529 current
->journal_info
= req
;
2530 err
= ceph_fill_trace(mdsc
->fsc
->sb
, req
, req
->r_session
);
2532 if (result
== 0 && (req
->r_op
== CEPH_MDS_OP_READDIR
||
2533 req
->r_op
== CEPH_MDS_OP_LSSNAP
))
2534 ceph_readdir_prepopulate(req
, req
->r_session
);
2535 ceph_unreserve_caps(mdsc
, &req
->r_caps_reservation
);
2537 current
->journal_info
= NULL
;
2538 mutex_unlock(&req
->r_fill_mutex
);
2540 up_read(&mdsc
->snap_rwsem
);
2542 ceph_put_snap_realm(mdsc
, realm
);
2544 if (err
== 0 && req
->r_got_unsafe
&& req
->r_target_inode
) {
2545 struct ceph_inode_info
*ci
= ceph_inode(req
->r_target_inode
);
2546 spin_lock(&ci
->i_unsafe_lock
);
2547 list_add_tail(&req
->r_unsafe_target_item
, &ci
->i_unsafe_iops
);
2548 spin_unlock(&ci
->i_unsafe_lock
);
2551 mutex_lock(&mdsc
->mutex
);
2552 if (!req
->r_aborted
) {
2556 req
->r_reply
= ceph_msg_get(msg
);
2557 req
->r_got_result
= true;
2560 dout("reply arrived after request %lld was aborted\n", tid
);
2562 mutex_unlock(&mdsc
->mutex
);
2564 mutex_unlock(&session
->s_mutex
);
2566 /* kick calling process */
2567 complete_request(mdsc
, req
);
2569 ceph_mdsc_put_request(req
);
2576 * handle mds notification that our request has been forwarded.
2578 static void handle_forward(struct ceph_mds_client
*mdsc
,
2579 struct ceph_mds_session
*session
,
2580 struct ceph_msg
*msg
)
2582 struct ceph_mds_request
*req
;
2583 u64 tid
= le64_to_cpu(msg
->hdr
.tid
);
2587 void *p
= msg
->front
.iov_base
;
2588 void *end
= p
+ msg
->front
.iov_len
;
2590 ceph_decode_need(&p
, end
, 2*sizeof(u32
), bad
);
2591 next_mds
= ceph_decode_32(&p
);
2592 fwd_seq
= ceph_decode_32(&p
);
2594 mutex_lock(&mdsc
->mutex
);
2595 req
= lookup_get_request(mdsc
, tid
);
2597 dout("forward tid %llu to mds%d - req dne\n", tid
, next_mds
);
2598 goto out
; /* dup reply? */
2601 if (req
->r_aborted
) {
2602 dout("forward tid %llu aborted, unregistering\n", tid
);
2603 __unregister_request(mdsc
, req
);
2604 } else if (fwd_seq
<= req
->r_num_fwd
) {
2605 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
2606 tid
, next_mds
, req
->r_num_fwd
, fwd_seq
);
2608 /* resend. forward race not possible; mds would drop */
2609 dout("forward tid %llu to mds%d (we resend)\n", tid
, next_mds
);
2611 BUG_ON(req
->r_got_result
);
2612 req
->r_attempts
= 0;
2613 req
->r_num_fwd
= fwd_seq
;
2614 req
->r_resend_mds
= next_mds
;
2615 put_request_session(req
);
2616 __do_request(mdsc
, req
);
2618 ceph_mdsc_put_request(req
);
2620 mutex_unlock(&mdsc
->mutex
);
2624 pr_err("mdsc_handle_forward decode error err=%d\n", err
);
2628 * handle a mds session control message
2630 static void handle_session(struct ceph_mds_session
*session
,
2631 struct ceph_msg
*msg
)
2633 struct ceph_mds_client
*mdsc
= session
->s_mdsc
;
2636 int mds
= session
->s_mds
;
2637 struct ceph_mds_session_head
*h
= msg
->front
.iov_base
;
2641 if (msg
->front
.iov_len
!= sizeof(*h
))
2643 op
= le32_to_cpu(h
->op
);
2644 seq
= le64_to_cpu(h
->seq
);
2646 mutex_lock(&mdsc
->mutex
);
2647 if (op
== CEPH_SESSION_CLOSE
)
2648 __unregister_session(mdsc
, session
);
2649 /* FIXME: this ttl calculation is generous */
2650 session
->s_ttl
= jiffies
+ HZ
*mdsc
->mdsmap
->m_session_autoclose
;
2651 mutex_unlock(&mdsc
->mutex
);
2653 mutex_lock(&session
->s_mutex
);
2655 dout("handle_session mds%d %s %p state %s seq %llu\n",
2656 mds
, ceph_session_op_name(op
), session
,
2657 ceph_session_state_name(session
->s_state
), seq
);
2659 if (session
->s_state
== CEPH_MDS_SESSION_HUNG
) {
2660 session
->s_state
= CEPH_MDS_SESSION_OPEN
;
2661 pr_info("mds%d came back\n", session
->s_mds
);
2665 case CEPH_SESSION_OPEN
:
2666 if (session
->s_state
== CEPH_MDS_SESSION_RECONNECTING
)
2667 pr_info("mds%d reconnect success\n", session
->s_mds
);
2668 session
->s_state
= CEPH_MDS_SESSION_OPEN
;
2669 renewed_caps(mdsc
, session
, 0);
2672 __close_session(mdsc
, session
);
2675 case CEPH_SESSION_RENEWCAPS
:
2676 if (session
->s_renew_seq
== seq
)
2677 renewed_caps(mdsc
, session
, 1);
2680 case CEPH_SESSION_CLOSE
:
2681 if (session
->s_state
== CEPH_MDS_SESSION_RECONNECTING
)
2682 pr_info("mds%d reconnect denied\n", session
->s_mds
);
2683 cleanup_session_requests(mdsc
, session
);
2684 remove_session_caps(session
);
2685 wake
= 2; /* for good measure */
2686 wake_up_all(&mdsc
->session_close_wq
);
2689 case CEPH_SESSION_STALE
:
2690 pr_info("mds%d caps went stale, renewing\n",
2692 spin_lock(&session
->s_gen_ttl_lock
);
2693 session
->s_cap_gen
++;
2694 session
->s_cap_ttl
= jiffies
- 1;
2695 spin_unlock(&session
->s_gen_ttl_lock
);
2696 send_renew_caps(mdsc
, session
);
2699 case CEPH_SESSION_RECALL_STATE
:
2700 trim_caps(mdsc
, session
, le32_to_cpu(h
->max_caps
));
2703 case CEPH_SESSION_FLUSHMSG
:
2704 send_flushmsg_ack(mdsc
, session
, seq
);
2707 case CEPH_SESSION_FORCE_RO
:
2708 dout("force_session_readonly %p\n", session
);
2709 spin_lock(&session
->s_cap_lock
);
2710 session
->s_readonly
= true;
2711 spin_unlock(&session
->s_cap_lock
);
2712 wake_up_session_caps(session
, 0);
2716 pr_err("mdsc_handle_session bad op %d mds%d\n", op
, mds
);
2720 mutex_unlock(&session
->s_mutex
);
2722 mutex_lock(&mdsc
->mutex
);
2723 __wake_requests(mdsc
, &session
->s_waiting
);
2725 kick_requests(mdsc
, mds
);
2726 mutex_unlock(&mdsc
->mutex
);
2731 pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds
,
2732 (int)msg
->front
.iov_len
);
2739 * called under session->mutex.
