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>
46 #include "current_stateid.h"
50 #define NFSDDBG_FACILITY NFSDDBG_PROC
52 #define all_ones {{~0,~0},~0}
53 static const stateid_t one_stateid = {
55 .si_opaque = all_ones,
57 static const stateid_t zero_stateid = {
60 static const stateid_t currentstateid = {
64 static u64 current_sessionid = 1;
66 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
67 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
68 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
70 /* forward declarations */
71 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
75 /* Currently used for almost all code touching nfsv4 state: */
76 static DEFINE_MUTEX(client_mutex);
79 * Currently used for the del_recall_lru and file hash table. In an
80 * effort to decrease the scope of the client_mutex, this spinlock may
81 * eventually cover more:
83 static DEFINE_SPINLOCK(recall_lock);
85 static struct kmem_cache *openowner_slab = NULL;
86 static struct kmem_cache *lockowner_slab = NULL;
87 static struct kmem_cache *file_slab = NULL;
88 static struct kmem_cache *stateid_slab = NULL;
89 static struct kmem_cache *deleg_slab = NULL;
94 mutex_lock(&client_mutex);
97 static void free_session(struct kref *);
99 /* Must be called under the client_lock */
100 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
102 kref_put(&ses->se_ref, free_session);
105 static void nfsd4_get_session(struct nfsd4_session *ses)
107 kref_get(&ses->se_ref);
111 nfs4_unlock_state(void)
113 mutex_unlock(&client_mutex);
117 opaque_hashval(const void *ptr, int nbytes)
119 unsigned char *cptr = (unsigned char *) ptr;
129 static struct list_head del_recall_lru;
131 static void nfsd4_free_file(struct nfs4_file *f)
133 kmem_cache_free(file_slab, f);
137 put_nfs4_file(struct nfs4_file *fi)
139 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
140 list_del(&fi->fi_hash);
141 spin_unlock(&recall_lock);
148 get_nfs4_file(struct nfs4_file *fi)
150 atomic_inc(&fi->fi_ref);
153 static int num_delegations;
154 unsigned long max_delegations;
157 * Open owner state (share locks)
160 /* hash tables for lock and open owners */
161 #define OWNER_HASH_BITS 8
162 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
163 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
165 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
169 ret = opaque_hashval(ownername->data, ownername->len);
171 return ret & OWNER_HASH_MASK;
174 /* hash table for nfs4_file */
175 #define FILE_HASH_BITS 8
176 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
178 static unsigned int file_hashval(struct inode *ino)
180 /* XXX: why are we hashing on inode pointer, anyway? */
181 return hash_ptr(ino, FILE_HASH_BITS);
184 static struct list_head file_hashtbl[FILE_HASH_SIZE];
186 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
188 WARN_ON_ONCE(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
189 atomic_inc(&fp->fi_access[oflag]);
192 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
194 if (oflag == O_RDWR) {
195 __nfs4_file_get_access(fp, O_RDONLY);
196 __nfs4_file_get_access(fp, O_WRONLY);
198 __nfs4_file_get_access(fp, oflag);
201 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
203 if (fp->fi_fds[oflag]) {
204 fput(fp->fi_fds[oflag]);
205 fp->fi_fds[oflag] = NULL;
209 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
211 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
212 nfs4_file_put_fd(fp, oflag);
214 * It's also safe to get rid of the RDWR open *if*
215 * we no longer have need of the other kind of access
216 * or if we already have the other kind of open:
218 if (fp->fi_fds[1-oflag]
219 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
220 nfs4_file_put_fd(fp, O_RDWR);
224 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
226 if (oflag == O_RDWR) {
227 __nfs4_file_put_access(fp, O_RDONLY);
228 __nfs4_file_put_access(fp, O_WRONLY);
230 __nfs4_file_put_access(fp, oflag);
233 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
236 struct idr *stateids = &cl->cl_stateids;
237 static int min_stateid = 0;
238 struct nfs4_stid *stid;
241 stid = kmem_cache_alloc(slab, GFP_KERNEL);
245 new_id = idr_alloc(stateids, stid, min_stateid, 0, GFP_KERNEL);
248 stid->sc_client = cl;
250 stid->sc_stateid.si_opaque.so_id = new_id;
251 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
252 /* Will be incremented before return to client: */
253 stid->sc_stateid.si_generation = 0;
256 * It shouldn't be a problem to reuse an opaque stateid value.
257 * I don't think it is for 4.1. But with 4.0 I worry that, for
258 * example, a stray write retransmission could be accepted by
259 * the server when it should have been rejected. Therefore,
260 * adopt a trick from the sctp code to attempt to maximize the
261 * amount of time until an id is reused, by ensuring they always
262 * "increase" (mod INT_MAX):
265 min_stateid = new_id+1;
266 if (min_stateid == INT_MAX)
274 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
276 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
279 static struct nfs4_delegation *
280 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
282 struct nfs4_delegation *dp;
283 struct nfs4_file *fp = stp->st_file;
285 dprintk("NFSD alloc_init_deleg\n");
287 * Major work on the lease subsystem (for example, to support
288 * calbacks on stat) will be required before we can support
289 * write delegations properly.
291 if (type != NFS4_OPEN_DELEGATE_READ)
293 if (fp->fi_had_conflict)
295 if (num_delegations > max_delegations)
297 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
300 dp->dl_stid.sc_type = NFS4_DELEG_STID;
302 * delegation seqid's are never incremented. The 4.1 special
303 * meaning of seqid 0 isn't meaningful, really, but let's avoid
304 * 0 anyway just for consistency and use 1:
306 dp->dl_stid.sc_stateid.si_generation = 1;
308 INIT_LIST_HEAD(&dp->dl_perfile);
309 INIT_LIST_HEAD(&dp->dl_perclnt);
310 INIT_LIST_HEAD(&dp->dl_recall_lru);
314 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
316 atomic_set(&dp->dl_count, 1);
317 nfsd4_init_callback(&dp->dl_recall);
321 static void free_stid(struct nfs4_stid *s, struct kmem_cache *slab)
323 struct idr *stateids = &s->sc_client->cl_stateids;
325 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
326 kmem_cache_free(slab, s);
330 nfs4_put_delegation(struct nfs4_delegation *dp)
332 if (atomic_dec_and_test(&dp->dl_count)) {
333 dprintk("NFSD: freeing dp %p\n",dp);
334 put_nfs4_file(dp->dl_file);
335 free_stid(&dp->dl_stid, deleg_slab);
340 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
342 if (atomic_dec_and_test(&fp->fi_delegees)) {
343 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
345 fput(fp->fi_deleg_file);
346 fp->fi_deleg_file = NULL;
350 static void unhash_stid(struct nfs4_stid *s)
355 /* Called under the state lock. */
357 unhash_delegation(struct nfs4_delegation *dp)
359 unhash_stid(&dp->dl_stid);
360 list_del_init(&dp->dl_perclnt);
361 spin_lock(&recall_lock);
362 list_del_init(&dp->dl_perfile);
363 list_del_init(&dp->dl_recall_lru);
364 spin_unlock(&recall_lock);
365 nfs4_put_deleg_lease(dp->dl_file);
366 nfs4_put_delegation(dp);
373 static unsigned int clientid_hashval(u32 id)
375 return id & CLIENT_HASH_MASK;
378 static unsigned int clientstr_hashval(const char *name)
380 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
384 * We store the NONE, READ, WRITE, and BOTH bits separately in the
385 * st_{access,deny}_bmap field of the stateid, in order to track not
386 * only what share bits are currently in force, but also what
387 * combinations of share bits previous opens have used. This allows us
388 * to enforce the recommendation of rfc 3530 14.2.19 that the server
389 * return an error if the client attempt to downgrade to a combination
390 * of share bits not explicable by closing some of its previous opens.
392 * XXX: This enforcement is actually incomplete, since we don't keep
393 * track of access/deny bit combinations; so, e.g., we allow:
395 * OPEN allow read, deny write
396 * OPEN allow both, deny none
397 * DOWNGRADE allow read, deny none
399 * which we should reject.
402 bmap_to_share_mode(unsigned long bmap) {
404 unsigned int access = 0;
406 for (i = 1; i < 4; i++) {
407 if (test_bit(i, &bmap))
414 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
415 unsigned int access, deny;
417 access = bmap_to_share_mode(stp->st_access_bmap);
418 deny = bmap_to_share_mode(stp->st_deny_bmap);
419 if ((access & open->op_share_deny) || (deny & open->op_share_access))
424 /* set share access for a given stateid */
426 set_access(u32 access, struct nfs4_ol_stateid *stp)
428 __set_bit(access, &stp->st_access_bmap);
431 /* clear share access for a given stateid */
433 clear_access(u32 access, struct nfs4_ol_stateid *stp)
435 __clear_bit(access, &stp->st_access_bmap);
438 /* test whether a given stateid has access */
440 test_access(u32 access, struct nfs4_ol_stateid *stp)
442 return test_bit(access, &stp->st_access_bmap);
445 /* set share deny for a given stateid */
447 set_deny(u32 access, struct nfs4_ol_stateid *stp)
449 __set_bit(access, &stp->st_deny_bmap);
452 /* clear share deny for a given stateid */
454 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
456 __clear_bit(access, &stp->st_deny_bmap);
459 /* test whether a given stateid is denying specific access */
461 test_deny(u32 access, struct nfs4_ol_stateid *stp)
463 return test_bit(access, &stp->st_deny_bmap);
466 static int nfs4_access_to_omode(u32 access)
468 switch (access & NFS4_SHARE_ACCESS_BOTH) {
469 case NFS4_SHARE_ACCESS_READ:
471 case NFS4_SHARE_ACCESS_WRITE:
473 case NFS4_SHARE_ACCESS_BOTH:
480 /* release all access and file references for a given stateid */
482 release_all_access(struct nfs4_ol_stateid *stp)
486 for (i = 1; i < 4; i++) {
487 if (test_access(i, stp))
488 nfs4_file_put_access(stp->st_file,
489 nfs4_access_to_omode(i));
490 clear_access(i, stp);
494 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
496 list_del(&stp->st_perfile);
497 list_del(&stp->st_perstateowner);
500 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
502 release_all_access(stp);
503 put_nfs4_file(stp->st_file);
507 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
509 free_stid(&stp->st_stid, stateid_slab);
512 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
516 unhash_generic_stateid(stp);
517 unhash_stid(&stp->st_stid);
518 file = find_any_file(stp->st_file);
520 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
521 close_generic_stateid(stp);
522 free_generic_stateid(stp);
525 static void unhash_lockowner(struct nfs4_lockowner *lo)
527 struct nfs4_ol_stateid *stp;
529 list_del(&lo->lo_owner.so_strhash);
530 list_del(&lo->lo_perstateid);
531 list_del(&lo->lo_owner_ino_hash);
532 while (!list_empty(&lo->lo_owner.so_stateids)) {
533 stp = list_first_entry(&lo->lo_owner.so_stateids,
534 struct nfs4_ol_stateid, st_perstateowner);
535 release_lock_stateid(stp);
539 static void release_lockowner(struct nfs4_lockowner *lo)
541 unhash_lockowner(lo);
542 nfs4_free_lockowner(lo);
546 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
548 struct nfs4_lockowner *lo;
550 while (!list_empty(&open_stp->st_lockowners)) {
551 lo = list_entry(open_stp->st_lockowners.next,
552 struct nfs4_lockowner, lo_perstateid);
553 release_lockowner(lo);
557 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
559 unhash_generic_stateid(stp);
560 release_stateid_lockowners(stp);
561 close_generic_stateid(stp);
564 static void release_open_stateid(struct nfs4_ol_stateid *stp)
566 unhash_open_stateid(stp);
567 unhash_stid(&stp->st_stid);
568 free_generic_stateid(stp);
571 static void unhash_openowner(struct nfs4_openowner *oo)
573 struct nfs4_ol_stateid *stp;
575 list_del(&oo->oo_owner.so_strhash);
576 list_del(&oo->oo_perclient);
577 while (!list_empty(&oo->oo_owner.so_stateids)) {
578 stp = list_first_entry(&oo->oo_owner.so_stateids,
579 struct nfs4_ol_stateid, st_perstateowner);
580 release_open_stateid(stp);
584 static void release_last_closed_stateid(struct nfs4_openowner *oo)
586 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
589 unhash_stid(&s->st_stid);
590 free_generic_stateid(s);
591 oo->oo_last_closed_stid = NULL;
595 static void release_openowner(struct nfs4_openowner *oo)
597 unhash_openowner(oo);
598 list_del(&oo->oo_close_lru);
599 release_last_closed_stateid(oo);
600 nfs4_free_openowner(oo);
604 hash_sessionid(struct nfs4_sessionid *sessionid)
606 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
608 return sid->sequence % SESSION_HASH_SIZE;
613 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
615 u32 *ptr = (u32 *)(&sessionid->data[0]);
616 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
620 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
627 gen_sessionid(struct nfsd4_session *ses)
629 struct nfs4_client *clp = ses->se_client;
630 struct nfsd4_sessionid *sid;
632 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
633 sid->clientid = clp->cl_clientid;
634 sid->sequence = current_sessionid++;
639 * The protocol defines ca_maxresponssize_cached to include the size of
640 * the rpc header, but all we need to cache is the data starting after
641 * the end of the initial SEQUENCE operation--the rest we regenerate
642 * each time. Therefore we can advertise a ca_maxresponssize_cached
643 * value that is the number of bytes in our cache plus a few additional
644 * bytes. In order to stay on the safe side, and not promise more than
645 * we can cache, those additional bytes must be the minimum possible: 24
646 * bytes of rpc header (xid through accept state, with AUTH_NULL
647 * verifier), 12 for the compound header (with zero-length tag), and 44
648 * for the SEQUENCE op response:
650 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
653 free_session_slots(struct nfsd4_session *ses)
657 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
658 kfree(ses->se_slots[i]);
662 * We don't actually need to cache the rpc and session headers, so we
663 * can allocate a little less for each slot:
665 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
667 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
670 static int nfsd4_sanitize_slot_size(u32 size)
672 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
673 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
679 * XXX: If we run out of reserved DRC memory we could (up to a point)
680 * re-negotiate active sessions and reduce their slot usage to make
681 * room for new connections. For now we just fail the create session.
