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>
47 #include "current_stateid.h"
51 #define NFSDDBG_FACILITY NFSDDBG_PROC
53 #define all_ones {{~0,~0},~0}
54 static const stateid_t one_stateid = {
56 .si_opaque = all_ones,
58 static const stateid_t zero_stateid = {
61 static const stateid_t currentstateid = {
65 static u64 current_sessionid = 1;
67 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
68 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
69 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
71 /* forward declarations */
72 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
76 /* Currently used for almost all code touching nfsv4 state: */
77 static DEFINE_MUTEX(client_mutex);
80 * Currently used for the del_recall_lru and file hash table. In an
81 * effort to decrease the scope of the client_mutex, this spinlock may
82 * eventually cover more:
84 static DEFINE_SPINLOCK(recall_lock);
86 static struct kmem_cache *openowner_slab = NULL;
87 static struct kmem_cache *lockowner_slab = NULL;
88 static struct kmem_cache *file_slab = NULL;
89 static struct kmem_cache *stateid_slab = NULL;
90 static struct kmem_cache *deleg_slab = NULL;
95 mutex_lock(&client_mutex);
98 static void free_session(struct nfsd4_session *);
100 static bool is_session_dead(struct nfsd4_session *ses)
102 return ses->se_flags & NFS4_SESSION_DEAD;
105 void nfsd4_put_session(struct nfsd4_session *ses)
107 if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
111 static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
113 if (atomic_read(&ses->se_ref) > ref_held_by_me)
114 return nfserr_jukebox;
115 ses->se_flags |= NFS4_SESSION_DEAD;
119 static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
121 if (is_session_dead(ses))
122 return nfserr_badsession;
123 atomic_inc(&ses->se_ref);
128 nfs4_unlock_state(void)
130 mutex_unlock(&client_mutex);
133 static bool is_client_expired(struct nfs4_client *clp)
135 return clp->cl_time == 0;
138 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
140 if (atomic_read(&clp->cl_refcount))
141 return nfserr_jukebox;
146 static __be32 mark_client_expired(struct nfs4_client *clp)
148 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
151 spin_lock(&nn->client_lock);
152 ret = mark_client_expired_locked(clp);
153 spin_unlock(&nn->client_lock);
157 static __be32 get_client_locked(struct nfs4_client *clp)
159 if (is_client_expired(clp))
160 return nfserr_expired;
161 atomic_inc(&clp->cl_refcount);
165 /* must be called under the client_lock */
167 renew_client_locked(struct nfs4_client *clp)
169 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
171 if (is_client_expired(clp)) {
173 printk("%s: client (clientid %08x/%08x) already expired\n",
175 clp->cl_clientid.cl_boot,
176 clp->cl_clientid.cl_id);
180 dprintk("renewing client (clientid %08x/%08x)\n",
181 clp->cl_clientid.cl_boot,
182 clp->cl_clientid.cl_id);
183 list_move_tail(&clp->cl_lru, &nn->client_lru);
184 clp->cl_time = get_seconds();
188 renew_client(struct nfs4_client *clp)
190 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
192 spin_lock(&nn->client_lock);
193 renew_client_locked(clp);
194 spin_unlock(&nn->client_lock);
197 static void put_client_renew_locked(struct nfs4_client *clp)
199 if (!atomic_dec_and_test(&clp->cl_refcount))
201 if (!is_client_expired(clp))
202 renew_client_locked(clp);
205 void put_client_renew(struct nfs4_client *clp)
207 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
209 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
211 if (!is_client_expired(clp))
212 renew_client_locked(clp);
213 spin_unlock(&nn->client_lock);
218 opaque_hashval(const void *ptr, int nbytes)
220 unsigned char *cptr = (unsigned char *) ptr;
230 static void nfsd4_free_file(struct nfs4_file *f)
232 kmem_cache_free(file_slab, f);
236 put_nfs4_file(struct nfs4_file *fi)
238 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
239 hlist_del(&fi->fi_hash);
240 spin_unlock(&recall_lock);
247 get_nfs4_file(struct nfs4_file *fi)
249 atomic_inc(&fi->fi_ref);
252 static int num_delegations;
253 unsigned long max_delegations;
256 * Open owner state (share locks)
259 /* hash tables for lock and open owners */
260 #define OWNER_HASH_BITS 8
261 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
262 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
264 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
268 ret = opaque_hashval(ownername->data, ownername->len);
270 return ret & OWNER_HASH_MASK;
273 /* hash table for nfs4_file */
274 #define FILE_HASH_BITS 8
275 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
277 static unsigned int file_hashval(struct inode *ino)
279 /* XXX: why are we hashing on inode pointer, anyway? */
280 return hash_ptr(ino, FILE_HASH_BITS);
283 static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
285 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
287 WARN_ON_ONCE(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
288 atomic_inc(&fp->fi_access[oflag]);
291 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
293 if (oflag == O_RDWR) {
294 __nfs4_file_get_access(fp, O_RDONLY);
295 __nfs4_file_get_access(fp, O_WRONLY);
297 __nfs4_file_get_access(fp, oflag);
300 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
302 if (fp->fi_fds[oflag]) {
303 fput(fp->fi_fds[oflag]);
304 fp->fi_fds[oflag] = NULL;
308 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
310 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
311 nfs4_file_put_fd(fp, oflag);
312 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
313 nfs4_file_put_fd(fp, O_RDWR);
317 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
319 if (oflag == O_RDWR) {
320 __nfs4_file_put_access(fp, O_RDONLY);
321 __nfs4_file_put_access(fp, O_WRONLY);
323 __nfs4_file_put_access(fp, oflag);
326 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
329 struct idr *stateids = &cl->cl_stateids;
330 struct nfs4_stid *stid;
333 stid = kmem_cache_alloc(slab, GFP_KERNEL);
337 new_id = idr_alloc_cyclic(stateids, stid, 0, 0, GFP_KERNEL);
340 stid->sc_client = cl;
342 stid->sc_stateid.si_opaque.so_id = new_id;
343 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
344 /* Will be incremented before return to client: */
345 stid->sc_stateid.si_generation = 0;
348 * It shouldn't be a problem to reuse an opaque stateid value.
349 * I don't think it is for 4.1. But with 4.0 I worry that, for
350 * example, a stray write retransmission could be accepted by
351 * the server when it should have been rejected. Therefore,
352 * adopt a trick from the sctp code to attempt to maximize the
353 * amount of time until an id is reused, by ensuring they always
354 * "increase" (mod INT_MAX):
358 kmem_cache_free(slab, stid);
362 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
364 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
367 static struct nfs4_delegation *
368 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh)
370 struct nfs4_delegation *dp;
372 dprintk("NFSD alloc_init_deleg\n");
373 if (num_delegations > max_delegations)
375 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
378 dp->dl_stid.sc_type = NFS4_DELEG_STID;
380 * delegation seqid's are never incremented. The 4.1 special
381 * meaning of seqid 0 isn't meaningful, really, but let's avoid
382 * 0 anyway just for consistency and use 1:
384 dp->dl_stid.sc_stateid.si_generation = 1;
386 INIT_LIST_HEAD(&dp->dl_perfile);
387 INIT_LIST_HEAD(&dp->dl_perclnt);
388 INIT_LIST_HEAD(&dp->dl_recall_lru);
390 dp->dl_type = NFS4_OPEN_DELEGATE_READ;
391 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
393 atomic_set(&dp->dl_count, 1);
394 nfsd4_init_callback(&dp->dl_recall);
398 static void remove_stid(struct nfs4_stid *s)
400 struct idr *stateids = &s->sc_client->cl_stateids;
402 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
405 static void nfs4_free_stid(struct kmem_cache *slab, struct nfs4_stid *s)
407 kmem_cache_free(slab, s);
411 nfs4_put_delegation(struct nfs4_delegation *dp)
413 if (atomic_dec_and_test(&dp->dl_count)) {
414 nfs4_free_stid(deleg_slab, &dp->dl_stid);
419 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
421 if (atomic_dec_and_test(&fp->fi_delegees)) {
422 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
424 fput(fp->fi_deleg_file);
425 fp->fi_deleg_file = NULL;
429 static void unhash_stid(struct nfs4_stid *s)
434 /* Called under the state lock. */
436 unhash_delegation(struct nfs4_delegation *dp)
438 list_del_init(&dp->dl_perclnt);
439 spin_lock(&recall_lock);
440 list_del_init(&dp->dl_perfile);
441 list_del_init(&dp->dl_recall_lru);
442 spin_unlock(&recall_lock);
443 nfs4_put_deleg_lease(dp->dl_file);
444 put_nfs4_file(dp->dl_file);
450 static void destroy_revoked_delegation(struct nfs4_delegation *dp)
452 list_del_init(&dp->dl_recall_lru);
453 remove_stid(&dp->dl_stid);
454 nfs4_put_delegation(dp);
457 static void destroy_delegation(struct nfs4_delegation *dp)
459 unhash_delegation(dp);
460 remove_stid(&dp->dl_stid);
461 nfs4_put_delegation(dp);
464 static void revoke_delegation(struct nfs4_delegation *dp)
466 struct nfs4_client *clp = dp->dl_stid.sc_client;
468 if (clp->cl_minorversion == 0)
469 destroy_delegation(dp);
471 unhash_delegation(dp);
472 dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
473 list_add(&dp->dl_recall_lru, &clp->cl_revoked);
481 static unsigned int clientid_hashval(u32 id)
483 return id & CLIENT_HASH_MASK;
486 static unsigned int clientstr_hashval(const char *name)
488 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
492 * We store the NONE, READ, WRITE, and BOTH bits separately in the
493 * st_{access,deny}_bmap field of the stateid, in order to track not
494 * only what share bits are currently in force, but also what
495 * combinations of share bits previous opens have used. This allows us
496 * to enforce the recommendation of rfc 3530 14.2.19 that the server
497 * return an error if the client attempt to downgrade to a combination
498 * of share bits not explicable by closing some of its previous opens.
500 * XXX: This enforcement is actually incomplete, since we don't keep
501 * track of access/deny bit combinations; so, e.g., we allow:
503 * OPEN allow read, deny write
504 * OPEN allow both, deny none
505 * DOWNGRADE allow read, deny none
507 * which we should reject.
510 bmap_to_share_mode(unsigned long bmap) {
512 unsigned int access = 0;
514 for (i = 1; i < 4; i++) {
515 if (test_bit(i, &bmap))
522 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
523 unsigned int access, deny;
525 access = bmap_to_share_mode(stp->st_access_bmap);
526 deny = bmap_to_share_mode(stp->st_deny_bmap);
527 if ((access & open->op_share_deny) || (deny & open->op_share_access))
532 /* set share access for a given stateid */
534 set_access(u32 access, struct nfs4_ol_stateid *stp)
536 __set_bit(access, &stp->st_access_bmap);
539 /* clear share access for a given stateid */
541 clear_access(u32 access, struct nfs4_ol_stateid *stp)
543 __clear_bit(access, &stp->st_access_bmap);
546 /* test whether a given stateid has access */
548 test_access(u32 access, struct nfs4_ol_stateid *stp)
550 return test_bit(access, &stp->st_access_bmap);
553 /* set share deny for a given stateid */
555 set_deny(u32 access, struct nfs4_ol_stateid *stp)
557 __set_bit(access, &stp->st_deny_bmap);
560 /* clear share deny for a given stateid */
562 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
564 __clear_bit(access, &stp->st_deny_bmap);
567 /* test whether a given stateid is denying specific access */
569 test_deny(u32 access, struct nfs4_ol_stateid *stp)
571 return test_bit(access, &stp->st_deny_bmap);
574 static int nfs4_access_to_omode(u32 access)
576 switch (access & NFS4_SHARE_ACCESS_BOTH) {
577 case NFS4_SHARE_ACCESS_READ:
579 case NFS4_SHARE_ACCESS_WRITE:
581 case NFS4_SHARE_ACCESS_BOTH:
588 /* release all access and file references for a given stateid */
590 release_all_access(struct nfs4_ol_stateid *stp)
594 for (i = 1; i < 4; i++) {
595 if (test_access(i, stp))
596 nfs4_file_put_access(stp->st_file,
597 nfs4_access_to_omode(i));
598 clear_access(i, stp);
602 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
604 list_del(&stp->st_perfile);
605 list_del(&stp->st_perstateowner);
608 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
610 release_all_access(stp);
611 put_nfs4_file(stp->st_file);
615 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
617 remove_stid(&stp->st_stid);
618 nfs4_free_stid(stateid_slab, &stp->st_stid);
621 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
625 unhash_generic_stateid(stp);
626 unhash_stid(&stp->st_stid);
627 file = find_any_file(stp->st_file);
629 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
630 close_generic_stateid(stp);
631 free_generic_stateid(stp);
634 static void unhash_lockowner(struct nfs4_lockowner *lo)
636 struct nfs4_ol_stateid *stp;
638 list_del(&lo->lo_owner.so_strhash);
639 list_del(&lo->lo_perstateid);
640 list_del(&lo->lo_owner_ino_hash);
641 while (!list_empty(&lo->lo_owner.so_stateids)) {
642 stp = list_first_entry(&lo->lo_owner.so_stateids,
643 struct nfs4_ol_stateid, st_perstateowner);
644 release_lock_stateid(stp);
648 static void release_lockowner(struct nfs4_lockowner *lo)
650 unhash_lockowner(lo);
651 nfs4_free_lockowner(lo);
655 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
657 struct nfs4_lockowner *lo;
659 while (!list_empty(&open_stp->st_lockowners)) {
660 lo = list_entry(open_stp->st_lockowners.next,
661 struct nfs4_lockowner, lo_perstateid);
662 release_lockowner(lo);
666 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
668 unhash_generic_stateid(stp);
669 release_stateid_lockowners(stp);
670 close_generic_stateid(stp);
673 static void release_open_stateid(struct nfs4_ol_stateid *stp)
675 unhash_open_stateid(stp);
676 free_generic_stateid(stp);
679 static void unhash_openowner(struct nfs4_openowner *oo)
681 struct nfs4_ol_stateid *stp;
683 list_del(&oo->oo_owner.so_strhash);
684 list_del(&oo->oo_perclient);
685 while (!list_empty(&oo->oo_owner.so_stateids)) {
686 stp = list_first_entry(&oo->oo_owner.so_stateids,
687 struct nfs4_ol_stateid, st_perstateowner);
688 release_open_stateid(stp);
692 static void release_last_closed_stateid(struct nfs4_openowner *oo)
694 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
697 free_generic_stateid(s);
698 oo->oo_last_closed_stid = NULL;
702 static void release_openowner(struct nfs4_openowner *oo)
704 unhash_openowner(oo);
705 list_del(&oo->oo_close_lru);
706 release_last_closed_stateid(oo);
707 nfs4_free_openowner(oo);
711 hash_sessionid(struct nfs4_sessionid *sessionid)
713 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
715 return sid->sequence % SESSION_HASH_SIZE;
720 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
722 u32 *ptr = (u32 *)(&sessionid->data[0]);
723 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
727 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
733 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
734 * won't be used for replay.
