1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
5 #include <linux/wait.h>
6 #include <linux/slab.h>
8 #include <linux/sched.h>
9 #include <linux/debugfs.h>
10 #include <linux/seq_file.h>
11 #include <linux/ratelimit.h>
12 #include <linux/bits.h>
13 #include <linux/ktime.h>
16 #include "mds_client.h"
18 #include <linux/ceph/ceph_features.h>
19 #include <linux/ceph/messenger.h>
20 #include <linux/ceph/decode.h>
21 #include <linux/ceph/pagelist.h>
22 #include <linux/ceph/auth.h>
23 #include <linux/ceph/debugfs.h>
25 #define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
28 * A cluster of MDS (metadata server) daemons is responsible for
29 * managing the file system namespace (the directory hierarchy and
30 * inodes) and for coordinating shared access to storage. Metadata is
31 * partitioning hierarchically across a number of servers, and that
32 * partition varies over time as the cluster adjusts the distribution
33 * in order to balance load.
35 * The MDS client is primarily responsible to managing synchronous
36 * metadata requests for operations like open, unlink, and so forth.
37 * If there is a MDS failure, we find out about it when we (possibly
38 * request and) receive a new MDS map, and can resubmit affected
41 * For the most part, though, we take advantage of a lossless
42 * communications channel to the MDS, and do not need to worry about
43 * timing out or resubmitting requests.
45 * We maintain a stateful "session" with each MDS we interact with.
46 * Within each session, we sent periodic heartbeat messages to ensure
47 * any capabilities or leases we have been issues remain valid. If
48 * the session times out and goes stale, our leases and capabilities
49 * are no longer valid.
52 struct ceph_reconnect_state {
53 struct ceph_mds_session *session;
54 int nr_caps, nr_realms;
55 struct ceph_pagelist *pagelist;
60 static void __wake_requests(struct ceph_mds_client *mdsc,
61 struct list_head *head);
62 static void ceph_cap_release_work(struct work_struct *work);
63 static void ceph_cap_reclaim_work(struct work_struct *work);
65 static const struct ceph_connection_operations mds_con_ops;
72 static int parse_reply_info_quota(void **p, void *end,
73 struct ceph_mds_reply_info_in *info)
75 u8 struct_v, struct_compat;
78 ceph_decode_8_safe(p, end, struct_v, bad);
79 ceph_decode_8_safe(p, end, struct_compat, bad);
80 /* struct_v is expected to be >= 1. we only
81 * understand encoding with struct_compat == 1. */
82 if (!struct_v || struct_compat != 1)
84 ceph_decode_32_safe(p, end, struct_len, bad);
85 ceph_decode_need(p, end, struct_len, bad);
86 end = *p + struct_len;
87 ceph_decode_64_safe(p, end, info->max_bytes, bad);
88 ceph_decode_64_safe(p, end, info->max_files, bad);
96 * parse individual inode info
98 static int parse_reply_info_in(void **p, void *end,
99 struct ceph_mds_reply_info_in *info,
105 if (features == (u64)-1) {
108 ceph_decode_8_safe(p, end, struct_v, bad);
109 ceph_decode_8_safe(p, end, struct_compat, bad);
110 /* struct_v is expected to be >= 1. we only understand
111 * encoding with struct_compat == 1. */
112 if (!struct_v || struct_compat != 1)
114 ceph_decode_32_safe(p, end, struct_len, bad);
115 ceph_decode_need(p, end, struct_len, bad);
116 end = *p + struct_len;
119 ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);
121 *p += sizeof(struct ceph_mds_reply_inode) +
122 sizeof(*info->in->fragtree.splits) *
123 le32_to_cpu(info->in->fragtree.nsplits);
125 ceph_decode_32_safe(p, end, info->symlink_len, bad);
126 ceph_decode_need(p, end, info->symlink_len, bad);
128 *p += info->symlink_len;
130 ceph_decode_copy_safe(p, end, &info->dir_layout,
131 sizeof(info->dir_layout), bad);
132 ceph_decode_32_safe(p, end, info->xattr_len, bad);
133 ceph_decode_need(p, end, info->xattr_len, bad);
134 info->xattr_data = *p;
135 *p += info->xattr_len;
137 if (features == (u64)-1) {
139 ceph_decode_64_safe(p, end, info->inline_version, bad);
140 ceph_decode_32_safe(p, end, info->inline_len, bad);
141 ceph_decode_need(p, end, info->inline_len, bad);
142 info->inline_data = *p;
143 *p += info->inline_len;
145 err = parse_reply_info_quota(p, end, info);
149 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
150 if (info->pool_ns_len > 0) {
151 ceph_decode_need(p, end, info->pool_ns_len, bad);
152 info->pool_ns_data = *p;
153 *p += info->pool_ns_len;
157 ceph_decode_need(p, end, sizeof(info->btime), bad);
158 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
160 /* change attribute */
161 ceph_decode_64_safe(p, end, info->change_attr, bad);
165 ceph_decode_32_safe(p, end, info->dir_pin, bad);
167 info->dir_pin = -ENODATA;
170 /* snapshot birth time, remains zero for v<=2 */
172 ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
173 ceph_decode_copy(p, &info->snap_btime,
174 sizeof(info->snap_btime));
176 memset(&info->snap_btime, 0, sizeof(info->snap_btime));
179 /* snapshot count, remains zero for v<=3 */
181 ceph_decode_64_safe(p, end, info->rsnaps, bad);
188 if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
189 ceph_decode_64_safe(p, end, info->inline_version, bad);
190 ceph_decode_32_safe(p, end, info->inline_len, bad);
191 ceph_decode_need(p, end, info->inline_len, bad);
192 info->inline_data = *p;
193 *p += info->inline_len;
195 info->inline_version = CEPH_INLINE_NONE;
197 if (features & CEPH_FEATURE_MDS_QUOTA) {
198 err = parse_reply_info_quota(p, end, info);
206 info->pool_ns_len = 0;
207 info->pool_ns_data = NULL;
208 if (features & CEPH_FEATURE_FS_FILE_LAYOUT_V2) {
209 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
210 if (info->pool_ns_len > 0) {
211 ceph_decode_need(p, end, info->pool_ns_len, bad);
212 info->pool_ns_data = *p;
213 *p += info->pool_ns_len;
217 if (features & CEPH_FEATURE_FS_BTIME) {
218 ceph_decode_need(p, end, sizeof(info->btime), bad);
219 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
220 ceph_decode_64_safe(p, end, info->change_attr, bad);
223 info->dir_pin = -ENODATA;
224 /* info->snap_btime and info->rsnaps remain zero */
233 static int parse_reply_info_dir(void **p, void *end,
234 struct ceph_mds_reply_dirfrag **dirfrag,
237 if (features == (u64)-1) {
238 u8 struct_v, struct_compat;
240 ceph_decode_8_safe(p, end, struct_v, bad);
241 ceph_decode_8_safe(p, end, struct_compat, bad);
242 /* struct_v is expected to be >= 1. we only understand
243 * encoding whose struct_compat == 1. */
244 if (!struct_v || struct_compat != 1)
246 ceph_decode_32_safe(p, end, struct_len, bad);
247 ceph_decode_need(p, end, struct_len, bad);
248 end = *p + struct_len;
251 ceph_decode_need(p, end, sizeof(**dirfrag), bad);
253 *p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);
254 if (unlikely(*p > end))
256 if (features == (u64)-1)
263 static int parse_reply_info_lease(void **p, void *end,
264 struct ceph_mds_reply_lease **lease,
267 if (features == (u64)-1) {
268 u8 struct_v, struct_compat;
270 ceph_decode_8_safe(p, end, struct_v, bad);
271 ceph_decode_8_safe(p, end, struct_compat, bad);
272 /* struct_v is expected to be >= 1. we only understand
273 * encoding whose struct_compat == 1. */
274 if (!struct_v || struct_compat != 1)
276 ceph_decode_32_safe(p, end, struct_len, bad);
277 ceph_decode_need(p, end, struct_len, bad);
278 end = *p + struct_len;
281 ceph_decode_need(p, end, sizeof(**lease), bad);
283 *p += sizeof(**lease);
284 if (features == (u64)-1)
292 * parse a normal reply, which may contain a (dir+)dentry and/or a
295 static int parse_reply_info_trace(void **p, void *end,
296 struct ceph_mds_reply_info_parsed *info,
301 if (info->head->is_dentry) {
302 err = parse_reply_info_in(p, end, &info->diri, features);
306 err = parse_reply_info_dir(p, end, &info->dirfrag, features);
310 ceph_decode_32_safe(p, end, info->dname_len, bad);
311 ceph_decode_need(p, end, info->dname_len, bad);
313 *p += info->dname_len;
315 err = parse_reply_info_lease(p, end, &info->dlease, features);
320 if (info->head->is_target) {
321 err = parse_reply_info_in(p, end, &info->targeti, features);
326 if (unlikely(*p != end))
333 pr_err("problem parsing mds trace %d\n", err);
338 * parse readdir results
340 static int parse_reply_info_readdir(void **p, void *end,
341 struct ceph_mds_reply_info_parsed *info,
347 err = parse_reply_info_dir(p, end, &info->dir_dir, features);
351 ceph_decode_need(p, end, sizeof(num) + 2, bad);
352 num = ceph_decode_32(p);
354 u16 flags = ceph_decode_16(p);
355 info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);
356 info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);
357 info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);
358 info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);
363 BUG_ON(!info->dir_entries);
364 if ((unsigned long)(info->dir_entries + num) >
365 (unsigned long)info->dir_entries + info->dir_buf_size) {
366 pr_err("dir contents are larger than expected\n");
373 struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
375 ceph_decode_32_safe(p, end, rde->name_len, bad);
376 ceph_decode_need(p, end, rde->name_len, bad);
379 dout("parsed dir dname '%.*s'\n", rde->name_len, rde->name);
382 err = parse_reply_info_lease(p, end, &rde->lease, features);
386 err = parse_reply_info_in(p, end, &rde->inode, features);
389 /* ceph_readdir_prepopulate() will update it */
396 /* Skip over any unrecognized fields */
403 pr_err("problem parsing dir contents %d\n", err);
408 * parse fcntl F_GETLK results
410 static int parse_reply_info_filelock(void **p, void *end,
411 struct ceph_mds_reply_info_parsed *info,
414 if (*p + sizeof(*info->filelock_reply) > end)
417 info->filelock_reply = *p;
419 /* Skip over any unrecognized fields */
427 #if BITS_PER_LONG == 64
429 #define DELEGATED_INO_AVAILABLE xa_mk_value(1)
431 static int ceph_parse_deleg_inos(void **p, void *end,
432 struct ceph_mds_session *s)
436 ceph_decode_32_safe(p, end, sets, bad);
437 dout("got %u sets of delegated inodes\n", sets);
441 ceph_decode_64_safe(p, end, start, bad);
442 ceph_decode_64_safe(p, end, len, bad);
444 /* Don't accept a delegation of system inodes */
445 if (start < CEPH_INO_SYSTEM_BASE) {
446 pr_warn_ratelimited("ceph: ignoring reserved inode range delegation (start=0x%llx len=0x%llx)\n",
451 int err = xa_insert(&s->s_delegated_inos, ino = start++,
452 DELEGATED_INO_AVAILABLE,
455 dout("added delegated inode 0x%llx\n",
457 } else if (err == -EBUSY) {
458 pr_warn("ceph: MDS delegated inode 0x%llx more than once.\n",
470 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
475 xa_for_each(&s->s_delegated_inos, ino, val) {
476 val = xa_erase(&s->s_delegated_inos, ino);
477 if (val == DELEGATED_INO_AVAILABLE)
483 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
485 return xa_insert(&s->s_delegated_inos, ino, DELEGATED_INO_AVAILABLE,
488 #else /* BITS_PER_LONG == 64 */
490 * FIXME: xarrays can't handle 64-bit indexes on a 32-bit arch. For now, just
491 * ignore delegated_inos on 32 bit arch. Maybe eventually add xarrays for top
494 static int ceph_parse_deleg_inos(void **p, void *end,
495 struct ceph_mds_session *s)
499 ceph_decode_32_safe(p, end, sets, bad);
501 ceph_decode_skip_n(p, end, sets * 2 * sizeof(__le64), bad);
507 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
512 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
516 #endif /* BITS_PER_LONG == 64 */
519 * parse create results
521 static int parse_reply_info_create(void **p, void *end,
522 struct ceph_mds_reply_info_parsed *info,
523 u64 features, struct ceph_mds_session *s)
527 if (features == (u64)-1 ||
528 (features & CEPH_FEATURE_REPLY_CREATE_INODE)) {
530 /* Malformed reply? */
531 info->has_create_ino = false;
532 } else if (test_bit(CEPHFS_FEATURE_DELEG_INO, &s->s_features)) {
533 info->has_create_ino = true;
534 /* struct_v, struct_compat, and len */
535 ceph_decode_skip_n(p, end, 2 + sizeof(u32), bad);
536 ceph_decode_64_safe(p, end, info->ino, bad);
537 ret = ceph_parse_deleg_inos(p, end, s);
542 ceph_decode_64_safe(p, end, info->ino, bad);
543 info->has_create_ino = true;
550 /* Skip over any unrecognized fields */
558 * parse extra results
560 static int parse_reply_info_extra(void **p, void *end,
561 struct ceph_mds_reply_info_parsed *info,
562 u64 features, struct ceph_mds_session *s)
564 u32 op = le32_to_cpu(info->head->op);
566 if (op == CEPH_MDS_OP_GETFILELOCK)
567 return parse_reply_info_filelock(p, end, info, features);
568 else if (op == CEPH_MDS_OP_READDIR || op == CEPH_MDS_OP_LSSNAP)
569 return parse_reply_info_readdir(p, end, info, features);
570 else if (op == CEPH_MDS_OP_CREATE)
571 return parse_reply_info_create(p, end, info, features, s);
577 * parse entire mds reply
579 static int parse_reply_info(struct ceph_mds_session *s, struct ceph_msg *msg,
580 struct ceph_mds_reply_info_parsed *info,
587 info->head = msg->front.iov_base;
588 p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
589 end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
592 ceph_decode_32_safe(&p, end, len, bad);
594 ceph_decode_need(&p, end, len, bad);
595 err = parse_reply_info_trace(&p, p+len, info, features);
601 ceph_decode_32_safe(&p, end, len, bad);
603 ceph_decode_need(&p, end, len, bad);
604 err = parse_reply_info_extra(&p, p+len, info, features, s);
610 ceph_decode_32_safe(&p, end, len, bad);
611 info->snapblob_len = len;
622 pr_err("mds parse_reply err %d\n", err);
626 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
628 if (!info->dir_entries)
630 free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));
637 const char *ceph_session_state_name(int s)
640 case CEPH_MDS_SESSION_NEW: return "new";
641 case CEPH_MDS_SESSION_OPENING: return "opening";
642 case CEPH_MDS_SESSION_OPEN: return "open";
643 case CEPH_MDS_SESSION_HUNG: return "hung";
644 case CEPH_MDS_SESSION_CLOSING: return "closing";
645 case CEPH_MDS_SESSION_CLOSED: return "closed";
646 case CEPH_MDS_SESSION_RESTARTING: return "restarting";
647 case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
648 case CEPH_MDS_SESSION_REJECTED: return "rejected";
649 default: return "???";
653 struct ceph_mds_session *ceph_get_mds_session(struct ceph_mds_session *s)
655 if (refcount_inc_not_zero(&s->s_ref)) {
656 dout("mdsc get_session %p %d -> %d\n", s,
657 refcount_read(&s->s_ref)-1, refcount_read(&s->s_ref));
660 dout("mdsc get_session %p 0 -- FAIL\n", s);
665 void ceph_put_mds_session(struct ceph_mds_session *s)
667 if (IS_ERR_OR_NULL(s))
670 dout("mdsc put_session %p %d -> %d\n", s,
671 refcount_read(&s->s_ref), refcount_read(&s->s_ref)-1);
672 if (refcount_dec_and_test(&s->s_ref)) {
673 if (s->s_auth.authorizer)
674 ceph_auth_destroy_authorizer(s->s_auth.authorizer);
675 WARN_ON(mutex_is_locked(&s->s_mutex));
676 xa_destroy(&s->s_delegated_inos);
682 * called under mdsc->mutex
684 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
687 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
689 return ceph_get_mds_session(mdsc->sessions[mds]);
692 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
694 if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
700 static int __verify_registered_session(struct ceph_mds_client *mdsc,
701 struct ceph_mds_session *s)
703 if (s->s_mds >= mdsc->max_sessions ||
704 mdsc->sessions[s->s_mds] != s)
710 * create+register a new session for given mds.
711 * called under mdsc->mutex.
