1 // SPDX-License-Identifier: GPL-2.0
25 #define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
72 static int parse_reply_info_quota(void **p, void *end,  in parse_reply_info_quota()  argument
78 	ceph_decode_8_safe(p, end, struct_v, bad);  in parse_reply_info_quota()
79 	ceph_decode_8_safe(p, end, struct_compat, bad);  in parse_reply_info_quota()
84 	ceph_decode_32_safe(p, end, struct_len, bad);  in parse_reply_info_quota()
85 	ceph_decode_need(p, end, struct_len, bad);  in parse_reply_info_quota()
86 	end = *p + struct_len;  in parse_reply_info_quota()
87 	ceph_decode_64_safe(p, end, info->max_bytes, bad);  in parse_reply_info_quota()
88 	ceph_decode_64_safe(p, end, info->max_files, bad);  in parse_reply_info_quota()
89 	*p = end;  in parse_reply_info_quota()
92 	return -EIO;  in parse_reply_info_quota()
98 static int parse_reply_info_in(void **p, void *end,  in parse_reply_info_in()  argument
105 	if (features == (u64)-1) {  in parse_reply_info_in()
108 		ceph_decode_8_safe(p, end, struct_v, bad);  in parse_reply_info_in()
109 		ceph_decode_8_safe(p, end, struct_compat, bad);  in parse_reply_info_in()
114 		ceph_decode_32_safe(p, end, struct_len, bad);  in parse_reply_info_in()
115 		ceph_decode_need(p, end, struct_len, bad);  in parse_reply_info_in()
116 		end = *p + struct_len;  in parse_reply_info_in()
119 	ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);  in parse_reply_info_in()
120 	info->in = *p;  in parse_reply_info_in()
122 		sizeof(*info->in->fragtree.splits) *  in parse_reply_info_in()
123 		le32_to_cpu(info->in->fragtree.nsplits);  in parse_reply_info_in()
125 	ceph_decode_32_safe(p, end, info->symlink_len, bad);  in parse_reply_info_in()
126 	ceph_decode_need(p, end, info->symlink_len, bad);  in parse_reply_info_in()
127 	info->symlink = *p;  in parse_reply_info_in()
128 	*p += info->symlink_len;  in parse_reply_info_in()
130 	ceph_decode_copy_safe(p, end, &info->dir_layout,  in parse_reply_info_in()
131 			      sizeof(info->dir_layout), bad);  in parse_reply_info_in()
132 	ceph_decode_32_safe(p, end, info->xattr_len, bad);  in parse_reply_info_in()
133 	ceph_decode_need(p, end, info->xattr_len, bad);  in parse_reply_info_in()
134 	info->xattr_data = *p;  in parse_reply_info_in()
135 	*p += info->xattr_len;  in parse_reply_info_in()
137 	if (features == (u64)-1) {  in parse_reply_info_in()
139 		ceph_decode_64_safe(p, end, info->inline_version, bad);  in parse_reply_info_in()
140 		ceph_decode_32_safe(p, end, info->inline_len, bad);  in parse_reply_info_in()
141 		ceph_decode_need(p, end, info->inline_len, bad);  in parse_reply_info_in()
142 		info->inline_data = *p;  in parse_reply_info_in()
143 		*p += info->inline_len;  in parse_reply_info_in()
145 		err = parse_reply_info_quota(p, end, info);  in parse_reply_info_in()
149 		ceph_decode_32_safe(p, end, info->pool_ns_len, bad);  in parse_reply_info_in()
150 		if (info->pool_ns_len > 0) {  in parse_reply_info_in()
151 			ceph_decode_need(p, end, info->pool_ns_len, bad);  in parse_reply_info_in()
152 			info->pool_ns_data = *p;  in parse_reply_info_in()
153 			*p += info->pool_ns_len;  in parse_reply_info_in()
157 		ceph_decode_need(p, end, sizeof(info->btime), bad);  in parse_reply_info_in()
158 		ceph_decode_copy(p, &info->btime, sizeof(info->btime));  in parse_reply_info_in()
161 		ceph_decode_64_safe(p, end, info->change_attr, bad);  in parse_reply_info_in()
165 			ceph_decode_32_safe(p, end, info->dir_pin, bad);  in parse_reply_info_in()
167 			info->dir_pin = -ENODATA;  in parse_reply_info_in()
172 			ceph_decode_need(p, end, sizeof(info->snap_btime), bad);  in parse_reply_info_in()
173 			ceph_decode_copy(p, &info->snap_btime,  in parse_reply_info_in()
174 					 sizeof(info->snap_btime));  in parse_reply_info_in()
176 			memset(&info->snap_btime, 0, sizeof(info->snap_btime));  in parse_reply_info_in()
179 		*p = end;  in parse_reply_info_in()
182 			ceph_decode_64_safe(p, end, info->inline_version, bad);  in parse_reply_info_in()
183 			ceph_decode_32_safe(p, end, info->inline_len, bad);  in parse_reply_info_in()
184 			ceph_decode_need(p, end, info->inline_len, bad);  in parse_reply_info_in()
185 			info->inline_data = *p;  in parse_reply_info_in()
186 			*p += info->inline_len;  in parse_reply_info_in()
188 			info->inline_version = CEPH_INLINE_NONE;  in parse_reply_info_in()
191 			err = parse_reply_info_quota(p, end, info);  in parse_reply_info_in()
195 			info->max_bytes = 0;  in parse_reply_info_in()
196 			info->max_files = 0;  in parse_reply_info_in()
199 		info->pool_ns_len = 0;  in parse_reply_info_in()
200 		info->pool_ns_data = NULL;  in parse_reply_info_in()
202 			ceph_decode_32_safe(p, end, info->pool_ns_len, bad);  in parse_reply_info_in()
203 			if (info->pool_ns_len > 0) {  in parse_reply_info_in()
204 				ceph_decode_need(p, end, info->pool_ns_len, bad);  in parse_reply_info_in()
205 				info->pool_ns_data = *p;  in parse_reply_info_in()
206 				*p += info->pool_ns_len;  in parse_reply_info_in()
211 			ceph_decode_need(p, end, sizeof(info->btime), bad);  in parse_reply_info_in()
212 			ceph_decode_copy(p, &info->btime, sizeof(info->btime));  in parse_reply_info_in()
213 			ceph_decode_64_safe(p, end, info->change_attr, bad);  in parse_reply_info_in()
216 		info->dir_pin = -ENODATA;  in parse_reply_info_in()
217 		/* info->snap_btime remains zero */  in parse_reply_info_in()
221 	err = -EIO;  in parse_reply_info_in()
226 static int parse_reply_info_dir(void **p, void *end,  in parse_reply_info_dir()  argument
230 	if (features == (u64)-1) {  in parse_reply_info_dir()
233 		ceph_decode_8_safe(p, end, struct_v, bad);  in parse_reply_info_dir()
234 		ceph_decode_8_safe(p, end, struct_compat, bad);  in parse_reply_info_dir()
239 		ceph_decode_32_safe(p, end, struct_len, bad);  in parse_reply_info_dir()
240 		ceph_decode_need(p, end, struct_len, bad);  in parse_reply_info_dir()
241 		end = *p + struct_len;  in parse_reply_info_dir()
244 	ceph_decode_need(p, end, sizeof(**dirfrag), bad);  in parse_reply_info_dir()
246 	*p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);  in parse_reply_info_dir()
247 	if (unlikely(*p > end))  in parse_reply_info_dir()
249 	if (features == (u64)-1)  in parse_reply_info_dir()
250 		*p = end;  in parse_reply_info_dir()
253 	return -EIO;  in parse_reply_info_dir()
256 static int parse_reply_info_lease(void **p, void *end,  in parse_reply_info_lease()  argument
260 	if (features == (u64)-1) {  in parse_reply_info_lease()
263 		ceph_decode_8_safe(p, end, struct_v, bad);  in parse_reply_info_lease()
264 		ceph_decode_8_safe(p, end, struct_compat, bad);  in parse_reply_info_lease()
269 		ceph_decode_32_safe(p, end, struct_len, bad);  in parse_reply_info_lease()
270 		ceph_decode_need(p, end, struct_len, bad);  in parse_reply_info_lease()
271 		end = *p + struct_len;  in parse_reply_info_lease()
274 	ceph_decode_need(p, end, sizeof(**lease), bad);  in parse_reply_info_lease()
277 	if (features == (u64)-1)  in parse_reply_info_lease()
278 		*p = end;  in parse_reply_info_lease()
281 	return -EIO;  in parse_reply_info_lease()
288 static int parse_reply_info_trace(void **p, void *end,  in parse_reply_info_trace()  argument
294 	if (info->head->is_dentry) {  in parse_reply_info_trace()
295 		err = parse_reply_info_in(p, end, &info->diri, features);  in parse_reply_info_trace()
299 		err = parse_reply_info_dir(p, end, &info->dirfrag, features);  in parse_reply_info_trace()
303 		ceph_decode_32_safe(p, end, info->dname_len, bad);  in parse_reply_info_trace()
304 		ceph_decode_need(p, end, info->dname_len, bad);  in parse_reply_info_trace()
305 		info->dname = *p;  in parse_reply_info_trace()
306 		*p += info->dname_len;  in parse_reply_info_trace()
308 		err = parse_reply_info_lease(p, end, &info->dlease, features);  in parse_reply_info_trace()
313 	if (info->head->is_target) {  in parse_reply_info_trace()
314 		err = parse_reply_info_in(p, end, &info->targeti, features);  in parse_reply_info_trace()
319 	if (unlikely(*p != end))  in parse_reply_info_trace()
324 	err = -EIO;  in parse_reply_info_trace()
333 static int parse_reply_info_readdir(void **p, void *end,  in parse_reply_info_readdir()  argument
340 	err = parse_reply_info_dir(p, end, &info->dir_dir, features);  in parse_reply_info_readdir()
344 	ceph_decode_need(p, end, sizeof(num) + 2, bad);  in parse_reply_info_readdir()
348 		info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);  in parse_reply_info_readdir()
349 		info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);  in parse_reply_info_readdir()
350 		info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);  in parse_reply_info_readdir()
351 		info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);  in parse_reply_info_readdir()
356 	BUG_ON(!info->dir_entries);  in parse_reply_info_readdir()
357 	if ((unsigned long)(info->dir_entries + num) >  in parse_reply_info_readdir()
358 	    (unsigned long)info->dir_entries + info->dir_buf_size) {  in parse_reply_info_readdir()
364 	info->dir_nr = num;  in parse_reply_info_readdir()
366 		struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;  in parse_reply_info_readdir()
368 		ceph_decode_32_safe(p, end, rde->name_len, bad);  in parse_reply_info_readdir()
369 		ceph_decode_need(p, end, rde->name_len, bad);  in parse_reply_info_readdir()
370 		rde->name = *p;  in parse_reply_info_readdir()
371 		*p += rde->name_len;  in parse_reply_info_readdir()
372 		dout("parsed dir dname '%.*s'\n", rde->name_len, rde->name);  in parse_reply_info_readdir()
375 		err = parse_reply_info_lease(p, end, &rde->lease, features);  in parse_reply_info_readdir()
379 		err = parse_reply_info_in(p, end, &rde->inode, features);  in parse_reply_info_readdir()
383 		rde->offset = 0;  in parse_reply_info_readdir()
385 		num--;  in parse_reply_info_readdir()
390 	*p = end;  in parse_reply_info_readdir()
394 	err = -EIO;  in parse_reply_info_readdir()
403 static int parse_reply_info_filelock(void **p, void *end,  in parse_reply_info_filelock()  argument
407 	if (*p + sizeof(*info->filelock_reply) > end)  in parse_reply_info_filelock()
410 	info->filelock_reply = *p;  in parse_reply_info_filelock()
413 	*p = end;  in parse_reply_info_filelock()
416 	return -EIO;  in parse_reply_info_filelock()
424 static int ceph_parse_deleg_inos(void **p, void *end,  in ceph_parse_deleg_inos()  argument
429 	ceph_decode_32_safe(p, end, sets, bad);  in ceph_parse_deleg_inos()
431 	while (sets--) {  in ceph_parse_deleg_inos()
434 		ceph_decode_64_safe(p, end, start, bad);  in ceph_parse_deleg_inos()
435 		ceph_decode_64_safe(p, end, len, bad);  in ceph_parse_deleg_inos()
436 		while (len--) {  in ceph_parse_deleg_inos()
437 			int err = xa_insert(&s->s_delegated_inos, ino = start++,  in ceph_parse_deleg_inos()
442 				     start - 1);  in ceph_parse_deleg_inos()
443 			} else if (err == -EBUSY) {  in ceph_parse_deleg_inos()
445 					start - 1);  in ceph_parse_deleg_inos()
453 	return -EIO;  in ceph_parse_deleg_inos()
461 	xa_for_each(&s->s_delegated_inos, ino, val) {  in ceph_get_deleg_ino()
462 		val = xa_erase(&s->s_delegated_inos, ino);  in ceph_get_deleg_ino()
471 	return xa_insert(&s->s_delegated_inos, ino, DELEGATED_INO_AVAILABLE,  in ceph_restore_deleg_ino()
476  * FIXME: xarrays can't handle 64-bit indexes on a 32-bit arch. For now, just
480 static int ceph_parse_deleg_inos(void **p, void *end,  in ceph_parse_deleg_inos()  argument
485 	ceph_decode_32_safe(p, end, sets, bad);  in ceph_parse_deleg_inos()
487 		ceph_decode_skip_n(p, end, sets * 2 * sizeof(__le64), bad);  in ceph_parse_deleg_inos()
490 	return -EIO;  in ceph_parse_deleg_inos()
507 static int parse_reply_info_create(void **p, void *end,  in parse_reply_info_create()  argument
513 	if (features == (u64)-1 ||  in parse_reply_info_create()
515 		if (*p == end) {  in parse_reply_info_create()
517 			info->has_create_ino = false;  in parse_reply_info_create()
518 		} else if (test_bit(CEPHFS_FEATURE_DELEG_INO, &s->s_features)) {  in parse_reply_info_create()
522 			info->has_create_ino = true;  in parse_reply_info_create()
523 			ceph_decode_8_safe(p, end, struct_v, bad);  in parse_reply_info_create()
524 			ceph_decode_8_safe(p, end, struct_compat, bad);  in parse_reply_info_create()
525 			ceph_decode_32_safe(p, end, len, bad);  in parse_reply_info_create()
526 			ceph_decode_64_safe(p, end, info->ino, bad);  in parse_reply_info_create()
527 			ret = ceph_parse_deleg_inos(p, end, s);  in parse_reply_info_create()
532 			ceph_decode_64_safe(p, end, info->ino, bad);  in parse_reply_info_create()
533 			info->has_create_ino = true;  in parse_reply_info_create()
536 		if (*p != end)  in parse_reply_info_create()
541 	*p = end;  in parse_reply_info_create()
544 	return -EIO;  in parse_reply_info_create()
550 static int parse_reply_info_extra(void **p, void *end,  in parse_reply_info_extra()  argument
554 	u32 op = le32_to_cpu(info->head->op);  in parse_reply_info_extra()
557 		return parse_reply_info_filelock(p, end, info, features);  in parse_reply_info_extra()
559 		return parse_reply_info_readdir(p, end, info, features);  in parse_reply_info_extra()
561 		return parse_reply_info_create(p, end, info, features, s);  in parse_reply_info_extra()
563 		return -EIO;  in parse_reply_info_extra()
573 	void *p, *end;  in parse_reply_info()  local
577 	info->head = msg->front.iov_base;  in parse_reply_info()
578 	p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);  in parse_reply_info()
579 	end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);  in parse_reply_info()
582 	ceph_decode_32_safe(&p, end, len, bad);  in parse_reply_info()
584 		ceph_decode_need(&p, end, len, bad);  in parse_reply_info()
591 	ceph_decode_32_safe(&p, end, len, bad);  in parse_reply_info()
593 		ceph_decode_need(&p, end, len, bad);  in parse_reply_info()
600 	ceph_decode_32_safe(&p, end, len, bad);  in parse_reply_info()
601 	info->snapblob_len = len;  in parse_reply_info()
602 	info->snapblob = p;  in parse_reply_info()
605 	if (p != end)  in parse_reply_info()
610 	err = -EIO;  in parse_reply_info()
618 	if (!info->dir_entries)  in destroy_reply_info()
620 	free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));  in destroy_reply_info()
645 	if (refcount_inc_not_zero(&s->s_ref)) {  in ceph_get_mds_session()
646 		dout("mdsc get_session %p %d -> %d\n", s,  in ceph_get_mds_session()
647 		     refcount_read(&s->s_ref)-1, refcount_read(&s->s_ref));  in ceph_get_mds_session()
650 		dout("mdsc get_session %p 0 -- FAIL\n", s);  in ceph_get_mds_session()
657 	dout("mdsc put_session %p %d -> %d\n", s,  in ceph_put_mds_session()
658 	     refcount_read(&s->s_ref), refcount_read(&s->s_ref)-1);  in ceph_put_mds_session()
659 	if (refcount_dec_and_test(&s->s_ref)) {  in ceph_put_mds_session()
660 		if (s->s_auth.authorizer)  in ceph_put_mds_session()
661 			ceph_auth_destroy_authorizer(s->s_auth.authorizer);  in ceph_put_mds_session()
662 		WARN_ON(mutex_is_locked(&s->s_mutex));  in ceph_put_mds_session()
663 		xa_destroy(&s->s_delegated_inos);  in ceph_put_mds_session()
669  * called under mdsc->mutex
674 	if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])  in __ceph_lookup_mds_session()
676 	return ceph_get_mds_session(mdsc->sessions[mds]);  in __ceph_lookup_mds_session()
681 	if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])  in __have_session()
690 	if (s->s_mds >= mdsc->max_sessions ||  in __verify_registered_session()
691 	    mdsc->sessions[s->s_mds] != s)  in __verify_registered_session()
692 		return -ENOENT;  in __verify_registered_session()
698  * called under mdsc->mutex.