2741 static void replay_unsafe_requests(struct ceph_mds_client
*mdsc
,
2742 struct ceph_mds_session
*session
)
2744 struct ceph_mds_request
*req
, *nreq
;
2748 dout("replay_unsafe_requests mds%d\n", session
->s_mds
);
2750 mutex_lock(&mdsc
->mutex
);
2751 list_for_each_entry_safe(req
, nreq
, &session
->s_unsafe
, r_unsafe_item
) {
2752 err
= __prepare_send_request(mdsc
, req
, session
->s_mds
, true);
2754 ceph_msg_get(req
->r_request
);
2755 ceph_con_send(&session
->s_con
, req
->r_request
);
2760 * also re-send old requests when MDS enters reconnect stage. So that MDS
2761 * can process completed request in clientreplay stage.
2763 p
= rb_first(&mdsc
->request_tree
);
2765 req
= rb_entry(p
, struct ceph_mds_request
, r_node
);
2767 if (req
->r_got_unsafe
)
2769 if (req
->r_attempts
== 0)
2770 continue; /* only old requests */
2771 if (req
->r_session
&&
2772 req
->r_session
->s_mds
== session
->s_mds
) {
2773 err
= __prepare_send_request(mdsc
, req
,
2774 session
->s_mds
, true);
2776 ceph_msg_get(req
->r_request
);
2777 ceph_con_send(&session
->s_con
, req
->r_request
);
2781 mutex_unlock(&mdsc
->mutex
);
2785 * Encode information about a cap for a reconnect with the MDS.
2787 static int encode_caps_cb(struct inode
*inode
, struct ceph_cap
*cap
,
2791 struct ceph_mds_cap_reconnect v2
;
2792 struct ceph_mds_cap_reconnect_v1 v1
;
2795 struct ceph_inode_info
*ci
;
2796 struct ceph_reconnect_state
*recon_state
= arg
;
2797 struct ceph_pagelist
*pagelist
= recon_state
->pagelist
;
2801 struct dentry
*dentry
;
2805 dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
2806 inode
, ceph_vinop(inode
), cap
, cap
->cap_id
,
2807 ceph_cap_string(cap
->issued
));
2808 err
= ceph_pagelist_encode_64(pagelist
, ceph_ino(inode
));
2812 dentry
= d_find_alias(inode
);
2814 path
= ceph_mdsc_build_path(dentry
, &pathlen
, &pathbase
, 0);
2816 err
= PTR_ERR(path
);
2823 err
= ceph_pagelist_encode_string(pagelist
, path
, pathlen
);
2827 spin_lock(&ci
->i_ceph_lock
);
2828 cap
->seq
= 0; /* reset cap seq */
2829 cap
->issue_seq
= 0; /* and issue_seq */
2830 cap
->mseq
= 0; /* and migrate_seq */
2831 cap
->cap_gen
= cap
->session
->s_cap_gen
;
2833 if (recon_state
->flock
) {
2834 rec
.v2
.cap_id
= cpu_to_le64(cap
->cap_id
);
2835 rec
.v2
.wanted
= cpu_to_le32(__ceph_caps_wanted(ci
));
2836 rec
.v2
.issued
= cpu_to_le32(cap
->issued
);
2837 rec
.v2
.snaprealm
= cpu_to_le64(ci
->i_snap_realm
->ino
);
2838 rec
.v2
.pathbase
= cpu_to_le64(pathbase
);
2839 rec
.v2
.flock_len
= 0;
2840 reclen
= sizeof(rec
.v2
);
2842 rec
.v1
.cap_id
= cpu_to_le64(cap
->cap_id
);
2843 rec
.v1
.wanted
= cpu_to_le32(__ceph_caps_wanted(ci
));
2844 rec
.v1
.issued
= cpu_to_le32(cap
->issued
);
2845 rec
.v1
.size
= cpu_to_le64(inode
->i_size
);
2846 ceph_encode_timespec(&rec
.v1
.mtime
, &inode
->i_mtime
);
2847 ceph_encode_timespec(&rec
.v1
.atime
, &inode
->i_atime
);
2848 rec
.v1
.snaprealm
= cpu_to_le64(ci
->i_snap_realm
->ino
);
2849 rec
.v1
.pathbase
= cpu_to_le64(pathbase
);
2850 reclen
= sizeof(rec
.v1
);
2852 spin_unlock(&ci
->i_ceph_lock
);
2854 if (recon_state
->flock
) {
2855 int num_fcntl_locks
, num_flock_locks
;
2856 struct ceph_filelock
*flocks
;
2859 ceph_count_locks(inode
, &num_fcntl_locks
, &num_flock_locks
);
2860 flocks
= kmalloc((num_fcntl_locks
+num_flock_locks
) *
2861 sizeof(struct ceph_filelock
), GFP_NOFS
);
2866 err
= ceph_encode_locks_to_buffer(inode
, flocks
,
2876 * number of encoded locks is stable, so copy to pagelist
2878 rec
.v2
.flock_len
= cpu_to_le32(2*sizeof(u32
) +
2879 (num_fcntl_locks
+num_flock_locks
) *
2880 sizeof(struct ceph_filelock
));
2881 err
= ceph_pagelist_append(pagelist
, &rec
, reclen
);
2883 err
= ceph_locks_to_pagelist(flocks
, pagelist
,
2888 err
= ceph_pagelist_append(pagelist
, &rec
, reclen
);
2891 recon_state
->nr_caps
++;
2901 * If an MDS fails and recovers, clients need to reconnect in order to
2902 * reestablish shared state. This includes all caps issued through
2903 * this session _and_ the snap_realm hierarchy. Because it's not
2904 * clear which snap realms the mds cares about, we send everything we
2905 * know about.. that ensures we'll then get any new info the
2906 * recovering MDS might have.
