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/clnt.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 int 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 inline int get_new_stid(struct nfs4_stid *stid)
235 static int min_stateid = 0;
236 struct idr *stateids = &stid->sc_client->cl_stateids;
240 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
242 * Note: the necessary preallocation was done in
243 * nfs4_alloc_stateid(). The idr code caps the number of
244 * preallocations that can exist at a time, but the state lock
245 * prevents anyone from using ours before we get here:
249 * It shouldn't be a problem to reuse an opaque stateid value.
250 * I don't think it is for 4.1. But with 4.0 I worry that, for
251 * example, a stray write retransmission could be accepted by
252 * the server when it should have been rejected. Therefore,
253 * adopt a trick from the sctp code to attempt to maximize the
254 * amount of time until an id is reused, by ensuring they always
255 * "increase" (mod INT_MAX):
258 min_stateid = new_stid+1;
259 if (min_stateid == INT_MAX)
264 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
266 stateid_t *s = &stid->sc_stateid;
269 stid->sc_type = type;
270 stid->sc_client = cl;
271 s->si_opaque.so_clid = cl->cl_clientid;
272 new_id = get_new_stid(stid);
273 s->si_opaque.so_id = (u32)new_id;
274 /* Will be incremented before return to client: */
275 s->si_generation = 0;
278 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
280 struct idr *stateids = &cl->cl_stateids;
282 if (!idr_pre_get(stateids, GFP_KERNEL))
285 * Note: if we fail here (or any time between now and the time
286 * we actually get the new idr), we won't need to undo the idr
287 * preallocation, since the idr code caps the number of
288 * preallocated entries.
290 return kmem_cache_alloc(slab, GFP_KERNEL);
293 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
295 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
298 static struct nfs4_delegation *
299 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
301 struct nfs4_delegation *dp;
302 struct nfs4_file *fp = stp->st_file;
304 dprintk("NFSD alloc_init_deleg\n");
306 * Major work on the lease subsystem (for example, to support
307 * calbacks on stat) will be required before we can support
308 * write delegations properly.
310 if (type != NFS4_OPEN_DELEGATE_READ)
312 if (fp->fi_had_conflict)
314 if (num_delegations > max_delegations)
316 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
319 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
321 * delegation seqid's are never incremented. The 4.1 special
322 * meaning of seqid 0 isn't meaningful, really, but let's avoid
323 * 0 anyway just for consistency and use 1:
325 dp->dl_stid.sc_stateid.si_generation = 1;
327 INIT_LIST_HEAD(&dp->dl_perfile);
328 INIT_LIST_HEAD(&dp->dl_perclnt);
329 INIT_LIST_HEAD(&dp->dl_recall_lru);
333 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
335 atomic_set(&dp->dl_count, 1);
336 nfsd4_init_callback(&dp->dl_recall);
341 nfs4_put_delegation(struct nfs4_delegation *dp)
343 if (atomic_dec_and_test(&dp->dl_count)) {
344 dprintk("NFSD: freeing dp %p\n",dp);
345 put_nfs4_file(dp->dl_file);
346 kmem_cache_free(deleg_slab, dp);
351 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
353 if (atomic_dec_and_test(&fp->fi_delegees)) {
354 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
356 fput(fp->fi_deleg_file);
357 fp->fi_deleg_file = NULL;
361 static void unhash_stid(struct nfs4_stid *s)
363 struct idr *stateids = &s->sc_client->cl_stateids;
365 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
368 /* Called under the state lock. */
370 unhash_delegation(struct nfs4_delegation *dp)
372 unhash_stid(&dp->dl_stid);
373 list_del_init(&dp->dl_perclnt);
374 spin_lock(&recall_lock);
375 list_del_init(&dp->dl_perfile);
376 list_del_init(&dp->dl_recall_lru);
377 spin_unlock(&recall_lock);
378 nfs4_put_deleg_lease(dp->dl_file);
379 nfs4_put_delegation(dp);
386 static unsigned int clientid_hashval(u32 id)
388 return id & CLIENT_HASH_MASK;
391 static unsigned int clientstr_hashval(const char *name)
393 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
397 * We store the NONE, READ, WRITE, and BOTH bits separately in the
398 * st_{access,deny}_bmap field of the stateid, in order to track not
399 * only what share bits are currently in force, but also what
400 * combinations of share bits previous opens have used. This allows us
401 * to enforce the recommendation of rfc 3530 14.2.19 that the server
402 * return an error if the client attempt to downgrade to a combination
403 * of share bits not explicable by closing some of its previous opens.