705 	if (mds >= mdsc->mdsmap->possible_max_rank)  in register_session()
706 		return ERR_PTR(-EINVAL);  in register_session()
710 		return ERR_PTR(-ENOMEM);  in register_session()
712 	if (mds >= mdsc->max_sessions) {  in register_session()
720 		if (mdsc->sessions) {  in register_session()
721 			memcpy(sa, mdsc->sessions,  in register_session()
722 			       mdsc->max_sessions * sizeof(void *));  in register_session()
723 			kfree(mdsc->sessions);  in register_session()
725 		mdsc->sessions = sa;  in register_session()
726 		mdsc->max_sessions = newmax;  in register_session()
730 	s->s_mdsc = mdsc;  in register_session()
731 	s->s_mds = mds;  in register_session()
732 	s->s_state = CEPH_MDS_SESSION_NEW;  in register_session()
733 	s->s_ttl = 0;  in register_session()
734 	s->s_seq = 0;  in register_session()
735 	mutex_init(&s->s_mutex);  in register_session()
737 	ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);  in register_session()
739 	spin_lock_init(&s->s_gen_ttl_lock);  in register_session()
740 	s->s_cap_gen = 1;  in register_session()
741 	s->s_cap_ttl = jiffies - 1;  in register_session()
743 	spin_lock_init(&s->s_cap_lock);  in register_session()
744 	s->s_renew_requested = 0;  in register_session()
745 	s->s_renew_seq = 0;  in register_session()
746 	INIT_LIST_HEAD(&s->s_caps);  in register_session()
747 	s->s_nr_caps = 0;  in register_session()
748 	refcount_set(&s->s_ref, 1);  in register_session()
749 	INIT_LIST_HEAD(&s->s_waiting);  in register_session()
750 	INIT_LIST_HEAD(&s->s_unsafe);  in register_session()
751 	xa_init(&s->s_delegated_inos);  in register_session()
752 	s->s_num_cap_releases = 0;  in register_session()
753 	s->s_cap_reconnect = 0;  in register_session()
754 	s->s_cap_iterator = NULL;  in register_session()
755 	INIT_LIST_HEAD(&s->s_cap_releases);  in register_session()
756 	INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);  in register_session()
758 	INIT_LIST_HEAD(&s->s_cap_dirty);  in register_session()
759 	INIT_LIST_HEAD(&s->s_cap_flushing);  in register_session()
761 	mdsc->sessions[mds] = s;  in register_session()
762 	atomic_inc(&mdsc->num_sessions);  in register_session()
763 	refcount_inc(&s->s_ref);  /* one ref to sessions[], one to caller */  in register_session()
765 	ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,  in register_session()
766 		      ceph_mdsmap_get_addr(mdsc->mdsmap, mds));  in register_session()
772 	return ERR_PTR(-ENOMEM);  in register_session()
776  * called under mdsc->mutex
781 	dout("__unregister_session mds%d %p\n", s->s_mds, s);  in __unregister_session()
782 	BUG_ON(mdsc->sessions[s->s_mds] != s);  in __unregister_session()
783 	mdsc->sessions[s->s_mds] = NULL;  in __unregister_session()
784 	ceph_con_close(&s->s_con);  in __unregister_session()
786 	atomic_dec(&mdsc->num_sessions);  in __unregister_session()
792  * should be last request ref, or hold mdsc->mutex
796 	if (req->r_session) {  in put_request_session()
797 		ceph_put_mds_session(req->r_session);  in put_request_session()
798 		req->r_session = NULL;  in put_request_session()
808 	destroy_reply_info(&req->r_reply_info);  in ceph_mdsc_release_request()
809 	if (req->r_request)  in ceph_mdsc_release_request()
810 		ceph_msg_put(req->r_request);  in ceph_mdsc_release_request()
811 	if (req->r_reply)  in ceph_mdsc_release_request()
812 		ceph_msg_put(req->r_reply);  in ceph_mdsc_release_request()
813 	if (req->r_inode) {  in ceph_mdsc_release_request()
814 		ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);  in ceph_mdsc_release_request()
816 		ceph_async_iput(req->r_inode);  in ceph_mdsc_release_request()
818 	if (req->r_parent) {  in ceph_mdsc_release_request()
819 		ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);  in ceph_mdsc_release_request()
820 		ceph_async_iput(req->r_parent);  in ceph_mdsc_release_request()
822 	ceph_async_iput(req->r_target_inode);  in ceph_mdsc_release_request()
823 	if (req->r_dentry)  in ceph_mdsc_release_request()
824 		dput(req->r_dentry);  in ceph_mdsc_release_request()
825 	if (req->r_old_dentry)  in ceph_mdsc_release_request()
826 		dput(req->r_old_dentry);  in ceph_mdsc_release_request()
827 	if (req->r_old_dentry_dir) {  in ceph_mdsc_release_request()
834 		ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),  in ceph_mdsc_release_request()
836 		ceph_async_iput(req->r_old_dentry_dir);  in ceph_mdsc_release_request()
838 	kfree(req->r_path1);  in ceph_mdsc_release_request()
839 	kfree(req->r_path2);  in ceph_mdsc_release_request()
840 	if (req->r_pagelist)  in ceph_mdsc_release_request()
841 		ceph_pagelist_release(req->r_pagelist);  in ceph_mdsc_release_request()
843 	ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);  in ceph_mdsc_release_request()
844 	WARN_ON_ONCE(!list_empty(&req->r_wait));  in ceph_mdsc_release_request()
853  * called under mdsc->mutex.  in DEFINE_RB_FUNCS()
860 	req = lookup_request(&mdsc->request_tree, tid);  in DEFINE_RB_FUNCS()
868  * Register an in-flight request, and assign a tid.  Link to directory
871  * Called under mdsc->mutex.
879 	req->r_tid = ++mdsc->last_tid;  in __register_request()
880 	if (req->r_num_caps) {  in __register_request()
881 		ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,  in __register_request()
882 					req->r_num_caps);  in __register_request()
886 			/* set req->r_err to fail early from __do_request */  in __register_request()
887 			req->r_err = ret;  in __register_request()
891 	dout("__register_request %p tid %lld\n", req, req->r_tid);  in __register_request()
893 	insert_request(&mdsc->request_tree, req);  in __register_request()
895 	req->r_uid = current_fsuid();  in __register_request()
896 	req->r_gid = current_fsgid();  in __register_request()
898 	if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)  in __register_request()
899 		mdsc->oldest_tid = req->r_tid;  in __register_request()
905 		req->r_unsafe_dir = dir;  in __register_request()
906 		spin_lock(&ci->i_unsafe_lock);  in __register_request()
907 		list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);  in __register_request()
908 		spin_unlock(&ci->i_unsafe_lock);  in __register_request()
915 	dout("__unregister_request %p tid %lld\n", req, req->r_tid);  in __unregister_request()
918 	list_del_init(&req->r_unsafe_item);  in __unregister_request()
920 	if (req->r_tid == mdsc->oldest_tid) {  in __unregister_request()
921 		struct rb_node *p = rb_next(&req->r_node);  in __unregister_request()
922 		mdsc->oldest_tid = 0;  in __unregister_request()
926 			if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {  in __unregister_request()
927 				mdsc->oldest_tid = next_req->r_tid;  in __unregister_request()
934 	erase_request(&mdsc->request_tree, req);  in __unregister_request()
936 	if (req->r_unsafe_dir) {  in __unregister_request()
937 		struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);  in __unregister_request()
938 		spin_lock(&ci->i_unsafe_lock);  in __unregister_request()
939 		list_del_init(&req->r_unsafe_dir_item);  in __unregister_request()
940 		spin_unlock(&ci->i_unsafe_lock);  in __unregister_request()
942 	if (req->r_target_inode &&  in __unregister_request()
943 	    test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {  in __unregister_request()
944 		struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);  in __unregister_request()
945 		spin_lock(&ci->i_unsafe_lock);  in __unregister_request()
946 		list_del_init(&req->r_unsafe_target_item);  in __unregister_request()
947 		spin_unlock(&ci->i_unsafe_lock);  in __unregister_request()
950 	if (req->r_unsafe_dir) {  in __unregister_request()
952 		ceph_async_iput(req->r_unsafe_dir);  in __unregister_request()
953 		req->r_unsafe_dir = NULL;  in __unregister_request()
956 	complete_all(&req->r_safe_completion);  in __unregister_request()
963  * non-snapshot inode. We do this using the rcu_read_lock (which must be held
976 		dentry = dentry->d_parent;  in get_nonsnap_parent()
989  * Called under mdsc->mutex.