736 void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
738 struct nfs4_stateowner *so = cstate->replay_owner;
740 if (nfserr == nfserr_replay_me)
743 if (!seqid_mutating_err(ntohl(nfserr))) {
744 cstate->replay_owner = NULL;
749 if (so->so_is_open_owner)
750 release_last_closed_stateid(openowner(so));
756 gen_sessionid(struct nfsd4_session *ses)
758 struct nfs4_client *clp = ses->se_client;
759 struct nfsd4_sessionid *sid;
761 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
762 sid->clientid = clp->cl_clientid;
763 sid->sequence = current_sessionid++;
768 * The protocol defines ca_maxresponssize_cached to include the size of
769 * the rpc header, but all we need to cache is the data starting after
770 * the end of the initial SEQUENCE operation--the rest we regenerate
771 * each time. Therefore we can advertise a ca_maxresponssize_cached
772 * value that is the number of bytes in our cache plus a few additional
773 * bytes. In order to stay on the safe side, and not promise more than
774 * we can cache, those additional bytes must be the minimum possible: 24
775 * bytes of rpc header (xid through accept state, with AUTH_NULL
776 * verifier), 12 for the compound header (with zero-length tag), and 44
777 * for the SEQUENCE op response:
779 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
782 free_session_slots(struct nfsd4_session *ses)
786 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
787 kfree(ses->se_slots[i]);
791 * We don't actually need to cache the rpc and session headers, so we
792 * can allocate a little less for each slot:
794 static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
798 if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
801 size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
802 return size + sizeof(struct nfsd4_slot);
806 * XXX: If we run out of reserved DRC memory we could (up to a point)
807 * re-negotiate active sessions and reduce their slot usage to make
808 * room for new connections. For now we just fail the create session.
810 static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
812 u32 slotsize = slot_bytes(ca);
813 u32 num = ca->maxreqs;
816 spin_lock(&nfsd_drc_lock);
817 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
818 nfsd_drc_max_mem - nfsd_drc_mem_used);
819 num = min_t(int, num, avail / slotsize);
820 nfsd_drc_mem_used += num * slotsize;
821 spin_unlock(&nfsd_drc_lock);
826 static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
828 int slotsize = slot_bytes(ca);
830 spin_lock(&nfsd_drc_lock);
831 nfsd_drc_mem_used -= slotsize * ca->maxreqs;
832 spin_unlock(&nfsd_drc_lock);
835 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
836 struct nfsd4_channel_attrs *battrs)
838 int numslots = fattrs->maxreqs;
839 int slotsize = slot_bytes(fattrs);
840 struct nfsd4_session *new;
843 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
844 + sizeof(struct nfsd4_session) > PAGE_SIZE);
845 mem = numslots * sizeof(struct nfsd4_slot *);
847 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
850 /* allocate each struct nfsd4_slot and data cache in one piece */
851 for (i = 0; i < numslots; i++) {
852 new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
853 if (!new->se_slots[i])
857 memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
858 memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
863 kfree(new->se_slots[i]);
868 static void free_conn(struct nfsd4_conn *c)
870 svc_xprt_put(c->cn_xprt);
874 static void nfsd4_conn_lost(struct svc_xpt_user *u)
876 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
877 struct nfs4_client *clp = c->cn_session->se_client;
879 spin_lock(&clp->cl_lock);
880 if (!list_empty(&c->cn_persession)) {
881 list_del(&c->cn_persession);
884 nfsd4_probe_callback(clp);
885 spin_unlock(&clp->cl_lock);
888 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
890 struct nfsd4_conn *conn;
892 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
895 svc_xprt_get(rqstp->rq_xprt);
896 conn->cn_xprt = rqstp->rq_xprt;
897 conn->cn_flags = flags;
898 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
902 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
904 conn->cn_session = ses;
905 list_add(&conn->cn_persession, &ses->se_conns);
908 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
910 struct nfs4_client *clp = ses->se_client;
912 spin_lock(&clp->cl_lock);
913 __nfsd4_hash_conn(conn, ses);
914 spin_unlock(&clp->cl_lock);
917 static int nfsd4_register_conn(struct nfsd4_conn *conn)
919 conn->cn_xpt_user.callback = nfsd4_conn_lost;
920 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
923 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
927 nfsd4_hash_conn(conn, ses);
928 ret = nfsd4_register_conn(conn);
930 /* oops; xprt is already down: */
931 nfsd4_conn_lost(&conn->cn_xpt_user);
932 if (conn->cn_flags & NFS4_CDFC4_BACK) {
933 /* callback channel may be back up */
934 nfsd4_probe_callback(ses->se_client);
938 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
940 u32 dir = NFS4_CDFC4_FORE;
942 if (cses->flags & SESSION4_BACK_CHAN)
943 dir |= NFS4_CDFC4_BACK;
944 return alloc_conn(rqstp, dir);
947 /* must be called under client_lock */
948 static void nfsd4_del_conns(struct nfsd4_session *s)
950 struct nfs4_client *clp = s->se_client;
951 struct nfsd4_conn *c;
953 spin_lock(&clp->cl_lock);
954 while (!list_empty(&s->se_conns)) {
955 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
956 list_del_init(&c->cn_persession);
957 spin_unlock(&clp->cl_lock);
959 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
962 spin_lock(&clp->cl_lock);
964 spin_unlock(&clp->cl_lock);
967 static void __free_session(struct nfsd4_session *ses)
969 free_session_slots(ses);
973 static void free_session(struct nfsd4_session *ses)
975 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
977 lockdep_assert_held(&nn->client_lock);
978 nfsd4_del_conns(ses);
979 nfsd4_put_drc_mem(&ses->se_fchannel);
983 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
986 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
988 new->se_client = clp;
991 INIT_LIST_HEAD(&new->se_conns);
993 new->se_cb_seq_nr = 1;
994 new->se_flags = cses->flags;
995 new->se_cb_prog = cses->callback_prog;
996 new->se_cb_sec = cses->cb_sec;
997 atomic_set(&new->se_ref, 0);
998 idx = hash_sessionid(&new->se_sessionid);
999 spin_lock(&nn->client_lock);
1000 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1001 spin_lock(&clp->cl_lock);
1002 list_add(&new->se_perclnt, &clp->cl_sessions);
1003 spin_unlock(&clp->cl_lock);
1004 spin_unlock(&nn->client_lock);
1006 if (cses->flags & SESSION4_BACK_CHAN) {
1007 struct sockaddr *sa = svc_addr(rqstp);
1009 * This is a little silly; with sessions there's no real
1010 * use for the callback address. Use the peer address
1011 * as a reasonable default for now, but consider fixing
1012 * the rpc client not to require an address in the
1015 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1016 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1020 /* caller must hold client_lock */
1021 static struct nfsd4_session *
1022 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1024 struct nfsd4_session *elem;
1026 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1028 dump_sessionid(__func__, sessionid);
1029 idx = hash_sessionid(sessionid);
1030 /* Search in the appropriate list */
1031 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1032 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1033 NFS4_MAX_SESSIONID_LEN)) {
1038 dprintk("%s: session not found\n", __func__);
1042 /* caller must hold client_lock */
1044 unhash_session(struct nfsd4_session *ses)
1046 list_del(&ses->se_hash);
1047 spin_lock(&ses->se_client->cl_lock);
1048 list_del(&ses->se_perclnt);
1049 spin_unlock(&ses->se_client->cl_lock);
1052 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1054 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1056 if (clid->cl_boot == nn->boot_time)
1058 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1059 clid->cl_boot, clid->cl_id, nn->boot_time);
1064 * XXX Should we use a slab cache ?
1065 * This type of memory management is somewhat inefficient, but we use it
1066 * anyway since SETCLIENTID is not a common operation.
1068 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1070 struct nfs4_client *clp;
1072 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1075 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1076 if (clp->cl_name.data == NULL) {
1080 clp->cl_name.len = name.len;
1085 free_client(struct nfs4_client *clp)
1087 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1089 lockdep_assert_held(&nn->client_lock);
1090 while (!list_empty(&clp->cl_sessions)) {
1091 struct nfsd4_session *ses;
1092 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1094 list_del(&ses->se_perclnt);
1095 WARN_ON_ONCE(atomic_read(&ses->se_ref));
1098 free_svc_cred(&clp->cl_cred);
1099 kfree(clp->cl_name.data);
1100 idr_destroy(&clp->cl_stateids);
1104 /* must be called under the client_lock */
1106 unhash_client_locked(struct nfs4_client *clp)
1108 struct nfsd4_session *ses;
1110 list_del(&clp->cl_lru);
1111 spin_lock(&clp->cl_lock);
1112 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1113 list_del_init(&ses->se_hash);
1114 spin_unlock(&clp->cl_lock);
1118 destroy_client(struct nfs4_client *clp)
1120 struct nfs4_openowner *oo;
1121 struct nfs4_delegation *dp;
1122 struct list_head reaplist;
1123 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1125 INIT_LIST_HEAD(&reaplist);
1126 spin_lock(&recall_lock);
1127 while (!list_empty(&clp->cl_delegations)) {
1128 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1129 list_del_init(&dp->dl_perclnt);
1130 list_move(&dp->dl_recall_lru, &reaplist);
1132 spin_unlock(&recall_lock);
1133 while (!list_empty(&reaplist)) {
1134 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1135 destroy_delegation(dp);
1137 list_splice_init(&clp->cl_revoked, &reaplist);
1138 while (!list_empty(&reaplist)) {
1139 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1140 destroy_revoked_delegation(dp);
1142 while (!list_empty(&clp->cl_openowners)) {
1143 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1144 release_openowner(oo);
1146 nfsd4_shutdown_callback(clp);
1147 if (clp->cl_cb_conn.cb_xprt)
1148 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1149 list_del(&clp->cl_idhash);
1150 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1151 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1153 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1154 spin_lock(&nn->client_lock);
1155 unhash_client_locked(clp);
1156 WARN_ON_ONCE(atomic_read(&clp->cl_refcount));
1158 spin_unlock(&nn->client_lock);
1161 static void expire_client(struct nfs4_client *clp)
1163 nfsd4_client_record_remove(clp);
1164 destroy_client(clp);
1167 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1169 memcpy(target->cl_verifier.data, source->data,
1170 sizeof(target->cl_verifier.data));
1173 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1175 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1176 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1179 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1181 if (source->cr_principal) {
1182 target->cr_principal =
1183 kstrdup(source->cr_principal, GFP_KERNEL);
1184 if (target->cr_principal == NULL)
1187 target->cr_principal = NULL;
1188 target->cr_flavor = source->cr_flavor;
1189 target->cr_uid = source->cr_uid;
1190 target->cr_gid = source->cr_gid;
1191 target->cr_group_info = source->cr_group_info;
1192 get_group_info(target->cr_group_info);
1193 target->cr_gss_mech = source->cr_gss_mech;
1194 if (source->cr_gss_mech)
1195 gss_mech_get(source->cr_gss_mech);
1200 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1204 res = o1->len - o2->len;
1207 return (long long)memcmp(o1->data, o2->data, o1->len);
1210 static int same_name(const char *n1, const char *n2)
1212 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1216 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1218 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1222 same_clid(clientid_t *cl1, clientid_t *cl2)
1224 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1227 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1231 if (g1->ngroups != g2->ngroups)
1233 for (i=0; i<g1->ngroups; i++)
1234 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1240 * RFC 3530 language requires clid_inuse be returned when the
1241 * "principal" associated with a requests differs from that previously
1242 * used. We use uid, gid's, and gss principal string as our best
1243 * approximation. We also don't want to allow non-gss use of a client
1244 * established using gss: in theory cr_principal should catch that
1245 * change, but in practice cr_principal can be null even in the gss case
1246 * since gssd doesn't always pass down a principal string.