405 * XXX: This enforcement is actually incomplete, since we don't keep
406 * track of access/deny bit combinations; so, e.g., we allow:
408 * OPEN allow read, deny write
409 * OPEN allow both, deny none
410 * DOWNGRADE allow read, deny none
412 * which we should reject.
415 bmap_to_share_mode(unsigned long bmap) {
417 unsigned int access = 0;
419 for (i = 1; i < 4; i++) {
420 if (test_bit(i, &bmap))
427 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
428 unsigned int access, deny;
430 access = bmap_to_share_mode(stp->st_access_bmap);
431 deny = bmap_to_share_mode(stp->st_deny_bmap);
432 if ((access & open->op_share_deny) || (deny & open->op_share_access))
437 /* set share access for a given stateid */
439 set_access(u32 access, struct nfs4_ol_stateid *stp)
441 __set_bit(access, &stp->st_access_bmap);
444 /* clear share access for a given stateid */
446 clear_access(u32 access, struct nfs4_ol_stateid *stp)
448 __clear_bit(access, &stp->st_access_bmap);
451 /* test whether a given stateid has access */
453 test_access(u32 access, struct nfs4_ol_stateid *stp)
455 return test_bit(access, &stp->st_access_bmap);
458 /* set share deny for a given stateid */
460 set_deny(u32 access, struct nfs4_ol_stateid *stp)
462 __set_bit(access, &stp->st_deny_bmap);
465 /* clear share deny for a given stateid */
467 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
469 __clear_bit(access, &stp->st_deny_bmap);
472 /* test whether a given stateid is denying specific access */
474 test_deny(u32 access, struct nfs4_ol_stateid *stp)
476 return test_bit(access, &stp->st_deny_bmap);
479 static int nfs4_access_to_omode(u32 access)
481 switch (access & NFS4_SHARE_ACCESS_BOTH) {
482 case NFS4_SHARE_ACCESS_READ:
484 case NFS4_SHARE_ACCESS_WRITE:
486 case NFS4_SHARE_ACCESS_BOTH:
493 /* release all access and file references for a given stateid */
495 release_all_access(struct nfs4_ol_stateid *stp)
499 for (i = 1; i < 4; i++) {
500 if (test_access(i, stp))
501 nfs4_file_put_access(stp->st_file,
502 nfs4_access_to_omode(i));
503 clear_access(i, stp);
507 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
509 list_del(&stp->st_perfile);
510 list_del(&stp->st_perstateowner);
513 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
515 release_all_access(stp);
516 put_nfs4_file(stp->st_file);
520 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
522 kmem_cache_free(stateid_slab, stp);
525 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
529 unhash_generic_stateid(stp);
530 unhash_stid(&stp->st_stid);
531 file = find_any_file(stp->st_file);
533 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
534 close_generic_stateid(stp);
535 free_generic_stateid(stp);
538 static void unhash_lockowner(struct nfs4_lockowner *lo)
540 struct nfs4_ol_stateid *stp;
542 list_del(&lo->lo_owner.so_strhash);
543 list_del(&lo->lo_perstateid);
544 list_del(&lo->lo_owner_ino_hash);
545 while (!list_empty(&lo->lo_owner.so_stateids)) {
546 stp = list_first_entry(&lo->lo_owner.so_stateids,
547 struct nfs4_ol_stateid, st_perstateowner);
548 release_lock_stateid(stp);
552 static void release_lockowner(struct nfs4_lockowner *lo)
554 unhash_lockowner(lo);
555 nfs4_free_lockowner(lo);
559 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
561 struct nfs4_lockowner *lo;
563 while (!list_empty(&open_stp->st_lockowners)) {
