2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/addr.h>
44 #include <linux/jhash.h>
45 #include <linux/string_helpers.h>
46 #include <linux/fsnotify.h>
47 #include <linux/nfs_ssc.h>
51 #include "current_stateid.h"
55 #include "filecache.h"
58 #define NFSDDBG_FACILITY NFSDDBG_PROC
60 #define all_ones {{~0,~0},~0}
61 static const stateid_t one_stateid = {
63 .si_opaque = all_ones,
65 static const stateid_t zero_stateid = {
68 static const stateid_t currentstateid = {
71 static const stateid_t close_stateid = {
72 .si_generation = 0xffffffffU,
75 static u64 current_sessionid = 1;
77 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
78 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
79 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
80 #define CLOSE_STATEID(stateid) (!memcmp((stateid), &close_stateid, sizeof(stateid_t)))
82 /* forward declarations */
83 static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
84 static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
85 void nfsd4_end_grace(struct nfsd_net *nn);
86 static void _free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps);
91 * Currently used for the del_recall_lru and file hash table. In an
92 * effort to decrease the scope of the client_mutex, this spinlock may
93 * eventually cover more:
95 static DEFINE_SPINLOCK(state_lock);
97 enum nfsd4_st_mutex_lock_subclass {
98 OPEN_STATEID_MUTEX = 0,
99 LOCK_STATEID_MUTEX = 1,
103 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
104 * the refcount on the open stateid to drop.
106 static DECLARE_WAIT_QUEUE_HEAD(close_wq);
109 * A waitqueue where a writer to clients/#/ctl destroying a client can
110 * wait for cl_rpc_users to drop to 0 and then for the client to be
113 static DECLARE_WAIT_QUEUE_HEAD(expiry_wq);
115 static struct kmem_cache *client_slab;
116 static struct kmem_cache *openowner_slab;
117 static struct kmem_cache *lockowner_slab;
118 static struct kmem_cache *file_slab;
119 static struct kmem_cache *stateid_slab;
120 static struct kmem_cache *deleg_slab;
121 static struct kmem_cache *odstate_slab;
123 static void free_session(struct nfsd4_session *);
125 static const struct nfsd4_callback_ops nfsd4_cb_recall_ops;
126 static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops;
128 static bool is_session_dead(struct nfsd4_session *ses)
130 return ses->se_flags & NFS4_SESSION_DEAD;
133 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
135 if (atomic_read(&ses->se_ref) > ref_held_by_me)
136 return nfserr_jukebox;
137 ses->se_flags |= NFS4_SESSION_DEAD;
141 static bool is_client_expired(struct nfs4_client *clp)
143 return clp->cl_time == 0;
146 static __be32 get_client_locked(struct nfs4_client *clp)
148 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
150 lockdep_assert_held(&nn->client_lock);
152 if (is_client_expired(clp))
153 return nfserr_expired;
154 atomic_inc(&clp->cl_rpc_users);
158 /* must be called under the client_lock */
160 renew_client_locked(struct nfs4_client *clp)
162 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
164 if (is_client_expired(clp)) {
166 printk("%s: client (clientid %08x/%08x) already expired\n",
168 clp->cl_clientid.cl_boot,
169 clp->cl_clientid.cl_id);
173 list_move_tail(&clp->cl_lru, &nn->client_lru);
174 clp->cl_time = ktime_get_boottime_seconds();
177 static void put_client_renew_locked(struct nfs4_client *clp)
179 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
181 lockdep_assert_held(&nn->client_lock);
183 if (!atomic_dec_and_test(&clp->cl_rpc_users))
185 if (!is_client_expired(clp))
186 renew_client_locked(clp);
188 wake_up_all(&expiry_wq);
191 static void put_client_renew(struct nfs4_client *clp)
193 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
195 if (!atomic_dec_and_lock(&clp->cl_rpc_users, &nn->client_lock))
197 if (!is_client_expired(clp))
198 renew_client_locked(clp);
200 wake_up_all(&expiry_wq);
201 spin_unlock(&nn->client_lock);
204 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
208 if (is_session_dead(ses))
209 return nfserr_badsession;
210 status = get_client_locked(ses->se_client);
213 atomic_inc(&ses->se_ref);
217 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
219 struct nfs4_client *clp = ses->se_client;
220 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
222 lockdep_assert_held(&nn->client_lock);
224 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
226 put_client_renew_locked(clp);
229 static void nfsd4_put_session(struct nfsd4_session *ses)
231 struct nfs4_client *clp = ses->se_client;
232 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
234 spin_lock(&nn->client_lock);
235 nfsd4_put_session_locked(ses);
236 spin_unlock(&nn->client_lock);
239 static struct nfsd4_blocked_lock *
240 find_blocked_lock(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
243 struct nfsd4_blocked_lock *cur, *found = NULL;
245 spin_lock(&nn->blocked_locks_lock);
246 list_for_each_entry(cur, &lo->lo_blocked, nbl_list) {
247 if (fh_match(fh, &cur->nbl_fh)) {
248 list_del_init(&cur->nbl_list);
249 WARN_ON(list_empty(&cur->nbl_lru));
250 list_del_init(&cur->nbl_lru);
255 spin_unlock(&nn->blocked_locks_lock);
257 locks_delete_block(&found->nbl_lock);
261 static struct nfsd4_blocked_lock *
262 find_or_allocate_block(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
265 struct nfsd4_blocked_lock *nbl;
267 nbl = find_blocked_lock(lo, fh, nn);
269 nbl= kmalloc(sizeof(*nbl), GFP_KERNEL);
271 INIT_LIST_HEAD(&nbl->nbl_list);
272 INIT_LIST_HEAD(&nbl->nbl_lru);
273 fh_copy_shallow(&nbl->nbl_fh, fh);
274 locks_init_lock(&nbl->nbl_lock);
275 kref_init(&nbl->nbl_kref);
276 nfsd4_init_cb(&nbl->nbl_cb, lo->lo_owner.so_client,
277 &nfsd4_cb_notify_lock_ops,
278 NFSPROC4_CLNT_CB_NOTIFY_LOCK);
285 free_nbl(struct kref *kref)
287 struct nfsd4_blocked_lock *nbl;
289 nbl = container_of(kref, struct nfsd4_blocked_lock, nbl_kref);
294 free_blocked_lock(struct nfsd4_blocked_lock *nbl)
296 locks_delete_block(&nbl->nbl_lock);
297 locks_release_private(&nbl->nbl_lock);
298 kref_put(&nbl->nbl_kref, free_nbl);
302 remove_blocked_locks(struct nfs4_lockowner *lo)
304 struct nfs4_client *clp = lo->lo_owner.so_client;
305 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
306 struct nfsd4_blocked_lock *nbl;
309 /* Dequeue all blocked locks */
310 spin_lock(&nn->blocked_locks_lock);
311 while (!list_empty(&lo->lo_blocked)) {
312 nbl = list_first_entry(&lo->lo_blocked,
313 struct nfsd4_blocked_lock,
315 list_del_init(&nbl->nbl_list);
316 WARN_ON(list_empty(&nbl->nbl_lru));
317 list_move(&nbl->nbl_lru, &reaplist);
319 spin_unlock(&nn->blocked_locks_lock);
322 while (!list_empty(&reaplist)) {
323 nbl = list_first_entry(&reaplist, struct nfsd4_blocked_lock,
325 list_del_init(&nbl->nbl_lru);
326 free_blocked_lock(nbl);
331 nfsd4_cb_notify_lock_prepare(struct nfsd4_callback *cb)
333 struct nfsd4_blocked_lock *nbl = container_of(cb,
334 struct nfsd4_blocked_lock, nbl_cb);
335 locks_delete_block(&nbl->nbl_lock);
339 nfsd4_cb_notify_lock_done(struct nfsd4_callback *cb, struct rpc_task *task)
342 * Since this is just an optimization, we don't try very hard if it
343 * turns out not to succeed. We'll requeue it on NFS4ERR_DELAY, and
344 * just quit trying on anything else.
346 switch (task->tk_status) {
348 rpc_delay(task, 1 * HZ);
356 nfsd4_cb_notify_lock_release(struct nfsd4_callback *cb)
358 struct nfsd4_blocked_lock *nbl = container_of(cb,
359 struct nfsd4_blocked_lock, nbl_cb);
361 free_blocked_lock(nbl);
364 static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops = {
365 .prepare = nfsd4_cb_notify_lock_prepare,
366 .done = nfsd4_cb_notify_lock_done,
367 .release = nfsd4_cb_notify_lock_release,
371 * We store the NONE, READ, WRITE, and BOTH bits separately in the
372 * st_{access,deny}_bmap field of the stateid, in order to track not
373 * only what share bits are currently in force, but also what
374 * combinations of share bits previous opens have used. This allows us
375 * to enforce the recommendation in
376 * https://datatracker.ietf.org/doc/html/rfc7530#section-16.19.4 that
377 * the server return an error if the client attempt to downgrade to a
378 * combination of share bits not explicable by closing some of its
381 * This enforcement is arguably incomplete, since we don't keep
382 * track of access/deny bit combinations; so, e.g., we allow:
384 * OPEN allow read, deny write
385 * OPEN allow both, deny none
386 * DOWNGRADE allow read, deny none
388 * which we should reject.
390 * But you could also argue that our current code is already overkill,
391 * since it only exists to return NFS4ERR_INVAL on incorrect client
395 bmap_to_share_mode(unsigned long bmap)
398 unsigned int access = 0;
400 for (i = 1; i < 4; i++) {
401 if (test_bit(i, &bmap))
407 /* set share access for a given stateid */
409 set_access(u32 access, struct nfs4_ol_stateid *stp)
411 unsigned char mask = 1 << access;
413 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
414 stp->st_access_bmap |= mask;
417 /* clear share access for a given stateid */
419 clear_access(u32 access, struct nfs4_ol_stateid *stp)
421 unsigned char mask = 1 << access;
423 WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
424 stp->st_access_bmap &= ~mask;
427 /* test whether a given stateid has access */
429 test_access(u32 access, struct nfs4_ol_stateid *stp)
431 unsigned char mask = 1 << access;
433 return (bool)(stp->st_access_bmap & mask);
436 /* set share deny for a given stateid */
438 set_deny(u32 deny, struct nfs4_ol_stateid *stp)
440 unsigned char mask = 1 << deny;
442 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
443 stp->st_deny_bmap |= mask;
446 /* clear share deny for a given stateid */
448 clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
450 unsigned char mask = 1 << deny;
452 WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
453 stp->st_deny_bmap &= ~mask;
456 /* test whether a given stateid is denying specific access */
458 test_deny(u32 deny, struct nfs4_ol_stateid *stp)
460 unsigned char mask = 1 << deny;
462 return (bool)(stp->st_deny_bmap & mask);
465 static int nfs4_access_to_omode(u32 access)
467 switch (access & NFS4_SHARE_ACCESS_BOTH) {
468 case NFS4_SHARE_ACCESS_READ:
470 case NFS4_SHARE_ACCESS_WRITE:
472 case NFS4_SHARE_ACCESS_BOTH:
480 access_permit_read(struct nfs4_ol_stateid *stp)
482 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
483 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
484 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
488 access_permit_write(struct nfs4_ol_stateid *stp)
490 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
491 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
494 static inline struct nfs4_stateowner *
495 nfs4_get_stateowner(struct nfs4_stateowner *sop)
497 atomic_inc(&sop->so_count);
502 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
504 return (sop->so_owner.len == owner->len) &&
505 0 == memcmp(sop->so_owner.data, owner->data, owner->len);
508 static struct nfs4_openowner *
509 find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
510 struct nfs4_client *clp)
512 struct nfs4_stateowner *so;
514 lockdep_assert_held(&clp->cl_lock);
516 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
518 if (!so->so_is_open_owner)
520 if (same_owner_str(so, &open->op_owner))
521 return openowner(nfs4_get_stateowner(so));
526 static struct nfs4_openowner *
527 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
528 struct nfs4_client *clp)
530 struct nfs4_openowner *oo;
532 spin_lock(&clp->cl_lock);
533 oo = find_openstateowner_str_locked(hashval, open, clp);
534 spin_unlock(&clp->cl_lock);
539 opaque_hashval(const void *ptr, int nbytes)
541 unsigned char *cptr = (unsigned char *) ptr;
551 static void nfsd4_free_file_rcu(struct rcu_head *rcu)
553 struct nfs4_file *fp = container_of(rcu, struct nfs4_file, fi_rcu);
555 kmem_cache_free(file_slab, fp);
559 put_nfs4_file(struct nfs4_file *fi)
561 might_lock(&state_lock);
563 if (refcount_dec_and_lock(&fi->fi_ref, &state_lock)) {
564 hlist_del_rcu(&fi->fi_hash);
565 spin_unlock(&state_lock);
566 WARN_ON_ONCE(!list_empty(&fi->fi_clnt_odstate));
567 WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
568 call_rcu(&fi->fi_rcu, nfsd4_free_file_rcu);
572 static struct nfsd_file *
573 __nfs4_get_fd(struct nfs4_file *f, int oflag)
575 if (f->fi_fds[oflag])
576 return nfsd_file_get(f->fi_fds[oflag]);
580 static struct nfsd_file *
581 find_writeable_file_locked(struct nfs4_file *f)
583 struct nfsd_file *ret;
585 lockdep_assert_held(&f->fi_lock);
587 ret = __nfs4_get_fd(f, O_WRONLY);
589 ret = __nfs4_get_fd(f, O_RDWR);
593 static struct nfsd_file *
594 find_writeable_file(struct nfs4_file *f)
596 struct nfsd_file *ret;
598 spin_lock(&f->fi_lock);
599 ret = find_writeable_file_locked(f);
600 spin_unlock(&f->fi_lock);
605 static struct nfsd_file *
606 find_readable_file_locked(struct nfs4_file *f)
608 struct nfsd_file *ret;
610 lockdep_assert_held(&f->fi_lock);
612 ret = __nfs4_get_fd(f, O_RDONLY);
614 ret = __nfs4_get_fd(f, O_RDWR);
618 static struct nfsd_file *
619 find_readable_file(struct nfs4_file *f)
621 struct nfsd_file *ret;
623 spin_lock(&f->fi_lock);
624 ret = find_readable_file_locked(f);
625 spin_unlock(&f->fi_lock);
631 find_any_file(struct nfs4_file *f)
633 struct nfsd_file *ret;
637 spin_lock(&f->fi_lock);
638 ret = __nfs4_get_fd(f, O_RDWR);
640 ret = __nfs4_get_fd(f, O_WRONLY);
642 ret = __nfs4_get_fd(f, O_RDONLY);
644 spin_unlock(&f->fi_lock);
648 static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
650 struct nfsd_file *ret = NULL;
652 spin_lock(&f->fi_lock);
653 if (f->fi_deleg_file)
654 ret = nfsd_file_get(f->fi_deleg_file);
655 spin_unlock(&f->fi_lock);
659 static atomic_long_t num_delegations;
660 unsigned long max_delegations;
663 * Open owner state (share locks)
666 /* hash tables for lock and open owners */
667 #define OWNER_HASH_BITS 8
668 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
669 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
671 static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
675 ret = opaque_hashval(ownername->data, ownername->len);
676 return ret & OWNER_HASH_MASK;
679 /* hash table for nfs4_file */
680 #define FILE_HASH_BITS 8
681 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
683 static unsigned int file_hashval(struct svc_fh *fh)
685 struct inode *inode = d_inode(fh->fh_dentry);
687 /* XXX: why not (here & in file cache) use inode? */
688 return (unsigned int)hash_long(inode->i_ino, FILE_HASH_BITS);
691 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
694 __nfs4_file_get_access(struct nfs4_file *fp, u32 access)
696 lockdep_assert_held(&fp->fi_lock);
698 if (access & NFS4_SHARE_ACCESS_WRITE)
699 atomic_inc(&fp->fi_access[O_WRONLY]);
700 if (access & NFS4_SHARE_ACCESS_READ)
701 atomic_inc(&fp->fi_access[O_RDONLY]);
705 nfs4_file_get_access(struct nfs4_file *fp, u32 access)
707 lockdep_assert_held(&fp->fi_lock);
709 /* Does this access mode make sense? */
710 if (access & ~NFS4_SHARE_ACCESS_BOTH)
713 /* Does it conflict with a deny mode already set? */
714 if ((access & fp->fi_share_deny) != 0)
715 return nfserr_share_denied;
717 __nfs4_file_get_access(fp, access);
721 static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
723 /* Common case is that there is no deny mode. */
725 /* Does this deny mode make sense? */
726 if (deny & ~NFS4_SHARE_DENY_BOTH)
729 if ((deny & NFS4_SHARE_DENY_READ) &&
730 atomic_read(&fp->fi_access[O_RDONLY]))
731 return nfserr_share_denied;
733 if ((deny & NFS4_SHARE_DENY_WRITE) &&
734 atomic_read(&fp->fi_access[O_WRONLY]))
735 return nfserr_share_denied;
740 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
742 might_lock(&fp->fi_lock);
744 if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
745 struct nfsd_file *f1 = NULL;
746 struct nfsd_file *f2 = NULL;
748 swap(f1, fp->fi_fds[oflag]);
749 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
750 swap(f2, fp->fi_fds[O_RDWR]);
751 spin_unlock(&fp->fi_lock);
759 static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
761 WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
763 if (access & NFS4_SHARE_ACCESS_WRITE)
764 __nfs4_file_put_access(fp, O_WRONLY);
765 if (access & NFS4_SHARE_ACCESS_READ)
766 __nfs4_file_put_access(fp, O_RDONLY);
770 * Allocate a new open/delegation state counter. This is needed for
771 * pNFS for proper return on close semantics.
773 * Note that we only allocate it for pNFS-enabled exports, otherwise
774 * all pointers to struct nfs4_clnt_odstate are always NULL.
776 static struct nfs4_clnt_odstate *
777 alloc_clnt_odstate(struct nfs4_client *clp)
779 struct nfs4_clnt_odstate *co;
781 co = kmem_cache_zalloc(odstate_slab, GFP_KERNEL);
784 refcount_set(&co->co_odcount, 1);
790 hash_clnt_odstate_locked(struct nfs4_clnt_odstate *co)
792 struct nfs4_file *fp = co->co_file;
794 lockdep_assert_held(&fp->fi_lock);
795 list_add(&co->co_perfile, &fp->fi_clnt_odstate);
799 get_clnt_odstate(struct nfs4_clnt_odstate *co)
802 refcount_inc(&co->co_odcount);
806 put_clnt_odstate(struct nfs4_clnt_odstate *co)
808 struct nfs4_file *fp;
814 if (refcount_dec_and_lock(&co->co_odcount, &fp->fi_lock)) {
815 list_del(&co->co_perfile);
816 spin_unlock(&fp->fi_lock);
818 nfsd4_return_all_file_layouts(co->co_client, fp);
819 kmem_cache_free(odstate_slab, co);
823 static struct nfs4_clnt_odstate *
824 find_or_hash_clnt_odstate(struct nfs4_file *fp, struct nfs4_clnt_odstate *new)
826 struct nfs4_clnt_odstate *co;
827 struct nfs4_client *cl;
834 spin_lock(&fp->fi_lock);
835 list_for_each_entry(co, &fp->fi_clnt_odstate, co_perfile) {
836 if (co->co_client == cl) {
837 get_clnt_odstate(co);
843 hash_clnt_odstate_locked(new);
845 spin_unlock(&fp->fi_lock);
849 struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab,
850 void (*sc_free)(struct nfs4_stid *))
852 struct nfs4_stid *stid;
855 stid = kmem_cache_zalloc(slab, GFP_KERNEL);
859 idr_preload(GFP_KERNEL);
860 spin_lock(&cl->cl_lock);
861 /* Reserving 0 for start of file in nfsdfs "states" file: */
862 new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 1, 0, GFP_NOWAIT);
863 spin_unlock(&cl->cl_lock);
868 stid->sc_free = sc_free;
869 stid->sc_client = cl;
870 stid->sc_stateid.si_opaque.so_id = new_id;
871 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
872 /* Will be incremented before return to client: */
873 refcount_set(&stid->sc_count, 1);
874 spin_lock_init(&stid->sc_lock);
875 INIT_LIST_HEAD(&stid->sc_cp_list);
878 * It shouldn't be a problem to reuse an opaque stateid value.
879 * I don't think it is for 4.1. But with 4.0 I worry that, for
880 * example, a stray write retransmission could be accepted by
881 * the server when it should have been rejected. Therefore,
882 * adopt a trick from the sctp code to attempt to maximize the
883 * amount of time until an id is reused, by ensuring they always
884 * "increase" (mod INT_MAX):
888 kmem_cache_free(slab, stid);
893 * Create a unique stateid_t to represent each COPY.
895 static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid,
896 unsigned char sc_type)
900 stid->stid.si_opaque.so_clid.cl_boot = (u32)nn->boot_time;
901 stid->stid.si_opaque.so_clid.cl_id = nn->s2s_cp_cl_id;
902 stid->sc_type = sc_type;
904 idr_preload(GFP_KERNEL);
905 spin_lock(&nn->s2s_cp_lock);
906 new_id = idr_alloc_cyclic(&nn->s2s_cp_stateids, stid, 0, 0, GFP_NOWAIT);
907 stid->stid.si_opaque.so_id = new_id;
908 stid->stid.si_generation = 1;
909 spin_unlock(&nn->s2s_cp_lock);
916 int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_copy *copy)
918 return nfs4_init_cp_state(nn, ©->cp_stateid, NFS4_COPY_STID);
921 struct nfs4_cpntf_state *nfs4_alloc_init_cpntf_state(struct nfsd_net *nn,
922 struct nfs4_stid *p_stid)
924 struct nfs4_cpntf_state *cps;
926 cps = kzalloc(sizeof(struct nfs4_cpntf_state), GFP_KERNEL);
929 cps->cpntf_time = ktime_get_boottime_seconds();
930 refcount_set(&cps->cp_stateid.sc_count, 1);
931 if (!nfs4_init_cp_state(nn, &cps->cp_stateid, NFS4_COPYNOTIFY_STID))
933 spin_lock(&nn->s2s_cp_lock);
934 list_add(&cps->cp_list, &p_stid->sc_cp_list);
935 spin_unlock(&nn->s2s_cp_lock);
942 void nfs4_free_copy_state(struct nfsd4_copy *copy)
946 WARN_ON_ONCE(copy->cp_stateid.sc_type != NFS4_COPY_STID);
947 nn = net_generic(copy->cp_clp->net, nfsd_net_id);
948 spin_lock(&nn->s2s_cp_lock);
949 idr_remove(&nn->s2s_cp_stateids,
950 copy->cp_stateid.stid.si_opaque.so_id);
951 spin_unlock(&nn->s2s_cp_lock);
954 static void nfs4_free_cpntf_statelist(struct net *net, struct nfs4_stid *stid)
956 struct nfs4_cpntf_state *cps;
959 nn = net_generic(net, nfsd_net_id);
960 spin_lock(&nn->s2s_cp_lock);
961 while (!list_empty(&stid->sc_cp_list)) {
962 cps = list_first_entry(&stid->sc_cp_list,
963 struct nfs4_cpntf_state, cp_list);
964 _free_cpntf_state_locked(nn, cps);
966 spin_unlock(&nn->s2s_cp_lock);
969 static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
971 struct nfs4_stid *stid;
973 stid = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_ol_stateid);
977 return openlockstateid(stid);
980 static void nfs4_free_deleg(struct nfs4_stid *stid)
982 kmem_cache_free(deleg_slab, stid);
983 atomic_long_dec(&num_delegations);
987 * When we recall a delegation, we should be careful not to hand it
988 * out again straight away.
989 * To ensure this we keep a pair of bloom filters ('new' and 'old')
990 * in which the filehandles of recalled delegations are "stored".
991 * If a filehandle appear in either filter, a delegation is blocked.
992 * When a delegation is recalled, the filehandle is stored in the "new"
994 * Every 30 seconds we swap the filters and clear the "new" one,
995 * unless both are empty of course.
997 * Each filter is 256 bits. We hash the filehandle to 32bit and use the
998 * low 3 bytes as hash-table indices.
1000 * 'blocked_delegations_lock', which is always taken in block_delegations(),
1001 * is used to manage concurrent access. Testing does not need the lock
1002 * except when swapping the two filters.