683 static int nfsd4_get_drc_mem(int slotsize, u32 num)
687 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
689 spin_lock(&nfsd_drc_lock);
690 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
691 nfsd_drc_max_mem - nfsd_drc_mem_used);
692 num = min_t(int, num, avail / slotsize);
693 nfsd_drc_mem_used += num * slotsize;
694 spin_unlock(&nfsd_drc_lock);
699 static void nfsd4_put_drc_mem(int slotsize, int num)
701 spin_lock(&nfsd_drc_lock);
702 nfsd_drc_mem_used -= slotsize * num;
703 spin_unlock(&nfsd_drc_lock);
706 static struct nfsd4_session *__alloc_session(int slotsize, int numslots)
708 struct nfsd4_session *new;
711 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
712 + sizeof(struct nfsd4_session) > PAGE_SIZE);
713 mem = numslots * sizeof(struct nfsd4_slot *);
715 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
718 /* allocate each struct nfsd4_slot and data cache in one piece */
719 for (i = 0; i < numslots; i++) {
720 mem = sizeof(struct nfsd4_slot) + slotsize;
721 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
722 if (!new->se_slots[i])
728 kfree(new->se_slots[i]);
733 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new,
734 struct nfsd4_channel_attrs *req,
735 int numslots, int slotsize,
738 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
740 new->maxreqs = numslots;
741 new->maxresp_cached = min_t(u32, req->maxresp_cached,
742 slotsize + NFSD_MIN_HDR_SEQ_SZ);
743 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
744 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
745 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
748 static void free_conn(struct nfsd4_conn *c)
750 svc_xprt_put(c->cn_xprt);
754 static void nfsd4_conn_lost(struct svc_xpt_user *u)
756 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
757 struct nfs4_client *clp = c->cn_session->se_client;
759 spin_lock(&clp->cl_lock);
760 if (!list_empty(&c->cn_persession)) {
761 list_del(&c->cn_persession);
764 nfsd4_probe_callback(clp);
765 spin_unlock(&clp->cl_lock);
768 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
770 struct nfsd4_conn *conn;
772 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
775 svc_xprt_get(rqstp->rq_xprt);
776 conn->cn_xprt = rqstp->rq_xprt;
777 conn->cn_flags = flags;
778 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
782 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
784 conn->cn_session = ses;
785 list_add(&conn->cn_persession, &ses->se_conns);
788 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
790 struct nfs4_client *clp = ses->se_client;
792 spin_lock(&clp->cl_lock);
793 __nfsd4_hash_conn(conn, ses);
794 spin_unlock(&clp->cl_lock);
797 static int nfsd4_register_conn(struct nfsd4_conn *conn)
799 conn->cn_xpt_user.callback = nfsd4_conn_lost;
800 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
803 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
807 nfsd4_hash_conn(conn, ses);
808 ret = nfsd4_register_conn(conn);
810 /* oops; xprt is already down: */
811 nfsd4_conn_lost(&conn->cn_xpt_user);
812 if (conn->cn_flags & NFS4_CDFC4_BACK) {
813 /* callback channel may be back up */
814 nfsd4_probe_callback(ses->se_client);
818 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
820 u32 dir = NFS4_CDFC4_FORE;
822 if (cses->flags & SESSION4_BACK_CHAN)
823 dir |= NFS4_CDFC4_BACK;
824 return alloc_conn(rqstp, dir);
827 /* must be called under client_lock */
828 static void nfsd4_del_conns(struct nfsd4_session *s)
830 struct nfs4_client *clp = s->se_client;
831 struct nfsd4_conn *c;
833 spin_lock(&clp->cl_lock);
834 while (!list_empty(&s->se_conns)) {
835 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
836 list_del_init(&c->cn_persession);
837 spin_unlock(&clp->cl_lock);
839 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
842 spin_lock(&clp->cl_lock);
844 spin_unlock(&clp->cl_lock);
847 static void __free_session(struct nfsd4_session *ses)
849 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
850 free_session_slots(ses);
854 static void free_session(struct kref *kref)
856 struct nfsd4_session *ses;
859 ses = container_of(kref, struct nfsd4_session, se_ref);
860 nn = net_generic(ses->se_client->net, nfsd_net_id);
862 lockdep_assert_held(&nn->client_lock);
863 nfsd4_del_conns(ses);
867 static void nfsd4_put_session(struct nfsd4_session *ses)
869 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
871 spin_lock(&nn->client_lock);
872 nfsd4_put_session_locked(ses);
873 spin_unlock(&nn->client_lock);
876 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fchan,
879 struct nfsd4_session *new;
880 int numslots, slotsize;
882 * Note decreasing slot size below client's request may
883 * make it difficult for client to function correctly, whereas
884 * decreasing the number of slots will (just?) affect
885 * performance. When short on memory we therefore prefer to
886 * decrease number of slots instead of their size.
888 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
889 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
893 new = __alloc_session(slotsize, numslots);
895 nfsd4_put_drc_mem(slotsize, numslots);
898 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize, nn);
902 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
905 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
907 new->se_client = clp;
910 INIT_LIST_HEAD(&new->se_conns);
912 new->se_cb_seq_nr = 1;
913 new->se_flags = cses->flags;
914 new->se_cb_prog = cses->callback_prog;
915 new->se_cb_sec = cses->cb_sec;
916 kref_init(&new->se_ref);
917 idx = hash_sessionid(&new->se_sessionid);
918 spin_lock(&nn->client_lock);
919 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
920 spin_lock(&clp->cl_lock);
921 list_add(&new->se_perclnt, &clp->cl_sessions);
922 spin_unlock(&clp->cl_lock);
923 spin_unlock(&nn->client_lock);
925 if (cses->flags & SESSION4_BACK_CHAN) {
926 struct sockaddr *sa = svc_addr(rqstp);
928 * This is a little silly; with sessions there's no real
929 * use for the callback address. Use the peer address
930 * as a reasonable default for now, but consider fixing
931 * the rpc client not to require an address in the
934 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
935 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
939 /* caller must hold client_lock */
940 static struct nfsd4_session *
941 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
943 struct nfsd4_session *elem;
945 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
947 dump_sessionid(__func__, sessionid);
948 idx = hash_sessionid(sessionid);
949 /* Search in the appropriate list */
950 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
951 if (!memcmp(elem->se_sessionid.data, sessionid->data,
952 NFS4_MAX_SESSIONID_LEN)) {
957 dprintk("%s: session not found\n", __func__);
961 /* caller must hold client_lock */
963 unhash_session(struct nfsd4_session *ses)
965 list_del(&ses->se_hash);
966 spin_lock(&ses->se_client->cl_lock);
967 list_del(&ses->se_perclnt);
968 spin_unlock(&ses->se_client->cl_lock);
971 /* must be called under the client_lock */
973 renew_client_locked(struct nfs4_client *clp)
975 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
977 if (is_client_expired(clp)) {
979 printk("%s: client (clientid %08x/%08x) already expired\n",
981 clp->cl_clientid.cl_boot,
982 clp->cl_clientid.cl_id);
986 dprintk("renewing client (clientid %08x/%08x)\n",
987 clp->cl_clientid.cl_boot,
988 clp->cl_clientid.cl_id);
989 list_move_tail(&clp->cl_lru, &nn->client_lru);
990 clp->cl_time = get_seconds();
994 renew_client(struct nfs4_client *clp)
996 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
998 spin_lock(&nn->client_lock);
999 renew_client_locked(clp);
1000 spin_unlock(&nn->client_lock);
1003 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1005 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1007 if (clid->cl_boot == nn->boot_time)
1009 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1010 clid->cl_boot, clid->cl_id, nn->boot_time);
1015 * XXX Should we use a slab cache ?
1016 * This type of memory management is somewhat inefficient, but we use it
1017 * anyway since SETCLIENTID is not a common operation.
1019 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1021 struct nfs4_client *clp;
1023 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1026 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1027 if (clp->cl_name.data == NULL) {
1031 clp->cl_name.len = name.len;
1036 free_client(struct nfs4_client *clp)
1038 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1040 lockdep_assert_held(&nn->client_lock);
1041 while (!list_empty(&clp->cl_sessions)) {
1042 struct nfsd4_session *ses;
1043 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1045 list_del(&ses->se_perclnt);
1046 nfsd4_put_session_locked(ses);
1048 free_svc_cred(&clp->cl_cred);
1049 kfree(clp->cl_name.data);
1050 idr_destroy(&clp->cl_stateids);
1055 release_session_client(struct nfsd4_session *session)
1057 struct nfs4_client *clp = session->se_client;
1058 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1060 nfsd4_put_session(session);
1061 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
1064 * At this point we also know all sessions have refcnt 1,
1065 * so free_client will delete them all if necessary:
1067 if (is_client_expired(clp))
1070 renew_client_locked(clp);
1071 spin_unlock(&nn->client_lock);
1074 /* must be called under the client_lock */
1076 unhash_client_locked(struct nfs4_client *clp)
1078 struct nfsd4_session *ses;
1080 mark_client_expired(clp);
1081 list_del(&clp->cl_lru);
1082 spin_lock(&clp->cl_lock);
1083 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1084 list_del_init(&ses->se_hash);
1085 spin_unlock(&clp->cl_lock);
1089 destroy_client(struct nfs4_client *clp)
1091 struct nfs4_openowner *oo;
1092 struct nfs4_delegation *dp;
1093 struct list_head reaplist;
1094 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1096 INIT_LIST_HEAD(&reaplist);
1097 spin_lock(&recall_lock);
1098 while (!list_empty(&clp->cl_delegations)) {
1099 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1100 list_del_init(&dp->dl_perclnt);
1101 list_move(&dp->dl_recall_lru, &reaplist);
1103 spin_unlock(&recall_lock);
1104 while (!list_empty(&reaplist)) {
1105 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1106 unhash_delegation(dp);
1108 while (!list_empty(&clp->cl_openowners)) {
1109 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1110 release_openowner(oo);
1112 nfsd4_shutdown_callback(clp);
1113 if (clp->cl_cb_conn.cb_xprt)
1114 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1115 list_del(&clp->cl_idhash);
1116 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1117 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1119 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1120 spin_lock(&nn->client_lock);
1121 unhash_client_locked(clp);
1122 if (atomic_read(&clp->cl_refcount) == 0)
1124 spin_unlock(&nn->client_lock);
1127 static void expire_client(struct nfs4_client *clp)
1129 nfsd4_client_record_remove(clp);
1130 destroy_client(clp);
1133 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1135 memcpy(target->cl_verifier.data, source->data,
1136 sizeof(target->cl_verifier.data));
1139 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1141 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1142 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1145 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1147 if (source->cr_principal) {
1148 target->cr_principal =
1149 kstrdup(source->cr_principal, GFP_KERNEL);
1150 if (target->cr_principal == NULL)
1153 target->cr_principal = NULL;
1154 target->cr_flavor = source->cr_flavor;
1155 target->cr_uid = source->cr_uid;
1156 target->cr_gid = source->cr_gid;
1157 target->cr_group_info = source->cr_group_info;
1158 get_group_info(target->cr_group_info);
1163 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1167 res = o1->len - o2->len;
1170 return (long long)memcmp(o1->data, o2->data, o1->len);
1173 static int same_name(const char *n1, const char *n2)
1175 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1179 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1181 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1185 same_clid(clientid_t *cl1, clientid_t *cl2)
1187 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1190 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1194 if (g1->ngroups != g2->ngroups)
1196 for (i=0; i<g1->ngroups; i++)
1197 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1203 * RFC 3530 language requires clid_inuse be returned when the
1204 * "principal" associated with a requests differs from that previously
1205 * used. We use uid, gid's, and gss principal string as our best
1206 * approximation. We also don't want to allow non-gss use of a client
1207 * established using gss: in theory cr_principal should catch that
1208 * change, but in practice cr_principal can be null even in the gss case
1209 * since gssd doesn't always pass down a principal string.