713 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
716 struct ceph_mds_session *s;
718 if (mds >= mdsc->mdsmap->possible_max_rank)
719 return ERR_PTR(-EINVAL);
721 s = kzalloc(sizeof(*s), GFP_NOFS);
723 return ERR_PTR(-ENOMEM);
725 if (mds >= mdsc->max_sessions) {
726 int newmax = 1 << get_count_order(mds + 1);
727 struct ceph_mds_session **sa;
729 dout("%s: realloc to %d\n", __func__, newmax);
730 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
733 if (mdsc->sessions) {
734 memcpy(sa, mdsc->sessions,
735 mdsc->max_sessions * sizeof(void *));
736 kfree(mdsc->sessions);
739 mdsc->max_sessions = newmax;
742 dout("%s: mds%d\n", __func__, mds);
745 s->s_state = CEPH_MDS_SESSION_NEW;
748 mutex_init(&s->s_mutex);
750 ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
752 atomic_set(&s->s_cap_gen, 1);
753 s->s_cap_ttl = jiffies - 1;
755 spin_lock_init(&s->s_cap_lock);
756 s->s_renew_requested = 0;
758 INIT_LIST_HEAD(&s->s_caps);
760 refcount_set(&s->s_ref, 1);
761 INIT_LIST_HEAD(&s->s_waiting);
762 INIT_LIST_HEAD(&s->s_unsafe);
763 xa_init(&s->s_delegated_inos);
764 s->s_num_cap_releases = 0;
765 s->s_cap_reconnect = 0;
766 s->s_cap_iterator = NULL;
767 INIT_LIST_HEAD(&s->s_cap_releases);
768 INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);
770 INIT_LIST_HEAD(&s->s_cap_dirty);
771 INIT_LIST_HEAD(&s->s_cap_flushing);
773 mdsc->sessions[mds] = s;
774 atomic_inc(&mdsc->num_sessions);
775 refcount_inc(&s->s_ref); /* one ref to sessions[], one to caller */
777 ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
778 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
784 return ERR_PTR(-ENOMEM);
788 * called under mdsc->mutex
790 static void __unregister_session(struct ceph_mds_client *mdsc,
791 struct ceph_mds_session *s)
793 dout("__unregister_session mds%d %p\n", s->s_mds, s);
794 BUG_ON(mdsc->sessions[s->s_mds] != s);
795 mdsc->sessions[s->s_mds] = NULL;
796 ceph_con_close(&s->s_con);
797 ceph_put_mds_session(s);
798 atomic_dec(&mdsc->num_sessions);
802 * drop session refs in request.
804 * should be last request ref, or hold mdsc->mutex
806 static void put_request_session(struct ceph_mds_request *req)
808 if (req->r_session) {
809 ceph_put_mds_session(req->r_session);
810 req->r_session = NULL;
814 void ceph_mdsc_release_request(struct kref *kref)
816 struct ceph_mds_request *req = container_of(kref,
817 struct ceph_mds_request,
819 ceph_mdsc_release_dir_caps_no_check(req);
820 destroy_reply_info(&req->r_reply_info);
822 ceph_msg_put(req->r_request);
824 ceph_msg_put(req->r_reply);
826 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
830 ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
833 iput(req->r_target_inode);
836 if (req->r_old_dentry)
837 dput(req->r_old_dentry);
838 if (req->r_old_dentry_dir) {
840 * track (and drop pins for) r_old_dentry_dir
841 * separately, since r_old_dentry's d_parent may have
842 * changed between the dir mutex being dropped and
843 * this request being freed.
845 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
847 iput(req->r_old_dentry_dir);
851 put_cred(req->r_cred);
853 ceph_pagelist_release(req->r_pagelist);
854 put_request_session(req);
855 ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
856 WARN_ON_ONCE(!list_empty(&req->r_wait));
857 kmem_cache_free(ceph_mds_request_cachep, req);
860 DEFINE_RB_FUNCS(request, struct ceph_mds_request, r_tid, r_node)
863 * lookup session, bump ref if found.
865 * called under mdsc->mutex.
867 static struct ceph_mds_request *
868 lookup_get_request(struct ceph_mds_client *mdsc, u64 tid)
870 struct ceph_mds_request *req;
872 req = lookup_request(&mdsc->request_tree, tid);
874 ceph_mdsc_get_request(req);
880 * Register an in-flight request, and assign a tid. Link to directory
881 * are modifying (if any).
883 * Called under mdsc->mutex.
885 static void __register_request(struct ceph_mds_client *mdsc,
886 struct ceph_mds_request *req,
891 req->r_tid = ++mdsc->last_tid;
892 if (req->r_num_caps) {
893 ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,
896 pr_err("__register_request %p "
897 "failed to reserve caps: %d\n", req, ret);
898 /* set req->r_err to fail early from __do_request */
903 dout("__register_request %p tid %lld\n", req, req->r_tid);
904 ceph_mdsc_get_request(req);
905 insert_request(&mdsc->request_tree, req);
907 req->r_cred = get_current_cred();
909 if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
910 mdsc->oldest_tid = req->r_tid;
913 struct ceph_inode_info *ci = ceph_inode(dir);
916 req->r_unsafe_dir = dir;
917 spin_lock(&ci->i_unsafe_lock);
918 list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
919 spin_unlock(&ci->i_unsafe_lock);
923 static void __unregister_request(struct ceph_mds_client *mdsc,
924 struct ceph_mds_request *req)
926 dout("__unregister_request %p tid %lld\n", req, req->r_tid);
928 /* Never leave an unregistered request on an unsafe list! */
929 list_del_init(&req->r_unsafe_item);
931 if (req->r_tid == mdsc->oldest_tid) {
932 struct rb_node *p = rb_next(&req->r_node);
933 mdsc->oldest_tid = 0;
935 struct ceph_mds_request *next_req =
936 rb_entry(p, struct ceph_mds_request, r_node);
937 if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
938 mdsc->oldest_tid = next_req->r_tid;
945 erase_request(&mdsc->request_tree, req);
947 if (req->r_unsafe_dir) {
948 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
949 spin_lock(&ci->i_unsafe_lock);
950 list_del_init(&req->r_unsafe_dir_item);
951 spin_unlock(&ci->i_unsafe_lock);
953 if (req->r_target_inode &&
954 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
955 struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
956 spin_lock(&ci->i_unsafe_lock);
957 list_del_init(&req->r_unsafe_target_item);
958 spin_unlock(&ci->i_unsafe_lock);
961 if (req->r_unsafe_dir) {
962 iput(req->r_unsafe_dir);
963 req->r_unsafe_dir = NULL;
966 complete_all(&req->r_safe_completion);
968 ceph_mdsc_put_request(req);
972 * Walk back up the dentry tree until we hit a dentry representing a
973 * non-snapshot inode. We do this using the rcu_read_lock (which must be held
974 * when calling this) to ensure that the objects won't disappear while we're
975 * working with them. Once we hit a candidate dentry, we attempt to take a
976 * reference to it, and return that as the result.
978 static struct inode *get_nonsnap_parent(struct dentry *dentry)
980 struct inode *inode = NULL;
982 while (dentry && !IS_ROOT(dentry)) {
983 inode = d_inode_rcu(dentry);
984 if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
986 dentry = dentry->d_parent;
989 inode = igrab(inode);
994 * Choose mds to send request to next. If there is a hint set in the
995 * request (e.g., due to a prior forward hint from the mds), use that.
996 * Otherwise, consult frag tree and/or caps to identify the
997 * appropriate mds. If all else fails, choose randomly.
999 * Called under mdsc->mutex.
1001 static int __choose_mds(struct ceph_mds_client *mdsc,
1002 struct ceph_mds_request *req,
1005 struct inode *inode;
1006 struct ceph_inode_info *ci;
1007 struct ceph_cap *cap;
1008 int mode = req->r_direct_mode;
1010 u32 hash = req->r_direct_hash;
1011 bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
1017 * is there a specific mds we should try? ignore hint if we have
1018 * no session and the mds is not up (active or recovering).
1020 if (req->r_resend_mds >= 0 &&
1021 (__have_session(mdsc, req->r_resend_mds) ||
1022 ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
1023 dout("%s using resend_mds mds%d\n", __func__,
1025 return req->r_resend_mds;
1028 if (mode == USE_RANDOM_MDS)
1033 if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {
1034 inode = req->r_inode;
1037 /* req->r_dentry is non-null for LSSNAP request */
1039 inode = get_nonsnap_parent(req->r_dentry);
1041 dout("%s using snapdir's parent %p\n", __func__, inode);
1043 } else if (req->r_dentry) {
1044 /* ignore race with rename; old or new d_parent is okay */
1045 struct dentry *parent;
1049 parent = READ_ONCE(req->r_dentry->d_parent);
1050 dir = req->r_parent ? : d_inode_rcu(parent);
1052 if (!dir || dir->i_sb != mdsc->fsc->sb) {
1053 /* not this fs or parent went negative */
1054 inode = d_inode(req->r_dentry);
1057 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
1058 /* direct snapped/virtual snapdir requests
1059 * based on parent dir inode */
1060 inode = get_nonsnap_parent(parent);
1061 dout("%s using nonsnap parent %p\n", __func__, inode);
1064 inode = d_inode(req->r_dentry);
1065 if (!inode || mode == USE_AUTH_MDS) {
1068 hash = ceph_dentry_hash(dir, req->r_dentry);
1077 dout("%s %p is_hash=%d (0x%x) mode %d\n", __func__, inode, (int)is_hash,
1081 ci = ceph_inode(inode);
1083 if (is_hash && S_ISDIR(inode->i_mode)) {
1084 struct ceph_inode_frag frag;
1087 ceph_choose_frag(ci, hash, &frag, &found);
1089 if (mode == USE_ANY_MDS && frag.ndist > 0) {
1092 /* choose a random replica */
1093 get_random_bytes(&r, 1);
1096 dout("%s %p %llx.%llx frag %u mds%d (%d/%d)\n",
1097 __func__, inode, ceph_vinop(inode),
1098 frag.frag, mds, (int)r, frag.ndist);
1099 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1100 CEPH_MDS_STATE_ACTIVE &&
1101 !ceph_mdsmap_is_laggy(mdsc->mdsmap, mds))
1105 /* since this file/dir wasn't known to be
1106 * replicated, then we want to look for the
1107 * authoritative mds. */
1108 if (frag.mds >= 0) {
1109 /* choose auth mds */
1111 dout("%s %p %llx.%llx frag %u mds%d (auth)\n",
1112 __func__, inode, ceph_vinop(inode),
1114 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1115 CEPH_MDS_STATE_ACTIVE) {
1116 if (!ceph_mdsmap_is_laggy(mdsc->mdsmap,
1121 mode = USE_AUTH_MDS;
1125 spin_lock(&ci->i_ceph_lock);
1127 if (mode == USE_AUTH_MDS)
1128 cap = ci->i_auth_cap;
1129 if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1130 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1132 spin_unlock(&ci->i_ceph_lock);
1136 mds = cap->session->s_mds;
1137 dout("%s %p %llx.%llx mds%d (%scap %p)\n", __func__,
1138 inode, ceph_vinop(inode), mds,
1139 cap == ci->i_auth_cap ? "auth " : "", cap);
1140 spin_unlock(&ci->i_ceph_lock);
1149 mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
1150 dout("%s chose random mds%d\n", __func__, mds);
1158 static struct ceph_msg *create_session_msg(u32 op, u64 seq)
1160 struct ceph_msg *msg;
1161 struct ceph_mds_session_head *h;
1163 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1166 pr_err("create_session_msg ENOMEM creating msg\n");
1169 h = msg->front.iov_base;
1170 h->op = cpu_to_le32(op);
1171 h->seq = cpu_to_le64(seq);
1176 static const unsigned char feature_bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1177 #define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
1178 static int encode_supported_features(void **p, void *end)
1180 static const size_t count = ARRAY_SIZE(feature_bits);
1184 size_t size = FEATURE_BYTES(count);
1186 if (WARN_ON_ONCE(*p + 4 + size > end))
1189 ceph_encode_32(p, size);
1190 memset(*p, 0, size);
1191 for (i = 0; i < count; i++)
1192 ((unsigned char*)(*p))[i / 8] |= BIT(feature_bits[i] % 8);
1195 if (WARN_ON_ONCE(*p + 4 > end))
1198 ceph_encode_32(p, 0);
1204 static const unsigned char metric_bits[] = CEPHFS_METRIC_SPEC_CLIENT_SUPPORTED;
1205 #define METRIC_BYTES(cnt) (DIV_ROUND_UP((size_t)metric_bits[cnt - 1] + 1, 64) * 8)
1206 static int encode_metric_spec(void **p, void *end)
1208 static const size_t count = ARRAY_SIZE(metric_bits);
1211 if (WARN_ON_ONCE(*p + 2 > end))
1214 ceph_encode_8(p, 1); /* version */
1215 ceph_encode_8(p, 1); /* compat */
1219 size_t size = METRIC_BYTES(count);
1221 if (WARN_ON_ONCE(*p + 4 + 4 + size > end))
1224 /* metric spec info length */
1225 ceph_encode_32(p, 4 + size);
1228 ceph_encode_32(p, size);
1229 memset(*p, 0, size);
1230 for (i = 0; i < count; i++)
1231 ((unsigned char *)(*p))[i / 8] |= BIT(metric_bits[i] % 8);
1234 if (WARN_ON_ONCE(*p + 4 + 4 > end))
1237 /* metric spec info length */
1238 ceph_encode_32(p, 4);
1240 ceph_encode_32(p, 0);
1247 * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1248 * to include additional client metadata fields.
1250 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1252 struct ceph_msg *msg;
1253 struct ceph_mds_session_head *h;
1255 int extra_bytes = 0;
1256 int metadata_key_count = 0;
1257 struct ceph_options *opt = mdsc->fsc->client->options;
1258 struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
1263 const char* metadata[][2] = {
1264 {"hostname", mdsc->nodename},
1265 {"kernel_version", init_utsname()->release},
1266 {"entity_id", opt->name ? : ""},
1267 {"root", fsopt->server_path ? : "/"},
1271 /* Calculate serialized length of metadata */
1272 extra_bytes = 4; /* map length */
1273 for (i = 0; metadata[i][0]; ++i) {
1274 extra_bytes += 8 + strlen(metadata[i][0]) +
1275 strlen(metadata[i][1]);
1276 metadata_key_count++;
1279 /* supported feature */
1281 count = ARRAY_SIZE(feature_bits);
1283 size = FEATURE_BYTES(count);
1284 extra_bytes += 4 + size;
1288 count = ARRAY_SIZE(metric_bits);
1290 size = METRIC_BYTES(count);
1291 extra_bytes += 2 + 4 + 4 + size;
1293 /* Allocate the message */
1294 msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
1297 pr_err("create_session_msg ENOMEM creating msg\n");
1298 return ERR_PTR(-ENOMEM);
1300 p = msg->front.iov_base;
1301 end = p + msg->front.iov_len;
1304 h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1305 h->seq = cpu_to_le64(seq);
1308 * Serialize client metadata into waiting buffer space, using
1309 * the format that userspace expects for map<string, string>
1311 * ClientSession messages with metadata are v4
1313 msg->hdr.version = cpu_to_le16(4);
1314 msg->hdr.compat_version = cpu_to_le16(1);
1316 /* The write pointer, following the session_head structure */
1319 /* Number of entries in the map */
1320 ceph_encode_32(&p, metadata_key_count);
1322 /* Two length-prefixed strings for each entry in the map */
1323 for (i = 0; metadata[i][0]; ++i) {
1324 size_t const key_len = strlen(metadata[i][0]);
1325 size_t const val_len = strlen(metadata[i][1]);
1327 ceph_encode_32(&p, key_len);
1328 memcpy(p, metadata[i][0], key_len);
1330 ceph_encode_32(&p, val_len);
1331 memcpy(p, metadata[i][1], val_len);
1335 ret = encode_supported_features(&p, end);
1337 pr_err("encode_supported_features failed!\n");
1339 return ERR_PTR(ret);
1342 ret = encode_metric_spec(&p, end);
1344 pr_err("encode_metric_spec failed!\n");
1346 return ERR_PTR(ret);
1349 msg->front.iov_len = p - msg->front.iov_base;
1350 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1356 * send session open request.