2908 * This is a relatively heavyweight operation, but it's rare.
2910 * called with mdsc->mutex held.
2912 static void send_mds_reconnect(struct ceph_mds_client
*mdsc
,
2913 struct ceph_mds_session
*session
)
2915 struct ceph_msg
*reply
;
2917 int mds
= session
->s_mds
;
2920 struct ceph_pagelist
*pagelist
;
2921 struct ceph_reconnect_state recon_state
;
2923 pr_info("mds%d reconnect start\n", mds
);
2925 pagelist
= kmalloc(sizeof(*pagelist
), GFP_NOFS
);
2927 goto fail_nopagelist
;
2928 ceph_pagelist_init(pagelist
);
2930 reply
= ceph_msg_new(CEPH_MSG_CLIENT_RECONNECT
, 0, GFP_NOFS
, false);
2934 mutex_lock(&session
->s_mutex
);
2935 session
->s_state
= CEPH_MDS_SESSION_RECONNECTING
;
2938 dout("session %p state %s\n", session
,
2939 ceph_session_state_name(session
->s_state
));
2941 spin_lock(&session
->s_gen_ttl_lock
);
2942 session
->s_cap_gen
++;
2943 spin_unlock(&session
->s_gen_ttl_lock
);
2945 spin_lock(&session
->s_cap_lock
);
2946 /* don't know if session is readonly */
2947 session
->s_readonly
= 0;
2949 * notify __ceph_remove_cap() that we are composing cap reconnect.
2950 * If a cap get released before being added to the cap reconnect,
2951 * __ceph_remove_cap() should skip queuing cap release.
2953 session
->s_cap_reconnect
= 1;
2954 /* drop old cap expires; we're about to reestablish that state */
2955 cleanup_cap_releases(mdsc
, session
);
2957 /* trim unused caps to reduce MDS's cache rejoin time */
2958 if (mdsc
->fsc
->sb
->s_root
)
2959 shrink_dcache_parent(mdsc
->fsc
->sb
->s_root
);
2961 ceph_con_close(&session
->s_con
);
2962 ceph_con_open(&session
->s_con
,
2963 CEPH_ENTITY_TYPE_MDS
, mds
,
2964 ceph_mdsmap_get_addr(mdsc
->mdsmap
, mds
));
2966 /* replay unsafe requests */
2967 replay_unsafe_requests(mdsc
, session
);
2969 down_read(&mdsc
->snap_rwsem
);
2971 /* traverse this session's caps */
2972 s_nr_caps
= session
->s_nr_caps
;
2973 err
= ceph_pagelist_encode_32(pagelist
, s_nr_caps
);
2977 recon_state
.nr_caps
= 0;
2978 recon_state
.pagelist
= pagelist
;
2979 recon_state
.flock
= session
->s_con
.peer_features
& CEPH_FEATURE_FLOCK
;
2980 err
= iterate_session_caps(session
, encode_caps_cb
, &recon_state
);
2984 spin_lock(&session
->s_cap_lock
);
2985 session
->s_cap_reconnect
= 0;
2986 spin_unlock(&session
->s_cap_lock
);
2989 * snaprealms. we provide mds with the ino, seq (version), and
2990 * parent for all of our realms. If the mds has any newer info,
2993 for (p
= rb_first(&mdsc
->snap_realms
); p
; p
= rb_next(p
)) {
2994 struct ceph_snap_realm
*realm
=
2995 rb_entry(p
, struct ceph_snap_realm
, node
);
2996 struct ceph_mds_snaprealm_reconnect sr_rec
;
2998 dout(" adding snap realm %llx seq %lld parent %llx\n",
2999 realm
->ino
, realm
->seq
, realm
->parent_ino
);
3000 sr_rec
.ino
= cpu_to_le64(realm
->ino
);
3001 sr_rec
.seq
= cpu_to_le64(realm
->seq
);
3002 sr_rec
.parent
= cpu_to_le64(realm
->parent_ino
);
3003 err
= ceph_pagelist_append(pagelist
, &sr_rec
, sizeof(sr_rec
));
3008 if (recon_state
.flock
)
3009 reply
->hdr
.version
= cpu_to_le16(2);
3011 /* raced with cap release? */
3012 if (s_nr_caps
!= recon_state
.nr_caps
) {
3013 struct page
*page
= list_first_entry(&pagelist
->head
,
3015 __le32
*addr
= kmap_atomic(page
);
3016 *addr
= cpu_to_le32(recon_state
.nr_caps
);
3017 kunmap_atomic(addr
);
3020 reply
->hdr
.data_len
= cpu_to_le32(pagelist
->length
);
3021 ceph_msg_data_add_pagelist(reply
, pagelist
);
3023 ceph_early_kick_flushing_caps(mdsc
, session
);
3025 ceph_con_send(&session
->s_con
, reply
);
3027 mutex_unlock(&session
->s_mutex
);
3029 mutex_lock(&mdsc
->mutex
);
3030 __wake_requests(mdsc
, &session
->s_waiting
);
3031 mutex_unlock(&mdsc
->mutex
);
3033 up_read(&mdsc
->snap_rwsem
);
3037 ceph_msg_put(reply
);
3038 up_read(&mdsc
->snap_rwsem
);
3039 mutex_unlock(&session
->s_mutex
);
3041 ceph_pagelist_release(pagelist
);
3043 pr_err("error %d preparing reconnect for mds%d\n", err
, mds
);
3049 * compare old and new mdsmaps, kicking requests
3050 * and closing out old connections as necessary
3052 * called under mdsc->mutex.