1004 static DEFINE_SPINLOCK(blocked_delegations_lock);
1005 static struct bloom_pair {
1006 int entries, old_entries;
1008 int new; /* index into 'set' */
1009 DECLARE_BITMAP(set[2], 256);
1010 } blocked_delegations;
1012 static int delegation_blocked(struct knfsd_fh *fh)
1015 struct bloom_pair *bd = &blocked_delegations;
1017 if (bd->entries == 0)
1019 if (ktime_get_seconds() - bd->swap_time > 30) {
1020 spin_lock(&blocked_delegations_lock);
1021 if (ktime_get_seconds() - bd->swap_time > 30) {
1022 bd->entries -= bd->old_entries;
1023 bd->old_entries = bd->entries;
1024 memset(bd->set[bd->new], 0,
1025 sizeof(bd->set[0]));
1026 bd->new = 1-bd->new;
1027 bd->swap_time = ktime_get_seconds();
1029 spin_unlock(&blocked_delegations_lock);
1031 hash = jhash(&fh->fh_raw, fh->fh_size, 0);
1032 if (test_bit(hash&255, bd->set[0]) &&
1033 test_bit((hash>>8)&255, bd->set[0]) &&
1034 test_bit((hash>>16)&255, bd->set[0]))
1037 if (test_bit(hash&255, bd->set[1]) &&
1038 test_bit((hash>>8)&255, bd->set[1]) &&
1039 test_bit((hash>>16)&255, bd->set[1]))
1045 static void block_delegations(struct knfsd_fh *fh)
1048 struct bloom_pair *bd = &blocked_delegations;
1050 hash = jhash(&fh->fh_raw, fh->fh_size, 0);
1052 spin_lock(&blocked_delegations_lock);
1053 __set_bit(hash&255, bd->set[bd->new]);
1054 __set_bit((hash>>8)&255, bd->set[bd->new]);
1055 __set_bit((hash>>16)&255, bd->set[bd->new]);
1056 if (bd->entries == 0)
1057 bd->swap_time = ktime_get_seconds();
1059 spin_unlock(&blocked_delegations_lock);
1062 static struct nfs4_delegation *
1063 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_file *fp,
1064 struct svc_fh *current_fh,
1065 struct nfs4_clnt_odstate *odstate)
1067 struct nfs4_delegation *dp;
1070 dprintk("NFSD alloc_init_deleg\n");
1071 n = atomic_long_inc_return(&num_delegations);
1072 if (n < 0 || n > max_delegations)
1074 if (delegation_blocked(¤t_fh->fh_handle))
1076 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab, nfs4_free_deleg));
1081 * delegation seqid's are never incremented. The 4.1 special
1082 * meaning of seqid 0 isn't meaningful, really, but let's avoid
1083 * 0 anyway just for consistency and use 1:
1085 dp->dl_stid.sc_stateid.si_generation = 1;
1086 INIT_LIST_HEAD(&dp->dl_perfile);
1087 INIT_LIST_HEAD(&dp->dl_perclnt);
1088 INIT_LIST_HEAD(&dp->dl_recall_lru);
1089 dp->dl_clnt_odstate = odstate;
1090 get_clnt_odstate(odstate);
1091 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
1093 nfsd4_init_cb(&dp->dl_recall, dp->dl_stid.sc_client,
1094 &nfsd4_cb_recall_ops, NFSPROC4_CLNT_CB_RECALL);
1096 dp->dl_stid.sc_file = fp;
1099 atomic_long_dec(&num_delegations);
1104 nfs4_put_stid(struct nfs4_stid *s)
1106 struct nfs4_file *fp = s->sc_file;
1107 struct nfs4_client *clp = s->sc_client;
1109 might_lock(&clp->cl_lock);
1111 if (!refcount_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
1112 wake_up_all(&close_wq);
1115 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1116 nfs4_free_cpntf_statelist(clp->net, s);
1117 spin_unlock(&clp->cl_lock);
1124 nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid)
1126 stateid_t *src = &stid->sc_stateid;
1128 spin_lock(&stid->sc_lock);
1129 if (unlikely(++src->si_generation == 0))
1130 src->si_generation = 1;
1131 memcpy(dst, src, sizeof(*dst));
1132 spin_unlock(&stid->sc_lock);
1135 static void put_deleg_file(struct nfs4_file *fp)
1137 struct nfsd_file *nf = NULL;
1139 spin_lock(&fp->fi_lock);
1140 if (--fp->fi_delegees == 0)
1141 swap(nf, fp->fi_deleg_file);
1142 spin_unlock(&fp->fi_lock);
1148 static void nfs4_unlock_deleg_lease(struct nfs4_delegation *dp)
1150 struct nfs4_file *fp = dp->dl_stid.sc_file;
1151 struct nfsd_file *nf = fp->fi_deleg_file;
1153 WARN_ON_ONCE(!fp->fi_delegees);
1155 vfs_setlease(nf->nf_file, F_UNLCK, NULL, (void **)&dp);
1159 static void destroy_unhashed_deleg(struct nfs4_delegation *dp)
1161 put_clnt_odstate(dp->dl_clnt_odstate);
1162 nfs4_unlock_deleg_lease(dp);
1163 nfs4_put_stid(&dp->dl_stid);
1166 void nfs4_unhash_stid(struct nfs4_stid *s)
1172 * nfs4_delegation_exists - Discover if this delegation already exists
1173 * @clp: a pointer to the nfs4_client we're granting a delegation to
1174 * @fp: a pointer to the nfs4_file we're granting a delegation on
1177 * On success: true iff an existing delegation is found
1181 nfs4_delegation_exists(struct nfs4_client *clp, struct nfs4_file *fp)
1183 struct nfs4_delegation *searchdp = NULL;
1184 struct nfs4_client *searchclp = NULL;
1186 lockdep_assert_held(&state_lock);
1187 lockdep_assert_held(&fp->fi_lock);
1189 list_for_each_entry(searchdp, &fp->fi_delegations, dl_perfile) {
1190 searchclp = searchdp->dl_stid.sc_client;
1191 if (clp == searchclp) {
1199 * hash_delegation_locked - Add a delegation to the appropriate lists
1200 * @dp: a pointer to the nfs4_delegation we are adding.
1201 * @fp: a pointer to the nfs4_file we're granting a delegation on
1204 * On success: NULL if the delegation was successfully hashed.
1206 * On error: -EAGAIN if one was previously granted to this
1207 * nfs4_client for this nfs4_file. Delegation is not hashed.
1212 hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
1214 struct nfs4_client *clp = dp->dl_stid.sc_client;
1216 lockdep_assert_held(&state_lock);
1217 lockdep_assert_held(&fp->fi_lock);
1219 if (nfs4_delegation_exists(clp, fp))
1221 refcount_inc(&dp->dl_stid.sc_count);
1222 dp->dl_stid.sc_type = NFS4_DELEG_STID;
1223 list_add(&dp->dl_perfile, &fp->fi_delegations);
1224 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1228 static bool delegation_hashed(struct nfs4_delegation *dp)
1230 return !(list_empty(&dp->dl_perfile));
1234 unhash_delegation_locked(struct nfs4_delegation *dp)
1236 struct nfs4_file *fp = dp->dl_stid.sc_file;
1238 lockdep_assert_held(&state_lock);
1240 if (!delegation_hashed(dp))
1243 dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
1244 /* Ensure that deleg break won't try to requeue it */
1246 spin_lock(&fp->fi_lock);
1247 list_del_init(&dp->dl_perclnt);
1248 list_del_init(&dp->dl_recall_lru);
1249 list_del_init(&dp->dl_perfile);
1250 spin_unlock(&fp->fi_lock);
1254 static void destroy_delegation(struct nfs4_delegation *dp)
1258 spin_lock(&state_lock);
1259 unhashed = unhash_delegation_locked(dp);
1260 spin_unlock(&state_lock);
1262 destroy_unhashed_deleg(dp);
1265 static void revoke_delegation(struct nfs4_delegation *dp)
1267 struct nfs4_client *clp = dp->dl_stid.sc_client;
1269 WARN_ON(!list_empty(&dp->dl_recall_lru));
1271 if (clp->cl_minorversion) {
1272 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
1273 refcount_inc(&dp->dl_stid.sc_count);
1274 spin_lock(&clp->cl_lock);
1275 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
1276 spin_unlock(&clp->cl_lock);
1278 destroy_unhashed_deleg(dp);
1285 static unsigned int clientid_hashval(u32 id)
1287 return id & CLIENT_HASH_MASK;
1290 static unsigned int clientstr_hashval(struct xdr_netobj name)
1292 return opaque_hashval(name.data, 8) & CLIENT_HASH_MASK;
1296 * A stateid that had a deny mode associated with it is being released
1297 * or downgraded. Recalculate the deny mode on the file.
1300 recalculate_deny_mode(struct nfs4_file *fp)
1302 struct nfs4_ol_stateid *stp;
1304 spin_lock(&fp->fi_lock);
1305 fp->fi_share_deny = 0;
1306 list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
1307 fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
1308 spin_unlock(&fp->fi_lock);
1312 reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
1315 bool change = false;
1317 for (i = 1; i < 4; i++) {
1318 if ((i & deny) != i) {
1324 /* Recalculate per-file deny mode if there was a change */
1326 recalculate_deny_mode(stp->st_stid.sc_file);
1329 /* release all access and file references for a given stateid */
1331 release_all_access(struct nfs4_ol_stateid *stp)
1334 struct nfs4_file *fp = stp->st_stid.sc_file;
1336 if (fp && stp->st_deny_bmap != 0)
1337 recalculate_deny_mode(fp);
1339 for (i = 1; i < 4; i++) {
1340 if (test_access(i, stp))
1341 nfs4_file_put_access(stp->st_stid.sc_file, i);
1342 clear_access(i, stp);
1346 static inline void nfs4_free_stateowner(struct nfs4_stateowner *sop)
1348 kfree(sop->so_owner.data);
1349 sop->so_ops->so_free(sop);
1352 static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
1354 struct nfs4_client *clp = sop->so_client;
1356 might_lock(&clp->cl_lock);
1358 if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1360 sop->so_ops->so_unhash(sop);
1361 spin_unlock(&clp->cl_lock);
1362 nfs4_free_stateowner(sop);
1366 nfs4_ol_stateid_unhashed(const struct nfs4_ol_stateid *stp)
1368 return list_empty(&stp->st_perfile);
1371 static bool unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1373 struct nfs4_file *fp = stp->st_stid.sc_file;
1375 lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1377 if (list_empty(&stp->st_perfile))
1380 spin_lock(&fp->fi_lock);
1381 list_del_init(&stp->st_perfile);
1382 spin_unlock(&fp->fi_lock);
1383 list_del(&stp->st_perstateowner);
1387 static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1389 struct nfs4_ol_stateid *stp = openlockstateid(stid);
1391 put_clnt_odstate(stp->st_clnt_odstate);
1392 release_all_access(stp);
1393 if (stp->st_stateowner)
1394 nfs4_put_stateowner(stp->st_stateowner);
1395 kmem_cache_free(stateid_slab, stid);
1398 static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1400 struct nfs4_ol_stateid *stp = openlockstateid(stid);
1401 struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
1402 struct nfsd_file *nf;
1404 nf = find_any_file(stp->st_stid.sc_file);
1406 get_file(nf->nf_file);
1407 filp_close(nf->nf_file, (fl_owner_t)lo);
1410 nfs4_free_ol_stateid(stid);
1414 * Put the persistent reference to an already unhashed generic stateid, while
1415 * holding the cl_lock. If it's the last reference, then put it onto the
1416 * reaplist for later destruction.
1418 static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1419 struct list_head *reaplist)
1421 struct nfs4_stid *s = &stp->st_stid;
1422 struct nfs4_client *clp = s->sc_client;
1424 lockdep_assert_held(&clp->cl_lock);
1426 WARN_ON_ONCE(!list_empty(&stp->st_locks));
1428 if (!refcount_dec_and_test(&s->sc_count)) {
1429 wake_up_all(&close_wq);
1433 idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1434 list_add(&stp->st_locks, reaplist);
1437 static bool unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1439 lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1441 if (!unhash_ol_stateid(stp))
1443 list_del_init(&stp->st_locks);
1444 nfs4_unhash_stid(&stp->st_stid);
1448 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1450 struct nfs4_client *clp = stp->st_stid.sc_client;
1453 spin_lock(&clp->cl_lock);
1454 unhashed = unhash_lock_stateid(stp);
1455 spin_unlock(&clp->cl_lock);
1457 nfs4_put_stid(&stp->st_stid);
1460 static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1462 struct nfs4_client *clp = lo->lo_owner.so_client;
1464 lockdep_assert_held(&clp->cl_lock);
1466 list_del_init(&lo->lo_owner.so_strhash);
1470 * Free a list of generic stateids that were collected earlier after being
1474 free_ol_stateid_reaplist(struct list_head *reaplist)
1476 struct nfs4_ol_stateid *stp;
1477 struct nfs4_file *fp;
1481 while (!list_empty(reaplist)) {
1482 stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1484 list_del(&stp->st_locks);
1485 fp = stp->st_stid.sc_file;
1486 stp->st_stid.sc_free(&stp->st_stid);
1492 static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1493 struct list_head *reaplist)
1495 struct nfs4_ol_stateid *stp;
1497 lockdep_assert_held(&open_stp->st_stid.sc_client->cl_lock);
1499 while (!list_empty(&open_stp->st_locks)) {
1500 stp = list_entry(open_stp->st_locks.next,
1501 struct nfs4_ol_stateid, st_locks);
1502 WARN_ON(!unhash_lock_stateid(stp));
1503 put_ol_stateid_locked(stp, reaplist);
1507 static bool unhash_open_stateid(struct nfs4_ol_stateid *stp,
1508 struct list_head *reaplist)
1510 lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1512 if (!unhash_ol_stateid(stp))
1514 release_open_stateid_locks(stp, reaplist);
1518 static void release_open_stateid(struct nfs4_ol_stateid *stp)
1520 LIST_HEAD(reaplist);
1522 spin_lock(&stp->st_stid.sc_client->cl_lock);
1523 if (unhash_open_stateid(stp, &reaplist))
1524 put_ol_stateid_locked(stp, &reaplist);
1525 spin_unlock(&stp->st_stid.sc_client->cl_lock);
1526 free_ol_stateid_reaplist(&reaplist);
1529 static void unhash_openowner_locked(struct nfs4_openowner *oo)
1531 struct nfs4_client *clp = oo->oo_owner.so_client;
1533 lockdep_assert_held(&clp->cl_lock);
1535 list_del_init(&oo->oo_owner.so_strhash);
1536 list_del_init(&oo->oo_perclient);
1539 static void release_last_closed_stateid(struct nfs4_openowner *oo)
1541 struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1543 struct nfs4_ol_stateid *s;
1545 spin_lock(&nn->client_lock);
1546 s = oo->oo_last_closed_stid;
1548 list_del_init(&oo->oo_close_lru);
1549 oo->oo_last_closed_stid = NULL;
1551 spin_unlock(&nn->client_lock);
1553 nfs4_put_stid(&s->st_stid);
1556 static void release_openowner(struct nfs4_openowner *oo)
1558 struct nfs4_ol_stateid *stp;
1559 struct nfs4_client *clp = oo->oo_owner.so_client;
1560 struct list_head reaplist;
1562 INIT_LIST_HEAD(&reaplist);
1564 spin_lock(&clp->cl_lock);
1565 unhash_openowner_locked(oo);
1566 while (!list_empty(&oo->oo_owner.so_stateids)) {
1567 stp = list_first_entry(&oo->oo_owner.so_stateids,
1568 struct nfs4_ol_stateid, st_perstateowner);
1569 if (unhash_open_stateid(stp, &reaplist))
1570 put_ol_stateid_locked(stp, &reaplist);
1572 spin_unlock(&clp->cl_lock);
1573 free_ol_stateid_reaplist(&reaplist);
1574 release_last_closed_stateid(oo);
1575 nfs4_put_stateowner(&oo->oo_owner);
1579 hash_sessionid(struct nfs4_sessionid *sessionid)
1581 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1583 return sid->sequence % SESSION_HASH_SIZE;
1586 #ifdef CONFIG_SUNRPC_DEBUG
1588 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1590 u32 *ptr = (u32 *)(&sessionid->data[0]);
1591 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1595 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1601 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1602 * won't be used for replay.
1604 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1606 struct nfs4_stateowner *so = cstate->replay_owner;
1608 if (nfserr == nfserr_replay_me)
1611 if (!seqid_mutating_err(ntohl(nfserr))) {
1612 nfsd4_cstate_clear_replay(cstate);
1617 if (so->so_is_open_owner)
1618 release_last_closed_stateid(openowner(so));
1624 gen_sessionid(struct nfsd4_session *ses)
1626 struct nfs4_client *clp = ses->se_client;
1627 struct nfsd4_sessionid *sid;
1629 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1630 sid->clientid = clp->cl_clientid;
1631 sid->sequence = current_sessionid++;
1636 * The protocol defines ca_maxresponssize_cached to include the size of
1637 * the rpc header, but all we need to cache is the data starting after
1638 * the end of the initial SEQUENCE operation--the rest we regenerate
1639 * each time. Therefore we can advertise a ca_maxresponssize_cached
1640 * value that is the number of bytes in our cache plus a few additional
1641 * bytes. In order to stay on the safe side, and not promise more than
1642 * we can cache, those additional bytes must be the minimum possible: 24
1643 * bytes of rpc header (xid through accept state, with AUTH_NULL
1644 * verifier), 12 for the compound header (with zero-length tag), and 44
1645 * for the SEQUENCE op response:
1647 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1650 free_session_slots(struct nfsd4_session *ses)
1654 for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
1655 free_svc_cred(&ses->se_slots[i]->sl_cred);
1656 kfree(ses->se_slots[i]);
1661 * We don't actually need to cache the rpc and session headers, so we
1662 * can allocate a little less for each slot:
1664 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1668 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1671 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1672 return size + sizeof(struct nfsd4_slot);
1676 * XXX: If we run out of reserved DRC memory we could (up to a point)
1677 * re-negotiate active sessions and reduce their slot usage to make
1678 * room for new connections. For now we just fail the create session.
1680 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1682 u32 slotsize = slot_bytes(ca);
1683 u32 num = ca->maxreqs;
1684 unsigned long avail, total_avail;
1685 unsigned int scale_factor;
1687 spin_lock(&nfsd_drc_lock);
1688 if (nfsd_drc_max_mem > nfsd_drc_mem_used)
1689 total_avail = nfsd_drc_max_mem - nfsd_drc_mem_used;
1691 /* We have handed out more space than we chose in
1692 * set_max_drc() to allow. That isn't really a
1693 * problem as long as that doesn't make us think we
1694 * have lots more due to integer overflow.
1697 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION, total_avail);
1699 * Never use more than a fraction of the remaining memory,
1700 * unless it's the only way to give this client a slot.
1701 * The chosen fraction is either 1/8 or 1/number of threads,
1702 * whichever is smaller. This ensures there are adequate
1703 * slots to support multiple clients per thread.
1704 * Give the client one slot even if that would require
1705 * over-allocation--it is better than failure.
1707 scale_factor = max_t(unsigned int, 8, nn->nfsd_serv->sv_nrthreads);
1709 avail = clamp_t(unsigned long, avail, slotsize,
1710 total_avail/scale_factor);
1711 num = min_t(int, num, avail / slotsize);
1712 num = max_t(int, num, 1);
1713 nfsd_drc_mem_used += num * slotsize;
1714 spin_unlock(&nfsd_drc_lock);
1719 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1721 int slotsize = slot_bytes(ca);
1723 spin_lock(&nfsd_drc_lock);
1724 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1725 spin_unlock(&nfsd_drc_lock);
1728 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1729 struct nfsd4_channel_attrs *battrs)
1731 int numslots = fattrs->maxreqs;
1732 int slotsize = slot_bytes(fattrs);
1733 struct nfsd4_session *new;
1736 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1737 + sizeof(struct nfsd4_session) > PAGE_SIZE);
1738 mem = numslots * sizeof(struct nfsd4_slot *);
1740 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1743 /* allocate each struct nfsd4_slot and data cache in one piece */
1744 for (i = 0; i < numslots; i++) {
1745 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1746 if (!new->se_slots[i])
1750 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1751 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1756 kfree(new->se_slots[i]);
1761 static void free_conn(struct nfsd4_conn *c)
1763 svc_xprt_put(c->cn_xprt);
1767 static void nfsd4_conn_lost(struct svc_xpt_user *u)
1769 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1770 struct nfs4_client *clp = c->cn_session->se_client;
1772 trace_nfsd_cb_lost(clp);
1774 spin_lock(&clp->cl_lock);
1775 if (!list_empty(&c->cn_persession)) {
1776 list_del(&c->cn_persession);
1779 nfsd4_probe_callback(clp);
1780 spin_unlock(&clp->cl_lock);
1783 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1785 struct nfsd4_conn *conn;
1787 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1790 svc_xprt_get(rqstp->rq_xprt);
1791 conn->cn_xprt = rqstp->rq_xprt;
1792 conn->cn_flags = flags;
1793 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1797 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1799 conn->cn_session = ses;
1800 list_add(&conn->cn_persession, &ses->se_conns);
1803 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1805 struct nfs4_client *clp = ses->se_client;
1807 spin_lock(&clp->cl_lock);
1808 __nfsd4_hash_conn(conn, ses);
1809 spin_unlock(&clp->cl_lock);
1812 static int nfsd4_register_conn(struct nfsd4_conn *conn)
1814 conn->cn_xpt_user.callback = nfsd4_conn_lost;
1815 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1818 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1822 nfsd4_hash_conn(conn, ses);
1823 ret = nfsd4_register_conn(conn);
1825 /* oops; xprt is already down: */
1826 nfsd4_conn_lost(&conn->cn_xpt_user);
1827 /* We may have gained or lost a callback channel: */
1828 nfsd4_probe_callback_sync(ses->se_client);
1831 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1833 u32 dir = NFS4_CDFC4_FORE;
1835 if (cses->flags & SESSION4_BACK_CHAN)
1836 dir |= NFS4_CDFC4_BACK;
1837 return alloc_conn(rqstp, dir);
1840 /* must be called under client_lock */
1841 static void nfsd4_del_conns(struct nfsd4_session *s)
1843 struct nfs4_client *clp = s->se_client;
1844 struct nfsd4_conn *c;
1846 spin_lock(&clp->cl_lock);
1847 while (!list_empty(&s->se_conns)) {
1848 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1849 list_del_init(&c->cn_persession);
1850 spin_unlock(&clp->cl_lock);
1852 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1855 spin_lock(&clp->cl_lock);
1857 spin_unlock(&clp->cl_lock);
1860 static void __free_session(struct nfsd4_session *ses)
1862 free_session_slots(ses);
1866 static void free_session(struct nfsd4_session *ses)
1868 nfsd4_del_conns(ses);
1869 nfsd4_put_drc_mem(&ses->se_fchannel);
1870 __free_session(ses);
1873 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1876 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1878 new->se_client = clp;
1881 INIT_LIST_HEAD(&new->se_conns);
1883 new->se_cb_seq_nr = 1;
1884 new->se_flags = cses->flags;
1885 new->se_cb_prog = cses->callback_prog;
1886 new->se_cb_sec = cses->cb_sec;
1887 atomic_set(&new->se_ref, 0);
1888 idx = hash_sessionid(&new->se_sessionid);
1889 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1890 spin_lock(&clp->cl_lock);
1891 list_add(&new->se_perclnt, &clp->cl_sessions);
1892 spin_unlock(&clp->cl_lock);
1895 struct sockaddr *sa = svc_addr(rqstp);
1897 * This is a little silly; with sessions there's no real
1898 * use for the callback address. Use the peer address
1899 * as a reasonable default for now, but consider fixing
1900 * the rpc client not to require an address in the
1903 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1904 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1908 /* caller must hold client_lock */
1909 static struct nfsd4_session *
1910 __find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1912 struct nfsd4_session *elem;
1914 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1916 lockdep_assert_held(&nn->client_lock);
1918 dump_sessionid(__func__, sessionid);
1919 idx = hash_sessionid(sessionid);
1920 /* Search in the appropriate list */
1921 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1922 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1923 NFS4_MAX_SESSIONID_LEN)) {
1928 dprintk("%s: session not found\n", __func__);
1932 static struct nfsd4_session *
1933 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1936 struct nfsd4_session *session;
1937 __be32 status = nfserr_badsession;
1939 session = __find_in_sessionid_hashtbl(sessionid, net);
1942 status = nfsd4_get_session_locked(session);
1950 /* caller must hold client_lock */
1952 unhash_session(struct nfsd4_session *ses)
1954 struct nfs4_client *clp = ses->se_client;
1955 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1957 lockdep_assert_held(&nn->client_lock);
1959 list_del(&ses->se_hash);
1960 spin_lock(&ses->se_client->cl_lock);
1961 list_del(&ses->se_perclnt);
1962 spin_unlock(&ses->se_client->cl_lock);
1965 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1967 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1970 * We're assuming the clid was not given out from a boot
1971 * precisely 2^32 (about 136 years) before this one. That seems
1972 * a safe assumption:
1974 if (clid->cl_boot == (u32)nn->boot_time)
1976 trace_nfsd_clid_stale(clid);
1981 * XXX Should we use a slab cache ?
1982 * This type of memory management is somewhat inefficient, but we use it
1983 * anyway since SETCLIENTID is not a common operation.