1248 static bool is_gss_cred(struct svc_cred *cr)
1250 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1251 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1256 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1258 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1259 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1260 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1261 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1263 if (cr1->cr_principal == cr2->cr_principal)
1265 if (!cr1->cr_principal || !cr2->cr_principal)
1267 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1270 static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1272 struct svc_cred *cr = &rqstp->rq_cred;
1275 if (!cr->cr_gss_mech)
1277 service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1278 return service == RPC_GSS_SVC_INTEGRITY ||
1279 service == RPC_GSS_SVC_PRIVACY;
1282 static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1284 struct svc_cred *cr = &rqstp->rq_cred;
1286 if (!cl->cl_mach_cred)
1288 if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1290 if (!svc_rqst_integrity_protected(rqstp))
1292 if (!cr->cr_principal)
1294 return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1297 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1299 static u32 current_clientid = 1;
1301 clp->cl_clientid.cl_boot = nn->boot_time;
1302 clp->cl_clientid.cl_id = current_clientid++;
1305 static void gen_confirm(struct nfs4_client *clp)
1310 verf[0] = (__be32)get_seconds();
1311 verf[1] = (__be32)i++;
1312 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1315 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1317 struct nfs4_stid *ret;
1319 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1320 if (!ret || !ret->sc_type)
1325 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1327 struct nfs4_stid *s;
1329 s = find_stateid(cl, t);
1332 if (typemask & s->sc_type)
1337 static struct nfs4_client *create_client(struct xdr_netobj name,
1338 struct svc_rqst *rqstp, nfs4_verifier *verf)
1340 struct nfs4_client *clp;
1341 struct sockaddr *sa = svc_addr(rqstp);
1343 struct net *net = SVC_NET(rqstp);
1344 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1346 clp = alloc_client(name);
1350 INIT_LIST_HEAD(&clp->cl_sessions);
1351 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1353 spin_lock(&nn->client_lock);
1355 spin_unlock(&nn->client_lock);
1358 idr_init(&clp->cl_stateids);
1359 atomic_set(&clp->cl_refcount, 0);
1360 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1361 INIT_LIST_HEAD(&clp->cl_idhash);
1362 INIT_LIST_HEAD(&clp->cl_openowners);
1363 INIT_LIST_HEAD(&clp->cl_delegations);
1364 INIT_LIST_HEAD(&clp->cl_lru);
1365 INIT_LIST_HEAD(&clp->cl_callbacks);
1366 INIT_LIST_HEAD(&clp->cl_revoked);
1367 spin_lock_init(&clp->cl_lock);
1368 nfsd4_init_callback(&clp->cl_cb_null);
1369 clp->cl_time = get_seconds();
1370 clear_bit(0, &clp->cl_cb_slot_busy);
1371 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1372 copy_verf(clp, verf);
1373 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1375 clp->cl_cb_session = NULL;
1381 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1383 struct rb_node **new = &(root->rb_node), *parent = NULL;
1384 struct nfs4_client *clp;
1387 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1390 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1391 new = &((*new)->rb_left);
1393 new = &((*new)->rb_right);
1396 rb_link_node(&new_clp->cl_namenode, parent, new);
1397 rb_insert_color(&new_clp->cl_namenode, root);
1400 static struct nfs4_client *
1401 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1404 struct rb_node *node = root->rb_node;
1405 struct nfs4_client *clp;
1408 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1409 cmp = compare_blob(&clp->cl_name, name);
1411 node = node->rb_left;
1413 node = node->rb_right;
1421 add_to_unconfirmed(struct nfs4_client *clp)
1423 unsigned int idhashval;
1424 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1426 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1427 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1428 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1429 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1434 move_to_confirmed(struct nfs4_client *clp)
1436 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1437 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1439 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1440 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1441 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1442 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1443 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1447 static struct nfs4_client *
1448 find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
1450 struct nfs4_client *clp;
1451 unsigned int idhashval = clientid_hashval(clid->cl_id);
1453 list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
1454 if (same_clid(&clp->cl_clientid, clid)) {
1455 if ((bool)clp->cl_minorversion != sessions)
1464 static struct nfs4_client *
1465 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1467 struct list_head *tbl = nn->conf_id_hashtbl;
1469 return find_client_in_id_table(tbl, clid, sessions);
1472 static struct nfs4_client *
1473 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1475 struct list_head *tbl = nn->unconf_id_hashtbl;
1477 return find_client_in_id_table(tbl, clid, sessions);
1480 static bool clp_used_exchangeid(struct nfs4_client *clp)
1482 return clp->cl_exchange_flags != 0;
1485 static struct nfs4_client *
1486 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1488 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1491 static struct nfs4_client *
1492 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1494 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1498 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1500 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1501 struct sockaddr *sa = svc_addr(rqstp);
1502 u32 scopeid = rpc_get_scope_id(sa);
1503 unsigned short expected_family;
1505 /* Currently, we only support tcp and tcp6 for the callback channel */
1506 if (se->se_callback_netid_len == 3 &&
1507 !memcmp(se->se_callback_netid_val, "tcp", 3))
1508 expected_family = AF_INET;
1509 else if (se->se_callback_netid_len == 4 &&
1510 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1511 expected_family = AF_INET6;
1515 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1516 se->se_callback_addr_len,
1517 (struct sockaddr *)&conn->cb_addr,
1518 sizeof(conn->cb_addr));
1520 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1523 if (conn->cb_addr.ss_family == AF_INET6)
1524 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1526 conn->cb_prog = se->se_callback_prog;
1527 conn->cb_ident = se->se_callback_ident;
1528 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1531 conn->cb_addr.ss_family = AF_UNSPEC;
1532 conn->cb_addrlen = 0;
1533 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1534 "will not receive delegations\n",
1535 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1541 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1544 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1546 struct nfsd4_slot *slot = resp->cstate.slot;
1549 dprintk("--> %s slot %p\n", __func__, slot);
1551 slot->sl_opcnt = resp->opcnt;
1552 slot->sl_status = resp->cstate.status;
1554 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1555 if (nfsd4_not_cached(resp)) {
1556 slot->sl_datalen = 0;
1559 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1560 base = (char *)resp->cstate.datap -
1561 (char *)resp->xbuf->head[0].iov_base;
1562 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1564 WARN("%s: sessions DRC could not cache compound\n", __func__);
1569 * Encode the replay sequence operation from the slot values.
1570 * If cachethis is FALSE encode the uncached rep error on the next
1571 * operation which sets resp->p and increments resp->opcnt for
1572 * nfs4svc_encode_compoundres.
1576 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1577 struct nfsd4_compoundres *resp)
1579 struct nfsd4_op *op;
1580 struct nfsd4_slot *slot = resp->cstate.slot;
1582 /* Encode the replayed sequence operation */
1583 op = &args->ops[resp->opcnt - 1];
1584 nfsd4_encode_operation(resp, op);
1586 /* Return nfserr_retry_uncached_rep in next operation. */
1587 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1588 op = &args->ops[resp->opcnt++];
1589 op->status = nfserr_retry_uncached_rep;
1590 nfsd4_encode_operation(resp, op);
1596 * The sequence operation is not cached because we can use the slot and
1600 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1601 struct nfsd4_sequence *seq)
1603 struct nfsd4_slot *slot = resp->cstate.slot;
1606 dprintk("--> %s slot %p\n", __func__, slot);
1608 /* Either returns 0 or nfserr_retry_uncached */
1609 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1610 if (status == nfserr_retry_uncached_rep)
1613 /* The sequence operation has been encoded, cstate->datap set. */
1614 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1616 resp->opcnt = slot->sl_opcnt;
1617 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1618 status = slot->sl_status;
1624 * Set the exchange_id flags returned by the server.
1627 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1629 /* pNFS is not supported */
1630 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1632 /* Referrals are supported, Migration is not. */
1633 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1635 /* set the wire flags to return to client. */
1636 clid->flags = new->cl_exchange_flags;
1639 static bool client_has_state(struct nfs4_client *clp)
1642 * Note clp->cl_openowners check isn't quite right: there's no
1643 * need to count owners without stateid's.
1645 * Also note we should probably be using this in 4.0 case too.
1647 return !list_empty(&clp->cl_openowners)
1648 || !list_empty(&clp->cl_delegations)
1649 || !list_empty(&clp->cl_sessions);
1653 nfsd4_exchange_id(struct svc_rqst *rqstp,
1654 struct nfsd4_compound_state *cstate,
1655 struct nfsd4_exchange_id *exid)
1657 struct nfs4_client *unconf, *conf, *new;
1659 char addr_str[INET6_ADDRSTRLEN];
1660 nfs4_verifier verf = exid->verifier;
1661 struct sockaddr *sa = svc_addr(rqstp);
1662 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1663 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1665 rpc_ntop(sa, addr_str, sizeof(addr_str));
1666 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1667 "ip_addr=%s flags %x, spa_how %d\n",
1668 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1669 addr_str, exid->flags, exid->spa_how);
1671 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1672 return nfserr_inval;
1674 switch (exid->spa_how) {
1676 if (!svc_rqst_integrity_protected(rqstp))
1677 return nfserr_inval;
1680 default: /* checked by xdr code */
1683 return nfserr_encr_alg_unsupp;
1686 /* Cases below refer to rfc 5661 section 18.35.4: */
1688 conf = find_confirmed_client_by_name(&exid->clname, nn);
1690 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1691 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1694 if (!clp_used_exchangeid(conf)) { /* buggy client */
1695 status = nfserr_inval;
1698 if (!mach_creds_match(conf, rqstp)) {
1699 status = nfserr_wrong_cred;
1702 if (!creds_match) { /* case 9 */
1703 status = nfserr_perm;
1706 if (!verfs_match) { /* case 8 */
1707 status = nfserr_not_same;
1711 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1715 if (!creds_match) { /* case 3 */
1716 if (client_has_state(conf)) {
1717 status = nfserr_clid_inuse;
1720 expire_client(conf);
1723 if (verfs_match) { /* case 2 */
1724 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1728 /* case 5, client reboot */
1732 if (update) { /* case 7 */
1733 status = nfserr_noent;
1737 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1738 if (unconf) /* case 4, possible retry or client restart */
1739 expire_client(unconf);
1741 /* case 1 (normal case) */
1743 new = create_client(exid->clname, rqstp, &verf);
1745 status = nfserr_jukebox;
1748 new->cl_minorversion = cstate->minorversion;
1749 new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
1752 add_to_unconfirmed(new);
1754 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1755 exid->clientid.cl_id = new->cl_clientid.cl_id;
1757 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1758 nfsd4_set_ex_flags(new, exid);
1760 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1761 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1765 nfs4_unlock_state();
1770 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1772 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1775 /* The slot is in use, and no response has been sent. */
1777 if (seqid == slot_seqid)
1778 return nfserr_jukebox;
1780 return nfserr_seq_misordered;
1782 /* Note unsigned 32-bit arithmetic handles wraparound: */
1783 if (likely(seqid == slot_seqid + 1))
1785 if (seqid == slot_seqid)
1786 return nfserr_replay_cache;
1787 return nfserr_seq_misordered;
1791 * Cache the create session result into the create session single DRC
1792 * slot cache by saving the xdr structure. sl_seqid has been set.
1793 * Do this for solo or embedded create session operations.