1358 * called under mdsc->mutex
1360 static int __open_session(struct ceph_mds_client *mdsc,
1361 struct ceph_mds_session *session)
1363 struct ceph_msg *msg;
1365 int mds = session->s_mds;
1367 /* wait for mds to go active? */
1368 mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1369 dout("open_session to mds%d (%s)\n", mds,
1370 ceph_mds_state_name(mstate));
1371 session->s_state = CEPH_MDS_SESSION_OPENING;
1372 session->s_renew_requested = jiffies;
1374 /* send connect message */
1375 msg = create_session_open_msg(mdsc, session->s_seq);
1377 return PTR_ERR(msg);
1378 ceph_con_send(&session->s_con, msg);
1383 * open sessions for any export targets for the given mds
1385 * called under mdsc->mutex
1387 static struct ceph_mds_session *
1388 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
1390 struct ceph_mds_session *session;
1393 session = __ceph_lookup_mds_session(mdsc, target);
1395 session = register_session(mdsc, target);
1396 if (IS_ERR(session))
1399 if (session->s_state == CEPH_MDS_SESSION_NEW ||
1400 session->s_state == CEPH_MDS_SESSION_CLOSING) {
1401 ret = __open_session(mdsc, session);
1403 return ERR_PTR(ret);
1409 struct ceph_mds_session *
1410 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1412 struct ceph_mds_session *session;
1414 dout("open_export_target_session to mds%d\n", target);
1416 mutex_lock(&mdsc->mutex);
1417 session = __open_export_target_session(mdsc, target);
1418 mutex_unlock(&mdsc->mutex);
1423 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1424 struct ceph_mds_session *session)
1426 struct ceph_mds_info *mi;
1427 struct ceph_mds_session *ts;
1428 int i, mds = session->s_mds;
1430 if (mds >= mdsc->mdsmap->possible_max_rank)
1433 mi = &mdsc->mdsmap->m_info[mds];
1434 dout("open_export_target_sessions for mds%d (%d targets)\n",
1435 session->s_mds, mi->num_export_targets);
1437 for (i = 0; i < mi->num_export_targets; i++) {
1438 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1439 ceph_put_mds_session(ts);
1443 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1444 struct ceph_mds_session *session)
1446 mutex_lock(&mdsc->mutex);
1447 __open_export_target_sessions(mdsc, session);
1448 mutex_unlock(&mdsc->mutex);
1455 static void detach_cap_releases(struct ceph_mds_session *session,
1456 struct list_head *target)
1458 lockdep_assert_held(&session->s_cap_lock);
1460 list_splice_init(&session->s_cap_releases, target);
1461 session->s_num_cap_releases = 0;
1462 dout("dispose_cap_releases mds%d\n", session->s_mds);
1465 static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1466 struct list_head *dispose)
1468 while (!list_empty(dispose)) {
1469 struct ceph_cap *cap;
1470 /* zero out the in-progress message */
1471 cap = list_first_entry(dispose, struct ceph_cap, session_caps);
1472 list_del(&cap->session_caps);
1473 ceph_put_cap(mdsc, cap);
1477 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1478 struct ceph_mds_session *session)
1480 struct ceph_mds_request *req;
1482 struct ceph_inode_info *ci;
1484 dout("cleanup_session_requests mds%d\n", session->s_mds);
1485 mutex_lock(&mdsc->mutex);
1486 while (!list_empty(&session->s_unsafe)) {
1487 req = list_first_entry(&session->s_unsafe,
1488 struct ceph_mds_request, r_unsafe_item);
1489 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1491 if (req->r_target_inode) {
1492 /* dropping unsafe change of inode's attributes */
1493 ci = ceph_inode(req->r_target_inode);
1494 errseq_set(&ci->i_meta_err, -EIO);
1496 if (req->r_unsafe_dir) {
1497 /* dropping unsafe directory operation */
1498 ci = ceph_inode(req->r_unsafe_dir);
1499 errseq_set(&ci->i_meta_err, -EIO);
1501 __unregister_request(mdsc, req);
1503 /* zero r_attempts, so kick_requests() will re-send requests */
1504 p = rb_first(&mdsc->request_tree);
1506 req = rb_entry(p, struct ceph_mds_request, r_node);
1508 if (req->r_session &&
1509 req->r_session->s_mds == session->s_mds)
1510 req->r_attempts = 0;
1512 mutex_unlock(&mdsc->mutex);
1516 * Helper to safely iterate over all caps associated with a session, with
1517 * special care taken to handle a racing __ceph_remove_cap().
1519 * Caller must hold session s_mutex.
1521 int ceph_iterate_session_caps(struct ceph_mds_session *session,
1522 int (*cb)(struct inode *, struct ceph_cap *,
1525 struct list_head *p;
1526 struct ceph_cap *cap;
1527 struct inode *inode, *last_inode = NULL;
1528 struct ceph_cap *old_cap = NULL;
1531 dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1532 spin_lock(&session->s_cap_lock);
1533 p = session->s_caps.next;
1534 while (p != &session->s_caps) {
1535 cap = list_entry(p, struct ceph_cap, session_caps);
1536 inode = igrab(&cap->ci->vfs_inode);
1541 session->s_cap_iterator = cap;
1542 spin_unlock(&session->s_cap_lock);
1549 ceph_put_cap(session->s_mdsc, old_cap);
1553 ret = cb(inode, cap, arg);
1556 spin_lock(&session->s_cap_lock);
1559 dout("iterate_session_caps finishing cap %p removal\n",
1561 BUG_ON(cap->session != session);
1562 cap->session = NULL;
1563 list_del_init(&cap->session_caps);
1564 session->s_nr_caps--;
1565 atomic64_dec(&session->s_mdsc->metric.total_caps);
1566 if (cap->queue_release)
1567 __ceph_queue_cap_release(session, cap);
1569 old_cap = cap; /* put_cap it w/o locks held */
1576 session->s_cap_iterator = NULL;
1577 spin_unlock(&session->s_cap_lock);
1581 ceph_put_cap(session->s_mdsc, old_cap);
1586 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1589 struct ceph_fs_client *fsc = (struct ceph_fs_client *)arg;
1590 struct ceph_inode_info *ci = ceph_inode(inode);
1591 LIST_HEAD(to_remove);
1592 bool dirty_dropped = false;
1593 bool invalidate = false;
1595 dout("removing cap %p, ci is %p, inode is %p\n",
1596 cap, ci, &ci->vfs_inode);
1597 spin_lock(&ci->i_ceph_lock);
1598 __ceph_remove_cap(cap, false);
1599 if (!ci->i_auth_cap) {
1600 struct ceph_cap_flush *cf;
1601 struct ceph_mds_client *mdsc = fsc->mdsc;
1603 if (READ_ONCE(fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN) {
1604 if (inode->i_data.nrpages > 0)
1606 if (ci->i_wrbuffer_ref > 0)
1607 mapping_set_error(&inode->i_data, -EIO);
1610 while (!list_empty(&ci->i_cap_flush_list)) {
1611 cf = list_first_entry(&ci->i_cap_flush_list,
1612 struct ceph_cap_flush, i_list);
1613 list_move(&cf->i_list, &to_remove);
1616 spin_lock(&mdsc->cap_dirty_lock);
1618 list_for_each_entry(cf, &to_remove, i_list)
1619 list_del(&cf->g_list);
1621 if (!list_empty(&ci->i_dirty_item)) {
1622 pr_warn_ratelimited(
1623 " dropping dirty %s state for %p %lld\n",
1624 ceph_cap_string(ci->i_dirty_caps),
1625 inode, ceph_ino(inode));
1626 ci->i_dirty_caps = 0;
1627 list_del_init(&ci->i_dirty_item);
1628 dirty_dropped = true;
1630 if (!list_empty(&ci->i_flushing_item)) {
1631 pr_warn_ratelimited(
1632 " dropping dirty+flushing %s state for %p %lld\n",
1633 ceph_cap_string(ci->i_flushing_caps),
1634 inode, ceph_ino(inode));
1635 ci->i_flushing_caps = 0;
1636 list_del_init(&ci->i_flushing_item);
1637 mdsc->num_cap_flushing--;
1638 dirty_dropped = true;
1640 spin_unlock(&mdsc->cap_dirty_lock);
1642 if (dirty_dropped) {
1643 errseq_set(&ci->i_meta_err, -EIO);
1645 if (ci->i_wrbuffer_ref_head == 0 &&
1646 ci->i_wr_ref == 0 &&
1647 ci->i_dirty_caps == 0 &&
1648 ci->i_flushing_caps == 0) {
1649 ceph_put_snap_context(ci->i_head_snapc);
1650 ci->i_head_snapc = NULL;
1654 if (atomic_read(&ci->i_filelock_ref) > 0) {
1655 /* make further file lock syscall return -EIO */
1656 ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;
1657 pr_warn_ratelimited(" dropping file locks for %p %lld\n",
1658 inode, ceph_ino(inode));
1661 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
1662 list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove);
1663 ci->i_prealloc_cap_flush = NULL;
1666 spin_unlock(&ci->i_ceph_lock);
1667 while (!list_empty(&to_remove)) {
1668 struct ceph_cap_flush *cf;
1669 cf = list_first_entry(&to_remove,
1670 struct ceph_cap_flush, i_list);
1671 list_del(&cf->i_list);
1672 ceph_free_cap_flush(cf);
1675 wake_up_all(&ci->i_cap_wq);
1677 ceph_queue_invalidate(inode);
1684 * caller must hold session s_mutex
1686 static void remove_session_caps(struct ceph_mds_session *session)
1688 struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1689 struct super_block *sb = fsc->sb;
1692 dout("remove_session_caps on %p\n", session);
1693 ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
1695 wake_up_all(&fsc->mdsc->cap_flushing_wq);
1697 spin_lock(&session->s_cap_lock);
1698 if (session->s_nr_caps > 0) {
1699 struct inode *inode;
1700 struct ceph_cap *cap, *prev = NULL;
1701 struct ceph_vino vino;
1703 * iterate_session_caps() skips inodes that are being
1704 * deleted, we need to wait until deletions are complete.
1705 * __wait_on_freeing_inode() is designed for the job,
1706 * but it is not exported, so use lookup inode function
1709 while (!list_empty(&session->s_caps)) {
1710 cap = list_entry(session->s_caps.next,
1711 struct ceph_cap, session_caps);
1715 vino = cap->ci->i_vino;
1716 spin_unlock(&session->s_cap_lock);
1718 inode = ceph_find_inode(sb, vino);
1721 spin_lock(&session->s_cap_lock);
1725 // drop cap expires and unlock s_cap_lock
1726 detach_cap_releases(session, &dispose);
1728 BUG_ON(session->s_nr_caps > 0);
1729 BUG_ON(!list_empty(&session->s_cap_flushing));
1730 spin_unlock(&session->s_cap_lock);
1731 dispose_cap_releases(session->s_mdsc, &dispose);
1741 * wake up any threads waiting on this session's caps. if the cap is
1742 * old (didn't get renewed on the client reconnect), remove it now.
1744 * caller must hold s_mutex.
1746 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1749 struct ceph_inode_info *ci = ceph_inode(inode);
1750 unsigned long ev = (unsigned long)arg;
1752 if (ev == RECONNECT) {
1753 spin_lock(&ci->i_ceph_lock);
1754 ci->i_wanted_max_size = 0;
1755 ci->i_requested_max_size = 0;
1756 spin_unlock(&ci->i_ceph_lock);
1757 } else if (ev == RENEWCAPS) {
1758 if (cap->cap_gen < atomic_read(&cap->session->s_cap_gen)) {
1759 /* mds did not re-issue stale cap */
1760 spin_lock(&ci->i_ceph_lock);
1761 cap->issued = cap->implemented = CEPH_CAP_PIN;
1762 spin_unlock(&ci->i_ceph_lock);
1764 } else if (ev == FORCE_RO) {
1766 wake_up_all(&ci->i_cap_wq);
1770 static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
1772 dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1773 ceph_iterate_session_caps(session, wake_up_session_cb,
1774 (void *)(unsigned long)ev);
1778 * Send periodic message to MDS renewing all currently held caps. The
1779 * ack will reset the expiration for all caps from this session.
1781 * caller holds s_mutex
1783 static int send_renew_caps(struct ceph_mds_client *mdsc,
1784 struct ceph_mds_session *session)
1786 struct ceph_msg *msg;
1789 if (time_after_eq(jiffies, session->s_cap_ttl) &&
1790 time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1791 pr_info("mds%d caps stale\n", session->s_mds);
1792 session->s_renew_requested = jiffies;
1794 /* do not try to renew caps until a recovering mds has reconnected
1795 * with its clients. */
1796 state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1797 if (state < CEPH_MDS_STATE_RECONNECT) {
1798 dout("send_renew_caps ignoring mds%d (%s)\n",
1799 session->s_mds, ceph_mds_state_name(state));
1803 dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1804 ceph_mds_state_name(state));
1805 msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1806 ++session->s_renew_seq);
1809 ceph_con_send(&session->s_con, msg);
1813 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1814 struct ceph_mds_session *session, u64 seq)
1816 struct ceph_msg *msg;
1818 dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1819 session->s_mds, ceph_session_state_name(session->s_state), seq);
1820 msg = create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1823 ceph_con_send(&session->s_con, msg);
1829 * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1831 * Called under session->s_mutex
1833 static void renewed_caps(struct ceph_mds_client *mdsc,
1834 struct ceph_mds_session *session, int is_renew)
1839 spin_lock(&session->s_cap_lock);
1840 was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1842 session->s_cap_ttl = session->s_renew_requested +
1843 mdsc->mdsmap->m_session_timeout*HZ;
1846 if (time_before(jiffies, session->s_cap_ttl)) {
1847 pr_info("mds%d caps renewed\n", session->s_mds);
1850 pr_info("mds%d caps still stale\n", session->s_mds);
1853 dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1854 session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1855 time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1856 spin_unlock(&session->s_cap_lock);
1859 wake_up_session_caps(session, RENEWCAPS);
1863 * send a session close request
1865 static int request_close_session(struct ceph_mds_session *session)
1867 struct ceph_msg *msg;
1869 dout("request_close_session mds%d state %s seq %lld\n",
1870 session->s_mds, ceph_session_state_name(session->s_state),
1872 msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);
1875 ceph_con_send(&session->s_con, msg);
1880 * Called with s_mutex held.
1882 static int __close_session(struct ceph_mds_client *mdsc,
1883 struct ceph_mds_session *session)
1885 if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1887 session->s_state = CEPH_MDS_SESSION_CLOSING;
1888 return request_close_session(session);
1891 static bool drop_negative_children(struct dentry *dentry)
1893 struct dentry *child;
1894 bool all_negative = true;
1896 if (!d_is_dir(dentry))
1899 spin_lock(&dentry->d_lock);
1900 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
1901 if (d_really_is_positive(child)) {
1902 all_negative = false;
1906 spin_unlock(&dentry->d_lock);
1909 shrink_dcache_parent(dentry);
1911 return all_negative;
1915 * Trim old(er) caps.
1917 * Because we can't cache an inode without one or more caps, we do
1918 * this indirectly: if a cap is unused, we prune its aliases, at which
1919 * point the inode will hopefully get dropped to.
1921 * Yes, this is a bit sloppy. Our only real goal here is to respond to
1922 * memory pressure from the MDS, though, so it needn't be perfect.
1924 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1926 int *remaining = arg;
1927 struct ceph_inode_info *ci = ceph_inode(inode);
1928 int used, wanted, oissued, mine;
1930 if (*remaining <= 0)
1933 spin_lock(&ci->i_ceph_lock);
1934 mine = cap->issued | cap->implemented;
1935 used = __ceph_caps_used(ci);
1936 wanted = __ceph_caps_file_wanted(ci);
1937 oissued = __ceph_caps_issued_other(ci, cap);
1939 dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1940 inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1941 ceph_cap_string(used), ceph_cap_string(wanted));
1942 if (cap == ci->i_auth_cap) {
1943 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1944 !list_empty(&ci->i_cap_snaps))
1946 if ((used | wanted) & CEPH_CAP_ANY_WR)
1948 /* Note: it's possible that i_filelock_ref becomes non-zero
1949 * after dropping auth caps. It doesn't hurt because reply
1950 * of lock mds request will re-add auth caps. */
1951 if (atomic_read(&ci->i_filelock_ref) > 0)
1954 /* The inode has cached pages, but it's no longer used.
1955 * we can safely drop it */
1956 if (S_ISREG(inode->i_mode) &&
1957 wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
1958 !(oissued & CEPH_CAP_FILE_CACHE)) {
1962 if ((used | wanted) & ~oissued & mine)
1963 goto out; /* we need these caps */
1966 /* we aren't the only cap.. just remove us */
1967 __ceph_remove_cap(cap, true);
1970 struct dentry *dentry;
1971 /* try dropping referring dentries */
1972 spin_unlock(&ci->i_ceph_lock);
1973 dentry = d_find_any_alias(inode);
1974 if (dentry && drop_negative_children(dentry)) {
1977 d_prune_aliases(inode);
1978 count = atomic_read(&inode->i_count);
1981 dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1990 spin_unlock(&ci->i_ceph_lock);
1995 * Trim session cap count down to some max number.
1997 int ceph_trim_caps(struct ceph_mds_client *mdsc,
1998 struct ceph_mds_session *session,
2001 int trim_caps = session->s_nr_caps - max_caps;
2003 dout("trim_caps mds%d start: %d / %d, trim %d\n",
2004 session->s_mds, session->s_nr_caps, max_caps, trim_caps);
2005 if (trim_caps > 0) {
2006 int remaining = trim_caps;
2008 ceph_iterate_session_caps(session, trim_caps_cb, &remaining);
2009 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
2010 session->s_mds, session->s_nr_caps, max_caps,
2011 trim_caps - remaining);
2014 ceph_flush_cap_releases(mdsc, session);
2018 static int check_caps_flush(struct ceph_mds_client *mdsc,
2023 spin_lock(&mdsc->cap_dirty_lock);
2024 if (!list_empty(&mdsc->cap_flush_list)) {
2025 struct ceph_cap_flush *cf =
2026 list_first_entry(&mdsc->cap_flush_list,
2027 struct ceph_cap_flush, g_list);
2028 if (cf->tid <= want_flush_tid) {
2029 dout("check_caps_flush still flushing tid "
2030 "%llu <= %llu\n", cf->tid, want_flush_tid);
2034 spin_unlock(&mdsc->cap_dirty_lock);
2039 * flush all dirty inode data to disk.