1211 static bool is_gss_cred(struct svc_cred *cr)
1213 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1214 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1219 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1221 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1222 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1223 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1224 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1226 if (cr1->cr_principal == cr2->cr_principal)
1228 if (!cr1->cr_principal || !cr2->cr_principal)
1230 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1233 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1235 static u32 current_clientid = 1;
1237 clp->cl_clientid.cl_boot = nn->boot_time;
1238 clp->cl_clientid.cl_id = current_clientid++;
1241 static void gen_confirm(struct nfs4_client *clp)
1246 verf[0] = (__be32)get_seconds();
1247 verf[1] = (__be32)i++;
1248 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1251 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1253 struct nfs4_stid *ret;
1255 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1256 if (!ret || !ret->sc_type)
1261 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1263 struct nfs4_stid *s;
1265 s = find_stateid(cl, t);
1268 if (typemask & s->sc_type)
1273 static struct nfs4_client *create_client(struct xdr_netobj name,
1274 struct svc_rqst *rqstp, nfs4_verifier *verf)
1276 struct nfs4_client *clp;
1277 struct sockaddr *sa = svc_addr(rqstp);
1279 struct net *net = SVC_NET(rqstp);
1280 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1282 clp = alloc_client(name);
1286 INIT_LIST_HEAD(&clp->cl_sessions);
1287 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1289 spin_lock(&nn->client_lock);
1291 spin_unlock(&nn->client_lock);
1294 idr_init(&clp->cl_stateids);
1295 atomic_set(&clp->cl_refcount, 0);
1296 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1297 INIT_LIST_HEAD(&clp->cl_idhash);
1298 INIT_LIST_HEAD(&clp->cl_openowners);
1299 INIT_LIST_HEAD(&clp->cl_delegations);
1300 INIT_LIST_HEAD(&clp->cl_lru);
1301 INIT_LIST_HEAD(&clp->cl_callbacks);
1302 spin_lock_init(&clp->cl_lock);
1303 nfsd4_init_callback(&clp->cl_cb_null);
1304 clp->cl_time = get_seconds();
1305 clear_bit(0, &clp->cl_cb_slot_busy);
1306 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1307 copy_verf(clp, verf);
1308 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1310 clp->cl_cb_session = NULL;
1316 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1318 struct rb_node **new = &(root->rb_node), *parent = NULL;
1319 struct nfs4_client *clp;
1322 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1325 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1326 new = &((*new)->rb_left);
1328 new = &((*new)->rb_right);
1331 rb_link_node(&new_clp->cl_namenode, parent, new);
1332 rb_insert_color(&new_clp->cl_namenode, root);
1335 static struct nfs4_client *
1336 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1339 struct rb_node *node = root->rb_node;
1340 struct nfs4_client *clp;
1343 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1344 cmp = compare_blob(&clp->cl_name, name);
1346 node = node->rb_left;
1348 node = node->rb_right;
1356 add_to_unconfirmed(struct nfs4_client *clp)
1358 unsigned int idhashval;
1359 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1361 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1362 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1363 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1364 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1369 move_to_confirmed(struct nfs4_client *clp)
1371 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1372 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1374 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1375 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1376 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1377 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1378 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1382 static struct nfs4_client *
1383 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1385 struct nfs4_client *clp;
1386 unsigned int idhashval = clientid_hashval(clid->cl_id);
1388 list_for_each_entry(clp, &nn->conf_id_hashtbl[idhashval], cl_idhash) {
1389 if (same_clid(&clp->cl_clientid, clid)) {
1390 if ((bool)clp->cl_minorversion != sessions)
1399 static struct nfs4_client *
1400 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1402 struct nfs4_client *clp;
1403 unsigned int idhashval = clientid_hashval(clid->cl_id);
1405 list_for_each_entry(clp, &nn->unconf_id_hashtbl[idhashval], cl_idhash) {
1406 if (same_clid(&clp->cl_clientid, clid)) {
1407 if ((bool)clp->cl_minorversion != sessions)
1415 static bool clp_used_exchangeid(struct nfs4_client *clp)
1417 return clp->cl_exchange_flags != 0;
1420 static struct nfs4_client *
1421 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1423 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1426 static struct nfs4_client *
1427 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1429 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1433 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1435 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1436 struct sockaddr *sa = svc_addr(rqstp);
1437 u32 scopeid = rpc_get_scope_id(sa);
1438 unsigned short expected_family;
1440 /* Currently, we only support tcp and tcp6 for the callback channel */
1441 if (se->se_callback_netid_len == 3 &&
1442 !memcmp(se->se_callback_netid_val, "tcp", 3))
1443 expected_family = AF_INET;
1444 else if (se->se_callback_netid_len == 4 &&
1445 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1446 expected_family = AF_INET6;
1450 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1451 se->se_callback_addr_len,
1452 (struct sockaddr *)&conn->cb_addr,
1453 sizeof(conn->cb_addr));
1455 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1458 if (conn->cb_addr.ss_family == AF_INET6)
1459 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1461 conn->cb_prog = se->se_callback_prog;
1462 conn->cb_ident = se->se_callback_ident;
1463 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1466 conn->cb_addr.ss_family = AF_UNSPEC;
1467 conn->cb_addrlen = 0;
1468 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1469 "will not receive delegations\n",
1470 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1476 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1479 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1481 struct nfsd4_slot *slot = resp->cstate.slot;
1484 dprintk("--> %s slot %p\n", __func__, slot);
1486 slot->sl_opcnt = resp->opcnt;
1487 slot->sl_status = resp->cstate.status;
1489 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1490 if (nfsd4_not_cached(resp)) {
1491 slot->sl_datalen = 0;
1494 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1495 base = (char *)resp->cstate.datap -
1496 (char *)resp->xbuf->head[0].iov_base;
1497 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1499 WARN("%s: sessions DRC could not cache compound\n", __func__);
1504 * Encode the replay sequence operation from the slot values.
1505 * If cachethis is FALSE encode the uncached rep error on the next
1506 * operation which sets resp->p and increments resp->opcnt for
1507 * nfs4svc_encode_compoundres.
1511 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1512 struct nfsd4_compoundres *resp)
1514 struct nfsd4_op *op;
1515 struct nfsd4_slot *slot = resp->cstate.slot;
1517 /* Encode the replayed sequence operation */
1518 op = &args->ops[resp->opcnt - 1];
1519 nfsd4_encode_operation(resp, op);
1521 /* Return nfserr_retry_uncached_rep in next operation. */
1522 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1523 op = &args->ops[resp->opcnt++];
1524 op->status = nfserr_retry_uncached_rep;
1525 nfsd4_encode_operation(resp, op);
1531 * The sequence operation is not cached because we can use the slot and
1535 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1536 struct nfsd4_sequence *seq)
1538 struct nfsd4_slot *slot = resp->cstate.slot;
1541 dprintk("--> %s slot %p\n", __func__, slot);
1543 /* Either returns 0 or nfserr_retry_uncached */
1544 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1545 if (status == nfserr_retry_uncached_rep)
1548 /* The sequence operation has been encoded, cstate->datap set. */
1549 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1551 resp->opcnt = slot->sl_opcnt;
1552 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1553 status = slot->sl_status;
1559 * Set the exchange_id flags returned by the server.
1562 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1564 /* pNFS is not supported */
1565 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1567 /* Referrals are supported, Migration is not. */
1568 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1570 /* set the wire flags to return to client. */
1571 clid->flags = new->cl_exchange_flags;
1574 static bool client_has_state(struct nfs4_client *clp)
1577 * Note clp->cl_openowners check isn't quite right: there's no
1578 * need to count owners without stateid's.
1580 * Also note we should probably be using this in 4.0 case too.
1582 return !list_empty(&clp->cl_openowners)
1583 || !list_empty(&clp->cl_delegations)
1584 || !list_empty(&clp->cl_sessions);
1588 nfsd4_exchange_id(struct svc_rqst *rqstp,
1589 struct nfsd4_compound_state *cstate,
1590 struct nfsd4_exchange_id *exid)
1592 struct nfs4_client *unconf, *conf, *new;
1594 char addr_str[INET6_ADDRSTRLEN];
1595 nfs4_verifier verf = exid->verifier;
1596 struct sockaddr *sa = svc_addr(rqstp);
1597 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1598 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1600 rpc_ntop(sa, addr_str, sizeof(addr_str));
1601 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1602 "ip_addr=%s flags %x, spa_how %d\n",
1603 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1604 addr_str, exid->flags, exid->spa_how);
1606 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1607 return nfserr_inval;
1609 /* Currently only support SP4_NONE */
1610 switch (exid->spa_how) {
1613 default: /* checked by xdr code */
1617 return nfserr_serverfault; /* no excuse :-/ */
1620 /* Cases below refer to rfc 5661 section 18.35.4: */
1622 conf = find_confirmed_client_by_name(&exid->clname, nn);
1624 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1625 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1628 if (!clp_used_exchangeid(conf)) { /* buggy client */
1629 status = nfserr_inval;
1632 if (!creds_match) { /* case 9 */
1633 status = nfserr_perm;
1636 if (!verfs_match) { /* case 8 */
1637 status = nfserr_not_same;
1641 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1645 if (!creds_match) { /* case 3 */
1646 if (client_has_state(conf)) {
1647 status = nfserr_clid_inuse;
1650 expire_client(conf);
1653 if (verfs_match) { /* case 2 */
1654 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1658 /* case 5, client reboot */
1662 if (update) { /* case 7 */
1663 status = nfserr_noent;
1667 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1668 if (unconf) /* case 4, possible retry or client restart */
1669 expire_client(unconf);
1671 /* case 1 (normal case) */
1673 new = create_client(exid->clname, rqstp, &verf);
1675 status = nfserr_jukebox;
1678 new->cl_minorversion = 1;
1681 add_to_unconfirmed(new);
1683 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1684 exid->clientid.cl_id = new->cl_clientid.cl_id;
1686 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1687 nfsd4_set_ex_flags(new, exid);
1689 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1690 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1694 nfs4_unlock_state();
1699 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1701 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1704 /* The slot is in use, and no response has been sent. */
1706 if (seqid == slot_seqid)
1707 return nfserr_jukebox;
1709 return nfserr_seq_misordered;
1711 /* Note unsigned 32-bit arithmetic handles wraparound: */
1712 if (likely(seqid == slot_seqid + 1))
1714 if (seqid == slot_seqid)
1715 return nfserr_replay_cache;
1716 return nfserr_seq_misordered;
1720 * Cache the create session result into the create session single DRC
1721 * slot cache by saving the xdr structure. sl_seqid has been set.
1722 * Do this for solo or embedded create session operations.
1725 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1726 struct nfsd4_clid_slot *slot, __be32 nfserr)
1728 slot->sl_status = nfserr;
1729 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1733 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1734 struct nfsd4_clid_slot *slot)
1736 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1737 return slot->sl_status;
1740 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1741 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1742 1 + /* MIN tag is length with zero, only length */ \
1743 3 + /* version, opcount, opcode */ \
1744 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1745 /* seqid, slotID, slotID, cache */ \
1746 4 ) * sizeof(__be32))
1748 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1749 2 + /* verifier: AUTH_NULL, length 0 */\
1751 1 + /* MIN tag is length with zero, only length */ \
1752 3 + /* opcount, opcode, opstatus*/ \
1753 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1754 /* seqid, slotID, slotID, slotID, status */ \
1755 5 ) * sizeof(__be32))
1757 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1759 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1760 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1764 nfsd4_create_session(struct svc_rqst *rqstp,
1765 struct nfsd4_compound_state *cstate,
1766 struct nfsd4_create_session *cr_ses)
1768 struct sockaddr *sa = svc_addr(rqstp);
1769 struct nfs4_client *conf, *unconf;
1770 struct nfsd4_session *new;
1771 struct nfsd4_conn *conn;
1772 struct nfsd4_clid_slot *cs_slot = NULL;
1774 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1776 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1777 return nfserr_inval;
1778 if (check_forechannel_attrs(cr_ses->fore_channel))
1779 return nfserr_toosmall;
1780 new = alloc_session(&cr_ses->fore_channel, nn);
1782 return nfserr_jukebox;
1783 status = nfserr_jukebox;
1784 conn = alloc_conn_from_crses(rqstp, cr_ses);
1786 goto out_free_session;
1789 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1790 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1791 WARN_ON_ONCE(conf && unconf);
1794 cs_slot = &conf->cl_cs_slot;
1795 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1796 if (status == nfserr_replay_cache) {
1797 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1799 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1800 status = nfserr_seq_misordered;
1803 } else if (unconf) {
1804 struct nfs4_client *old;
1805 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1806 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1807 status = nfserr_clid_inuse;
1810 cs_slot = &unconf->cl_cs_slot;
1811 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1813 /* an unconfirmed replay returns misordered */
1814 status = nfserr_seq_misordered;
1817 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1820 move_to_confirmed(unconf);
1823 status = nfserr_stale_clientid;
1828 * We do not support RDMA or persistent sessions
1830 cr_ses->flags &= ~SESSION4_PERSIST;
1831 cr_ses->flags &= ~SESSION4_RDMA;
1833 init_session(rqstp, new, conf, cr_ses);
1834 nfsd4_init_conn(rqstp, conn, new);
1836 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1837 NFS4_MAX_SESSIONID_LEN);
1838 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1839 sizeof(struct nfsd4_channel_attrs));
1840 cs_slot->sl_seqid++;
1841 cr_ses->seqid = cs_slot->sl_seqid;
1843 /* cache solo and embedded create sessions under the state lock */
1844 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1845 nfs4_unlock_state();
1848 nfs4_unlock_state();
1851 __free_session(new);
1855 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1858 case NFS4_CDFC4_FORE:
1859 case NFS4_CDFC4_BACK:
1861 case NFS4_CDFC4_FORE_OR_BOTH:
1862 case NFS4_CDFC4_BACK_OR_BOTH:
1863 *dir = NFS4_CDFC4_BOTH;
1866 return nfserr_inval;
1869 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1871 struct nfsd4_session *session = cstate->session;
1872 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1874 spin_lock(&nn->client_lock);
1875 session->se_cb_prog = bc->bc_cb_program;
1876 session->se_cb_sec = bc->bc_cb_sec;
1877 spin_unlock(&nn->client_lock);
1879 nfsd4_probe_callback(session->se_client);
1884 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1885 struct nfsd4_compound_state *cstate,
1886 struct nfsd4_bind_conn_to_session *bcts)
1889 struct nfsd4_conn *conn;
1890 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1892 if (!nfsd4_last_compound_op(rqstp))
1893 return nfserr_not_only_op;
1894 spin_lock(&nn->client_lock);
1895 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
1896 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1897 * client_lock iself: */
1898 if (cstate->session) {
1899 nfsd4_get_session(cstate->session);
1900 atomic_inc(&cstate->session->se_client->cl_refcount);
1902 spin_unlock(&nn->client_lock);
1903 if (!cstate->session)
1904 return nfserr_badsession;
1906 status = nfsd4_map_bcts_dir(&bcts->dir);
1909 conn = alloc_conn(rqstp, bcts->dir);
1911 return nfserr_jukebox;
1912 nfsd4_init_conn(rqstp, conn, cstate->session);
1916 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1920 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1924 nfsd4_destroy_session(struct svc_rqst *r,
1925 struct nfsd4_compound_state *cstate,
1926 struct nfsd4_destroy_session *sessionid)
1928 struct nfsd4_session *ses;
1929 __be32 status = nfserr_badsession;
1930 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
1933 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1934 * - Should we return nfserr_back_chan_busy if waiting for
1935 * callbacks on to-be-destroyed session?