564 lo = list_entry(open_stp->st_lockowners.next,
565 struct nfs4_lockowner, lo_perstateid);
566 release_lockowner(lo);
570 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
572 unhash_generic_stateid(stp);
573 release_stateid_lockowners(stp);
574 close_generic_stateid(stp);
577 static void release_open_stateid(struct nfs4_ol_stateid *stp)
579 unhash_open_stateid(stp);
580 unhash_stid(&stp->st_stid);
581 free_generic_stateid(stp);
584 static void unhash_openowner(struct nfs4_openowner *oo)
586 struct nfs4_ol_stateid *stp;
588 list_del(&oo->oo_owner.so_strhash);
589 list_del(&oo->oo_perclient);
590 while (!list_empty(&oo->oo_owner.so_stateids)) {
591 stp = list_first_entry(&oo->oo_owner.so_stateids,
592 struct nfs4_ol_stateid, st_perstateowner);
593 release_open_stateid(stp);
597 static void release_last_closed_stateid(struct nfs4_openowner *oo)
599 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
602 unhash_stid(&s->st_stid);
603 free_generic_stateid(s);
604 oo->oo_last_closed_stid = NULL;
608 static void release_openowner(struct nfs4_openowner *oo)
610 unhash_openowner(oo);
611 list_del(&oo->oo_close_lru);
612 release_last_closed_stateid(oo);
613 nfs4_free_openowner(oo);
617 hash_sessionid(struct nfs4_sessionid *sessionid)
619 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
621 return sid->sequence % SESSION_HASH_SIZE;
626 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
628 u32 *ptr = (u32 *)(&sessionid->data[0]);
629 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
633 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
640 gen_sessionid(struct nfsd4_session *ses)
642 struct nfs4_client *clp = ses->se_client;
643 struct nfsd4_sessionid *sid;
645 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
646 sid->clientid = clp->cl_clientid;
647 sid->sequence = current_sessionid++;
652 * The protocol defines ca_maxresponssize_cached to include the size of
653 * the rpc header, but all we need to cache is the data starting after
654 * the end of the initial SEQUENCE operation--the rest we regenerate
655 * each time. Therefore we can advertise a ca_maxresponssize_cached
656 * value that is the number of bytes in our cache plus a few additional
657 * bytes. In order to stay on the safe side, and not promise more than
658 * we can cache, those additional bytes must be the minimum possible: 24
659 * bytes of rpc header (xid through accept state, with AUTH_NULL
660 * verifier), 12 for the compound header (with zero-length tag), and 44
661 * for the SEQUENCE op response:
663 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
666 free_session_slots(struct nfsd4_session *ses)
670 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
671 kfree(ses->se_slots[i]);
675 * We don't actually need to cache the rpc and session headers, so we
676 * can allocate a little less for each slot:
678 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
680 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
683 static int nfsd4_sanitize_slot_size(u32 size)
685 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
686 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
692 * XXX: If we run out of reserved DRC memory we could (up to a point)
693 * re-negotiate active sessions and reduce their slot usage to make
694 * room for new connections. For now we just fail the create session.