2041 * returns true if we've flushed through want_flush_tid
2043 static void wait_caps_flush(struct ceph_mds_client *mdsc,
2046 dout("check_caps_flush want %llu\n", want_flush_tid);
2048 wait_event(mdsc->cap_flushing_wq,
2049 check_caps_flush(mdsc, want_flush_tid));
2051 dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
2055 * called under s_mutex
2057 static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
2058 struct ceph_mds_session *session)
2060 struct ceph_msg *msg = NULL;
2061 struct ceph_mds_cap_release *head;
2062 struct ceph_mds_cap_item *item;
2063 struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
2064 struct ceph_cap *cap;
2065 LIST_HEAD(tmp_list);
2066 int num_cap_releases;
2067 __le32 barrier, *cap_barrier;
2069 down_read(&osdc->lock);
2070 barrier = cpu_to_le32(osdc->epoch_barrier);
2071 up_read(&osdc->lock);
2073 spin_lock(&session->s_cap_lock);
2075 list_splice_init(&session->s_cap_releases, &tmp_list);
2076 num_cap_releases = session->s_num_cap_releases;
2077 session->s_num_cap_releases = 0;
2078 spin_unlock(&session->s_cap_lock);
2080 while (!list_empty(&tmp_list)) {
2082 msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
2083 PAGE_SIZE, GFP_NOFS, false);
2086 head = msg->front.iov_base;
2087 head->num = cpu_to_le32(0);
2088 msg->front.iov_len = sizeof(*head);
2090 msg->hdr.version = cpu_to_le16(2);
2091 msg->hdr.compat_version = cpu_to_le16(1);
2094 cap = list_first_entry(&tmp_list, struct ceph_cap,
2096 list_del(&cap->session_caps);
2099 head = msg->front.iov_base;
2100 put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
2102 item = msg->front.iov_base + msg->front.iov_len;
2103 item->ino = cpu_to_le64(cap->cap_ino);
2104 item->cap_id = cpu_to_le64(cap->cap_id);
2105 item->migrate_seq = cpu_to_le32(cap->mseq);
2106 item->seq = cpu_to_le32(cap->issue_seq);
2107 msg->front.iov_len += sizeof(*item);
2109 ceph_put_cap(mdsc, cap);
2111 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
2112 // Append cap_barrier field
2113 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2114 *cap_barrier = barrier;
2115 msg->front.iov_len += sizeof(*cap_barrier);
2117 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2118 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2119 ceph_con_send(&session->s_con, msg);
2124 BUG_ON(num_cap_releases != 0);
2126 spin_lock(&session->s_cap_lock);
2127 if (!list_empty(&session->s_cap_releases))
2129 spin_unlock(&session->s_cap_lock);
2132 // Append cap_barrier field
2133 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2134 *cap_barrier = barrier;
2135 msg->front.iov_len += sizeof(*cap_barrier);
2137 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2138 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2139 ceph_con_send(&session->s_con, msg);
2143 pr_err("send_cap_releases mds%d, failed to allocate message\n",
2145 spin_lock(&session->s_cap_lock);
2146 list_splice(&tmp_list, &session->s_cap_releases);
2147 session->s_num_cap_releases += num_cap_releases;
2148 spin_unlock(&session->s_cap_lock);
2151 static void ceph_cap_release_work(struct work_struct *work)
2153 struct ceph_mds_session *session =
2154 container_of(work, struct ceph_mds_session, s_cap_release_work);
2156 mutex_lock(&session->s_mutex);
2157 if (session->s_state == CEPH_MDS_SESSION_OPEN ||
2158 session->s_state == CEPH_MDS_SESSION_HUNG)
2159 ceph_send_cap_releases(session->s_mdsc, session);
2160 mutex_unlock(&session->s_mutex);
2161 ceph_put_mds_session(session);
2164 void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
2165 struct ceph_mds_session *session)
2170 ceph_get_mds_session(session);
2171 if (queue_work(mdsc->fsc->cap_wq,
2172 &session->s_cap_release_work)) {
2173 dout("cap release work queued\n");
2175 ceph_put_mds_session(session);
2176 dout("failed to queue cap release work\n");
2181 * caller holds session->s_cap_lock
2183 void __ceph_queue_cap_release(struct ceph_mds_session *session,
2184 struct ceph_cap *cap)
2186 list_add_tail(&cap->session_caps, &session->s_cap_releases);
2187 session->s_num_cap_releases++;
2189 if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
2190 ceph_flush_cap_releases(session->s_mdsc, session);
2193 static void ceph_cap_reclaim_work(struct work_struct *work)
2195 struct ceph_mds_client *mdsc =
2196 container_of(work, struct ceph_mds_client, cap_reclaim_work);
2197 int ret = ceph_trim_dentries(mdsc);
2199 ceph_queue_cap_reclaim_work(mdsc);
2202 void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
2207 if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
2208 dout("caps reclaim work queued\n");
2210 dout("failed to queue caps release work\n");
2214 void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
2219 val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
2220 if ((val % CEPH_CAPS_PER_RELEASE) < nr) {
2221 atomic_set(&mdsc->cap_reclaim_pending, 0);
2222 ceph_queue_cap_reclaim_work(mdsc);
2230 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2233 struct ceph_inode_info *ci = ceph_inode(dir);
2234 struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2235 struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2236 size_t size = sizeof(struct ceph_mds_reply_dir_entry);
2237 unsigned int num_entries;
2240 spin_lock(&ci->i_ceph_lock);
2241 num_entries = ci->i_files + ci->i_subdirs;
2242 spin_unlock(&ci->i_ceph_lock);
2243 num_entries = max(num_entries, 1U);
2244 num_entries = min(num_entries, opt->max_readdir);
2246 order = get_order(size * num_entries);
2247 while (order >= 0) {
2248 rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2251 if (rinfo->dir_entries)
2255 if (!rinfo->dir_entries)
2258 num_entries = (PAGE_SIZE << order) / size;
2259 num_entries = min(num_entries, opt->max_readdir);
2261 rinfo->dir_buf_size = PAGE_SIZE << order;
2262 req->r_num_caps = num_entries + 1;
2263 req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2264 req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2269 * Create an mds request.
2271 struct ceph_mds_request *
2272 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2274 struct ceph_mds_request *req;
2276 req = kmem_cache_zalloc(ceph_mds_request_cachep, GFP_NOFS);
2278 return ERR_PTR(-ENOMEM);
2280 mutex_init(&req->r_fill_mutex);
2282 req->r_started = jiffies;
2283 req->r_start_latency = ktime_get();
2284 req->r_resend_mds = -1;
2285 INIT_LIST_HEAD(&req->r_unsafe_dir_item);
2286 INIT_LIST_HEAD(&req->r_unsafe_target_item);
2288 kref_init(&req->r_kref);
2289 RB_CLEAR_NODE(&req->r_node);
2290 INIT_LIST_HEAD(&req->r_wait);
2291 init_completion(&req->r_completion);
2292 init_completion(&req->r_safe_completion);
2293 INIT_LIST_HEAD(&req->r_unsafe_item);
2295 ktime_get_coarse_real_ts64(&req->r_stamp);
2298 req->r_direct_mode = mode;
2303 * return oldest (lowest) request, tid in request tree, 0 if none.
2305 * called under mdsc->mutex.
2307 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2309 if (RB_EMPTY_ROOT(&mdsc->request_tree))
2311 return rb_entry(rb_first(&mdsc->request_tree),
2312 struct ceph_mds_request, r_node);
2315 static inline u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2317 return mdsc->oldest_tid;
2321 * Build a dentry's path. Allocate on heap; caller must kfree. Based
2322 * on build_path_from_dentry in fs/cifs/dir.c.
2324 * If @stop_on_nosnap, generate path relative to the first non-snapped
2327 * Encode hidden .snap dirs as a double /, i.e.
2328 * foo/.snap/bar -> foo//bar
2330 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *pbase,
2333 struct dentry *temp;
2340 return ERR_PTR(-EINVAL);
2344 return ERR_PTR(-ENOMEM);
2349 seq = read_seqbegin(&rename_lock);
2353 struct inode *inode;
2355 spin_lock(&temp->d_lock);
2356 inode = d_inode(temp);
2357 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
2358 dout("build_path path+%d: %p SNAPDIR\n",
2360 } else if (stop_on_nosnap && inode && dentry != temp &&
2361 ceph_snap(inode) == CEPH_NOSNAP) {
2362 spin_unlock(&temp->d_lock);
2363 pos++; /* get rid of any prepended '/' */
2366 pos -= temp->d_name.len;
2368 spin_unlock(&temp->d_lock);
2371 memcpy(path + pos, temp->d_name.name, temp->d_name.len);
2373 spin_unlock(&temp->d_lock);
2374 temp = READ_ONCE(temp->d_parent);
2376 /* Are we at the root? */
2380 /* Are we out of buffer? */
2386 base = ceph_ino(d_inode(temp));
2389 if (read_seqretry(&rename_lock, seq))
2394 * A rename didn't occur, but somehow we didn't end up where
2395 * we thought we would. Throw a warning and try again.
2397 pr_warn("build_path did not end path lookup where "
2398 "expected, pos is %d\n", pos);
2403 *plen = PATH_MAX - 1 - pos;
2404 dout("build_path on %p %d built %llx '%.*s'\n",
2405 dentry, d_count(dentry), base, *plen, path + pos);
2409 static int build_dentry_path(struct dentry *dentry, struct inode *dir,
2410 const char **ppath, int *ppathlen, u64 *pino,
2411 bool *pfreepath, bool parent_locked)
2417 dir = d_inode_rcu(dentry->d_parent);
2418 if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP) {
2419 *pino = ceph_ino(dir);
2421 *ppath = dentry->d_name.name;
2422 *ppathlen = dentry->d_name.len;
2426 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2428 return PTR_ERR(path);
2434 static int build_inode_path(struct inode *inode,
2435 const char **ppath, int *ppathlen, u64 *pino,
2438 struct dentry *dentry;
2441 if (ceph_snap(inode) == CEPH_NOSNAP) {
2442 *pino = ceph_ino(inode);
2446 dentry = d_find_alias(inode);
2447 path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2450 return PTR_ERR(path);
2457 * request arguments may be specified via an inode *, a dentry *, or
2458 * an explicit ino+path.
2460 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
2461 struct inode *rdiri, const char *rpath,
2462 u64 rino, const char **ppath, int *pathlen,
2463 u64 *ino, bool *freepath, bool parent_locked)
2468 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2469 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2471 } else if (rdentry) {
2472 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
2473 freepath, parent_locked);
2474 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
2476 } else if (rpath || rino) {
2479 *pathlen = rpath ? strlen(rpath) : 0;
2480 dout(" path %.*s\n", *pathlen, rpath);
2486 static void encode_timestamp_and_gids(void **p,
2487 const struct ceph_mds_request *req)
2489 struct ceph_timespec ts;
2492 ceph_encode_timespec64(&ts, &req->r_stamp);
2493 ceph_encode_copy(p, &ts, sizeof(ts));
2496 ceph_encode_32(p, req->r_cred->group_info->ngroups);
2497 for (i = 0; i < req->r_cred->group_info->ngroups; i++)
2498 ceph_encode_64(p, from_kgid(&init_user_ns,
2499 req->r_cred->group_info->gid[i]));
2503 * called under mdsc->mutex
2505 static struct ceph_msg *create_request_message(struct ceph_mds_session *session,
2506 struct ceph_mds_request *req,
2507 bool drop_cap_releases)
2509 int mds = session->s_mds;
2510 struct ceph_mds_client *mdsc = session->s_mdsc;
2511 struct ceph_msg *msg;
2512 struct ceph_mds_request_head_old *head;
2513 const char *path1 = NULL;
2514 const char *path2 = NULL;
2515 u64 ino1 = 0, ino2 = 0;
2516 int pathlen1 = 0, pathlen2 = 0;
2517 bool freepath1 = false, freepath2 = false;
2522 bool legacy = !(session->s_con.peer_features & CEPH_FEATURE_FS_BTIME);
2524 ret = set_request_path_attr(req->r_inode, req->r_dentry,
2525 req->r_parent, req->r_path1, req->r_ino1.ino,
2526 &path1, &pathlen1, &ino1, &freepath1,
2527 test_bit(CEPH_MDS_R_PARENT_LOCKED,
2528 &req->r_req_flags));
2534 /* If r_old_dentry is set, then assume that its parent is locked */
2535 ret = set_request_path_attr(NULL, req->r_old_dentry,
2536 req->r_old_dentry_dir,
2537 req->r_path2, req->r_ino2.ino,
2538 &path2, &pathlen2, &ino2, &freepath2, true);
2544 len = legacy ? sizeof(*head) : sizeof(struct ceph_mds_request_head);
2545 len += pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
2546 sizeof(struct ceph_timespec);
2547 len += sizeof(u32) + (sizeof(u64) * req->r_cred->group_info->ngroups);
2549 /* calculate (max) length for cap releases */
2550 len += sizeof(struct ceph_mds_request_release) *
2551 (!!req->r_inode_drop + !!req->r_dentry_drop +
2552 !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2554 if (req->r_dentry_drop)
2556 if (req->r_old_dentry_drop)
2559 msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
2561 msg = ERR_PTR(-ENOMEM);
2565 msg->hdr.tid = cpu_to_le64(req->r_tid);
2568 * The old ceph_mds_request_head didn't contain a version field, and
2569 * one was added when we moved the message version from 3->4.
2572 msg->hdr.version = cpu_to_le16(3);
2573 head = msg->front.iov_base;
2574 p = msg->front.iov_base + sizeof(*head);
2576 struct ceph_mds_request_head *new_head = msg->front.iov_base;
2578 msg->hdr.version = cpu_to_le16(4);
2579 new_head->version = cpu_to_le16(CEPH_MDS_REQUEST_HEAD_VERSION);
2580 head = (struct ceph_mds_request_head_old *)&new_head->oldest_client_tid;
2581 p = msg->front.iov_base + sizeof(*new_head);
2584 end = msg->front.iov_base + msg->front.iov_len;
2586 head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
2587 head->op = cpu_to_le32(req->r_op);
2588 head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns,
2589 req->r_cred->fsuid));
2590 head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns,
2591 req->r_cred->fsgid));
2592 head->ino = cpu_to_le64(req->r_deleg_ino);
2593 head->args = req->r_args;
2595 ceph_encode_filepath(&p, end, ino1, path1);
2596 ceph_encode_filepath(&p, end, ino2, path2);
2598 /* make note of release offset, in case we need to replay */
2599 req->r_request_release_offset = p - msg->front.iov_base;
2603 if (req->r_inode_drop)
2604 releases += ceph_encode_inode_release(&p,
2605 req->r_inode ? req->r_inode : d_inode(req->r_dentry),
2606 mds, req->r_inode_drop, req->r_inode_unless,
2607 req->r_op == CEPH_MDS_OP_READDIR);
2608 if (req->r_dentry_drop)
2609 releases += ceph_encode_dentry_release(&p, req->r_dentry,
2610 req->r_parent, mds, req->r_dentry_drop,
2611 req->r_dentry_unless);
2612 if (req->r_old_dentry_drop)
2613 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
2614 req->r_old_dentry_dir, mds,
2615 req->r_old_dentry_drop,
2616 req->r_old_dentry_unless);
2617 if (req->r_old_inode_drop)
2618 releases += ceph_encode_inode_release(&p,
2619 d_inode(req->r_old_dentry),
2620 mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
2622 if (drop_cap_releases) {
2624 p = msg->front.iov_base + req->r_request_release_offset;
2627 head->num_releases = cpu_to_le16(releases);
2629 encode_timestamp_and_gids(&p, req);
2631 if (WARN_ON_ONCE(p > end)) {
2633 msg = ERR_PTR(-ERANGE);
2637 msg->front.iov_len = p - msg->front.iov_base;
2638 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2640 if (req->r_pagelist) {
2641 struct ceph_pagelist *pagelist = req->r_pagelist;
2642 ceph_msg_data_add_pagelist(msg, pagelist);
2643 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2645 msg->hdr.data_len = 0;
2648 msg->hdr.data_off = cpu_to_le16(0);
2652 ceph_mdsc_free_path((char *)path2, pathlen2);
2655 ceph_mdsc_free_path((char *)path1, pathlen1);
2661 * called under mdsc->mutex if error, under no mutex if
2664 static void complete_request(struct ceph_mds_client *mdsc,
2665 struct ceph_mds_request *req)
2667 req->r_end_latency = ktime_get();
2669 if (req->r_callback)
2670 req->r_callback(mdsc, req);
2671 complete_all(&req->r_completion);
2674 static struct ceph_mds_request_head_old *
2675 find_old_request_head(void *p, u64 features)
2677 bool legacy = !(features & CEPH_FEATURE_FS_BTIME);
2678 struct ceph_mds_request_head *new_head;
2681 return (struct ceph_mds_request_head_old *)p;
2682 new_head = (struct ceph_mds_request_head *)p;
2683 return (struct ceph_mds_request_head_old *)&new_head->oldest_client_tid;
2687 * called under mdsc->mutex
2689 static int __prepare_send_request(struct ceph_mds_session *session,
2690 struct ceph_mds_request *req,
2691 bool drop_cap_releases)
2693 int mds = session->s_mds;
2694 struct ceph_mds_client *mdsc = session->s_mdsc;
2695 struct ceph_mds_request_head_old *rhead;
2696 struct ceph_msg *msg;
2701 struct ceph_cap *cap =
2702 ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2705 req->r_sent_on_mseq = cap->mseq;
2707 req->r_sent_on_mseq = -1;
2709 dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2710 req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2712 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2716 * Replay. Do not regenerate message (and rebuild
2717 * paths, etc.); just use the original message.