998 	int mode = req->r_direct_mode;  in __choose_mds()
999 	int mds = -1;  in __choose_mds()
1000 	u32 hash = req->r_direct_hash;  in __choose_mds()
1001 	bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);  in __choose_mds()
1010 	if (req->r_resend_mds >= 0 &&  in __choose_mds()
1011 	    (__have_session(mdsc, req->r_resend_mds) ||  in __choose_mds()
1012 	     ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {  in __choose_mds()
1014 		     req->r_resend_mds);  in __choose_mds()
1015 		return req->r_resend_mds;  in __choose_mds()
1022 	if (req->r_inode) {  in __choose_mds()
1023 		if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {  in __choose_mds()
1024 			inode = req->r_inode;  in __choose_mds()
1027 			/* req->r_dentry is non-null for LSSNAP request */  in __choose_mds()
1029 			inode = get_nonsnap_parent(req->r_dentry);  in __choose_mds()
1033 	} else if (req->r_dentry) {  in __choose_mds()
1039 		parent = READ_ONCE(req->r_dentry->d_parent);  in __choose_mds()
1040 		dir = req->r_parent ? : d_inode_rcu(parent);  in __choose_mds()
1042 		if (!dir || dir->i_sb != mdsc->fsc->sb) {  in __choose_mds()
1044 			inode = d_inode(req->r_dentry);  in __choose_mds()
1054 			inode = d_inode(req->r_dentry);  in __choose_mds()
1058 				hash = ceph_dentry_hash(dir, req->r_dentry);  in __choose_mds()
1073 	if (is_hash && S_ISDIR(inode->i_mode)) {  in __choose_mds()
1089 				if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=  in __choose_mds()
1091 				    !ceph_mdsmap_is_laggy(mdsc->mdsmap, mds))  in __choose_mds()
1104 				if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=  in __choose_mds()
1106 					if (!ceph_mdsmap_is_laggy(mdsc->mdsmap,  in __choose_mds()
1115 	spin_lock(&ci->i_ceph_lock);  in __choose_mds()
1118 		cap = ci->i_auth_cap;  in __choose_mds()
1119 	if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))  in __choose_mds()
1120 		cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);  in __choose_mds()
1122 		spin_unlock(&ci->i_ceph_lock);  in __choose_mds()
1126 	mds = cap->session->s_mds;  in __choose_mds()
1129 	     cap == ci->i_auth_cap ? "auth " : "", cap);  in __choose_mds()
1130 	spin_unlock(&ci->i_ceph_lock);  in __choose_mds()
1132 	/* avoid calling iput_final() while holding mdsc->mutex or  in __choose_mds()
1141 	mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);  in __choose_mds()
1161 	h = msg->front.iov_base;  in create_session_msg()
1162 	h->op = cpu_to_le32(op);  in create_session_msg()
1163 	h->seq = cpu_to_le64(seq);  in create_session_msg()
1169 #define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
1170 static int encode_supported_features(void **p, void *end)  in encode_supported_features()  argument
1178 		if (WARN_ON_ONCE(*p + 4 + size > end))  in encode_supported_features()
1179 			return -ERANGE;  in encode_supported_features()
1187 		if (WARN_ON_ONCE(*p + 4 > end))  in encode_supported_features()
1188 			return -ERANGE;  in encode_supported_features()
1197 #define METRIC_BYTES(cnt) (DIV_ROUND_UP((size_t)metric_bits[cnt - 1] + 1, 64) * 8)
1198 static int encode_metric_spec(void **p, void *end)  in encode_metric_spec()  argument
1203 	if (WARN_ON_ONCE(*p + 2 > end))  in encode_metric_spec()
1204 		return -ERANGE;  in encode_metric_spec()
1213 		if (WARN_ON_ONCE(*p + 4 + 4 + size > end))  in encode_metric_spec()
1214 			return -ERANGE;  in encode_metric_spec()
1226 		if (WARN_ON_ONCE(*p + 4 + 4 > end))  in encode_metric_spec()
1227 			return -ERANGE;  in encode_metric_spec()
1246 	int i = -1;  in create_session_open_msg()
1249 	struct ceph_options *opt = mdsc->fsc->client->options;  in create_session_open_msg()
1250 	struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;  in create_session_open_msg()
1252 	void *p, *end;  in create_session_open_msg()  local
1256 		{"hostname", mdsc->nodename},  in create_session_open_msg()
1257 		{"kernel_version", init_utsname()->release},  in create_session_open_msg()
1258 		{"entity_id", opt->name ? : ""},  in create_session_open_msg()
1259 		{"root", fsopt->server_path ? : "/"},  in create_session_open_msg()
1290 		return ERR_PTR(-ENOMEM);  in create_session_open_msg()
1292 	p = msg->front.iov_base;  in create_session_open_msg()
1293 	end = p + msg->front.iov_len;  in create_session_open_msg()
1296 	h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);  in create_session_open_msg()
1297 	h->seq = cpu_to_le64(seq);  in create_session_open_msg()
1305 	msg->hdr.version = cpu_to_le16(4);  in create_session_open_msg()
1306 	msg->hdr.compat_version = cpu_to_le16(1);  in create_session_open_msg()
1314 	/* Two length-prefixed strings for each entry in the map */  in create_session_open_msg()
1327 	ret = encode_supported_features(&p, end);  in create_session_open_msg()
1334 	ret = encode_metric_spec(&p, end);  in create_session_open_msg()
1341 	msg->front.iov_len = p - msg->front.iov_base;  in create_session_open_msg()
1342 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);  in create_session_open_msg()
1350  * called under mdsc->mutex
1357 	int mds = session->s_mds;  in __open_session()
1360 	mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);  in __open_session()
1363 	session->s_state = CEPH_MDS_SESSION_OPENING;  in __open_session()
1364 	session->s_renew_requested = jiffies;  in __open_session()
1367 	msg = create_session_open_msg(mdsc, session->s_seq);  in __open_session()
1370 	ceph_con_send(&session->s_con, msg);  in __open_session()
1377  * called under mdsc->mutex
1391 	if (session->s_state == CEPH_MDS_SESSION_NEW ||  in __open_export_target_session()
1392 	    session->s_state == CEPH_MDS_SESSION_CLOSING) {  in __open_export_target_session()
1408 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_open_export_target_session()
1410 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_open_export_target_session()
1420 	int i, mds = session->s_mds;  in __open_export_target_sessions()
1422 	if (mds >= mdsc->mdsmap->possible_max_rank)  in __open_export_target_sessions()
1425 	mi = &mdsc->mdsmap->m_info[mds];  in __open_export_target_sessions()
1427 	     session->s_mds, mi->num_export_targets);  in __open_export_target_sessions()
1429 	for (i = 0; i < mi->num_export_targets; i++) {  in __open_export_target_sessions()
1430 		ts = __open_export_target_session(mdsc, mi->export_targets[i]);  in __open_export_target_sessions()
1439 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_open_export_target_sessions()
1441 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_open_export_target_sessions()
1451 	lockdep_assert_held(&session->s_cap_lock);  in detach_cap_releases()
1453 	list_splice_init(&session->s_cap_releases, target);  in detach_cap_releases()
1454 	session->s_num_cap_releases = 0;  in detach_cap_releases()
1455 	dout("dispose_cap_releases mds%d\n", session->s_mds);  in detach_cap_releases()
1463 		/* zero out the in-progress message */  in dispose_cap_releases()
1465 		list_del(&cap->session_caps);  in dispose_cap_releases()
1477 	dout("cleanup_session_requests mds%d\n", session->s_mds);  in cleanup_session_requests()
1478 	mutex_lock(&mdsc->mutex);  in cleanup_session_requests()
1479 	while (!list_empty(&session->s_unsafe)) {  in cleanup_session_requests()
1480 		req = list_first_entry(&session->s_unsafe,  in cleanup_session_requests()
1483 				    req->r_tid);  in cleanup_session_requests()
1484 		if (req->r_target_inode) {  in cleanup_session_requests()
1486 			ci = ceph_inode(req->r_target_inode);  in cleanup_session_requests()
1487 			errseq_set(&ci->i_meta_err, -EIO);  in cleanup_session_requests()
1489 		if (req->r_unsafe_dir) {  in cleanup_session_requests()
1491 			ci = ceph_inode(req->r_unsafe_dir);  in cleanup_session_requests()
1492 			errseq_set(&ci->i_meta_err, -EIO);  in cleanup_session_requests()
1496 	/* zero r_attempts, so kick_requests() will re-send requests */  in cleanup_session_requests()
1497 	p = rb_first(&mdsc->request_tree);  in cleanup_session_requests()
1501 		if (req->r_session &&  in cleanup_session_requests()
1502 		    req->r_session->s_mds == session->s_mds)  in cleanup_session_requests()
1503 			req->r_attempts = 0;  in cleanup_session_requests()
1505 	mutex_unlock(&mdsc->mutex);  in cleanup_session_requests()
1524 	dout("iterate_session_caps %p mds%d\n", session, session->s_mds);  in ceph_iterate_session_caps()
1525 	spin_lock(&session->s_cap_lock);  in ceph_iterate_session_caps()
1526 	p = session->s_caps.next;  in ceph_iterate_session_caps()
1527 	while (p != &session->s_caps) {  in ceph_iterate_session_caps()
1529 		inode = igrab(&cap->ci->vfs_inode);  in ceph_iterate_session_caps()
1531 			p = p->next;  in ceph_iterate_session_caps()
1534 		session->s_cap_iterator = cap;  in ceph_iterate_session_caps()
1535 		spin_unlock(&session->s_cap_lock);  in ceph_iterate_session_caps()
1544 			ceph_put_cap(session->s_mdsc, old_cap);  in ceph_iterate_session_caps()
1551 		spin_lock(&session->s_cap_lock);  in ceph_iterate_session_caps()
1552 		p = p->next;  in ceph_iterate_session_caps()
1553 		if (!cap->ci) {  in ceph_iterate_session_caps()
1556 			BUG_ON(cap->session != session);  in ceph_iterate_session_caps()
1557 			cap->session = NULL;  in ceph_iterate_session_caps()
1558 			list_del_init(&cap->session_caps);  in ceph_iterate_session_caps()
1559 			session->s_nr_caps--;  in ceph_iterate_session_caps()
1560 			atomic64_dec(&session->s_mdsc->metric.total_caps);  in ceph_iterate_session_caps()
1561 			if (cap->queue_release)  in ceph_iterate_session_caps()
1571 	session->s_cap_iterator = NULL;  in ceph_iterate_session_caps()
1572 	spin_unlock(&session->s_cap_lock);  in ceph_iterate_session_caps()
1576 		ceph_put_cap(session->s_mdsc, old_cap);  in ceph_iterate_session_caps()
1591 	     cap, ci, &ci->vfs_inode);  in remove_session_caps_cb()
1592 	spin_lock(&ci->i_ceph_lock);  in remove_session_caps_cb()
1594 	if (!ci->i_auth_cap) {  in remove_session_caps_cb()
1596 		struct ceph_mds_client *mdsc = fsc->mdsc;  in remove_session_caps_cb()
1598 		if (READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {  in remove_session_caps_cb()
1599 			if (inode->i_data.nrpages > 0)  in remove_session_caps_cb()
1601 			if (ci->i_wrbuffer_ref > 0)  in remove_session_caps_cb()
1602 				mapping_set_error(&inode->i_data, -EIO);  in remove_session_caps_cb()
1605 		while (!list_empty(&ci->i_cap_flush_list)) {  in remove_session_caps_cb()
1606 			cf = list_first_entry(&ci->i_cap_flush_list,  in remove_session_caps_cb()
1608 			list_move(&cf->i_list, &to_remove);  in remove_session_caps_cb()
1611 		spin_lock(&mdsc->cap_dirty_lock);  in remove_session_caps_cb()
1614 			list_del(&cf->g_list);  in remove_session_caps_cb()
1616 		if (!list_empty(&ci->i_dirty_item)) {  in remove_session_caps_cb()
1619 				ceph_cap_string(ci->i_dirty_caps),  in remove_session_caps_cb()
1621 			ci->i_dirty_caps = 0;  in remove_session_caps_cb()
1622 			list_del_init(&ci->i_dirty_item);  in remove_session_caps_cb()
1625 		if (!list_empty(&ci->i_flushing_item)) {  in remove_session_caps_cb()
1628 				ceph_cap_string(ci->i_flushing_caps),  in remove_session_caps_cb()
1630 			ci->i_flushing_caps = 0;  in remove_session_caps_cb()
1631 			list_del_init(&ci->i_flushing_item);  in remove_session_caps_cb()
1632 			mdsc->num_cap_flushing--;  in remove_session_caps_cb()
1635 		spin_unlock(&mdsc->cap_dirty_lock);  in remove_session_caps_cb()
1638 			errseq_set(&ci->i_meta_err, -EIO);  in remove_session_caps_cb()
1640 			if (ci->i_wrbuffer_ref_head == 0 &&  in remove_session_caps_cb()
1641 			    ci->i_wr_ref == 0 &&  in remove_session_caps_cb()
1642 			    ci->i_dirty_caps == 0 &&  in remove_session_caps_cb()
1643 			    ci->i_flushing_caps == 0) {  in remove_session_caps_cb()
1644 				ceph_put_snap_context(ci->i_head_snapc);  in remove_session_caps_cb()
1645 				ci->i_head_snapc = NULL;  in remove_session_caps_cb()
1649 		if (atomic_read(&ci->i_filelock_ref) > 0) {  in remove_session_caps_cb()
1650 			/* make further file lock syscall return -EIO */  in remove_session_caps_cb()
1651 			ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;  in remove_session_caps_cb()
1656 		if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {  in remove_session_caps_cb()
1657 			list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove);  in remove_session_caps_cb()
1658 			ci->i_prealloc_cap_flush = NULL;  in remove_session_caps_cb()
1661 	spin_unlock(&ci->i_ceph_lock);  in remove_session_caps_cb()
1666 		list_del(&cf->i_list);  in remove_session_caps_cb()
1670 	wake_up_all(&ci->i_cap_wq);  in remove_session_caps_cb()
1683 	struct ceph_fs_client *fsc = session->s_mdsc->fsc;  in remove_session_caps()
1684 	struct super_block *sb = fsc->sb;  in remove_session_caps()
1690 	wake_up_all(&fsc->mdsc->cap_flushing_wq);  in remove_session_caps()
1692 	spin_lock(&session->s_cap_lock);  in remove_session_caps()
1693 	if (session->s_nr_caps > 0) {  in remove_session_caps()