1985 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1987 struct nfs4_client *clp;
1990 clp = kmem_cache_zalloc(client_slab, GFP_KERNEL);
1993 xdr_netobj_dup(&clp->cl_name, &name, GFP_KERNEL);
1994 if (clp->cl_name.data == NULL)
1996 clp->cl_ownerstr_hashtbl = kmalloc_array(OWNER_HASH_SIZE,
1997 sizeof(struct list_head),
1999 if (!clp->cl_ownerstr_hashtbl)
2000 goto err_no_hashtbl;
2001 for (i = 0; i < OWNER_HASH_SIZE; i++)
2002 INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
2003 INIT_LIST_HEAD(&clp->cl_sessions);
2004 idr_init(&clp->cl_stateids);
2005 atomic_set(&clp->cl_rpc_users, 0);
2006 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
2007 INIT_LIST_HEAD(&clp->cl_idhash);
2008 INIT_LIST_HEAD(&clp->cl_openowners);
2009 INIT_LIST_HEAD(&clp->cl_delegations);
2010 INIT_LIST_HEAD(&clp->cl_lru);
2011 INIT_LIST_HEAD(&clp->cl_revoked);
2012 #ifdef CONFIG_NFSD_PNFS
2013 INIT_LIST_HEAD(&clp->cl_lo_states);
2015 INIT_LIST_HEAD(&clp->async_copies);
2016 spin_lock_init(&clp->async_lock);
2017 spin_lock_init(&clp->cl_lock);
2018 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
2021 kfree(clp->cl_name.data);
2023 kmem_cache_free(client_slab, clp);
2027 static void __free_client(struct kref *k)
2029 struct nfsdfs_client *c = container_of(k, struct nfsdfs_client, cl_ref);
2030 struct nfs4_client *clp = container_of(c, struct nfs4_client, cl_nfsdfs);
2032 free_svc_cred(&clp->cl_cred);
2033 kfree(clp->cl_ownerstr_hashtbl);
2034 kfree(clp->cl_name.data);
2035 kfree(clp->cl_nii_domain.data);
2036 kfree(clp->cl_nii_name.data);
2037 idr_destroy(&clp->cl_stateids);
2038 kmem_cache_free(client_slab, clp);
2041 static void drop_client(struct nfs4_client *clp)
2043 kref_put(&clp->cl_nfsdfs.cl_ref, __free_client);
2047 free_client(struct nfs4_client *clp)
2049 while (!list_empty(&clp->cl_sessions)) {
2050 struct nfsd4_session *ses;
2051 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
2053 list_del(&ses->se_perclnt);
2054 WARN_ON_ONCE(atomic_read(&ses->se_ref));
2057 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
2058 if (clp->cl_nfsd_dentry) {
2059 nfsd_client_rmdir(clp->cl_nfsd_dentry);
2060 clp->cl_nfsd_dentry = NULL;
2061 wake_up_all(&expiry_wq);
2066 /* must be called under the client_lock */
2068 unhash_client_locked(struct nfs4_client *clp)
2070 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2071 struct nfsd4_session *ses;
2073 lockdep_assert_held(&nn->client_lock);
2075 /* Mark the client as expired! */
2077 /* Make it invisible */
2078 if (!list_empty(&clp->cl_idhash)) {
2079 list_del_init(&clp->cl_idhash);
2080 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2081 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
2083 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2085 list_del_init(&clp->cl_lru);
2086 spin_lock(&clp->cl_lock);
2087 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
2088 list_del_init(&ses->se_hash);
2089 spin_unlock(&clp->cl_lock);
2093 unhash_client(struct nfs4_client *clp)
2095 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2097 spin_lock(&nn->client_lock);
2098 unhash_client_locked(clp);
2099 spin_unlock(&nn->client_lock);
2102 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
2104 if (atomic_read(&clp->cl_rpc_users))
2105 return nfserr_jukebox;
2106 unhash_client_locked(clp);
2111 __destroy_client(struct nfs4_client *clp)
2114 struct nfs4_openowner *oo;
2115 struct nfs4_delegation *dp;
2116 struct list_head reaplist;
2118 INIT_LIST_HEAD(&reaplist);
2119 spin_lock(&state_lock);
2120 while (!list_empty(&clp->cl_delegations)) {
2121 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
2122 WARN_ON(!unhash_delegation_locked(dp));
2123 list_add(&dp->dl_recall_lru, &reaplist);
2125 spin_unlock(&state_lock);
2126 while (!list_empty(&reaplist)) {
2127 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
2128 list_del_init(&dp->dl_recall_lru);
2129 destroy_unhashed_deleg(dp);
2131 while (!list_empty(&clp->cl_revoked)) {
2132 dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
2133 list_del_init(&dp->dl_recall_lru);
2134 nfs4_put_stid(&dp->dl_stid);
2136 while (!list_empty(&clp->cl_openowners)) {
2137 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
2138 nfs4_get_stateowner(&oo->oo_owner);
2139 release_openowner(oo);
2141 for (i = 0; i < OWNER_HASH_SIZE; i++) {
2142 struct nfs4_stateowner *so, *tmp;
2144 list_for_each_entry_safe(so, tmp, &clp->cl_ownerstr_hashtbl[i],
2146 /* Should be no openowners at this point */
2147 WARN_ON_ONCE(so->so_is_open_owner);
2148 remove_blocked_locks(lockowner(so));
2151 nfsd4_return_all_client_layouts(clp);
2152 nfsd4_shutdown_copy(clp);
2153 nfsd4_shutdown_callback(clp);
2154 if (clp->cl_cb_conn.cb_xprt)
2155 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
2157 wake_up_all(&expiry_wq);
2161 destroy_client(struct nfs4_client *clp)
2164 __destroy_client(clp);
2167 static void inc_reclaim_complete(struct nfs4_client *clp)
2169 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2171 if (!nn->track_reclaim_completes)
2173 if (!nfsd4_find_reclaim_client(clp->cl_name, nn))
2175 if (atomic_inc_return(&nn->nr_reclaim_complete) ==
2176 nn->reclaim_str_hashtbl_size) {
2177 printk(KERN_INFO "NFSD: all clients done reclaiming, ending NFSv4 grace period (net %x)\n",
2179 nfsd4_end_grace(nn);
2183 static void expire_client(struct nfs4_client *clp)
2186 nfsd4_client_record_remove(clp);
2187 __destroy_client(clp);
2190 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
2192 memcpy(target->cl_verifier.data, source->data,
2193 sizeof(target->cl_verifier.data));
2196 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
2198 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
2199 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
2202 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
2204 target->cr_principal = kstrdup(source->cr_principal, GFP_KERNEL);
2205 target->cr_raw_principal = kstrdup(source->cr_raw_principal,
2207 target->cr_targ_princ = kstrdup(source->cr_targ_princ, GFP_KERNEL);
2208 if ((source->cr_principal && !target->cr_principal) ||
2209 (source->cr_raw_principal && !target->cr_raw_principal) ||
2210 (source->cr_targ_princ && !target->cr_targ_princ))
2213 target->cr_flavor = source->cr_flavor;
2214 target->cr_uid = source->cr_uid;
2215 target->cr_gid = source->cr_gid;
2216 target->cr_group_info = source->cr_group_info;
2217 get_group_info(target->cr_group_info);
2218 target->cr_gss_mech = source->cr_gss_mech;
2219 if (source->cr_gss_mech)
2220 gss_mech_get(source->cr_gss_mech);
2225 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
2227 if (o1->len < o2->len)
2229 if (o1->len > o2->len)
2231 return memcmp(o1->data, o2->data, o1->len);
2235 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
2237 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
2241 same_clid(clientid_t *cl1, clientid_t *cl2)
2243 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
2246 static bool groups_equal(struct group_info *g1, struct group_info *g2)
2250 if (g1->ngroups != g2->ngroups)
2252 for (i=0; i<g1->ngroups; i++)
2253 if (!gid_eq(g1->gid[i], g2->gid[i]))
2259 * RFC 3530 language requires clid_inuse be returned when the
2260 * "principal" associated with a requests differs from that previously
2261 * used. We use uid, gid's, and gss principal string as our best
2262 * approximation. We also don't want to allow non-gss use of a client
2263 * established using gss: in theory cr_principal should catch that
2264 * change, but in practice cr_principal can be null even in the gss case
2265 * since gssd doesn't always pass down a principal string.
2267 static bool is_gss_cred(struct svc_cred *cr)
2269 /* Is cr_flavor one of the gss "pseudoflavors"?: */
2270 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
2275 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
2277 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
2278 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
2279 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
2280 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
2282 /* XXX: check that cr_targ_princ fields match ? */
2283 if (cr1->cr_principal == cr2->cr_principal)
2285 if (!cr1->cr_principal || !cr2->cr_principal)
2287 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
2290 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
2292 struct svc_cred *cr = &rqstp->rq_cred;
2295 if (!cr->cr_gss_mech)
2297 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
2298 return service == RPC_GSS_SVC_INTEGRITY ||
2299 service == RPC_GSS_SVC_PRIVACY;
2302 bool nfsd4_mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
2304 struct svc_cred *cr = &rqstp->rq_cred;
2306 if (!cl->cl_mach_cred)
2308 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
2310 if (!svc_rqst_integrity_protected(rqstp))
2312 if (cl->cl_cred.cr_raw_principal)
2313 return 0 == strcmp(cl->cl_cred.cr_raw_principal,
2314 cr->cr_raw_principal);
2315 if (!cr->cr_principal)
2317 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
2320 static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
2325 * This is opaque to client, so no need to byte-swap. Use
2326 * __force to keep sparse happy
2328 verf[0] = (__force __be32)(u32)ktime_get_real_seconds();
2329 verf[1] = (__force __be32)nn->clverifier_counter++;
2330 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
2333 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
2335 clp->cl_clientid.cl_boot = (u32)nn->boot_time;
2336 clp->cl_clientid.cl_id = nn->clientid_counter++;
2337 gen_confirm(clp, nn);
2340 static struct nfs4_stid *
2341 find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
2343 struct nfs4_stid *ret;
2345 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
2346 if (!ret || !ret->sc_type)
2351 static struct nfs4_stid *
2352 find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
2354 struct nfs4_stid *s;
2356 spin_lock(&cl->cl_lock);
2357 s = find_stateid_locked(cl, t);
2359 if (typemask & s->sc_type)
2360 refcount_inc(&s->sc_count);
2364 spin_unlock(&cl->cl_lock);
2368 static struct nfs4_client *get_nfsdfs_clp(struct inode *inode)
2370 struct nfsdfs_client *nc;
2371 nc = get_nfsdfs_client(inode);
2374 return container_of(nc, struct nfs4_client, cl_nfsdfs);
2377 static void seq_quote_mem(struct seq_file *m, char *data, int len)
2379 seq_printf(m, "\"");
2380 seq_escape_mem(m, data, len, ESCAPE_HEX | ESCAPE_NAP | ESCAPE_APPEND, "\"\\");
2381 seq_printf(m, "\"");
2384 static const char *cb_state2str(int state)
2389 case NFSD4_CB_UNKNOWN:
2393 case NFSD4_CB_FAULT:
2399 static int client_info_show(struct seq_file *m, void *v)
2401 struct inode *inode = m->private;
2402 struct nfs4_client *clp;
2405 clp = get_nfsdfs_clp(inode);
2408 memcpy(&clid, &clp->cl_clientid, sizeof(clid));
2409 seq_printf(m, "clientid: 0x%llx\n", clid);
2410 seq_printf(m, "address: \"%pISpc\"\n", (struct sockaddr *)&clp->cl_addr);
2411 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2412 seq_puts(m, "status: confirmed\n");
2414 seq_puts(m, "status: unconfirmed\n");
2415 seq_printf(m, "name: ");
2416 seq_quote_mem(m, clp->cl_name.data, clp->cl_name.len);
2417 seq_printf(m, "\nminor version: %d\n", clp->cl_minorversion);
2418 if (clp->cl_nii_domain.data) {
2419 seq_printf(m, "Implementation domain: ");
2420 seq_quote_mem(m, clp->cl_nii_domain.data,
2421 clp->cl_nii_domain.len);
2422 seq_printf(m, "\nImplementation name: ");
2423 seq_quote_mem(m, clp->cl_nii_name.data, clp->cl_nii_name.len);
2424 seq_printf(m, "\nImplementation time: [%lld, %ld]\n",
2425 clp->cl_nii_time.tv_sec, clp->cl_nii_time.tv_nsec);
2427 seq_printf(m, "callback state: %s\n", cb_state2str(clp->cl_cb_state));
2428 seq_printf(m, "callback address: %pISpc\n", &clp->cl_cb_conn.cb_addr);
2434 static int client_info_open(struct inode *inode, struct file *file)
2436 return single_open(file, client_info_show, inode);
2439 static const struct file_operations client_info_fops = {
2440 .open = client_info_open,
2442 .llseek = seq_lseek,
2443 .release = single_release,
2446 static void *states_start(struct seq_file *s, loff_t *pos)
2447 __acquires(&clp->cl_lock)
2449 struct nfs4_client *clp = s->private;
2450 unsigned long id = *pos;
2453 spin_lock(&clp->cl_lock);
2454 ret = idr_get_next_ul(&clp->cl_stateids, &id);
2459 static void *states_next(struct seq_file *s, void *v, loff_t *pos)
2461 struct nfs4_client *clp = s->private;
2462 unsigned long id = *pos;
2467 ret = idr_get_next_ul(&clp->cl_stateids, &id);
2472 static void states_stop(struct seq_file *s, void *v)
2473 __releases(&clp->cl_lock)
2475 struct nfs4_client *clp = s->private;
2477 spin_unlock(&clp->cl_lock);
2480 static void nfs4_show_fname(struct seq_file *s, struct nfsd_file *f)
2482 seq_printf(s, "filename: \"%pD2\"", f->nf_file);
2485 static void nfs4_show_superblock(struct seq_file *s, struct nfsd_file *f)
2487 struct inode *inode = f->nf_inode;
2489 seq_printf(s, "superblock: \"%02x:%02x:%ld\"",
2490 MAJOR(inode->i_sb->s_dev),
2491 MINOR(inode->i_sb->s_dev),
2495 static void nfs4_show_owner(struct seq_file *s, struct nfs4_stateowner *oo)
2497 seq_printf(s, "owner: ");
2498 seq_quote_mem(s, oo->so_owner.data, oo->so_owner.len);
2501 static void nfs4_show_stateid(struct seq_file *s, stateid_t *stid)
2503 seq_printf(s, "0x%.8x", stid->si_generation);
2504 seq_printf(s, "%12phN", &stid->si_opaque);
2507 static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
2509 struct nfs4_ol_stateid *ols;
2510 struct nfs4_file *nf;
2511 struct nfsd_file *file;
2512 struct nfs4_stateowner *oo;
2513 unsigned int access, deny;
2515 if (st->sc_type != NFS4_OPEN_STID && st->sc_type != NFS4_LOCK_STID)
2516 return 0; /* XXX: or SEQ_SKIP? */
2517 ols = openlockstateid(st);
2518 oo = ols->st_stateowner;
2520 file = find_any_file(nf);
2524 seq_printf(s, "- ");
2525 nfs4_show_stateid(s, &st->sc_stateid);
2526 seq_printf(s, ": { type: open, ");
2528 access = bmap_to_share_mode(ols->st_access_bmap);
2529 deny = bmap_to_share_mode(ols->st_deny_bmap);
2531 seq_printf(s, "access: %s%s, ",
2532 access & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2533 access & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2534 seq_printf(s, "deny: %s%s, ",
2535 deny & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2536 deny & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2538 nfs4_show_superblock(s, file);
2539 seq_printf(s, ", ");
2540 nfs4_show_fname(s, file);
2541 seq_printf(s, ", ");
2542 nfs4_show_owner(s, oo);
2543 seq_printf(s, " }\n");
2544 nfsd_file_put(file);
2549 static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
2551 struct nfs4_ol_stateid *ols;
2552 struct nfs4_file *nf;
2553 struct nfsd_file *file;
2554 struct nfs4_stateowner *oo;
2556 ols = openlockstateid(st);
2557 oo = ols->st_stateowner;
2559 file = find_any_file(nf);
2563 seq_printf(s, "- ");
2564 nfs4_show_stateid(s, &st->sc_stateid);
2565 seq_printf(s, ": { type: lock, ");
2568 * Note: a lock stateid isn't really the same thing as a lock,
2569 * it's the locking state held by one owner on a file, and there
2570 * may be multiple (or no) lock ranges associated with it.
2571 * (Same for the matter is true of open stateids.)
2574 nfs4_show_superblock(s, file);
2575 /* XXX: open stateid? */
2576 seq_printf(s, ", ");
2577 nfs4_show_fname(s, file);
2578 seq_printf(s, ", ");
2579 nfs4_show_owner(s, oo);
2580 seq_printf(s, " }\n");
2581 nfsd_file_put(file);
2586 static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
2588 struct nfs4_delegation *ds;
2589 struct nfs4_file *nf;
2590 struct nfsd_file *file;
2592 ds = delegstateid(st);
2594 file = find_deleg_file(nf);
2598 seq_printf(s, "- ");
2599 nfs4_show_stateid(s, &st->sc_stateid);
2600 seq_printf(s, ": { type: deleg, ");
2602 /* Kinda dead code as long as we only support read delegs: */
2603 seq_printf(s, "access: %s, ",
2604 ds->dl_type == NFS4_OPEN_DELEGATE_READ ? "r" : "w");
2606 /* XXX: lease time, whether it's being recalled. */
2608 nfs4_show_superblock(s, file);
2609 seq_printf(s, ", ");
2610 nfs4_show_fname(s, file);
2611 seq_printf(s, " }\n");
2612 nfsd_file_put(file);
2617 static int nfs4_show_layout(struct seq_file *s, struct nfs4_stid *st)
2619 struct nfs4_layout_stateid *ls;
2620 struct nfsd_file *file;
2622 ls = container_of(st, struct nfs4_layout_stateid, ls_stid);
2625 seq_printf(s, "- ");
2626 nfs4_show_stateid(s, &st->sc_stateid);
2627 seq_printf(s, ": { type: layout, ");
2629 /* XXX: What else would be useful? */
2631 nfs4_show_superblock(s, file);
2632 seq_printf(s, ", ");
2633 nfs4_show_fname(s, file);
2634 seq_printf(s, " }\n");
2639 static int states_show(struct seq_file *s, void *v)
2641 struct nfs4_stid *st = v;
2643 switch (st->sc_type) {
2644 case NFS4_OPEN_STID:
2645 return nfs4_show_open(s, st);
2646 case NFS4_LOCK_STID:
2647 return nfs4_show_lock(s, st);
2648 case NFS4_DELEG_STID:
2649 return nfs4_show_deleg(s, st);
2650 case NFS4_LAYOUT_STID:
2651 return nfs4_show_layout(s, st);
2653 return 0; /* XXX: or SEQ_SKIP? */
2655 /* XXX: copy stateids? */
2658 static struct seq_operations states_seq_ops = {
2659 .start = states_start,
2660 .next = states_next,
2661 .stop = states_stop,
2665 static int client_states_open(struct inode *inode, struct file *file)
2668 struct nfs4_client *clp;
2671 clp = get_nfsdfs_clp(inode);
2675 ret = seq_open(file, &states_seq_ops);
2678 s = file->private_data;
2683 static int client_opens_release(struct inode *inode, struct file *file)
2685 struct seq_file *m = file->private_data;
2686 struct nfs4_client *clp = m->private;
2688 /* XXX: alternatively, we could get/drop in seq start/stop */
2693 static const struct file_operations client_states_fops = {
2694 .open = client_states_open,
2696 .llseek = seq_lseek,
2697 .release = client_opens_release,
2701 * Normally we refuse to destroy clients that are in use, but here the
2702 * administrator is telling us to just do it. We also want to wait
2703 * so the caller has a guarantee that the client's locks are gone by
2704 * the time the write returns:
2706 static void force_expire_client(struct nfs4_client *clp)
2708 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2709 bool already_expired;
2711 trace_nfsd_clid_admin_expired(&clp->cl_clientid);
2713 spin_lock(&nn->client_lock);
2715 spin_unlock(&nn->client_lock);
2717 wait_event(expiry_wq, atomic_read(&clp->cl_rpc_users) == 0);
2718 spin_lock(&nn->client_lock);
2719 already_expired = list_empty(&clp->cl_lru);
2720 if (!already_expired)
2721 unhash_client_locked(clp);
2722 spin_unlock(&nn->client_lock);
2724 if (!already_expired)
2727 wait_event(expiry_wq, clp->cl_nfsd_dentry == NULL);
2730 static ssize_t client_ctl_write(struct file *file, const char __user *buf,
2731 size_t size, loff_t *pos)
2734 struct nfs4_client *clp;
2736 data = simple_transaction_get(file, buf, size);
2738 return PTR_ERR(data);
2739 if (size != 7 || 0 != memcmp(data, "expire\n", 7))
2741 clp = get_nfsdfs_clp(file_inode(file));
2744 force_expire_client(clp);
2749 static const struct file_operations client_ctl_fops = {
2750 .write = client_ctl_write,
2751 .release = simple_transaction_release,
2754 static const struct tree_descr client_files[] = {
2755 [0] = {"info", &client_info_fops, S_IRUSR},
2756 [1] = {"states", &client_states_fops, S_IRUSR},
2757 [2] = {"ctl", &client_ctl_fops, S_IWUSR},
2761 static struct nfs4_client *create_client(struct xdr_netobj name,
2762 struct svc_rqst *rqstp, nfs4_verifier *verf)
2764 struct nfs4_client *clp;
2765 struct sockaddr *sa = svc_addr(rqstp);
2767 struct net *net = SVC_NET(rqstp);
2768 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2769 struct dentry *dentries[ARRAY_SIZE(client_files)];
2771 clp = alloc_client(name);
2775 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
2781 kref_init(&clp->cl_nfsdfs.cl_ref);
2782 nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
2783 clp->cl_time = ktime_get_boottime_seconds();
2784 clear_bit(0, &clp->cl_cb_slot_busy);
2785 copy_verf(clp, verf);
2786 memcpy(&clp->cl_addr, sa, sizeof(struct sockaddr_storage));
2787 clp->cl_cb_session = NULL;
2789 clp->cl_nfsd_dentry = nfsd_client_mkdir(
2790 nn, &clp->cl_nfsdfs,
2791 clp->cl_clientid.cl_id - nn->clientid_base,
2792 client_files, dentries);
2793 clp->cl_nfsd_info_dentry = dentries[0];
2794 if (!clp->cl_nfsd_dentry) {
2802 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
2804 struct rb_node **new = &(root->rb_node), *parent = NULL;
2805 struct nfs4_client *clp;
2808 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
2811 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
2812 new = &((*new)->rb_left);
2814 new = &((*new)->rb_right);
2817 rb_link_node(&new_clp->cl_namenode, parent, new);
2818 rb_insert_color(&new_clp->cl_namenode, root);
2821 static struct nfs4_client *
2822 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
2825 struct rb_node *node = root->rb_node;
2826 struct nfs4_client *clp;
2829 clp = rb_entry(node, struct nfs4_client, cl_namenode);
2830 cmp = compare_blob(&clp->cl_name, name);
2832 node = node->rb_left;
2834 node = node->rb_right;
2842 add_to_unconfirmed(struct nfs4_client *clp)
2844 unsigned int idhashval;
2845 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2847 lockdep_assert_held(&nn->client_lock);
2849 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2850 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
2851 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2852 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
2853 renew_client_locked(clp);
2857 move_to_confirmed(struct nfs4_client *clp)
2859 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2860 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2862 lockdep_assert_held(&nn->client_lock);
2864 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
2865 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2866 add_clp_to_name_tree(clp, &nn->conf_name_tree);
2867 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2868 trace_nfsd_clid_confirmed(&clp->cl_clientid);
2869 renew_client_locked(clp);
2872 static struct nfs4_client *
2873 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
2875 struct nfs4_client *clp;
2876 unsigned int idhashval = clientid_hashval(clid->cl_id);
2878 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
2879 if (same_clid(&clp->cl_clientid, clid)) {
2880 if ((bool)clp->cl_minorversion != sessions)
2882 renew_client_locked(clp);
2889 static struct nfs4_client *
2890 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2892 struct list_head *tbl = nn->conf_id_hashtbl;
2894 lockdep_assert_held(&nn->client_lock);
2895 return find_client_in_id_table(tbl, clid, sessions);
2898 static struct nfs4_client *
2899 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2901 struct list_head *tbl = nn->unconf_id_hashtbl;
2903 lockdep_assert_held(&nn->client_lock);
2904 return find_client_in_id_table(tbl, clid, sessions);
2907 static bool clp_used_exchangeid(struct nfs4_client *clp)
2909 return clp->cl_exchange_flags != 0;
2912 static struct nfs4_client *
2913 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2915 lockdep_assert_held(&nn->client_lock);
2916 return find_clp_in_name_tree(name, &nn->conf_name_tree);
2919 static struct nfs4_client *
2920 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2922 lockdep_assert_held(&nn->client_lock);
2923 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
2927 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
2929 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
2930 struct sockaddr *sa = svc_addr(rqstp);
2931 u32 scopeid = rpc_get_scope_id(sa);
2932 unsigned short expected_family;
2934 /* Currently, we only support tcp and tcp6 for the callback channel */
2935 if (se->se_callback_netid_len == 3 &&
2936 !memcmp(se->se_callback_netid_val, "tcp", 3))
2937 expected_family = AF_INET;
2938 else if (se->se_callback_netid_len == 4 &&
2939 !memcmp(se->se_callback_netid_val, "tcp6", 4))
2940 expected_family = AF_INET6;
2944 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
2945 se->se_callback_addr_len,
2946 (struct sockaddr *)&conn->cb_addr,
2947 sizeof(conn->cb_addr));
2949 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
2952 if (conn->cb_addr.ss_family == AF_INET6)
2953 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
2955 conn->cb_prog = se->se_callback_prog;
2956 conn->cb_ident = se->se_callback_ident;
2957 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
2958 trace_nfsd_cb_args(clp, conn);
2961 conn->cb_addr.ss_family = AF_UNSPEC;
2962 conn->cb_addrlen = 0;
2963 trace_nfsd_cb_nodelegs(clp);
2968 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2971 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
2973 struct xdr_buf *buf = resp->xdr->buf;
2974 struct nfsd4_slot *slot = resp->cstate.slot;
2977 dprintk("--> %s slot %p\n", __func__, slot);
2979 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
2980 slot->sl_opcnt = resp->opcnt;
2981 slot->sl_status = resp->cstate.status;
2982 free_svc_cred(&slot->sl_cred);
2983 copy_cred(&slot->sl_cred, &resp->rqstp->rq_cred);
2985 if (!nfsd4_cache_this(resp)) {
2986 slot->sl_flags &= ~NFSD4_SLOT_CACHED;
2989 slot->sl_flags |= NFSD4_SLOT_CACHED;
2991 base = resp->cstate.data_offset;
2992 slot->sl_datalen = buf->len - base;
2993 if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
2994 WARN(1, "%s: sessions DRC could not cache compound\n",
3000 * Encode the replay sequence operation from the slot values.