1796 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1797 struct nfsd4_clid_slot *slot, __be32 nfserr)
1799 slot->sl_status = nfserr;
1800 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1804 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1805 struct nfsd4_clid_slot *slot)
1807 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1808 return slot->sl_status;
1811 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1812 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1813 1 + /* MIN tag is length with zero, only length */ \
1814 3 + /* version, opcount, opcode */ \
1815 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1816 /* seqid, slotID, slotID, cache */ \
1817 4 ) * sizeof(__be32))
1819 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1820 2 + /* verifier: AUTH_NULL, length 0 */\
1822 1 + /* MIN tag is length with zero, only length */ \
1823 3 + /* opcount, opcode, opstatus*/ \
1824 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1825 /* seqid, slotID, slotID, slotID, status */ \
1826 5 ) * sizeof(__be32))
1828 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1830 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
1832 if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
1833 return nfserr_toosmall;
1834 if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
1835 return nfserr_toosmall;
1836 ca->headerpadsz = 0;
1837 ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
1838 ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
1839 ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
1840 ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
1841 NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
1842 ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
1844 * Note decreasing slot size below client's request may make it
1845 * difficult for client to function correctly, whereas
1846 * decreasing the number of slots will (just?) affect
1847 * performance. When short on memory we therefore prefer to
1848 * decrease number of slots instead of their size. Clients that
1849 * request larger slots than they need will get poor results:
1851 ca->maxreqs = nfsd4_get_drc_mem(ca);
1853 return nfserr_jukebox;
1858 #define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
1859 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
1860 #define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
1861 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
1863 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
1865 ca->headerpadsz = 0;
1868 * These RPC_MAX_HEADER macros are overkill, especially since we
1869 * don't even do gss on the backchannel yet. But this is still
1870 * less than 1k. Tighten up this estimate in the unlikely event
1871 * it turns out to be a problem for some client:
1873 if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
1874 return nfserr_toosmall;
1875 if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
1876 return nfserr_toosmall;
1877 ca->maxresp_cached = 0;
1879 return nfserr_toosmall;
1884 static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
1886 switch (cbs->flavor) {
1892 * GSS case: the spec doesn't allow us to return this
1893 * error. But it also doesn't allow us not to support
1895 * I'd rather this fail hard than return some error the
1896 * client might think it can already handle:
1898 return nfserr_encr_alg_unsupp;
1903 nfsd4_create_session(struct svc_rqst *rqstp,
1904 struct nfsd4_compound_state *cstate,
1905 struct nfsd4_create_session *cr_ses)
1907 struct sockaddr *sa = svc_addr(rqstp);
1908 struct nfs4_client *conf, *unconf;
1909 struct nfsd4_session *new;
1910 struct nfsd4_conn *conn;
1911 struct nfsd4_clid_slot *cs_slot = NULL;
1913 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1915 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1916 return nfserr_inval;
1917 status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
1920 status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
1923 status = check_backchannel_attrs(&cr_ses->back_channel);
1925 goto out_release_drc_mem;
1926 status = nfserr_jukebox;
1927 new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
1929 goto out_release_drc_mem;
1930 conn = alloc_conn_from_crses(rqstp, cr_ses);
1932 goto out_free_session;
1935 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1936 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1937 WARN_ON_ONCE(conf && unconf);
1940 status = nfserr_wrong_cred;
1941 if (!mach_creds_match(conf, rqstp))
1943 cs_slot = &conf->cl_cs_slot;
1944 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1945 if (status == nfserr_replay_cache) {
1946 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1948 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1949 status = nfserr_seq_misordered;
1952 } else if (unconf) {
1953 struct nfs4_client *old;
1954 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1955 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1956 status = nfserr_clid_inuse;
1959 status = nfserr_wrong_cred;
1960 if (!mach_creds_match(unconf, rqstp))
1962 cs_slot = &unconf->cl_cs_slot;
1963 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1965 /* an unconfirmed replay returns misordered */
1966 status = nfserr_seq_misordered;
1969 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1971 status = mark_client_expired(old);
1976 move_to_confirmed(unconf);
1979 status = nfserr_stale_clientid;
1984 * We do not support RDMA or persistent sessions
1986 cr_ses->flags &= ~SESSION4_PERSIST;
1987 cr_ses->flags &= ~SESSION4_RDMA;
1989 init_session(rqstp, new, conf, cr_ses);
1990 nfsd4_init_conn(rqstp, conn, new);
1992 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1993 NFS4_MAX_SESSIONID_LEN);
1994 cs_slot->sl_seqid++;
1995 cr_ses->seqid = cs_slot->sl_seqid;
1997 /* cache solo and embedded create sessions under the state lock */
1998 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1999 nfs4_unlock_state();
2002 nfs4_unlock_state();
2005 __free_session(new);
2006 out_release_drc_mem:
2007 nfsd4_put_drc_mem(&cr_ses->fore_channel);
2011 static __be32 nfsd4_map_bcts_dir(u32 *dir)
2014 case NFS4_CDFC4_FORE:
2015 case NFS4_CDFC4_BACK:
2017 case NFS4_CDFC4_FORE_OR_BOTH:
2018 case NFS4_CDFC4_BACK_OR_BOTH:
2019 *dir = NFS4_CDFC4_BOTH;
2022 return nfserr_inval;
2025 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2027 struct nfsd4_session *session = cstate->session;
2028 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2031 status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2034 spin_lock(&nn->client_lock);
2035 session->se_cb_prog = bc->bc_cb_program;
2036 session->se_cb_sec = bc->bc_cb_sec;
2037 spin_unlock(&nn->client_lock);
2039 nfsd4_probe_callback(session->se_client);
2044 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2045 struct nfsd4_compound_state *cstate,
2046 struct nfsd4_bind_conn_to_session *bcts)
2049 struct nfsd4_conn *conn;
2050 struct nfsd4_session *session;
2051 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2053 if (!nfsd4_last_compound_op(rqstp))
2054 return nfserr_not_only_op;
2056 spin_lock(&nn->client_lock);
2057 session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
2058 spin_unlock(&nn->client_lock);
2059 status = nfserr_badsession;
2062 status = nfserr_wrong_cred;
2063 if (!mach_creds_match(session->se_client, rqstp))
2065 status = nfsd4_map_bcts_dir(&bcts->dir);
2068 conn = alloc_conn(rqstp, bcts->dir);
2069 status = nfserr_jukebox;
2072 nfsd4_init_conn(rqstp, conn, session);
2075 nfs4_unlock_state();
2079 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2083 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2087 nfsd4_destroy_session(struct svc_rqst *r,
2088 struct nfsd4_compound_state *cstate,
2089 struct nfsd4_destroy_session *sessionid)
2091 struct nfsd4_session *ses;
2093 int ref_held_by_me = 0;
2094 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
2097 status = nfserr_not_only_op;
2098 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2099 if (!nfsd4_last_compound_op(r))
2103 dump_sessionid(__func__, &sessionid->sessionid);
2104 spin_lock(&nn->client_lock);
2105 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
2106 status = nfserr_badsession;
2108 goto out_client_lock;
2109 status = nfserr_wrong_cred;
2110 if (!mach_creds_match(ses->se_client, r))
2111 goto out_client_lock;
2112 nfsd4_get_session_locked(ses);
2113 status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2115 goto out_put_session;
2116 unhash_session(ses);
2117 spin_unlock(&nn->client_lock);
2119 nfsd4_probe_callback_sync(ses->se_client);
2121 spin_lock(&nn->client_lock);
2124 nfsd4_put_session(ses);
2126 spin_unlock(&nn->client_lock);
2128 nfs4_unlock_state();
2132 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2134 struct nfsd4_conn *c;
2136 list_for_each_entry(c, &s->se_conns, cn_persession) {
2137 if (c->cn_xprt == xpt) {
2144 static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2146 struct nfs4_client *clp = ses->se_client;
2147 struct nfsd4_conn *c;
2148 __be32 status = nfs_ok;
2151 spin_lock(&clp->cl_lock);
2152 c = __nfsd4_find_conn(new->cn_xprt, ses);
2155 status = nfserr_conn_not_bound_to_session;
2156 if (clp->cl_mach_cred)
2158 __nfsd4_hash_conn(new, ses);
2159 spin_unlock(&clp->cl_lock);
2160 ret = nfsd4_register_conn(new);
2162 /* oops; xprt is already down: */
2163 nfsd4_conn_lost(&new->cn_xpt_user);
2166 spin_unlock(&clp->cl_lock);
2171 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2173 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2175 return args->opcnt > session->se_fchannel.maxops;
2178 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2179 struct nfsd4_session *session)
2181 struct xdr_buf *xb = &rqstp->rq_arg;
2183 return xb->len > session->se_fchannel.maxreq_sz;
2187 nfsd4_sequence(struct svc_rqst *rqstp,
2188 struct nfsd4_compound_state *cstate,
2189 struct nfsd4_sequence *seq)
2191 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2192 struct nfsd4_session *session;
2193 struct nfs4_client *clp;
2194 struct nfsd4_slot *slot;
2195 struct nfsd4_conn *conn;
2197 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2199 if (resp->opcnt != 1)
2200 return nfserr_sequence_pos;
2203 * Will be either used or freed by nfsd4_sequence_check_conn
2206 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2208 return nfserr_jukebox;
2210 spin_lock(&nn->client_lock);
2211 status = nfserr_badsession;
2212 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2214 goto out_no_session;
2215 clp = session->se_client;
2216 status = get_client_locked(clp);
2218 goto out_no_session;
2219 status = nfsd4_get_session_locked(session);
2221 goto out_put_client;
2223 status = nfserr_too_many_ops;
2224 if (nfsd4_session_too_many_ops(rqstp, session))
2225 goto out_put_session;
2227 status = nfserr_req_too_big;
2228 if (nfsd4_request_too_big(rqstp, session))
2229 goto out_put_session;
2231 status = nfserr_badslot;
2232 if (seq->slotid >= session->se_fchannel.maxreqs)
2233 goto out_put_session;
2235 slot = session->se_slots[seq->slotid];
2236 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2238 /* We do not negotiate the number of slots yet, so set the
2239 * maxslots to the session maxreqs which is used to encode
2240 * sr_highest_slotid and the sr_target_slot id to maxslots */
2241 seq->maxslots = session->se_fchannel.maxreqs;
2243 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2244 slot->sl_flags & NFSD4_SLOT_INUSE);
2245 if (status == nfserr_replay_cache) {
2246 status = nfserr_seq_misordered;
2247 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2248 goto out_put_session;
2249 cstate->slot = slot;
2250 cstate->session = session;
2251 /* Return the cached reply status and set cstate->status
2252 * for nfsd4_proc_compound processing */
2253 status = nfsd4_replay_cache_entry(resp, seq);
2254 cstate->status = nfserr_replay_cache;
2258 goto out_put_session;
2260 status = nfsd4_sequence_check_conn(conn, session);
2263 goto out_put_session;
2265 /* Success! bump slot seqid */
2266 slot->sl_seqid = seq->seqid;
2267 slot->sl_flags |= NFSD4_SLOT_INUSE;
2269 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2271 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2273 cstate->slot = slot;
2274 cstate->session = session;
2277 switch (clp->cl_cb_state) {
2279 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2281 case NFSD4_CB_FAULT:
2282 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2285 seq->status_flags = 0;
2287 if (!list_empty(&clp->cl_revoked))
2288 seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2292 spin_unlock(&nn->client_lock);
2295 nfsd4_put_session(session);
2297 put_client_renew_locked(clp);
2298 goto out_no_session;
2302 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2304 struct nfs4_client *conf, *unconf, *clp;
2306 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2309 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2310 conf = find_confirmed_client(&dc->clientid, true, nn);
2311 WARN_ON_ONCE(conf && unconf);
2316 if (client_has_state(conf)) {
2317 status = nfserr_clientid_busy;
2323 status = nfserr_stale_clientid;
2326 if (!mach_creds_match(clp, rqstp)) {
2327 status = nfserr_wrong_cred;
2332 nfs4_unlock_state();
2337 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2341 if (rc->rca_one_fs) {
2342 if (!cstate->current_fh.fh_dentry)
2343 return nfserr_nofilehandle;
2345 * We don't take advantage of the rca_one_fs case.
2346 * That's OK, it's optional, we can safely ignore it.
2352 status = nfserr_complete_already;
2353 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2354 &cstate->session->se_client->cl_flags))
2357 status = nfserr_stale_clientid;
2358 if (is_client_expired(cstate->session->se_client))
2360 * The following error isn't really legal.
2361 * But we only get here if the client just explicitly
2362 * destroyed the client. Surely it no longer cares what
2363 * error it gets back on an operation for the dead
2369 nfsd4_client_record_create(cstate->session->se_client);
2371 nfs4_unlock_state();
2376 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2377 struct nfsd4_setclientid *setclid)
2379 struct xdr_netobj clname = setclid->se_name;
2380 nfs4_verifier clverifier = setclid->se_verf;
2381 struct nfs4_client *conf, *unconf, *new;
2383 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2385 /* Cases below refer to rfc 3530 section 14.2.33: */
2387 conf = find_confirmed_client_by_name(&clname, nn);
2390 status = nfserr_clid_inuse;
2391 if (clp_used_exchangeid(conf))
2393 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2394 char addr_str[INET6_ADDRSTRLEN];
2395 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2397 dprintk("NFSD: setclientid: string in use by client "
2398 "at %s\n", addr_str);
2402 unconf = find_unconfirmed_client_by_name(&clname, nn);
2404 expire_client(unconf);
2405 status = nfserr_jukebox;
2406 new = create_client(clname, rqstp, &clverifier);
2409 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2410 /* case 1: probable callback update */
2411 copy_clid(new, conf);
2412 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2414 new->cl_minorversion = 0;
2415 gen_callback(new, setclid, rqstp);
2416 add_to_unconfirmed(new);
2417 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2418 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2419 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2422 nfs4_unlock_state();
2428 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2429 struct nfsd4_compound_state *cstate,
2430 struct nfsd4_setclientid_confirm *setclientid_confirm)
2432 struct nfs4_client *conf, *unconf;
2433 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2434 clientid_t * clid = &setclientid_confirm->sc_clientid;
2436 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2438 if (STALE_CLIENTID(clid, nn))
2439 return nfserr_stale_clientid;
2442 conf = find_confirmed_client(clid, false, nn);
2443 unconf = find_unconfirmed_client(clid, false, nn);
2445 * We try hard to give out unique clientid's, so if we get an
2446 * attempt to confirm the same clientid with a different cred,
2447 * there's a bug somewhere. Let's charitably assume it's our
2450 status = nfserr_serverfault;
2451 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2453 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2455 /* cases below refer to rfc 3530 section 14.2.34: */
2456 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2457 if (conf && !unconf) /* case 2: probable retransmit */
2459 else /* case 4: client hasn't noticed we rebooted yet? */
2460 status = nfserr_stale_clientid;
2464 if (conf) { /* case 1: callback update */
2465 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2466 nfsd4_probe_callback(conf);
2467 expire_client(unconf);
2468 } else { /* case 3: normal case; new or rebooted client */
2469 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2471 status = mark_client_expired(conf);
2474 expire_client(conf);
2476 move_to_confirmed(unconf);
2477 nfsd4_probe_callback(unconf);
2480 nfs4_unlock_state();
2484 static struct nfs4_file *nfsd4_alloc_file(void)
2486 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2489 /* OPEN Share state helper functions */
2490 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2492 unsigned int hashval = file_hashval(ino);
2494 atomic_set(&fp->fi_ref, 1);
2495 INIT_LIST_HEAD(&fp->fi_stateids);