1936 * - Do we need to clear any callback info from previous session?
1939 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1940 if (!nfsd4_last_compound_op(r))
1941 return nfserr_not_only_op;
1943 dump_sessionid(__func__, &sessionid->sessionid);
1944 spin_lock(&nn->client_lock);
1945 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
1947 spin_unlock(&nn->client_lock);
1951 unhash_session(ses);
1952 spin_unlock(&nn->client_lock);
1955 nfsd4_probe_callback_sync(ses->se_client);
1956 nfs4_unlock_state();
1958 spin_lock(&nn->client_lock);
1959 nfsd4_del_conns(ses);
1960 nfsd4_put_session_locked(ses);
1961 spin_unlock(&nn->client_lock);
1967 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1969 struct nfsd4_conn *c;
1971 list_for_each_entry(c, &s->se_conns, cn_persession) {
1972 if (c->cn_xprt == xpt) {
1979 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1981 struct nfs4_client *clp = ses->se_client;
1982 struct nfsd4_conn *c;
1985 spin_lock(&clp->cl_lock);
1986 c = __nfsd4_find_conn(new->cn_xprt, ses);
1988 spin_unlock(&clp->cl_lock);
1992 __nfsd4_hash_conn(new, ses);
1993 spin_unlock(&clp->cl_lock);
1994 ret = nfsd4_register_conn(new);
1996 /* oops; xprt is already down: */
1997 nfsd4_conn_lost(&new->cn_xpt_user);
2001 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2003 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2005 return args->opcnt > session->se_fchannel.maxops;
2008 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2009 struct nfsd4_session *session)
2011 struct xdr_buf *xb = &rqstp->rq_arg;
2013 return xb->len > session->se_fchannel.maxreq_sz;
2017 nfsd4_sequence(struct svc_rqst *rqstp,
2018 struct nfsd4_compound_state *cstate,
2019 struct nfsd4_sequence *seq)
2021 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2022 struct nfsd4_session *session;
2023 struct nfsd4_slot *slot;
2024 struct nfsd4_conn *conn;
2026 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2028 if (resp->opcnt != 1)
2029 return nfserr_sequence_pos;
2032 * Will be either used or freed by nfsd4_sequence_check_conn
2035 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2037 return nfserr_jukebox;
2039 spin_lock(&nn->client_lock);
2040 status = nfserr_badsession;
2041 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2045 status = nfserr_too_many_ops;
2046 if (nfsd4_session_too_many_ops(rqstp, session))
2049 status = nfserr_req_too_big;
2050 if (nfsd4_request_too_big(rqstp, session))
2053 status = nfserr_badslot;
2054 if (seq->slotid >= session->se_fchannel.maxreqs)
2057 slot = session->se_slots[seq->slotid];
2058 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2060 /* We do not negotiate the number of slots yet, so set the
2061 * maxslots to the session maxreqs which is used to encode
2062 * sr_highest_slotid and the sr_target_slot id to maxslots */
2063 seq->maxslots = session->se_fchannel.maxreqs;
2065 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2066 slot->sl_flags & NFSD4_SLOT_INUSE);
2067 if (status == nfserr_replay_cache) {
2068 status = nfserr_seq_misordered;
2069 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2071 cstate->slot = slot;
2072 cstate->session = session;
2073 /* Return the cached reply status and set cstate->status
2074 * for nfsd4_proc_compound processing */
2075 status = nfsd4_replay_cache_entry(resp, seq);
2076 cstate->status = nfserr_replay_cache;
2082 nfsd4_sequence_check_conn(conn, session);
2085 /* Success! bump slot seqid */
2086 slot->sl_seqid = seq->seqid;
2087 slot->sl_flags |= NFSD4_SLOT_INUSE;
2089 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2091 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2093 cstate->slot = slot;
2094 cstate->session = session;
2097 /* Hold a session reference until done processing the compound. */
2098 if (cstate->session) {
2099 struct nfs4_client *clp = session->se_client;
2101 nfsd4_get_session(cstate->session);
2102 atomic_inc(&clp->cl_refcount);
2103 switch (clp->cl_cb_state) {
2105 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2107 case NFSD4_CB_FAULT:
2108 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2111 seq->status_flags = 0;
2115 spin_unlock(&nn->client_lock);
2120 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2122 struct nfs4_client *conf, *unconf, *clp;
2124 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2127 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2128 conf = find_confirmed_client(&dc->clientid, true, nn);
2129 WARN_ON_ONCE(conf && unconf);
2134 if (!is_client_expired(conf) && client_has_state(conf)) {
2135 status = nfserr_clientid_busy;
2139 /* rfc5661 18.50.3 */
2140 if (cstate->session && conf == cstate->session->se_client) {
2141 status = nfserr_clientid_busy;
2147 status = nfserr_stale_clientid;
2153 nfs4_unlock_state();
2158 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2162 if (rc->rca_one_fs) {
2163 if (!cstate->current_fh.fh_dentry)
2164 return nfserr_nofilehandle;
2166 * We don't take advantage of the rca_one_fs case.
2167 * That's OK, it's optional, we can safely ignore it.
2173 status = nfserr_complete_already;
2174 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2175 &cstate->session->se_client->cl_flags))
2178 status = nfserr_stale_clientid;
2179 if (is_client_expired(cstate->session->se_client))
2181 * The following error isn't really legal.
2182 * But we only get here if the client just explicitly
2183 * destroyed the client. Surely it no longer cares what
2184 * error it gets back on an operation for the dead
2190 nfsd4_client_record_create(cstate->session->se_client);
2192 nfs4_unlock_state();
2197 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2198 struct nfsd4_setclientid *setclid)
2200 struct xdr_netobj clname = setclid->se_name;
2201 nfs4_verifier clverifier = setclid->se_verf;
2202 struct nfs4_client *conf, *unconf, *new;
2204 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2206 /* Cases below refer to rfc 3530 section 14.2.33: */
2208 conf = find_confirmed_client_by_name(&clname, nn);
2211 status = nfserr_clid_inuse;
2212 if (clp_used_exchangeid(conf))
2214 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2215 char addr_str[INET6_ADDRSTRLEN];
2216 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2218 dprintk("NFSD: setclientid: string in use by client "
2219 "at %s\n", addr_str);
2223 unconf = find_unconfirmed_client_by_name(&clname, nn);
2225 expire_client(unconf);
2226 status = nfserr_jukebox;
2227 new = create_client(clname, rqstp, &clverifier);
2230 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2231 /* case 1: probable callback update */
2232 copy_clid(new, conf);
2233 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2235 new->cl_minorversion = 0;
2236 gen_callback(new, setclid, rqstp);
2237 add_to_unconfirmed(new);
2238 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2239 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2240 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2243 nfs4_unlock_state();
2249 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2250 struct nfsd4_compound_state *cstate,
2251 struct nfsd4_setclientid_confirm *setclientid_confirm)
2253 struct nfs4_client *conf, *unconf;
2254 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2255 clientid_t * clid = &setclientid_confirm->sc_clientid;
2257 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2259 if (STALE_CLIENTID(clid, nn))
2260 return nfserr_stale_clientid;
2263 conf = find_confirmed_client(clid, false, nn);
2264 unconf = find_unconfirmed_client(clid, false, nn);
2266 * We try hard to give out unique clientid's, so if we get an
2267 * attempt to confirm the same clientid with a different cred,
2268 * there's a bug somewhere. Let's charitably assume it's our
2271 status = nfserr_serverfault;
2272 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2274 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2276 /* cases below refer to rfc 3530 section 14.2.34: */
2277 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2278 if (conf && !unconf) /* case 2: probable retransmit */
2280 else /* case 4: client hasn't noticed we rebooted yet? */
2281 status = nfserr_stale_clientid;
2285 if (conf) { /* case 1: callback update */
2286 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2287 nfsd4_probe_callback(conf);
2288 expire_client(unconf);
2289 } else { /* case 3: normal case; new or rebooted client */
2290 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2292 expire_client(conf);
2293 move_to_confirmed(unconf);
2294 nfsd4_probe_callback(unconf);
2297 nfs4_unlock_state();
2301 static struct nfs4_file *nfsd4_alloc_file(void)
2303 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2306 /* OPEN Share state helper functions */
2307 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2309 unsigned int hashval = file_hashval(ino);
2311 atomic_set(&fp->fi_ref, 1);
2312 INIT_LIST_HEAD(&fp->fi_hash);
2313 INIT_LIST_HEAD(&fp->fi_stateids);
2314 INIT_LIST_HEAD(&fp->fi_delegations);
2315 fp->fi_inode = igrab(ino);
2316 fp->fi_had_conflict = false;
2317 fp->fi_lease = NULL;
2318 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2319 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2320 spin_lock(&recall_lock);
2321 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2322 spin_unlock(&recall_lock);
2326 nfsd4_free_slab(struct kmem_cache **slab)
2330 kmem_cache_destroy(*slab);
2335 nfsd4_free_slabs(void)
2337 nfsd4_free_slab(&openowner_slab);
2338 nfsd4_free_slab(&lockowner_slab);
2339 nfsd4_free_slab(&file_slab);
2340 nfsd4_free_slab(&stateid_slab);
2341 nfsd4_free_slab(&deleg_slab);
2345 nfsd4_init_slabs(void)
2347 openowner_slab = kmem_cache_create("nfsd4_openowners",
2348 sizeof(struct nfs4_openowner), 0, 0, NULL);
2349 if (openowner_slab == NULL)
2351 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2352 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2353 if (lockowner_slab == NULL)
2355 file_slab = kmem_cache_create("nfsd4_files",
2356 sizeof(struct nfs4_file), 0, 0, NULL);
2357 if (file_slab == NULL)
2359 stateid_slab = kmem_cache_create("nfsd4_stateids",
2360 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2361 if (stateid_slab == NULL)
2363 deleg_slab = kmem_cache_create("nfsd4_delegations",
2364 sizeof(struct nfs4_delegation), 0, 0, NULL);
2365 if (deleg_slab == NULL)
2370 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2374 void nfs4_free_openowner(struct nfs4_openowner *oo)
2376 kfree(oo->oo_owner.so_owner.data);
2377 kmem_cache_free(openowner_slab, oo);
2380 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2382 kfree(lo->lo_owner.so_owner.data);
2383 kmem_cache_free(lockowner_slab, lo);
2386 static void init_nfs4_replay(struct nfs4_replay *rp)
2388 rp->rp_status = nfserr_serverfault;
2390 rp->rp_buf = rp->rp_ibuf;
2393 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2395 struct nfs4_stateowner *sop;
2397 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2401 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2402 if (!sop->so_owner.data) {
2403 kmem_cache_free(slab, sop);
2406 sop->so_owner.len = owner->len;
2408 INIT_LIST_HEAD(&sop->so_stateids);
2409 sop->so_client = clp;
2410 init_nfs4_replay(&sop->so_replay);
2414 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2416 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2418 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2419 list_add(&oo->oo_perclient, &clp->cl_openowners);
2422 static struct nfs4_openowner *
2423 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2424 struct nfs4_openowner *oo;
2426 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2429 oo->oo_owner.so_is_open_owner = 1;
2430 oo->oo_owner.so_seqid = open->op_seqid;
2431 oo->oo_flags = NFS4_OO_NEW;
2433 oo->oo_last_closed_stid = NULL;
2434 INIT_LIST_HEAD(&oo->oo_close_lru);
2435 hash_openowner(oo, clp, strhashval);
2439 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2440 struct nfs4_openowner *oo = open->op_openowner;
2442 stp->st_stid.sc_type = NFS4_OPEN_STID;
2443 INIT_LIST_HEAD(&stp->st_lockowners);
2444 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2445 list_add(&stp->st_perfile, &fp->fi_stateids);
2446 stp->st_stateowner = &oo->oo_owner;
2449 stp->st_access_bmap = 0;
2450 stp->st_deny_bmap = 0;
2451 set_access(open->op_share_access, stp);
2452 set_deny(open->op_share_deny, stp);
2453 stp->st_openstp = NULL;
2457 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2459 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2461 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2463 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2464 oo->oo_time = get_seconds();
2468 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2471 return (sop->so_owner.len == owner->len) &&
2472 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2473 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2476 static struct nfs4_openowner *
2477 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2478 bool sessions, struct nfsd_net *nn)
2480 struct nfs4_stateowner *so;
2481 struct nfs4_openowner *oo;
2482 struct nfs4_client *clp;
2484 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2485 if (!so->so_is_open_owner)
2487 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2489 clp = oo->oo_owner.so_client;
2490 if ((bool)clp->cl_minorversion != sessions)
2492 renew_client(oo->oo_owner.so_client);
2499 /* search file_hashtbl[] for file */
2500 static struct nfs4_file *
2501 find_file(struct inode *ino)
2503 unsigned int hashval = file_hashval(ino);
2504 struct nfs4_file *fp;
2506 spin_lock(&recall_lock);
2507 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2508 if (fp->fi_inode == ino) {