696 static int nfsd4_get_drc_mem(int slotsize, u32 num)
700 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
702 spin_lock(&nfsd_drc_lock);
703 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
704 nfsd_drc_max_mem - nfsd_drc_mem_used);
705 num = min_t(int, num, avail / slotsize);
706 nfsd_drc_mem_used += num * slotsize;
707 spin_unlock(&nfsd_drc_lock);
712 static void nfsd4_put_drc_mem(int slotsize, int num)
714 spin_lock(&nfsd_drc_lock);
715 nfsd_drc_mem_used -= slotsize * num;
716 spin_unlock(&nfsd_drc_lock);
719 static struct nfsd4_session *__alloc_session(int slotsize, int numslots)
721 struct nfsd4_session *new;
724 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
725 + sizeof(struct nfsd4_session) > PAGE_SIZE);
726 mem = numslots * sizeof(struct nfsd4_slot *);
728 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
731 /* allocate each struct nfsd4_slot and data cache in one piece */
732 for (i = 0; i < numslots; i++) {
733 mem = sizeof(struct nfsd4_slot) + slotsize;
734 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
735 if (!new->se_slots[i])
741 kfree(new->se_slots[i]);
746 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
748 u32 maxrpc = nfsd_serv->sv_max_mesg;
750 new->maxreqs = numslots;
751 new->maxresp_cached = min_t(u32, req->maxresp_cached,
752 slotsize + NFSD_MIN_HDR_SEQ_SZ);
753 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
754 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
755 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
758 static void free_conn(struct nfsd4_conn *c)
760 svc_xprt_put(c->cn_xprt);
764 static void nfsd4_conn_lost(struct svc_xpt_user *u)
766 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
767 struct nfs4_client *clp = c->cn_session->se_client;
769 spin_lock(&clp->cl_lock);
770 if (!list_empty(&c->cn_persession)) {
771 list_del(&c->cn_persession);
774 spin_unlock(&clp->cl_lock);
775 nfsd4_probe_callback(clp);
778 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
780 struct nfsd4_conn *conn;
782 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
785 svc_xprt_get(rqstp->rq_xprt);
786 conn->cn_xprt = rqstp->rq_xprt;
787 conn->cn_flags = flags;
788 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
792 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
794 conn->cn_session = ses;
795 list_add(&conn->cn_persession, &ses->se_conns);
798 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
800 struct nfs4_client *clp = ses->se_client;
802 spin_lock(&clp->cl_lock);
803 __nfsd4_hash_conn(conn, ses);
804 spin_unlock(&clp->cl_lock);
807 static int nfsd4_register_conn(struct nfsd4_conn *conn)
809 conn->cn_xpt_user.callback = nfsd4_conn_lost;
810 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
813 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
817 nfsd4_hash_conn(conn, ses);
818 ret = nfsd4_register_conn(conn);
820 /* oops; xprt is already down: */
821 nfsd4_conn_lost(&conn->cn_xpt_user);
822 if (conn->cn_flags & NFS4_CDFC4_BACK) {
823 /* callback channel may be back up */
824 nfsd4_probe_callback(ses->se_client);
828 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
830 u32 dir = NFS4_CDFC4_FORE;
832 if (cses->flags & SESSION4_BACK_CHAN)
833 dir |= NFS4_CDFC4_BACK;
834 return alloc_conn(rqstp, dir);
837 /* must be called under client_lock */
838 static void nfsd4_del_conns(struct nfsd4_session *s)
840 struct nfs4_client *clp = s->se_client;
841 struct nfsd4_conn *c;
843 spin_lock(&clp->cl_lock);
844 while (!list_empty(&s->se_conns)) {
845 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
846 list_del_init(&c->cn_persession);
847 spin_unlock(&clp->cl_lock);
849 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
852 spin_lock(&clp->cl_lock);
854 spin_unlock(&clp->cl_lock);
857 static void __free_session(struct nfsd4_session *ses)
859 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
860 free_session_slots(ses);
864 static void free_session(struct kref *kref)
866 struct nfsd4_session *ses;
869 ses = container_of(kref, struct nfsd4_session, se_ref);
870 nn = net_generic(ses->se_client->net, nfsd_net_id);
872 lockdep_assert_held(&nn->client_lock);
873 nfsd4_del_conns(ses);
877 void nfsd4_put_session(struct nfsd4_session *ses)
879 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
881 spin_lock(&nn->client_lock);
882 nfsd4_put_session_locked(ses);
883 spin_unlock(&nn->client_lock);
886 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fchan)
888 struct nfsd4_session *new;
889 int numslots, slotsize;
891 * Note decreasing slot size below client's request may
892 * make it difficult for client to function correctly, whereas
893 * decreasing the number of slots will (just?) affect
894 * performance. When short on memory we therefore prefer to
895 * decrease number of slots instead of their size.