2496 INIT_LIST_HEAD(&fp->fi_delegations);
2497 fp->fi_inode = igrab(ino);
2498 fp->fi_had_conflict = false;
2499 fp->fi_lease = NULL;
2500 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2501 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2502 spin_lock(&recall_lock);
2503 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]);
2504 spin_unlock(&recall_lock);
2508 nfsd4_free_slab(struct kmem_cache **slab)
2512 kmem_cache_destroy(*slab);
2517 nfsd4_free_slabs(void)
2519 nfsd4_free_slab(&openowner_slab);
2520 nfsd4_free_slab(&lockowner_slab);
2521 nfsd4_free_slab(&file_slab);
2522 nfsd4_free_slab(&stateid_slab);
2523 nfsd4_free_slab(&deleg_slab);
2527 nfsd4_init_slabs(void)
2529 openowner_slab = kmem_cache_create("nfsd4_openowners",
2530 sizeof(struct nfs4_openowner), 0, 0, NULL);
2531 if (openowner_slab == NULL)
2533 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2534 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2535 if (lockowner_slab == NULL)
2537 file_slab = kmem_cache_create("nfsd4_files",
2538 sizeof(struct nfs4_file), 0, 0, NULL);
2539 if (file_slab == NULL)
2541 stateid_slab = kmem_cache_create("nfsd4_stateids",
2542 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2543 if (stateid_slab == NULL)
2545 deleg_slab = kmem_cache_create("nfsd4_delegations",
2546 sizeof(struct nfs4_delegation), 0, 0, NULL);
2547 if (deleg_slab == NULL)
2552 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2556 void nfs4_free_openowner(struct nfs4_openowner *oo)
2558 kfree(oo->oo_owner.so_owner.data);
2559 kmem_cache_free(openowner_slab, oo);
2562 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2564 kfree(lo->lo_owner.so_owner.data);
2565 kmem_cache_free(lockowner_slab, lo);
2568 static void init_nfs4_replay(struct nfs4_replay *rp)
2570 rp->rp_status = nfserr_serverfault;
2572 rp->rp_buf = rp->rp_ibuf;
2575 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2577 struct nfs4_stateowner *sop;
2579 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2583 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2584 if (!sop->so_owner.data) {
2585 kmem_cache_free(slab, sop);
2588 sop->so_owner.len = owner->len;
2590 INIT_LIST_HEAD(&sop->so_stateids);
2591 sop->so_client = clp;
2592 init_nfs4_replay(&sop->so_replay);
2596 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2598 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2600 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2601 list_add(&oo->oo_perclient, &clp->cl_openowners);
2604 static struct nfs4_openowner *
2605 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2606 struct nfs4_openowner *oo;
2608 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2611 oo->oo_owner.so_is_open_owner = 1;
2612 oo->oo_owner.so_seqid = open->op_seqid;
2613 oo->oo_flags = NFS4_OO_NEW;
2615 oo->oo_last_closed_stid = NULL;
2616 INIT_LIST_HEAD(&oo->oo_close_lru);
2617 hash_openowner(oo, clp, strhashval);
2621 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2622 struct nfs4_openowner *oo = open->op_openowner;
2624 stp->st_stid.sc_type = NFS4_OPEN_STID;
2625 INIT_LIST_HEAD(&stp->st_lockowners);
2626 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2627 list_add(&stp->st_perfile, &fp->fi_stateids);
2628 stp->st_stateowner = &oo->oo_owner;
2631 stp->st_access_bmap = 0;
2632 stp->st_deny_bmap = 0;
2633 set_access(open->op_share_access, stp);
2634 set_deny(open->op_share_deny, stp);
2635 stp->st_openstp = NULL;
2639 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2641 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2643 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2645 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2646 oo->oo_time = get_seconds();
2650 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2653 return (sop->so_owner.len == owner->len) &&
2654 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2655 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2658 static struct nfs4_openowner *
2659 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2660 bool sessions, struct nfsd_net *nn)
2662 struct nfs4_stateowner *so;
2663 struct nfs4_openowner *oo;
2664 struct nfs4_client *clp;
2666 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2667 if (!so->so_is_open_owner)
2669 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2671 clp = oo->oo_owner.so_client;
2672 if ((bool)clp->cl_minorversion != sessions)
2674 renew_client(oo->oo_owner.so_client);
2681 /* search file_hashtbl[] for file */
2682 static struct nfs4_file *
2683 find_file(struct inode *ino)
2685 unsigned int hashval = file_hashval(ino);
2686 struct nfs4_file *fp;
2688 spin_lock(&recall_lock);
2689 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2690 if (fp->fi_inode == ino) {
2692 spin_unlock(&recall_lock);
2696 spin_unlock(&recall_lock);
2701 * Called to check deny when READ with all zero stateid or
2702 * WRITE with all zero or all one stateid
2705 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2707 struct inode *ino = current_fh->fh_dentry->d_inode;
2708 struct nfs4_file *fp;
2709 struct nfs4_ol_stateid *stp;
2712 fp = find_file(ino);
2715 ret = nfserr_locked;
2716 /* Search for conflicting share reservations */
2717 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2718 if (test_deny(deny_type, stp) ||
2719 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2728 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2730 struct nfs4_client *clp = dp->dl_stid.sc_client;
2731 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2733 /* We're assuming the state code never drops its reference
2734 * without first removing the lease. Since we're in this lease
2735 * callback (and since the lease code is serialized by the kernel
2736 * lock) we know the server hasn't removed the lease yet, we know
2737 * it's safe to take a reference: */
2738 atomic_inc(&dp->dl_count);
2740 list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
2742 /* Only place dl_time is set; protected by i_lock: */
2743 dp->dl_time = get_seconds();
2745 nfsd4_cb_recall(dp);
2748 /* Called from break_lease() with i_lock held. */
2749 static void nfsd_break_deleg_cb(struct file_lock *fl)
2751 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2752 struct nfs4_delegation *dp;
2755 WARN(1, "(%p)->fl_owner NULL\n", fl);
2758 if (fp->fi_had_conflict) {
2759 WARN(1, "duplicate break on %p\n", fp);
2763 * We don't want the locks code to timeout the lease for us;
2764 * we'll remove it ourself if a delegation isn't returned
2767 fl->fl_break_time = 0;
2769 spin_lock(&recall_lock);
2770 fp->fi_had_conflict = true;
2771 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2772 nfsd_break_one_deleg(dp);
2773 spin_unlock(&recall_lock);
2777 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2780 return lease_modify(onlist, arg);
2785 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2786 .lm_break = nfsd_break_deleg_cb,
2787 .lm_change = nfsd_change_deleg_cb,
2790 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2792 if (nfsd4_has_session(cstate))
2794 if (seqid == so->so_seqid - 1)
2795 return nfserr_replay_me;
2796 if (seqid == so->so_seqid)
2798 return nfserr_bad_seqid;
2802 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2803 struct nfsd4_open *open, struct nfsd_net *nn)
2805 clientid_t *clientid = &open->op_clientid;
2806 struct nfs4_client *clp = NULL;
2807 unsigned int strhashval;
2808 struct nfs4_openowner *oo = NULL;
2811 if (STALE_CLIENTID(&open->op_clientid, nn))
2812 return nfserr_stale_clientid;
2814 * In case we need it later, after we've already created the
2815 * file and don't want to risk a further failure:
2817 open->op_file = nfsd4_alloc_file();
2818 if (open->op_file == NULL)
2819 return nfserr_jukebox;
2821 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2822 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2823 open->op_openowner = oo;
2825 clp = find_confirmed_client(clientid, cstate->minorversion,
2828 return nfserr_expired;
2831 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2832 /* Replace unconfirmed owners without checking for replay. */
2833 clp = oo->oo_owner.so_client;
2834 release_openowner(oo);
2835 open->op_openowner = NULL;
2838 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2841 clp = oo->oo_owner.so_client;
2844 oo = alloc_init_open_stateowner(strhashval, clp, open);
2846 return nfserr_jukebox;
2847 open->op_openowner = oo;
2849 open->op_stp = nfs4_alloc_stateid(clp);
2851 return nfserr_jukebox;
2855 static inline __be32
2856 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2858 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2859 return nfserr_openmode;
2864 static int share_access_to_flags(u32 share_access)
2866 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2869 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2871 struct nfs4_stid *ret;
2873 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2876 return delegstateid(ret);
2879 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2881 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2882 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2886 nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
2887 struct nfs4_delegation **dp)
2890 __be32 status = nfserr_bad_stateid;
2892 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2895 flags = share_access_to_flags(open->op_share_access);
2896 status = nfs4_check_delegmode(*dp, flags);
2900 if (!nfsd4_is_deleg_cur(open))
2904 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2909 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2911 struct nfs4_ol_stateid *local;
2912 struct nfs4_openowner *oo = open->op_openowner;
2914 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2915 /* ignore lock owners */
2916 if (local->st_stateowner->so_is_open_owner == 0)
2918 /* remember if we have seen this open owner */
2919 if (local->st_stateowner == &oo->oo_owner)
2921 /* check for conflicting share reservations */
2922 if (!test_share(local, open))
2923 return nfserr_share_denied;
2928 static inline int nfs4_access_to_access(u32 nfs4_access)
2932 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2933 flags |= NFSD_MAY_READ;
2934 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2935 flags |= NFSD_MAY_WRITE;
2939 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2940 struct svc_fh *cur_fh, struct nfsd4_open *open)
2943 int oflag = nfs4_access_to_omode(open->op_share_access);
2944 int access = nfs4_access_to_access(open->op_share_access);
2946 if (!fp->fi_fds[oflag]) {
2947 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2948 &fp->fi_fds[oflag]);
2952 nfs4_file_get_access(fp, oflag);
2957 static inline __be32
2958 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2959 struct nfsd4_open *open)
2961 struct iattr iattr = {
2962 .ia_valid = ATTR_SIZE,
2965 if (!open->op_truncate)
2967 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2968 return nfserr_inval;
2969 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2973 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)
2975 u32 op_share_access = open->op_share_access;
2979 new_access = !test_access(op_share_access, stp);
2981 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2985 status = nfsd4_truncate(rqstp, cur_fh, open);
2988 int oflag = nfs4_access_to_omode(op_share_access);
2989 nfs4_file_put_access(fp, oflag);
2993 /* remember the open */
2994 set_access(op_share_access, stp);
2995 set_deny(open->op_share_deny, stp);
3002 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
3004 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3007 /* Should we give out recallable state?: */
3008 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3010 if (clp->cl_cb_state == NFSD4_CB_UP)
3013 * In the sessions case, since we don't have to establish a
3014 * separate connection for callbacks, we assume it's OK
3015 * until we hear otherwise:
3017 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3020 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
3022 struct file_lock *fl;
3024 fl = locks_alloc_lock();
3027 locks_init_lock(fl);
3028 fl->fl_lmops = &nfsd_lease_mng_ops;
3029 fl->fl_flags = FL_DELEG;
3030 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3031 fl->fl_end = OFFSET_MAX;
3032 fl->fl_owner = (fl_owner_t)(dp->dl_file);
3033 fl->fl_pid = current->tgid;
3037 static int nfs4_setlease(struct nfs4_delegation *dp)
3039 struct nfs4_file *fp = dp->dl_file;
3040 struct file_lock *fl;
3043 fl = nfs4_alloc_init_lease(dp, NFS4_OPEN_DELEGATE_READ);
3046 fl->fl_file = find_readable_file(fp);
3047 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
3050 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
3052 fp->fi_deleg_file = get_file(fl->fl_file);
3053 atomic_set(&fp->fi_delegees, 1);
3054 list_add(&dp->dl_perfile, &fp->fi_delegations);
3057 locks_free_lock(fl);
3061 static int nfs4_set_delegation(struct nfs4_delegation *dp, struct nfs4_file *fp)
3065 if (fp->fi_had_conflict)
3069 if (!fp->fi_lease) {
3070 status = nfs4_setlease(dp);
3075 spin_lock(&recall_lock);
3076 if (fp->fi_had_conflict) {
3077 spin_unlock(&recall_lock);
3081 atomic_inc(&fp->fi_delegees);
3082 list_add(&dp->dl_perfile, &fp->fi_delegations);
3083 spin_unlock(&recall_lock);
3084 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
3092 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
3094 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3095 if (status == -EAGAIN)
3096 open->op_why_no_deleg = WND4_CONTENTION;
3098 open->op_why_no_deleg = WND4_RESOURCE;
3099 switch (open->op_deleg_want) {
3100 case NFS4_SHARE_WANT_READ_DELEG:
3101 case NFS4_SHARE_WANT_WRITE_DELEG:
3102 case NFS4_SHARE_WANT_ANY_DELEG:
3104 case NFS4_SHARE_WANT_CANCEL:
3105 open->op_why_no_deleg = WND4_CANCELLED;
3107 case NFS4_SHARE_WANT_NO_DELEG:
3114 * Attempt to hand out a delegation.
3116 * Note we don't support write delegations, and won't until the vfs has
3117 * proper support for them.
3120 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
3121 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
3123 struct nfs4_delegation *dp;
3124 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
3128 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
3129 open->op_recall = 0;
3130 switch (open->op_claim_type) {
3131 case NFS4_OPEN_CLAIM_PREVIOUS:
3133 open->op_recall = 1;
3134 if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
3137 case NFS4_OPEN_CLAIM_NULL:
3138 case NFS4_OPEN_CLAIM_FH:
3140 * Let's not give out any delegations till everyone's
3141 * had the chance to reclaim theirs....
3143 if (locks_in_grace(net))
3145 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
3148 * Also, if the file was opened for write or
3149 * create, there's a good chance the client's
3150 * about to write to it, resulting in an
3151 * immediate recall (since we don't support
3152 * write delegations):
3154 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
3156 if (open->op_create == NFS4_OPEN_CREATE)
3162 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh);
3165 status = nfs4_set_delegation(dp, stp->st_file);
3169 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3171 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3172 STATEID_VAL(&dp->dl_stid.sc_stateid));
3173 open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
3176 remove_stid(&dp->dl_stid);
3177 nfs4_put_delegation(dp);
3179 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
3180 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3181 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
3182 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3183 open->op_recall = 1;
3186 /* 4.1 client asking for a delegation? */
3187 if (open->op_deleg_want)
3188 nfsd4_open_deleg_none_ext(open, status);
3192 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3193 struct nfs4_delegation *dp)
3195 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3196 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3197 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3198 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3199 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3200 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3201 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3202 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3204 /* Otherwise the client must be confused wanting a delegation
3205 * it already has, therefore we don't return
3206 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3211 * called with nfs4_lock_state() held.
3214 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3216 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3217 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3218 struct nfs4_file *fp = NULL;
3219 struct inode *ino = current_fh->fh_dentry->d_inode;
3220 struct nfs4_ol_stateid *stp = NULL;
3221 struct nfs4_delegation *dp = NULL;
3225 * Lookup file; if found, lookup stateid and check open request,
3226 * and check for delegations in the process of being recalled.
3227 * If not found, create the nfs4_file struct
3229 fp = find_file(ino);
3231 if ((status = nfs4_check_open(fp, open, &stp)))
3233 status = nfs4_check_deleg(cl, open, &dp);
3237 status = nfserr_bad_stateid;
3238 if (nfsd4_is_deleg_cur(open))
3240 status = nfserr_jukebox;
3242 open->op_file = NULL;
3243 nfsd4_init_file(fp, ino);
3247 * OPEN the file, or upgrade an existing OPEN.