2510 spin_unlock(&recall_lock);
2514 spin_unlock(&recall_lock);
2519 * Called to check deny when READ with all zero stateid or
2520 * WRITE with all zero or all one stateid
2523 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2525 struct inode *ino = current_fh->fh_dentry->d_inode;
2526 struct nfs4_file *fp;
2527 struct nfs4_ol_stateid *stp;
2530 fp = find_file(ino);
2533 ret = nfserr_locked;
2534 /* Search for conflicting share reservations */
2535 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2536 if (test_deny(deny_type, stp) ||
2537 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2546 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2548 /* We're assuming the state code never drops its reference
2549 * without first removing the lease. Since we're in this lease
2550 * callback (and since the lease code is serialized by the kernel
2551 * lock) we know the server hasn't removed the lease yet, we know
2552 * it's safe to take a reference: */
2553 atomic_inc(&dp->dl_count);
2555 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2557 /* only place dl_time is set. protected by lock_flocks*/
2558 dp->dl_time = get_seconds();
2560 nfsd4_cb_recall(dp);
2563 /* Called from break_lease() with lock_flocks() held. */
2564 static void nfsd_break_deleg_cb(struct file_lock *fl)
2566 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2567 struct nfs4_delegation *dp;
2570 WARN(1, "(%p)->fl_owner NULL\n", fl);
2573 if (fp->fi_had_conflict) {
2574 WARN(1, "duplicate break on %p\n", fp);
2578 * We don't want the locks code to timeout the lease for us;
2579 * we'll remove it ourself if a delegation isn't returned
2582 fl->fl_break_time = 0;
2584 spin_lock(&recall_lock);
2585 fp->fi_had_conflict = true;
2586 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2587 nfsd_break_one_deleg(dp);
2588 spin_unlock(&recall_lock);
2592 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2595 return lease_modify(onlist, arg);
2600 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2601 .lm_break = nfsd_break_deleg_cb,
2602 .lm_change = nfsd_change_deleg_cb,
2605 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2607 if (nfsd4_has_session(cstate))
2609 if (seqid == so->so_seqid - 1)
2610 return nfserr_replay_me;
2611 if (seqid == so->so_seqid)
2613 return nfserr_bad_seqid;
2617 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2618 struct nfsd4_open *open, struct nfsd_net *nn)
2620 clientid_t *clientid = &open->op_clientid;
2621 struct nfs4_client *clp = NULL;
2622 unsigned int strhashval;
2623 struct nfs4_openowner *oo = NULL;
2626 if (STALE_CLIENTID(&open->op_clientid, nn))
2627 return nfserr_stale_clientid;
2629 * In case we need it later, after we've already created the
2630 * file and don't want to risk a further failure:
2632 open->op_file = nfsd4_alloc_file();
2633 if (open->op_file == NULL)
2634 return nfserr_jukebox;
2636 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2637 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2638 open->op_openowner = oo;
2640 clp = find_confirmed_client(clientid, cstate->minorversion,
2643 return nfserr_expired;
2646 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2647 /* Replace unconfirmed owners without checking for replay. */
2648 clp = oo->oo_owner.so_client;
2649 release_openowner(oo);
2650 open->op_openowner = NULL;
2653 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2656 clp = oo->oo_owner.so_client;
2659 oo = alloc_init_open_stateowner(strhashval, clp, open);
2661 return nfserr_jukebox;
2662 open->op_openowner = oo;
2664 open->op_stp = nfs4_alloc_stateid(clp);
2666 return nfserr_jukebox;
2670 static inline __be32
2671 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2673 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2674 return nfserr_openmode;
2679 static int share_access_to_flags(u32 share_access)
2681 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2684 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2686 struct nfs4_stid *ret;
2688 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2691 return delegstateid(ret);
2694 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2696 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2697 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2701 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2702 struct nfs4_delegation **dp)
2705 __be32 status = nfserr_bad_stateid;
2707 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2710 flags = share_access_to_flags(open->op_share_access);
2711 status = nfs4_check_delegmode(*dp, flags);
2715 if (!nfsd4_is_deleg_cur(open))
2719 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2724 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2726 struct nfs4_ol_stateid *local;
2727 struct nfs4_openowner *oo = open->op_openowner;
2729 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2730 /* ignore lock owners */
2731 if (local->st_stateowner->so_is_open_owner == 0)
2733 /* remember if we have seen this open owner */
2734 if (local->st_stateowner == &oo->oo_owner)
2736 /* check for conflicting share reservations */
2737 if (!test_share(local, open))
2738 return nfserr_share_denied;
2743 static inline int nfs4_access_to_access(u32 nfs4_access)
2747 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2748 flags |= NFSD_MAY_READ;
2749 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2750 flags |= NFSD_MAY_WRITE;
2754 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2755 struct svc_fh *cur_fh, struct nfsd4_open *open)
2758 int oflag = nfs4_access_to_omode(open->op_share_access);
2759 int access = nfs4_access_to_access(open->op_share_access);
2761 if (!fp->fi_fds[oflag]) {
2762 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2763 &fp->fi_fds[oflag]);
2767 nfs4_file_get_access(fp, oflag);
2772 static inline __be32
2773 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2774 struct nfsd4_open *open)
2776 struct iattr iattr = {
2777 .ia_valid = ATTR_SIZE,
2780 if (!open->op_truncate)
2782 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2783 return nfserr_inval;
2784 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2788 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)
2790 u32 op_share_access = open->op_share_access;
2794 new_access = !test_access(op_share_access, stp);
2796 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2800 status = nfsd4_truncate(rqstp, cur_fh, open);
2803 int oflag = nfs4_access_to_omode(op_share_access);
2804 nfs4_file_put_access(fp, oflag);
2808 /* remember the open */
2809 set_access(op_share_access, stp);
2810 set_deny(open->op_share_deny, stp);
2817 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2819 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2822 /* Should we give out recallable state?: */
2823 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2825 if (clp->cl_cb_state == NFSD4_CB_UP)
2828 * In the sessions case, since we don't have to establish a
2829 * separate connection for callbacks, we assume it's OK
2830 * until we hear otherwise:
2832 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2835 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2837 struct file_lock *fl;
2839 fl = locks_alloc_lock();
2842 locks_init_lock(fl);
2843 fl->fl_lmops = &nfsd_lease_mng_ops;
2844 fl->fl_flags = FL_LEASE;
2845 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2846 fl->fl_end = OFFSET_MAX;
2847 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2848 fl->fl_pid = current->tgid;
2852 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2854 struct nfs4_file *fp = dp->dl_file;
2855 struct file_lock *fl;
2858 fl = nfs4_alloc_init_lease(dp, flag);
2861 fl->fl_file = find_readable_file(fp);
2862 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2863 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2865 list_del_init(&dp->dl_perclnt);
2866 locks_free_lock(fl);
2870 fp->fi_deleg_file = get_file(fl->fl_file);
2871 atomic_set(&fp->fi_delegees, 1);
2872 list_add(&dp->dl_perfile, &fp->fi_delegations);
2876 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2878 struct nfs4_file *fp = dp->dl_file;
2881 return nfs4_setlease(dp, flag);
2882 spin_lock(&recall_lock);
2883 if (fp->fi_had_conflict) {
2884 spin_unlock(&recall_lock);
2887 atomic_inc(&fp->fi_delegees);
2888 list_add(&dp->dl_perfile, &fp->fi_delegations);
2889 spin_unlock(&recall_lock);
2890 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2894 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2896 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2897 if (status == -EAGAIN)
2898 open->op_why_no_deleg = WND4_CONTENTION;
2900 open->op_why_no_deleg = WND4_RESOURCE;
2901 switch (open->op_deleg_want) {
2902 case NFS4_SHARE_WANT_READ_DELEG:
2903 case NFS4_SHARE_WANT_WRITE_DELEG:
2904 case NFS4_SHARE_WANT_ANY_DELEG:
2906 case NFS4_SHARE_WANT_CANCEL:
2907 open->op_why_no_deleg = WND4_CANCELLED;
2909 case NFS4_SHARE_WANT_NO_DELEG:
2916 * Attempt to hand out a delegation.
2919 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2920 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2922 struct nfs4_delegation *dp;
2923 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2925 int status = 0, flag = 0;
2927 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2928 flag = NFS4_OPEN_DELEGATE_NONE;
2929 open->op_recall = 0;
2930 switch (open->op_claim_type) {
2931 case NFS4_OPEN_CLAIM_PREVIOUS:
2933 open->op_recall = 1;
2934 flag = open->op_delegate_type;
2935 if (flag == NFS4_OPEN_DELEGATE_NONE)
2938 case NFS4_OPEN_CLAIM_NULL:
2939 /* Let's not give out any delegations till everyone's
2940 * had the chance to reclaim theirs.... */
2941 if (locks_in_grace(net))
2943 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2945 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2946 flag = NFS4_OPEN_DELEGATE_WRITE;
2948 flag = NFS4_OPEN_DELEGATE_READ;
2954 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2957 status = nfs4_set_delegation(dp, flag);
2961 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2963 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2964 STATEID_VAL(&dp->dl_stid.sc_stateid));
2966 open->op_delegate_type = flag;
2967 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2968 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2969 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2970 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2972 /* 4.1 client asking for a delegation? */
2973 if (open->op_deleg_want)
2974 nfsd4_open_deleg_none_ext(open, status);
2978 unhash_stid(&dp->dl_stid);
2979 nfs4_put_delegation(dp);
2981 flag = NFS4_OPEN_DELEGATE_NONE;
2985 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2986 struct nfs4_delegation *dp)
2988 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2989 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2990 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2991 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2992 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2993 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2994 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2995 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
2997 /* Otherwise the client must be confused wanting a delegation
2998 * it already has, therefore we don't return
2999 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3004 * called with nfs4_lock_state() held.
3007 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3009 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3010 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3011 struct nfs4_file *fp = NULL;
3012 struct inode *ino = current_fh->fh_dentry->d_inode;
3013 struct nfs4_ol_stateid *stp = NULL;
3014 struct nfs4_delegation *dp = NULL;
3018 * Lookup file; if found, lookup stateid and check open request,
3019 * and check for delegations in the process of being recalled.
3020 * If not found, create the nfs4_file struct
3022 fp = find_file(ino);
3024 if ((status = nfs4_check_open(fp, open, &stp)))
3026 status = nfs4_check_deleg(cl, fp, open, &dp);
3030 status = nfserr_bad_stateid;
3031 if (nfsd4_is_deleg_cur(open))
3033 status = nfserr_jukebox;
3035 open->op_file = NULL;
3036 nfsd4_init_file(fp, ino);
3040 * OPEN the file, or upgrade an existing OPEN.
3041 * If truncate fails, the OPEN fails.
3044 /* Stateid was found, this is an OPEN upgrade */
3045 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3049 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3052 status = nfsd4_truncate(rqstp, current_fh, open);
3056 open->op_stp = NULL;
3057 init_open_stateid(stp, fp, open);
3059 update_stateid(&stp->st_stid.sc_stateid);
3060 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3062 if (nfsd4_has_session(&resp->cstate)) {
3063 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3065 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3066 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3067 open->op_why_no_deleg = WND4_NOT_WANTED;
3073 * Attempt to hand out a delegation. No error return, because the
3074 * OPEN succeeds even if we fail.
3076 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3080 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3081 STATEID_VAL(&stp->st_stid.sc_stateid));
3083 /* 4.1 client trying to upgrade/downgrade delegation? */
3084 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3085 open->op_deleg_want)
3086 nfsd4_deleg_xgrade_none_ext(open, dp);
3090 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3091 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3093 * To finish the open response, we just need to set the rflags.