1704 		while (!list_empty(&session->s_caps)) {  in remove_session_caps()
1705 			cap = list_entry(session->s_caps.next,  in remove_session_caps()
1710 			vino = cap->ci->i_vino;  in remove_session_caps()
1711 			spin_unlock(&session->s_cap_lock);  in remove_session_caps()
1717 			spin_lock(&session->s_cap_lock);  in remove_session_caps()
1724 	BUG_ON(session->s_nr_caps > 0);  in remove_session_caps()
1725 	BUG_ON(!list_empty(&session->s_cap_flushing));  in remove_session_caps()
1726 	spin_unlock(&session->s_cap_lock);  in remove_session_caps()
1727 	dispose_cap_releases(session->s_mdsc, &dispose);  in remove_session_caps()
1749 		spin_lock(&ci->i_ceph_lock);  in wake_up_session_cb()
1750 		ci->i_wanted_max_size = 0;  in wake_up_session_cb()
1751 		ci->i_requested_max_size = 0;  in wake_up_session_cb()
1752 		spin_unlock(&ci->i_ceph_lock);  in wake_up_session_cb()
1754 		if (cap->cap_gen < cap->session->s_cap_gen) {  in wake_up_session_cb()
1755 			/* mds did not re-issue stale cap */  in wake_up_session_cb()
1756 			spin_lock(&ci->i_ceph_lock);  in wake_up_session_cb()
1757 			cap->issued = cap->implemented = CEPH_CAP_PIN;  in wake_up_session_cb()
1758 			spin_unlock(&ci->i_ceph_lock);  in wake_up_session_cb()
1762 	wake_up_all(&ci->i_cap_wq);  in wake_up_session_cb()
1768 	dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);  in wake_up_session_caps()
1785 	if (time_after_eq(jiffies, session->s_cap_ttl) &&  in send_renew_caps()
1786 	    time_after_eq(session->s_cap_ttl, session->s_renew_requested))  in send_renew_caps()
1787 		pr_info("mds%d caps stale\n", session->s_mds);  in send_renew_caps()
1788 	session->s_renew_requested = jiffies;  in send_renew_caps()
1792 	state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);  in send_renew_caps()
1795 		     session->s_mds, ceph_mds_state_name(state));  in send_renew_caps()
1799 	dout("send_renew_caps to mds%d (%s)\n", session->s_mds,  in send_renew_caps()
1802 				 ++session->s_renew_seq);  in send_renew_caps()
1804 		return -ENOMEM;  in send_renew_caps()
1805 	ceph_con_send(&session->s_con, msg);  in send_renew_caps()
1815 	     session->s_mds, ceph_session_state_name(session->s_state), seq);  in send_flushmsg_ack()
1818 		return -ENOMEM;  in send_flushmsg_ack()
1819 	ceph_con_send(&session->s_con, msg);  in send_flushmsg_ack()
1825  * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1827  * Called under session->s_mutex
1835 	spin_lock(&session->s_cap_lock);  in renewed_caps()
1836 	was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);  in renewed_caps()
1838 	session->s_cap_ttl = session->s_renew_requested +  in renewed_caps()
1839 		mdsc->mdsmap->m_session_timeout*HZ;  in renewed_caps()
1842 		if (time_before(jiffies, session->s_cap_ttl)) {  in renewed_caps()
1843 			pr_info("mds%d caps renewed\n", session->s_mds);  in renewed_caps()
1846 			pr_info("mds%d caps still stale\n", session->s_mds);  in renewed_caps()
1850 	     session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",  in renewed_caps()
1851 	     time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");  in renewed_caps()
1852 	spin_unlock(&session->s_cap_lock);  in renewed_caps()
1866 	     session->s_mds, ceph_session_state_name(session->s_state),  in request_close_session()
1867 	     session->s_seq);  in request_close_session()
1868 	msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);  in request_close_session()
1870 		return -ENOMEM;  in request_close_session()
1871 	ceph_con_send(&session->s_con, msg);  in request_close_session()
1881 	if (session->s_state >= CEPH_MDS_SESSION_CLOSING)  in __close_session()
1883 	session->s_state = CEPH_MDS_SESSION_CLOSING;  in __close_session()
1895 	spin_lock(&dentry->d_lock);  in drop_negative_children()
1896 	list_for_each_entry(child, &dentry->d_subdirs, d_child) {  in drop_negative_children()
1902 	spin_unlock(&dentry->d_lock);  in drop_negative_children()
1927 		return -1;  in trim_caps_cb()
1929 	spin_lock(&ci->i_ceph_lock);  in trim_caps_cb()
1930 	mine = cap->issued | cap->implemented;  in trim_caps_cb()
1938 	if (cap == ci->i_auth_cap) {  in trim_caps_cb()
1939 		if (ci->i_dirty_caps || ci->i_flushing_caps ||  in trim_caps_cb()
1940 		    !list_empty(&ci->i_cap_snaps))  in trim_caps_cb()
1944 		/* Note: it's possible that i_filelock_ref becomes non-zero  in trim_caps_cb()
1946 		 * of lock mds request will re-add auth caps. */  in trim_caps_cb()
1947 		if (atomic_read(&ci->i_filelock_ref) > 0)  in trim_caps_cb()
1952 	if (S_ISREG(inode->i_mode) &&  in trim_caps_cb()
1964 		(*remaining)--;  in trim_caps_cb()
1968 		spin_unlock(&ci->i_ceph_lock);  in trim_caps_cb()
1974 			count = atomic_read(&inode->i_count);  in trim_caps_cb()
1976 				(*remaining)--;  in trim_caps_cb()
1986 	spin_unlock(&ci->i_ceph_lock);  in trim_caps_cb()
1997 	int trim_caps = session->s_nr_caps - max_caps;  in ceph_trim_caps()
2000 	     session->s_mds, session->s_nr_caps, max_caps, trim_caps);  in ceph_trim_caps()
2006 		     session->s_mds, session->s_nr_caps, max_caps,  in ceph_trim_caps()
2007 			trim_caps - remaining);  in ceph_trim_caps()
2019 	spin_lock(&mdsc->cap_dirty_lock);  in check_caps_flush()
2020 	if (!list_empty(&mdsc->cap_flush_list)) {  in check_caps_flush()
2022 			list_first_entry(&mdsc->cap_flush_list,  in check_caps_flush()
2024 		if (cf->tid <= want_flush_tid) {  in check_caps_flush()
2026 			     "%llu <= %llu\n", cf->tid, want_flush_tid);  in check_caps_flush()
2030 	spin_unlock(&mdsc->cap_dirty_lock);  in check_caps_flush()
2044 	wait_event(mdsc->cap_flushing_wq,  in wait_caps_flush()
2059 	struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;  in ceph_send_cap_releases()
2065 	down_read(&osdc->lock);  in ceph_send_cap_releases()
2066 	barrier = cpu_to_le32(osdc->epoch_barrier);  in ceph_send_cap_releases()
2067 	up_read(&osdc->lock);  in ceph_send_cap_releases()
2069 	spin_lock(&session->s_cap_lock);  in ceph_send_cap_releases()
2071 	list_splice_init(&session->s_cap_releases, &tmp_list);  in ceph_send_cap_releases()
2072 	num_cap_releases = session->s_num_cap_releases;  in ceph_send_cap_releases()
2073 	session->s_num_cap_releases = 0;  in ceph_send_cap_releases()
2074 	spin_unlock(&session->s_cap_lock);  in ceph_send_cap_releases()
2082 			head = msg->front.iov_base;  in ceph_send_cap_releases()
2083 			head->num = cpu_to_le32(0);  in ceph_send_cap_releases()
2084 			msg->front.iov_len = sizeof(*head);  in ceph_send_cap_releases()
2086 			msg->hdr.version = cpu_to_le16(2);  in ceph_send_cap_releases()
2087 			msg->hdr.compat_version = cpu_to_le16(1);  in ceph_send_cap_releases()
2092 		list_del(&cap->session_caps);  in ceph_send_cap_releases()
2093 		num_cap_releases--;  in ceph_send_cap_releases()
2095 		head = msg->front.iov_base;  in ceph_send_cap_releases()
2096 		put_unaligned_le32(get_unaligned_le32(&head->num) + 1,  in ceph_send_cap_releases()
2097 				   &head->num);  in ceph_send_cap_releases()
2098 		item = msg->front.iov_base + msg->front.iov_len;  in ceph_send_cap_releases()
2099 		item->ino = cpu_to_le64(cap->cap_ino);  in ceph_send_cap_releases()
2100 		item->cap_id = cpu_to_le64(cap->cap_id);  in ceph_send_cap_releases()
2101 		item->migrate_seq = cpu_to_le32(cap->mseq);  in ceph_send_cap_releases()
2102 		item->seq = cpu_to_le32(cap->issue_seq);  in ceph_send_cap_releases()
2103 		msg->front.iov_len += sizeof(*item);  in ceph_send_cap_releases()
2107 		if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {  in ceph_send_cap_releases()
2109 			cap_barrier = msg->front.iov_base + msg->front.iov_len;  in ceph_send_cap_releases()
2111 			msg->front.iov_len += sizeof(*cap_barrier);  in ceph_send_cap_releases()
2113 			msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);  in ceph_send_cap_releases()
2114 			dout("send_cap_releases mds%d %p\n", session->s_mds, msg);  in ceph_send_cap_releases()
2115 			ceph_con_send(&session->s_con, msg);  in ceph_send_cap_releases()
2122 	spin_lock(&session->s_cap_lock);  in ceph_send_cap_releases()
2123 	if (!list_empty(&session->s_cap_releases))  in ceph_send_cap_releases()
2125 	spin_unlock(&session->s_cap_lock);  in ceph_send_cap_releases()
2129 		cap_barrier = msg->front.iov_base + msg->front.iov_len;  in ceph_send_cap_releases()
2131 		msg->front.iov_len += sizeof(*cap_barrier);  in ceph_send_cap_releases()
2133 		msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);  in ceph_send_cap_releases()
2134 		dout("send_cap_releases mds%d %p\n", session->s_mds, msg);  in ceph_send_cap_releases()
2135 		ceph_con_send(&session->s_con, msg);  in ceph_send_cap_releases()
2140 		session->s_mds);  in ceph_send_cap_releases()
2141 	spin_lock(&session->s_cap_lock);  in ceph_send_cap_releases()
2142 	list_splice(&tmp_list, &session->s_cap_releases);  in ceph_send_cap_releases()
2143 	session->s_num_cap_releases += num_cap_releases;  in ceph_send_cap_releases()
2144 	spin_unlock(&session->s_cap_lock);  in ceph_send_cap_releases()
2152 	mutex_lock(&session->s_mutex);  in ceph_cap_release_work()
2153 	if (session->s_state == CEPH_MDS_SESSION_OPEN ||  in ceph_cap_release_work()
2154 	    session->s_state == CEPH_MDS_SESSION_HUNG)  in ceph_cap_release_work()
2155 		ceph_send_cap_releases(session->s_mdsc, session);  in ceph_cap_release_work()
2156 	mutex_unlock(&session->s_mutex);  in ceph_cap_release_work()
2163 	if (mdsc->stopping)  in ceph_flush_cap_releases()
2167 	if (queue_work(mdsc->fsc->cap_wq,  in ceph_flush_cap_releases()
2168 		       &session->s_cap_release_work)) {  in ceph_flush_cap_releases()
2177  * caller holds session->s_cap_lock
2182 	list_add_tail(&cap->session_caps, &session->s_cap_releases);  in __ceph_queue_cap_release()
2183 	session->s_num_cap_releases++;  in __ceph_queue_cap_release()
2185 	if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))  in __ceph_queue_cap_release()
2186 		ceph_flush_cap_releases(session->s_mdsc, session);  in __ceph_queue_cap_release()
2194 	if (ret == -EAGAIN)  in ceph_cap_reclaim_work()
2200 	if (mdsc->stopping)  in ceph_queue_cap_reclaim_work()
2203         if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {  in ceph_queue_cap_reclaim_work()
2215 	val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);  in ceph_reclaim_caps_nr()
2217 		atomic_set(&mdsc->cap_reclaim_pending, 0);  in ceph_reclaim_caps_nr()
2230 	struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;  in ceph_alloc_readdir_reply_buffer()
2231 	struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;  in ceph_alloc_readdir_reply_buffer()
2236 	spin_lock(&ci->i_ceph_lock);  in ceph_alloc_readdir_reply_buffer()
2237 	num_entries = ci->i_files + ci->i_subdirs;  in ceph_alloc_readdir_reply_buffer()
2238 	spin_unlock(&ci->i_ceph_lock);  in ceph_alloc_readdir_reply_buffer()
2240 	num_entries = min(num_entries, opt->max_readdir);  in ceph_alloc_readdir_reply_buffer()
2244 		rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |  in ceph_alloc_readdir_reply_buffer()
2247 		if (rinfo->dir_entries)  in ceph_alloc_readdir_reply_buffer()
2249 		order--;  in ceph_alloc_readdir_reply_buffer()
2251 	if (!rinfo->dir_entries)  in ceph_alloc_readdir_reply_buffer()
2252 		return -ENOMEM;  in ceph_alloc_readdir_reply_buffer()
2255 	num_entries = min(num_entries, opt->max_readdir);  in ceph_alloc_readdir_reply_buffer()
2257 	rinfo->dir_buf_size = PAGE_SIZE << order;  in ceph_alloc_readdir_reply_buffer()
2258 	req->r_num_caps = num_entries + 1;  in ceph_alloc_readdir_reply_buffer()
2259 	req->r_args.readdir.max_entries = cpu_to_le32(num_entries);  in ceph_alloc_readdir_reply_buffer()
2260 	req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);  in ceph_alloc_readdir_reply_buffer()
2274 		return ERR_PTR(-ENOMEM);  in ceph_mdsc_create_request()
2276 	mutex_init(&req->r_fill_mutex);  in ceph_mdsc_create_request()
2277 	req->r_mdsc = mdsc;  in ceph_mdsc_create_request()
2278 	req->r_started = jiffies;  in ceph_mdsc_create_request()
2279 	req->r_start_latency = ktime_get();  in ceph_mdsc_create_request()
2280 	req->r_resend_mds = -1;  in ceph_mdsc_create_request()
2281 	INIT_LIST_HEAD(&req->r_unsafe_dir_item);  in ceph_mdsc_create_request()
2282 	INIT_LIST_HEAD(&req->r_unsafe_target_item);  in ceph_mdsc_create_request()
2283 	req->r_fmode = -1;  in ceph_mdsc_create_request()
2284 	kref_init(&req->r_kref);  in ceph_mdsc_create_request()
2285 	RB_CLEAR_NODE(&req->r_node);  in ceph_mdsc_create_request()
2286 	INIT_LIST_HEAD(&req->r_wait);  in ceph_mdsc_create_request()
2287 	init_completion(&req->r_completion);  in ceph_mdsc_create_request()
2288 	init_completion(&req->r_safe_completion);  in ceph_mdsc_create_request()
2289 	INIT_LIST_HEAD(&req->r_unsafe_item);  in ceph_mdsc_create_request()
2291 	ktime_get_coarse_real_ts64(&req->r_stamp);  in ceph_mdsc_create_request()
2293 	req->r_op = op;  in ceph_mdsc_create_request()
2294 	req->r_direct_mode = mode;  in ceph_mdsc_create_request()
2301  * called under mdsc->mutex.