3248 * If truncate fails, the OPEN fails.
3251 /* Stateid was found, this is an OPEN upgrade */
3252 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3256 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3259 status = nfsd4_truncate(rqstp, current_fh, open);
3263 open->op_stp = NULL;
3264 init_open_stateid(stp, fp, open);
3266 update_stateid(&stp->st_stid.sc_stateid);
3267 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3269 if (nfsd4_has_session(&resp->cstate)) {
3270 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3272 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3273 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3274 open->op_why_no_deleg = WND4_NOT_WANTED;
3280 * Attempt to hand out a delegation. No error return, because the
3281 * OPEN succeeds even if we fail.
3283 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3287 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3288 STATEID_VAL(&stp->st_stid.sc_stateid));
3290 /* 4.1 client trying to upgrade/downgrade delegation? */
3291 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3292 open->op_deleg_want)
3293 nfsd4_deleg_xgrade_none_ext(open, dp);
3297 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3298 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3300 * To finish the open response, we just need to set the rflags.
3302 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3303 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3304 !nfsd4_has_session(&resp->cstate))
3305 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3310 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3312 if (open->op_openowner) {
3313 struct nfs4_openowner *oo = open->op_openowner;
3315 if (!list_empty(&oo->oo_owner.so_stateids))
3316 list_del_init(&oo->oo_close_lru);
3317 if (oo->oo_flags & NFS4_OO_NEW) {
3319 release_openowner(oo);
3320 open->op_openowner = NULL;
3322 oo->oo_flags &= ~NFS4_OO_NEW;
3326 nfsd4_free_file(open->op_file);
3328 free_generic_stateid(open->op_stp);
3331 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3333 struct nfs4_client *found;
3335 if (STALE_CLIENTID(clid, nn))
3336 return nfserr_stale_clientid;
3337 found = find_confirmed_client(clid, session, nn);
3340 return found ? nfs_ok : nfserr_expired;
3344 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3347 struct nfs4_client *clp;
3349 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3352 dprintk("process_renew(%08x/%08x): starting\n",
3353 clid->cl_boot, clid->cl_id);
3354 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3357 status = nfserr_cb_path_down;
3358 if (!list_empty(&clp->cl_delegations)
3359 && clp->cl_cb_state != NFSD4_CB_UP)
3363 nfs4_unlock_state();
3368 nfsd4_end_grace(struct nfsd_net *nn)
3370 /* do nothing if grace period already ended */
3371 if (nn->grace_ended)
3374 dprintk("NFSD: end of grace period\n");
3375 nn->grace_ended = true;
3376 nfsd4_record_grace_done(nn, nn->boot_time);
3377 locks_end_grace(&nn->nfsd4_manager);
3379 * Now that every NFSv4 client has had the chance to recover and
3380 * to see the (possibly new, possibly shorter) lease time, we
3381 * can safely set the next grace time to the current lease time:
3383 nn->nfsd4_grace = nn->nfsd4_lease;
3387 nfs4_laundromat(struct nfsd_net *nn)
3389 struct nfs4_client *clp;
3390 struct nfs4_openowner *oo;
3391 struct nfs4_delegation *dp;
3392 struct list_head *pos, *next, reaplist;
3393 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3394 time_t t, clientid_val = nn->nfsd4_lease;
3395 time_t u, test_val = nn->nfsd4_lease;
3399 dprintk("NFSD: laundromat service - starting\n");
3400 nfsd4_end_grace(nn);
3401 INIT_LIST_HEAD(&reaplist);
3402 spin_lock(&nn->client_lock);
3403 list_for_each_safe(pos, next, &nn->client_lru) {
3404 clp = list_entry(pos, struct nfs4_client, cl_lru);
3405 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3406 t = clp->cl_time - cutoff;
3407 if (clientid_val > t)
3411 if (mark_client_expired_locked(clp)) {
3412 dprintk("NFSD: client in use (clientid %08x)\n",
3413 clp->cl_clientid.cl_id);
3416 list_move(&clp->cl_lru, &reaplist);
3418 spin_unlock(&nn->client_lock);
3419 list_for_each_safe(pos, next, &reaplist) {
3420 clp = list_entry(pos, struct nfs4_client, cl_lru);
3421 dprintk("NFSD: purging unused client (clientid %08x)\n",
3422 clp->cl_clientid.cl_id);
3425 spin_lock(&recall_lock);
3426 list_for_each_safe(pos, next, &nn->del_recall_lru) {
3427 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3428 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3430 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3431 u = dp->dl_time - cutoff;
3436 list_move(&dp->dl_recall_lru, &reaplist);
3438 spin_unlock(&recall_lock);
3439 list_for_each_safe(pos, next, &reaplist) {
3440 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3441 revoke_delegation(dp);
3443 test_val = nn->nfsd4_lease;
3444 list_for_each_safe(pos, next, &nn->close_lru) {
3445 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3446 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3447 u = oo->oo_time - cutoff;
3452 release_openowner(oo);
3454 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3455 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3456 nfs4_unlock_state();
3457 return clientid_val;
3460 static struct workqueue_struct *laundry_wq;
3461 static void laundromat_main(struct work_struct *);
3464 laundromat_main(struct work_struct *laundry)
3467 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3469 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3472 t = nfs4_laundromat(nn);
3473 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3474 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3477 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3479 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3480 return nfserr_bad_stateid;
3485 access_permit_read(struct nfs4_ol_stateid *stp)
3487 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3488 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3489 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3493 access_permit_write(struct nfs4_ol_stateid *stp)
3495 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3496 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3500 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3502 __be32 status = nfserr_openmode;
3504 /* For lock stateid's, we test the parent open, not the lock: */
3505 if (stp->st_openstp)
3506 stp = stp->st_openstp;
3507 if ((flags & WR_STATE) && !access_permit_write(stp))
3509 if ((flags & RD_STATE) && !access_permit_read(stp))
3516 static inline __be32
3517 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3519 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3521 else if (locks_in_grace(net)) {
3522 /* Answer in remaining cases depends on existence of
3523 * conflicting state; so we must wait out the grace period. */
3524 return nfserr_grace;
3525 } else if (flags & WR_STATE)
3526 return nfs4_share_conflict(current_fh,
3527 NFS4_SHARE_DENY_WRITE);
3528 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3529 return nfs4_share_conflict(current_fh,
3530 NFS4_SHARE_DENY_READ);
3534 * Allow READ/WRITE during grace period on recovered state only for files
3535 * that are not able to provide mandatory locking.
3538 grace_disallows_io(struct net *net, struct inode *inode)
3540 return locks_in_grace(net) && mandatory_lock(inode);
3543 /* Returns true iff a is later than b: */
3544 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3546 return (s32)(a->si_generation - b->si_generation) > 0;
3549 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3552 * When sessions are used the stateid generation number is ignored
3555 if (has_session && in->si_generation == 0)
3558 if (in->si_generation == ref->si_generation)
3561 /* If the client sends us a stateid from the future, it's buggy: */
3562 if (stateid_generation_after(in, ref))
3563 return nfserr_bad_stateid;
3565 * However, we could see a stateid from the past, even from a
3566 * non-buggy client. For example, if the client sends a lock
3567 * while some IO is outstanding, the lock may bump si_generation
3568 * while the IO is still in flight. The client could avoid that
3569 * situation by waiting for responses on all the IO requests,
3570 * but better performance may result in retrying IO that
3571 * receives an old_stateid error if requests are rarely
3572 * reordered in flight:
3574 return nfserr_old_stateid;
3577 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3579 struct nfs4_stid *s;
3580 struct nfs4_ol_stateid *ols;
3583 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3584 return nfserr_bad_stateid;
3585 /* Client debugging aid. */
3586 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3587 char addr_str[INET6_ADDRSTRLEN];
3588 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3590 pr_warn_ratelimited("NFSD: client %s testing state ID "
3591 "with incorrect client ID\n", addr_str);
3592 return nfserr_bad_stateid;
3594 s = find_stateid(cl, stateid);
3596 return nfserr_bad_stateid;
3597 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3600 switch (s->sc_type) {
3601 case NFS4_DELEG_STID:
3603 case NFS4_REVOKED_DELEG_STID:
3604 return nfserr_deleg_revoked;
3605 case NFS4_OPEN_STID:
3606 case NFS4_LOCK_STID:
3607 ols = openlockstateid(s);
3608 if (ols->st_stateowner->so_is_open_owner
3609 && !(openowner(ols->st_stateowner)->oo_flags
3610 & NFS4_OO_CONFIRMED))
3611 return nfserr_bad_stateid;
3614 printk("unknown stateid type %x\n", s->sc_type);
3615 case NFS4_CLOSED_STID:
3616 return nfserr_bad_stateid;
3620 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3621 struct nfs4_stid **s, bool sessions,
3622 struct nfsd_net *nn)
3624 struct nfs4_client *cl;
3627 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3628 return nfserr_bad_stateid;
3629 status = lookup_clientid(&stateid->si_opaque.so_clid, sessions,
3631 if (status == nfserr_stale_clientid)
3632 return nfserr_stale_stateid;
3635 *s = find_stateid_by_type(cl, stateid, typemask);
3637 return nfserr_bad_stateid;
3642 * Checks for stateid operations
3645 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3646 stateid_t *stateid, int flags, struct file **filpp)
3648 struct nfs4_stid *s;
3649 struct nfs4_ol_stateid *stp = NULL;
3650 struct nfs4_delegation *dp = NULL;
3651 struct svc_fh *current_fh = &cstate->current_fh;
3652 struct inode *ino = current_fh->fh_dentry->d_inode;
3653 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3659 if (grace_disallows_io(net, ino))
3660 return nfserr_grace;
3662 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3663 return check_special_stateids(net, current_fh, stateid, flags);
3665 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3666 &s, cstate->minorversion, nn);
3669 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3672 switch (s->sc_type) {
3673 case NFS4_DELEG_STID:
3674 dp = delegstateid(s);
3675 status = nfs4_check_delegmode(dp, flags);
3679 *filpp = dp->dl_file->fi_deleg_file;
3682 status = nfserr_serverfault;
3687 case NFS4_OPEN_STID:
3688 case NFS4_LOCK_STID:
3689 stp = openlockstateid(s);
3690 status = nfs4_check_fh(current_fh, stp);
3693 if (stp->st_stateowner->so_is_open_owner
3694 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3696 status = nfs4_check_openmode(stp, flags);
3700 if (flags & RD_STATE)
3701 *filpp = find_readable_file(stp->st_file);
3703 *filpp = find_writeable_file(stp->st_file);
3707 return nfserr_bad_stateid;
3715 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3717 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3718 return nfserr_locks_held;
3719 release_lock_stateid(stp);
3724 * Test if the stateid is valid
3727 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3728 struct nfsd4_test_stateid *test_stateid)
3730 struct nfsd4_test_stateid_id *stateid;
3731 struct nfs4_client *cl = cstate->session->se_client;
3734 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3735 stateid->ts_id_status =
3736 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3737 nfs4_unlock_state();
3743 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3744 struct nfsd4_free_stateid *free_stateid)
3746 stateid_t *stateid = &free_stateid->fr_stateid;
3747 struct nfs4_stid *s;
3748 struct nfs4_delegation *dp;
3749 struct nfs4_client *cl = cstate->session->se_client;
3750 __be32 ret = nfserr_bad_stateid;
3753 s = find_stateid(cl, stateid);
3756 switch (s->sc_type) {
3757 case NFS4_DELEG_STID:
3758 ret = nfserr_locks_held;
3760 case NFS4_OPEN_STID:
3761 case NFS4_LOCK_STID:
3762 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3765 if (s->sc_type == NFS4_LOCK_STID)
3766 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3768 ret = nfserr_locks_held;
3770 case NFS4_REVOKED_DELEG_STID:
3771 dp = delegstateid(s);
3772 destroy_revoked_delegation(dp);
3776 ret = nfserr_bad_stateid;
3779 nfs4_unlock_state();
3786 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3787 RD_STATE : WR_STATE;
3790 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3792 struct svc_fh *current_fh = &cstate->current_fh;
3793 struct nfs4_stateowner *sop = stp->st_stateowner;
3796 status = nfsd4_check_seqid(cstate, sop, seqid);
3799 if (stp->st_stid.sc_type == NFS4_CLOSED_STID
3800 || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
3802 * "Closed" stateid's exist *only* to return
3803 * nfserr_replay_me from the previous step, and
3804 * revoked delegations are kept only for free_stateid.
3806 return nfserr_bad_stateid;
3807 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3810 return nfs4_check_fh(current_fh, stp);
3814 * Checks for sequence id mutating operations.