3054 static void check_new_map(struct ceph_mds_client
*mdsc
,
3055 struct ceph_mdsmap
*newmap
,
3056 struct ceph_mdsmap
*oldmap
)
3059 int oldstate
, newstate
;
3060 struct ceph_mds_session
*s
;
3062 dout("check_new_map new %u old %u\n",
3063 newmap
->m_epoch
, oldmap
->m_epoch
);
3065 for (i
= 0; i
< oldmap
->m_max_mds
&& i
< mdsc
->max_sessions
; i
++) {
3066 if (mdsc
->sessions
[i
] == NULL
)
3068 s
= mdsc
->sessions
[i
];
3069 oldstate
= ceph_mdsmap_get_state(oldmap
, i
);
3070 newstate
= ceph_mdsmap_get_state(newmap
, i
);
3072 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
3073 i
, ceph_mds_state_name(oldstate
),
3074 ceph_mdsmap_is_laggy(oldmap
, i
) ? " (laggy)" : "",
3075 ceph_mds_state_name(newstate
),
3076 ceph_mdsmap_is_laggy(newmap
, i
) ? " (laggy)" : "",
3077 ceph_session_state_name(s
->s_state
));
3079 if (i
>= newmap
->m_max_mds
||
3080 memcmp(ceph_mdsmap_get_addr(oldmap
, i
),
3081 ceph_mdsmap_get_addr(newmap
, i
),
3082 sizeof(struct ceph_entity_addr
))) {
3083 if (s
->s_state
== CEPH_MDS_SESSION_OPENING
) {
3084 /* the session never opened, just close it
3086 __wake_requests(mdsc
, &s
->s_waiting
);
3087 __unregister_session(mdsc
, s
);
3090 mutex_unlock(&mdsc
->mutex
);
3091 mutex_lock(&s
->s_mutex
);
3092 mutex_lock(&mdsc
->mutex
);
3093 ceph_con_close(&s
->s_con
);
3094 mutex_unlock(&s
->s_mutex
);
3095 s
->s_state
= CEPH_MDS_SESSION_RESTARTING
;
3097 } else if (oldstate
== newstate
) {
3098 continue; /* nothing new with this mds */
3104 if (s
->s_state
== CEPH_MDS_SESSION_RESTARTING
&&
3105 newstate
>= CEPH_MDS_STATE_RECONNECT
) {
3106 mutex_unlock(&mdsc
->mutex
);
3107 send_mds_reconnect(mdsc
, s
);
3108 mutex_lock(&mdsc
->mutex
);
3112 * kick request on any mds that has gone active.
3114 if (oldstate
< CEPH_MDS_STATE_ACTIVE
&&
3115 newstate
>= CEPH_MDS_STATE_ACTIVE
) {
3116 if (oldstate
!= CEPH_MDS_STATE_CREATING
&&
3117 oldstate
!= CEPH_MDS_STATE_STARTING
)
3118 pr_info("mds%d recovery completed\n", s
->s_mds
);
3119 kick_requests(mdsc
, i
);
3120 ceph_kick_flushing_caps(mdsc
, s
);
3121 wake_up_session_caps(s
, 1);
3125 for (i
= 0; i
< newmap
->m_max_mds
&& i
< mdsc
->max_sessions
; i
++) {
3126 s
= mdsc
->sessions
[i
];
3129 if (!ceph_mdsmap_is_laggy(newmap
, i
))
3131 if (s
->s_state
== CEPH_MDS_SESSION_OPEN
||
3132 s
->s_state
== CEPH_MDS_SESSION_HUNG
||
3133 s
->s_state
== CEPH_MDS_SESSION_CLOSING
) {
3134 dout(" connecting to export targets of laggy mds%d\n",
3136 __open_export_target_sessions(mdsc
, s
);
3148 * caller must hold session s_mutex, dentry->d_lock
3150 void __ceph_mdsc_drop_dentry_lease(struct dentry
*dentry
)
3152 struct ceph_dentry_info
*di
= ceph_dentry(dentry
);
3154 ceph_put_mds_session(di
->lease_session
);
3155 di
->lease_session
= NULL
;
3158 static void handle_lease(struct ceph_mds_client
*mdsc
,
3159 struct ceph_mds_session
*session
,
3160 struct ceph_msg
*msg
)
3162 struct super_block
*sb
= mdsc
->fsc
->sb
;
3163 struct inode
*inode
;
3164 struct dentry
*parent
, *dentry
;
3165 struct ceph_dentry_info
*di
;
3166 int mds
= session
->s_mds
;
3167 struct ceph_mds_lease
*h
= msg
->front
.iov_base
;
3169 struct ceph_vino vino
;
3173 dout("handle_lease from mds%d\n", mds
);
3176 if (msg
->front
.iov_len
< sizeof(*h
) + sizeof(u32
))
3178 vino
.ino
= le64_to_cpu(h
->ino
);
3179 vino
.snap
= CEPH_NOSNAP
;
3180 seq
= le32_to_cpu(h
->seq
);
3181 dname
.name
= (void *)h
+ sizeof(*h
) + sizeof(u32
);
3182 dname
.len
= msg
->front
.iov_len
- sizeof(*h
) - sizeof(u32
);
3183 if (dname
.len
!= get_unaligned_le32(h
+1))
3187 inode
= ceph_find_inode(sb
, vino
);
3188 dout("handle_lease %s, ino %llx %p %.*s\n",
3189 ceph_lease_op_name(h
->action
), vino
.ino
, inode
,
3190 dname
.len
, dname
.name
);
3192 mutex_lock(&session
->s_mutex
);
3195 if (inode
== NULL
) {
3196 dout("handle_lease no inode %llx\n", vino
.ino
);
3201 parent
= d_find_alias(inode
);
3203 dout("no parent dentry on inode %p\n", inode
);
3205 goto release
; /* hrm... */
3207 dname
.hash
= full_name_hash(dname
.name
, dname
.len
);
3208 dentry
= d_lookup(parent
, &dname
);
3213 spin_lock(&dentry
->d_lock
);
3214 di
= ceph_dentry(dentry
);
3215 switch (h
->action
) {
3216 case CEPH_MDS_LEASE_REVOKE
:
3217 if (di
->lease_session
== session
) {
3218 if (ceph_seq_cmp(di
->lease_seq
, seq
) > 0)
3219 h
->seq
= cpu_to_le32(di
->lease_seq
);
3220 __ceph_mdsc_drop_dentry_lease(dentry
);
3225 case CEPH_MDS_LEASE_RENEW
:
3226 if (di
->lease_session
== session
&&
3227 di
->lease_gen
== session
->s_cap_gen
&&
3228 di
->lease_renew_from
&&
3229 di
->lease_renew_after
== 0) {
3230 unsigned long duration
=
3231 msecs_to_jiffies(le32_to_cpu(h
->duration_ms
));
3233 di
->lease_seq
= seq
;
3234 dentry
->d_time
= di
->lease_renew_from
+ duration
;
3235 di
->lease_renew_after
= di
->lease_renew_from
+
3237 di
->lease_renew_from
= 0;
3241 spin_unlock(&dentry
->d_lock
);
3248 /* let's just reuse the same message */
3249 h
->action
= CEPH_MDS_LEASE_REVOKE_ACK
;
3251 ceph_con_send(&session
->s_con
, msg
);
3255 mutex_unlock(&session
->s_mutex
);
3259 pr_err("corrupt lease message\n");
3263 void ceph_mdsc_lease_send_msg(struct ceph_mds_session
*session
,
3264 struct inode
*inode
,
3265 struct dentry
*dentry
, char action
,
3268 struct ceph_msg
*msg
;
3269 struct ceph_mds_lease
*lease
;
3270 int len
= sizeof(*lease
) + sizeof(u32
);
3273 dout("lease_send_msg inode %p dentry %p %s to mds%d\n",
3274 inode
, dentry
, ceph_lease_op_name(action
), session
->s_mds
);
3275 dnamelen
= dentry
->d_name
.len
;
3278 msg
= ceph_msg_new(CEPH_MSG_CLIENT_LEASE
, len
, GFP_NOFS
, false);
3281 lease
= msg
->front
.iov_base
;
3282 lease
->action
= action
;
3283 lease
->ino
= cpu_to_le64(ceph_vino(inode
).ino
);
3284 lease
->first
= lease
->last
= cpu_to_le64(ceph_vino(inode
).snap
);
3285 lease
->seq
= cpu_to_le32(seq
);
3286 put_unaligned_le32(dnamelen
, lease
+ 1);
3287 memcpy((void *)(lease
+ 1) + 4, dentry
->d_name
.name
, dnamelen
);
3290 * if this is a preemptive lease RELEASE, no need to
3291 * flush request stream, since the actual request will
3294 msg
->more_to_follow
= (action
== CEPH_MDS_LEASE_RELEASE
);
3296 ceph_con_send(&session
->s_con
, msg
);
3300 * Preemptively release a lease we expect to invalidate anyway.