2718 * Rebuilding paths will break for renames because
2719 * d_move mangles the src name.
2721 msg = req->r_request;
2722 rhead = find_old_request_head(msg->front.iov_base,
2723 session->s_con.peer_features);
2725 flags = le32_to_cpu(rhead->flags);
2726 flags |= CEPH_MDS_FLAG_REPLAY;
2727 rhead->flags = cpu_to_le32(flags);
2729 if (req->r_target_inode)
2730 rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2732 rhead->num_retry = req->r_attempts - 1;
2734 /* remove cap/dentry releases from message */
2735 rhead->num_releases = 0;
2737 p = msg->front.iov_base + req->r_request_release_offset;
2738 encode_timestamp_and_gids(&p, req);
2740 msg->front.iov_len = p - msg->front.iov_base;
2741 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2745 if (req->r_request) {
2746 ceph_msg_put(req->r_request);
2747 req->r_request = NULL;
2749 msg = create_request_message(session, req, drop_cap_releases);
2751 req->r_err = PTR_ERR(msg);
2752 return PTR_ERR(msg);
2754 req->r_request = msg;
2756 rhead = find_old_request_head(msg->front.iov_base,
2757 session->s_con.peer_features);
2758 rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2759 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2760 flags |= CEPH_MDS_FLAG_REPLAY;
2761 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))
2762 flags |= CEPH_MDS_FLAG_ASYNC;
2764 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2765 rhead->flags = cpu_to_le32(flags);
2766 rhead->num_fwd = req->r_num_fwd;
2767 rhead->num_retry = req->r_attempts - 1;
2769 dout(" r_parent = %p\n", req->r_parent);
2774 * called under mdsc->mutex
2776 static int __send_request(struct ceph_mds_session *session,
2777 struct ceph_mds_request *req,
2778 bool drop_cap_releases)
2782 err = __prepare_send_request(session, req, drop_cap_releases);
2784 ceph_msg_get(req->r_request);
2785 ceph_con_send(&session->s_con, req->r_request);
2792 * send request, or put it on the appropriate wait list.
2794 static void __do_request(struct ceph_mds_client *mdsc,
2795 struct ceph_mds_request *req)
2797 struct ceph_mds_session *session = NULL;
2802 if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2803 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
2804 __unregister_request(mdsc, req);
2808 if (req->r_timeout &&
2809 time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2810 dout("do_request timed out\n");
2814 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
2815 dout("do_request forced umount\n");
2819 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
2820 if (mdsc->mdsmap_err) {
2821 err = mdsc->mdsmap_err;
2822 dout("do_request mdsmap err %d\n", err);
2825 if (mdsc->mdsmap->m_epoch == 0) {
2826 dout("do_request no mdsmap, waiting for map\n");
2827 list_add(&req->r_wait, &mdsc->waiting_for_map);
2830 if (!(mdsc->fsc->mount_options->flags &
2831 CEPH_MOUNT_OPT_MOUNTWAIT) &&
2832 !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
2833 err = -EHOSTUNREACH;
2838 put_request_session(req);
2840 mds = __choose_mds(mdsc, req, &random);
2842 ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2843 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2847 dout("do_request no mds or not active, waiting for map\n");
2848 list_add(&req->r_wait, &mdsc->waiting_for_map);
2852 /* get, open session */
2853 session = __ceph_lookup_mds_session(mdsc, mds);
2855 session = register_session(mdsc, mds);
2856 if (IS_ERR(session)) {
2857 err = PTR_ERR(session);
2861 req->r_session = ceph_get_mds_session(session);
2863 dout("do_request mds%d session %p state %s\n", mds, session,
2864 ceph_session_state_name(session->s_state));
2865 if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2866 session->s_state != CEPH_MDS_SESSION_HUNG) {
2868 * We cannot queue async requests since the caps and delegated
2869 * inodes are bound to the session. Just return -EJUKEBOX and
2870 * let the caller retry a sync request in that case.
2872 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2878 * If the session has been REJECTED, then return a hard error,
2879 * unless it's a CLEANRECOVER mount, in which case we'll queue
2880 * it to the mdsc queue.
2882 if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
2883 if (ceph_test_mount_opt(mdsc->fsc, CLEANRECOVER))
2884 list_add(&req->r_wait, &mdsc->waiting_for_map);
2890 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2891 session->s_state == CEPH_MDS_SESSION_CLOSING) {
2892 err = __open_session(mdsc, session);
2895 /* retry the same mds later */
2897 req->r_resend_mds = mds;
2899 list_add(&req->r_wait, &session->s_waiting);
2904 req->r_resend_mds = -1; /* forget any previous mds hint */
2906 if (req->r_request_started == 0) /* note request start time */
2907 req->r_request_started = jiffies;
2909 err = __send_request(session, req, false);
2912 ceph_put_mds_session(session);
2915 dout("__do_request early error %d\n", err);
2917 complete_request(mdsc, req);
2918 __unregister_request(mdsc, req);
2924 * called under mdsc->mutex
2926 static void __wake_requests(struct ceph_mds_client *mdsc,
2927 struct list_head *head)
2929 struct ceph_mds_request *req;
2930 LIST_HEAD(tmp_list);
2932 list_splice_init(head, &tmp_list);
2934 while (!list_empty(&tmp_list)) {
2935 req = list_entry(tmp_list.next,
2936 struct ceph_mds_request, r_wait);
2937 list_del_init(&req->r_wait);
2938 dout(" wake request %p tid %llu\n", req, req->r_tid);
2939 __do_request(mdsc, req);
2944 * Wake up threads with requests pending for @mds, so that they can
2945 * resubmit their requests to a possibly different mds.
2947 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2949 struct ceph_mds_request *req;
2950 struct rb_node *p = rb_first(&mdsc->request_tree);
2952 dout("kick_requests mds%d\n", mds);
2954 req = rb_entry(p, struct ceph_mds_request, r_node);
2956 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2958 if (req->r_attempts > 0)
2959 continue; /* only new requests */
2960 if (req->r_session &&
2961 req->r_session->s_mds == mds) {
2962 dout(" kicking tid %llu\n", req->r_tid);
2963 list_del_init(&req->r_wait);
2964 __do_request(mdsc, req);
2969 int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
2970 struct ceph_mds_request *req)
2974 /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
2976 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
2977 if (req->r_parent) {
2978 struct ceph_inode_info *ci = ceph_inode(req->r_parent);
2979 int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?
2980 CEPH_FILE_MODE_WR : CEPH_FILE_MODE_RD;
2981 spin_lock(&ci->i_ceph_lock);
2982 ceph_take_cap_refs(ci, CEPH_CAP_PIN, false);
2983 __ceph_touch_fmode(ci, mdsc, fmode);
2984 spin_unlock(&ci->i_ceph_lock);
2986 if (req->r_old_dentry_dir)
2987 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2991 err = ceph_wait_on_async_create(req->r_inode);
2993 dout("%s: wait for async create returned: %d\n",
2999 if (!err && req->r_old_inode) {
3000 err = ceph_wait_on_async_create(req->r_old_inode);
3002 dout("%s: wait for async create returned: %d\n",
3008 dout("submit_request on %p for inode %p\n", req, dir);
3009 mutex_lock(&mdsc->mutex);
3010 __register_request(mdsc, req, dir);
3011 __do_request(mdsc, req);
3013 mutex_unlock(&mdsc->mutex);
3017 static int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
3018 struct ceph_mds_request *req)
3023 dout("do_request waiting\n");
3024 if (!req->r_timeout && req->r_wait_for_completion) {
3025 err = req->r_wait_for_completion(mdsc, req);
3027 long timeleft = wait_for_completion_killable_timeout(
3029 ceph_timeout_jiffies(req->r_timeout));
3033 err = -ETIMEDOUT; /* timed out */
3035 err = timeleft; /* killed */
3037 dout("do_request waited, got %d\n", err);
3038 mutex_lock(&mdsc->mutex);
3040 /* only abort if we didn't race with a real reply */
3041 if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
3042 err = le32_to_cpu(req->r_reply_info.head->result);
3043 } else if (err < 0) {
3044 dout("aborted request %lld with %d\n", req->r_tid, err);
3047 * ensure we aren't running concurrently with
3048 * ceph_fill_trace or ceph_readdir_prepopulate, which
3049 * rely on locks (dir mutex) held by our caller.
3051 mutex_lock(&req->r_fill_mutex);
3053 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3054 mutex_unlock(&req->r_fill_mutex);
3056 if (req->r_parent &&
3057 (req->r_op & CEPH_MDS_OP_WRITE))
3058 ceph_invalidate_dir_request(req);
3063 mutex_unlock(&mdsc->mutex);
3068 * Synchrously perform an mds request. Take care of all of the
3069 * session setup, forwarding, retry details.
3071 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
3073 struct ceph_mds_request *req)
3077 dout("do_request on %p\n", req);
3080 err = ceph_mdsc_submit_request(mdsc, dir, req);
3082 err = ceph_mdsc_wait_request(mdsc, req);
3083 dout("do_request %p done, result %d\n", req, err);
3088 * Invalidate dir's completeness, dentry lease state on an aborted MDS
3089 * namespace request.
3091 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
3093 struct inode *dir = req->r_parent;
3094 struct inode *old_dir = req->r_old_dentry_dir;
3096 dout("invalidate_dir_request %p %p (complete, lease(s))\n", dir, old_dir);
3098 ceph_dir_clear_complete(dir);
3100 ceph_dir_clear_complete(old_dir);
3102 ceph_invalidate_dentry_lease(req->r_dentry);
3103 if (req->r_old_dentry)
3104 ceph_invalidate_dentry_lease(req->r_old_dentry);
3110 * We take the session mutex and parse and process the reply immediately.
3111 * This preserves the logical ordering of replies, capabilities, etc., sent
3112 * by the MDS as they are applied to our local cache.
3114 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
3116 struct ceph_mds_client *mdsc = session->s_mdsc;
3117 struct ceph_mds_request *req;
3118 struct ceph_mds_reply_head *head = msg->front.iov_base;
3119 struct ceph_mds_reply_info_parsed *rinfo; /* parsed reply info */
3120 struct ceph_snap_realm *realm;
3123 int mds = session->s_mds;
3125 if (msg->front.iov_len < sizeof(*head)) {
3126 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
3131 /* get request, session */
3132 tid = le64_to_cpu(msg->hdr.tid);
3133 mutex_lock(&mdsc->mutex);
3134 req = lookup_get_request(mdsc, tid);
3136 dout("handle_reply on unknown tid %llu\n", tid);
3137 mutex_unlock(&mdsc->mutex);
3140 dout("handle_reply %p\n", req);
3142 /* correct session? */
3143 if (req->r_session != session) {
3144 pr_err("mdsc_handle_reply got %llu on session mds%d"
3145 " not mds%d\n", tid, session->s_mds,
3146 req->r_session ? req->r_session->s_mds : -1);
3147 mutex_unlock(&mdsc->mutex);
3152 if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
3153 (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
3154 pr_warn("got a dup %s reply on %llu from mds%d\n",
3155 head->safe ? "safe" : "unsafe", tid, mds);
3156 mutex_unlock(&mdsc->mutex);
3159 if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
3160 pr_warn("got unsafe after safe on %llu from mds%d\n",
3162 mutex_unlock(&mdsc->mutex);
3166 result = le32_to_cpu(head->result);
3170 * if we're not talking to the authority, send to them
3171 * if the authority has changed while we weren't looking,
3172 * send to new authority
3173 * Otherwise we just have to return an ESTALE
3175 if (result == -ESTALE) {
3176 dout("got ESTALE on request %llu\n", req->r_tid);
3177 req->r_resend_mds = -1;
3178 if (req->r_direct_mode != USE_AUTH_MDS) {
3179 dout("not using auth, setting for that now\n");
3180 req->r_direct_mode = USE_AUTH_MDS;
3181 __do_request(mdsc, req);
3182 mutex_unlock(&mdsc->mutex);
3185 int mds = __choose_mds(mdsc, req, NULL);
3186 if (mds >= 0 && mds != req->r_session->s_mds) {
3187 dout("but auth changed, so resending\n");
3188 __do_request(mdsc, req);
3189 mutex_unlock(&mdsc->mutex);
3193 dout("have to return ESTALE on request %llu\n", req->r_tid);
3198 set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
3199 __unregister_request(mdsc, req);
3201 /* last request during umount? */
3202 if (mdsc->stopping && !__get_oldest_req(mdsc))
3203 complete_all(&mdsc->safe_umount_waiters);
3205 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3207 * We already handled the unsafe response, now do the
3208 * cleanup. No need to examine the response; the MDS
3209 * doesn't include any result info in the safe
3210 * response. And even if it did, there is nothing
3211 * useful we could do with a revised return value.
3213 dout("got safe reply %llu, mds%d\n", tid, mds);
3215 mutex_unlock(&mdsc->mutex);
3219 set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
3220 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
3223 dout("handle_reply tid %lld result %d\n", tid, result);
3224 rinfo = &req->r_reply_info;
3225 if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
3226 err = parse_reply_info(session, msg, rinfo, (u64)-1);
3228 err = parse_reply_info(session, msg, rinfo, session->s_con.peer_features);
3229 mutex_unlock(&mdsc->mutex);
3231 /* Must find target inode outside of mutexes to avoid deadlocks */
3232 if ((err >= 0) && rinfo->head->is_target) {
3234 struct ceph_vino tvino = {
3235 .ino = le64_to_cpu(rinfo->targeti.in->ino),
3236 .snap = le64_to_cpu(rinfo->targeti.in->snapid)
3239 in = ceph_get_inode(mdsc->fsc->sb, tvino);
3242 mutex_lock(&session->s_mutex);
3245 req->r_target_inode = in;
3248 mutex_lock(&session->s_mutex);
3250 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
3257 if (rinfo->snapblob_len) {
3258 down_write(&mdsc->snap_rwsem);
3259 ceph_update_snap_trace(mdsc, rinfo->snapblob,
3260 rinfo->snapblob + rinfo->snapblob_len,
3261 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
3263 downgrade_write(&mdsc->snap_rwsem);
3265 down_read(&mdsc->snap_rwsem);
3268 /* insert trace into our cache */
3269 mutex_lock(&req->r_fill_mutex);
3270 current->journal_info = req;
3271 err = ceph_fill_trace(mdsc->fsc->sb, req);
3273 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
3274 req->r_op == CEPH_MDS_OP_LSSNAP))
3275 ceph_readdir_prepopulate(req, req->r_session);
3277 current->journal_info = NULL;
3278 mutex_unlock(&req->r_fill_mutex);
3280 up_read(&mdsc->snap_rwsem);
3282 ceph_put_snap_realm(mdsc, realm);
3285 if (req->r_target_inode &&
3286 test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3287 struct ceph_inode_info *ci =
3288 ceph_inode(req->r_target_inode);
3289 spin_lock(&ci->i_unsafe_lock);
3290 list_add_tail(&req->r_unsafe_target_item,
3291 &ci->i_unsafe_iops);
3292 spin_unlock(&ci->i_unsafe_lock);
3295 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
3298 mutex_lock(&mdsc->mutex);
3299 if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3303 req->r_reply = ceph_msg_get(msg);
3304 set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
3307 dout("reply arrived after request %lld was aborted\n", tid);
3309 mutex_unlock(&mdsc->mutex);
3311 mutex_unlock(&session->s_mutex);
3313 /* kick calling process */
3314 complete_request(mdsc, req);
3316 ceph_update_metadata_metrics(&mdsc->metric, req->r_start_latency,
3317 req->r_end_latency, err);
3319 ceph_mdsc_put_request(req);
3326 * handle mds notification that our request has been forwarded.
3328 static void handle_forward(struct ceph_mds_client *mdsc,
3329 struct ceph_mds_session *session,
3330 struct ceph_msg *msg)
3332 struct ceph_mds_request *req;
3333 u64 tid = le64_to_cpu(msg->hdr.tid);
3337 void *p = msg->front.iov_base;
3338 void *end = p + msg->front.iov_len;
3340 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3341 next_mds = ceph_decode_32(&p);
3342 fwd_seq = ceph_decode_32(&p);
3344 mutex_lock(&mdsc->mutex);
3345 req = lookup_get_request(mdsc, tid);
3347 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
3348 goto out; /* dup reply? */
3351 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3352 dout("forward tid %llu aborted, unregistering\n", tid);
3353 __unregister_request(mdsc, req);
3354 } else if (fwd_seq <= req->r_num_fwd) {
3355 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
3356 tid, next_mds, req->r_num_fwd, fwd_seq);
3358 /* resend. forward race not possible; mds would drop */
3359 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3361 BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3362 req->r_attempts = 0;
3363 req->r_num_fwd = fwd_seq;
3364 req->r_resend_mds = next_mds;
3365 put_request_session(req);
3366 __do_request(mdsc, req);
3368 ceph_mdsc_put_request(req);
3370 mutex_unlock(&mdsc->mutex);
3374 pr_err("mdsc_handle_forward decode error err=%d\n", err);
3377 static int __decode_session_metadata(void **p, void *end,
3380 /* map<string,string> */
3383 ceph_decode_32_safe(p, end, n, bad);
3386 ceph_decode_32_safe(p, end, len, bad);
3387 ceph_decode_need(p, end, len, bad);
3388 err_str = !strncmp(*p, "error_string", len);
3390 ceph_decode_32_safe(p, end, len, bad);
3391 ceph_decode_need(p, end, len, bad);
3393 * Match "blocklisted (blacklisted)" from newer MDSes,
3394 * or "blacklisted" from older MDSes.