3095 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3096 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3097 !nfsd4_has_session(&resp->cstate))
3098 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3103 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3105 if (open->op_openowner) {
3106 struct nfs4_openowner *oo = open->op_openowner;
3108 if (!list_empty(&oo->oo_owner.so_stateids))
3109 list_del_init(&oo->oo_close_lru);
3110 if (oo->oo_flags & NFS4_OO_NEW) {
3112 release_openowner(oo);
3113 open->op_openowner = NULL;
3115 oo->oo_flags &= ~NFS4_OO_NEW;
3119 nfsd4_free_file(open->op_file);
3121 free_generic_stateid(open->op_stp);
3124 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3126 struct nfs4_client *found;
3128 if (STALE_CLIENTID(clid, nn))
3129 return nfserr_stale_clientid;
3130 found = find_confirmed_client(clid, session, nn);
3133 return found ? nfs_ok : nfserr_expired;
3137 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3140 struct nfs4_client *clp;
3142 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3145 dprintk("process_renew(%08x/%08x): starting\n",
3146 clid->cl_boot, clid->cl_id);
3147 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3150 status = nfserr_cb_path_down;
3151 if (!list_empty(&clp->cl_delegations)
3152 && clp->cl_cb_state != NFSD4_CB_UP)
3156 nfs4_unlock_state();
3161 nfsd4_end_grace(struct nfsd_net *nn)
3163 /* do nothing if grace period already ended */
3164 if (nn->grace_ended)
3167 dprintk("NFSD: end of grace period\n");
3168 nn->grace_ended = true;
3169 nfsd4_record_grace_done(nn, nn->boot_time);
3170 locks_end_grace(&nn->nfsd4_manager);
3172 * Now that every NFSv4 client has had the chance to recover and
3173 * to see the (possibly new, possibly shorter) lease time, we
3174 * can safely set the next grace time to the current lease time:
3176 nn->nfsd4_grace = nn->nfsd4_lease;
3180 nfs4_laundromat(struct nfsd_net *nn)
3182 struct nfs4_client *clp;
3183 struct nfs4_openowner *oo;
3184 struct nfs4_delegation *dp;
3185 struct list_head *pos, *next, reaplist;
3186 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3187 time_t t, clientid_val = nn->nfsd4_lease;
3188 time_t u, test_val = nn->nfsd4_lease;
3192 dprintk("NFSD: laundromat service - starting\n");
3193 nfsd4_end_grace(nn);
3194 INIT_LIST_HEAD(&reaplist);
3195 spin_lock(&nn->client_lock);
3196 list_for_each_safe(pos, next, &nn->client_lru) {
3197 clp = list_entry(pos, struct nfs4_client, cl_lru);
3198 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3199 t = clp->cl_time - cutoff;
3200 if (clientid_val > t)
3204 if (atomic_read(&clp->cl_refcount)) {
3205 dprintk("NFSD: client in use (clientid %08x)\n",
3206 clp->cl_clientid.cl_id);
3209 unhash_client_locked(clp);
3210 list_add(&clp->cl_lru, &reaplist);
3212 spin_unlock(&nn->client_lock);
3213 list_for_each_safe(pos, next, &reaplist) {
3214 clp = list_entry(pos, struct nfs4_client, cl_lru);
3215 dprintk("NFSD: purging unused client (clientid %08x)\n",
3216 clp->cl_clientid.cl_id);
3219 spin_lock(&recall_lock);
3220 list_for_each_safe(pos, next, &del_recall_lru) {
3221 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3222 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3224 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3225 u = dp->dl_time - cutoff;
3230 list_move(&dp->dl_recall_lru, &reaplist);
3232 spin_unlock(&recall_lock);
3233 list_for_each_safe(pos, next, &reaplist) {
3234 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3235 unhash_delegation(dp);
3237 test_val = nn->nfsd4_lease;
3238 list_for_each_safe(pos, next, &nn->close_lru) {
3239 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3240 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3241 u = oo->oo_time - cutoff;
3246 release_openowner(oo);
3248 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3249 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3250 nfs4_unlock_state();
3251 return clientid_val;
3254 static struct workqueue_struct *laundry_wq;
3255 static void laundromat_main(struct work_struct *);
3258 laundromat_main(struct work_struct *laundry)
3261 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3263 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3266 t = nfs4_laundromat(nn);
3267 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3268 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3271 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3273 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3274 return nfserr_bad_stateid;
3279 access_permit_read(struct nfs4_ol_stateid *stp)
3281 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3282 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3283 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3287 access_permit_write(struct nfs4_ol_stateid *stp)
3289 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3290 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3294 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3296 __be32 status = nfserr_openmode;
3298 /* For lock stateid's, we test the parent open, not the lock: */
3299 if (stp->st_openstp)
3300 stp = stp->st_openstp;
3301 if ((flags & WR_STATE) && !access_permit_write(stp))
3303 if ((flags & RD_STATE) && !access_permit_read(stp))
3310 static inline __be32
3311 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3313 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3315 else if (locks_in_grace(net)) {
3316 /* Answer in remaining cases depends on existence of
3317 * conflicting state; so we must wait out the grace period. */
3318 return nfserr_grace;
3319 } else if (flags & WR_STATE)
3320 return nfs4_share_conflict(current_fh,
3321 NFS4_SHARE_DENY_WRITE);
3322 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3323 return nfs4_share_conflict(current_fh,
3324 NFS4_SHARE_DENY_READ);
3328 * Allow READ/WRITE during grace period on recovered state only for files
3329 * that are not able to provide mandatory locking.
3332 grace_disallows_io(struct net *net, struct inode *inode)
3334 return locks_in_grace(net) && mandatory_lock(inode);
3337 /* Returns true iff a is later than b: */
3338 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3340 return (s32)a->si_generation - (s32)b->si_generation > 0;
3343 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3346 * When sessions are used the stateid generation number is ignored
3349 if (has_session && in->si_generation == 0)
3352 if (in->si_generation == ref->si_generation)
3355 /* If the client sends us a stateid from the future, it's buggy: */
3356 if (stateid_generation_after(in, ref))
3357 return nfserr_bad_stateid;
3359 * However, we could see a stateid from the past, even from a
3360 * non-buggy client. For example, if the client sends a lock
3361 * while some IO is outstanding, the lock may bump si_generation
3362 * while the IO is still in flight. The client could avoid that
3363 * situation by waiting for responses on all the IO requests,
3364 * but better performance may result in retrying IO that
3365 * receives an old_stateid error if requests are rarely
3366 * reordered in flight:
3368 return nfserr_old_stateid;
3371 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3373 struct nfs4_stid *s;
3374 struct nfs4_ol_stateid *ols;
3377 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3378 return nfserr_bad_stateid;
3379 /* Client debugging aid. */
3380 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3381 char addr_str[INET6_ADDRSTRLEN];
3382 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3384 pr_warn_ratelimited("NFSD: client %s testing state ID "
3385 "with incorrect client ID\n", addr_str);
3386 return nfserr_bad_stateid;
3388 s = find_stateid(cl, stateid);
3390 return nfserr_bad_stateid;
3391 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3394 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3396 ols = openlockstateid(s);
3397 if (ols->st_stateowner->so_is_open_owner
3398 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3399 return nfserr_bad_stateid;
3403 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3404 struct nfs4_stid **s, bool sessions,
3405 struct nfsd_net *nn)
3407 struct nfs4_client *cl;
3410 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3411 return nfserr_bad_stateid;
3412 status = lookup_clientid(&stateid->si_opaque.so_clid, sessions,
3414 if (status == nfserr_stale_clientid)
3415 return nfserr_stale_stateid;
3418 *s = find_stateid_by_type(cl, stateid, typemask);
3420 return nfserr_bad_stateid;
3425 * Checks for stateid operations
3428 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3429 stateid_t *stateid, int flags, struct file **filpp)
3431 struct nfs4_stid *s;
3432 struct nfs4_ol_stateid *stp = NULL;
3433 struct nfs4_delegation *dp = NULL;
3434 struct svc_fh *current_fh = &cstate->current_fh;
3435 struct inode *ino = current_fh->fh_dentry->d_inode;
3436 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3442 if (grace_disallows_io(net, ino))
3443 return nfserr_grace;
3445 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3446 return check_special_stateids(net, current_fh, stateid, flags);
3448 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3449 &s, cstate->minorversion, nn);
3452 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3455 switch (s->sc_type) {
3456 case NFS4_DELEG_STID:
3457 dp = delegstateid(s);
3458 status = nfs4_check_delegmode(dp, flags);
3462 *filpp = dp->dl_file->fi_deleg_file;
3465 status = nfserr_serverfault;
3470 case NFS4_OPEN_STID:
3471 case NFS4_LOCK_STID:
3472 stp = openlockstateid(s);
3473 status = nfs4_check_fh(current_fh, stp);
3476 if (stp->st_stateowner->so_is_open_owner
3477 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3479 status = nfs4_check_openmode(stp, flags);
3483 if (flags & RD_STATE)
3484 *filpp = find_readable_file(stp->st_file);
3486 *filpp = find_writeable_file(stp->st_file);
3490 return nfserr_bad_stateid;
3498 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3500 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3501 return nfserr_locks_held;
3502 release_lock_stateid(stp);
3507 * Test if the stateid is valid
3510 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3511 struct nfsd4_test_stateid *test_stateid)
3513 struct nfsd4_test_stateid_id *stateid;
3514 struct nfs4_client *cl = cstate->session->se_client;
3517 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3518 stateid->ts_id_status =
3519 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3520 nfs4_unlock_state();
3526 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3527 struct nfsd4_free_stateid *free_stateid)
3529 stateid_t *stateid = &free_stateid->fr_stateid;
3530 struct nfs4_stid *s;
3531 struct nfs4_client *cl = cstate->session->se_client;
3532 __be32 ret = nfserr_bad_stateid;
3535 s = find_stateid(cl, stateid);
3538 switch (s->sc_type) {
3539 case NFS4_DELEG_STID:
3540 ret = nfserr_locks_held;
3542 case NFS4_OPEN_STID:
3543 case NFS4_LOCK_STID:
3544 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3547 if (s->sc_type == NFS4_LOCK_STID)
3548 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3550 ret = nfserr_locks_held;
3553 ret = nfserr_bad_stateid;
3556 nfs4_unlock_state();
3563 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3564 RD_STATE : WR_STATE;
3567 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3569 struct svc_fh *current_fh = &cstate->current_fh;
3570 struct nfs4_stateowner *sop = stp->st_stateowner;
3573 status = nfsd4_check_seqid(cstate, sop, seqid);
3576 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3578 * "Closed" stateid's exist *only* to return
3579 * nfserr_replay_me from the previous step.
3581 return nfserr_bad_stateid;
3582 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3585 return nfs4_check_fh(current_fh, stp);
3589 * Checks for sequence id mutating operations.
3592 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3593 stateid_t *stateid, char typemask,
3594 struct nfs4_ol_stateid **stpp,
3595 struct nfsd_net *nn)
3598 struct nfs4_stid *s;
3600 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3601 seqid, STATEID_VAL(stateid));
3604 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3605 cstate->minorversion, nn);
3608 *stpp = openlockstateid(s);
3609 cstate->replay_owner = (*stpp)->st_stateowner;
3611 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3614 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3615 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3618 struct nfs4_openowner *oo;
3620 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3621 NFS4_OPEN_STID, stpp, nn);
3624 oo = openowner((*stpp)->st_stateowner);
3625 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3626 return nfserr_bad_stateid;
3631 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3632 struct nfsd4_open_confirm *oc)
3635 struct nfs4_openowner *oo;
3636 struct nfs4_ol_stateid *stp;
3637 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3639 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3640 (int)cstate->current_fh.fh_dentry->d_name.len,
3641 cstate->current_fh.fh_dentry->d_name.name);
3643 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3649 status = nfs4_preprocess_seqid_op(cstate,
3650 oc->oc_seqid, &oc->oc_req_stateid,
3651 NFS4_OPEN_STID, &stp, nn);
3654 oo = openowner(stp->st_stateowner);
3655 status = nfserr_bad_stateid;
3656 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3658 oo->oo_flags |= NFS4_OO_CONFIRMED;
3659 update_stateid(&stp->st_stid.sc_stateid);
3660 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3661 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3662 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3664 nfsd4_client_record_create(oo->oo_owner.so_client);
3667 if (!cstate->replay_owner)
3668 nfs4_unlock_state();
3672 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3674 if (!test_access(access, stp))
3676 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3677 clear_access(access, stp);
3680 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3682 switch (to_access) {
3683 case NFS4_SHARE_ACCESS_READ:
3684 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3685 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3687 case NFS4_SHARE_ACCESS_WRITE:
3688 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3689 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3691 case NFS4_SHARE_ACCESS_BOTH:
3699 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3702 for (i = 0; i < 4; i++) {
3703 if ((i & deny) != i)
3709 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3710 struct nfsd4_compound_state *cstate,
3711 struct nfsd4_open_downgrade *od)
3714 struct nfs4_ol_stateid *stp;
3715 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3717 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3718 (int)cstate->current_fh.fh_dentry->d_name.len,
3719 cstate->current_fh.fh_dentry->d_name.name);
3721 /* We don't yet support WANT bits: */
3722 if (od->od_deleg_want)
3723 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3727 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3728 &od->od_stateid, &stp, nn);
3731 status = nfserr_inval;
3732 if (!test_access(od->od_share_access, stp)) {
3733 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3734 stp->st_access_bmap, od->od_share_access);
3737 if (!test_deny(od->od_share_deny, stp)) {
3738 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3739 stp->st_deny_bmap, od->od_share_deny);
3742 nfs4_stateid_downgrade(stp, od->od_share_access);
3744 reset_union_bmap_deny(od->od_share_deny, stp);
3746 update_stateid(&stp->st_stid.sc_stateid);
3747 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3750 if (!cstate->replay_owner)
3751 nfs4_unlock_state();
3755 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3757 struct nfs4_openowner *oo;
3758 struct nfs4_ol_stateid *s;
3760 if (!so->so_is_open_owner)
3763 s = oo->oo_last_closed_stid;
3766 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3767 /* Release the last_closed_stid on the next seqid bump: */
3768 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3771 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3772 release_last_closed_stateid(oo);
3775 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3777 unhash_open_stateid(s);
3778 s->st_stid.sc_type = NFS4_CLOSED_STID;
3782 * nfs4_unlock_state() called after encode
3785 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3786 struct nfsd4_close *close)
3789 struct nfs4_openowner *oo;
3790 struct nfs4_ol_stateid *stp;
3791 struct net *net = SVC_NET(rqstp);
3792 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3794 dprintk("NFSD: nfsd4_close on file %.*s\n",
3795 (int)cstate->current_fh.fh_dentry->d_name.len,
3796 cstate->current_fh.fh_dentry->d_name.name);
3799 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3801 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3805 oo = openowner(stp->st_stateowner);
3807 update_stateid(&stp->st_stid.sc_stateid);
3808 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3810 nfsd4_close_open_stateid(stp);
3811 release_last_closed_stateid(oo);
3812 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3813 oo->oo_last_closed_stid = stp;
3815 if (list_empty(&oo->oo_owner.so_stateids)) {
3816 if (cstate->minorversion) {
3817 release_openowner(oo);
3818 cstate->replay_owner = NULL;
3821 * In the 4.0 case we need to keep the owners around a
3822 * little while to handle CLOSE replay.