2305 	if (RB_EMPTY_ROOT(&mdsc->request_tree))  in __get_oldest_req()
2307 	return rb_entry(rb_first(&mdsc->request_tree),  in __get_oldest_req()
2313 	return mdsc->oldest_tid;  in __get_oldest_tid()
2320  * If @stop_on_nosnap, generate path relative to the first non-snapped
2324  *   foo/.snap/bar -> foo//bar
2336 		return ERR_PTR(-EINVAL);  in ceph_mdsc_build_path()
2340 		return ERR_PTR(-ENOMEM);  in ceph_mdsc_build_path()
2342 	pos = PATH_MAX - 1;  in ceph_mdsc_build_path()
2351 		spin_lock(&temp->d_lock);  in ceph_mdsc_build_path()
2358 			spin_unlock(&temp->d_lock);  in ceph_mdsc_build_path()
2362 			pos -= temp->d_name.len;  in ceph_mdsc_build_path()
2364 				spin_unlock(&temp->d_lock);  in ceph_mdsc_build_path()
2367 			memcpy(path + pos, temp->d_name.name, temp->d_name.len);  in ceph_mdsc_build_path()
2369 		spin_unlock(&temp->d_lock);  in ceph_mdsc_build_path()
2370 		temp = READ_ONCE(temp->d_parent);  in ceph_mdsc_build_path()
2377 		if (--pos < 0)  in ceph_mdsc_build_path()
2390 		 * A rename didn't occur, but somehow we didn't end up where  in ceph_mdsc_build_path()
2393 		pr_warn("build_path did not end path lookup where "  in ceph_mdsc_build_path()
2399 	*plen = PATH_MAX - 1 - pos;  in ceph_mdsc_build_path()
2413 		dir = d_inode_rcu(dentry->d_parent);  in build_dentry_path()
2417 		*ppath = dentry->d_name.name;  in build_dentry_path()
2418 		*ppathlen = dentry->d_name.len;  in build_dentry_path()
2483  * called under mdsc->mutex
2498 	void *p, *end;  in create_request_message()  local
2501 	ret = set_request_path_attr(req->r_inode, req->r_dentry,  in create_request_message()
2502 			      req->r_parent, req->r_path1, req->r_ino1.ino,  in create_request_message()
2505 					&req->r_req_flags));  in create_request_message()
2512 	ret = set_request_path_attr(NULL, req->r_old_dentry,  in create_request_message()
2513 			      req->r_old_dentry_dir,  in create_request_message()
2514 			      req->r_path2, req->r_ino2.ino,  in create_request_message()
2527 		(!!req->r_inode_drop + !!req->r_dentry_drop +  in create_request_message()
2528 		 !!req->r_old_inode_drop + !!req->r_old_dentry_drop);  in create_request_message()
2529 	if (req->r_dentry_drop)  in create_request_message()
2531 	if (req->r_old_dentry_drop)  in create_request_message()
2536 		msg = ERR_PTR(-ENOMEM);  in create_request_message()
2540 	msg->hdr.version = cpu_to_le16(2);  in create_request_message()
2541 	msg->hdr.tid = cpu_to_le64(req->r_tid);  in create_request_message()
2543 	head = msg->front.iov_base;  in create_request_message()
2544 	p = msg->front.iov_base + sizeof(*head);  in create_request_message()
2545 	end = msg->front.iov_base + msg->front.iov_len;  in create_request_message()
2547 	head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);  in create_request_message()
2548 	head->op = cpu_to_le32(req->r_op);  in create_request_message()
2549 	head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns, req->r_uid));  in create_request_message()
2550 	head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns, req->r_gid));  in create_request_message()
2551 	head->ino = cpu_to_le64(req->r_deleg_ino);  in create_request_message()
2552 	head->args = req->r_args;  in create_request_message()
2554 	ceph_encode_filepath(&p, end, ino1, path1);  in create_request_message()
2555 	ceph_encode_filepath(&p, end, ino2, path2);  in create_request_message()
2558 	req->r_request_release_offset = p - msg->front.iov_base;  in create_request_message()
2562 	if (req->r_inode_drop)  in create_request_message()
2564 		      req->r_inode ? req->r_inode : d_inode(req->r_dentry),  in create_request_message()
2565 		      mds, req->r_inode_drop, req->r_inode_unless,  in create_request_message()
2566 		      req->r_op == CEPH_MDS_OP_READDIR);  in create_request_message()
2567 	if (req->r_dentry_drop)  in create_request_message()
2568 		releases += ceph_encode_dentry_release(&p, req->r_dentry,  in create_request_message()
2569 				req->r_parent, mds, req->r_dentry_drop,  in create_request_message()
2570 				req->r_dentry_unless);  in create_request_message()
2571 	if (req->r_old_dentry_drop)  in create_request_message()
2572 		releases += ceph_encode_dentry_release(&p, req->r_old_dentry,  in create_request_message()
2573 				req->r_old_dentry_dir, mds,  in create_request_message()
2574 				req->r_old_dentry_drop,  in create_request_message()
2575 				req->r_old_dentry_unless);  in create_request_message()
2576 	if (req->r_old_inode_drop)  in create_request_message()
2578 		      d_inode(req->r_old_dentry),  in create_request_message()
2579 		      mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);  in create_request_message()
2583 		p = msg->front.iov_base + req->r_request_release_offset;  in create_request_message()
2586 	head->num_releases = cpu_to_le16(releases);  in create_request_message()
2591 		ceph_encode_timespec64(&ts, &req->r_stamp);  in create_request_message()
2595 	if (WARN_ON_ONCE(p > end)) {  in create_request_message()
2597 		msg = ERR_PTR(-ERANGE);  in create_request_message()
2601 	msg->front.iov_len = p - msg->front.iov_base;  in create_request_message()
2602 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);  in create_request_message()
2604 	if (req->r_pagelist) {  in create_request_message()
2605 		struct ceph_pagelist *pagelist = req->r_pagelist;  in create_request_message()
2607 		msg->hdr.data_len = cpu_to_le32(pagelist->length);  in create_request_message()
2609 		msg->hdr.data_len = 0;  in create_request_message()
2612 	msg->hdr.data_off = cpu_to_le16(0);  in create_request_message()
2625  * called under mdsc->mutex if error, under no mutex if
2631 	req->r_end_latency = ktime_get();  in complete_request()
2633 	if (req->r_callback)  in complete_request()
2634 		req->r_callback(mdsc, req);  in complete_request()
2635 	complete_all(&req->r_completion);  in complete_request()
2639  * called under mdsc->mutex
2649 	req->r_attempts++;  in __prepare_send_request()
2650 	if (req->r_inode) {  in __prepare_send_request()
2652 			ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);  in __prepare_send_request()
2655 			req->r_sent_on_mseq = cap->mseq;  in __prepare_send_request()
2657 			req->r_sent_on_mseq = -1;  in __prepare_send_request()
2660 	     req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);  in __prepare_send_request()
2662 	if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {  in __prepare_send_request()
2670 		msg = req->r_request;  in __prepare_send_request()
2671 		rhead = msg->front.iov_base;  in __prepare_send_request()
2673 		flags = le32_to_cpu(rhead->flags);  in __prepare_send_request()
2675 		rhead->flags = cpu_to_le32(flags);  in __prepare_send_request()
2677 		if (req->r_target_inode)  in __prepare_send_request()
2678 			rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));  in __prepare_send_request()
2680 		rhead->num_retry = req->r_attempts - 1;  in __prepare_send_request()
2683 		rhead->num_releases = 0;  in __prepare_send_request()
2686 		p = msg->front.iov_base + req->r_request_release_offset;  in __prepare_send_request()
2689 			ceph_encode_timespec64(&ts, &req->r_stamp);  in __prepare_send_request()
2693 		msg->front.iov_len = p - msg->front.iov_base;  in __prepare_send_request()
2694 		msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);  in __prepare_send_request()
2698 	if (req->r_request) {  in __prepare_send_request()
2699 		ceph_msg_put(req->r_request);  in __prepare_send_request()
2700 		req->r_request = NULL;  in __prepare_send_request()
2704 		req->r_err = PTR_ERR(msg);  in __prepare_send_request()
2707 	req->r_request = msg;  in __prepare_send_request()
2709 	rhead = msg->front.iov_base;  in __prepare_send_request()
2710 	rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));  in __prepare_send_request()
2711 	if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))  in __prepare_send_request()
2713 	if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))  in __prepare_send_request()
2715 	if (req->r_parent)  in __prepare_send_request()
2717 	rhead->flags = cpu_to_le32(flags);  in __prepare_send_request()
2718 	rhead->num_fwd = req->r_num_fwd;  in __prepare_send_request()
2719 	rhead->num_retry = req->r_attempts - 1;  in __prepare_send_request()
2721 	dout(" r_parent = %p\n", req->r_parent);  in __prepare_send_request()
2726  * called under mdsc->mutex
2735 	err = __prepare_send_request(mdsc, req, session->s_mds,  in __send_request()
2738 		ceph_msg_get(req->r_request);  in __send_request()
2739 		ceph_con_send(&session->s_con, req->r_request);  in __send_request()
2752 	int mds = -1;  in __do_request()
2756 	if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {  in __do_request()
2757 		if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))  in __do_request()
2762 	if (req->r_timeout &&  in __do_request()
2763 	    time_after_eq(jiffies, req->r_started + req->r_timeout)) {  in __do_request()
2765 		err = -ETIMEDOUT;  in __do_request()
2768 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {  in __do_request()
2770 		err = -EIO;  in __do_request()
2773 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {  in __do_request()
2774 		if (mdsc->mdsmap_err) {  in __do_request()
2775 			err = mdsc->mdsmap_err;  in __do_request()
2779 		if (mdsc->mdsmap->m_epoch == 0) {  in __do_request()
2781 			list_add(&req->r_wait, &mdsc->waiting_for_map);  in __do_request()
2784 		if (!(mdsc->fsc->mount_options->flags &  in __do_request()
2786 		    !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {  in __do_request()
2787 			err = -EHOSTUNREACH;  in __do_request()
2796 	    ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {  in __do_request()
2797 		if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {  in __do_request()
2798 			err = -EJUKEBOX;  in __do_request()
2802 		list_add(&req->r_wait, &mdsc->waiting_for_map);  in __do_request()
2815 	req->r_session = ceph_get_mds_session(session);  in __do_request()
2818 	     ceph_session_state_name(session->s_state));  in __do_request()
2819 	if (session->s_state != CEPH_MDS_SESSION_OPEN &&  in __do_request()
2820 	    session->s_state != CEPH_MDS_SESSION_HUNG) {  in __do_request()
2821 		if (session->s_state == CEPH_MDS_SESSION_REJECTED) {  in __do_request()
2822 			err = -EACCES;  in __do_request()
2827 		 * inodes are bound to the session. Just return -EJUKEBOX and  in __do_request()
2830 		if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {  in __do_request()
2831 			err = -EJUKEBOX;  in __do_request()
2834 		if (session->s_state == CEPH_MDS_SESSION_NEW ||  in __do_request()
2835 		    session->s_state == CEPH_MDS_SESSION_CLOSING) {  in __do_request()
2841 				req->r_resend_mds = mds;  in __do_request()
2843 		list_add(&req->r_wait, &session->s_waiting);  in __do_request()
2848 	req->r_resend_mds = -1;   /* forget any previous mds hint */  in __do_request()
2850 	if (req->r_request_started == 0)   /* note request start time */  in __do_request()
2851 		req->r_request_started = jiffies;  in __do_request()
2860 		req->r_err = err;  in __do_request()
2868  * called under mdsc->mutex
2881 		list_del_init(&req->r_wait);  in __wake_requests()
2882 		dout(" wake request %p tid %llu\n", req, req->r_tid);  in __wake_requests()
2894 	struct rb_node *p = rb_first(&mdsc->request_tree);  in kick_requests()
2900 		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))  in kick_requests()
2902 		if (req->r_attempts > 0)  in kick_requests()
2904 		if (req->r_session &&  in kick_requests()
2905 		    req->r_session->s_mds == mds) {  in kick_requests()
2906 			dout(" kicking tid %llu\n", req->r_tid);  in kick_requests()
2907 			list_del_init(&req->r_wait);  in kick_requests()
2919 	if (req->r_inode)  in ceph_mdsc_submit_request()
2920 		ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);  in ceph_mdsc_submit_request()
2921 	if (req->r_parent) {  in ceph_mdsc_submit_request()
2922 		struct ceph_inode_info *ci = ceph_inode(req->r_parent);  in ceph_mdsc_submit_request()
2923 		int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?  in ceph_mdsc_submit_request()
2925 		spin_lock(&ci->i_ceph_lock);  in ceph_mdsc_submit_request()
2928 		spin_unlock(&ci->i_ceph_lock);  in ceph_mdsc_submit_request()
2929 		ihold(req->r_parent);  in ceph_mdsc_submit_request()
2931 	if (req->r_old_dentry_dir)  in ceph_mdsc_submit_request()
2932 		ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),  in ceph_mdsc_submit_request()
2935 	if (req->r_inode) {  in ceph_mdsc_submit_request()
2936 		err = ceph_wait_on_async_create(req->r_inode);  in ceph_mdsc_submit_request()
2944 	if (!err && req->r_old_inode) {  in ceph_mdsc_submit_request()
2945 		err = ceph_wait_on_async_create(req->r_old_inode);  in ceph_mdsc_submit_request()
2954 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_submit_request()
2957 	err = req->r_err;  in ceph_mdsc_submit_request()
2958 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_submit_request()
2969 	if (!req->r_timeout && req->r_wait_for_completion) {  in ceph_mdsc_wait_request()
2970 		err = req->r_wait_for_completion(mdsc, req);  in ceph_mdsc_wait_request()
2973 					&req->r_completion,  in ceph_mdsc_wait_request()
2974 					ceph_timeout_jiffies(req->r_timeout));  in ceph_mdsc_wait_request()
2978 			err = -ETIMEDOUT;  /* timed out */  in ceph_mdsc_wait_request()
2983 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_wait_request()
2986 	if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {  in ceph_mdsc_wait_request()
2987 		err = le32_to_cpu(req->r_reply_info.head->result);  in ceph_mdsc_wait_request()
2989 		dout("aborted request %lld with %d\n", req->r_tid, err);  in ceph_mdsc_wait_request()
2996 		mutex_lock(&req->r_fill_mutex);  in ceph_mdsc_wait_request()
2997 		req->r_err = err;  in ceph_mdsc_wait_request()
2998 		set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);  in ceph_mdsc_wait_request()
2999 		mutex_unlock(&req->r_fill_mutex);  in ceph_mdsc_wait_request()
3001 		if (req->r_parent &&  in ceph_mdsc_wait_request()
3002 		    (req->r_op & CEPH_MDS_OP_WRITE))  in ceph_mdsc_wait_request()
3005 		err = req->r_err;  in ceph_mdsc_wait_request()
3008 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_wait_request()
3038 	struct inode *dir = req->r_parent;  in ceph_invalidate_dir_request()
3039 	struct inode *old_dir = req->r_old_dentry_dir;  in ceph_invalidate_dir_request()
3046 	if (req->r_dentry)  in ceph_invalidate_dir_request()
3047 		ceph_invalidate_dentry_lease(req->r_dentry);  in ceph_invalidate_dir_request()
3048 	if (req->r_old_dentry)  in ceph_invalidate_dir_request()
3049 		ceph_invalidate_dentry_lease(req->r_old_dentry);  in ceph_invalidate_dir_request()
3061 	struct ceph_mds_client *mdsc = session->s_mdsc;  in handle_reply()
3063 	struct ceph_mds_reply_head *head = msg->front.iov_base;  in handle_reply()
3068 	int mds = session->s_mds;  in handle_reply()
3070 	if (msg->front.iov_len < sizeof(*head)) {  in handle_reply()
3077 	tid = le64_to_cpu(msg->hdr.tid);  in handle_reply()
3078 	mutex_lock(&mdsc->mutex);  in handle_reply()
3082 		mutex_unlock(&mdsc->mutex);  in handle_reply()
3088 	if (req->r_session != session) {  in handle_reply()
3090 		       " not mds%d\n", tid, session->s_mds,  in handle_reply()
3091 		       req->r_session ? req->r_session->s_mds : -1);  in handle_reply()
3092 		mutex_unlock(&mdsc->mutex);  in handle_reply()
3097 	if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||  in handle_reply()
3098 	    (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {  in handle_reply()
3100 			   head->safe ? "safe" : "unsafe", tid, mds);  in handle_reply()
3101 		mutex_unlock(&mdsc->mutex);  in handle_reply()
3104 	if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {  in handle_reply()
3107 		mutex_unlock(&mdsc->mutex);  in handle_reply()
3111 	result = le32_to_cpu(head->result);  in handle_reply()
3120 	if (result == -ESTALE) {  in handle_reply()
3121 		dout("got ESTALE on request %llu\n", req->r_tid);  in handle_reply()
3122 		req->r_resend_mds = -1;  in handle_reply()
3123 		if (req->r_direct_mode != USE_AUTH_MDS) {  in handle_reply()
3125 			req->r_direct_mode = USE_AUTH_MDS;  in handle_reply()
3127 			mutex_unlock(&mdsc->mutex);  in handle_reply()
3131 			if (mds >= 0 && mds != req->r_session->s_mds) {  in handle_reply()
3134 				mutex_unlock(&mdsc->mutex);  in handle_reply()
3138 		dout("have to return ESTALE on request %llu\n", req->r_tid);  in handle_reply()
3142 	if (head->safe) {  in handle_reply()
3143 		set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);  in handle_reply()
3147 		if (mdsc->stopping && !__get_oldest_req(mdsc))  in handle_reply()
3148 			complete_all(&mdsc->safe_umount_waiters);  in handle_reply()
3150 		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {  in handle_reply()
3160 			mutex_unlock(&mdsc->mutex);  in handle_reply()
3164 		set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);  in handle_reply()
3165 		list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);  in handle_reply()
3169 	rinfo = &req->r_reply_info;  in handle_reply()
3170 	if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))  in handle_reply()
3171 		err = parse_reply_info(session, msg, rinfo, (u64)-1);  in handle_reply()
3173 		err = parse_reply_info(session, msg, rinfo, session->s_con.peer_features);  in handle_reply()
3174 	mutex_unlock(&mdsc->mutex);  in handle_reply()
3176 	mutex_lock(&session->s_mutex);  in handle_reply()