3001 * If cachethis is FALSE encode the uncached rep error on the next
3002 * operation which sets resp->p and increments resp->opcnt for
3003 * nfs4svc_encode_compoundres.
3007 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
3008 struct nfsd4_compoundres *resp)
3010 struct nfsd4_op *op;
3011 struct nfsd4_slot *slot = resp->cstate.slot;
3013 /* Encode the replayed sequence operation */
3014 op = &args->ops[resp->opcnt - 1];
3015 nfsd4_encode_operation(resp, op);
3017 if (slot->sl_flags & NFSD4_SLOT_CACHED)
3019 if (args->opcnt == 1) {
3021 * The original operation wasn't a solo sequence--we
3022 * always cache those--so this retry must not match the
3025 op->status = nfserr_seq_false_retry;
3027 op = &args->ops[resp->opcnt++];
3028 op->status = nfserr_retry_uncached_rep;
3029 nfsd4_encode_operation(resp, op);
3035 * The sequence operation is not cached because we can use the slot and
3039 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
3040 struct nfsd4_sequence *seq)
3042 struct nfsd4_slot *slot = resp->cstate.slot;
3043 struct xdr_stream *xdr = resp->xdr;
3047 dprintk("--> %s slot %p\n", __func__, slot);
3049 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
3053 p = xdr_reserve_space(xdr, slot->sl_datalen);
3056 return nfserr_serverfault;
3058 xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
3059 xdr_commit_encode(xdr);
3061 resp->opcnt = slot->sl_opcnt;
3062 return slot->sl_status;
3066 * Set the exchange_id flags returned by the server.
3069 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
3071 #ifdef CONFIG_NFSD_PNFS
3072 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
3074 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
3077 /* Referrals are supported, Migration is not. */
3078 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
3080 /* set the wire flags to return to client. */
3081 clid->flags = new->cl_exchange_flags;
3084 static bool client_has_openowners(struct nfs4_client *clp)
3086 struct nfs4_openowner *oo;
3088 list_for_each_entry(oo, &clp->cl_openowners, oo_perclient) {
3089 if (!list_empty(&oo->oo_owner.so_stateids))
3095 static bool client_has_state(struct nfs4_client *clp)
3097 return client_has_openowners(clp)
3098 #ifdef CONFIG_NFSD_PNFS
3099 || !list_empty(&clp->cl_lo_states)
3101 || !list_empty(&clp->cl_delegations)
3102 || !list_empty(&clp->cl_sessions)
3103 || !list_empty(&clp->async_copies);
3106 static __be32 copy_impl_id(struct nfs4_client *clp,
3107 struct nfsd4_exchange_id *exid)
3109 if (!exid->nii_domain.data)
3111 xdr_netobj_dup(&clp->cl_nii_domain, &exid->nii_domain, GFP_KERNEL);
3112 if (!clp->cl_nii_domain.data)
3113 return nfserr_jukebox;
3114 xdr_netobj_dup(&clp->cl_nii_name, &exid->nii_name, GFP_KERNEL);
3115 if (!clp->cl_nii_name.data)
3116 return nfserr_jukebox;
3117 clp->cl_nii_time = exid->nii_time;
3122 nfsd4_exchange_id(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3123 union nfsd4_op_u *u)
3125 struct nfsd4_exchange_id *exid = &u->exchange_id;
3126 struct nfs4_client *conf, *new;
3127 struct nfs4_client *unconf = NULL;
3129 char addr_str[INET6_ADDRSTRLEN];
3130 nfs4_verifier verf = exid->verifier;
3131 struct sockaddr *sa = svc_addr(rqstp);
3132 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
3133 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3135 rpc_ntop(sa, addr_str, sizeof(addr_str));
3136 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
3137 "ip_addr=%s flags %x, spa_how %u\n",
3138 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
3139 addr_str, exid->flags, exid->spa_how);
3141 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
3142 return nfserr_inval;
3144 new = create_client(exid->clname, rqstp, &verf);
3146 return nfserr_jukebox;
3147 status = copy_impl_id(new, exid);
3151 switch (exid->spa_how) {
3153 exid->spo_must_enforce[0] = 0;
3154 exid->spo_must_enforce[1] = (
3155 1 << (OP_BIND_CONN_TO_SESSION - 32) |
3156 1 << (OP_EXCHANGE_ID - 32) |
3157 1 << (OP_CREATE_SESSION - 32) |
3158 1 << (OP_DESTROY_SESSION - 32) |
3159 1 << (OP_DESTROY_CLIENTID - 32));
3161 exid->spo_must_allow[0] &= (1 << (OP_CLOSE) |
3162 1 << (OP_OPEN_DOWNGRADE) |
3164 1 << (OP_DELEGRETURN));
3166 exid->spo_must_allow[1] &= (
3167 1 << (OP_TEST_STATEID - 32) |
3168 1 << (OP_FREE_STATEID - 32));
3169 if (!svc_rqst_integrity_protected(rqstp)) {
3170 status = nfserr_inval;
3174 * Sometimes userspace doesn't give us a principal.
3175 * Which is a bug, really. Anyway, we can't enforce
3176 * MACH_CRED in that case, better to give up now:
3178 if (!new->cl_cred.cr_principal &&
3179 !new->cl_cred.cr_raw_principal) {
3180 status = nfserr_serverfault;
3183 new->cl_mach_cred = true;
3187 default: /* checked by xdr code */
3191 status = nfserr_encr_alg_unsupp;
3195 /* Cases below refer to rfc 5661 section 18.35.4: */
3196 spin_lock(&nn->client_lock);
3197 conf = find_confirmed_client_by_name(&exid->clname, nn);
3199 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
3200 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
3203 if (!clp_used_exchangeid(conf)) { /* buggy client */
3204 status = nfserr_inval;
3207 if (!nfsd4_mach_creds_match(conf, rqstp)) {
3208 status = nfserr_wrong_cred;
3211 if (!creds_match) { /* case 9 */
3212 status = nfserr_perm;
3215 if (!verfs_match) { /* case 8 */
3216 status = nfserr_not_same;
3220 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
3221 trace_nfsd_clid_confirmed_r(conf);
3224 if (!creds_match) { /* case 3 */
3225 if (client_has_state(conf)) {
3226 status = nfserr_clid_inuse;
3227 trace_nfsd_clid_cred_mismatch(conf, rqstp);
3232 if (verfs_match) { /* case 2 */
3233 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
3234 trace_nfsd_clid_confirmed_r(conf);
3237 /* case 5, client reboot */
3238 trace_nfsd_clid_verf_mismatch(conf, rqstp, &verf);
3243 if (update) { /* case 7 */
3244 status = nfserr_noent;
3248 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
3249 if (unconf) /* case 4, possible retry or client restart */
3250 unhash_client_locked(unconf);
3252 /* case 1, new owner ID */
3253 trace_nfsd_clid_fresh(new);
3257 status = mark_client_expired_locked(conf);
3260 trace_nfsd_clid_replaced(&conf->cl_clientid);
3262 new->cl_minorversion = cstate->minorversion;
3263 new->cl_spo_must_allow.u.words[0] = exid->spo_must_allow[0];
3264 new->cl_spo_must_allow.u.words[1] = exid->spo_must_allow[1];
3266 add_to_unconfirmed(new);
3269 exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
3270 exid->clientid.cl_id = conf->cl_clientid.cl_id;
3272 exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
3273 nfsd4_set_ex_flags(conf, exid);
3275 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
3276 conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
3280 spin_unlock(&nn->client_lock);
3285 trace_nfsd_clid_expire_unconf(&unconf->cl_clientid);
3286 expire_client(unconf);
3292 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
3294 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
3297 /* The slot is in use, and no response has been sent. */
3299 if (seqid == slot_seqid)
3300 return nfserr_jukebox;
3302 return nfserr_seq_misordered;
3304 /* Note unsigned 32-bit arithmetic handles wraparound: */
3305 if (likely(seqid == slot_seqid + 1))
3307 if (seqid == slot_seqid)
3308 return nfserr_replay_cache;
3309 return nfserr_seq_misordered;
3313 * Cache the create session result into the create session single DRC
3314 * slot cache by saving the xdr structure. sl_seqid has been set.
3315 * Do this for solo or embedded create session operations.
3318 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
3319 struct nfsd4_clid_slot *slot, __be32 nfserr)
3321 slot->sl_status = nfserr;
3322 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
3326 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
3327 struct nfsd4_clid_slot *slot)
3329 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
3330 return slot->sl_status;
3333 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
3334 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
3335 1 + /* MIN tag is length with zero, only length */ \
3336 3 + /* version, opcount, opcode */ \
3337 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3338 /* seqid, slotID, slotID, cache */ \
3339 4 ) * sizeof(__be32))
3341 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
3342 2 + /* verifier: AUTH_NULL, length 0 */\
3344 1 + /* MIN tag is length with zero, only length */ \
3345 3 + /* opcount, opcode, opstatus*/ \
3346 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3347 /* seqid, slotID, slotID, slotID, status */ \
3348 5 ) * sizeof(__be32))
3350 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
3352 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
3354 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
3355 return nfserr_toosmall;
3356 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
3357 return nfserr_toosmall;
3358 ca->headerpadsz = 0;
3359 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
3360 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
3361 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
3362 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
3363 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
3364 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
3366 * Note decreasing slot size below client's request may make it
3367 * difficult for client to function correctly, whereas
3368 * decreasing the number of slots will (just?) affect
3369 * performance. When short on memory we therefore prefer to
3370 * decrease number of slots instead of their size. Clients that
3371 * request larger slots than they need will get poor results:
3372 * Note that we always allow at least one slot, because our
3373 * accounting is soft and provides no guarantees either way.
3375 ca->maxreqs = nfsd4_get_drc_mem(ca, nn);
3381 * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
3382 * These are based on similar macros in linux/sunrpc/msg_prot.h .
3384 #define RPC_MAX_HEADER_WITH_AUTH_SYS \
3385 (RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
3387 #define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
3388 (RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
3390 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
3391 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
3392 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
3393 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
3396 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
3398 ca->headerpadsz = 0;
3400 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
3401 return nfserr_toosmall;
3402 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
3403 return nfserr_toosmall;
3404 ca->maxresp_cached = 0;
3406 return nfserr_toosmall;
3411 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
3413 switch (cbs->flavor) {
3419 * GSS case: the spec doesn't allow us to return this
3420 * error. But it also doesn't allow us not to support
3422 * I'd rather this fail hard than return some error the
3423 * client might think it can already handle:
3425 return nfserr_encr_alg_unsupp;
3430 nfsd4_create_session(struct svc_rqst *rqstp,
3431 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
3433 struct nfsd4_create_session *cr_ses = &u->create_session;
3434 struct sockaddr *sa = svc_addr(rqstp);
3435 struct nfs4_client *conf, *unconf;
3436 struct nfs4_client *old = NULL;
3437 struct nfsd4_session *new;
3438 struct nfsd4_conn *conn;
3439 struct nfsd4_clid_slot *cs_slot = NULL;
3441 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3443 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
3444 return nfserr_inval;
3445 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
3448 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
3451 status = check_backchannel_attrs(&cr_ses->back_channel);
3453 goto out_release_drc_mem;
3454 status = nfserr_jukebox;
3455 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
3457 goto out_release_drc_mem;
3458 conn = alloc_conn_from_crses(rqstp, cr_ses);
3460 goto out_free_session;
3462 spin_lock(&nn->client_lock);
3463 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
3464 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
3465 WARN_ON_ONCE(conf && unconf);
3468 status = nfserr_wrong_cred;
3469 if (!nfsd4_mach_creds_match(conf, rqstp))
3471 cs_slot = &conf->cl_cs_slot;
3472 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3474 if (status == nfserr_replay_cache)
3475 status = nfsd4_replay_create_session(cr_ses, cs_slot);
3478 } else if (unconf) {
3479 status = nfserr_clid_inuse;
3480 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
3481 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
3482 trace_nfsd_clid_cred_mismatch(unconf, rqstp);
3485 status = nfserr_wrong_cred;
3486 if (!nfsd4_mach_creds_match(unconf, rqstp))
3488 cs_slot = &unconf->cl_cs_slot;
3489 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3491 /* an unconfirmed replay returns misordered */
3492 status = nfserr_seq_misordered;
3495 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3497 status = mark_client_expired_locked(old);
3502 trace_nfsd_clid_replaced(&old->cl_clientid);
3504 move_to_confirmed(unconf);
3507 status = nfserr_stale_clientid;
3511 /* Persistent sessions are not supported */
3512 cr_ses->flags &= ~SESSION4_PERSIST;
3513 /* Upshifting from TCP to RDMA is not supported */
3514 cr_ses->flags &= ~SESSION4_RDMA;
3516 init_session(rqstp, new, conf, cr_ses);
3517 nfsd4_get_session_locked(new);
3519 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
3520 NFS4_MAX_SESSIONID_LEN);
3521 cs_slot->sl_seqid++;
3522 cr_ses->seqid = cs_slot->sl_seqid;
3524 /* cache solo and embedded create sessions under the client_lock */
3525 nfsd4_cache_create_session(cr_ses, cs_slot, status);
3526 spin_unlock(&nn->client_lock);
3528 fsnotify_dentry(conf->cl_nfsd_info_dentry, FS_MODIFY);
3529 /* init connection and backchannel */
3530 nfsd4_init_conn(rqstp, conn, new);
3531 nfsd4_put_session(new);
3536 spin_unlock(&nn->client_lock);
3541 __free_session(new);
3542 out_release_drc_mem:
3543 nfsd4_put_drc_mem(&cr_ses->fore_channel);
3547 static __be32 nfsd4_map_bcts_dir(u32 *dir)
3550 case NFS4_CDFC4_FORE:
3551 case NFS4_CDFC4_BACK:
3553 case NFS4_CDFC4_FORE_OR_BOTH:
3554 case NFS4_CDFC4_BACK_OR_BOTH:
3555 *dir = NFS4_CDFC4_BOTH;
3558 return nfserr_inval;
3561 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp,
3562 struct nfsd4_compound_state *cstate,
3563 union nfsd4_op_u *u)
3565 struct nfsd4_backchannel_ctl *bc = &u->backchannel_ctl;
3566 struct nfsd4_session *session = cstate->session;
3567 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3570 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
3573 spin_lock(&nn->client_lock);
3574 session->se_cb_prog = bc->bc_cb_program;
3575 session->se_cb_sec = bc->bc_cb_sec;
3576 spin_unlock(&nn->client_lock);
3578 nfsd4_probe_callback(session->se_client);
3583 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
3585 struct nfsd4_conn *c;
3587 list_for_each_entry(c, &s->se_conns, cn_persession) {
3588 if (c->cn_xprt == xpt) {
3595 static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,
3596 struct nfsd4_session *session, u32 req, struct nfsd4_conn **conn)
3598 struct nfs4_client *clp = session->se_client;
3599 struct svc_xprt *xpt = rqst->rq_xprt;
3600 struct nfsd4_conn *c;
3603 /* Following the last paragraph of RFC 5661 Section 18.34.3: */
3604 spin_lock(&clp->cl_lock);
3605 c = __nfsd4_find_conn(xpt, session);
3607 status = nfserr_noent;
3608 else if (req == c->cn_flags)
3610 else if (req == NFS4_CDFC4_FORE_OR_BOTH &&
3611 c->cn_flags != NFS4_CDFC4_BACK)
3613 else if (req == NFS4_CDFC4_BACK_OR_BOTH &&
3614 c->cn_flags != NFS4_CDFC4_FORE)
3617 status = nfserr_inval;
3618 spin_unlock(&clp->cl_lock);
3619 if (status == nfs_ok && conn)
3624 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
3625 struct nfsd4_compound_state *cstate,
3626 union nfsd4_op_u *u)
3628 struct nfsd4_bind_conn_to_session *bcts = &u->bind_conn_to_session;
3630 struct nfsd4_conn *conn;
3631 struct nfsd4_session *session;
3632 struct net *net = SVC_NET(rqstp);
3633 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3635 if (!nfsd4_last_compound_op(rqstp))
3636 return nfserr_not_only_op;
3637 spin_lock(&nn->client_lock);
3638 session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
3639 spin_unlock(&nn->client_lock);
3641 goto out_no_session;
3642 status = nfserr_wrong_cred;
3643 if (!nfsd4_mach_creds_match(session->se_client, rqstp))
3645 status = nfsd4_match_existing_connection(rqstp, session,
3647 if (status == nfs_ok) {
3648 if (bcts->dir == NFS4_CDFC4_FORE_OR_BOTH ||
3649 bcts->dir == NFS4_CDFC4_BACK)
3650 conn->cn_flags |= NFS4_CDFC4_BACK;
3651 nfsd4_probe_callback(session->se_client);
3654 if (status == nfserr_inval)
3656 status = nfsd4_map_bcts_dir(&bcts->dir);
3659 conn = alloc_conn(rqstp, bcts->dir);
3660 status = nfserr_jukebox;
3663 nfsd4_init_conn(rqstp, conn, session);
3666 nfsd4_put_session(session);
3671 static bool nfsd4_compound_in_session(struct nfsd4_compound_state *cstate, struct nfs4_sessionid *sid)
3673 if (!cstate->session)
3675 return !memcmp(sid, &cstate->session->se_sessionid, sizeof(*sid));
3679 nfsd4_destroy_session(struct svc_rqst *r, struct nfsd4_compound_state *cstate,
3680 union nfsd4_op_u *u)
3682 struct nfs4_sessionid *sessionid = &u->destroy_session.sessionid;
3683 struct nfsd4_session *ses;
3685 int ref_held_by_me = 0;
3686 struct net *net = SVC_NET(r);
3687 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3689 status = nfserr_not_only_op;
3690 if (nfsd4_compound_in_session(cstate, sessionid)) {
3691 if (!nfsd4_last_compound_op(r))
3695 dump_sessionid(__func__, sessionid);
3696 spin_lock(&nn->client_lock);
3697 ses = find_in_sessionid_hashtbl(sessionid, net, &status);
3699 goto out_client_lock;
3700 status = nfserr_wrong_cred;
3701 if (!nfsd4_mach_creds_match(ses->se_client, r))
3702 goto out_put_session;
3703 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
3705 goto out_put_session;
3706 unhash_session(ses);
3707 spin_unlock(&nn->client_lock);
3709 nfsd4_probe_callback_sync(ses->se_client);
3711 spin_lock(&nn->client_lock);
3714 nfsd4_put_session_locked(ses);
3716 spin_unlock(&nn->client_lock);
3721 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
3723 struct nfs4_client *clp = ses->se_client;
3724 struct nfsd4_conn *c;
3725 __be32 status = nfs_ok;
3728 spin_lock(&clp->cl_lock);
3729 c = __nfsd4_find_conn(new->cn_xprt, ses);
3732 status = nfserr_conn_not_bound_to_session;
3733 if (clp->cl_mach_cred)
3735 __nfsd4_hash_conn(new, ses);
3736 spin_unlock(&clp->cl_lock);
3737 ret = nfsd4_register_conn(new);
3739 /* oops; xprt is already down: */
3740 nfsd4_conn_lost(&new->cn_xpt_user);
3743 spin_unlock(&clp->cl_lock);
3748 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
3750 struct nfsd4_compoundargs *args = rqstp->rq_argp;
3752 return args->opcnt > session->se_fchannel.maxops;
3755 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
3756 struct nfsd4_session *session)
3758 struct xdr_buf *xb = &rqstp->rq_arg;
3760 return xb->len > session->se_fchannel.maxreq_sz;
3763 static bool replay_matches_cache(struct svc_rqst *rqstp,
3764 struct nfsd4_sequence *seq, struct nfsd4_slot *slot)
3766 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3768 if ((bool)(slot->sl_flags & NFSD4_SLOT_CACHETHIS) !=
3769 (bool)seq->cachethis)
3772 * If there's an error then the reply can have fewer ops than
3775 if (slot->sl_opcnt < argp->opcnt && !slot->sl_status)
3778 * But if we cached a reply with *more* ops than the call you're
3779 * sending us now, then this new call is clearly not really a
3780 * replay of the old one:
3782 if (slot->sl_opcnt > argp->opcnt)
3784 /* This is the only check explicitly called by spec: */
3785 if (!same_creds(&rqstp->rq_cred, &slot->sl_cred))
3788 * There may be more comparisons we could actually do, but the
3789 * spec doesn't require us to catch every case where the calls
3790 * don't match (that would require caching the call as well as
3791 * the reply), so we don't bother.
3797 nfsd4_sequence(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3798 union nfsd4_op_u *u)
3800 struct nfsd4_sequence *seq = &u->sequence;
3801 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3802 struct xdr_stream *xdr = resp->xdr;
3803 struct nfsd4_session *session;
3804 struct nfs4_client *clp;
3805 struct nfsd4_slot *slot;
3806 struct nfsd4_conn *conn;
3809 struct net *net = SVC_NET(rqstp);
3810 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3812 if (resp->opcnt != 1)
3813 return nfserr_sequence_pos;
3816 * Will be either used or freed by nfsd4_sequence_check_conn
3819 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
3821 return nfserr_jukebox;
3823 spin_lock(&nn->client_lock);
3824 session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
3826 goto out_no_session;
3827 clp = session->se_client;
3829 status = nfserr_too_many_ops;
3830 if (nfsd4_session_too_many_ops(rqstp, session))
3831 goto out_put_session;
3833 status = nfserr_req_too_big;
3834 if (nfsd4_request_too_big(rqstp, session))
3835 goto out_put_session;
3837 status = nfserr_badslot;
3838 if (seq->slotid >= session->se_fchannel.maxreqs)
3839 goto out_put_session;
3841 slot = session->se_slots[seq->slotid];
3842 dprintk("%s: slotid %d\n", __func__, seq->slotid);
3844 /* We do not negotiate the number of slots yet, so set the
3845 * maxslots to the session maxreqs which is used to encode
3846 * sr_highest_slotid and the sr_target_slot id to maxslots */
3847 seq->maxslots = session->se_fchannel.maxreqs;
3849 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
3850 slot->sl_flags & NFSD4_SLOT_INUSE);
3851 if (status == nfserr_replay_cache) {
3852 status = nfserr_seq_misordered;
3853 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
3854 goto out_put_session;
3855 status = nfserr_seq_false_retry;
3856 if (!replay_matches_cache(rqstp, seq, slot))
3857 goto out_put_session;
3858 cstate->slot = slot;
3859 cstate->session = session;
3861 /* Return the cached reply status and set cstate->status
3862 * for nfsd4_proc_compound processing */
3863 status = nfsd4_replay_cache_entry(resp, seq);
3864 cstate->status = nfserr_replay_cache;
3868 goto out_put_session;
3870 status = nfsd4_sequence_check_conn(conn, session);
3873 goto out_put_session;
3875 buflen = (seq->cachethis) ?