3824 if (list_empty(&oo->oo_owner.so_stateids))
3825 move_to_close_lru(oo, SVC_NET(rqstp));
3829 if (!cstate->replay_owner)
3830 nfs4_unlock_state();
3835 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3836 struct nfsd4_delegreturn *dr)
3838 struct nfs4_delegation *dp;
3839 stateid_t *stateid = &dr->dr_stateid;
3840 struct nfs4_stid *s;
3842 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3844 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3848 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
3849 cstate->minorversion, nn);
3852 dp = delegstateid(s);
3853 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3857 unhash_delegation(dp);
3859 nfs4_unlock_state();
3865 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3867 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3870 end_offset(u64 start, u64 len)
3875 return end >= start ? end: NFS4_MAX_UINT64;
3878 /* last octet in a range */
3880 last_byte_offset(u64 start, u64 len)
3886 return end > start ? end - 1: NFS4_MAX_UINT64;
3889 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3891 return (file_hashval(inode) + cl_id
3892 + opaque_hashval(ownername->data, ownername->len))
3893 & LOCKOWNER_INO_HASH_MASK;
3897 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3898 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3899 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3900 * locking, this prevents us from being completely protocol-compliant. The
3901 * real solution to this problem is to start using unsigned file offsets in
3902 * the VFS, but this is a very deep change!
3905 nfs4_transform_lock_offset(struct file_lock *lock)
3907 if (lock->fl_start < 0)
3908 lock->fl_start = OFFSET_MAX;
3909 if (lock->fl_end < 0)
3910 lock->fl_end = OFFSET_MAX;
3913 /* Hack!: For now, we're defining this just so we can use a pointer to it
3914 * as a unique cookie to identify our (NFSv4's) posix locks. */
3915 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3919 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3921 struct nfs4_lockowner *lo;
3923 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3924 lo = (struct nfs4_lockowner *) fl->fl_owner;
3925 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3926 lo->lo_owner.so_owner.len, GFP_KERNEL);
3927 if (!deny->ld_owner.data)
3928 /* We just don't care that much */
3930 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3931 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3934 deny->ld_owner.len = 0;
3935 deny->ld_owner.data = NULL;
3936 deny->ld_clientid.cl_boot = 0;
3937 deny->ld_clientid.cl_id = 0;
3939 deny->ld_start = fl->fl_start;
3940 deny->ld_length = NFS4_MAX_UINT64;
3941 if (fl->fl_end != NFS4_MAX_UINT64)
3942 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3943 deny->ld_type = NFS4_READ_LT;
3944 if (fl->fl_type != F_RDLCK)
3945 deny->ld_type = NFS4_WRITE_LT;
3948 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3950 struct nfs4_ol_stateid *lst;
3952 if (!same_owner_str(&lo->lo_owner, owner, clid))
3954 lst = list_first_entry(&lo->lo_owner.so_stateids,
3955 struct nfs4_ol_stateid, st_perstateowner);
3956 return lst->st_file->fi_inode == inode;
3959 static struct nfs4_lockowner *
3960 find_lockowner_str(struct inode *inode, clientid_t *clid,
3961 struct xdr_netobj *owner, struct nfsd_net *nn)
3963 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3964 struct nfs4_lockowner *lo;
3966 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3967 if (same_lockowner_ino(lo, inode, clid, owner))
3973 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3975 struct inode *inode = open_stp->st_file->fi_inode;
3976 unsigned int inohash = lockowner_ino_hashval(inode,
3977 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3978 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
3980 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
3981 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
3982 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3986 * Alloc a lock owner structure.
3987 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3990 * strhashval = ownerstr_hashval
3993 static struct nfs4_lockowner *
3994 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3995 struct nfs4_lockowner *lo;
3997 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4000 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4001 lo->lo_owner.so_is_open_owner = 0;
4002 /* It is the openowner seqid that will be incremented in encode in the
4003 * case of new lockowners; so increment the lock seqid manually: */
4004 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4005 hash_lockowner(lo, strhashval, clp, open_stp);
4009 static struct nfs4_ol_stateid *
4010 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4012 struct nfs4_ol_stateid *stp;
4013 struct nfs4_client *clp = lo->lo_owner.so_client;
4015 stp = nfs4_alloc_stateid(clp);
4018 stp->st_stid.sc_type = NFS4_LOCK_STID;
4019 list_add(&stp->st_perfile, &fp->fi_stateids);
4020 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4021 stp->st_stateowner = &lo->lo_owner;
4024 stp->st_access_bmap = 0;
4025 stp->st_deny_bmap = open_stp->st_deny_bmap;
4026 stp->st_openstp = open_stp;
4031 check_lock_length(u64 offset, u64 length)
4033 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4034 LOFF_OVERFLOW(offset, length)));
4037 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4039 struct nfs4_file *fp = lock_stp->st_file;
4040 int oflag = nfs4_access_to_omode(access);
4042 if (test_access(access, lock_stp))
4044 nfs4_file_get_access(fp, oflag);
4045 set_access(access, lock_stp);
4048 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)
4050 struct nfs4_file *fi = ost->st_file;
4051 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4052 struct nfs4_client *cl = oo->oo_owner.so_client;
4053 struct nfs4_lockowner *lo;
4054 unsigned int strhashval;
4055 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4057 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4058 &lock->v.new.owner, nn);
4060 if (!cstate->minorversion)
4061 return nfserr_bad_seqid;
4062 /* XXX: a lockowner always has exactly one stateid: */
4063 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4064 struct nfs4_ol_stateid, st_perstateowner);
4067 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4068 &lock->v.new.owner);
4069 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4071 return nfserr_jukebox;
4072 *lst = alloc_init_lock_stateid(lo, fi, ost);
4074 release_lockowner(lo);
4075 return nfserr_jukebox;
4085 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4086 struct nfsd4_lock *lock)
4088 struct nfs4_openowner *open_sop = NULL;
4089 struct nfs4_lockowner *lock_sop = NULL;
4090 struct nfs4_ol_stateid *lock_stp;
4091 struct file *filp = NULL;
4092 struct file_lock *file_lock = NULL;
4093 struct file_lock *conflock = NULL;
4095 bool new_state = false;
4098 struct net *net = SVC_NET(rqstp);
4099 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4101 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4102 (long long) lock->lk_offset,
4103 (long long) lock->lk_length);
4105 if (check_lock_length(lock->lk_offset, lock->lk_length))
4106 return nfserr_inval;
4108 if ((status = fh_verify(rqstp, &cstate->current_fh,
4109 S_IFREG, NFSD_MAY_LOCK))) {
4110 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4116 if (lock->lk_is_new) {
4117 struct nfs4_ol_stateid *open_stp = NULL;
4119 if (nfsd4_has_session(cstate))
4120 /* See rfc 5661 18.10.3: given clientid is ignored: */
4121 memcpy(&lock->v.new.clientid,
4122 &cstate->session->se_client->cl_clientid,
4123 sizeof(clientid_t));
4125 status = nfserr_stale_clientid;
4126 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4129 /* validate and update open stateid and open seqid */
4130 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4131 lock->lk_new_open_seqid,
4132 &lock->lk_new_open_stateid,
4136 open_sop = openowner(open_stp->st_stateowner);
4137 status = nfserr_bad_stateid;
4138 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4139 &lock->v.new.clientid))
4141 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4142 &lock_stp, &new_state);
4144 status = nfs4_preprocess_seqid_op(cstate,
4145 lock->lk_old_lock_seqid,
4146 &lock->lk_old_lock_stateid,
4147 NFS4_LOCK_STID, &lock_stp, nn);
4150 lock_sop = lockowner(lock_stp->st_stateowner);
4152 lkflg = setlkflg(lock->lk_type);
4153 status = nfs4_check_openmode(lock_stp, lkflg);
4157 status = nfserr_grace;
4158 if (locks_in_grace(net) && !lock->lk_reclaim)
4160 status = nfserr_no_grace;
4161 if (!locks_in_grace(net) && lock->lk_reclaim)
4164 file_lock = locks_alloc_lock();
4166 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4167 status = nfserr_jukebox;
4171 locks_init_lock(file_lock);
4172 switch (lock->lk_type) {
4175 filp = find_readable_file(lock_stp->st_file);
4177 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4178 file_lock->fl_type = F_RDLCK;
4181 case NFS4_WRITEW_LT:
4182 filp = find_writeable_file(lock_stp->st_file);
4184 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4185 file_lock->fl_type = F_WRLCK;
4188 status = nfserr_inval;
4192 status = nfserr_openmode;
4195 file_lock->fl_owner = (fl_owner_t)lock_sop;
4196 file_lock->fl_pid = current->tgid;
4197 file_lock->fl_file = filp;
4198 file_lock->fl_flags = FL_POSIX;
4199 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4200 file_lock->fl_start = lock->lk_offset;
4201 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4202 nfs4_transform_lock_offset(file_lock);
4204 conflock = locks_alloc_lock();
4206 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4207 status = nfserr_jukebox;
4211 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4213 case 0: /* success! */
4214 update_stateid(&lock_stp->st_stid.sc_stateid);
4215 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4219 case (EAGAIN): /* conflock holds conflicting lock */
4220 status = nfserr_denied;
4221 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4222 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4225 status = nfserr_deadlock;
4228 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4229 status = nfserrno(err);
4233 if (status && new_state)
4234 release_lockowner(lock_sop);
4235 if (!cstate->replay_owner)
4236 nfs4_unlock_state();
4238 locks_free_lock(file_lock);
4240 locks_free_lock(conflock);
4245 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4246 * so we do a temporary open here just to get an open file to pass to
4247 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4250 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4253 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4255 err = nfserrno(vfs_test_lock(file, lock));
4265 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4266 struct nfsd4_lockt *lockt)
4268 struct inode *inode;
4269 struct file_lock *file_lock = NULL;
4270 struct nfs4_lockowner *lo;
4272 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4274 if (locks_in_grace(SVC_NET(rqstp)))
4275 return nfserr_grace;
4277 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4278 return nfserr_inval;
4282 if (!nfsd4_has_session(cstate)) {
4283 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4288 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4291 inode = cstate->current_fh.fh_dentry->d_inode;
4292 file_lock = locks_alloc_lock();
4294 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4295 status = nfserr_jukebox;
4298 locks_init_lock(file_lock);
4299 switch (lockt->lt_type) {
4302 file_lock->fl_type = F_RDLCK;
4305 case NFS4_WRITEW_LT:
4306 file_lock->fl_type = F_WRLCK;
4309 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4310 status = nfserr_inval;
4314 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4316 file_lock->fl_owner = (fl_owner_t)lo;
4317 file_lock->fl_pid = current->tgid;
4318 file_lock->fl_flags = FL_POSIX;
4320 file_lock->fl_start = lockt->lt_offset;
4321 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4323 nfs4_transform_lock_offset(file_lock);
4325 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4329 if (file_lock->fl_type != F_UNLCK) {
4330 status = nfserr_denied;
4331 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4334 nfs4_unlock_state();
4336 locks_free_lock(file_lock);
4341 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4342 struct nfsd4_locku *locku)
4344 struct nfs4_ol_stateid *stp;
4345 struct file *filp = NULL;
4346 struct file_lock *file_lock = NULL;
4349 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4351 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4352 (long long) locku->lu_offset,
4353 (long long) locku->lu_length);
4355 if (check_lock_length(locku->lu_offset, locku->lu_length))
4356 return nfserr_inval;
4360 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4361 &locku->lu_stateid, NFS4_LOCK_STID,
4365 filp = find_any_file(stp->st_file);
4367 status = nfserr_lock_range;
4370 file_lock = locks_alloc_lock();
4372 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4373 status = nfserr_jukebox;
4376 locks_init_lock(file_lock);
4377 file_lock->fl_type = F_UNLCK;
4378 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4379 file_lock->fl_pid = current->tgid;
4380 file_lock->fl_file = filp;
4381 file_lock->fl_flags = FL_POSIX;
4382 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4383 file_lock->fl_start = locku->lu_offset;
4385 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4387 nfs4_transform_lock_offset(file_lock);
4390 * Try to unlock the file in the VFS.