3817 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3818 stateid_t *stateid, char typemask,
3819 struct nfs4_ol_stateid **stpp,
3820 struct nfsd_net *nn)
3823 struct nfs4_stid *s;
3825 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3826 seqid, STATEID_VAL(stateid));
3829 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3830 cstate->minorversion, nn);
3833 *stpp = openlockstateid(s);
3834 if (!nfsd4_has_session(cstate))
3835 cstate->replay_owner = (*stpp)->st_stateowner;
3837 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3840 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3841 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3844 struct nfs4_openowner *oo;
3846 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3847 NFS4_OPEN_STID, stpp, nn);
3850 oo = openowner((*stpp)->st_stateowner);
3851 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3852 return nfserr_bad_stateid;
3857 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3858 struct nfsd4_open_confirm *oc)
3861 struct nfs4_openowner *oo;
3862 struct nfs4_ol_stateid *stp;
3863 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3865 dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
3866 cstate->current_fh.fh_dentry);
3868 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3874 status = nfs4_preprocess_seqid_op(cstate,
3875 oc->oc_seqid, &oc->oc_req_stateid,
3876 NFS4_OPEN_STID, &stp, nn);
3879 oo = openowner(stp->st_stateowner);
3880 status = nfserr_bad_stateid;
3881 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3883 oo->oo_flags |= NFS4_OO_CONFIRMED;
3884 update_stateid(&stp->st_stid.sc_stateid);
3885 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3886 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3887 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3889 nfsd4_client_record_create(oo->oo_owner.so_client);
3892 nfsd4_bump_seqid(cstate, status);
3893 if (!cstate->replay_owner)
3894 nfs4_unlock_state();
3898 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3900 if (!test_access(access, stp))
3902 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3903 clear_access(access, stp);
3906 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3908 switch (to_access) {
3909 case NFS4_SHARE_ACCESS_READ:
3910 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3911 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3913 case NFS4_SHARE_ACCESS_WRITE:
3914 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3915 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3917 case NFS4_SHARE_ACCESS_BOTH:
3925 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3928 for (i = 0; i < 4; i++) {
3929 if ((i & deny) != i)
3935 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3936 struct nfsd4_compound_state *cstate,
3937 struct nfsd4_open_downgrade *od)
3940 struct nfs4_ol_stateid *stp;
3941 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3943 dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
3944 cstate->current_fh.fh_dentry);
3946 /* We don't yet support WANT bits: */
3947 if (od->od_deleg_want)
3948 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3952 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3953 &od->od_stateid, &stp, nn);
3956 status = nfserr_inval;
3957 if (!test_access(od->od_share_access, stp)) {
3958 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3959 stp->st_access_bmap, od->od_share_access);
3962 if (!test_deny(od->od_share_deny, stp)) {
3963 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3964 stp->st_deny_bmap, od->od_share_deny);
3967 nfs4_stateid_downgrade(stp, od->od_share_access);
3969 reset_union_bmap_deny(od->od_share_deny, stp);
3971 update_stateid(&stp->st_stid.sc_stateid);
3972 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3975 nfsd4_bump_seqid(cstate, status);
3976 if (!cstate->replay_owner)
3977 nfs4_unlock_state();
3981 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3983 unhash_open_stateid(s);
3984 s->st_stid.sc_type = NFS4_CLOSED_STID;
3988 * nfs4_unlock_state() called after encode
3991 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3992 struct nfsd4_close *close)
3995 struct nfs4_openowner *oo;
3996 struct nfs4_ol_stateid *stp;
3997 struct net *net = SVC_NET(rqstp);
3998 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4000 dprintk("NFSD: nfsd4_close on file %pd\n",
4001 cstate->current_fh.fh_dentry);
4004 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
4006 NFS4_OPEN_STID|NFS4_CLOSED_STID,
4008 nfsd4_bump_seqid(cstate, status);
4011 oo = openowner(stp->st_stateowner);
4012 update_stateid(&stp->st_stid.sc_stateid);
4013 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4015 nfsd4_close_open_stateid(stp);
4017 if (cstate->minorversion)
4018 free_generic_stateid(stp);
4020 oo->oo_last_closed_stid = stp;
4022 if (list_empty(&oo->oo_owner.so_stateids)) {
4023 if (cstate->minorversion)
4024 release_openowner(oo);
4027 * In the 4.0 case we need to keep the owners around a
4028 * little while to handle CLOSE replay.
4030 move_to_close_lru(oo, SVC_NET(rqstp));
4034 if (!cstate->replay_owner)
4035 nfs4_unlock_state();
4040 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4041 struct nfsd4_delegreturn *dr)
4043 struct nfs4_delegation *dp;
4044 stateid_t *stateid = &dr->dr_stateid;
4045 struct nfs4_stid *s;
4047 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4049 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4053 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
4054 cstate->minorversion, nn);
4057 dp = delegstateid(s);
4058 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4062 destroy_delegation(dp);
4064 nfs4_unlock_state();
4070 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
4072 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
4075 end_offset(u64 start, u64 len)
4080 return end >= start ? end: NFS4_MAX_UINT64;
4083 /* last octet in a range */
4085 last_byte_offset(u64 start, u64 len)
4091 return end > start ? end - 1: NFS4_MAX_UINT64;
4094 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
4096 return (file_hashval(inode) + cl_id
4097 + opaque_hashval(ownername->data, ownername->len))
4098 & LOCKOWNER_INO_HASH_MASK;
4102 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
4103 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
4104 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
4105 * locking, this prevents us from being completely protocol-compliant. The
4106 * real solution to this problem is to start using unsigned file offsets in
4107 * the VFS, but this is a very deep change!
4110 nfs4_transform_lock_offset(struct file_lock *lock)
4112 if (lock->fl_start < 0)
4113 lock->fl_start = OFFSET_MAX;
4114 if (lock->fl_end < 0)
4115 lock->fl_end = OFFSET_MAX;
4118 /* Hack!: For now, we're defining this just so we can use a pointer to it
4119 * as a unique cookie to identify our (NFSv4's) posix locks. */
4120 static const struct lock_manager_operations nfsd_posix_mng_ops = {
4124 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
4126 struct nfs4_lockowner *lo;
4128 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
4129 lo = (struct nfs4_lockowner *) fl->fl_owner;
4130 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
4131 lo->lo_owner.so_owner.len, GFP_KERNEL);
4132 if (!deny->ld_owner.data)
4133 /* We just don't care that much */
4135 deny->ld_owner.len = lo->lo_owner.so_owner.len;
4136 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
4139 deny->ld_owner.len = 0;
4140 deny->ld_owner.data = NULL;
4141 deny->ld_clientid.cl_boot = 0;
4142 deny->ld_clientid.cl_id = 0;
4144 deny->ld_start = fl->fl_start;
4145 deny->ld_length = NFS4_MAX_UINT64;
4146 if (fl->fl_end != NFS4_MAX_UINT64)
4147 deny->ld_length = fl->fl_end - fl->fl_start + 1;
4148 deny->ld_type = NFS4_READ_LT;
4149 if (fl->fl_type != F_RDLCK)
4150 deny->ld_type = NFS4_WRITE_LT;
4153 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
4155 struct nfs4_ol_stateid *lst;
4157 if (!same_owner_str(&lo->lo_owner, owner, clid))
4159 lst = list_first_entry(&lo->lo_owner.so_stateids,
4160 struct nfs4_ol_stateid, st_perstateowner);
4161 return lst->st_file->fi_inode == inode;
4164 static struct nfs4_lockowner *
4165 find_lockowner_str(struct inode *inode, clientid_t *clid,
4166 struct xdr_netobj *owner, struct nfsd_net *nn)
4168 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
4169 struct nfs4_lockowner *lo;
4171 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
4172 if (same_lockowner_ino(lo, inode, clid, owner))
4178 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
4180 struct inode *inode = open_stp->st_file->fi_inode;
4181 unsigned int inohash = lockowner_ino_hashval(inode,
4182 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
4183 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
4185 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4186 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
4187 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4191 * Alloc a lock owner structure.
4192 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4195 * strhashval = ownerstr_hashval
4198 static struct nfs4_lockowner *
4199 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4200 struct nfs4_lockowner *lo;
4202 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4205 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4206 lo->lo_owner.so_is_open_owner = 0;
4207 /* It is the openowner seqid that will be incremented in encode in the
4208 * case of new lockowners; so increment the lock seqid manually: */
4209 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4210 hash_lockowner(lo, strhashval, clp, open_stp);
4214 static struct nfs4_ol_stateid *
4215 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4217 struct nfs4_ol_stateid *stp;
4218 struct nfs4_client *clp = lo->lo_owner.so_client;
4220 stp = nfs4_alloc_stateid(clp);
4223 stp->st_stid.sc_type = NFS4_LOCK_STID;
4224 list_add(&stp->st_perfile, &fp->fi_stateids);
4225 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4226 stp->st_stateowner = &lo->lo_owner;
4229 stp->st_access_bmap = 0;
4230 stp->st_deny_bmap = open_stp->st_deny_bmap;
4231 stp->st_openstp = open_stp;
4236 check_lock_length(u64 offset, u64 length)
4238 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4239 LOFF_OVERFLOW(offset, length)));
4242 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4244 struct nfs4_file *fp = lock_stp->st_file;
4245 int oflag = nfs4_access_to_omode(access);
4247 if (test_access(access, lock_stp))
4249 nfs4_file_get_access(fp, oflag);
4250 set_access(access, lock_stp);
4253 static __be32 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate, struct nfs4_ol_stateid *ost, struct nfsd4_lock *lock, struct nfs4_ol_stateid **lst, bool *new)
4255 struct nfs4_file *fi = ost->st_file;
4256 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4257 struct nfs4_client *cl = oo->oo_owner.so_client;
4258 struct nfs4_lockowner *lo;
4259 unsigned int strhashval;
4260 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4262 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4263 &lock->v.new.owner, nn);
4265 if (!cstate->minorversion)
4266 return nfserr_bad_seqid;
4267 /* XXX: a lockowner always has exactly one stateid: */
4268 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4269 struct nfs4_ol_stateid, st_perstateowner);
4272 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4273 &lock->v.new.owner);
4274 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4276 return nfserr_jukebox;
4277 *lst = alloc_init_lock_stateid(lo, fi, ost);
4279 release_lockowner(lo);
4280 return nfserr_jukebox;
4290 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4291 struct nfsd4_lock *lock)
4293 struct nfs4_openowner *open_sop = NULL;
4294 struct nfs4_lockowner *lock_sop = NULL;
4295 struct nfs4_ol_stateid *lock_stp;
4296 struct file *filp = NULL;
4297 struct file_lock *file_lock = NULL;
4298 struct file_lock *conflock = NULL;
4300 bool new_state = false;
4303 struct net *net = SVC_NET(rqstp);
4304 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4306 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4307 (long long) lock->lk_offset,
4308 (long long) lock->lk_length);
4310 if (check_lock_length(lock->lk_offset, lock->lk_length))
4311 return nfserr_inval;
4313 if ((status = fh_verify(rqstp, &cstate->current_fh,
4314 S_IFREG, NFSD_MAY_LOCK))) {
4315 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4321 if (lock->lk_is_new) {
4322 struct nfs4_ol_stateid *open_stp = NULL;
4324 if (nfsd4_has_session(cstate))
4325 /* See rfc 5661 18.10.3: given clientid is ignored: */
4326 memcpy(&lock->v.new.clientid,
4327 &cstate->session->se_client->cl_clientid,
4328 sizeof(clientid_t));
4330 status = nfserr_stale_clientid;
4331 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4334 /* validate and update open stateid and open seqid */
4335 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4336 lock->lk_new_open_seqid,
4337 &lock->lk_new_open_stateid,
4341 open_sop = openowner(open_stp->st_stateowner);
4342 status = nfserr_bad_stateid;
4343 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4344 &lock->v.new.clientid))
4346 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4347 &lock_stp, &new_state);
4349 status = nfs4_preprocess_seqid_op(cstate,
4350 lock->lk_old_lock_seqid,
4351 &lock->lk_old_lock_stateid,
4352 NFS4_LOCK_STID, &lock_stp, nn);
4355 lock_sop = lockowner(lock_stp->st_stateowner);
4357 lkflg = setlkflg(lock->lk_type);
4358 status = nfs4_check_openmode(lock_stp, lkflg);
4362 status = nfserr_grace;
4363 if (locks_in_grace(net) && !lock->lk_reclaim)
4365 status = nfserr_no_grace;
4366 if (!locks_in_grace(net) && lock->lk_reclaim)
4369 file_lock = locks_alloc_lock();
4371 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4372 status = nfserr_jukebox;
4376 locks_init_lock(file_lock);
4377 switch (lock->lk_type) {
4380 filp = find_readable_file(lock_stp->st_file);
4382 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4383 file_lock->fl_type = F_RDLCK;
4386 case NFS4_WRITEW_LT:
4387 filp = find_writeable_file(lock_stp->st_file);
4389 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4390 file_lock->fl_type = F_WRLCK;
4393 status = nfserr_inval;
4397 status = nfserr_openmode;
4400 file_lock->fl_owner = (fl_owner_t)lock_sop;
4401 file_lock->fl_pid = current->tgid;
4402 file_lock->fl_file = filp;
4403 file_lock->fl_flags = FL_POSIX;
4404 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4405 file_lock->fl_start = lock->lk_offset;
4406 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4407 nfs4_transform_lock_offset(file_lock);
4409 conflock = locks_alloc_lock();
4411 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4412 status = nfserr_jukebox;
4416 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4418 case 0: /* success! */
4419 update_stateid(&lock_stp->st_stid.sc_stateid);
4420 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4424 case (EAGAIN): /* conflock holds conflicting lock */
4425 status = nfserr_denied;
4426 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4427 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4430 status = nfserr_deadlock;
4433 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4434 status = nfserrno(err);
4438 if (status && new_state)
4439 release_lockowner(lock_sop);
4440 nfsd4_bump_seqid(cstate, status);
4441 if (!cstate->replay_owner)
4442 nfs4_unlock_state();
4444 locks_free_lock(file_lock);
4446 locks_free_lock(conflock);
4451 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4452 * so we do a temporary open here just to get an open file to pass to
4453 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4456 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4459 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4461 err = nfserrno(vfs_test_lock(file, lock));
4471 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4472 struct nfsd4_lockt *lockt)
4474 struct inode *inode;
4475 struct file_lock *file_lock = NULL;
4476 struct nfs4_lockowner *lo;
4478 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4480 if (locks_in_grace(SVC_NET(rqstp)))
4481 return nfserr_grace;
4483 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4484 return nfserr_inval;
4488 if (!nfsd4_has_session(cstate)) {
4489 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4494 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4497 inode = cstate->current_fh.fh_dentry->d_inode;
4498 file_lock = locks_alloc_lock();
4500 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4501 status = nfserr_jukebox;
4504 locks_init_lock(file_lock);
4505 switch (lockt->lt_type) {
4508 file_lock->fl_type = F_RDLCK;
4511 case NFS4_WRITEW_LT:
4512 file_lock->fl_type = F_WRLCK;
4515 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4516 status = nfserr_inval;
4520 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4522 file_lock->fl_owner = (fl_owner_t)lo;
4523 file_lock->fl_pid = current->tgid;
4524 file_lock->fl_flags = FL_POSIX;
4526 file_lock->fl_start = lockt->lt_offset;
4527 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4529 nfs4_transform_lock_offset(file_lock);
4531 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4535 if (file_lock->fl_type != F_UNLCK) {
4536 status = nfserr_denied;
4537 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4540 nfs4_unlock_state();
4542 locks_free_lock(file_lock);
4547 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4548 struct nfsd4_locku *locku)
4550 struct nfs4_ol_stateid *stp;
4551 struct file *filp = NULL;
4552 struct file_lock *file_lock = NULL;
4555 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4557 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4558 (long long) locku->lu_offset,
4559 (long long) locku->lu_length);
4561 if (check_lock_length(locku->lu_offset, locku->lu_length))
4562 return nfserr_inval;
4566 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4567 &locku->lu_stateid, NFS4_LOCK_STID,
4571 filp = find_any_file(stp->st_file);
4573 status = nfserr_lock_range;
4576 file_lock = locks_alloc_lock();
4578 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4579 status = nfserr_jukebox;
4582 locks_init_lock(file_lock);
4583 file_lock->fl_type = F_UNLCK;
4584 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4585 file_lock->fl_pid = current->tgid;
4586 file_lock->fl_file = filp;
4587 file_lock->fl_flags = FL_POSIX;
4588 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4589 file_lock->fl_start = locku->lu_offset;
4591 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4593 nfs4_transform_lock_offset(file_lock);
4595 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4597 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4600 update_stateid(&stp->st_stid.sc_stateid);
4601 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4604 nfsd4_bump_seqid(cstate, status);
4605 if (!cstate->replay_owner)
4606 nfs4_unlock_state();
4608 locks_free_lock(file_lock);
4612 status = nfserrno(err);
4618 * 1: locks held by lockowner
4619 * 0: no locks held by lockowner
4622 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4624 struct file_lock **flpp;
4625 struct inode *inode = filp->fi_inode;
4628 spin_lock(&inode->i_lock);
4629 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4630 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4636 spin_unlock(&inode->i_lock);
4641 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4642 struct nfsd4_compound_state *cstate,
4643 struct nfsd4_release_lockowner *rlockowner)
4645 clientid_t *clid = &rlockowner->rl_clientid;
4646 struct nfs4_stateowner *sop;
4647 struct nfs4_lockowner *lo;
4648 struct nfs4_ol_stateid *stp;
4649 struct xdr_netobj *owner = &rlockowner->rl_owner;
4650 struct list_head matches;
4651 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4653 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4655 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4656 clid->cl_boot, clid->cl_id);
4660 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4664 status = nfserr_locks_held;
4665 INIT_LIST_HEAD(&matches);
4667 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4668 if (sop->so_is_open_owner)
4670 if (!same_owner_str(sop, owner, clid))
4672 list_for_each_entry(stp, &sop->so_stateids,
4674 lo = lockowner(sop);
4675 if (check_for_locks(stp->st_file, lo))
4677 list_add(&lo->lo_list, &matches);
4680 /* Clients probably won't expect us to return with some (but not all)
4681 * of the lockowner state released; so don't release any until all
4682 * have been checked. */
4684 while (!list_empty(&matches)) {
4685 lo = list_entry(matches.next, struct nfs4_lockowner,
4687 /* unhash_stateowner deletes so_perclient only
4688 * for openowners. */
4689 list_del(&lo->lo_list);
4690 release_lockowner(lo);
4693 nfs4_unlock_state();
4697 static inline struct nfs4_client_reclaim *
4700 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4704 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4706 struct nfs4_client_reclaim *crp;
4708 crp = nfsd4_find_reclaim_client(name, nn);
4709 return (crp && crp->cr_clp);
4713 * failure => all reset bets are off, nfserr_no_grace...