897 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
898 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
902 new = __alloc_session(slotsize, numslots);
904 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
907 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
911 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
914 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
916 new->se_client = clp;
919 INIT_LIST_HEAD(&new->se_conns);
921 new->se_cb_seq_nr = 1;
922 new->se_flags = cses->flags;
923 new->se_cb_prog = cses->callback_prog;
924 new->se_cb_sec = cses->cb_sec;
925 kref_init(&new->se_ref);
926 idx = hash_sessionid(&new->se_sessionid);
927 spin_lock(&nn->client_lock);
928 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
929 spin_lock(&clp->cl_lock);
930 list_add(&new->se_perclnt, &clp->cl_sessions);
931 spin_unlock(&clp->cl_lock);
932 spin_unlock(&nn->client_lock);
934 if (cses->flags & SESSION4_BACK_CHAN) {
935 struct sockaddr *sa = svc_addr(rqstp);
937 * This is a little silly; with sessions there's no real
938 * use for the callback address. Use the peer address
939 * as a reasonable default for now, but consider fixing
940 * the rpc client not to require an address in the
943 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
944 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
948 /* caller must hold client_lock */
949 static struct nfsd4_session *
950 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
952 struct nfsd4_session *elem;
954 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
956 dump_sessionid(__func__, sessionid);
957 idx = hash_sessionid(sessionid);
958 /* Search in the appropriate list */
959 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
960 if (!memcmp(elem->se_sessionid.data, sessionid->data,
961 NFS4_MAX_SESSIONID_LEN)) {
966 dprintk("%s: session not found\n", __func__);
970 /* caller must hold client_lock */
972 unhash_session(struct nfsd4_session *ses)
974 list_del(&ses->se_hash);
975 spin_lock(&ses->se_client->cl_lock);
976 list_del(&ses->se_perclnt);
977 spin_unlock(&ses->se_client->cl_lock);
980 /* must be called under the client_lock */
982 renew_client_locked(struct nfs4_client *clp)
984 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
986 if (is_client_expired(clp)) {
988 printk("%s: client (clientid %08x/%08x) already expired\n",
990 clp->cl_clientid.cl_boot,
991 clp->cl_clientid.cl_id);
995 dprintk("renewing client (clientid %08x/%08x)\n",
996 clp->cl_clientid.cl_boot,
997 clp->cl_clientid.cl_id);
998 list_move_tail(&clp->cl_lru, &nn->client_lru);
999 clp->cl_time = get_seconds();
1003 renew_client(struct nfs4_client *clp)
1005 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1007 spin_lock(&nn->client_lock);
1008 renew_client_locked(clp);
1009 spin_unlock(&nn->client_lock);
1012 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1014 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1016 if (clid->cl_boot == nn->boot_time)
1018 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1019 clid->cl_boot, clid->cl_id, nn->boot_time);
1024 * XXX Should we use a slab cache ?
1025 * This type of memory management is somewhat inefficient, but we use it
1026 * anyway since SETCLIENTID is not a common operation.
1028 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1030 struct nfs4_client *clp;
1032 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1035 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1036 if (clp->cl_name.data == NULL) {
1040 clp->cl_name.len = name.len;
1045 free_client(struct nfs4_client *clp)
1047 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1049 lockdep_assert_held(&nn->client_lock);
1050 while (!list_empty(&clp->cl_sessions)) {
1051 struct nfsd4_session *ses;
1052 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1054 list_del(&ses->se_perclnt);
1055 nfsd4_put_session_locked(ses);
1057 free_svc_cred(&clp->cl_cred);
1058 kfree(clp->cl_name.data);
1063 release_session_client(struct nfsd4_session *session)
1065 struct nfs4_client *clp = session->se_client;
1066 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1068 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
1070 if (is_client_expired(clp)) {
1072 session->se_client = NULL;
1074 renew_client_locked(clp);
1075 spin_unlock(&nn->client_lock);
1078 /* must be called under the client_lock */
1080 unhash_client_locked(struct nfs4_client *clp)
1082 struct nfsd4_session *ses;
1084 mark_client_expired(clp);
1085 list_del(&clp->cl_lru);
1086 spin_lock(&clp->cl_lock);
1087 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1088 list_del_init(&ses->se_hash);
1089 spin_unlock(&clp->cl_lock);
1093 destroy_client(struct nfs4_client *clp)
1095 struct nfs4_openowner *oo;
1096 struct nfs4_delegation *dp;
1097 struct list_head reaplist;
1098 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1100 INIT_LIST_HEAD(&reaplist);
1101 spin_lock(&recall_lock);
1102 while (!list_empty(&clp->cl_delegations)) {