4392 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4394 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4398 * OK, unlock succeeded; the only thing left to do is update the stateid.
4400 update_stateid(&stp->st_stid.sc_stateid);
4401 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4404 if (!cstate->replay_owner)
4405 nfs4_unlock_state();
4407 locks_free_lock(file_lock);
4411 status = nfserrno(err);
4417 * 1: locks held by lockowner
4418 * 0: no locks held by lockowner
4421 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4423 struct file_lock **flpp;
4424 struct inode *inode = filp->fi_inode;
4428 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4429 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4440 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4441 struct nfsd4_compound_state *cstate,
4442 struct nfsd4_release_lockowner *rlockowner)
4444 clientid_t *clid = &rlockowner->rl_clientid;
4445 struct nfs4_stateowner *sop;
4446 struct nfs4_lockowner *lo;
4447 struct nfs4_ol_stateid *stp;
4448 struct xdr_netobj *owner = &rlockowner->rl_owner;
4449 struct list_head matches;
4450 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4452 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4454 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4455 clid->cl_boot, clid->cl_id);
4459 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4463 status = nfserr_locks_held;
4464 INIT_LIST_HEAD(&matches);
4466 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4467 if (sop->so_is_open_owner)
4469 if (!same_owner_str(sop, owner, clid))
4471 list_for_each_entry(stp, &sop->so_stateids,
4473 lo = lockowner(sop);
4474 if (check_for_locks(stp->st_file, lo))
4476 list_add(&lo->lo_list, &matches);
4479 /* Clients probably won't expect us to return with some (but not all)
4480 * of the lockowner state released; so don't release any until all
4481 * have been checked. */
4483 while (!list_empty(&matches)) {
4484 lo = list_entry(matches.next, struct nfs4_lockowner,
4486 /* unhash_stateowner deletes so_perclient only
4487 * for openowners. */
4488 list_del(&lo->lo_list);
4489 release_lockowner(lo);
4492 nfs4_unlock_state();
4496 static inline struct nfs4_client_reclaim *
4499 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4503 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4505 struct nfs4_client_reclaim *crp;
4507 crp = nfsd4_find_reclaim_client(name, nn);
4508 return (crp && crp->cr_clp);
4512 * failure => all reset bets are off, nfserr_no_grace...
4514 struct nfs4_client_reclaim *
4515 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4517 unsigned int strhashval;
4518 struct nfs4_client_reclaim *crp;
4520 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4521 crp = alloc_reclaim();
4523 strhashval = clientstr_hashval(name);
4524 INIT_LIST_HEAD(&crp->cr_strhash);
4525 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4526 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4528 nn->reclaim_str_hashtbl_size++;
4534 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4536 list_del(&crp->cr_strhash);
4538 nn->reclaim_str_hashtbl_size--;
4542 nfs4_release_reclaim(struct nfsd_net *nn)
4544 struct nfs4_client_reclaim *crp = NULL;
4547 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4548 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4549 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4550 struct nfs4_client_reclaim, cr_strhash);
4551 nfs4_remove_reclaim_record(crp, nn);
4554 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4558 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4559 struct nfs4_client_reclaim *
4560 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4562 unsigned int strhashval;
4563 struct nfs4_client_reclaim *crp = NULL;
4565 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4567 strhashval = clientstr_hashval(recdir);
4568 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4569 if (same_name(crp->cr_recdir, recdir)) {
4577 * Called from OPEN. Look for clientid in reclaim list.
4580 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4582 struct nfs4_client *clp;
4584 /* find clientid in conf_id_hashtbl */
4585 clp = find_confirmed_client(clid, sessions, nn);
4587 return nfserr_reclaim_bad;
4589 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4592 #ifdef CONFIG_NFSD_FAULT_INJECTION
4594 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4600 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4602 char buf[INET6_ADDRSTRLEN];
4603 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4604 printk(KERN_INFO "NFS Client: %s\n", buf);
4608 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4611 char buf[INET6_ADDRSTRLEN];
4612 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4613 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4616 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_lockowner *))
4618 struct nfs4_openowner *oop;
4619 struct nfs4_lockowner *lop, *lo_next;
4620 struct nfs4_ol_stateid *stp, *st_next;
4623 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4624 list_for_each_entry_safe(stp, st_next, &oop->oo_owner.so_stateids, st_perstateowner) {
4625 list_for_each_entry_safe(lop, lo_next, &stp->st_lockowners, lo_perstateid) {
4637 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
4639 return nfsd_foreach_client_lock(clp, max, release_lockowner);
4642 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
4644 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
4645 nfsd_print_count(clp, count, "locked files");
4649 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
4651 struct nfs4_openowner *oop, *next;
4654 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
4664 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
4666 return nfsd_foreach_client_open(clp, max, release_openowner);
4669 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
4671 u64 count = nfsd_foreach_client_open(clp, max, NULL);
4672 nfsd_print_count(clp, count, "open files");
4676 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
4677 struct list_head *victims)
4679 struct nfs4_delegation *dp, *next;
4682 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
4684 list_move(&dp->dl_recall_lru, victims);
4691 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
4693 struct nfs4_delegation *dp, *next;
4697 spin_lock(&recall_lock);
4698 count = nfsd_find_all_delegations(clp, max, &victims);
4699 spin_unlock(&recall_lock);
4701 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4702 unhash_delegation(dp);
4707 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
4709 struct nfs4_delegation *dp, *next;
4713 spin_lock(&recall_lock);
4714 count = nfsd_find_all_delegations(clp, max, &victims);
4715 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4716 nfsd_break_one_deleg(dp);
4717 spin_unlock(&recall_lock);
4722 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
4726 spin_lock(&recall_lock);
4727 count = nfsd_find_all_delegations(clp, max, NULL);
4728 spin_unlock(&recall_lock);
4730 nfsd_print_count(clp, count, "delegations");
4734 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
4736 struct nfs4_client *clp, *next;
4738 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4740 if (!nfsd_netns_ready(nn))
4743 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4744 count += func(clp, max - count);
4745 if ((max != 0) && (count >= max))
4752 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
4754 struct nfs4_client *clp;
4755 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4757 if (!nfsd_netns_ready(nn))
4760 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
4761 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
4767 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4769 /* initialization to perform at module load time: */
4772 nfs4_state_init(void)
4776 for (i = 0; i < FILE_HASH_SIZE; i++) {
4777 INIT_LIST_HEAD(&file_hashtbl[i]);
4779 INIT_LIST_HEAD(&del_recall_lru);
4783 * Since the lifetime of a delegation isn't limited to that of an open, a
4784 * client may quite reasonably hang on to a delegation as long as it has
4785 * the inode cached. This becomes an obvious problem the first time a
4786 * client's inode cache approaches the size of the server's total memory.
4788 * For now we avoid this problem by imposing a hard limit on the number
4789 * of delegations, which varies according to the server's memory size.
4792 set_max_delegations(void)
4795 * Allow at most 4 delegations per megabyte of RAM. Quick
4796 * estimates suggest that in the worst case (where every delegation
4797 * is for a different inode), a delegation could take about 1.5K,
4798 * giving a worst case usage of about 6% of memory.
4800 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4803 static int nfs4_state_create_net(struct net *net)
4805 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4808 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4809 CLIENT_HASH_SIZE, GFP_KERNEL);
4810 if (!nn->conf_id_hashtbl)
4812 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4813 CLIENT_HASH_SIZE, GFP_KERNEL);
4814 if (!nn->unconf_id_hashtbl)
4816 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
4817 OWNER_HASH_SIZE, GFP_KERNEL);
4818 if (!nn->ownerstr_hashtbl)
4820 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
4821 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
4822 if (!nn->lockowner_ino_hashtbl)
4823 goto err_lockowner_ino;
4824 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
4825 SESSION_HASH_SIZE, GFP_KERNEL);
4826 if (!nn->sessionid_hashtbl)
4829 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4830 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
4831 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
4833 for (i = 0; i < OWNER_HASH_SIZE; i++)
4834 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
4835 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4836 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
4837 for (i = 0; i < SESSION_HASH_SIZE; i++)
4838 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
4839 nn->conf_name_tree = RB_ROOT;
4840 nn->unconf_name_tree = RB_ROOT;
4841 INIT_LIST_HEAD(&nn->client_lru);
4842 INIT_LIST_HEAD(&nn->close_lru);
4843 spin_lock_init(&nn->client_lock);
4845 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
4851 kfree(nn->lockowner_ino_hashtbl);
4853 kfree(nn->ownerstr_hashtbl);
4855 kfree(nn->unconf_id_hashtbl);
4857 kfree(nn->conf_id_hashtbl);
4863 nfs4_state_destroy_net(struct net *net)
4866 struct nfs4_client *clp = NULL;
4867 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4868 struct rb_node *node, *tmp;
4870 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4871 while (!list_empty(&nn->conf_id_hashtbl[i])) {
4872 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4873 destroy_client(clp);
4877 node = rb_first(&nn->unconf_name_tree);
4878 while (node != NULL) {
4880 node = rb_next(tmp);
4881 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4882 rb_erase(tmp, &nn->unconf_name_tree);
4883 destroy_client(clp);
4886 kfree(nn->sessionid_hashtbl);
4887 kfree(nn->lockowner_ino_hashtbl);
4888 kfree(nn->ownerstr_hashtbl);
4889 kfree(nn->unconf_id_hashtbl);
4890 kfree(nn->conf_id_hashtbl);
4895 nfs4_state_start_net(struct net *net)
4897 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4900 ret = nfs4_state_create_net(net);
4903 nfsd4_client_tracking_init(net);
4904 nn->boot_time = get_seconds();
4905 locks_start_grace(net, &nn->nfsd4_manager);
4906 nn->grace_ended = false;
4907 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
4908 nn->nfsd4_grace, net);
4909 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
4913 /* initialization to perform when the nfsd service is started: */
4916 nfs4_state_start(void)
4920 ret = set_callback_cred();
4923 laundry_wq = create_singlethread_workqueue("nfsd4");
4924 if (laundry_wq == NULL) {
4928 ret = nfsd4_create_callback_queue();
4930 goto out_free_laundry;
4932 set_max_delegations();
4937 destroy_workqueue(laundry_wq);
4942 /* should be called with the state lock held */
4944 nfs4_state_shutdown_net(struct net *net)
4946 struct nfs4_delegation *dp = NULL;
4947 struct list_head *pos, *next, reaplist;
4948 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4950 cancel_delayed_work_sync(&nn->laundromat_work);
4951 locks_end_grace(&nn->nfsd4_manager);
4953 INIT_LIST_HEAD(&reaplist);
4954 spin_lock(&recall_lock);
4955 list_for_each_safe(pos, next, &del_recall_lru) {
4956 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4957 if (dp->dl_stid.sc_client->net != net)
4959 list_move(&dp->dl_recall_lru, &reaplist);
4961 spin_unlock(&recall_lock);
4962 list_for_each_safe(pos, next, &reaplist) {
4963 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4964 unhash_delegation(dp);
4967 nfsd4_client_tracking_exit(net);
4968 nfs4_state_destroy_net(net);
4972 nfs4_state_shutdown(void)
4974 destroy_workqueue(laundry_wq);
4975 nfsd4_destroy_callback_queue();
4979 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4981 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4982 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4986 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4988 if (cstate->minorversion) {
4989 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4990 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4995 clear_current_stateid(struct nfsd4_compound_state *cstate)
4997 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5001 * functions to set current state id
5004 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5006 put_stateid(cstate, &odp->od_stateid);
5010 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5012 put_stateid(cstate, &open->op_stateid);
5016 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5018 put_stateid(cstate, &close->cl_stateid);
5022 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5024 put_stateid(cstate, &lock->lk_resp_stateid);
5028 * functions to consume current state id
5032 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5034 get_stateid(cstate, &odp->od_stateid);
5038 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5040 get_stateid(cstate, &drp->dr_stateid);
5044 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5046 get_stateid(cstate, &fsp->fr_stateid);
5050 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5052 get_stateid(cstate, &setattr->sa_stateid);
5056 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5058 get_stateid(cstate, &close->cl_stateid);
5062 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5064 get_stateid(cstate, &locku->lu_stateid);
5068 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5070 get_stateid(cstate, &read->rd_stateid);
5074 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5076 get_stateid(cstate, &write->wr_stateid);