3301 * Pass @inode always, @dentry is optional.
3303 void ceph_mdsc_lease_release(struct ceph_mds_client
*mdsc
, struct inode
*inode
,
3304 struct dentry
*dentry
)
3306 struct ceph_dentry_info
*di
;
3307 struct ceph_mds_session
*session
;
3310 BUG_ON(inode
== NULL
);
3311 BUG_ON(dentry
== NULL
);
3313 /* is dentry lease valid? */
3314 spin_lock(&dentry
->d_lock
);
3315 di
= ceph_dentry(dentry
);
3316 if (!di
|| !di
->lease_session
||
3317 di
->lease_session
->s_mds
< 0 ||
3318 di
->lease_gen
!= di
->lease_session
->s_cap_gen
||
3319 !time_before(jiffies
, dentry
->d_time
)) {
3320 dout("lease_release inode %p dentry %p -- "
3323 spin_unlock(&dentry
->d_lock
);
3327 /* we do have a lease on this dentry; note mds and seq */
3328 session
= ceph_get_mds_session(di
->lease_session
);
3329 seq
= di
->lease_seq
;
3330 __ceph_mdsc_drop_dentry_lease(dentry
);
3331 spin_unlock(&dentry
->d_lock
);
3333 dout("lease_release inode %p dentry %p to mds%d\n",
3334 inode
, dentry
, session
->s_mds
);
3335 ceph_mdsc_lease_send_msg(session
, inode
, dentry
,
3336 CEPH_MDS_LEASE_RELEASE
, seq
);
3337 ceph_put_mds_session(session
);
3341 * drop all leases (and dentry refs) in preparation for umount
3343 static void drop_leases(struct ceph_mds_client
*mdsc
)
3347 dout("drop_leases\n");
3348 mutex_lock(&mdsc
->mutex
);
3349 for (i
= 0; i
< mdsc
->max_sessions
; i
++) {
3350 struct ceph_mds_session
*s
= __ceph_lookup_mds_session(mdsc
, i
);
3353 mutex_unlock(&mdsc
->mutex
);
3354 mutex_lock(&s
->s_mutex
);
3355 mutex_unlock(&s
->s_mutex
);
3356 ceph_put_mds_session(s
);
3357 mutex_lock(&mdsc
->mutex
);
3359 mutex_unlock(&mdsc
->mutex
);
3365 * delayed work -- periodically trim expired leases, renew caps with mds
3367 static void schedule_delayed(struct ceph_mds_client
*mdsc
)
3370 unsigned hz
= round_jiffies_relative(HZ
* delay
);
3371 schedule_delayed_work(&mdsc
->delayed_work
, hz
);
3374 static void delayed_work(struct work_struct
*work
)
3377 struct ceph_mds_client
*mdsc
=
3378 container_of(work
, struct ceph_mds_client
, delayed_work
.work
);
3382 dout("mdsc delayed_work\n");
3383 ceph_check_delayed_caps(mdsc
);
3385 mutex_lock(&mdsc
->mutex
);
3386 renew_interval
= mdsc
->mdsmap
->m_session_timeout
>> 2;
3387 renew_caps
= time_after_eq(jiffies
, HZ
*renew_interval
+
3388 mdsc
->last_renew_caps
);
3390 mdsc
->last_renew_caps
= jiffies
;
3392 for (i
= 0; i
< mdsc
->max_sessions
; i
++) {
3393 struct ceph_mds_session
*s
= __ceph_lookup_mds_session(mdsc
, i
);
3396 if (s
->s_state
== CEPH_MDS_SESSION_CLOSING
) {
3397 dout("resending session close request for mds%d\n",
3399 request_close_session(mdsc
, s
);
3400 ceph_put_mds_session(s
);
3403 if (s
->s_ttl
&& time_after(jiffies
, s
->s_ttl
)) {
3404 if (s
->s_state
== CEPH_MDS_SESSION_OPEN
) {
3405 s
->s_state
= CEPH_MDS_SESSION_HUNG
;
3406 pr_info("mds%d hung\n", s
->s_mds
);
3409 if (s
->s_state
< CEPH_MDS_SESSION_OPEN
) {
3410 /* this mds is failed or recovering, just wait */
3411 ceph_put_mds_session(s
);
3414 mutex_unlock(&mdsc
->mutex
);
3416 mutex_lock(&s
->s_mutex
);
3418 send_renew_caps(mdsc
, s
);
3420 ceph_con_keepalive(&s
->s_con
);
3421 if (s
->s_state
== CEPH_MDS_SESSION_OPEN
||
3422 s
->s_state
== CEPH_MDS_SESSION_HUNG
)
3423 ceph_send_cap_releases(mdsc
, s
);
3424 mutex_unlock(&s
->s_mutex
);
3425 ceph_put_mds_session(s
);
3427 mutex_lock(&mdsc
->mutex
);
3429 mutex_unlock(&mdsc
->mutex
);
3431 schedule_delayed(mdsc
);
3434 int ceph_mdsc_init(struct ceph_fs_client
*fsc
)
3437 struct ceph_mds_client
*mdsc
;
3439 mdsc
= kzalloc(sizeof(struct ceph_mds_client
), GFP_NOFS
);
3444 mutex_init(&mdsc
->mutex
);
3445 mdsc
->mdsmap
= kzalloc(sizeof(*mdsc
->mdsmap
), GFP_NOFS
);
3446 if (mdsc
->mdsmap
== NULL
) {
3451 init_completion(&mdsc
->safe_umount_waiters
);
3452 init_waitqueue_head(&mdsc
->session_close_wq
);
3453 INIT_LIST_HEAD(&mdsc
->waiting_for_map
);
3454 mdsc
->sessions
= NULL
;
3455 atomic_set(&mdsc
->num_sessions
, 0);
3456 mdsc
->max_sessions
= 0;
3458 mdsc
->last_snap_seq
= 0;
3459 init_rwsem(&mdsc
->snap_rwsem
);
3460 mdsc
->snap_realms
= RB_ROOT
;
3461 INIT_LIST_HEAD(&mdsc
->snap_empty
);
3462 spin_lock_init(&mdsc
->snap_empty_lock
);
3464 mdsc
->oldest_tid
= 0;
3465 mdsc
->request_tree
= RB_ROOT
;
3466 INIT_DELAYED_WORK(&mdsc
->delayed_work