3185 	if (rinfo->snapblob_len) {  in handle_reply()
3186 		down_write(&mdsc->snap_rwsem);  in handle_reply()
3187 		ceph_update_snap_trace(mdsc, rinfo->snapblob,  in handle_reply()
3188 				rinfo->snapblob + rinfo->snapblob_len,  in handle_reply()
3189 				le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,  in handle_reply()
3191 		downgrade_write(&mdsc->snap_rwsem);  in handle_reply()
3193 		down_read(&mdsc->snap_rwsem);  in handle_reply()
3197 	mutex_lock(&req->r_fill_mutex);  in handle_reply()
3198 	current->journal_info = req;  in handle_reply()
3199 	err = ceph_fill_trace(mdsc->fsc->sb, req);  in handle_reply()
3201 		if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||  in handle_reply()
3202 				    req->r_op == CEPH_MDS_OP_LSSNAP))  in handle_reply()
3203 			ceph_readdir_prepopulate(req, req->r_session);  in handle_reply()
3205 	current->journal_info = NULL;  in handle_reply()
3206 	mutex_unlock(&req->r_fill_mutex);  in handle_reply()
3208 	up_read(&mdsc->snap_rwsem);  in handle_reply()
3213 		if (req->r_target_inode &&  in handle_reply()
3214 		    test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {  in handle_reply()
3216 				ceph_inode(req->r_target_inode);  in handle_reply()
3217 			spin_lock(&ci->i_unsafe_lock);  in handle_reply()
3218 			list_add_tail(&req->r_unsafe_target_item,  in handle_reply()
3219 				      &ci->i_unsafe_iops);  in handle_reply()
3220 			spin_unlock(&ci->i_unsafe_lock);  in handle_reply()
3223 		ceph_unreserve_caps(mdsc, &req->r_caps_reservation);  in handle_reply()
3226 	mutex_lock(&mdsc->mutex);  in handle_reply()
3227 	if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {  in handle_reply()
3229 			req->r_err = err;  in handle_reply()
3231 			req->r_reply =  ceph_msg_get(msg);  in handle_reply()
3232 			set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);  in handle_reply()
3237 	mutex_unlock(&mdsc->mutex);  in handle_reply()
3239 	mutex_unlock(&session->s_mutex);  in handle_reply()
3244 	ceph_update_metadata_latency(&mdsc->metric, req->r_start_latency,  in handle_reply()
3245 				     req->r_end_latency, err);  in handle_reply()
3261 	u64 tid = le64_to_cpu(msg->hdr.tid);  in handle_forward()
3264 	int err = -EINVAL;  in handle_forward()
3265 	void *p = msg->front.iov_base;  in handle_forward()
3266 	void *end = p + msg->front.iov_len;  in handle_forward()  local
3268 	ceph_decode_need(&p, end, 2*sizeof(u32), bad);  in handle_forward()
3272 	mutex_lock(&mdsc->mutex);  in handle_forward()
3275 		dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);  in handle_forward()
3279 	if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {  in handle_forward()
3282 	} else if (fwd_seq <= req->r_num_fwd) {  in handle_forward()
3283 		dout("forward tid %llu to mds%d - old seq %d <= %d\n",  in handle_forward()
3284 		     tid, next_mds, req->r_num_fwd, fwd_seq);  in handle_forward()
3288 		BUG_ON(req->r_err);  in handle_forward()
3289 		BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));  in handle_forward()
3290 		req->r_attempts = 0;  in handle_forward()
3291 		req->r_num_fwd = fwd_seq;  in handle_forward()
3292 		req->r_resend_mds = next_mds;  in handle_forward()
3298 	mutex_unlock(&mdsc->mutex);  in handle_forward()
3305 static int __decode_session_metadata(void **p, void *end,  in __decode_session_metadata()  argument
3311 	ceph_decode_32_safe(p, end, n, bad);  in __decode_session_metadata()
3312 	while (n-- > 0) {  in __decode_session_metadata()
3314 		ceph_decode_32_safe(p, end, len, bad);  in __decode_session_metadata()
3315 		ceph_decode_need(p, end, len, bad);  in __decode_session_metadata()
3318 		ceph_decode_32_safe(p, end, len, bad);  in __decode_session_metadata()
3319 		ceph_decode_need(p, end, len, bad);  in __decode_session_metadata()
3330 	return -1;  in __decode_session_metadata()
3339 	struct ceph_mds_client *mdsc = session->s_mdsc;  in handle_session()
3340 	int mds = session->s_mds;  in handle_session()
3341 	int msg_version = le16_to_cpu(msg->hdr.version);  in handle_session()
3342 	void *p = msg->front.iov_base;  in handle_session()
3343 	void *end = p + msg->front.iov_len;  in handle_session()  local
3351 	ceph_decode_need(&p, end, sizeof(*h), bad);  in handle_session()
3355 	op = le32_to_cpu(h->op);  in handle_session()
3356 	seq = le64_to_cpu(h->seq);  in handle_session()
3361 		if (__decode_session_metadata(&p, end, &blocklisted) < 0)  in handle_session()
3364 		ceph_decode_32_safe(&p, end, len, bad);  in handle_session()
3366 			ceph_decode_64_safe(&p, end, features, bad);  in handle_session()
3367 			p += len - sizeof(features);  in handle_session()
3371 	mutex_lock(&mdsc->mutex);  in handle_session()
3377 	session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;  in handle_session()
3378 	mutex_unlock(&mdsc->mutex);  in handle_session()
3380 	mutex_lock(&session->s_mutex);  in handle_session()
3384 	     ceph_session_state_name(session->s_state), seq);  in handle_session()
3386 	if (session->s_state == CEPH_MDS_SESSION_HUNG) {  in handle_session()
3387 		session->s_state = CEPH_MDS_SESSION_OPEN;  in handle_session()
3388 		pr_info("mds%d came back\n", session->s_mds);  in handle_session()
3393 		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)  in handle_session()
3394 			pr_info("mds%d reconnect success\n", session->s_mds);  in handle_session()
3395 		session->s_state = CEPH_MDS_SESSION_OPEN;  in handle_session()
3396 		session->s_features = features;  in handle_session()
3398 		if (test_bit(CEPHFS_FEATURE_METRIC_COLLECT, &session->s_features))  in handle_session()
3399 			metric_schedule_delayed(&mdsc->metric);  in handle_session()
3401 		if (mdsc->stopping)  in handle_session()
3406 		if (session->s_renew_seq == seq)  in handle_session()
3411 		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)  in handle_session()
3412 			pr_info("mds%d reconnect denied\n", session->s_mds);  in handle_session()
3413 		session->s_state = CEPH_MDS_SESSION_CLOSED;  in handle_session()
3417 		wake_up_all(&mdsc->session_close_wq);  in handle_session()
3422 			session->s_mds);  in handle_session()
3423 		spin_lock(&session->s_gen_ttl_lock);  in handle_session()
3424 		session->s_cap_gen++;  in handle_session()
3425 		session->s_cap_ttl = jiffies - 1;  in handle_session()
3426 		spin_unlock(&session->s_gen_ttl_lock);  in handle_session()
3431 		ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));  in handle_session()
3440 		spin_lock(&session->s_cap_lock);  in handle_session()
3441 		session->s_readonly = true;  in handle_session()
3442 		spin_unlock(&session->s_cap_lock);  in handle_session()
3447 		WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);  in handle_session()
3448 		pr_info("mds%d rejected session\n", session->s_mds);  in handle_session()
3449 		session->s_state = CEPH_MDS_SESSION_REJECTED;  in handle_session()
3453 			mdsc->fsc->blocklisted = true;  in handle_session()
3462 	mutex_unlock(&session->s_mutex);  in handle_session()
3464 		mutex_lock(&mdsc->mutex);  in handle_session()
3465 		__wake_requests(mdsc, &session->s_waiting);  in handle_session()
3468 		mutex_unlock(&mdsc->mutex);  in handle_session()
3476 	       (int)msg->front.iov_len);  in handle_session()
3485 	dcaps = xchg(&req->r_dir_caps, 0);  in ceph_mdsc_release_dir_caps()
3488 		ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);  in ceph_mdsc_release_dir_caps()
3496 	dcaps = xchg(&req->r_dir_caps, 0);  in ceph_mdsc_release_dir_caps_no_check()
3499 		ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),  in ceph_mdsc_release_dir_caps_no_check()
3505  * called under session->mutex.
3513 	dout("replay_unsafe_requests mds%d\n", session->s_mds);  in replay_unsafe_requests()
3515 	mutex_lock(&mdsc->mutex);  in replay_unsafe_requests()
3516 	list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)  in replay_unsafe_requests()
3520 	 * also re-send old requests when MDS enters reconnect stage. So that MDS  in replay_unsafe_requests()
3523 	p = rb_first(&mdsc->request_tree);  in replay_unsafe_requests()
3527 		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))  in replay_unsafe_requests()
3529 		if (req->r_attempts == 0)  in replay_unsafe_requests()
3531 		if (!req->r_session)  in replay_unsafe_requests()
3533 		if (req->r_session->s_mds != session->s_mds)  in replay_unsafe_requests()
3540 	mutex_unlock(&mdsc->mutex);  in replay_unsafe_requests()
3549 	int err = -ENOMEM;  in send_reconnect_partial()
3551 	if (!recon_state->allow_multi)  in send_reconnect_partial()
3552 		return -ENOSPC;  in send_reconnect_partial()
3555 	BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);  in send_reconnect_partial()
3557 	/* pre-allocate new pagelist */  in send_reconnect_partial()
3560 		return -ENOMEM;  in send_reconnect_partial()
3571 	if (recon_state->nr_caps) {  in send_reconnect_partial()
3573 		err = ceph_pagelist_encode_32(recon_state->pagelist, 0);  in send_reconnect_partial()
3583 	err = ceph_pagelist_encode_8(recon_state->pagelist, 1);  in send_reconnect_partial()
3587 	page = list_first_entry(&recon_state->pagelist->head, struct page, lru);  in send_reconnect_partial()
3589 	if (recon_state->nr_caps) {  in send_reconnect_partial()
3591 		*addr = cpu_to_le32(recon_state->nr_caps);  in send_reconnect_partial()
3594 		*(addr + 1) = cpu_to_le32(recon_state->nr_realms);  in send_reconnect_partial()
3598 	reply->hdr.version = cpu_to_le16(5);  in send_reconnect_partial()
3599 	reply->hdr.compat_version = cpu_to_le16(4);  in send_reconnect_partial()
3601 	reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);  in send_reconnect_partial()
3602 	ceph_msg_data_add_pagelist(reply, recon_state->pagelist);  in send_reconnect_partial()
3604 	ceph_con_send(&recon_state->session->s_con, reply);  in send_reconnect_partial()
3605 	ceph_pagelist_release(recon_state->pagelist);  in send_reconnect_partial()
3607 	recon_state->pagelist = _pagelist;  in send_reconnect_partial()
3608 	recon_state->nr_caps = 0;  in send_reconnect_partial()
3609 	recon_state->nr_realms = 0;  in send_reconnect_partial()
3610 	recon_state->msg_version = 5;  in send_reconnect_partial()
3623 	if (hlist_empty(&inode->i_dentry))  in d_find_primary()
3626 	spin_lock(&inode->i_lock);  in d_find_primary()
3627 	if (hlist_empty(&inode->i_dentry))  in d_find_primary()
3630 	if (S_ISDIR(inode->i_mode)) {  in d_find_primary()
3631 		alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);  in d_find_primary()
3637 	hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {  in d_find_primary()
3638 		spin_lock(&alias->d_lock);  in d_find_primary()
3640 		    (ceph_dentry(alias)->flags & CEPH_DENTRY_PRIMARY_LINK)) {  in d_find_primary()
3643 		spin_unlock(&alias->d_lock);  in d_find_primary()
3648 	spin_unlock(&inode->i_lock);  in d_find_primary()
3662 	struct ceph_inode_info *ci = cap->ci;  in reconnect_caps_cb()
3664 	struct ceph_pagelist *pagelist = recon_state->pagelist;  in reconnect_caps_cb()
3672 	     inode, ceph_vinop(inode), cap, cap->cap_id,  in reconnect_caps_cb()
3673 	     ceph_cap_string(cap->issued));  in reconnect_caps_cb()
3679 					    recon_state->msg_version >= 2);  in reconnect_caps_cb()
3691 	spin_lock(&ci->i_ceph_lock);  in reconnect_caps_cb()
3692 	cap->seq = 0;        /* reset cap seq */  in reconnect_caps_cb()
3693 	cap->issue_seq = 0;  /* and issue_seq */  in reconnect_caps_cb()
3694 	cap->mseq = 0;       /* and migrate_seq */  in reconnect_caps_cb()
3695 	cap->cap_gen = cap->session->s_cap_gen;  in reconnect_caps_cb()
3698 	if (S_ISDIR(inode->i_mode)) {  in reconnect_caps_cb()
3699 		if (cap->issued & CEPH_CAP_DIR_CREATE) {  in reconnect_caps_cb()
3700 			ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));  in reconnect_caps_cb()
3701 			memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));  in reconnect_caps_cb()
3703 		cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;  in reconnect_caps_cb()
3706 	if (recon_state->msg_version >= 2) {  in reconnect_caps_cb()
3707 		rec.v2.cap_id = cpu_to_le64(cap->cap_id);  in reconnect_caps_cb()
3709 		rec.v2.issued = cpu_to_le32(cap->issued);  in reconnect_caps_cb()
3710 		rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);  in reconnect_caps_cb()
3713 			((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);  in reconnect_caps_cb()
3715 		rec.v1.cap_id = cpu_to_le64(cap->cap_id);  in reconnect_caps_cb()
3717 		rec.v1.issued = cpu_to_le32(cap->issued);  in reconnect_caps_cb()
3718 		rec.v1.size = cpu_to_le64(inode->i_size);  in reconnect_caps_cb()
3719 		ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);  in reconnect_caps_cb()
3720 		ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);  in reconnect_caps_cb()
3721 		rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);  in reconnect_caps_cb()
3725 	if (list_empty(&ci->i_cap_snaps)) {  in reconnect_caps_cb()
3726 		snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;  in reconnect_caps_cb()
3729 			list_first_entry(&ci->i_cap_snaps,  in reconnect_caps_cb()
3731 		snap_follows = capsnap->follows;  in reconnect_caps_cb()
3733 	spin_unlock(&ci->i_ceph_lock);  in reconnect_caps_cb()
3735 	if (recon_state->msg_version >= 2) {  in reconnect_caps_cb()
3753 				err = -ENOMEM;  in reconnect_caps_cb()
3762 				if (err == -ENOSPC)  in reconnect_caps_cb()
3771 		if (recon_state->msg_version >= 3) {  in reconnect_caps_cb()
3791 		if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {  in reconnect_caps_cb()
3795 			pagelist = recon_state->pagelist;  in reconnect_caps_cb()
3803 		if (recon_state->msg_version >= 3) {  in reconnect_caps_cb()
3831 		recon_state->nr_caps++;  in reconnect_caps_cb()
3839 	struct ceph_pagelist *pagelist = recon_state->pagelist;  in encode_snap_realms()
3842 	if (recon_state->msg_version >= 4) {  in encode_snap_realms()
3843 		err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);  in encode_snap_realms()
3853 	for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {  in encode_snap_realms()
3858 		if (recon_state->msg_version >= 4) {  in encode_snap_realms()
3862 			if (pagelist->length + need > RECONNECT_MAX_SIZE) {  in encode_snap_realms()
3866 				pagelist = recon_state->pagelist;  in encode_snap_realms()
3879 		     realm->ino, realm->seq, realm->parent_ino);  in encode_snap_realms()
3880 		sr_rec.ino = cpu_to_le64(realm->ino);  in encode_snap_realms()
3881 		sr_rec.seq = cpu_to_le64(realm->seq);  in encode_snap_realms()
3882 		sr_rec.parent = cpu_to_le64(realm->parent_ino);  in encode_snap_realms()
3888 		recon_state->nr_realms++;  in encode_snap_realms()
3909 	int mds = session->s_mds;  in send_mds_reconnect()
3910 	int err = -ENOMEM;  in send_mds_reconnect()
3926 	xa_destroy(&session->s_delegated_inos);  in send_mds_reconnect()
3928 	mutex_lock(&session->s_mutex);  in send_mds_reconnect()
3929 	session->s_state = CEPH_MDS_SESSION_RECONNECTING;  in send_mds_reconnect()
3930 	session->s_seq = 0;  in send_mds_reconnect()
3933 	     ceph_session_state_name(session->s_state));  in send_mds_reconnect()
3935 	spin_lock(&session->s_gen_ttl_lock);  in send_mds_reconnect()
3936 	session->s_cap_gen++;  in send_mds_reconnect()
3937 	spin_unlock(&session->s_gen_ttl_lock);  in send_mds_reconnect()
3939 	spin_lock(&session->s_cap_lock);  in send_mds_reconnect()
3941 	session->s_readonly = 0;  in send_mds_reconnect()
3947 	session->s_cap_reconnect = 1;  in send_mds_reconnect()
3950 	spin_unlock(&session->s_cap_lock);  in send_mds_reconnect()
3954 	if (mdsc->fsc->sb->s_root)  in send_mds_reconnect()
3955 		shrink_dcache_parent(mdsc->fsc->sb->s_root);  in send_mds_reconnect()
3957 	ceph_con_close(&session->s_con);  in send_mds_reconnect()
3958 	ceph_con_open(&session->s_con,  in send_mds_reconnect()
3960 		      ceph_mdsmap_get_addr(mdsc->mdsmap, mds));  in send_mds_reconnect()
3967 	down_read(&mdsc->snap_rwsem);  in send_mds_reconnect()
3974 	if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {  in send_mds_reconnect()
3977 	} else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {  in send_mds_reconnect()
3985 	spin_lock(&session->s_cap_lock);  in send_mds_reconnect()
3986 	session->s_cap_reconnect = 0;  in send_mds_reconnect()
3987 	spin_unlock(&session->s_cap_lock);  in send_mds_reconnect()
3993 	if (mdsc->num_snap_realms) {  in send_mds_reconnect()
3995 			recon_state.pagelist->length +  in send_mds_reconnect()
3996 			mdsc->num_snap_realms *  in send_mds_reconnect()
4002 			total_len += mdsc->num_snap_realms *  in send_mds_reconnect()
4007 				err = -ENOSPC;  in send_mds_reconnect()
4031 			list_first_entry(&recon_state.pagelist->head,  in send_mds_reconnect()
4035 			WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);  in send_mds_reconnect()
4043 	reply->hdr.version = cpu_to_le16(recon_state.msg_version);  in send_mds_reconnect()
4045 		reply->hdr.compat_version = cpu_to_le16(4);  in send_mds_reconnect()
4047 	reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);  in send_mds_reconnect()
4050 	ceph_con_send(&session->s_con, reply);  in send_mds_reconnect()
4052 	mutex_unlock(&session->s_mutex);  in send_mds_reconnect()
4054 	mutex_lock(&mdsc->mutex);  in send_mds_reconnect()
4055 	__wake_requests(mdsc, &session->s_waiting);  in send_mds_reconnect()
4056 	mutex_unlock(&mdsc->mutex);  in send_mds_reconnect()
4058 	up_read(&mdsc->snap_rwsem);  in send_mds_reconnect()
4064 	up_read(&mdsc->snap_rwsem);  in send_mds_reconnect()
4065 	mutex_unlock(&session->s_mutex);  in send_mds_reconnect()
4078  * called under mdsc->mutex.