3876 session->se_fchannel.maxresp_cached :
3877 session->se_fchannel.maxresp_sz;
3878 status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
3880 if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
3881 goto out_put_session;
3882 svc_reserve(rqstp, buflen);
3885 /* Success! bump slot seqid */
3886 slot->sl_seqid = seq->seqid;
3887 slot->sl_flags |= NFSD4_SLOT_INUSE;
3889 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
3891 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
3893 cstate->slot = slot;
3894 cstate->session = session;
3898 switch (clp->cl_cb_state) {
3900 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
3902 case NFSD4_CB_FAULT:
3903 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
3906 seq->status_flags = 0;
3908 if (!list_empty(&clp->cl_revoked))
3909 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
3913 spin_unlock(&nn->client_lock);
3916 nfsd4_put_session_locked(session);
3917 goto out_no_session;
3921 nfsd4_sequence_done(struct nfsd4_compoundres *resp)
3923 struct nfsd4_compound_state *cs = &resp->cstate;
3925 if (nfsd4_has_session(cs)) {
3926 if (cs->status != nfserr_replay_cache) {
3927 nfsd4_store_cache_entry(resp);
3928 cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
3930 /* Drop session reference that was taken in nfsd4_sequence() */
3931 nfsd4_put_session(cs->session);
3933 put_client_renew(cs->clp);
3937 nfsd4_destroy_clientid(struct svc_rqst *rqstp,
3938 struct nfsd4_compound_state *cstate,
3939 union nfsd4_op_u *u)
3941 struct nfsd4_destroy_clientid *dc = &u->destroy_clientid;
3942 struct nfs4_client *conf, *unconf;
3943 struct nfs4_client *clp = NULL;
3945 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3947 spin_lock(&nn->client_lock);
3948 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
3949 conf = find_confirmed_client(&dc->clientid, true, nn);
3950 WARN_ON_ONCE(conf && unconf);
3953 if (client_has_state(conf)) {
3954 status = nfserr_clientid_busy;
3957 status = mark_client_expired_locked(conf);
3964 status = nfserr_stale_clientid;
3967 if (!nfsd4_mach_creds_match(clp, rqstp)) {
3969 status = nfserr_wrong_cred;
3972 trace_nfsd_clid_destroyed(&clp->cl_clientid);
3973 unhash_client_locked(clp);
3975 spin_unlock(&nn->client_lock);
3982 nfsd4_reclaim_complete(struct svc_rqst *rqstp,
3983 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
3985 struct nfsd4_reclaim_complete *rc = &u->reclaim_complete;
3986 struct nfs4_client *clp = cstate->clp;
3989 if (rc->rca_one_fs) {
3990 if (!cstate->current_fh.fh_dentry)
3991 return nfserr_nofilehandle;
3993 * We don't take advantage of the rca_one_fs case.
3994 * That's OK, it's optional, we can safely ignore it.
3999 status = nfserr_complete_already;
4000 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
4003 status = nfserr_stale_clientid;
4004 if (is_client_expired(clp))
4006 * The following error isn't really legal.
4007 * But we only get here if the client just explicitly
4008 * destroyed the client. Surely it no longer cares what
4009 * error it gets back on an operation for the dead
4015 trace_nfsd_clid_reclaim_complete(&clp->cl_clientid);
4016 nfsd4_client_record_create(clp);
4017 inc_reclaim_complete(clp);
4023 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4024 union nfsd4_op_u *u)
4026 struct nfsd4_setclientid *setclid = &u->setclientid;
4027 struct xdr_netobj clname = setclid->se_name;
4028 nfs4_verifier clverifier = setclid->se_verf;
4029 struct nfs4_client *conf, *new;
4030 struct nfs4_client *unconf = NULL;
4032 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4034 new = create_client(clname, rqstp, &clverifier);
4036 return nfserr_jukebox;
4037 spin_lock(&nn->client_lock);
4038 conf = find_confirmed_client_by_name(&clname, nn);
4039 if (conf && client_has_state(conf)) {
4040 status = nfserr_clid_inuse;
4041 if (clp_used_exchangeid(conf))
4043 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
4044 trace_nfsd_clid_cred_mismatch(conf, rqstp);
4048 unconf = find_unconfirmed_client_by_name(&clname, nn);
4050 unhash_client_locked(unconf);
4052 if (same_verf(&conf->cl_verifier, &clverifier)) {
4053 copy_clid(new, conf);
4054 gen_confirm(new, nn);
4056 trace_nfsd_clid_verf_mismatch(conf, rqstp,
4059 trace_nfsd_clid_fresh(new);
4060 new->cl_minorversion = 0;
4061 gen_callback(new, setclid, rqstp);
4062 add_to_unconfirmed(new);
4063 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
4064 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
4065 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
4069 spin_unlock(&nn->client_lock);
4073 trace_nfsd_clid_expire_unconf(&unconf->cl_clientid);
4074 expire_client(unconf);
4080 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
4081 struct nfsd4_compound_state *cstate,
4082 union nfsd4_op_u *u)
4084 struct nfsd4_setclientid_confirm *setclientid_confirm =
4085 &u->setclientid_confirm;
4086 struct nfs4_client *conf, *unconf;
4087 struct nfs4_client *old = NULL;
4088 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
4089 clientid_t * clid = &setclientid_confirm->sc_clientid;
4091 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4093 if (STALE_CLIENTID(clid, nn))
4094 return nfserr_stale_clientid;
4096 spin_lock(&nn->client_lock);
4097 conf = find_confirmed_client(clid, false, nn);
4098 unconf = find_unconfirmed_client(clid, false, nn);
4100 * We try hard to give out unique clientid's, so if we get an
4101 * attempt to confirm the same clientid with a different cred,
4102 * the client may be buggy; this should never happen.
4104 * Nevertheless, RFC 7530 recommends INUSE for this case:
4106 status = nfserr_clid_inuse;
4107 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
4108 trace_nfsd_clid_cred_mismatch(unconf, rqstp);
4111 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
4112 trace_nfsd_clid_cred_mismatch(conf, rqstp);
4115 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
4116 if (conf && same_verf(&confirm, &conf->cl_confirm)) {
4119 status = nfserr_stale_clientid;
4125 unhash_client_locked(old);
4126 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
4128 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
4130 status = nfserr_clid_inuse;
4131 if (client_has_state(old)
4132 && !same_creds(&unconf->cl_cred,
4135 status = mark_client_expired_locked(old);
4140 trace_nfsd_clid_replaced(&old->cl_clientid);
4142 move_to_confirmed(unconf);
4145 get_client_locked(conf);
4146 spin_unlock(&nn->client_lock);
4148 fsnotify_dentry(conf->cl_nfsd_info_dentry, FS_MODIFY);
4149 nfsd4_probe_callback(conf);
4150 spin_lock(&nn->client_lock);
4151 put_client_renew_locked(conf);
4153 spin_unlock(&nn->client_lock);
4159 static struct nfs4_file *nfsd4_alloc_file(void)
4161 return kmem_cache_alloc(file_slab, GFP_KERNEL);
4164 /* OPEN Share state helper functions */
4165 static void nfsd4_init_file(struct svc_fh *fh, unsigned int hashval,
4166 struct nfs4_file *fp)
4168 lockdep_assert_held(&state_lock);
4170 refcount_set(&fp->fi_ref, 1);
4171 spin_lock_init(&fp->fi_lock);
4172 INIT_LIST_HEAD(&fp->fi_stateids);
4173 INIT_LIST_HEAD(&fp->fi_delegations);
4174 INIT_LIST_HEAD(&fp->fi_clnt_odstate);
4175 fh_copy_shallow(&fp->fi_fhandle, &fh->fh_handle);
4176 fp->fi_deleg_file = NULL;
4177 fp->fi_had_conflict = false;
4178 fp->fi_share_deny = 0;
4179 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
4180 memset(fp->fi_access, 0, sizeof(fp->fi_access));
4181 fp->fi_aliased = false;
4182 fp->fi_inode = d_inode(fh->fh_dentry);
4183 #ifdef CONFIG_NFSD_PNFS
4184 INIT_LIST_HEAD(&fp->fi_lo_states);
4185 atomic_set(&fp->fi_lo_recalls, 0);
4187 hlist_add_head_rcu(&fp->fi_hash, &file_hashtbl[hashval]);
4191 nfsd4_free_slabs(void)
4193 kmem_cache_destroy(client_slab);
4194 kmem_cache_destroy(openowner_slab);
4195 kmem_cache_destroy(lockowner_slab);
4196 kmem_cache_destroy(file_slab);
4197 kmem_cache_destroy(stateid_slab);
4198 kmem_cache_destroy(deleg_slab);
4199 kmem_cache_destroy(odstate_slab);
4203 nfsd4_init_slabs(void)
4205 client_slab = kmem_cache_create("nfsd4_clients",
4206 sizeof(struct nfs4_client), 0, 0, NULL);
4207 if (client_slab == NULL)
4209 openowner_slab = kmem_cache_create("nfsd4_openowners",
4210 sizeof(struct nfs4_openowner), 0, 0, NULL);
4211 if (openowner_slab == NULL)
4212 goto out_free_client_slab;
4213 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
4214 sizeof(struct nfs4_lockowner), 0, 0, NULL);
4215 if (lockowner_slab == NULL)
4216 goto out_free_openowner_slab;
4217 file_slab = kmem_cache_create("nfsd4_files",
4218 sizeof(struct nfs4_file), 0, 0, NULL);
4219 if (file_slab == NULL)
4220 goto out_free_lockowner_slab;
4221 stateid_slab = kmem_cache_create("nfsd4_stateids",
4222 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
4223 if (stateid_slab == NULL)
4224 goto out_free_file_slab;
4225 deleg_slab = kmem_cache_create("nfsd4_delegations",
4226 sizeof(struct nfs4_delegation), 0, 0, NULL);
4227 if (deleg_slab == NULL)
4228 goto out_free_stateid_slab;
4229 odstate_slab = kmem_cache_create("nfsd4_odstate",
4230 sizeof(struct nfs4_clnt_odstate), 0, 0, NULL);
4231 if (odstate_slab == NULL)
4232 goto out_free_deleg_slab;
4235 out_free_deleg_slab:
4236 kmem_cache_destroy(deleg_slab);
4237 out_free_stateid_slab:
4238 kmem_cache_destroy(stateid_slab);
4240 kmem_cache_destroy(file_slab);
4241 out_free_lockowner_slab:
4242 kmem_cache_destroy(lockowner_slab);
4243 out_free_openowner_slab:
4244 kmem_cache_destroy(openowner_slab);
4245 out_free_client_slab:
4246 kmem_cache_destroy(client_slab);
4251 static void init_nfs4_replay(struct nfs4_replay *rp)
4253 rp->rp_status = nfserr_serverfault;
4255 rp->rp_buf = rp->rp_ibuf;
4256 mutex_init(&rp->rp_mutex);
4259 static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
4260 struct nfs4_stateowner *so)
4262 if (!nfsd4_has_session(cstate)) {
4263 mutex_lock(&so->so_replay.rp_mutex);
4264 cstate->replay_owner = nfs4_get_stateowner(so);
4268 void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
4270 struct nfs4_stateowner *so = cstate->replay_owner;
4273 cstate->replay_owner = NULL;
4274 mutex_unlock(&so->so_replay.rp_mutex);
4275 nfs4_put_stateowner(so);
4279 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
4281 struct nfs4_stateowner *sop;
4283 sop = kmem_cache_alloc(slab, GFP_KERNEL);
4287 xdr_netobj_dup(&sop->so_owner, owner, GFP_KERNEL);
4288 if (!sop->so_owner.data) {
4289 kmem_cache_free(slab, sop);
4293 INIT_LIST_HEAD(&sop->so_stateids);
4294 sop->so_client = clp;
4295 init_nfs4_replay(&sop->so_replay);
4296 atomic_set(&sop->so_count, 1);
4300 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
4302 lockdep_assert_held(&clp->cl_lock);
4304 list_add(&oo->oo_owner.so_strhash,
4305 &clp->cl_ownerstr_hashtbl[strhashval]);
4306 list_add(&oo->oo_perclient, &clp->cl_openowners);
4309 static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
4311 unhash_openowner_locked(openowner(so));
4314 static void nfs4_free_openowner(struct nfs4_stateowner *so)
4316 struct nfs4_openowner *oo = openowner(so);
4318 kmem_cache_free(openowner_slab, oo);
4321 static const struct nfs4_stateowner_operations openowner_ops = {
4322 .so_unhash = nfs4_unhash_openowner,
4323 .so_free = nfs4_free_openowner,
4326 static struct nfs4_ol_stateid *
4327 nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4329 struct nfs4_ol_stateid *local, *ret = NULL;
4330 struct nfs4_openowner *oo = open->op_openowner;
4332 lockdep_assert_held(&fp->fi_lock);
4334 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
4335 /* ignore lock owners */
4336 if (local->st_stateowner->so_is_open_owner == 0)
4338 if (local->st_stateowner != &oo->oo_owner)
4340 if (local->st_stid.sc_type == NFS4_OPEN_STID) {
4342 refcount_inc(&ret->st_stid.sc_count);
4350 nfsd4_verify_open_stid(struct nfs4_stid *s)
4352 __be32 ret = nfs_ok;
4354 switch (s->sc_type) {
4358 case NFS4_CLOSED_STID:
4359 case NFS4_CLOSED_DELEG_STID:
4360 ret = nfserr_bad_stateid;
4362 case NFS4_REVOKED_DELEG_STID:
4363 ret = nfserr_deleg_revoked;
4368 /* Lock the stateid st_mutex, and deal with races with CLOSE */
4370 nfsd4_lock_ol_stateid(struct nfs4_ol_stateid *stp)
4374 mutex_lock_nested(&stp->st_mutex, LOCK_STATEID_MUTEX);
4375 ret = nfsd4_verify_open_stid(&stp->st_stid);
4377 mutex_unlock(&stp->st_mutex);
4381 static struct nfs4_ol_stateid *
4382 nfsd4_find_and_lock_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4384 struct nfs4_ol_stateid *stp;
4386 spin_lock(&fp->fi_lock);
4387 stp = nfsd4_find_existing_open(fp, open);
4388 spin_unlock(&fp->fi_lock);
4389 if (!stp || nfsd4_lock_ol_stateid(stp) == nfs_ok)
4391 nfs4_put_stid(&stp->st_stid);
4396 static struct nfs4_openowner *
4397 alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
4398 struct nfsd4_compound_state *cstate)
4400 struct nfs4_client *clp = cstate->clp;
4401 struct nfs4_openowner *oo, *ret;
4403 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
4406 oo->oo_owner.so_ops = &openowner_ops;
4407 oo->oo_owner.so_is_open_owner = 1;
4408 oo->oo_owner.so_seqid = open->op_seqid;
4410 if (nfsd4_has_session(cstate))
4411 oo->oo_flags |= NFS4_OO_CONFIRMED;
4413 oo->oo_last_closed_stid = NULL;
4414 INIT_LIST_HEAD(&oo->oo_close_lru);
4415 spin_lock(&clp->cl_lock);
4416 ret = find_openstateowner_str_locked(strhashval, open, clp);
4418 hash_openowner(oo, clp, strhashval);
4421 nfs4_free_stateowner(&oo->oo_owner);
4423 spin_unlock(&clp->cl_lock);
4427 static struct nfs4_ol_stateid *
4428 init_open_stateid(struct nfs4_file *fp, struct nfsd4_open *open)
4431 struct nfs4_openowner *oo = open->op_openowner;
4432 struct nfs4_ol_stateid *retstp = NULL;
4433 struct nfs4_ol_stateid *stp;
4436 /* We are moving these outside of the spinlocks to avoid the warnings */
4437 mutex_init(&stp->st_mutex);
4438 mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
4441 spin_lock(&oo->oo_owner.so_client->cl_lock);
4442 spin_lock(&fp->fi_lock);
4444 retstp = nfsd4_find_existing_open(fp, open);
4448 open->op_stp = NULL;
4449 refcount_inc(&stp->st_stid.sc_count);
4450 stp->st_stid.sc_type = NFS4_OPEN_STID;
4451 INIT_LIST_HEAD(&stp->st_locks);
4452 stp->st_stateowner = nfs4_get_stateowner(&oo->oo_owner);
4454 stp->st_stid.sc_file = fp;
4455 stp->st_access_bmap = 0;
4456 stp->st_deny_bmap = 0;
4457 stp->st_openstp = NULL;
4458 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
4459 list_add(&stp->st_perfile, &fp->fi_stateids);
4462 spin_unlock(&fp->fi_lock);
4463 spin_unlock(&oo->oo_owner.so_client->cl_lock);
4465 /* Handle races with CLOSE */
4466 if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
4467 nfs4_put_stid(&retstp->st_stid);
4470 /* To keep mutex tracking happy */
4471 mutex_unlock(&stp->st_mutex);
4478 * In the 4.0 case we need to keep the owners around a little while to handle
4479 * CLOSE replay. We still do need to release any file access that is held by
4480 * them before returning however.
4483 move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
4485 struct nfs4_ol_stateid *last;
4486 struct nfs4_openowner *oo = openowner(s->st_stateowner);
4487 struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
4490 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
4493 * We know that we hold one reference via nfsd4_close, and another
4494 * "persistent" reference for the client. If the refcount is higher
4495 * than 2, then there are still calls in progress that are using this
4496 * stateid. We can't put the sc_file reference until they are finished.
4497 * Wait for the refcount to drop to 2. Since it has been unhashed,
4498 * there should be no danger of the refcount going back up again at
4501 wait_event(close_wq, refcount_read(&s->st_stid.sc_count) == 2);
4503 release_all_access(s);
4504 if (s->st_stid.sc_file) {
4505 put_nfs4_file(s->st_stid.sc_file);
4506 s->st_stid.sc_file = NULL;
4509 spin_lock(&nn->client_lock);
4510 last = oo->oo_last_closed_stid;
4511 oo->oo_last_closed_stid = s;
4512 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
4513 oo->oo_time = ktime_get_boottime_seconds();
4514 spin_unlock(&nn->client_lock);
4516 nfs4_put_stid(&last->st_stid);
4519 /* search file_hashtbl[] for file */
4520 static struct nfs4_file *
4521 find_file_locked(struct svc_fh *fh, unsigned int hashval)
4523 struct nfs4_file *fp;
4525 hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash,
4526 lockdep_is_held(&state_lock)) {
4527 if (fh_match(&fp->fi_fhandle, &fh->fh_handle)) {
4528 if (refcount_inc_not_zero(&fp->fi_ref))
4535 static struct nfs4_file *insert_file(struct nfs4_file *new, struct svc_fh *fh,
4536 unsigned int hashval)
4538 struct nfs4_file *fp;
4539 struct nfs4_file *ret = NULL;
4540 bool alias_found = false;
4542 spin_lock(&state_lock);
4543 hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash,
4544 lockdep_is_held(&state_lock)) {
4545 if (fh_match(&fp->fi_fhandle, &fh->fh_handle)) {
4546 if (refcount_inc_not_zero(&fp->fi_ref))
4548 } else if (d_inode(fh->fh_dentry) == fp->fi_inode)
4549 fp->fi_aliased = alias_found = true;
4551 if (likely(ret == NULL)) {
4552 nfsd4_init_file(fh, hashval, new);
4553 new->fi_aliased = alias_found;
4556 spin_unlock(&state_lock);
4560 static struct nfs4_file * find_file(struct svc_fh *fh)
4562 struct nfs4_file *fp;
4563 unsigned int hashval = file_hashval(fh);
4566 fp = find_file_locked(fh, hashval);
4571 static struct nfs4_file *
4572 find_or_add_file(struct nfs4_file *new, struct svc_fh *fh)
4574 struct nfs4_file *fp;
4575 unsigned int hashval = file_hashval(fh);
4578 fp = find_file_locked(fh, hashval);
4583 return insert_file(new, fh, hashval);
4587 * Called to check deny when READ with all zero stateid or
4588 * WRITE with all zero or all one stateid
4591 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
4593 struct nfs4_file *fp;
4594 __be32 ret = nfs_ok;
4596 fp = find_file(current_fh);
4599 /* Check for conflicting share reservations */
4600 spin_lock(&fp->fi_lock);
4601 if (fp->fi_share_deny & deny_type)
4602 ret = nfserr_locked;
4603 spin_unlock(&fp->fi_lock);
4608 static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
4610 struct nfs4_delegation *dp = cb_to_delegation(cb);
4611 struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
4614 block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
4617 * We can't do this in nfsd_break_deleg_cb because it is
4618 * already holding inode->i_lock.
4620 * If the dl_time != 0, then we know that it has already been
4621 * queued for a lease break. Don't queue it again.
4623 spin_lock(&state_lock);
4624 if (delegation_hashed(dp) && dp->dl_time == 0) {
4625 dp->dl_time = ktime_get_boottime_seconds();
4626 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
4628 spin_unlock(&state_lock);
4631 static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
4632 struct rpc_task *task)
4634 struct nfs4_delegation *dp = cb_to_delegation(cb);
4636 if (dp->dl_stid.sc_type == NFS4_CLOSED_DELEG_STID ||
4637 dp->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4640 switch (task->tk_status) {
4643 case -NFS4ERR_DELAY:
4644 rpc_delay(task, 2 * HZ);
4647 case -NFS4ERR_BAD_STATEID:
4649 * Race: client probably got cb_recall before open reply
4650 * granting delegation.
4652 if (dp->dl_retries--) {
4653 rpc_delay(task, 2 * HZ);
4662 static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
4664 struct nfs4_delegation *dp = cb_to_delegation(cb);
4666 nfs4_put_stid(&dp->dl_stid);
4669 static const struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
4670 .prepare = nfsd4_cb_recall_prepare,
4671 .done = nfsd4_cb_recall_done,
4672 .release = nfsd4_cb_recall_release,
4675 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
4678 * We're assuming the state code never drops its reference
4679 * without first removing the lease. Since we're in this lease
4680 * callback (and since the lease code is serialized by the
4681 * i_lock) we know the server hasn't removed the lease yet, and
4682 * we know it's safe to take a reference.