3396 if (err_str && strnstr(*p, "blacklisted", len))
3397 *blocklisted = true;
3406 * handle a mds session control message
3408 static void handle_session(struct ceph_mds_session *session,
3409 struct ceph_msg *msg)
3411 struct ceph_mds_client *mdsc = session->s_mdsc;
3412 int mds = session->s_mds;
3413 int msg_version = le16_to_cpu(msg->hdr.version);
3414 void *p = msg->front.iov_base;
3415 void *end = p + msg->front.iov_len;
3416 struct ceph_mds_session_head *h;
3418 u64 seq, features = 0;
3420 bool blocklisted = false;
3423 ceph_decode_need(&p, end, sizeof(*h), bad);
3427 op = le32_to_cpu(h->op);
3428 seq = le64_to_cpu(h->seq);
3430 if (msg_version >= 3) {
3432 /* version >= 2, metadata */
3433 if (__decode_session_metadata(&p, end, &blocklisted) < 0)
3435 /* version >= 3, feature bits */
3436 ceph_decode_32_safe(&p, end, len, bad);
3438 ceph_decode_64_safe(&p, end, features, bad);
3439 p += len - sizeof(features);
3443 mutex_lock(&mdsc->mutex);
3444 if (op == CEPH_SESSION_CLOSE) {
3445 ceph_get_mds_session(session);
3446 __unregister_session(mdsc, session);
3448 /* FIXME: this ttl calculation is generous */
3449 session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
3450 mutex_unlock(&mdsc->mutex);
3452 mutex_lock(&session->s_mutex);
3454 dout("handle_session mds%d %s %p state %s seq %llu\n",
3455 mds, ceph_session_op_name(op), session,
3456 ceph_session_state_name(session->s_state), seq);
3458 if (session->s_state == CEPH_MDS_SESSION_HUNG) {
3459 session->s_state = CEPH_MDS_SESSION_OPEN;
3460 pr_info("mds%d came back\n", session->s_mds);
3464 case CEPH_SESSION_OPEN:
3465 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3466 pr_info("mds%d reconnect success\n", session->s_mds);
3467 session->s_state = CEPH_MDS_SESSION_OPEN;
3468 session->s_features = features;
3469 renewed_caps(mdsc, session, 0);
3470 if (test_bit(CEPHFS_FEATURE_METRIC_COLLECT, &session->s_features))
3471 metric_schedule_delayed(&mdsc->metric);
3474 __close_session(mdsc, session);
3477 case CEPH_SESSION_RENEWCAPS:
3478 if (session->s_renew_seq == seq)
3479 renewed_caps(mdsc, session, 1);
3482 case CEPH_SESSION_CLOSE:
3483 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3484 pr_info("mds%d reconnect denied\n", session->s_mds);
3485 session->s_state = CEPH_MDS_SESSION_CLOSED;
3486 cleanup_session_requests(mdsc, session);
3487 remove_session_caps(session);
3488 wake = 2; /* for good measure */
3489 wake_up_all(&mdsc->session_close_wq);
3492 case CEPH_SESSION_STALE:
3493 pr_info("mds%d caps went stale, renewing\n",
3495 atomic_inc(&session->s_cap_gen);
3496 session->s_cap_ttl = jiffies - 1;
3497 send_renew_caps(mdsc, session);
3500 case CEPH_SESSION_RECALL_STATE:
3501 ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
3504 case CEPH_SESSION_FLUSHMSG:
3505 send_flushmsg_ack(mdsc, session, seq);
3508 case CEPH_SESSION_FORCE_RO:
3509 dout("force_session_readonly %p\n", session);
3510 spin_lock(&session->s_cap_lock);
3511 session->s_readonly = true;
3512 spin_unlock(&session->s_cap_lock);
3513 wake_up_session_caps(session, FORCE_RO);
3516 case CEPH_SESSION_REJECT:
3517 WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
3518 pr_info("mds%d rejected session\n", session->s_mds);
3519 session->s_state = CEPH_MDS_SESSION_REJECTED;
3520 cleanup_session_requests(mdsc, session);
3521 remove_session_caps(session);
3523 mdsc->fsc->blocklisted = true;
3524 wake = 2; /* for good measure */
3528 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
3532 mutex_unlock(&session->s_mutex);
3534 mutex_lock(&mdsc->mutex);
3535 __wake_requests(mdsc, &session->s_waiting);
3537 kick_requests(mdsc, mds);
3538 mutex_unlock(&mdsc->mutex);
3540 if (op == CEPH_SESSION_CLOSE)
3541 ceph_put_mds_session(session);
3545 pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
3546 (int)msg->front.iov_len);
3551 void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req)
3555 dcaps = xchg(&req->r_dir_caps, 0);
3557 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3558 ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);
3562 void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req)
3566 dcaps = xchg(&req->r_dir_caps, 0);
3568 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3569 ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),
3575 * called under session->mutex.
3577 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
3578 struct ceph_mds_session *session)
3580 struct ceph_mds_request *req, *nreq;
3583 dout("replay_unsafe_requests mds%d\n", session->s_mds);
3585 mutex_lock(&mdsc->mutex);
3586 list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)
3587 __send_request(session, req, true);
3590 * also re-send old requests when MDS enters reconnect stage. So that MDS
3591 * can process completed request in clientreplay stage.
3593 p = rb_first(&mdsc->request_tree);
3595 req = rb_entry(p, struct ceph_mds_request, r_node);
3597 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3599 if (req->r_attempts == 0)
3600 continue; /* only old requests */
3601 if (!req->r_session)
3603 if (req->r_session->s_mds != session->s_mds)
3606 ceph_mdsc_release_dir_caps_no_check(req);
3608 __send_request(session, req, true);
3610 mutex_unlock(&mdsc->mutex);
3613 static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
3615 struct ceph_msg *reply;
3616 struct ceph_pagelist *_pagelist;
3621 if (!recon_state->allow_multi)
3624 /* can't handle message that contains both caps and realm */
3625 BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
3627 /* pre-allocate new pagelist */
3628 _pagelist = ceph_pagelist_alloc(GFP_NOFS);
3632 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3636 /* placeholder for nr_caps */
3637 err = ceph_pagelist_encode_32(_pagelist, 0);
3641 if (recon_state->nr_caps) {
3642 /* currently encoding caps */
3643 err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
3647 /* placeholder for nr_realms (currently encoding relams) */
3648 err = ceph_pagelist_encode_32(_pagelist, 0);
3653 err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
3657 page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
3658 addr = kmap_atomic(page);
3659 if (recon_state->nr_caps) {
3660 /* currently encoding caps */
3661 *addr = cpu_to_le32(recon_state->nr_caps);
3663 /* currently encoding relams */
3664 *(addr + 1) = cpu_to_le32(recon_state->nr_realms);
3666 kunmap_atomic(addr);
3668 reply->hdr.version = cpu_to_le16(5);
3669 reply->hdr.compat_version = cpu_to_le16(4);
3671 reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
3672 ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
3674 ceph_con_send(&recon_state->session->s_con, reply);
3675 ceph_pagelist_release(recon_state->pagelist);
3677 recon_state->pagelist = _pagelist;
3678 recon_state->nr_caps = 0;
3679 recon_state->nr_realms = 0;
3680 recon_state->msg_version = 5;
3683 ceph_msg_put(reply);
3685 ceph_pagelist_release(_pagelist);
3689 static struct dentry* d_find_primary(struct inode *inode)
3691 struct dentry *alias, *dn = NULL;
3693 if (hlist_empty(&inode->i_dentry))
3696 spin_lock(&inode->i_lock);
3697 if (hlist_empty(&inode->i_dentry))
3700 if (S_ISDIR(inode->i_mode)) {
3701 alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
3702 if (!IS_ROOT(alias))
3707 hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
3708 spin_lock(&alias->d_lock);
3709 if (!d_unhashed(alias) &&
3710 (ceph_dentry(alias)->flags & CEPH_DENTRY_PRIMARY_LINK)) {
3711 dn = dget_dlock(alias);
3713 spin_unlock(&alias->d_lock);
3718 spin_unlock(&inode->i_lock);
3723 * Encode information about a cap for a reconnect with the MDS.
3725 static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
3729 struct ceph_mds_cap_reconnect v2;
3730 struct ceph_mds_cap_reconnect_v1 v1;
3732 struct ceph_inode_info *ci = cap->ci;
3733 struct ceph_reconnect_state *recon_state = arg;
3734 struct ceph_pagelist *pagelist = recon_state->pagelist;
3735 struct dentry *dentry;
3741 dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
3742 inode, ceph_vinop(inode), cap, cap->cap_id,
3743 ceph_cap_string(cap->issued));
3745 dentry = d_find_primary(inode);
3747 /* set pathbase to parent dir when msg_version >= 2 */
3748 path = ceph_mdsc_build_path(dentry, &pathlen, &pathbase,
3749 recon_state->msg_version >= 2);
3752 err = PTR_ERR(path);
3761 spin_lock(&ci->i_ceph_lock);
3762 cap->seq = 0; /* reset cap seq */
3763 cap->issue_seq = 0; /* and issue_seq */
3764 cap->mseq = 0; /* and migrate_seq */
3765 cap->cap_gen = atomic_read(&cap->session->s_cap_gen);
3767 /* These are lost when the session goes away */
3768 if (S_ISDIR(inode->i_mode)) {
3769 if (cap->issued & CEPH_CAP_DIR_CREATE) {
3770 ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
3771 memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
3773 cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;
3776 if (recon_state->msg_version >= 2) {
3777 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
3778 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3779 rec.v2.issued = cpu_to_le32(cap->issued);
3780 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3781 rec.v2.pathbase = cpu_to_le64(pathbase);
3782 rec.v2.flock_len = (__force __le32)
3783 ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
3785 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
3786 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3787 rec.v1.issued = cpu_to_le32(cap->issued);
3788 rec.v1.size = cpu_to_le64(i_size_read(inode));
3789 ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
3790 ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
3791 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3792 rec.v1.pathbase = cpu_to_le64(pathbase);
3795 if (list_empty(&ci->i_cap_snaps)) {
3796 snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
3798 struct ceph_cap_snap *capsnap =
3799 list_first_entry(&ci->i_cap_snaps,
3800 struct ceph_cap_snap, ci_item);
3801 snap_follows = capsnap->follows;
3803 spin_unlock(&ci->i_ceph_lock);
3805 if (recon_state->msg_version >= 2) {
3806 int num_fcntl_locks, num_flock_locks;
3807 struct ceph_filelock *flocks = NULL;
3808 size_t struct_len, total_len = sizeof(u64);
3812 if (rec.v2.flock_len) {
3813 ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
3815 num_fcntl_locks = 0;
3816 num_flock_locks = 0;
3818 if (num_fcntl_locks + num_flock_locks > 0) {
3819 flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
3820 sizeof(struct ceph_filelock),
3826 err = ceph_encode_locks_to_buffer(inode, flocks,
3841 if (recon_state->msg_version >= 3) {
3842 /* version, compat_version and struct_len */
3843 total_len += 2 * sizeof(u8) + sizeof(u32);
3847 * number of encoded locks is stable, so copy to pagelist
3849 struct_len = 2 * sizeof(u32) +
3850 (num_fcntl_locks + num_flock_locks) *
3851 sizeof(struct ceph_filelock);
3852 rec.v2.flock_len = cpu_to_le32(struct_len);
3854 struct_len += sizeof(u32) + pathlen + sizeof(rec.v2);
3857 struct_len += sizeof(u64); /* snap_follows */
3859 total_len += struct_len;
3861 if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
3862 err = send_reconnect_partial(recon_state);
3864 goto out_freeflocks;
3865 pagelist = recon_state->pagelist;
3868 err = ceph_pagelist_reserve(pagelist, total_len);
3870 goto out_freeflocks;
3872 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3873 if (recon_state->msg_version >= 3) {
3874 ceph_pagelist_encode_8(pagelist, struct_v);
3875 ceph_pagelist_encode_8(pagelist, 1);
3876 ceph_pagelist_encode_32(pagelist, struct_len);
3878 ceph_pagelist_encode_string(pagelist, path, pathlen);
3879 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
3880 ceph_locks_to_pagelist(flocks, pagelist,
3881 num_fcntl_locks, num_flock_locks);
3883 ceph_pagelist_encode_64(pagelist, snap_follows);
3887 err = ceph_pagelist_reserve(pagelist,
3888 sizeof(u64) + sizeof(u32) +
3889 pathlen + sizeof(rec.v1));
3893 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3894 ceph_pagelist_encode_string(pagelist, path, pathlen);
3895 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
3899 ceph_mdsc_free_path(path, pathlen);
3901 recon_state->nr_caps++;
3905 static int encode_snap_realms(struct ceph_mds_client *mdsc,
3906 struct ceph_reconnect_state *recon_state)
3909 struct ceph_pagelist *pagelist = recon_state->pagelist;
3912 if (recon_state->msg_version >= 4) {
3913 err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
3919 * snaprealms. we provide mds with the ino, seq (version), and
3920 * parent for all of our realms. If the mds has any newer info,
3923 for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
3924 struct ceph_snap_realm *realm =
3925 rb_entry(p, struct ceph_snap_realm, node);
3926 struct ceph_mds_snaprealm_reconnect sr_rec;
3928 if (recon_state->msg_version >= 4) {
3929 size_t need = sizeof(u8) * 2 + sizeof(u32) +
3932 if (pagelist->length + need > RECONNECT_MAX_SIZE) {
3933 err = send_reconnect_partial(recon_state);
3936 pagelist = recon_state->pagelist;
3939 err = ceph_pagelist_reserve(pagelist, need);
3943 ceph_pagelist_encode_8(pagelist, 1);
3944 ceph_pagelist_encode_8(pagelist, 1);
3945 ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
3948 dout(" adding snap realm %llx seq %lld parent %llx\n",
3949 realm->ino, realm->seq, realm->parent_ino);
3950 sr_rec.ino = cpu_to_le64(realm->ino);
3951 sr_rec.seq = cpu_to_le64(realm->seq);
3952 sr_rec.parent = cpu_to_le64(realm->parent_ino);
3954 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
3958 recon_state->nr_realms++;
3966 * If an MDS fails and recovers, clients need to reconnect in order to
3967 * reestablish shared state. This includes all caps issued through
3968 * this session _and_ the snap_realm hierarchy. Because it's not
3969 * clear which snap realms the mds cares about, we send everything we
3970 * know about.. that ensures we'll then get any new info the
3971 * recovering MDS might have.
3973 * This is a relatively heavyweight operation, but it's rare.
3975 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
3976 struct ceph_mds_session *session)
3978 struct ceph_msg *reply;
3979 int mds = session->s_mds;
3981 struct ceph_reconnect_state recon_state = {
3986 pr_info("mds%d reconnect start\n", mds);
3988 recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
3989 if (!recon_state.pagelist)
3990 goto fail_nopagelist;
3992 reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3996 xa_destroy(&session->s_delegated_inos);
3998 mutex_lock(&session->s_mutex);
3999 session->s_state = CEPH_MDS_SESSION_RECONNECTING;
4002 dout("session %p state %s\n", session,
4003 ceph_session_state_name(session->s_state));
4005 atomic_inc(&session->s_cap_gen);
4007 spin_lock(&session->s_cap_lock);
4008 /* don't know if session is readonly */
4009 session->s_readonly = 0;
4011 * notify __ceph_remove_cap() that we are composing cap reconnect.
4012 * If a cap get released before being added to the cap reconnect,
4013 * __ceph_remove_cap() should skip queuing cap release.