, delayed_work
);
3467 mdsc
->last_renew_caps
= jiffies
;
3468 INIT_LIST_HEAD(&mdsc
->cap_delay_list
);
3469 spin_lock_init(&mdsc
->cap_delay_lock
);
3470 INIT_LIST_HEAD(&mdsc
->snap_flush_list
);
3471 spin_lock_init(&mdsc
->snap_flush_lock
);
3472 mdsc
->last_cap_flush_tid
= 1;
3473 mdsc
->cap_flush_tree
= RB_ROOT
;
3474 INIT_LIST_HEAD(&mdsc
->cap_dirty
);
3475 INIT_LIST_HEAD(&mdsc
->cap_dirty_migrating
);
3476 mdsc
->num_cap_flushing
= 0;
3477 spin_lock_init(&mdsc
->cap_dirty_lock
);
3478 init_waitqueue_head(&mdsc
->cap_flushing_wq
);
3479 spin_lock_init(&mdsc
->dentry_lru_lock
);
3480 INIT_LIST_HEAD(&mdsc
->dentry_lru
);
3482 ceph_caps_init(mdsc
);
3483 ceph_adjust_min_caps(mdsc
, fsc
->min_caps
);
3485 init_rwsem(&mdsc
->pool_perm_rwsem
);
3486 mdsc
->pool_perm_tree
= RB_ROOT
;
3492 * Wait for safe replies on open mds requests. If we time out, drop
3493 * all requests from the tree to avoid dangling dentry refs.
3495 static void wait_requests(struct ceph_mds_client
*mdsc
)
3497 struct ceph_options
*opts
= mdsc
->fsc
->client
->options
;
3498 struct ceph_mds_request
*req
;
3500 mutex_lock(&mdsc
->mutex
);
3501 if (__get_oldest_req(mdsc
)) {
3502 mutex_unlock(&mdsc
->mutex
);
3504 dout("wait_requests waiting for requests\n");
3505 wait_for_completion_timeout(&mdsc
->safe_umount_waiters
,
3506 ceph_timeout_jiffies(opts
->mount_timeout
));
3508 /* tear down remaining requests */
3509 mutex_lock(&mdsc
->mutex
);
3510 while ((req
= __get_oldest_req(mdsc
))) {
3511 dout("wait_requests timed out on tid %llu\n",
3513 __unregister_request(mdsc
, req
);
3516 mutex_unlock(&mdsc
->mutex
);
3517 dout("wait_requests done\n");
3521 * called before mount is ro, and before dentries are torn down.
3522 * (hmm, does this still race with new lookups?)
3524 void ceph_mdsc_pre_umount(struct ceph_mds_client
*mdsc
)
3526 dout("pre_umount\n");
3530 ceph_flush_dirty_caps(mdsc
);
3531 wait_requests(mdsc
);
3534 * wait for reply handlers to drop their request refs and
3535 * their inode/dcache refs
3541 * wait for all write mds requests to flush.
3543 static void wait_unsafe_requests(struct ceph_mds_client
*mdsc
, u64 want_tid
)
3545 struct ceph_mds_request
*req
= NULL
, *nextreq
;
3548 mutex_lock(&mdsc
->mutex
);
3549 dout("wait_unsafe_requests want %lld\n", want_tid
);
3551 req
= __get_oldest_req(mdsc
);
3552 while (req
&& req
->r_tid
<= want_tid
) {
3553 /* find next request */
3554 n
= rb_next(&req
->r_node
);
3556 nextreq
= rb_entry(n
, struct ceph_mds_request
, r_node
);
3559 if (req
->r_op
!= CEPH_MDS_OP_SETFILELOCK
&&
3560 (req
->r_op
& CEPH_MDS_OP_WRITE
)) {
3562 ceph_mdsc_get_request(req
);
3564 ceph_mdsc_get_request(nextreq
);
3565 mutex_unlock(&mdsc
->mutex
);
3566 dout("wait_unsafe_requests wait on %llu (want %llu)\n",
3567 req
->r_tid
, want_tid
);
3568 wait_for_completion(&req
->r_safe_completion
);
3569 mutex_lock(&mdsc
->mutex
);
3570 ceph_mdsc_put_request(req
);
3572 break; /* next dne before, so we're done! */
3573 if (RB_EMPTY_NODE(&nextreq
->r_node
)) {
3574 /* next request was removed from tree */
3575 ceph_mdsc_put_request(nextreq
);
3578 ceph_mdsc_put_request(nextreq
); /* won't go away */
3582 mutex_unlock(&mdsc
->mutex
);
3583 dout("wait_unsafe_requests done\n");
3586 void ceph_mdsc_sync(struct ceph_mds_client
*mdsc
)
3588 u64 want_tid
, want_flush
, want_snap
;
3590 if (ACCESS_ONCE(mdsc
->fsc
->mount_state
) == CEPH_MOUNT_SHUTDOWN
)
3594 mutex_lock(&mdsc
->mutex
);
3595 want_tid
= mdsc
->last_tid
;
3596 mutex_unlock(&mdsc
->mutex
);
3598 ceph_flush_dirty_caps(mdsc
);
3599 spin_lock(&mdsc
->cap_dirty_lock
);
3600 want_flush
= mdsc
->last_cap_flush_tid
;
3601 spin_unlock(&mdsc
->cap_dirty_lock
);
3603 down_read(&mdsc
->snap_rwsem
);
3604 want_snap
= mdsc
->last_snap_seq
;
3605 up_read(&mdsc
->snap_rwsem
);
3607 dout("sync want tid %lld flush_seq %lld snap_seq %lld\n",
3608 want_tid
, want_flush
, want_snap
);
3610 wait_unsafe_requests(mdsc
, want_tid
);
3611 wait_caps_flush(mdsc
, want_flush
, want_snap
);
3615 * true if all sessions are closed, or we force unmount
3617 static bool done_closing_sessions(struct ceph_mds_client
*mdsc
)
3619 if (ACCESS_ONCE(mdsc
->fsc
->mount_state
) == CEPH_MOUNT_SHUTDOWN
)
3621 return atomic_read(&mdsc
->num_sessions
) == 0;
3625 * called after sb is ro.