4684 refcount_inc(&dp->dl_stid.sc_count);
4685 nfsd4_run_cb(&dp->dl_recall);
4688 /* Called from break_lease() with i_lock held. */
4690 nfsd_break_deleg_cb(struct file_lock *fl)
4693 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
4694 struct nfs4_file *fp = dp->dl_stid.sc_file;
4696 trace_nfsd_cb_recall(&dp->dl_stid);
4699 * We don't want the locks code to timeout the lease for us;
4700 * we'll remove it ourself if a delegation isn't returned
4703 fl->fl_break_time = 0;
4705 spin_lock(&fp->fi_lock);
4706 fp->fi_had_conflict = true;
4707 nfsd_break_one_deleg(dp);
4708 spin_unlock(&fp->fi_lock);
4712 static bool nfsd_breaker_owns_lease(struct file_lock *fl)
4714 struct nfs4_delegation *dl = fl->fl_owner;
4715 struct svc_rqst *rqst;
4716 struct nfs4_client *clp;
4720 rqst = kthread_data(current);
4721 /* Note rq_prog == NFS_ACL_PROGRAM is also possible: */
4722 if (rqst->rq_prog != NFS_PROGRAM || rqst->rq_vers < 4)
4724 clp = *(rqst->rq_lease_breaker);
4725 return dl->dl_stid.sc_client == clp;
4729 nfsd_change_deleg_cb(struct file_lock *onlist, int arg,
4730 struct list_head *dispose)
4733 return lease_modify(onlist, arg, dispose);
4738 static const struct lock_manager_operations nfsd_lease_mng_ops = {
4739 .lm_breaker_owns_lease = nfsd_breaker_owns_lease,
4740 .lm_break = nfsd_break_deleg_cb,
4741 .lm_change = nfsd_change_deleg_cb,
4744 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
4746 if (nfsd4_has_session(cstate))
4748 if (seqid == so->so_seqid - 1)
4749 return nfserr_replay_me;
4750 if (seqid == so->so_seqid)
4752 return nfserr_bad_seqid;
4755 static struct nfs4_client *lookup_clientid(clientid_t *clid, bool sessions,
4756 struct nfsd_net *nn)
4758 struct nfs4_client *found;
4760 spin_lock(&nn->client_lock);
4761 found = find_confirmed_client(clid, sessions, nn);
4763 atomic_inc(&found->cl_rpc_users);
4764 spin_unlock(&nn->client_lock);
4768 static __be32 set_client(clientid_t *clid,
4769 struct nfsd4_compound_state *cstate,
4770 struct nfsd_net *nn)
4773 if (!same_clid(&cstate->clp->cl_clientid, clid))
4774 return nfserr_stale_clientid;
4777 if (STALE_CLIENTID(clid, nn))
4778 return nfserr_stale_clientid;
4780 * We're in the 4.0 case (otherwise the SEQUENCE op would have
4781 * set cstate->clp), so session = false:
4783 cstate->clp = lookup_clientid(clid, false, nn);
4785 return nfserr_expired;
4790 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
4791 struct nfsd4_open *open, struct nfsd_net *nn)
4793 clientid_t *clientid = &open->op_clientid;
4794 struct nfs4_client *clp = NULL;
4795 unsigned int strhashval;
4796 struct nfs4_openowner *oo = NULL;
4800 * In case we need it later, after we've already created the
4801 * file and don't want to risk a further failure:
4803 open->op_file = nfsd4_alloc_file();
4804 if (open->op_file == NULL)
4805 return nfserr_jukebox;
4807 status = set_client(clientid, cstate, nn);
4812 strhashval = ownerstr_hashval(&open->op_owner);
4813 oo = find_openstateowner_str(strhashval, open, clp);
4814 open->op_openowner = oo;
4818 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4819 /* Replace unconfirmed owners without checking for replay. */
4820 release_openowner(oo);
4821 open->op_openowner = NULL;
4824 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
4829 oo = alloc_init_open_stateowner(strhashval, open, cstate);
4831 return nfserr_jukebox;
4832 open->op_openowner = oo;
4834 open->op_stp = nfs4_alloc_open_stateid(clp);
4836 return nfserr_jukebox;
4838 if (nfsd4_has_session(cstate) &&
4839 (cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
4840 open->op_odstate = alloc_clnt_odstate(clp);
4841 if (!open->op_odstate)
4842 return nfserr_jukebox;
4848 static inline __be32
4849 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
4851 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
4852 return nfserr_openmode;
4857 static int share_access_to_flags(u32 share_access)
4859 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
4862 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
4864 struct nfs4_stid *ret;
4866 ret = find_stateid_by_type(cl, s,
4867 NFS4_DELEG_STID|NFS4_REVOKED_DELEG_STID);
4870 return delegstateid(ret);
4873 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
4875 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
4876 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
4880 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
4881 struct nfs4_delegation **dp)
4884 __be32 status = nfserr_bad_stateid;
4885 struct nfs4_delegation *deleg;
4887 deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
4890 if (deleg->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID) {
4891 nfs4_put_stid(&deleg->dl_stid);
4892 if (cl->cl_minorversion)
4893 status = nfserr_deleg_revoked;
4896 flags = share_access_to_flags(open->op_share_access);
4897 status = nfs4_check_delegmode(deleg, flags);
4899 nfs4_put_stid(&deleg->dl_stid);
4904 if (!nfsd4_is_deleg_cur(open))
4908 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
4912 static inline int nfs4_access_to_access(u32 nfs4_access)
4916 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
4917 flags |= NFSD_MAY_READ;
4918 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
4919 flags |= NFSD_MAY_WRITE;
4923 static inline __be32
4924 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
4925 struct nfsd4_open *open)
4927 struct iattr iattr = {
4928 .ia_valid = ATTR_SIZE,
4931 if (!open->op_truncate)
4933 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
4934 return nfserr_inval;
4935 return nfsd_setattr(rqstp, fh, &iattr, 0, (time64_t)0);
4938 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
4939 struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
4940 struct nfsd4_open *open)
4942 struct nfsd_file *nf = NULL;
4944 int oflag = nfs4_access_to_omode(open->op_share_access);
4945 int access = nfs4_access_to_access(open->op_share_access);
4946 unsigned char old_access_bmap, old_deny_bmap;
4948 spin_lock(&fp->fi_lock);
4951 * Are we trying to set a deny mode that would conflict with
4954 status = nfs4_file_check_deny(fp, open->op_share_deny);
4955 if (status != nfs_ok) {
4956 spin_unlock(&fp->fi_lock);
4960 /* set access to the file */
4961 status = nfs4_file_get_access(fp, open->op_share_access);
4962 if (status != nfs_ok) {
4963 spin_unlock(&fp->fi_lock);
4967 /* Set access bits in stateid */
4968 old_access_bmap = stp->st_access_bmap;
4969 set_access(open->op_share_access, stp);
4971 /* Set new deny mask */
4972 old_deny_bmap = stp->st_deny_bmap;
4973 set_deny(open->op_share_deny, stp);
4974 fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
4976 if (!fp->fi_fds[oflag]) {
4977 spin_unlock(&fp->fi_lock);
4978 status = nfsd_file_acquire(rqstp, cur_fh, access, &nf);
4980 goto out_put_access;
4981 spin_lock(&fp->fi_lock);
4982 if (!fp->fi_fds[oflag]) {
4983 fp->fi_fds[oflag] = nf;
4987 spin_unlock(&fp->fi_lock);
4991 status = nfserrno(nfsd_open_break_lease(cur_fh->fh_dentry->d_inode,
4994 goto out_put_access;
4996 status = nfsd4_truncate(rqstp, cur_fh, open);
4998 goto out_put_access;
5002 stp->st_access_bmap = old_access_bmap;
5003 nfs4_file_put_access(fp, open->op_share_access);
5004 reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
5009 nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp, struct nfsd4_open *open)
5012 unsigned char old_deny_bmap = stp->st_deny_bmap;
5014 if (!test_access(open->op_share_access, stp))
5015 return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
5017 /* test and set deny mode */
5018 spin_lock(&fp->fi_lock);
5019 status = nfs4_file_check_deny(fp, open->op_share_deny);
5020 if (status == nfs_ok) {
5021 set_deny(open->op_share_deny, stp);
5022 fp->fi_share_deny |=
5023 (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
5025 spin_unlock(&fp->fi_lock);
5027 if (status != nfs_ok)
5030 status = nfsd4_truncate(rqstp, cur_fh, open);
5031 if (status != nfs_ok)
5032 reset_union_bmap_deny(old_deny_bmap, stp);
5036 /* Should we give out recallable state?: */
5037 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
5039 if (clp->cl_cb_state == NFSD4_CB_UP)
5042 * In the sessions case, since we don't have to establish a
5043 * separate connection for callbacks, we assume it's OK
5044 * until we hear otherwise:
5046 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
5049 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp,
5052 struct file_lock *fl;
5054 fl = locks_alloc_lock();
5057 fl->fl_lmops = &nfsd_lease_mng_ops;
5058 fl->fl_flags = FL_DELEG;
5059 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
5060 fl->fl_end = OFFSET_MAX;
5061 fl->fl_owner = (fl_owner_t)dp;
5062 fl->fl_pid = current->tgid;
5063 fl->fl_file = dp->dl_stid.sc_file->fi_deleg_file->nf_file;
5067 static int nfsd4_check_conflicting_opens(struct nfs4_client *clp,
5068 struct nfs4_file *fp)
5070 struct nfs4_ol_stateid *st;
5071 struct file *f = fp->fi_deleg_file->nf_file;
5072 struct inode *ino = locks_inode(f);
5075 writes = atomic_read(&ino->i_writecount);
5079 * There could be multiple filehandles (hence multiple
5080 * nfs4_files) referencing this file, but that's not too
5081 * common; let's just give up in that case rather than
5082 * trying to go look up all the clients using that other
5083 * nfs4_file as well:
5088 * If there's a close in progress, make sure that we see it
5089 * clear any fi_fds[] entries before we see it decrement
5092 smp_mb__after_atomic();
5094 if (fp->fi_fds[O_WRONLY])
5096 if (fp->fi_fds[O_RDWR])
5099 return -EAGAIN; /* There may be non-NFSv4 writers */
5101 * It's possible there are non-NFSv4 write opens in progress,
5102 * but if they haven't incremented i_writecount yet then they
5103 * also haven't called break lease yet; so, they'll break this
5104 * lease soon enough. So, all that's left to check for is NFSv4
5107 spin_lock(&fp->fi_lock);
5108 list_for_each_entry(st, &fp->fi_stateids, st_perfile) {
5109 if (st->st_openstp == NULL /* it's an open */ &&
5110 access_permit_write(st) &&
5111 st->st_stid.sc_client != clp) {
5112 spin_unlock(&fp->fi_lock);
5116 spin_unlock(&fp->fi_lock);
5118 * There's a small chance that we could be racing with another
5119 * NFSv4 open. However, any open that hasn't added itself to
5120 * the fi_stateids list also hasn't called break_lease yet; so,
5121 * they'll break this lease soon enough.
5126 static struct nfs4_delegation *
5127 nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
5128 struct nfs4_file *fp, struct nfs4_clnt_odstate *odstate)
5131 struct nfs4_delegation *dp;
5132 struct nfsd_file *nf;
5133 struct file_lock *fl;
5136 * The fi_had_conflict and nfs_get_existing_delegation checks
5137 * here are just optimizations; we'll need to recheck them at
5140 if (fp->fi_had_conflict)
5141 return ERR_PTR(-EAGAIN);
5143 nf = find_readable_file(fp);
5146 * We probably could attempt another open and get a read
5147 * delegation, but for now, don't bother until the
5148 * client actually sends us one.
5150 return ERR_PTR(-EAGAIN);
5152 spin_lock(&state_lock);
5153 spin_lock(&fp->fi_lock);
5154 if (nfs4_delegation_exists(clp, fp))
5156 else if (!fp->fi_deleg_file) {
5157 fp->fi_deleg_file = nf;
5158 /* increment early to prevent fi_deleg_file from being
5160 fp->fi_delegees = 1;
5164 spin_unlock(&fp->fi_lock);
5165 spin_unlock(&state_lock);
5169 return ERR_PTR(status);
5172 dp = alloc_init_deleg(clp, fp, fh, odstate);
5176 fl = nfs4_alloc_init_lease(dp, NFS4_OPEN_DELEGATE_READ);
5178 goto out_clnt_odstate;
5180 status = vfs_setlease(fp->fi_deleg_file->nf_file, fl->fl_type, &fl, NULL);
5182 locks_free_lock(fl);
5184 goto out_clnt_odstate;
5185 status = nfsd4_check_conflicting_opens(clp, fp);
5189 spin_lock(&state_lock);
5190 spin_lock(&fp->fi_lock);
5191 if (fp->fi_had_conflict)
5194 status = hash_delegation_locked(dp, fp);
5195 spin_unlock(&fp->fi_lock);
5196 spin_unlock(&state_lock);
5203 vfs_setlease(fp->fi_deleg_file->nf_file, F_UNLCK, NULL, (void **)&dp);
5205 put_clnt_odstate(dp->dl_clnt_odstate);
5206 nfs4_put_stid(&dp->dl_stid);
5209 return ERR_PTR(status);
5212 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
5214 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5215 if (status == -EAGAIN)
5216 open->op_why_no_deleg = WND4_CONTENTION;
5218 open->op_why_no_deleg = WND4_RESOURCE;
5219 switch (open->op_deleg_want) {
5220 case NFS4_SHARE_WANT_READ_DELEG:
5221 case NFS4_SHARE_WANT_WRITE_DELEG:
5222 case NFS4_SHARE_WANT_ANY_DELEG:
5224 case NFS4_SHARE_WANT_CANCEL:
5225 open->op_why_no_deleg = WND4_CANCELLED;
5227 case NFS4_SHARE_WANT_NO_DELEG:
5234 * Attempt to hand out a delegation.
5236 * Note we don't support write delegations, and won't until the vfs has
5237 * proper support for them.
5240 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
5241 struct nfs4_ol_stateid *stp)
5243 struct nfs4_delegation *dp;
5244 struct nfs4_openowner *oo = openowner(stp->st_stateowner);
5245 struct nfs4_client *clp = stp->st_stid.sc_client;
5249 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
5250 open->op_recall = 0;
5251 switch (open->op_claim_type) {
5252 case NFS4_OPEN_CLAIM_PREVIOUS:
5254 open->op_recall = 1;
5255 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
5258 case NFS4_OPEN_CLAIM_NULL:
5259 case NFS4_OPEN_CLAIM_FH:
5261 * Let's not give out any delegations till everyone's
5262 * had the chance to reclaim theirs, *and* until
5263 * NLM locks have all been reclaimed:
5265 if (locks_in_grace(clp->net))
5267 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
5273 dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file, stp->st_clnt_odstate);
5277 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
5279 trace_nfsd_deleg_read(&dp->dl_stid.sc_stateid);
5280 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
5281 nfs4_put_stid(&dp->dl_stid);
5284 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
5285 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
5286 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
5287 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
5288 open->op_recall = 1;
5291 /* 4.1 client asking for a delegation? */
5292 if (open->op_deleg_want)
5293 nfsd4_open_deleg_none_ext(open, status);
5297 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
5298 struct nfs4_delegation *dp)
5300 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
5301 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5302 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5303 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
5304 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
5305 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5306 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5307 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
5309 /* Otherwise the client must be confused wanting a delegation
5310 * it already has, therefore we don't return
5311 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
5316 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
5318 struct nfsd4_compoundres *resp = rqstp->rq_resp;
5319 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
5320 struct nfs4_file *fp = NULL;
5321 struct nfs4_ol_stateid *stp = NULL;
5322 struct nfs4_delegation *dp = NULL;
5324 bool new_stp = false;
5327 * Lookup file; if found, lookup stateid and check open request,
5328 * and check for delegations in the process of being recalled.
5329 * If not found, create the nfs4_file struct
5331 fp = find_or_add_file(open->op_file, current_fh);
5332 if (fp != open->op_file) {
5333 status = nfs4_check_deleg(cl, open, &dp);
5336 stp = nfsd4_find_and_lock_existing_open(fp, open);
5338 open->op_file = NULL;
5339 status = nfserr_bad_stateid;
5340 if (nfsd4_is_deleg_cur(open))
5345 stp = init_open_stateid(fp, open);
5351 * OPEN the file, or upgrade an existing OPEN.
5352 * If truncate fails, the OPEN fails.
5354 * stp is already locked.
5357 /* Stateid was found, this is an OPEN upgrade */
5358 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
5360 mutex_unlock(&stp->st_mutex);
5364 status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
5366 stp->st_stid.sc_type = NFS4_CLOSED_STID;
5367 release_open_stateid(stp);
5368 mutex_unlock(&stp->st_mutex);
5372 stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
5374 if (stp->st_clnt_odstate == open->op_odstate)
5375 open->op_odstate = NULL;
5378 nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
5379 mutex_unlock(&stp->st_mutex);
5381 if (nfsd4_has_session(&resp->cstate)) {
5382 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
5383 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5384 open->op_why_no_deleg = WND4_NOT_WANTED;
5390 * Attempt to hand out a delegation. No error return, because the
5391 * OPEN succeeds even if we fail.
5393 nfs4_open_delegation(current_fh, open, stp);
5396 trace_nfsd_open(&stp->st_stid.sc_stateid);
5398 /* 4.1 client trying to upgrade/downgrade delegation? */
5399 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
5400 open->op_deleg_want)
5401 nfsd4_deleg_xgrade_none_ext(open, dp);
5405 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
5406 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
5408 * To finish the open response, we just need to set the rflags.
5410 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
5411 if (nfsd4_has_session(&resp->cstate))
5412 open->op_rflags |= NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK;
5413 else if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED))
5414 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
5417 nfs4_put_stid(&dp->dl_stid);
5419 nfs4_put_stid(&stp->st_stid);
5424 void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
5425 struct nfsd4_open *open)
5427 if (open->op_openowner) {
5428 struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
5430 nfsd4_cstate_assign_replay(cstate, so);
5431 nfs4_put_stateowner(so);
5434 kmem_cache_free(file_slab, open->op_file);
5436 nfs4_put_stid(&open->op_stp->st_stid);
5437 if (open->op_odstate)
5438 kmem_cache_free(odstate_slab, open->op_odstate);
5442 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5443 union nfsd4_op_u *u)
5445 clientid_t *clid = &u->renew;
5446 struct nfs4_client *clp;
5448 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5450 trace_nfsd_clid_renew(clid);
5451 status = set_client(clid, cstate, nn);
5455 if (!list_empty(&clp->cl_delegations)
5456 && clp->cl_cb_state != NFSD4_CB_UP)
5457 return nfserr_cb_path_down;
5462 nfsd4_end_grace(struct nfsd_net *nn)
5464 /* do nothing if grace period already ended */
5465 if (nn->grace_ended)
5468 trace_nfsd_grace_complete(nn);
5469 nn->grace_ended = true;
5471 * If the server goes down again right now, an NFSv4
5472 * client will still be allowed to reclaim after it comes back up,
5473 * even if it hasn't yet had a chance to reclaim state this time.
5476 nfsd4_record_grace_done(nn);
5478 * At this point, NFSv4 clients can still reclaim. But if the
5479 * server crashes, any that have not yet reclaimed will be out
5480 * of luck on the next boot.
5482 * (NFSv4.1+ clients are considered to have reclaimed once they
5483 * call RECLAIM_COMPLETE. NFSv4.0 clients are considered to
5484 * have reclaimed after their first OPEN.)
5486 locks_end_grace(&nn->nfsd4_manager);
5488 * At this point, and once lockd and/or any other containers
5489 * exit their grace period, further reclaims will fail and
5490 * regular locking can resume.
5495 * If we've waited a lease period but there are still clients trying to
5496 * reclaim, wait a little longer to give them a chance to finish.
5498 static bool clients_still_reclaiming(struct nfsd_net *nn)
5500 time64_t double_grace_period_end = nn->boot_time +
5501 2 * nn->nfsd4_lease;
5503 if (nn->track_reclaim_completes &&
5504 atomic_read(&nn->nr_reclaim_complete) ==
5505 nn->reclaim_str_hashtbl_size)
5507 if (!nn->somebody_reclaimed)
5509 nn->somebody_reclaimed = false;
5511 * If we've given them *two* lease times to reclaim, and they're
5512 * still not done, give up:
5514 if (ktime_get_boottime_seconds() > double_grace_period_end)
5519 struct laundry_time {
5524 static bool state_expired(struct laundry_time *lt, time64_t last_refresh)
5526 time64_t time_remaining;
5528 if (last_refresh < lt->cutoff)
5530 time_remaining = last_refresh - lt->cutoff;
5531 lt->new_timeo = min(lt->new_timeo, time_remaining);
5535 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
5536 void nfsd4_ssc_init_umount_work(struct nfsd_net *nn)
5538 spin_lock_init(&nn->nfsd_ssc_lock);
5539 INIT_LIST_HEAD(&nn->nfsd_ssc_mount_list);
5540 init_waitqueue_head(&nn->nfsd_ssc_waitq);
5542 EXPORT_SYMBOL_GPL(nfsd4_ssc_init_umount_work);
5545 * This is called when nfsd is being shutdown, after all inter_ssc
5546 * cleanup were done, to destroy the ssc delayed unmount list.
5548 static void nfsd4_ssc_shutdown_umount(struct nfsd_net *nn)
5550 struct nfsd4_ssc_umount_item *ni = NULL;
5551 struct nfsd4_ssc_umount_item *tmp;
5553 spin_lock(&nn->nfsd_ssc_lock);
5554 list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
5555 list_del(&ni->nsui_list);
5556 spin_unlock(&nn->nfsd_ssc_lock);
5557 mntput(ni->nsui_vfsmount);
5559 spin_lock(&nn->nfsd_ssc_lock);
5561 spin_unlock(&nn->nfsd_ssc_lock);
5564 static void nfsd4_ssc_expire_umount(struct nfsd_net *nn)
5566 bool do_wakeup = false;
5567 struct nfsd4_ssc_umount_item *ni = NULL;
5568 struct nfsd4_ssc_umount_item *tmp;
5570 spin_lock(&nn->nfsd_ssc_lock);
5571 list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
5572 if (time_after(jiffies, ni->nsui_expire)) {
5573 if (refcount_read(&ni->nsui_refcnt) > 1)
5576 /* mark being unmount */
5577 ni->nsui_busy = true;
5578 spin_unlock(&nn->nfsd_ssc_lock);
5579 mntput(ni->nsui_vfsmount);
5580 spin_lock(&nn->nfsd_ssc_lock);
5582 /* waiters need to start from begin of list */
5583 list_del(&ni->nsui_list);
5586 /* wakeup ssc_connect waiters */
5593 wake_up_all(&nn->nfsd_ssc_waitq);
5594 spin_unlock(&nn->nfsd_ssc_lock);
5599 nfs4_laundromat(struct nfsd_net *nn)
5601 struct nfs4_client *clp;
5602 struct nfs4_openowner *oo;
5603 struct nfs4_delegation *dp;
5604 struct nfs4_ol_stateid *stp;
5605 struct nfsd4_blocked_lock *nbl;
5606 struct list_head *pos, *next, reaplist;
5607 struct laundry_time lt = {
5608 .cutoff = ktime_get_boottime_seconds() - nn->nfsd4_lease,
5609 .new_timeo = nn->nfsd4_lease
5611 struct nfs4_cpntf_state *cps;
5612 copy_stateid_t *cps_t;
5615 if (clients_still_reclaiming(nn)) {
5619 nfsd4_end_grace(nn);
5620 INIT_LIST_HEAD(&reaplist);
5622 spin_lock(&nn->s2s_cp_lock);
5623 idr_for_each_entry(&nn->s2s_cp_stateids, cps_t, i) {
5624 cps = container_of(cps_t, struct nfs4_cpntf_state, cp_stateid);
5625 if (cps->cp_stateid.sc_type == NFS4_COPYNOTIFY_STID &&
5626 state_expired(<, cps->cpntf_time))
5627 _free_cpntf_state_locked(nn, cps);
5629 spin_unlock(&nn->s2s_cp_lock);
5631 spin_lock(&nn->client_lock);
5632 list_for_each_safe(pos, next, &nn->client_lru) {
5633 clp = list_entry(pos, struct nfs4_client, cl_lru);
5634 if (!state_expired(<, clp->cl_time))
5636 if (mark_client_expired_locked(clp))
5638 list_add(&clp->cl_lru, &reaplist);
5640 spin_unlock(&nn->client_lock);
5641 list_for_each_safe(pos, next, &reaplist) {
5642 clp = list_entry(pos, struct nfs4_client, cl_lru);
5643 trace_nfsd_clid_purged(&clp->cl_clientid);
5644 list_del_init(&clp->cl_lru);
5647 spin_lock(&state_lock);
5648 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5649 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5650 if (!state_expired(<, dp->dl_time))
5652 WARN_ON(!unhash_delegation_locked(dp));
5653 list_add(&dp->dl_recall_lru, &reaplist);
5655 spin_unlock(&state_lock);
5656 while (!list_empty(&reaplist)) {
5657 dp = list_first_entry(&reaplist, struct nfs4_delegation,
5659 list_del_init(&dp->dl_recall_lru);
5660 revoke_delegation(dp);
5663 spin_lock(&nn->client_lock);
5664 while (!list_empty(&nn->close_lru)) {
5665 oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
5667 if (!state_expired(<, oo->oo_time))
5669 list_del_init(&oo->oo_close_lru);
5670 stp = oo->oo_last_closed_stid;
5671 oo->oo_last_closed_stid = NULL;
5672 spin_unlock(&nn->client_lock);
5673 nfs4_put_stid(&stp->st_stid);
5674 spin_lock(&nn->client_lock);
5676 spin_unlock(&nn->client_lock);
5679 * It's possible for a client to try and acquire an already held lock
5680 * that is being held for a long time, and then lose interest in it.
5681 * So, we clean out any un-revisited request after a lease period
5682 * under the assumption that the client is no longer interested.
5684 * RFC5661, sec. 9.6 states that the client must not rely on getting
5685 * notifications and must continue to poll for locks, even when the
5686 * server supports them. Thus this shouldn't lead to clients blocking
5687 * indefinitely once the lock does become free.