1103 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1104 list_del_init(&dp->dl_perclnt);
1105 list_move(&dp->dl_recall_lru, &reaplist);
1107 spin_unlock(&recall_lock);
1108 while (!list_empty(&reaplist)) {
1109 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1110 unhash_delegation(dp);
1112 while (!list_empty(&clp->cl_openowners)) {
1113 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1114 release_openowner(oo);
1116 nfsd4_shutdown_callback(clp);
1117 if (clp->cl_cb_conn.cb_xprt)
1118 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1119 list_del(&clp->cl_idhash);
1120 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1121 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1123 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1124 spin_lock(&nn->client_lock);
1125 unhash_client_locked(clp);
1126 if (atomic_read(&clp->cl_refcount) == 0)
1128 spin_unlock(&nn->client_lock);
1131 static void expire_client(struct nfs4_client *clp)
1133 nfsd4_client_record_remove(clp);
1134 destroy_client(clp);
1137 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1139 memcpy(target->cl_verifier.data, source->data,
1140 sizeof(target->cl_verifier.data));
1143 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1145 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1146 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1149 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1151 if (source->cr_principal) {
1152 target->cr_principal =
1153 kstrdup(source->cr_principal, GFP_KERNEL);
1154 if (target->cr_principal == NULL)
1157 target->cr_principal = NULL;
1158 target->cr_flavor = source->cr_flavor;
1159 target->cr_uid = source->cr_uid;
1160 target->cr_gid = source->cr_gid;
1161 target->cr_group_info = source->cr_group_info;
1162 get_group_info(target->cr_group_info);
1167 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1171 res = o1->len - o2->len;
1174 return (long long)memcmp(o1->data, o2->data, o1->len);
1177 static int same_name(const char *n1, const char *n2)
1179 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1183 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1185 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1189 same_clid(clientid_t *cl1, clientid_t *cl2)
1191 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1194 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1198 if (g1->ngroups != g2->ngroups)
1200 for (i=0; i<g1->ngroups; i++)
1201 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1207 * RFC 3530 language requires clid_inuse be returned when the
1208 * "principal" associated with a requests differs from that previously
1209 * used. We use uid, gid's, and gss principal string as our best
1210 * approximation. We also don't want to allow non-gss use of a client
1211 * established using gss: in theory cr_principal should catch that
1212 * change, but in practice cr_principal can be null even in the gss case
1213 * since gssd doesn't always pass down a principal string.
1215 static bool is_gss_cred(struct svc_cred *cr)
1217 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1218 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1223 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1225 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1226 || (cr1->cr_uid != cr2->cr_uid)
1227 || (cr1->cr_gid != cr2->cr_gid)
1228 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1230 if (cr1->cr_principal == cr2->cr_principal)
1232 if (!cr1->cr_principal || !cr2->cr_principal)
1234 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1237 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1239 static u32 current_clientid = 1;
1241 clp->cl_clientid.cl_boot = nn->boot_time;
1242 clp->cl_clientid.cl_id = current_clientid++;
1245 static void gen_confirm(struct nfs4_client *clp)
1250 verf[0] = (__be32)get_seconds();
1251 verf[1] = (__be32)i++;
1252 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1255 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1257 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1260 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1262 struct nfs4_stid *s;
1264 s = find_stateid(cl, t);
1267 if (typemask & s->sc_type)
1272 static struct nfs4_client *create_client(struct xdr_netobj name,
1273 struct svc_rqst *rqstp, nfs4_verifier *verf)
1275 struct nfs4_client *clp;
1276 struct sockaddr *sa = svc_addr(rqstp);
1278 struct net *net = SVC_NET(rqstp);
1279 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1281 clp = alloc_client(name);
1285 INIT_LIST_HEAD(&clp->cl_sessions);
1286 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1288 spin_lock(&nn->client_lock);
1290 spin_unlock(&nn->client_lock);
1293 idr_init(&clp->cl_stateids);
1294 atomic_set(&clp->cl_refcount, 0);
1295 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1296 INIT_LIST_HEAD(&clp->cl_idhash);
1297 INIT_LIST_HEAD(&clp->cl_openowners);
1298 INIT_LIST_HEAD(&clp->cl_delegations);
1299 INIT_LIST_HEAD(&clp->cl_lru);