4715 struct nfs4_client_reclaim *
4716 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4718 unsigned int strhashval;
4719 struct nfs4_client_reclaim *crp;
4721 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4722 crp = alloc_reclaim();
4724 strhashval = clientstr_hashval(name);
4725 INIT_LIST_HEAD(&crp->cr_strhash);
4726 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4727 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4729 nn->reclaim_str_hashtbl_size++;
4735 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4737 list_del(&crp->cr_strhash);
4739 nn->reclaim_str_hashtbl_size--;
4743 nfs4_release_reclaim(struct nfsd_net *nn)
4745 struct nfs4_client_reclaim *crp = NULL;
4748 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4749 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4750 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4751 struct nfs4_client_reclaim, cr_strhash);
4752 nfs4_remove_reclaim_record(crp, nn);
4755 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4759 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4760 struct nfs4_client_reclaim *
4761 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4763 unsigned int strhashval;
4764 struct nfs4_client_reclaim *crp = NULL;
4766 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4768 strhashval = clientstr_hashval(recdir);
4769 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4770 if (same_name(crp->cr_recdir, recdir)) {
4778 * Called from OPEN. Look for clientid in reclaim list.
4781 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4783 struct nfs4_client *clp;
4785 /* find clientid in conf_id_hashtbl */
4786 clp = find_confirmed_client(clid, sessions, nn);
4788 return nfserr_reclaim_bad;
4790 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4793 #ifdef CONFIG_NFSD_FAULT_INJECTION
4795 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4797 if (mark_client_expired(clp))
4803 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4805 char buf[INET6_ADDRSTRLEN];
4806 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4807 printk(KERN_INFO "NFS Client: %s\n", buf);
4811 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4814 char buf[INET6_ADDRSTRLEN];
4815 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4816 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4819 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_lockowner *))
4821 struct nfs4_openowner *oop;
4822 struct nfs4_lockowner *lop, *lo_next;
4823 struct nfs4_ol_stateid *stp, *st_next;
4826 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4827 list_for_each_entry_safe(stp, st_next, &oop->oo_owner.so_stateids, st_perstateowner) {
4828 list_for_each_entry_safe(lop, lo_next, &stp->st_lockowners, lo_perstateid) {
4840 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
4842 return nfsd_foreach_client_lock(clp, max, release_lockowner);
4845 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
4847 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
4848 nfsd_print_count(clp, count, "locked files");
4852 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
4854 struct nfs4_openowner *oop, *next;
4857 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
4867 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
4869 return nfsd_foreach_client_open(clp, max, release_openowner);
4872 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
4874 u64 count = nfsd_foreach_client_open(clp, max, NULL);
4875 nfsd_print_count(clp, count, "open files");
4879 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
4880 struct list_head *victims)
4882 struct nfs4_delegation *dp, *next;
4885 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
4887 list_move(&dp->dl_recall_lru, victims);
4894 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
4896 struct nfs4_delegation *dp, *next;
4900 spin_lock(&recall_lock);
4901 count = nfsd_find_all_delegations(clp, max, &victims);
4902 spin_unlock(&recall_lock);
4904 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4905 revoke_delegation(dp);
4910 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
4912 struct nfs4_delegation *dp, *next;
4916 spin_lock(&recall_lock);
4917 count = nfsd_find_all_delegations(clp, max, &victims);
4918 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4919 nfsd_break_one_deleg(dp);
4920 spin_unlock(&recall_lock);
4925 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
4929 spin_lock(&recall_lock);
4930 count = nfsd_find_all_delegations(clp, max, NULL);
4931 spin_unlock(&recall_lock);
4933 nfsd_print_count(clp, count, "delegations");
4937 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
4939 struct nfs4_client *clp, *next;
4941 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4943 if (!nfsd_netns_ready(nn))
4946 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4947 count += func(clp, max - count);
4948 if ((max != 0) && (count >= max))
4955 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
4957 struct nfs4_client *clp;
4958 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4960 if (!nfsd_netns_ready(nn))
4963 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
4964 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
4970 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4972 /* initialization to perform at module load time: */
4975 nfs4_state_init(void)
4980 * Since the lifetime of a delegation isn't limited to that of an open, a
4981 * client may quite reasonably hang on to a delegation as long as it has
4982 * the inode cached. This becomes an obvious problem the first time a
4983 * client's inode cache approaches the size of the server's total memory.
4985 * For now we avoid this problem by imposing a hard limit on the number
4986 * of delegations, which varies according to the server's memory size.
4989 set_max_delegations(void)
4992 * Allow at most 4 delegations per megabyte of RAM. Quick
4993 * estimates suggest that in the worst case (where every delegation
4994 * is for a different inode), a delegation could take about 1.5K,
4995 * giving a worst case usage of about 6% of memory.
4997 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
5000 static int nfs4_state_create_net(struct net *net)
5002 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5005 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5006 CLIENT_HASH_SIZE, GFP_KERNEL);
5007 if (!nn->conf_id_hashtbl)
5009 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
5010 CLIENT_HASH_SIZE, GFP_KERNEL);
5011 if (!nn->unconf_id_hashtbl)
5013 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
5014 OWNER_HASH_SIZE, GFP_KERNEL);
5015 if (!nn->ownerstr_hashtbl)
5017 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
5018 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
5019 if (!nn->lockowner_ino_hashtbl)
5020 goto err_lockowner_ino;
5021 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
5022 SESSION_HASH_SIZE, GFP_KERNEL);
5023 if (!nn->sessionid_hashtbl)
5026 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5027 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
5028 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
5030 for (i = 0; i < OWNER_HASH_SIZE; i++)
5031 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
5032 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
5033 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
5034 for (i = 0; i < SESSION_HASH_SIZE; i++)
5035 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
5036 nn->conf_name_tree = RB_ROOT;
5037 nn->unconf_name_tree = RB_ROOT;
5038 INIT_LIST_HEAD(&nn->client_lru);
5039 INIT_LIST_HEAD(&nn->close_lru);
5040 INIT_LIST_HEAD(&nn->del_recall_lru);
5041 spin_lock_init(&nn->client_lock);
5043 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
5049 kfree(nn->lockowner_ino_hashtbl);
5051 kfree(nn->ownerstr_hashtbl);
5053 kfree(nn->unconf_id_hashtbl);
5055 kfree(nn->conf_id_hashtbl);
5061 nfs4_state_destroy_net(struct net *net)
5064 struct nfs4_client *clp = NULL;
5065 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5067 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5068 while (!list_empty(&nn->conf_id_hashtbl[i])) {
5069 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5070 destroy_client(clp);
5074 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5075 while (!list_empty(&nn->unconf_id_hashtbl[i])) {
5076 clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
5077 destroy_client(clp);
5081 kfree(nn->sessionid_hashtbl);
5082 kfree(nn->lockowner_ino_hashtbl);
5083 kfree(nn->ownerstr_hashtbl);
5084 kfree(nn->unconf_id_hashtbl);
5085 kfree(nn->conf_id_hashtbl);
5090 nfs4_state_start_net(struct net *net)
5092 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5095 ret = nfs4_state_create_net(net);
5098 nfsd4_client_tracking_init(net);
5099 nn->boot_time = get_seconds();
5100 locks_start_grace(net, &nn->nfsd4_manager);
5101 nn->grace_ended = false;
5102 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
5103 nn->nfsd4_grace, net);
5104 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
5108 /* initialization to perform when the nfsd service is started: */
5111 nfs4_state_start(void)
5115 ret = set_callback_cred();
5118 laundry_wq = create_singlethread_workqueue("nfsd4");
5119 if (laundry_wq == NULL) {
5123 ret = nfsd4_create_callback_queue();
5125 goto out_free_laundry;
5127 set_max_delegations();
5132 destroy_workqueue(laundry_wq);
5138 nfs4_state_shutdown_net(struct net *net)
5140 struct nfs4_delegation *dp = NULL;
5141 struct list_head *pos, *next, reaplist;
5142 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5144 cancel_delayed_work_sync(&nn->laundromat_work);
5145 locks_end_grace(&nn->nfsd4_manager);
5148 INIT_LIST_HEAD(&reaplist);
5149 spin_lock(&recall_lock);
5150 list_for_each_safe(pos, next, &nn->del_recall_lru) {
5151 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5152 list_move(&dp->dl_recall_lru, &reaplist);
5154 spin_unlock(&recall_lock);
5155 list_for_each_safe(pos, next, &reaplist) {
5156 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5157 destroy_delegation(dp);
5160 nfsd4_client_tracking_exit(net);
5161 nfs4_state_destroy_net(net);
5162 nfs4_unlock_state();
5166 nfs4_state_shutdown(void)
5168 destroy_workqueue(laundry_wq);
5169 nfsd4_destroy_callback_queue();
5173 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5175 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5176 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5180 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5182 if (cstate->minorversion) {
5183 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5184 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5189 clear_current_stateid(struct nfsd4_compound_state *cstate)
5191 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5195 * functions to set current state id
5198 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5200 put_stateid(cstate, &odp->od_stateid);
5204 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5206 put_stateid(cstate, &open->op_stateid);
5210 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5212 put_stateid(cstate, &close->cl_stateid);
5216 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5218 put_stateid(cstate, &lock->lk_resp_stateid);
5222 * functions to consume current state id
5226 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5228 get_stateid(cstate, &odp->od_stateid);
5232 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5234 get_stateid(cstate, &drp->dr_stateid);
5238 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5240 get_stateid(cstate, &fsp->fr_stateid);
5244 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5246 get_stateid(cstate, &setattr->sa_stateid);
5250 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5252 get_stateid(cstate, &close->cl_stateid);
5256 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5258 get_stateid(cstate, &locku->lu_stateid);
5262 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5264 get_stateid(cstate, &read->rd_stateid);
5268 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5270 get_stateid(cstate, &write->wr_stateid);