5689 BUG_ON(!list_empty(&reaplist));
5690 spin_lock(&nn->blocked_locks_lock);
5691 while (!list_empty(&nn->blocked_locks_lru)) {
5692 nbl = list_first_entry(&nn->blocked_locks_lru,
5693 struct nfsd4_blocked_lock, nbl_lru);
5694 if (!state_expired(<, nbl->nbl_time))
5696 list_move(&nbl->nbl_lru, &reaplist);
5697 list_del_init(&nbl->nbl_list);
5699 spin_unlock(&nn->blocked_locks_lock);
5701 while (!list_empty(&reaplist)) {
5702 nbl = list_first_entry(&reaplist,
5703 struct nfsd4_blocked_lock, nbl_lru);
5704 list_del_init(&nbl->nbl_lru);
5705 free_blocked_lock(nbl);
5707 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
5708 /* service the server-to-server copy delayed unmount list */
5709 nfsd4_ssc_expire_umount(nn);
5712 return max_t(time64_t, lt.new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
5715 static struct workqueue_struct *laundry_wq;
5716 static void laundromat_main(struct work_struct *);
5719 laundromat_main(struct work_struct *laundry)
5722 struct delayed_work *dwork = to_delayed_work(laundry);
5723 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
5726 t = nfs4_laundromat(nn);
5727 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
5730 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid *stp)
5732 if (!fh_match(&fhp->fh_handle, &stp->sc_file->fi_fhandle))
5733 return nfserr_bad_stateid;
5738 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
5740 __be32 status = nfserr_openmode;
5742 /* For lock stateid's, we test the parent open, not the lock: */
5743 if (stp->st_openstp)
5744 stp = stp->st_openstp;
5745 if ((flags & WR_STATE) && !access_permit_write(stp))
5747 if ((flags & RD_STATE) && !access_permit_read(stp))
5754 static inline __be32
5755 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
5757 if (ONE_STATEID(stateid) && (flags & RD_STATE))
5759 else if (opens_in_grace(net)) {
5760 /* Answer in remaining cases depends on existence of
5761 * conflicting state; so we must wait out the grace period. */
5762 return nfserr_grace;
5763 } else if (flags & WR_STATE)
5764 return nfs4_share_conflict(current_fh,
5765 NFS4_SHARE_DENY_WRITE);
5766 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
5767 return nfs4_share_conflict(current_fh,
5768 NFS4_SHARE_DENY_READ);
5771 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
5774 * When sessions are used the stateid generation number is ignored
5777 if (has_session && in->si_generation == 0)
5780 if (in->si_generation == ref->si_generation)
5783 /* If the client sends us a stateid from the future, it's buggy: */
5784 if (nfsd4_stateid_generation_after(in, ref))
5785 return nfserr_bad_stateid;
5787 * However, we could see a stateid from the past, even from a
5788 * non-buggy client. For example, if the client sends a lock
5789 * while some IO is outstanding, the lock may bump si_generation
5790 * while the IO is still in flight. The client could avoid that
5791 * situation by waiting for responses on all the IO requests,
5792 * but better performance may result in retrying IO that
5793 * receives an old_stateid error if requests are rarely
5794 * reordered in flight:
5796 return nfserr_old_stateid;
5799 static __be32 nfsd4_stid_check_stateid_generation(stateid_t *in, struct nfs4_stid *s, bool has_session)
5803 spin_lock(&s->sc_lock);
5804 ret = nfsd4_verify_open_stid(s);
5806 ret = check_stateid_generation(in, &s->sc_stateid, has_session);
5807 spin_unlock(&s->sc_lock);
5811 static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
5813 if (ols->st_stateowner->so_is_open_owner &&
5814 !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
5815 return nfserr_bad_stateid;
5819 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
5821 struct nfs4_stid *s;
5822 __be32 status = nfserr_bad_stateid;
5824 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
5825 CLOSE_STATEID(stateid))
5827 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid))
5829 spin_lock(&cl->cl_lock);
5830 s = find_stateid_locked(cl, stateid);
5833 status = nfsd4_stid_check_stateid_generation(stateid, s, 1);
5836 switch (s->sc_type) {
5837 case NFS4_DELEG_STID:
5840 case NFS4_REVOKED_DELEG_STID:
5841 status = nfserr_deleg_revoked;
5843 case NFS4_OPEN_STID:
5844 case NFS4_LOCK_STID:
5845 status = nfsd4_check_openowner_confirmed(openlockstateid(s));
5848 printk("unknown stateid type %x\n", s->sc_type);
5850 case NFS4_CLOSED_STID:
5851 case NFS4_CLOSED_DELEG_STID:
5852 status = nfserr_bad_stateid;
5855 spin_unlock(&cl->cl_lock);
5860 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
5861 stateid_t *stateid, unsigned char typemask,
5862 struct nfs4_stid **s, struct nfsd_net *nn)
5865 bool return_revoked = false;
5868 * only return revoked delegations if explicitly asked.
5869 * otherwise we report revoked or bad_stateid status.
5871 if (typemask & NFS4_REVOKED_DELEG_STID)
5872 return_revoked = true;
5873 else if (typemask & NFS4_DELEG_STID)
5874 typemask |= NFS4_REVOKED_DELEG_STID;
5876 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
5877 CLOSE_STATEID(stateid))
5878 return nfserr_bad_stateid;
5879 status = set_client(&stateid->si_opaque.so_clid, cstate, nn);
5880 if (status == nfserr_stale_clientid) {
5881 if (cstate->session)
5882 return nfserr_bad_stateid;
5883 return nfserr_stale_stateid;
5887 *s = find_stateid_by_type(cstate->clp, stateid, typemask);
5889 return nfserr_bad_stateid;
5890 if (((*s)->sc_type == NFS4_REVOKED_DELEG_STID) && !return_revoked) {
5892 if (cstate->minorversion)
5893 return nfserr_deleg_revoked;
5894 return nfserr_bad_stateid;
5899 static struct nfsd_file *
5900 nfs4_find_file(struct nfs4_stid *s, int flags)
5905 switch (s->sc_type) {
5906 case NFS4_DELEG_STID:
5907 if (WARN_ON_ONCE(!s->sc_file->fi_deleg_file))
5909 return nfsd_file_get(s->sc_file->fi_deleg_file);
5910 case NFS4_OPEN_STID:
5911 case NFS4_LOCK_STID:
5912 if (flags & RD_STATE)
5913 return find_readable_file(s->sc_file);
5915 return find_writeable_file(s->sc_file);
5922 nfs4_check_olstateid(struct nfs4_ol_stateid *ols, int flags)
5926 status = nfsd4_check_openowner_confirmed(ols);
5929 return nfs4_check_openmode(ols, flags);
5933 nfs4_check_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct nfs4_stid *s,
5934 struct nfsd_file **nfp, int flags)
5936 int acc = (flags & RD_STATE) ? NFSD_MAY_READ : NFSD_MAY_WRITE;
5937 struct nfsd_file *nf;
5940 nf = nfs4_find_file(s, flags);
5942 status = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
5943 acc | NFSD_MAY_OWNER_OVERRIDE);
5949 status = nfsd_file_acquire(rqstp, fhp, acc, &nf);
5958 _free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
5960 WARN_ON_ONCE(cps->cp_stateid.sc_type != NFS4_COPYNOTIFY_STID);
5961 if (!refcount_dec_and_test(&cps->cp_stateid.sc_count))
5963 list_del(&cps->cp_list);
5964 idr_remove(&nn->s2s_cp_stateids,
5965 cps->cp_stateid.stid.si_opaque.so_id);
5969 * A READ from an inter server to server COPY will have a
5970 * copy stateid. Look up the copy notify stateid from the
5971 * idr structure and take a reference on it.
5973 __be32 manage_cpntf_state(struct nfsd_net *nn, stateid_t *st,
5974 struct nfs4_client *clp,
5975 struct nfs4_cpntf_state **cps)
5977 copy_stateid_t *cps_t;
5978 struct nfs4_cpntf_state *state = NULL;
5980 if (st->si_opaque.so_clid.cl_id != nn->s2s_cp_cl_id)
5981 return nfserr_bad_stateid;
5982 spin_lock(&nn->s2s_cp_lock);
5983 cps_t = idr_find(&nn->s2s_cp_stateids, st->si_opaque.so_id);
5985 state = container_of(cps_t, struct nfs4_cpntf_state,
5987 if (state->cp_stateid.sc_type != NFS4_COPYNOTIFY_STID) {
5992 refcount_inc(&state->cp_stateid.sc_count);
5994 _free_cpntf_state_locked(nn, state);
5997 spin_unlock(&nn->s2s_cp_lock);
5999 return nfserr_bad_stateid;
6005 static __be32 find_cpntf_state(struct nfsd_net *nn, stateid_t *st,
6006 struct nfs4_stid **stid)
6009 struct nfs4_cpntf_state *cps = NULL;
6010 struct nfs4_client *found;
6012 status = manage_cpntf_state(nn, st, NULL, &cps);
6016 cps->cpntf_time = ktime_get_boottime_seconds();
6018 status = nfserr_expired;
6019 found = lookup_clientid(&cps->cp_p_clid, true, nn);
6023 *stid = find_stateid_by_type(found, &cps->cp_p_stateid,
6024 NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID);
6028 status = nfserr_bad_stateid;
6030 put_client_renew(found);
6032 nfs4_put_cpntf_state(nn, cps);
6036 void nfs4_put_cpntf_state(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
6038 spin_lock(&nn->s2s_cp_lock);
6039 _free_cpntf_state_locked(nn, cps);
6040 spin_unlock(&nn->s2s_cp_lock);
6044 * Checks for stateid operations
6047 nfs4_preprocess_stateid_op(struct svc_rqst *rqstp,
6048 struct nfsd4_compound_state *cstate, struct svc_fh *fhp,
6049 stateid_t *stateid, int flags, struct nfsd_file **nfp,
6050 struct nfs4_stid **cstid)
6052 struct net *net = SVC_NET(rqstp);
6053 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6054 struct nfs4_stid *s = NULL;
6060 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
6062 status = nfserr_bad_stateid;
6064 status = check_special_stateids(net, fhp, stateid,
6069 status = nfsd4_lookup_stateid(cstate, stateid,
6070 NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
6072 if (status == nfserr_bad_stateid)
6073 status = find_cpntf_state(nn, stateid, &s);
6076 status = nfsd4_stid_check_stateid_generation(stateid, s,
6077 nfsd4_has_session(cstate));
6081 switch (s->sc_type) {
6082 case NFS4_DELEG_STID:
6083 status = nfs4_check_delegmode(delegstateid(s), flags);
6085 case NFS4_OPEN_STID:
6086 case NFS4_LOCK_STID:
6087 status = nfs4_check_olstateid(openlockstateid(s), flags);
6090 status = nfserr_bad_stateid;
6095 status = nfs4_check_fh(fhp, s);
6098 if (status == nfs_ok && nfp)
6099 status = nfs4_check_file(rqstp, fhp, s, nfp, flags);
6102 if (!status && cstid)
6111 * Test if the stateid is valid
6114 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6115 union nfsd4_op_u *u)
6117 struct nfsd4_test_stateid *test_stateid = &u->test_stateid;
6118 struct nfsd4_test_stateid_id *stateid;
6119 struct nfs4_client *cl = cstate->clp;
6121 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
6122 stateid->ts_id_status =
6123 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
6129 nfsd4_free_lock_stateid(stateid_t *stateid, struct nfs4_stid *s)
6131 struct nfs4_ol_stateid *stp = openlockstateid(s);
6134 ret = nfsd4_lock_ol_stateid(stp);
6138 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
6142 ret = nfserr_locks_held;
6143 if (check_for_locks(stp->st_stid.sc_file,
6144 lockowner(stp->st_stateowner)))
6147 release_lock_stateid(stp);
6151 mutex_unlock(&stp->st_mutex);
6158 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6159 union nfsd4_op_u *u)
6161 struct nfsd4_free_stateid *free_stateid = &u->free_stateid;
6162 stateid_t *stateid = &free_stateid->fr_stateid;
6163 struct nfs4_stid *s;
6164 struct nfs4_delegation *dp;
6165 struct nfs4_client *cl = cstate->clp;
6166 __be32 ret = nfserr_bad_stateid;
6168 spin_lock(&cl->cl_lock);
6169 s = find_stateid_locked(cl, stateid);
6172 spin_lock(&s->sc_lock);
6173 switch (s->sc_type) {
6174 case NFS4_DELEG_STID:
6175 ret = nfserr_locks_held;
6177 case NFS4_OPEN_STID:
6178 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
6181 ret = nfserr_locks_held;
6183 case NFS4_LOCK_STID:
6184 spin_unlock(&s->sc_lock);
6185 refcount_inc(&s->sc_count);
6186 spin_unlock(&cl->cl_lock);
6187 ret = nfsd4_free_lock_stateid(stateid, s);
6189 case NFS4_REVOKED_DELEG_STID:
6190 spin_unlock(&s->sc_lock);
6191 dp = delegstateid(s);
6192 list_del_init(&dp->dl_recall_lru);
6193 spin_unlock(&cl->cl_lock);
6197 /* Default falls through and returns nfserr_bad_stateid */
6199 spin_unlock(&s->sc_lock);
6201 spin_unlock(&cl->cl_lock);
6209 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
6210 RD_STATE : WR_STATE;
6213 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
6215 struct svc_fh *current_fh = &cstate->current_fh;
6216 struct nfs4_stateowner *sop = stp->st_stateowner;
6219 status = nfsd4_check_seqid(cstate, sop, seqid);
6222 status = nfsd4_lock_ol_stateid(stp);
6223 if (status != nfs_ok)
6225 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
6226 if (status == nfs_ok)
6227 status = nfs4_check_fh(current_fh, &stp->st_stid);
6228 if (status != nfs_ok)
6229 mutex_unlock(&stp->st_mutex);
6234 * Checks for sequence id mutating operations.
6237 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6238 stateid_t *stateid, char typemask,
6239 struct nfs4_ol_stateid **stpp,
6240 struct nfsd_net *nn)
6243 struct nfs4_stid *s;
6244 struct nfs4_ol_stateid *stp = NULL;
6246 trace_nfsd_preprocess(seqid, stateid);
6249 status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
6252 stp = openlockstateid(s);
6253 nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
6255 status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
6259 nfs4_put_stid(&stp->st_stid);
6263 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6264 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
6267 struct nfs4_openowner *oo;
6268 struct nfs4_ol_stateid *stp;
6270 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
6271 NFS4_OPEN_STID, &stp, nn);
6274 oo = openowner(stp->st_stateowner);
6275 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
6276 mutex_unlock(&stp->st_mutex);
6277 nfs4_put_stid(&stp->st_stid);
6278 return nfserr_bad_stateid;
6285 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6286 union nfsd4_op_u *u)
6288 struct nfsd4_open_confirm *oc = &u->open_confirm;
6290 struct nfs4_openowner *oo;
6291 struct nfs4_ol_stateid *stp;
6292 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6294 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
6295 cstate->current_fh.fh_dentry);
6297 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
6301 status = nfs4_preprocess_seqid_op(cstate,
6302 oc->oc_seqid, &oc->oc_req_stateid,
6303 NFS4_OPEN_STID, &stp, nn);
6306 oo = openowner(stp->st_stateowner);
6307 status = nfserr_bad_stateid;
6308 if (oo->oo_flags & NFS4_OO_CONFIRMED) {
6309 mutex_unlock(&stp->st_mutex);
6312 oo->oo_flags |= NFS4_OO_CONFIRMED;
6313 nfs4_inc_and_copy_stateid(&oc->oc_resp_stateid, &stp->st_stid);
6314 mutex_unlock(&stp->st_mutex);
6315 trace_nfsd_open_confirm(oc->oc_seqid, &stp->st_stid.sc_stateid);
6316 nfsd4_client_record_create(oo->oo_owner.so_client);
6319 nfs4_put_stid(&stp->st_stid);
6321 nfsd4_bump_seqid(cstate, status);
6325 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
6327 if (!test_access(access, stp))
6329 nfs4_file_put_access(stp->st_stid.sc_file, access);
6330 clear_access(access, stp);
6333 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
6335 switch (to_access) {
6336 case NFS4_SHARE_ACCESS_READ:
6337 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
6338 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6340 case NFS4_SHARE_ACCESS_WRITE:
6341 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
6342 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6344 case NFS4_SHARE_ACCESS_BOTH:
6352 nfsd4_open_downgrade(struct svc_rqst *rqstp,
6353 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
6355 struct nfsd4_open_downgrade *od = &u->open_downgrade;
6357 struct nfs4_ol_stateid *stp;
6358 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6360 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
6361 cstate->current_fh.fh_dentry);
6363 /* We don't yet support WANT bits: */
6364 if (od->od_deleg_want)
6365 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
6368 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
6369 &od->od_stateid, &stp, nn);
6372 status = nfserr_inval;
6373 if (!test_access(od->od_share_access, stp)) {
6374 dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
6375 stp->st_access_bmap, od->od_share_access);
6378 if (!test_deny(od->od_share_deny, stp)) {
6379 dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
6380 stp->st_deny_bmap, od->od_share_deny);
6383 nfs4_stateid_downgrade(stp, od->od_share_access);
6384 reset_union_bmap_deny(od->od_share_deny, stp);
6385 nfs4_inc_and_copy_stateid(&od->od_stateid, &stp->st_stid);
6388 mutex_unlock(&stp->st_mutex);
6389 nfs4_put_stid(&stp->st_stid);
6391 nfsd4_bump_seqid(cstate, status);
6395 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
6397 struct nfs4_client *clp = s->st_stid.sc_client;
6399 LIST_HEAD(reaplist);
6401 spin_lock(&clp->cl_lock);
6402 unhashed = unhash_open_stateid(s, &reaplist);
6404 if (clp->cl_minorversion) {
6406 put_ol_stateid_locked(s, &reaplist);
6407 spin_unlock(&clp->cl_lock);
6408 free_ol_stateid_reaplist(&reaplist);
6410 spin_unlock(&clp->cl_lock);
6411 free_ol_stateid_reaplist(&reaplist);
6413 move_to_close_lru(s, clp->net);
6418 * nfs4_unlock_state() called after encode
6421 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6422 union nfsd4_op_u *u)
6424 struct nfsd4_close *close = &u->close;
6426 struct nfs4_ol_stateid *stp;
6427 struct net *net = SVC_NET(rqstp);
6428 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6430 dprintk("NFSD: nfsd4_close on file %pd\n",
6431 cstate->current_fh.fh_dentry);
6433 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
6435 NFS4_OPEN_STID|NFS4_CLOSED_STID,
6437 nfsd4_bump_seqid(cstate, status);
6441 stp->st_stid.sc_type = NFS4_CLOSED_STID;
6444 * Technically we don't _really_ have to increment or copy it, since
6445 * it should just be gone after this operation and we clobber the
6446 * copied value below, but we continue to do so here just to ensure
6447 * that racing ops see that there was a state change.
6449 nfs4_inc_and_copy_stateid(&close->cl_stateid, &stp->st_stid);
6451 nfsd4_close_open_stateid(stp);
6452 mutex_unlock(&stp->st_mutex);
6454 /* v4.1+ suggests that we send a special stateid in here, since the
6455 * clients should just ignore this anyway. Since this is not useful
6456 * for v4.0 clients either, we set it to the special close_stateid
6459 * See RFC5661 section 18.2.4, and RFC7530 section 16.2.5
6461 memcpy(&close->cl_stateid, &close_stateid, sizeof(close->cl_stateid));
6463 /* put reference from nfs4_preprocess_seqid_op */
6464 nfs4_put_stid(&stp->st_stid);
6470 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6471 union nfsd4_op_u *u)
6473 struct nfsd4_delegreturn *dr = &u->delegreturn;
6474 struct nfs4_delegation *dp;
6475 stateid_t *stateid = &dr->dr_stateid;
6476 struct nfs4_stid *s;
6478 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6480 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
6483 status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
6486 dp = delegstateid(s);
6487 status = nfsd4_stid_check_stateid_generation(stateid, &dp->dl_stid, nfsd4_has_session(cstate));
6491 destroy_delegation(dp);
6493 nfs4_put_stid(&dp->dl_stid);
6498 /* last octet in a range */
6500 last_byte_offset(u64 start, u64 len)
6506 return end > start ? end - 1: NFS4_MAX_UINT64;
6510 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
6511 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
6512 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
6513 * locking, this prevents us from being completely protocol-compliant. The
6514 * real solution to this problem is to start using unsigned file offsets in
6515 * the VFS, but this is a very deep change!
6518 nfs4_transform_lock_offset(struct file_lock *lock)
6520 if (lock->fl_start < 0)
6521 lock->fl_start = OFFSET_MAX;
6522 if (lock->fl_end < 0)
6523 lock->fl_end = OFFSET_MAX;
6527 nfsd4_fl_get_owner(fl_owner_t owner)
6529 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
6531 nfs4_get_stateowner(&lo->lo_owner);
6536 nfsd4_fl_put_owner(fl_owner_t owner)
6538 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
6541 nfs4_put_stateowner(&lo->lo_owner);
6545 nfsd4_lm_notify(struct file_lock *fl)
6547 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)fl->fl_owner;
6548 struct net *net = lo->lo_owner.so_client->net;
6549 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6550 struct nfsd4_blocked_lock *nbl = container_of(fl,
6551 struct nfsd4_blocked_lock, nbl_lock);
6554 /* An empty list means that something else is going to be using it */
6555 spin_lock(&nn->blocked_locks_lock);
6556 if (!list_empty(&nbl->nbl_list)) {
6557 list_del_init(&nbl->nbl_list);
6558 list_del_init(&nbl->nbl_lru);
6561 spin_unlock(&nn->blocked_locks_lock);
6564 trace_nfsd_cb_notify_lock(lo, nbl);
6565 nfsd4_run_cb(&nbl->nbl_cb);
6569 static const struct lock_manager_operations nfsd_posix_mng_ops = {
6570 .lm_notify = nfsd4_lm_notify,
6571 .lm_get_owner = nfsd4_fl_get_owner,
6572 .lm_put_owner = nfsd4_fl_put_owner,
6576 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
6578 struct nfs4_lockowner *lo;
6580 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
6581 lo = (struct nfs4_lockowner *) fl->fl_owner;
6582 xdr_netobj_dup(&deny->ld_owner, &lo->lo_owner.so_owner,
6584 if (!deny->ld_owner.data)
6585 /* We just don't care that much */
6587 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
6590 deny->ld_owner.len = 0;
6591 deny->ld_owner.data = NULL;
6592 deny->ld_clientid.cl_boot = 0;
6593 deny->ld_clientid.cl_id = 0;
6595 deny->ld_start = fl->fl_start;
6596 deny->ld_length = NFS4_MAX_UINT64;
6597 if (fl->fl_end != NFS4_MAX_UINT64)
6598 deny->ld_length = fl->fl_end - fl->fl_start + 1;
6599 deny->ld_type = NFS4_READ_LT;
6600 if (fl->fl_type != F_RDLCK)
6601 deny->ld_type = NFS4_WRITE_LT;
6604 static struct nfs4_lockowner *
6605 find_lockowner_str_locked(struct nfs4_client *clp, struct xdr_netobj *owner)
6607 unsigned int strhashval = ownerstr_hashval(owner);
6608 struct nfs4_stateowner *so;
6610 lockdep_assert_held(&clp->cl_lock);
6612 list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
6614 if (so->so_is_open_owner)
6616 if (same_owner_str(so, owner))
6617 return lockowner(nfs4_get_stateowner(so));
6622 static struct nfs4_lockowner *
6623 find_lockowner_str(struct nfs4_client *clp, struct xdr_netobj *owner)
6625 struct nfs4_lockowner *lo;
6627 spin_lock(&clp->cl_lock);
6628 lo = find_lockowner_str_locked(clp, owner);
6629 spin_unlock(&clp->cl_lock);
6633 static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
6635 unhash_lockowner_locked(lockowner(sop));
6638 static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
6640 struct nfs4_lockowner *lo = lockowner(sop);
6642 kmem_cache_free(lockowner_slab, lo);
6645 static const struct nfs4_stateowner_operations lockowner_ops = {
6646 .so_unhash = nfs4_unhash_lockowner,
6647 .so_free = nfs4_free_lockowner,
6651 * Alloc a lock owner structure.