4015 session->s_cap_reconnect = 1;
4016 /* drop old cap expires; we're about to reestablish that state */
4017 detach_cap_releases(session, &dispose);
4018 spin_unlock(&session->s_cap_lock);
4019 dispose_cap_releases(mdsc, &dispose);
4021 /* trim unused caps to reduce MDS's cache rejoin time */
4022 if (mdsc->fsc->sb->s_root)
4023 shrink_dcache_parent(mdsc->fsc->sb->s_root);
4025 ceph_con_close(&session->s_con);
4026 ceph_con_open(&session->s_con,
4027 CEPH_ENTITY_TYPE_MDS, mds,
4028 ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
4030 /* replay unsafe requests */
4031 replay_unsafe_requests(mdsc, session);
4033 ceph_early_kick_flushing_caps(mdsc, session);
4035 down_read(&mdsc->snap_rwsem);
4037 /* placeholder for nr_caps */
4038 err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
4042 if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
4043 recon_state.msg_version = 3;
4044 recon_state.allow_multi = true;
4045 } else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
4046 recon_state.msg_version = 3;
4048 recon_state.msg_version = 2;
4050 /* trsaverse this session's caps */
4051 err = ceph_iterate_session_caps(session, reconnect_caps_cb, &recon_state);
4053 spin_lock(&session->s_cap_lock);
4054 session->s_cap_reconnect = 0;
4055 spin_unlock(&session->s_cap_lock);
4060 /* check if all realms can be encoded into current message */
4061 if (mdsc->num_snap_realms) {
4063 recon_state.pagelist->length +
4064 mdsc->num_snap_realms *
4065 sizeof(struct ceph_mds_snaprealm_reconnect);
4066 if (recon_state.msg_version >= 4) {
4067 /* number of realms */
4068 total_len += sizeof(u32);
4069 /* version, compat_version and struct_len */
4070 total_len += mdsc->num_snap_realms *
4071 (2 * sizeof(u8) + sizeof(u32));
4073 if (total_len > RECONNECT_MAX_SIZE) {
4074 if (!recon_state.allow_multi) {
4078 if (recon_state.nr_caps) {
4079 err = send_reconnect_partial(&recon_state);
4083 recon_state.msg_version = 5;
4087 err = encode_snap_realms(mdsc, &recon_state);
4091 if (recon_state.msg_version >= 5) {
4092 err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
4097 if (recon_state.nr_caps || recon_state.nr_realms) {
4099 list_first_entry(&recon_state.pagelist->head,
4101 __le32 *addr = kmap_atomic(page);
4102 if (recon_state.nr_caps) {
4103 WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
4104 *addr = cpu_to_le32(recon_state.nr_caps);
4105 } else if (recon_state.msg_version >= 4) {
4106 *(addr + 1) = cpu_to_le32(recon_state.nr_realms);
4108 kunmap_atomic(addr);
4111 reply->hdr.version = cpu_to_le16(recon_state.msg_version);
4112 if (recon_state.msg_version >= 4)
4113 reply->hdr.compat_version = cpu_to_le16(4);
4115 reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
4116 ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
4118 ceph_con_send(&session->s_con, reply);
4120 mutex_unlock(&session->s_mutex);
4122 mutex_lock(&mdsc->mutex);
4123 __wake_requests(mdsc, &session->s_waiting);
4124 mutex_unlock(&mdsc->mutex);
4126 up_read(&mdsc->snap_rwsem);
4127 ceph_pagelist_release(recon_state.pagelist);
4131 ceph_msg_put(reply);
4132 up_read(&mdsc->snap_rwsem);
4133 mutex_unlock(&session->s_mutex);
4135 ceph_pagelist_release(recon_state.pagelist);
4137 pr_err("error %d preparing reconnect for mds%d\n", err, mds);
4143 * compare old and new mdsmaps, kicking requests
4144 * and closing out old connections as necessary
4146 * called under mdsc->mutex.
4148 static void check_new_map(struct ceph_mds_client *mdsc,
4149 struct ceph_mdsmap *newmap,
4150 struct ceph_mdsmap *oldmap)
4153 int oldstate, newstate;
4154 struct ceph_mds_session *s;
4156 dout("check_new_map new %u old %u\n",
4157 newmap->m_epoch, oldmap->m_epoch);
4159 for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4160 if (!mdsc->sessions[i])
4162 s = mdsc->sessions[i];
4163 oldstate = ceph_mdsmap_get_state(oldmap, i);
4164 newstate = ceph_mdsmap_get_state(newmap, i);
4166 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
4167 i, ceph_mds_state_name(oldstate),
4168 ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
4169 ceph_mds_state_name(newstate),
4170 ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
4171 ceph_session_state_name(s->s_state));
4173 if (i >= newmap->possible_max_rank) {
4174 /* force close session for stopped mds */
4175 ceph_get_mds_session(s);
4176 __unregister_session(mdsc, s);
4177 __wake_requests(mdsc, &s->s_waiting);
4178 mutex_unlock(&mdsc->mutex);
4180 mutex_lock(&s->s_mutex);
4181 cleanup_session_requests(mdsc, s);
4182 remove_session_caps(s);
4183 mutex_unlock(&s->s_mutex);
4185 ceph_put_mds_session(s);
4187 mutex_lock(&mdsc->mutex);
4188 kick_requests(mdsc, i);
4192 if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
4193 ceph_mdsmap_get_addr(newmap, i),
4194 sizeof(struct ceph_entity_addr))) {
4196 mutex_unlock(&mdsc->mutex);
4197 mutex_lock(&s->s_mutex);
4198 mutex_lock(&mdsc->mutex);
4199 ceph_con_close(&s->s_con);
4200 mutex_unlock(&s->s_mutex);
4201 s->s_state = CEPH_MDS_SESSION_RESTARTING;
4202 } else if (oldstate == newstate) {
4203 continue; /* nothing new with this mds */
4209 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
4210 newstate >= CEPH_MDS_STATE_RECONNECT) {
4211 mutex_unlock(&mdsc->mutex);
4212 send_mds_reconnect(mdsc, s);
4213 mutex_lock(&mdsc->mutex);
4217 * kick request on any mds that has gone active.
4219 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
4220 newstate >= CEPH_MDS_STATE_ACTIVE) {
4221 if (oldstate != CEPH_MDS_STATE_CREATING &&
4222 oldstate != CEPH_MDS_STATE_STARTING)
4223 pr_info("mds%d recovery completed\n", s->s_mds);
4224 kick_requests(mdsc, i);
4225 mutex_unlock(&mdsc->mutex);
4226 mutex_lock(&s->s_mutex);
4227 mutex_lock(&mdsc->mutex);
4228 ceph_kick_flushing_caps(mdsc, s);
4229 mutex_unlock(&s->s_mutex);
4230 wake_up_session_caps(s, RECONNECT);
4234 for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4235 s = mdsc->sessions[i];
4238 if (!ceph_mdsmap_is_laggy(newmap, i))
4240 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4241 s->s_state == CEPH_MDS_SESSION_HUNG ||
4242 s->s_state == CEPH_MDS_SESSION_CLOSING) {
4243 dout(" connecting to export targets of laggy mds%d\n",
4245 __open_export_target_sessions(mdsc, s);
4257 * caller must hold session s_mutex, dentry->d_lock
4259 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
4261 struct ceph_dentry_info *di = ceph_dentry(dentry);
4263 ceph_put_mds_session(di->lease_session);
4264 di->lease_session = NULL;
4267 static void handle_lease(struct ceph_mds_client *mdsc,
4268 struct ceph_mds_session *session,
4269 struct ceph_msg *msg)
4271 struct super_block *sb = mdsc->fsc->sb;
4272 struct inode *inode;
4273 struct dentry *parent, *dentry;
4274 struct ceph_dentry_info *di;
4275 int mds = session->s_mds;
4276 struct ceph_mds_lease *h = msg->front.iov_base;
4278 struct ceph_vino vino;
4282 dout("handle_lease from mds%d\n", mds);
4285 if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
4287 vino.ino = le64_to_cpu(h->ino);
4288 vino.snap = CEPH_NOSNAP;
4289 seq = le32_to_cpu(h->seq);
4290 dname.len = get_unaligned_le32(h + 1);
4291 if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
4293 dname.name = (void *)(h + 1) + sizeof(u32);
4296 inode = ceph_find_inode(sb, vino);
4297 dout("handle_lease %s, ino %llx %p %.*s\n",
4298 ceph_lease_op_name(h->action), vino.ino, inode,
4299 dname.len, dname.name);
4301 mutex_lock(&session->s_mutex);
4302 inc_session_sequence(session);
4305 dout("handle_lease no inode %llx\n", vino.ino);
4310 parent = d_find_alias(inode);
4312 dout("no parent dentry on inode %p\n", inode);
4314 goto release; /* hrm... */
4316 dname.hash = full_name_hash(parent, dname.name, dname.len);
4317 dentry = d_lookup(parent, &dname);
4322 spin_lock(&dentry->d_lock);
4323 di = ceph_dentry(dentry);
4324 switch (h->action) {
4325 case CEPH_MDS_LEASE_REVOKE:
4326 if (di->lease_session == session) {
4327 if (ceph_seq_cmp(di->lease_seq, seq) > 0)
4328 h->seq = cpu_to_le32(di->lease_seq);
4329 __ceph_mdsc_drop_dentry_lease(dentry);
4334 case CEPH_MDS_LEASE_RENEW:
4335 if (di->lease_session == session &&
4336 di->lease_gen == atomic_read(&session->s_cap_gen) &&
4337 di->lease_renew_from &&
4338 di->lease_renew_after == 0) {
4339 unsigned long duration =
4340 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
4342 di->lease_seq = seq;
4343 di->time = di->lease_renew_from + duration;
4344 di->lease_renew_after = di->lease_renew_from +
4346 di->lease_renew_from = 0;
4350 spin_unlock(&dentry->d_lock);
4357 /* let's just reuse the same message */
4358 h->action = CEPH_MDS_LEASE_REVOKE_ACK;
4360 ceph_con_send(&session->s_con, msg);
4363 mutex_unlock(&session->s_mutex);
4368 pr_err("corrupt lease message\n");
4372 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
4373 struct dentry *dentry, char action,
4376 struct ceph_msg *msg;
4377 struct ceph_mds_lease *lease;
4379 int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
4381 dout("lease_send_msg identry %p %s to mds%d\n",
4382 dentry, ceph_lease_op_name(action), session->s_mds);
4384 msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
4387 lease = msg->front.iov_base;
4388 lease->action = action;
4389 lease->seq = cpu_to_le32(seq);
4391 spin_lock(&dentry->d_lock);
4392 dir = d_inode(dentry->d_parent);
4393 lease->ino = cpu_to_le64(ceph_ino(dir));
4394 lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
4396 put_unaligned_le32(dentry->d_name.len, lease + 1);
4397 memcpy((void *)(lease + 1) + 4,
4398 dentry->d_name.name, dentry->d_name.len);
4399 spin_unlock(&dentry->d_lock);
4401 * if this is a preemptive lease RELEASE, no need to
4402 * flush request stream, since the actual request will
4405 msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
4407 ceph_con_send(&session->s_con, msg);
4411 * lock unlock sessions, to wait ongoing session activities
4413 static void lock_unlock_sessions(struct ceph_mds_client *mdsc)
4417 mutex_lock(&mdsc->mutex);
4418 for (i = 0; i < mdsc->max_sessions; i++) {
4419 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4422 mutex_unlock(&mdsc->mutex);
4423 mutex_lock(&s->s_mutex);
4424 mutex_unlock(&s->s_mutex);
4425 ceph_put_mds_session(s);
4426 mutex_lock(&mdsc->mutex);
4428 mutex_unlock(&mdsc->mutex);
4431 static void maybe_recover_session(struct ceph_mds_client *mdsc)
4433 struct ceph_fs_client *fsc = mdsc->fsc;
4435 if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
4438 if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
4441 if (!READ_ONCE(fsc->blocklisted))
4444 pr_info("auto reconnect after blocklisted\n");
4445 ceph_force_reconnect(fsc->sb);
4448 bool check_session_state(struct ceph_mds_session *s)
4450 switch (s->s_state) {
4451 case CEPH_MDS_SESSION_OPEN:
4452 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
4453 s->s_state = CEPH_MDS_SESSION_HUNG;
4454 pr_info("mds%d hung\n", s->s_mds);
4457 case CEPH_MDS_SESSION_CLOSING:
4458 /* Should never reach this when we're unmounting */
4461 case CEPH_MDS_SESSION_NEW:
4462 case CEPH_MDS_SESSION_RESTARTING:
4463 case CEPH_MDS_SESSION_CLOSED:
4464 case CEPH_MDS_SESSION_REJECTED:
4472 * If the sequence is incremented while we're waiting on a REQUEST_CLOSE reply,
4473 * then we need to retransmit that request.
4475 void inc_session_sequence(struct ceph_mds_session *s)
4477 lockdep_assert_held(&s->s_mutex);
4481 if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
4484 dout("resending session close request for mds%d\n", s->s_mds);
4485 ret = request_close_session(s);
4487 pr_err("unable to close session to mds%d: %d\n",
4493 * delayed work -- periodically trim expired leases, renew caps with mds
4495 static void schedule_delayed(struct ceph_mds_client *mdsc)
4498 unsigned hz = round_jiffies_relative(HZ * delay);
4499 schedule_delayed_work(&mdsc->delayed_work, hz);
4502 static void delayed_work(struct work_struct *work)
4505 struct ceph_mds_client *mdsc =
4506 container_of(work, struct ceph_mds_client, delayed_work.work);
4510 dout("mdsc delayed_work\n");
4515 mutex_lock(&mdsc->mutex);
4516 renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
4517 renew_caps = time_after_eq(jiffies, HZ*renew_interval +
4518 mdsc->last_renew_caps);
4520 mdsc->last_renew_caps = jiffies;
4522 for (i = 0; i < mdsc->max_sessions; i++) {
4523 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4527 if (!check_session_state(s)) {
4528 ceph_put_mds_session(s);
4531 mutex_unlock(&mdsc->mutex);
4533 mutex_lock(&s->s_mutex);
4535 send_renew_caps(mdsc, s);
4537 ceph_con_keepalive(&s->s_con);
4538 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4539 s->s_state == CEPH_MDS_SESSION_HUNG)
4540 ceph_send_cap_releases(mdsc, s);
4541 mutex_unlock(&s->s_mutex);
4542 ceph_put_mds_session(s);
4544 mutex_lock(&mdsc->mutex);
4546 mutex_unlock(&mdsc->mutex);
4548 ceph_check_delayed_caps(mdsc);
4550 ceph_queue_cap_reclaim_work(mdsc);
4552 ceph_trim_snapid_map(mdsc);
4554 maybe_recover_session(mdsc);
4556 schedule_delayed(mdsc);
4559 int ceph_mdsc_init(struct ceph_fs_client *fsc)
4562 struct ceph_mds_client *mdsc;
4565 mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
4569 mutex_init(&mdsc->mutex);
4570 mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
4571 if (!mdsc->mdsmap) {
4576 init_completion(&mdsc->safe_umount_waiters);
4577 init_waitqueue_head(&mdsc->session_close_wq);
4578 INIT_LIST_HEAD(&mdsc->waiting_for_map);
4579 mdsc->sessions = NULL;
4580 atomic_set(&mdsc->num_sessions, 0);
4581 mdsc->max_sessions = 0;
4583 atomic64_set(&mdsc->quotarealms_count, 0);
4584 mdsc->quotarealms_inodes = RB_ROOT;
4585 mutex_init(&mdsc->quotarealms_inodes_mutex);
4586 mdsc->last_snap_seq = 0;
4587 init_rwsem(&mdsc->snap_rwsem);
4588 mdsc->snap_realms = RB_ROOT;
4589 INIT_LIST_HEAD(&mdsc->snap_empty);
4590 mdsc->num_snap_realms = 0;
4591 spin_lock_init(&mdsc->snap_empty_lock);
4593 mdsc->oldest_tid = 0;
4594 mdsc->request_tree = RB_ROOT;
4595 INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
4596 mdsc->last_renew_caps = jiffies;
4597 INIT_LIST_HEAD(&mdsc->cap_delay_list);
4598 INIT_LIST_HEAD(&mdsc->cap_wait_list);
4599 spin_lock_init(&mdsc->cap_delay_lock);
4600 INIT_LIST_HEAD(&mdsc->snap_flush_list);
4601 spin_lock_init(&mdsc->snap_flush_lock);
4602 mdsc->last_cap_flush_tid = 1;
4603 INIT_LIST_HEAD(&mdsc->cap_flush_list);
4604 INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
4605 mdsc->num_cap_flushing = 0;
4606 spin_lock_init(&mdsc->cap_dirty_lock);
4607 init_waitqueue_head(&mdsc->cap_flushing_wq);
4608 INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
4609 atomic_set(&mdsc->cap_reclaim_pending, 0);
4610 err = ceph_metric_init(&mdsc->metric);
4614 spin_lock_init(&mdsc->dentry_list_lock);
4615 INIT_LIST_HEAD(&mdsc->dentry_leases);
4616 INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
4618 ceph_caps_init(mdsc);
4619 ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
4621 spin_lock_init(&mdsc->snapid_map_lock);
4622 mdsc->snapid_map_tree = RB_ROOT;
4623 INIT_LIST_HEAD(&mdsc->snapid_map_lru);
4625 init_rwsem(&mdsc->pool_perm_rwsem);
4626 mdsc->pool_perm_tree = RB_ROOT;
4628 strscpy(mdsc->nodename, utsname()->nodename,
4629 sizeof(mdsc->nodename));
4635 kfree(mdsc->mdsmap);
4642 * Wait for safe replies on open mds requests. If we time out, drop
4643 * all requests from the tree to avoid dangling dentry refs.
4645 static void wait_requests(struct ceph_mds_client *mdsc)
4647 struct ceph_options *opts = mdsc->fsc->client->options;
4648 struct ceph_mds_request *req;
4650 mutex_lock(&mdsc->mutex);
4651 if (__get_oldest_req(mdsc)) {
4652 mutex_unlock(&mdsc->mutex);
4654 dout("wait_requests waiting for requests\n");
4655 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
4656 ceph_timeout_jiffies(opts->mount_timeout));
4658 /* tear down remaining requests */
4659 mutex_lock(&mdsc->mutex);
4660 while ((req = __get_oldest_req(mdsc))) {
4661 dout("wait_requests timed out on tid %llu\n",
4663 list_del_init(&req->r_wait);
4664 __unregister_request(mdsc, req);
4667 mutex_unlock(&mdsc->mutex);
4668 dout("wait_requests done\n");
4672 * called before mount is ro, and before dentries are torn down.
4673 * (hmm, does this still race with new lookups?)