4089 	     newmap->m_epoch, oldmap->m_epoch);  in check_new_map()
4091 	for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {  in check_new_map()
4092 		if (!mdsc->sessions[i])  in check_new_map()
4094 		s = mdsc->sessions[i];  in check_new_map()
4098 		dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",  in check_new_map()
4103 		     ceph_session_state_name(s->s_state));  in check_new_map()
4105 		if (i >= newmap->possible_max_rank) {  in check_new_map()
4109 			__wake_requests(mdsc, &s->s_waiting);  in check_new_map()
4110 			mutex_unlock(&mdsc->mutex);  in check_new_map()
4112 			mutex_lock(&s->s_mutex);  in check_new_map()
4115 			mutex_unlock(&s->s_mutex);  in check_new_map()
4119 			mutex_lock(&mdsc->mutex);  in check_new_map()
4128 			mutex_unlock(&mdsc->mutex);  in check_new_map()
4129 			mutex_lock(&s->s_mutex);  in check_new_map()
4130 			mutex_lock(&mdsc->mutex);  in check_new_map()
4131 			ceph_con_close(&s->s_con);  in check_new_map()
4132 			mutex_unlock(&s->s_mutex);  in check_new_map()
4133 			s->s_state = CEPH_MDS_SESSION_RESTARTING;  in check_new_map()
4141 		if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&  in check_new_map()
4143 			mutex_unlock(&mdsc->mutex);  in check_new_map()
4145 			mutex_lock(&mdsc->mutex);  in check_new_map()
4155 				pr_info("mds%d recovery completed\n", s->s_mds);  in check_new_map()
4157 			mutex_unlock(&mdsc->mutex);  in check_new_map()
4158 			mutex_lock(&s->s_mutex);  in check_new_map()
4159 			mutex_lock(&mdsc->mutex);  in check_new_map()
4161 			mutex_unlock(&s->s_mutex);  in check_new_map()
4166 	for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {  in check_new_map()
4167 		s = mdsc->sessions[i];  in check_new_map()
4172 		if (s->s_state == CEPH_MDS_SESSION_OPEN ||  in check_new_map()
4173 		    s->s_state == CEPH_MDS_SESSION_HUNG ||  in check_new_map()
4174 		    s->s_state == CEPH_MDS_SESSION_CLOSING) {  in check_new_map()
4189  * caller must hold session s_mutex, dentry->d_lock
4195 	ceph_put_mds_session(di->lease_session);  in __ceph_mdsc_drop_dentry_lease()
4196 	di->lease_session = NULL;  in __ceph_mdsc_drop_dentry_lease()
4203 	struct super_block *sb = mdsc->fsc->sb;  in handle_lease()
4207 	int mds = session->s_mds;  in handle_lease()
4208 	struct ceph_mds_lease *h = msg->front.iov_base;  in handle_lease()
4217 	if (msg->front.iov_len < sizeof(*h) + sizeof(u32))  in handle_lease()
4219 	vino.ino = le64_to_cpu(h->ino);  in handle_lease()
4221 	seq = le32_to_cpu(h->seq);  in handle_lease()
4223 	if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)  in handle_lease()
4230 	     ceph_lease_op_name(h->action), vino.ino, inode,  in handle_lease()
4233 	mutex_lock(&session->s_mutex);  in handle_lease()
4254 	spin_lock(&dentry->d_lock);  in handle_lease()
4256 	switch (h->action) {  in handle_lease()
4258 		if (di->lease_session == session) {  in handle_lease()
4259 			if (ceph_seq_cmp(di->lease_seq, seq) > 0)  in handle_lease()
4260 				h->seq = cpu_to_le32(di->lease_seq);  in handle_lease()
4267 		if (di->lease_session == session &&  in handle_lease()
4268 		    di->lease_gen == session->s_cap_gen &&  in handle_lease()
4269 		    di->lease_renew_from &&  in handle_lease()
4270 		    di->lease_renew_after == 0) {  in handle_lease()
4272 				msecs_to_jiffies(le32_to_cpu(h->duration_ms));  in handle_lease()
4274 			di->lease_seq = seq;  in handle_lease()
4275 			di->time = di->lease_renew_from + duration;  in handle_lease()
4276 			di->lease_renew_after = di->lease_renew_from +  in handle_lease()
4278 			di->lease_renew_from = 0;  in handle_lease()
4282 	spin_unlock(&dentry->d_lock);  in handle_lease()
4290 	h->action = CEPH_MDS_LEASE_REVOKE_ACK;  in handle_lease()
4292 	ceph_con_send(&session->s_con, msg);  in handle_lease()
4295 	mutex_unlock(&session->s_mutex);  in handle_lease()
4315 	     dentry, ceph_lease_op_name(action), session->s_mds);  in ceph_mdsc_lease_send_msg()
4320 	lease = msg->front.iov_base;  in ceph_mdsc_lease_send_msg()
4321 	lease->action = action;  in ceph_mdsc_lease_send_msg()
4322 	lease->seq = cpu_to_le32(seq);  in ceph_mdsc_lease_send_msg()
4324 	spin_lock(&dentry->d_lock);  in ceph_mdsc_lease_send_msg()
4325 	dir = d_inode(dentry->d_parent);  in ceph_mdsc_lease_send_msg()
4326 	lease->ino = cpu_to_le64(ceph_ino(dir));  in ceph_mdsc_lease_send_msg()
4327 	lease->first = lease->last = cpu_to_le64(ceph_snap(dir));  in ceph_mdsc_lease_send_msg()
4329 	put_unaligned_le32(dentry->d_name.len, lease + 1);  in ceph_mdsc_lease_send_msg()
4331 	       dentry->d_name.name, dentry->d_name.len);  in ceph_mdsc_lease_send_msg()
4332 	spin_unlock(&dentry->d_lock);  in ceph_mdsc_lease_send_msg()
4338 	msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);  in ceph_mdsc_lease_send_msg()
4340 	ceph_con_send(&session->s_con, msg);  in ceph_mdsc_lease_send_msg()
4350 	mutex_lock(&mdsc->mutex);  in lock_unlock_sessions()
4351 	for (i = 0; i < mdsc->max_sessions; i++) {  in lock_unlock_sessions()
4355 		mutex_unlock(&mdsc->mutex);  in lock_unlock_sessions()
4356 		mutex_lock(&s->s_mutex);  in lock_unlock_sessions()
4357 		mutex_unlock(&s->s_mutex);  in lock_unlock_sessions()
4359 		mutex_lock(&mdsc->mutex);  in lock_unlock_sessions()
4361 	mutex_unlock(&mdsc->mutex);  in lock_unlock_sessions()
4366 	struct ceph_fs_client *fsc = mdsc->fsc;  in maybe_recover_session()
4371 	if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)  in maybe_recover_session()
4374 	if (!READ_ONCE(fsc->blocklisted))  in maybe_recover_session()
4377 	if (fsc->last_auto_reconnect &&  in maybe_recover_session()
4378 	    time_before(jiffies, fsc->last_auto_reconnect + HZ * 60 * 30))  in maybe_recover_session()
4382 	fsc->last_auto_reconnect = jiffies;  in maybe_recover_session()
4383 	ceph_force_reconnect(fsc->sb);  in maybe_recover_session()
4388 	switch (s->s_state) {  in check_session_state()
4390 		if (s->s_ttl && time_after(jiffies, s->s_ttl)) {  in check_session_state()
4391 			s->s_state = CEPH_MDS_SESSION_HUNG;  in check_session_state()
4392 			pr_info("mds%d hung\n", s->s_mds);  in check_session_state()
4415 	lockdep_assert_held(&s->s_mutex);  in inc_session_sequence()
4417 	s->s_seq++;  in inc_session_sequence()
4419 	if (s->s_state == CEPH_MDS_SESSION_CLOSING) {  in inc_session_sequence()
4422 		dout("resending session close request for mds%d\n", s->s_mds);  in inc_session_sequence()
4426 			       s->s_mds, ret);  in inc_session_sequence()
4431  * delayed work -- periodically trim expired leases, renew caps with mds
4437 	schedule_delayed_work(&mdsc->delayed_work, hz);  in schedule_delayed()
4450 	if (mdsc->stopping)  in delayed_work()
4453 	mutex_lock(&mdsc->mutex);  in delayed_work()
4454 	renew_interval = mdsc->mdsmap->m_session_timeout >> 2;  in delayed_work()
4456 				   mdsc->last_renew_caps);  in delayed_work()
4458 		mdsc->last_renew_caps = jiffies;  in delayed_work()
4460 	for (i = 0; i < mdsc->max_sessions; i++) {  in delayed_work()
4469 		mutex_unlock(&mdsc->mutex);  in delayed_work()
4471 		mutex_lock(&s->s_mutex);  in delayed_work()
4475 			ceph_con_keepalive(&s->s_con);  in delayed_work()
4476 		if (s->s_state == CEPH_MDS_SESSION_OPEN ||  in delayed_work()
4477 		    s->s_state == CEPH_MDS_SESSION_HUNG)  in delayed_work()
4479 		mutex_unlock(&s->s_mutex);  in delayed_work()
4482 		mutex_lock(&mdsc->mutex);  in delayed_work()
4484 	mutex_unlock(&mdsc->mutex);  in delayed_work()
4505 		return -ENOMEM;  in ceph_mdsc_init()
4506 	mdsc->fsc = fsc;  in ceph_mdsc_init()
4507 	mutex_init(&mdsc->mutex);  in ceph_mdsc_init()
4508 	mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);  in ceph_mdsc_init()
4509 	if (!mdsc->mdsmap) {  in ceph_mdsc_init()
4510 		err = -ENOMEM;  in ceph_mdsc_init()
4514 	init_completion(&mdsc->safe_umount_waiters);  in ceph_mdsc_init()
4515 	init_waitqueue_head(&mdsc->session_close_wq);  in ceph_mdsc_init()
4516 	INIT_LIST_HEAD(&mdsc->waiting_for_map);  in ceph_mdsc_init()
4517 	mdsc->sessions = NULL;  in ceph_mdsc_init()
4518 	atomic_set(&mdsc->num_sessions, 0);  in ceph_mdsc_init()
4519 	mdsc->max_sessions = 0;  in ceph_mdsc_init()
4520 	mdsc->stopping = 0;  in ceph_mdsc_init()
4521 	atomic64_set(&mdsc->quotarealms_count, 0);  in ceph_mdsc_init()
4522 	mdsc->quotarealms_inodes = RB_ROOT;  in ceph_mdsc_init()
4523 	mutex_init(&mdsc->quotarealms_inodes_mutex);  in ceph_mdsc_init()
4524 	mdsc->last_snap_seq = 0;  in ceph_mdsc_init()
4525 	init_rwsem(&mdsc->snap_rwsem);  in ceph_mdsc_init()
4526 	mdsc->snap_realms = RB_ROOT;  in ceph_mdsc_init()
4527 	INIT_LIST_HEAD(&mdsc->snap_empty);  in ceph_mdsc_init()
4528 	mdsc->num_snap_realms = 0;  in ceph_mdsc_init()
4529 	spin_lock_init(&mdsc->snap_empty_lock);  in ceph_mdsc_init()
4530 	mdsc->last_tid = 0;  in ceph_mdsc_init()
4531 	mdsc->oldest_tid = 0;  in ceph_mdsc_init()
4532 	mdsc->request_tree = RB_ROOT;  in ceph_mdsc_init()
4533 	INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);  in ceph_mdsc_init()
4534 	mdsc->last_renew_caps = jiffies;  in ceph_mdsc_init()
4535 	INIT_LIST_HEAD(&mdsc->cap_delay_list);  in ceph_mdsc_init()
4536 	INIT_LIST_HEAD(&mdsc->cap_wait_list);  in ceph_mdsc_init()
4537 	spin_lock_init(&mdsc->cap_delay_lock);  in ceph_mdsc_init()
4538 	INIT_LIST_HEAD(&mdsc->snap_flush_list);  in ceph_mdsc_init()
4539 	spin_lock_init(&mdsc->snap_flush_lock);  in ceph_mdsc_init()
4540 	mdsc->last_cap_flush_tid = 1;  in ceph_mdsc_init()
4541 	INIT_LIST_HEAD(&mdsc->cap_flush_list);  in ceph_mdsc_init()
4542 	INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);  in ceph_mdsc_init()
4543 	mdsc->num_cap_flushing = 0;  in ceph_mdsc_init()
4544 	spin_lock_init(&mdsc->cap_dirty_lock);  in ceph_mdsc_init()
4545 	init_waitqueue_head(&mdsc->cap_flushing_wq);  in ceph_mdsc_init()
4546 	INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);  in ceph_mdsc_init()
4547 	atomic_set(&mdsc->cap_reclaim_pending, 0);  in ceph_mdsc_init()
4548 	err = ceph_metric_init(&mdsc->metric);  in ceph_mdsc_init()
4552 	spin_lock_init(&mdsc->dentry_list_lock);  in ceph_mdsc_init()
4553 	INIT_LIST_HEAD(&mdsc->dentry_leases);  in ceph_mdsc_init()
4554 	INIT_LIST_HEAD(&mdsc->dentry_dir_leases);  in ceph_mdsc_init()
4557 	ceph_adjust_caps_max_min(mdsc, fsc->mount_options);  in ceph_mdsc_init()
4559 	spin_lock_init(&mdsc->snapid_map_lock);  in ceph_mdsc_init()
4560 	mdsc->snapid_map_tree = RB_ROOT;  in ceph_mdsc_init()
4561 	INIT_LIST_HEAD(&mdsc->snapid_map_lru);  in ceph_mdsc_init()
4563 	init_rwsem(&mdsc->pool_perm_rwsem);  in ceph_mdsc_init()
4564 	mdsc->pool_perm_tree = RB_ROOT;  in ceph_mdsc_init()
4566 	strscpy(mdsc->nodename, utsname()->nodename,  in ceph_mdsc_init()
4567 		sizeof(mdsc->nodename));  in ceph_mdsc_init()
4569 	fsc->mdsc = mdsc;  in ceph_mdsc_init()
4573 	kfree(mdsc->mdsmap);  in ceph_mdsc_init()
4585 	struct ceph_options *opts = mdsc->fsc->client->options;  in wait_requests()
4588 	mutex_lock(&mdsc->mutex);  in wait_requests()
4590 		mutex_unlock(&mdsc->mutex);  in wait_requests()
4593 		wait_for_completion_timeout(&mdsc->safe_umount_waiters,  in wait_requests()