3627 void ceph_mdsc_close_sessions(struct ceph_mds_client
*mdsc
)
3629 struct ceph_options
*opts
= mdsc
->fsc
->client
->options
;
3630 struct ceph_mds_session
*session
;
3633 dout("close_sessions\n");
3635 /* close sessions */
3636 mutex_lock(&mdsc
->mutex
);
3637 for (i
= 0; i
< mdsc
->max_sessions
; i
++) {
3638 session
= __ceph_lookup_mds_session(mdsc
, i
);
3641 mutex_unlock(&mdsc
->mutex
);
3642 mutex_lock(&session
->s_mutex
);
3643 __close_session(mdsc
, session
);
3644 mutex_unlock(&session
->s_mutex
);
3645 ceph_put_mds_session(session
);
3646 mutex_lock(&mdsc
->mutex
);
3648 mutex_unlock(&mdsc
->mutex
);
3650 dout("waiting for sessions to close\n");
3651 wait_event_timeout(mdsc
->session_close_wq
, done_closing_sessions(mdsc
),
3652 ceph_timeout_jiffies(opts
->mount_timeout
));
3654 /* tear down remaining sessions */
3655 mutex_lock(&mdsc
->mutex
);
3656 for (i
= 0; i
< mdsc
->max_sessions
; i
++) {
3657 if (mdsc
->sessions
[i
]) {
3658 session
= get_session(mdsc
->sessions
[i
]);
3659 __unregister_session(mdsc
, session
);
3660 mutex_unlock(&mdsc
->mutex
);
3661 mutex_lock(&session
->s_mutex
);
3662 remove_session_caps(session
);
3663 mutex_unlock(&session
->s_mutex
);
3664 ceph_put_mds_session(session
);
3665 mutex_lock(&mdsc
->mutex
);
3668 WARN_ON(!list_empty(&mdsc
->cap_delay_list
));
3669 mutex_unlock(&mdsc
->mutex
);
3671 ceph_cleanup_empty_realms(mdsc
);
3673 cancel_delayed_work_sync(&mdsc
->delayed_work
); /* cancel timer */
3678 void ceph_mdsc_force_umount(struct ceph_mds_client
*mdsc
)
3680 struct ceph_mds_session
*session
;
3683 dout("force umount\n");
3685 mutex_lock(&mdsc
->mutex
);
3686 for (mds
= 0; mds
< mdsc
->max_sessions
; mds
++) {
3687 session
= __ceph_lookup_mds_session(mdsc
, mds
);
3690 mutex_unlock(&mdsc
->mutex
);
3691 mutex_lock(&session
->s_mutex
);
3692 __close_session(mdsc
, session
);
3693 if (session
->s_state
== CEPH_MDS_SESSION_CLOSING
) {
3694 cleanup_session_requests(mdsc
, session
);
3695 remove_session_caps(session
);
3697 mutex_unlock(&session
->s_mutex
);
3698 ceph_put_mds_session(session
);
3699 mutex_lock(&mdsc
->mutex
);
3700 kick_requests(mdsc
, mds
);
3702 __wake_requests(mdsc
, &mdsc
->waiting_for_map
);
3703 mutex_unlock(&mdsc
->mutex
);
3706 static void ceph_mdsc_stop(struct ceph_mds_client
*mdsc
)
3709 cancel_delayed_work_sync(&mdsc
->delayed_work
); /* cancel timer */
3711 ceph_mdsmap_destroy(mdsc
->mdsmap
);
3712 kfree(mdsc
->sessions
);
3713 ceph_caps_finalize(mdsc
);
3714 ceph_pool_perm_destroy(mdsc
);
3717 void ceph_mdsc_destroy(struct ceph_fs_client
*fsc
)
3719 struct ceph_mds_client
*mdsc
= fsc
->mdsc
;
3721 dout("mdsc_destroy %p\n", mdsc
);
3722 ceph_mdsc_stop(mdsc
);
3724 /* flush out any connection work with references to us */
3729 dout("mdsc_destroy %p done\n", mdsc
);
3734 * handle mds map update.
3736 void ceph_mdsc_handle_map(struct ceph_mds_client
*mdsc
, struct ceph_msg
*msg
)
3740 void *p
= msg
->front
.iov_base
;
3741 void *end
= p
+ msg
->front
.iov_len
;
3742 struct ceph_mdsmap
*newmap
, *oldmap
;
3743 struct ceph_fsid fsid
;
3746 ceph_decode_need(&p
, end
, sizeof(fsid
)+2*sizeof(u32
), bad
);
3747 ceph_decode_copy(&p
, &fsid
, sizeof(fsid
));
3748 if (ceph_check_fsid(mdsc
->fsc
->client
, &fsid
) < 0)
3750 epoch
= ceph_decode_32(&p
);
3751 maplen
= ceph_decode_32(&p
);
3752 dout("handle_map epoch %u len %d\n", epoch
, (int)maplen
);
3754 /* do we need it? */
3755 mutex_lock(&mdsc
->mutex
);
3756 if (mdsc
->mdsmap
&& epoch
<= mdsc
->mdsmap
->m_epoch
) {
3757 dout("handle_map epoch %u <= our %u\n",
3758 epoch
, mdsc
->mdsmap
->m_epoch
);
3759 mutex_unlock(&mdsc
->mutex
);
3763 newmap
= ceph_mdsmap_decode(&p
, end
);
3764 if (IS_ERR(newmap
)) {
3765 err
= PTR_ERR(newmap
);
3769 /* swap into place */
3771 oldmap
= mdsc
->mdsmap
;
3772 mdsc
->mdsmap
= newmap
;
3773 check_new_map(mdsc
, newmap
, oldmap
);
3774 ceph_mdsmap_destroy(oldmap
);
3776 mdsc
->mdsmap
= newmap
; /* first mds map */
3778 mdsc
->fsc
->sb
->s_maxbytes
= mdsc
->mdsmap
->m_max_file_size
;
3780 __wake_requests(mdsc
, &mdsc
->waiting_for_map
);
3781 ceph_monc_got_map(&mdsc
->fsc
->client
->monc
, CEPH_SUB_MDSMAP
,
3782 mdsc
->mdsmap
->m_epoch
);
3784 mutex_unlock(&mdsc
->mutex
);
3785 schedule_delayed(mdsc
);
3789 mutex_unlock(&mdsc
->mutex
);
3791 pr_err("error decoding mdsmap %d\n", err
);
3795 static struct ceph_connection
*con_get(struct ceph_connection
*con
)
3797 struct ceph_mds_session
*s
= con
->private;
3799 if (get_session(s
)) {
3800 dout("mdsc con_get %p ok (%d)\n", s
, atomic_read(&s
->s_ref
));
3803 dout("mdsc con_get %p FAIL\n", s
);
3807 static void con_put(struct ceph_connection
*con
)
3809 struct ceph_mds_session
*s
= con
->private;
3811 dout("mdsc con_put %p (%d)\n", s
, atomic_read(&s
->s_ref
) - 1);
3812 ceph_put_mds_session(s
);
3816 * if the client is unresponsive for long enough, the mds will kill
3817 * the session entirely.