4675 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
4677 dout("pre_umount\n");
4680 lock_unlock_sessions(mdsc);
4681 ceph_flush_dirty_caps(mdsc);
4682 wait_requests(mdsc);
4685 * wait for reply handlers to drop their request refs and
4686 * their inode/dcache refs
4690 ceph_cleanup_quotarealms_inodes(mdsc);
4694 * wait for all write mds requests to flush.
4696 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
4698 struct ceph_mds_request *req = NULL, *nextreq;
4701 mutex_lock(&mdsc->mutex);
4702 dout("wait_unsafe_requests want %lld\n", want_tid);
4704 req = __get_oldest_req(mdsc);
4705 while (req && req->r_tid <= want_tid) {
4706 /* find next request */
4707 n = rb_next(&req->r_node);
4709 nextreq = rb_entry(n, struct ceph_mds_request, r_node);
4712 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
4713 (req->r_op & CEPH_MDS_OP_WRITE)) {
4715 ceph_mdsc_get_request(req);
4717 ceph_mdsc_get_request(nextreq);
4718 mutex_unlock(&mdsc->mutex);
4719 dout("wait_unsafe_requests wait on %llu (want %llu)\n",
4720 req->r_tid, want_tid);
4721 wait_for_completion(&req->r_safe_completion);
4722 mutex_lock(&mdsc->mutex);
4723 ceph_mdsc_put_request(req);
4725 break; /* next dne before, so we're done! */
4726 if (RB_EMPTY_NODE(&nextreq->r_node)) {
4727 /* next request was removed from tree */
4728 ceph_mdsc_put_request(nextreq);
4731 ceph_mdsc_put_request(nextreq); /* won't go away */
4735 mutex_unlock(&mdsc->mutex);
4736 dout("wait_unsafe_requests done\n");
4739 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
4741 u64 want_tid, want_flush;
4743 if (READ_ONCE(mdsc->fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN)
4747 mutex_lock(&mdsc->mutex);
4748 want_tid = mdsc->last_tid;
4749 mutex_unlock(&mdsc->mutex);
4751 ceph_flush_dirty_caps(mdsc);
4752 spin_lock(&mdsc->cap_dirty_lock);
4753 want_flush = mdsc->last_cap_flush_tid;
4754 if (!list_empty(&mdsc->cap_flush_list)) {
4755 struct ceph_cap_flush *cf =
4756 list_last_entry(&mdsc->cap_flush_list,
4757 struct ceph_cap_flush, g_list);
4760 spin_unlock(&mdsc->cap_dirty_lock);
4762 dout("sync want tid %lld flush_seq %lld\n",
4763 want_tid, want_flush);
4765 wait_unsafe_requests(mdsc, want_tid);
4766 wait_caps_flush(mdsc, want_flush);
4770 * true if all sessions are closed, or we force unmount
4772 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
4774 if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4776 return atomic_read(&mdsc->num_sessions) <= skipped;
4780 * called after sb is ro.
4782 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
4784 struct ceph_options *opts = mdsc->fsc->client->options;
4785 struct ceph_mds_session *session;
4789 dout("close_sessions\n");
4791 /* close sessions */
4792 mutex_lock(&mdsc->mutex);
4793 for (i = 0; i < mdsc->max_sessions; i++) {
4794 session = __ceph_lookup_mds_session(mdsc, i);
4797 mutex_unlock(&mdsc->mutex);
4798 mutex_lock(&session->s_mutex);
4799 if (__close_session(mdsc, session) <= 0)
4801 mutex_unlock(&session->s_mutex);
4802 ceph_put_mds_session(session);
4803 mutex_lock(&mdsc->mutex);
4805 mutex_unlock(&mdsc->mutex);
4807 dout("waiting for sessions to close\n");
4808 wait_event_timeout(mdsc->session_close_wq,
4809 done_closing_sessions(mdsc, skipped),
4810 ceph_timeout_jiffies(opts->mount_timeout));
4812 /* tear down remaining sessions */
4813 mutex_lock(&mdsc->mutex);
4814 for (i = 0; i < mdsc->max_sessions; i++) {
4815 if (mdsc->sessions[i]) {
4816 session = ceph_get_mds_session(mdsc->sessions[i]);
4817 __unregister_session(mdsc, session);
4818 mutex_unlock(&mdsc->mutex);
4819 mutex_lock(&session->s_mutex);
4820 remove_session_caps(session);
4821 mutex_unlock(&session->s_mutex);
4822 ceph_put_mds_session(session);
4823 mutex_lock(&mdsc->mutex);
4826 WARN_ON(!list_empty(&mdsc->cap_delay_list));
4827 mutex_unlock(&mdsc->mutex);
4829 ceph_cleanup_snapid_map(mdsc);
4830 ceph_cleanup_empty_realms(mdsc);
4832 cancel_work_sync(&mdsc->cap_reclaim_work);
4833 cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4838 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
4840 struct ceph_mds_session *session;
4843 dout("force umount\n");
4845 mutex_lock(&mdsc->mutex);
4846 for (mds = 0; mds < mdsc->max_sessions; mds++) {
4847 session = __ceph_lookup_mds_session(mdsc, mds);
4851 if (session->s_state == CEPH_MDS_SESSION_REJECTED)
4852 __unregister_session(mdsc, session);
4853 __wake_requests(mdsc, &session->s_waiting);
4854 mutex_unlock(&mdsc->mutex);
4856 mutex_lock(&session->s_mutex);
4857 __close_session(mdsc, session);
4858 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
4859 cleanup_session_requests(mdsc, session);
4860 remove_session_caps(session);
4862 mutex_unlock(&session->s_mutex);
4863 ceph_put_mds_session(session);
4865 mutex_lock(&mdsc->mutex);
4866 kick_requests(mdsc, mds);
4868 __wake_requests(mdsc, &mdsc->waiting_for_map);
4869 mutex_unlock(&mdsc->mutex);
4872 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
4876 * Make sure the delayed work stopped before releasing
4879 * Because the cancel_delayed_work_sync() will only
4880 * guarantee that the work finishes executing. But the
4881 * delayed work will re-arm itself again after that.
4883 flush_delayed_work(&mdsc->delayed_work);
4886 ceph_mdsmap_destroy(mdsc->mdsmap);
4887 kfree(mdsc->sessions);
4888 ceph_caps_finalize(mdsc);
4889 ceph_pool_perm_destroy(mdsc);
4892 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
4894 struct ceph_mds_client *mdsc = fsc->mdsc;
4895 dout("mdsc_destroy %p\n", mdsc);
4900 /* flush out any connection work with references to us */
4903 ceph_mdsc_stop(mdsc);
4905 ceph_metric_destroy(&mdsc->metric);
4907 flush_delayed_work(&mdsc->metric.delayed_work);
4910 dout("mdsc_destroy %p done\n", mdsc);
4913 void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4915 struct ceph_fs_client *fsc = mdsc->fsc;
4916 const char *mds_namespace = fsc->mount_options->mds_namespace;
4917 void *p = msg->front.iov_base;
4918 void *end = p + msg->front.iov_len;
4921 u32 mount_fscid = (u32)-1;
4924 ceph_decode_need(&p, end, sizeof(u32), bad);
4925 epoch = ceph_decode_32(&p);
4927 dout("handle_fsmap epoch %u\n", epoch);
4929 /* struct_v, struct_cv, map_len, epoch, legacy_client_fscid */
4930 ceph_decode_skip_n(&p, end, 2 + sizeof(u32) * 3, bad);
4932 ceph_decode_32_safe(&p, end, num_fs, bad);
4933 while (num_fs-- > 0) {
4934 void *info_p, *info_end;
4938 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4939 p += 2; // info_v, info_cv
4940 info_len = ceph_decode_32(&p);
4941 ceph_decode_need(&p, end, info_len, bad);
4943 info_end = p + info_len;
4946 ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
4947 fscid = ceph_decode_32(&info_p);
4948 namelen = ceph_decode_32(&info_p);
4949 ceph_decode_need(&info_p, info_end, namelen, bad);
4951 if (mds_namespace &&
4952 strlen(mds_namespace) == namelen &&
4953 !strncmp(mds_namespace, (char *)info_p, namelen)) {
4954 mount_fscid = fscid;
4959 ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
4960 if (mount_fscid != (u32)-1) {
4961 fsc->client->monc.fs_cluster_id = mount_fscid;
4962 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
4964 ceph_monc_renew_subs(&fsc->client->monc);
4972 pr_err("error decoding fsmap\n");
4974 mutex_lock(&mdsc->mutex);
4975 mdsc->mdsmap_err = err;
4976 __wake_requests(mdsc, &mdsc->waiting_for_map);
4977 mutex_unlock(&mdsc->mutex);
4981 * handle mds map update.
4983 void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4987 void *p = msg->front.iov_base;
4988 void *end = p + msg->front.iov_len;
4989 struct ceph_mdsmap *newmap, *oldmap;
4990 struct ceph_fsid fsid;
4993 ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
4994 ceph_decode_copy(&p, &fsid, sizeof(fsid));
4995 if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
4997 epoch = ceph_decode_32(&p);
4998 maplen = ceph_decode_32(&p);
4999 dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
5001 /* do we need it? */
5002 mutex_lock(&mdsc->mutex);
5003 if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
5004 dout("handle_map epoch %u <= our %u\n",
5005 epoch, mdsc->mdsmap->m_epoch);
5006 mutex_unlock(&mdsc->mutex);
5010 newmap = ceph_mdsmap_decode(&p, end, ceph_msgr2(mdsc->fsc->client));
5011 if (IS_ERR(newmap)) {
5012 err = PTR_ERR(newmap);
5016 /* swap into place */
5018 oldmap = mdsc->mdsmap;
5019 mdsc->mdsmap = newmap;
5020 check_new_map(mdsc, newmap, oldmap);
5021 ceph_mdsmap_destroy(oldmap);
5023 mdsc->mdsmap = newmap; /* first mds map */
5025 mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
5028 __wake_requests(mdsc, &mdsc->waiting_for_map);
5029 ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
5030 mdsc->mdsmap->m_epoch);
5032 mutex_unlock(&mdsc->mutex);
5033 schedule_delayed(mdsc);
5037 mutex_unlock(&mdsc->mutex);
5039 pr_err("error decoding mdsmap %d\n", err);
5043 static struct ceph_connection *mds_get_con(struct ceph_connection *con)
5045 struct ceph_mds_session *s = con->private;
5047 if (ceph_get_mds_session(s))
5052 static void mds_put_con(struct ceph_connection *con)
5054 struct ceph_mds_session *s = con->private;
5056 ceph_put_mds_session(s);
5060 * if the client is unresponsive for long enough, the mds will kill
5061 * the session entirely.
5063 static void mds_peer_reset(struct ceph_connection *con)
5065 struct ceph_mds_session *s = con->private;
5066 struct ceph_mds_client *mdsc = s->s_mdsc;
5068 pr_warn("mds%d closed our session\n", s->s_mds);
5069 send_mds_reconnect(mdsc, s);
5072 static void mds_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5074 struct ceph_mds_session *s = con->private;
5075 struct ceph_mds_client *mdsc = s->s_mdsc;
5076 int type = le16_to_cpu(msg->hdr.type);
5078 mutex_lock(&mdsc->mutex);
5079 if (__verify_registered_session(mdsc, s) < 0) {
5080 mutex_unlock(&mdsc->mutex);
5083 mutex_unlock(&mdsc->mutex);
5086 case CEPH_MSG_MDS_MAP:
5087 ceph_mdsc_handle_mdsmap(mdsc, msg);
5089 case CEPH_MSG_FS_MAP_USER:
5090 ceph_mdsc_handle_fsmap(mdsc, msg);
5092 case CEPH_MSG_CLIENT_SESSION:
5093 handle_session(s, msg);
5095 case CEPH_MSG_CLIENT_REPLY:
5096 handle_reply(s, msg);
5098 case CEPH_MSG_CLIENT_REQUEST_FORWARD:
5099 handle_forward(mdsc, s, msg);
5101 case CEPH_MSG_CLIENT_CAPS:
5102 ceph_handle_caps(s, msg);
5104 case CEPH_MSG_CLIENT_SNAP:
5105 ceph_handle_snap(mdsc, s, msg);
5107 case CEPH_MSG_CLIENT_LEASE:
5108 handle_lease(mdsc, s, msg);
5110 case CEPH_MSG_CLIENT_QUOTA:
5111 ceph_handle_quota(mdsc, s, msg);
5115 pr_err("received unknown message type %d %s\n", type,
5116 ceph_msg_type_name(type));
5127 * Note: returned pointer is the address of a structure that's
5128 * managed separately. Caller must *not* attempt to free it.
5130 static struct ceph_auth_handshake *
5131 mds_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
5133 struct ceph_mds_session *s = con->private;
5134 struct ceph_mds_client *mdsc = s->s_mdsc;
5135 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5136 struct ceph_auth_handshake *auth = &s->s_auth;
5139 ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5140 force_new, proto, NULL, NULL);
5142 return ERR_PTR(ret);
5147 static int mds_add_authorizer_challenge(struct ceph_connection *con,
5148 void *challenge_buf, int challenge_buf_len)
5150 struct ceph_mds_session *s = con->private;
5151 struct ceph_mds_client *mdsc = s->s_mdsc;
5152 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5154 return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
5155 challenge_buf, challenge_buf_len);
5158 static int mds_verify_authorizer_reply(struct ceph_connection *con)
5160 struct ceph_mds_session *s = con->private;
5161 struct ceph_mds_client *mdsc = s->s_mdsc;
5162 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5163 struct ceph_auth_handshake *auth = &s->s_auth;
5165 return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5166 auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5167 NULL, NULL, NULL, NULL);
5170 static int mds_invalidate_authorizer(struct ceph_connection *con)
5172 struct ceph_mds_session *s = con->private;
5173 struct ceph_mds_client *mdsc = s->s_mdsc;
5174 struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5176 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
5178 return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
5181 static int mds_get_auth_request(struct ceph_connection *con,
5182 void *buf, int *buf_len,
5183 void **authorizer, int *authorizer_len)
5185 struct ceph_mds_session *s = con->private;
5186 struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5187 struct ceph_auth_handshake *auth = &s->s_auth;
5190 ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5195 *authorizer = auth->authorizer_buf;
5196 *authorizer_len = auth->authorizer_buf_len;
5200 static int mds_handle_auth_reply_more(struct ceph_connection *con,
5201 void *reply, int reply_len,
5202 void *buf, int *buf_len,
5203 void **authorizer, int *authorizer_len)
5205 struct ceph_mds_session *s = con->private;
5206 struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5207 struct ceph_auth_handshake *auth = &s->s_auth;
5210 ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
5215 *authorizer = auth->authorizer_buf;
5216 *authorizer_len = auth->authorizer_buf_len;
5220 static int mds_handle_auth_done(struct ceph_connection *con,
5221 u64 global_id, void *reply, int reply_len,
5222 u8 *session_key, int *session_key_len,
5223 u8 *con_secret, int *con_secret_len)
5225 struct ceph_mds_session *s = con->private;
5226 struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5227 struct ceph_auth_handshake *auth = &s->s_auth;
5229 return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
5230 session_key, session_key_len,
5231 con_secret, con_secret_len);
5234 static int mds_handle_auth_bad_method(struct ceph_connection *con,
5235 int used_proto, int result,
5236 const int *allowed_protos, int proto_cnt,
5237 const int *allowed_modes, int mode_cnt)
5239 struct ceph_mds_session *s = con->private;
5240 struct ceph_mon_client *monc = &s->s_mdsc->fsc->client->monc;
5243 if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_MDS,
5245 allowed_protos, proto_cnt,
5246 allowed_modes, mode_cnt)) {
5247 ret = ceph_monc_validate_auth(monc);
5255 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
5256 struct ceph_msg_header *hdr, int *skip)
5258 struct ceph_msg *msg;
5259 int type = (int) le16_to_cpu(hdr->type);
5260 int front_len = (int) le32_to_cpu(hdr->front_len);
5266 msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
5268 pr_err("unable to allocate msg type %d len %d\n",
5276 static int mds_sign_message(struct ceph_msg *msg)
5278 struct ceph_mds_session *s = msg->con->private;
5279 struct ceph_auth_handshake *auth = &s->s_auth;
5281 return ceph_auth_sign_message(auth, msg);
5284 static int mds_check_message_signature(struct ceph_msg *msg)
5286 struct ceph_mds_session *s = msg->con->private;
5287 struct ceph_auth_handshake *auth = &s->s_auth;
5289 return ceph_auth_check_message_signature(auth, msg);
5292 static const struct ceph_connection_operations mds_con_ops = {
5295 .alloc_msg = mds_alloc_msg,
5296 .dispatch = mds_dispatch,
5297 .peer_reset = mds_peer_reset,
5298 .get_authorizer = mds_get_authorizer,
5299 .add_authorizer_challenge = mds_add_authorizer_challenge,
5300 .verify_authorizer_reply = mds_verify_authorizer_reply,
5301 .invalidate_authorizer = mds_invalidate_authorizer,
5302 .sign_message = mds_sign_message,
5303 .check_message_signature = mds_check_message_signature,
5304 .get_auth_request = mds_get_auth_request,
5305 .handle_auth_reply_more = mds_handle_auth_reply_more,
5306 .handle_auth_done = mds_handle_auth_done,
5307 .handle_auth_bad_method = mds_handle_auth_bad_method,