6652 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
6655 * strhashval = ownerstr_hashval
6657 static struct nfs4_lockowner *
6658 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
6659 struct nfs4_ol_stateid *open_stp,
6660 struct nfsd4_lock *lock)
6662 struct nfs4_lockowner *lo, *ret;
6664 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
6667 INIT_LIST_HEAD(&lo->lo_blocked);
6668 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
6669 lo->lo_owner.so_is_open_owner = 0;
6670 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
6671 lo->lo_owner.so_ops = &lockowner_ops;
6672 spin_lock(&clp->cl_lock);
6673 ret = find_lockowner_str_locked(clp, &lock->lk_new_owner);
6675 list_add(&lo->lo_owner.so_strhash,
6676 &clp->cl_ownerstr_hashtbl[strhashval]);
6679 nfs4_free_stateowner(&lo->lo_owner);
6681 spin_unlock(&clp->cl_lock);
6685 static struct nfs4_ol_stateid *
6686 find_lock_stateid(const struct nfs4_lockowner *lo,
6687 const struct nfs4_ol_stateid *ost)
6689 struct nfs4_ol_stateid *lst;
6691 lockdep_assert_held(&ost->st_stid.sc_client->cl_lock);
6693 /* If ost is not hashed, ost->st_locks will not be valid */
6694 if (!nfs4_ol_stateid_unhashed(ost))
6695 list_for_each_entry(lst, &ost->st_locks, st_locks) {
6696 if (lst->st_stateowner == &lo->lo_owner) {
6697 refcount_inc(&lst->st_stid.sc_count);
6704 static struct nfs4_ol_stateid *
6705 init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
6706 struct nfs4_file *fp, struct inode *inode,
6707 struct nfs4_ol_stateid *open_stp)
6709 struct nfs4_client *clp = lo->lo_owner.so_client;
6710 struct nfs4_ol_stateid *retstp;
6712 mutex_init(&stp->st_mutex);
6713 mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
6715 spin_lock(&clp->cl_lock);
6716 if (nfs4_ol_stateid_unhashed(open_stp))
6718 retstp = find_lock_stateid(lo, open_stp);
6721 refcount_inc(&stp->st_stid.sc_count);
6722 stp->st_stid.sc_type = NFS4_LOCK_STID;
6723 stp->st_stateowner = nfs4_get_stateowner(&lo->lo_owner);
6725 stp->st_stid.sc_file = fp;
6726 stp->st_access_bmap = 0;
6727 stp->st_deny_bmap = open_stp->st_deny_bmap;
6728 stp->st_openstp = open_stp;
6729 spin_lock(&fp->fi_lock);
6730 list_add(&stp->st_locks, &open_stp->st_locks);
6731 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
6732 list_add(&stp->st_perfile, &fp->fi_stateids);
6733 spin_unlock(&fp->fi_lock);
6734 spin_unlock(&clp->cl_lock);
6737 spin_unlock(&clp->cl_lock);
6738 if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
6739 nfs4_put_stid(&retstp->st_stid);
6742 /* To keep mutex tracking happy */
6743 mutex_unlock(&stp->st_mutex);
6746 spin_unlock(&clp->cl_lock);
6747 mutex_unlock(&stp->st_mutex);
6751 static struct nfs4_ol_stateid *
6752 find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
6753 struct inode *inode, struct nfs4_ol_stateid *ost,
6756 struct nfs4_stid *ns = NULL;
6757 struct nfs4_ol_stateid *lst;
6758 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
6759 struct nfs4_client *clp = oo->oo_owner.so_client;
6762 spin_lock(&clp->cl_lock);
6763 lst = find_lock_stateid(lo, ost);
6764 spin_unlock(&clp->cl_lock);
6766 if (nfsd4_lock_ol_stateid(lst) == nfs_ok)
6768 nfs4_put_stid(&lst->st_stid);
6770 ns = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_lock_stateid);
6774 lst = init_lock_stateid(openlockstateid(ns), lo, fi, inode, ost);
6775 if (lst == openlockstateid(ns))
6784 check_lock_length(u64 offset, u64 length)
6786 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
6787 (length > ~offset)));
6790 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
6792 struct nfs4_file *fp = lock_stp->st_stid.sc_file;
6794 lockdep_assert_held(&fp->fi_lock);
6796 if (test_access(access, lock_stp))
6798 __nfs4_file_get_access(fp, access);
6799 set_access(access, lock_stp);
6803 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
6804 struct nfs4_ol_stateid *ost,
6805 struct nfsd4_lock *lock,
6806 struct nfs4_ol_stateid **plst, bool *new)
6809 struct nfs4_file *fi = ost->st_stid.sc_file;
6810 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
6811 struct nfs4_client *cl = oo->oo_owner.so_client;
6812 struct inode *inode = d_inode(cstate->current_fh.fh_dentry);
6813 struct nfs4_lockowner *lo;
6814 struct nfs4_ol_stateid *lst;
6815 unsigned int strhashval;
6817 lo = find_lockowner_str(cl, &lock->lk_new_owner);
6819 strhashval = ownerstr_hashval(&lock->lk_new_owner);
6820 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
6822 return nfserr_jukebox;
6824 /* with an existing lockowner, seqids must be the same */
6825 status = nfserr_bad_seqid;
6826 if (!cstate->minorversion &&
6827 lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
6831 lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
6833 status = nfserr_jukebox;
6840 nfs4_put_stateowner(&lo->lo_owner);
6848 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6849 union nfsd4_op_u *u)
6851 struct nfsd4_lock *lock = &u->lock;
6852 struct nfs4_openowner *open_sop = NULL;
6853 struct nfs4_lockowner *lock_sop = NULL;
6854 struct nfs4_ol_stateid *lock_stp = NULL;
6855 struct nfs4_ol_stateid *open_stp = NULL;
6856 struct nfs4_file *fp;
6857 struct nfsd_file *nf = NULL;
6858 struct nfsd4_blocked_lock *nbl = NULL;
6859 struct file_lock *file_lock = NULL;
6860 struct file_lock *conflock = NULL;
6865 unsigned char fl_type;
6866 unsigned int fl_flags = FL_POSIX;
6867 struct net *net = SVC_NET(rqstp);
6868 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6870 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
6871 (long long) lock->lk_offset,
6872 (long long) lock->lk_length);
6874 if (check_lock_length(lock->lk_offset, lock->lk_length))
6875 return nfserr_inval;
6877 if ((status = fh_verify(rqstp, &cstate->current_fh,
6878 S_IFREG, NFSD_MAY_LOCK))) {
6879 dprintk("NFSD: nfsd4_lock: permission denied!\n");
6883 if (lock->lk_is_new) {
6884 if (nfsd4_has_session(cstate))
6885 /* See rfc 5661 18.10.3: given clientid is ignored: */
6886 memcpy(&lock->lk_new_clientid,
6887 &cstate->clp->cl_clientid,
6888 sizeof(clientid_t));
6890 /* validate and update open stateid and open seqid */
6891 status = nfs4_preprocess_confirmed_seqid_op(cstate,
6892 lock->lk_new_open_seqid,
6893 &lock->lk_new_open_stateid,
6897 mutex_unlock(&open_stp->st_mutex);
6898 open_sop = openowner(open_stp->st_stateowner);
6899 status = nfserr_bad_stateid;
6900 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
6901 &lock->lk_new_clientid))
6903 status = lookup_or_create_lock_state(cstate, open_stp, lock,
6906 status = nfs4_preprocess_seqid_op(cstate,
6907 lock->lk_old_lock_seqid,
6908 &lock->lk_old_lock_stateid,
6909 NFS4_LOCK_STID, &lock_stp, nn);
6913 lock_sop = lockowner(lock_stp->st_stateowner);
6915 lkflg = setlkflg(lock->lk_type);
6916 status = nfs4_check_openmode(lock_stp, lkflg);
6920 status = nfserr_grace;
6921 if (locks_in_grace(net) && !lock->lk_reclaim)
6923 status = nfserr_no_grace;
6924 if (!locks_in_grace(net) && lock->lk_reclaim)
6927 if (lock->lk_reclaim)
6928 fl_flags |= FL_RECLAIM;
6930 fp = lock_stp->st_stid.sc_file;
6931 switch (lock->lk_type) {
6933 if (nfsd4_has_session(cstate))
6934 fl_flags |= FL_SLEEP;
6937 spin_lock(&fp->fi_lock);
6938 nf = find_readable_file_locked(fp);
6940 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
6941 spin_unlock(&fp->fi_lock);
6944 case NFS4_WRITEW_LT:
6945 if (nfsd4_has_session(cstate))
6946 fl_flags |= FL_SLEEP;
6949 spin_lock(&fp->fi_lock);
6950 nf = find_writeable_file_locked(fp);
6952 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
6953 spin_unlock(&fp->fi_lock);
6957 status = nfserr_inval;
6962 status = nfserr_openmode;
6967 * Most filesystems with their own ->lock operations will block
6968 * the nfsd thread waiting to acquire the lock. That leads to
6969 * deadlocks (we don't want every nfsd thread tied up waiting
6970 * for file locks), so don't attempt blocking lock notifications
6971 * on those filesystems:
6973 if (nf->nf_file->f_op->lock)
6974 fl_flags &= ~FL_SLEEP;
6976 nbl = find_or_allocate_block(lock_sop, &fp->fi_fhandle, nn);
6978 dprintk("NFSD: %s: unable to allocate block!\n", __func__);
6979 status = nfserr_jukebox;
6983 file_lock = &nbl->nbl_lock;
6984 file_lock->fl_type = fl_type;
6985 file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(&lock_sop->lo_owner));
6986 file_lock->fl_pid = current->tgid;
6987 file_lock->fl_file = nf->nf_file;
6988 file_lock->fl_flags = fl_flags;
6989 file_lock->fl_lmops = &nfsd_posix_mng_ops;
6990 file_lock->fl_start = lock->lk_offset;
6991 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
6992 nfs4_transform_lock_offset(file_lock);
6994 conflock = locks_alloc_lock();
6996 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
6997 status = nfserr_jukebox;
7001 if (fl_flags & FL_SLEEP) {
7002 nbl->nbl_time = ktime_get_boottime_seconds();
7003 spin_lock(&nn->blocked_locks_lock);
7004 list_add_tail(&nbl->nbl_list, &lock_sop->lo_blocked);
7005 list_add_tail(&nbl->nbl_lru, &nn->blocked_locks_lru);
7006 kref_get(&nbl->nbl_kref);
7007 spin_unlock(&nn->blocked_locks_lock);
7010 err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, conflock);
7012 case 0: /* success! */
7013 nfs4_inc_and_copy_stateid(&lock->lk_resp_stateid, &lock_stp->st_stid);
7015 if (lock->lk_reclaim)
7016 nn->somebody_reclaimed = true;
7018 case FILE_LOCK_DEFERRED:
7019 kref_put(&nbl->nbl_kref, free_nbl);
7022 case -EAGAIN: /* conflock holds conflicting lock */
7023 status = nfserr_denied;
7024 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
7025 nfs4_set_lock_denied(conflock, &lock->lk_denied);
7028 status = nfserr_deadlock;
7031 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
7032 status = nfserrno(err);
7037 /* dequeue it if we queued it before */
7038 if (fl_flags & FL_SLEEP) {
7039 spin_lock(&nn->blocked_locks_lock);
7040 if (!list_empty(&nbl->nbl_list) &&
7041 !list_empty(&nbl->nbl_lru)) {
7042 list_del_init(&nbl->nbl_list);
7043 list_del_init(&nbl->nbl_lru);
7044 kref_put(&nbl->nbl_kref, free_nbl);
7046 /* nbl can use one of lists to be linked to reaplist */
7047 spin_unlock(&nn->blocked_locks_lock);
7049 free_blocked_lock(nbl);
7054 /* Bump seqid manually if the 4.0 replay owner is openowner */
7055 if (cstate->replay_owner &&
7056 cstate->replay_owner != &lock_sop->lo_owner &&
7057 seqid_mutating_err(ntohl(status)))
7058 lock_sop->lo_owner.so_seqid++;
7061 * If this is a new, never-before-used stateid, and we are
7062 * returning an error, then just go ahead and release it.
7065 release_lock_stateid(lock_stp);
7067 mutex_unlock(&lock_stp->st_mutex);
7069 nfs4_put_stid(&lock_stp->st_stid);
7072 nfs4_put_stid(&open_stp->st_stid);
7073 nfsd4_bump_seqid(cstate, status);
7075 locks_free_lock(conflock);
7080 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
7081 * so we do a temporary open here just to get an open file to pass to
7084 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
7086 struct nfsd_file *nf;
7089 err = nfsd_file_acquire(rqstp, fhp, NFSD_MAY_READ, &nf);
7092 fh_lock(fhp); /* to block new leases till after test_lock: */
7093 err = nfserrno(nfsd_open_break_lease(fhp->fh_dentry->d_inode,
7097 lock->fl_file = nf->nf_file;
7098 err = nfserrno(vfs_test_lock(nf->nf_file, lock));
7099 lock->fl_file = NULL;
7110 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7111 union nfsd4_op_u *u)
7113 struct nfsd4_lockt *lockt = &u->lockt;
7114 struct file_lock *file_lock = NULL;
7115 struct nfs4_lockowner *lo = NULL;
7117 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7119 if (locks_in_grace(SVC_NET(rqstp)))
7120 return nfserr_grace;
7122 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
7123 return nfserr_inval;
7125 if (!nfsd4_has_session(cstate)) {
7126 status = set_client(&lockt->lt_clientid, cstate, nn);
7131 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
7134 file_lock = locks_alloc_lock();
7136 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7137 status = nfserr_jukebox;
7141 switch (lockt->lt_type) {
7144 file_lock->fl_type = F_RDLCK;
7147 case NFS4_WRITEW_LT:
7148 file_lock->fl_type = F_WRLCK;
7151 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
7152 status = nfserr_inval;
7156 lo = find_lockowner_str(cstate->clp, &lockt->lt_owner);
7158 file_lock->fl_owner = (fl_owner_t)lo;
7159 file_lock->fl_pid = current->tgid;
7160 file_lock->fl_flags = FL_POSIX;
7162 file_lock->fl_start = lockt->lt_offset;
7163 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
7165 nfs4_transform_lock_offset(file_lock);
7167 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
7171 if (file_lock->fl_type != F_UNLCK) {
7172 status = nfserr_denied;
7173 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
7177 nfs4_put_stateowner(&lo->lo_owner);
7179 locks_free_lock(file_lock);
7184 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7185 union nfsd4_op_u *u)
7187 struct nfsd4_locku *locku = &u->locku;
7188 struct nfs4_ol_stateid *stp;
7189 struct nfsd_file *nf = NULL;
7190 struct file_lock *file_lock = NULL;
7193 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7195 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
7196 (long long) locku->lu_offset,
7197 (long long) locku->lu_length);
7199 if (check_lock_length(locku->lu_offset, locku->lu_length))
7200 return nfserr_inval;
7202 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
7203 &locku->lu_stateid, NFS4_LOCK_STID,
7207 nf = find_any_file(stp->st_stid.sc_file);
7209 status = nfserr_lock_range;
7212 file_lock = locks_alloc_lock();
7214 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7215 status = nfserr_jukebox;
7219 file_lock->fl_type = F_UNLCK;
7220 file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(stp->st_stateowner));
7221 file_lock->fl_pid = current->tgid;
7222 file_lock->fl_file = nf->nf_file;
7223 file_lock->fl_flags = FL_POSIX;
7224 file_lock->fl_lmops = &nfsd_posix_mng_ops;
7225 file_lock->fl_start = locku->lu_offset;
7227 file_lock->fl_end = last_byte_offset(locku->lu_offset,
7229 nfs4_transform_lock_offset(file_lock);
7231 err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, NULL);
7233 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
7236 nfs4_inc_and_copy_stateid(&locku->lu_stateid, &stp->st_stid);
7240 mutex_unlock(&stp->st_mutex);
7241 nfs4_put_stid(&stp->st_stid);
7243 nfsd4_bump_seqid(cstate, status);
7245 locks_free_lock(file_lock);
7249 status = nfserrno(err);
7255 * true: locks held by lockowner
7256 * false: no locks held by lockowner
7259 check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
7261 struct file_lock *fl;
7263 struct nfsd_file *nf = find_any_file(fp);
7264 struct inode *inode;
7265 struct file_lock_context *flctx;
7268 /* Any valid lock stateid should have some sort of access */
7273 inode = locks_inode(nf->nf_file);
7274 flctx = inode->i_flctx;
7276 if (flctx && !list_empty_careful(&flctx->flc_posix)) {
7277 spin_lock(&flctx->flc_lock);
7278 list_for_each_entry(fl, &flctx->flc_posix, fl_list) {
7279 if (fl->fl_owner == (fl_owner_t)lowner) {
7284 spin_unlock(&flctx->flc_lock);
7291 nfsd4_release_lockowner(struct svc_rqst *rqstp,
7292 struct nfsd4_compound_state *cstate,
7293 union nfsd4_op_u *u)
7295 struct nfsd4_release_lockowner *rlockowner = &u->release_lockowner;
7296 clientid_t *clid = &rlockowner->rl_clientid;
7297 struct nfs4_stateowner *sop;
7298 struct nfs4_lockowner *lo = NULL;
7299 struct nfs4_ol_stateid *stp;
7300 struct xdr_netobj *owner = &rlockowner->rl_owner;
7301 unsigned int hashval = ownerstr_hashval(owner);
7303 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7304 struct nfs4_client *clp;
7305 LIST_HEAD (reaplist);
7307 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
7308 clid->cl_boot, clid->cl_id);
7310 status = set_client(clid, cstate, nn);
7315 /* Find the matching lock stateowner */
7316 spin_lock(&clp->cl_lock);
7317 list_for_each_entry(sop, &clp->cl_ownerstr_hashtbl[hashval],
7320 if (sop->so_is_open_owner || !same_owner_str(sop, owner))
7323 /* see if there are still any locks associated with it */
7324 lo = lockowner(sop);
7325 list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
7326 if (check_for_locks(stp->st_stid.sc_file, lo)) {
7327 status = nfserr_locks_held;
7328 spin_unlock(&clp->cl_lock);
7333 nfs4_get_stateowner(sop);
7337 spin_unlock(&clp->cl_lock);
7341 unhash_lockowner_locked(lo);
7342 while (!list_empty(&lo->lo_owner.so_stateids)) {
7343 stp = list_first_entry(&lo->lo_owner.so_stateids,
7344 struct nfs4_ol_stateid,
7346 WARN_ON(!unhash_lock_stateid(stp));
7347 put_ol_stateid_locked(stp, &reaplist);
7349 spin_unlock(&clp->cl_lock);
7350 free_ol_stateid_reaplist(&reaplist);
7351 remove_blocked_locks(lo);
7352 nfs4_put_stateowner(&lo->lo_owner);
7357 static inline struct nfs4_client_reclaim *
7360 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
7364 nfs4_has_reclaimed_state(struct xdr_netobj name, struct nfsd_net *nn)
7366 struct nfs4_client_reclaim *crp;
7368 crp = nfsd4_find_reclaim_client(name, nn);
7369 return (crp && crp->cr_clp);
7373 * failure => all reset bets are off, nfserr_no_grace...
7375 * The caller is responsible for freeing name.data if NULL is returned (it
7376 * will be freed in nfs4_remove_reclaim_record in the normal case).
7378 struct nfs4_client_reclaim *
7379 nfs4_client_to_reclaim(struct xdr_netobj name, struct xdr_netobj princhash,
7380 struct nfsd_net *nn)
7382 unsigned int strhashval;
7383 struct nfs4_client_reclaim *crp;
7385 crp = alloc_reclaim();
7387 strhashval = clientstr_hashval(name);
7388 INIT_LIST_HEAD(&crp->cr_strhash);
7389 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
7390 crp->cr_name.data = name.data;
7391 crp->cr_name.len = name.len;
7392 crp->cr_princhash.data = princhash.data;
7393 crp->cr_princhash.len = princhash.len;
7395 nn->reclaim_str_hashtbl_size++;
7401 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
7403 list_del(&crp->cr_strhash);
7404 kfree(crp->cr_name.data);
7405 kfree(crp->cr_princhash.data);
7407 nn->reclaim_str_hashtbl_size--;
7411 nfs4_release_reclaim(struct nfsd_net *nn)
7413 struct nfs4_client_reclaim *crp = NULL;
7416 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7417 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
7418 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
7419 struct nfs4_client_reclaim, cr_strhash);
7420 nfs4_remove_reclaim_record(crp, nn);
7423 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
7427 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
7428 struct nfs4_client_reclaim *
7429 nfsd4_find_reclaim_client(struct xdr_netobj name, struct nfsd_net *nn)
7431 unsigned int strhashval;
7432 struct nfs4_client_reclaim *crp = NULL;
7434 strhashval = clientstr_hashval(name);
7435 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
7436 if (compare_blob(&crp->cr_name, &name) == 0) {
7444 nfs4_check_open_reclaim(struct nfs4_client *clp)
7446 if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
7447 return nfserr_no_grace;
7449 if (nfsd4_client_record_check(clp))
7450 return nfserr_reclaim_bad;
7456 * Since the lifetime of a delegation isn't limited to that of an open, a
7457 * client may quite reasonably hang on to a delegation as long as it has
7458 * the inode cached. This becomes an obvious problem the first time a
7459 * client's inode cache approaches the size of the server's total memory.
7461 * For now we avoid this problem by imposing a hard limit on the number
7462 * of delegations, which varies according to the server's memory size.
7465 set_max_delegations(void)
7468 * Allow at most 4 delegations per megabyte of RAM. Quick
7469 * estimates suggest that in the worst case (where every delegation
7470 * is for a different inode), a delegation could take about 1.5K,
7471 * giving a worst case usage of about 6% of memory.
7473 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
7476 static int nfs4_state_create_net(struct net *net)
7478 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7481 nn->conf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
7482 sizeof(struct list_head),
7484 if (!nn->conf_id_hashtbl)
7486 nn->unconf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
7487 sizeof(struct list_head),
7489 if (!nn->unconf_id_hashtbl)
7491 nn->sessionid_hashtbl = kmalloc_array(SESSION_HASH_SIZE,
7492 sizeof(struct list_head),
7494 if (!nn->sessionid_hashtbl)
7497 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7498 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
7499 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
7501 for (i = 0; i < SESSION_HASH_SIZE; i++)
7502 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
7503 nn->conf_name_tree = RB_ROOT;
7504 nn->unconf_name_tree = RB_ROOT;
7505 nn->boot_time = ktime_get_real_seconds();
7506 nn->grace_ended = false;
7507 nn->nfsd4_manager.block_opens = true;
7508 INIT_LIST_HEAD(&nn->nfsd4_manager.list);
7509 INIT_LIST_HEAD(&nn->client_lru);
7510 INIT_LIST_HEAD(&nn->close_lru);
7511 INIT_LIST_HEAD(&nn->del_recall_lru);
7512 spin_lock_init(&nn->client_lock);
7513 spin_lock_init(&nn->s2s_cp_lock);
7514 idr_init(&nn->s2s_cp_stateids);
7516 spin_lock_init(&nn->blocked_locks_lock);
7517 INIT_LIST_HEAD(&nn->blocked_locks_lru);
7519 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
7525 kfree(nn->unconf_id_hashtbl);
7527 kfree(nn->conf_id_hashtbl);
7533 nfs4_state_destroy_net(struct net *net)
7536 struct nfs4_client *clp = NULL;
7537 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7539 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7540 while (!list_empty(&nn->conf_id_hashtbl[i])) {
7541 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
7542 destroy_client(clp);
7546 WARN_ON(!list_empty(&nn->blocked_locks_lru));
7548 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7549 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
7550 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
7551 destroy_client(clp);
7555 kfree(nn->sessionid_hashtbl);
7556 kfree(nn->unconf_id_hashtbl);
7557 kfree(nn->conf_id_hashtbl);
7562 nfs4_state_start_net(struct net *net)
7564 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7567 ret = nfs4_state_create_net(net);
7570 locks_start_grace(net, &nn->nfsd4_manager);
7571 nfsd4_client_tracking_init(net);
7572 if (nn->track_reclaim_completes && nn->reclaim_str_hashtbl_size == 0)
7574 printk(KERN_INFO "NFSD: starting %lld-second grace period (net %x)\n",
7575 nn->nfsd4_grace, net->ns.inum);
7576 trace_nfsd_grace_start(nn);
7577 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
7581 printk(KERN_INFO "NFSD: no clients to reclaim, skipping NFSv4 grace period (net %x)\n",
7583 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_lease * HZ);
7584 nfsd4_end_grace(nn);
7588 /* initialization to perform when the nfsd service is started: */
7591 nfs4_state_start(void)
7595 laundry_wq = alloc_workqueue("%s", WQ_UNBOUND, 0, "nfsd4");
7596 if (laundry_wq == NULL) {
7600 ret = nfsd4_create_callback_queue();
7602 goto out_free_laundry;
7604 set_max_delegations();
7608 destroy_workqueue(laundry_wq);
7614 nfs4_state_shutdown_net(struct net *net)
7616 struct nfs4_delegation *dp = NULL;
7617 struct list_head *pos, *next, reaplist;
7618 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7620 cancel_delayed_work_sync(&nn->laundromat_work);
7621 locks_end_grace(&nn->nfsd4_manager);
7623 INIT_LIST_HEAD(&reaplist);
7624 spin_lock(&state_lock);
7625 list_for_each_safe(pos, next, &nn->del_recall_lru) {
7626 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
7627 WARN_ON(!unhash_delegation_locked(dp));
7628 list_add(&dp->dl_recall_lru, &reaplist);
7630 spin_unlock(&state_lock);
7631 list_for_each_safe(pos, next, &reaplist) {
7632 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
7633 list_del_init(&dp->dl_recall_lru);
7634 destroy_unhashed_deleg(dp);
7637 nfsd4_client_tracking_exit(net);
7638 nfs4_state_destroy_net(net);
7639 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
7640 nfsd4_ssc_shutdown_umount(nn);
7645 nfs4_state_shutdown(void)
7647 destroy_workqueue(laundry_wq);
7648 nfsd4_destroy_callback_queue();
7652 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
7654 if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG) &&
7655 CURRENT_STATEID(stateid))
7656 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
7660 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
7662 if (cstate->minorversion) {
7663 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
7664 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
7669 clear_current_stateid(struct nfsd4_compound_state *cstate)
7671 CLEAR_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
7675 * functions to set current state id
7678 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate,
7679 union nfsd4_op_u *u)
7681 put_stateid(cstate, &u->open_downgrade.od_stateid);
7685 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate,
7686 union nfsd4_op_u *u)
7688 put_stateid(cstate, &u->open.op_stateid);
7692 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate,
7693 union nfsd4_op_u *u)
7695 put_stateid(cstate, &u->close.cl_stateid);
7699 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate,
7700 union nfsd4_op_u *u)
7702 put_stateid(cstate, &u->lock.lk_resp_stateid);
7706 * functions to consume current state id
7710 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate,
7711 union nfsd4_op_u *u)
7713 get_stateid(cstate, &u->open_downgrade.od_stateid);
7717 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate,
7718 union nfsd4_op_u *u)
7720 get_stateid(cstate, &u->delegreturn.dr_stateid);
7724 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate,
7725 union nfsd4_op_u *u)
7727 get_stateid(cstate, &u->free_stateid.fr_stateid);
7731 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate,
7732 union nfsd4_op_u *u)
7734 get_stateid(cstate, &u->setattr.sa_stateid);
7738 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate,
7739 union nfsd4_op_u *u)
7741 get_stateid(cstate, &u->close.cl_stateid);
7745 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate,
7746 union nfsd4_op_u *u)
7748 get_stateid(cstate, &u->locku.lu_stateid);
7752 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate,
7753 union nfsd4_op_u *u)
7755 get_stateid(cstate, &u->read.rd_stateid);
7759 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate,
7760 union nfsd4_op_u *u)
7762 get_stateid(cstate, &u->write.wr_stateid);