3819 static void peer_reset(struct ceph_connection
*con
)
3821 struct ceph_mds_session
*s
= con
->private;
3822 struct ceph_mds_client
*mdsc
= s
->s_mdsc
;
3824 pr_warn("mds%d closed our session\n", s
->s_mds
);
3825 send_mds_reconnect(mdsc
, s
);
3828 static void dispatch(struct ceph_connection
*con
, struct ceph_msg
*msg
)
3830 struct ceph_mds_session
*s
= con
->private;
3831 struct ceph_mds_client
*mdsc
= s
->s_mdsc
;
3832 int type
= le16_to_cpu(msg
->hdr
.type
);
3834 mutex_lock(&mdsc
->mutex
);
3835 if (__verify_registered_session(mdsc
, s
) < 0) {
3836 mutex_unlock(&mdsc
->mutex
);
3839 mutex_unlock(&mdsc
->mutex
);
3842 case CEPH_MSG_MDS_MAP
:
3843 ceph_mdsc_handle_map(mdsc
, msg
);
3845 case CEPH_MSG_CLIENT_SESSION
:
3846 handle_session(s
, msg
);
3848 case CEPH_MSG_CLIENT_REPLY
:
3849 handle_reply(s
, msg
);
3851 case CEPH_MSG_CLIENT_REQUEST_FORWARD
:
3852 handle_forward(mdsc
, s
, msg
);
3854 case CEPH_MSG_CLIENT_CAPS
:
3855 ceph_handle_caps(s
, msg
);
3857 case CEPH_MSG_CLIENT_SNAP
:
3858 ceph_handle_snap(mdsc
, s
, msg
);
3860 case CEPH_MSG_CLIENT_LEASE
:
3861 handle_lease(mdsc
, s
, msg
);
3865 pr_err("received unknown message type %d %s\n", type
,
3866 ceph_msg_type_name(type
));
3877 * Note: returned pointer is the address of a structure that's
3878 * managed separately. Caller must *not* attempt to free it.
3880 static struct ceph_auth_handshake
*get_authorizer(struct ceph_connection
*con
,
3881 int *proto
, int force_new
)
3883 struct ceph_mds_session
*s
= con
->private;
3884 struct ceph_mds_client
*mdsc
= s
->s_mdsc
;
3885 struct ceph_auth_client
*ac
= mdsc
->fsc
->client
->monc
.auth
;
3886 struct ceph_auth_handshake
*auth
= &s
->s_auth
;
3888 if (force_new
&& auth
->authorizer
) {
3889 ceph_auth_destroy_authorizer(auth
->authorizer
);
3890 auth
->authorizer
= NULL
;
3892 if (!auth
->authorizer
) {
3893 int ret
= ceph_auth_create_authorizer(ac
, CEPH_ENTITY_TYPE_MDS
,
3896 return ERR_PTR(ret
);
3898 int ret
= ceph_auth_update_authorizer(ac
, CEPH_ENTITY_TYPE_MDS
,
3901 return ERR_PTR(ret
);
3903 *proto
= ac
->protocol
;
3909 static int verify_authorizer_reply(struct ceph_connection
*con
, int len
)
3911 struct ceph_mds_session
*s
= con
->private;
3912 struct ceph_mds_client
*mdsc
= s
->s_mdsc
;
3913 struct ceph_auth_client
*ac
= mdsc
->fsc
->client
->monc
.auth
;
3915 return ceph_auth_verify_authorizer_reply(ac
, s
->s_auth
.authorizer
, len
);
3918 static int invalidate_authorizer(struct ceph_connection
*con
)
3920 struct ceph_mds_session
*s
= con
->private;
3921 struct ceph_mds_client
*mdsc
= s
->s_mdsc
;
3922 struct ceph_auth_client
*ac
= mdsc
->fsc
->client
->monc
.auth
;
3924 ceph_auth_invalidate_authorizer(ac
, CEPH_ENTITY_TYPE_MDS
);
3926 return ceph_monc_validate_auth(&mdsc
->fsc
->client
->monc
);
3929 static struct ceph_msg
*mds_alloc_msg(struct ceph_connection
*con
,
3930 struct ceph_msg_header
*hdr
, int *skip
)
3932 struct ceph_msg
*msg
;
3933 int type
= (int) le16_to_cpu(hdr
->type
);
3934 int front_len
= (int) le32_to_cpu(hdr
->front_len
);
3940 msg
= ceph_msg_new(type
, front_len
, GFP_NOFS
, false);
3942 pr_err("unable to allocate msg type %d len %d\n",
3950 static int mds_sign_message(struct ceph_msg
*msg
)
3952 struct ceph_mds_session
*s
= msg
->con
->private;
3953 struct ceph_auth_handshake
*auth
= &s
->s_auth
;
3955 return ceph_auth_sign_message(auth
, msg
);
3958 static int mds_check_message_signature(struct ceph_msg
*msg
)
3960 struct ceph_mds_session
*s
= msg
->con
->private;
3961 struct ceph_auth_handshake
*auth
= &s
->s_auth
;
3963 return ceph_auth_check_message_signature(auth
, msg
);
3966 static const struct ceph_connection_operations mds_con_ops
= {
3969 .dispatch
= dispatch
,
3970 .get_authorizer
= get_authorizer
,
3971 .verify_authorizer_reply
= verify_authorizer_reply
,
3972 .invalidate_authorizer
= invalidate_authorizer
,
3973 .peer_reset
= peer_reset
,
3974 .alloc_msg
= mds_alloc_msg
,
3975 .sign_message
= mds_sign_message
,
3976 .check_message_signature
= mds_check_message_signature
,