4594 				    ceph_timeout_jiffies(opts->mount_timeout));  in wait_requests()
4597 		mutex_lock(&mdsc->mutex);  in wait_requests()
4600 			     req->r_tid);  in wait_requests()
4601 			list_del_init(&req->r_wait);  in wait_requests()
4605 	mutex_unlock(&mdsc->mutex);  in wait_requests()
4616 	mdsc->stopping = 1;  in ceph_mdsc_pre_umount()
4639 	mutex_lock(&mdsc->mutex);  in wait_unsafe_requests()
4643 	while (req && req->r_tid <= want_tid) {  in wait_unsafe_requests()
4645 		n = rb_next(&req->r_node);  in wait_unsafe_requests()
4650 		if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&  in wait_unsafe_requests()
4651 		    (req->r_op & CEPH_MDS_OP_WRITE)) {  in wait_unsafe_requests()
4656 			mutex_unlock(&mdsc->mutex);  in wait_unsafe_requests()
4658 			     req->r_tid, want_tid);  in wait_unsafe_requests()
4659 			wait_for_completion(&req->r_safe_completion);  in wait_unsafe_requests()
4660 			mutex_lock(&mdsc->mutex);  in wait_unsafe_requests()
4664 			if (RB_EMPTY_NODE(&nextreq->r_node)) {  in wait_unsafe_requests()
4673 	mutex_unlock(&mdsc->mutex);  in wait_unsafe_requests()
4681 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)  in ceph_mdsc_sync()
4685 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_sync()
4686 	want_tid = mdsc->last_tid;  in ceph_mdsc_sync()
4687 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_sync()
4690 	spin_lock(&mdsc->cap_dirty_lock);  in ceph_mdsc_sync()
4691 	want_flush = mdsc->last_cap_flush_tid;  in ceph_mdsc_sync()
4692 	if (!list_empty(&mdsc->cap_flush_list)) {  in ceph_mdsc_sync()
4694 			list_last_entry(&mdsc->cap_flush_list,  in ceph_mdsc_sync()
4696 		cf->wake = true;  in ceph_mdsc_sync()
4698 	spin_unlock(&mdsc->cap_dirty_lock);  in ceph_mdsc_sync()
4712 	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)  in done_closing_sessions()
4714 	return atomic_read(&mdsc->num_sessions) <= skipped;  in done_closing_sessions()
4722 	struct ceph_options *opts = mdsc->fsc->client->options;  in ceph_mdsc_close_sessions()
4730 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4731 	for (i = 0; i < mdsc->max_sessions; i++) {  in ceph_mdsc_close_sessions()
4735 		mutex_unlock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4736 		mutex_lock(&session->s_mutex);  in ceph_mdsc_close_sessions()
4739 		mutex_unlock(&session->s_mutex);  in ceph_mdsc_close_sessions()
4741 		mutex_lock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4743 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4746 	wait_event_timeout(mdsc->session_close_wq,  in ceph_mdsc_close_sessions()
4748 			   ceph_timeout_jiffies(opts->mount_timeout));  in ceph_mdsc_close_sessions()
4751 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4752 	for (i = 0; i < mdsc->max_sessions; i++) {  in ceph_mdsc_close_sessions()
4753 		if (mdsc->sessions[i]) {  in ceph_mdsc_close_sessions()
4754 			session = ceph_get_mds_session(mdsc->sessions[i]);  in ceph_mdsc_close_sessions()
4756 			mutex_unlock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4757 			mutex_lock(&session->s_mutex);  in ceph_mdsc_close_sessions()
4759 			mutex_unlock(&session->s_mutex);  in ceph_mdsc_close_sessions()
4761 			mutex_lock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4764 	WARN_ON(!list_empty(&mdsc->cap_delay_list));  in ceph_mdsc_close_sessions()
4765 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_close_sessions()
4770 	cancel_work_sync(&mdsc->cap_reclaim_work);  in ceph_mdsc_close_sessions()
4771 	cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */  in ceph_mdsc_close_sessions()
4783 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_force_umount()
4784 	for (mds = 0; mds < mdsc->max_sessions; mds++) {  in ceph_mdsc_force_umount()
4789 		if (session->s_state == CEPH_MDS_SESSION_REJECTED)  in ceph_mdsc_force_umount()
4791 		__wake_requests(mdsc, &session->s_waiting);  in ceph_mdsc_force_umount()
4792 		mutex_unlock(&mdsc->mutex);  in ceph_mdsc_force_umount()
4794 		mutex_lock(&session->s_mutex);  in ceph_mdsc_force_umount()
4796 		if (session->s_state == CEPH_MDS_SESSION_CLOSING) {  in ceph_mdsc_force_umount()
4800 		mutex_unlock(&session->s_mutex);  in ceph_mdsc_force_umount()
4803 		mutex_lock(&mdsc->mutex);  in ceph_mdsc_force_umount()
4806 	__wake_requests(mdsc, &mdsc->waiting_for_map);  in ceph_mdsc_force_umount()
4807 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_force_umount()
4819 	 * delayed work will re-arm itself again after that.  in ceph_mdsc_stop()
4821 	flush_delayed_work(&mdsc->delayed_work);  in ceph_mdsc_stop()
4823 	if (mdsc->mdsmap)  in ceph_mdsc_stop()
4824 		ceph_mdsmap_destroy(mdsc->mdsmap);  in ceph_mdsc_stop()
4825 	kfree(mdsc->sessions);  in ceph_mdsc_stop()
4832 	struct ceph_mds_client *mdsc = fsc->mdsc;  in ceph_mdsc_destroy()
4843 	ceph_metric_destroy(&mdsc->metric);  in ceph_mdsc_destroy()
4845 	flush_delayed_work(&mdsc->metric.delayed_work);  in ceph_mdsc_destroy()
4846 	fsc->mdsc = NULL;  in ceph_mdsc_destroy()
4853 	struct ceph_fs_client *fsc = mdsc->fsc;  in ceph_mdsc_handle_fsmap()
4854 	const char *mds_namespace = fsc->mount_options->mds_namespace;  in ceph_mdsc_handle_fsmap()
4855 	void *p = msg->front.iov_base;  in ceph_mdsc_handle_fsmap()
4856 	void *end = p + msg->front.iov_len;  in ceph_mdsc_handle_fsmap()  local
4860 	u32 mount_fscid = (u32)-1;  in ceph_mdsc_handle_fsmap()
4862 	int err = -EINVAL;  in ceph_mdsc_handle_fsmap()
4864 	ceph_decode_need(&p, end, sizeof(u32), bad);  in ceph_mdsc_handle_fsmap()
4869 	ceph_decode_need(&p, end, 2 + sizeof(u32), bad);  in ceph_mdsc_handle_fsmap()
4874 	ceph_decode_need(&p, end, sizeof(u32) * 3, bad);  in ceph_mdsc_handle_fsmap()
4878 	while (num_fs-- > 0) {  in ceph_mdsc_handle_fsmap()
4884 		ceph_decode_need(&p, end, 2 + sizeof(u32), bad);  in ceph_mdsc_handle_fsmap()
4888 		ceph_decode_need(&p, end, info_len, bad);  in ceph_mdsc_handle_fsmap()
4906 	ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);  in ceph_mdsc_handle_fsmap()
4907 	if (mount_fscid != (u32)-1) {  in ceph_mdsc_handle_fsmap()
4908 		fsc->client->monc.fs_cluster_id = mount_fscid;  in ceph_mdsc_handle_fsmap()
4909 		ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,  in ceph_mdsc_handle_fsmap()
4911 		ceph_monc_renew_subs(&fsc->client->monc);  in ceph_mdsc_handle_fsmap()
4913 		err = -ENOENT;  in ceph_mdsc_handle_fsmap()
4921 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_handle_fsmap()
4922 	mdsc->mdsmap_err = err;  in ceph_mdsc_handle_fsmap()
4923 	__wake_requests(mdsc, &mdsc->waiting_for_map);  in ceph_mdsc_handle_fsmap()
4924 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_handle_fsmap()
4934 	void *p = msg->front.iov_base;  in ceph_mdsc_handle_mdsmap()
4935 	void *end = p + msg->front.iov_len;  in ceph_mdsc_handle_mdsmap()  local
4938 	int err = -EINVAL;  in ceph_mdsc_handle_mdsmap()
4940 	ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);  in ceph_mdsc_handle_mdsmap()
4942 	if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)  in ceph_mdsc_handle_mdsmap()
4949 	mutex_lock(&mdsc->mutex);  in ceph_mdsc_handle_mdsmap()
4950 	if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {  in ceph_mdsc_handle_mdsmap()
4952 		     epoch, mdsc->mdsmap->m_epoch);  in ceph_mdsc_handle_mdsmap()
4953 		mutex_unlock(&mdsc->mutex);  in ceph_mdsc_handle_mdsmap()
4957 	newmap = ceph_mdsmap_decode(&p, end);  in ceph_mdsc_handle_mdsmap()
4964 	if (mdsc->mdsmap) {  in ceph_mdsc_handle_mdsmap()
4965 		oldmap = mdsc->mdsmap;  in ceph_mdsc_handle_mdsmap()
4966 		mdsc->mdsmap = newmap;  in ceph_mdsc_handle_mdsmap()
4970 		mdsc->mdsmap = newmap;  /* first mds map */  in ceph_mdsc_handle_mdsmap()
4972 	mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,  in ceph_mdsc_handle_mdsmap()
4975 	__wake_requests(mdsc, &mdsc->waiting_for_map);  in ceph_mdsc_handle_mdsmap()
4976 	ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,  in ceph_mdsc_handle_mdsmap()
4977 			  mdsc->mdsmap->m_epoch);  in ceph_mdsc_handle_mdsmap()
4979 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_handle_mdsmap()
4984 	mutex_unlock(&mdsc->mutex);  in ceph_mdsc_handle_mdsmap()
4992 	struct ceph_mds_session *s = con->private;  in con_get()
5001 	struct ceph_mds_session *s = con->private;  in con_put()
5012 	struct ceph_mds_session *s = con->private;  in peer_reset()
5013 	struct ceph_mds_client *mdsc = s->s_mdsc;  in peer_reset()
5015 	pr_warn("mds%d closed our session\n", s->s_mds);  in peer_reset()
5021 	struct ceph_mds_session *s = con->private;  in dispatch()
5022 	struct ceph_mds_client *mdsc = s->s_mdsc;  in dispatch()
5023 	int type = le16_to_cpu(msg->hdr.type);  in dispatch()
5025 	mutex_lock(&mdsc->mutex);  in dispatch()
5027 		mutex_unlock(&mdsc->mutex);  in dispatch()
5030 	mutex_unlock(&mdsc->mutex);  in dispatch()
5080 	struct ceph_mds_session *s = con->private;  in get_authorizer()
5081 	struct ceph_mds_client *mdsc = s->s_mdsc;  in get_authorizer()
5082 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;  in get_authorizer()
5083 	struct ceph_auth_handshake *auth = &s->s_auth;  in get_authorizer()
5085 	if (force_new && auth->authorizer) {  in get_authorizer()
5086 		ceph_auth_destroy_authorizer(auth->authorizer);  in get_authorizer()
5087 		auth->authorizer = NULL;  in get_authorizer()
5089 	if (!auth->authorizer) {  in get_authorizer()
5100 	*proto = ac->protocol;  in get_authorizer()
5108 	struct ceph_mds_session *s = con->private;  in add_authorizer_challenge()
5109 	struct ceph_mds_client *mdsc = s->s_mdsc;  in add_authorizer_challenge()
5110 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;  in add_authorizer_challenge()
5112 	return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,  in add_authorizer_challenge()
5118 	struct ceph_mds_session *s = con->private;  in verify_authorizer_reply()
5119 	struct ceph_mds_client *mdsc = s->s_mdsc;  in verify_authorizer_reply()
5120 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;  in verify_authorizer_reply()
5122 	return ceph_auth_verify_authorizer_reply(ac, s->s_auth.authorizer);  in verify_authorizer_reply()
5127 	struct ceph_mds_session *s = con->private;  in invalidate_authorizer()
5128 	struct ceph_mds_client *mdsc = s->s_mdsc;  in invalidate_authorizer()
5129 	struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;  in invalidate_authorizer()
5133 	return ceph_monc_validate_auth(&mdsc->fsc->client->monc);  in invalidate_authorizer()
5140 	int type = (int) le16_to_cpu(hdr->type);  in mds_alloc_msg()
5141 	int front_len = (int) le32_to_cpu(hdr->front_len);  in mds_alloc_msg()
5143 	if (con->in_msg)  in mds_alloc_msg()
5144 		return con->in_msg;  in mds_alloc_msg()
5159        struct ceph_mds_session *s = msg->con->private;  in mds_sign_message()
5160        struct ceph_auth_handshake *auth = &s->s_auth;  in mds_sign_message()
5167        struct ceph_mds_session *s = msg->con->private;  in mds_check_message_signature()
5168        struct ceph_auth_handshake *auth = &s->s_auth;  in mds_check_message_signature()