1300 INIT_LIST_HEAD(&clp->cl_callbacks);
1301 spin_lock_init(&clp->cl_lock);
1302 nfsd4_init_callback(&clp->cl_cb_null);
1303 clp->cl_time = get_seconds();
1304 clear_bit(0, &clp->cl_cb_slot_busy);
1305 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1306 copy_verf(clp, verf);
1307 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1309 clp->cl_cb_session = NULL;
1315 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1317 struct rb_node **new = &(root->rb_node), *parent = NULL;
1318 struct nfs4_client *clp;
1321 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1324 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1325 new = &((*new)->rb_left);
1327 new = &((*new)->rb_right);
1330 rb_link_node(&new_clp->cl_namenode, parent, new);
1331 rb_insert_color(&new_clp->cl_namenode, root);
1334 static struct nfs4_client *
1335 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1338 struct rb_node *node = root->rb_node;
1339 struct nfs4_client *clp;
1342 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1343 cmp = compare_blob(&clp->cl_name, name);
1345 node = node->rb_left;
1347 node = node->rb_right;
1355 add_to_unconfirmed(struct nfs4_client *clp)
1357 unsigned int idhashval;
1358 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1360 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1361 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1362 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1363 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1368 move_to_confirmed(struct nfs4_client *clp)
1370 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1371 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1373 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1374 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1375 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1376 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1377 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1381 static struct nfs4_client *
1382 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1384 struct nfs4_client *clp;
1385 unsigned int idhashval = clientid_hashval(clid->cl_id);
1387 list_for_each_entry(clp, &nn->conf_id_hashtbl[idhashval], cl_idhash) {
1388 if (same_clid(&clp->cl_clientid, clid)) {
1389 if ((bool)clp->cl_minorversion != sessions)
1398 static struct nfs4_client *
1399 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1401 struct nfs4_client *clp;
1402 unsigned int idhashval = clientid_hashval(clid->cl_id);
1404 list_for_each_entry(clp, &nn->unconf_id_hashtbl[idhashval], cl_idhash) {
1405 if (same_clid(&clp->cl_clientid, clid)) {
1406 if ((bool)clp->cl_minorversion != sessions)
1414 static bool clp_used_exchangeid(struct nfs4_client *clp)
1416 return clp->cl_exchange_flags != 0;
1419 static struct nfs4_client *
1420 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1422 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1425 static struct nfs4_client *
1426 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1428 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1432 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1434 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1435 struct sockaddr *sa = svc_addr(rqstp);
1436 u32 scopeid = rpc_get_scope_id(sa);
1437 unsigned short expected_family;
1439 /* Currently, we only support tcp and tcp6 for the callback channel */
1440 if (se->se_callback_netid_len == 3 &&
1441 !memcmp(se->se_callback_netid_val, "tcp", 3))
1442 expected_family = AF_INET;
1443 else if (se->se_callback_netid_len == 4 &&
1444 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1445 expected_family = AF_INET6;
1449 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1450 se->se_callback_addr_len,
1451 (struct sockaddr *)&conn->cb_addr,
1452 sizeof(conn->cb_addr));
1454 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1457 if (conn->cb_addr.ss_family == AF_INET6)
1458 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1460 conn->cb_prog = se->se_callback_prog;
1461 conn->cb_ident = se->se_callback_ident;
1462 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1465 conn->cb_addr.ss_family = AF_UNSPEC;
1466 conn->cb_addrlen = 0;
1467 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1468 "will not receive delegations\n",
1469 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1475 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1478 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1480 struct nfsd4_slot *slot = resp->cstate.slot;
1483 dprintk("--> %s slot %p\n", __func__, slot);
1485 slot->sl_opcnt = resp->opcnt;
1486 slot->sl_status = resp->cstate.status;
1488 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1489 if (nfsd4_not_cached(resp)) {
1490 slot->sl_datalen = 0;
1493 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1494 base = (char *)resp->cstate.datap -
1495 (char *)resp->xbuf->head[0].iov_base;
1496 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1498 WARN("%s: sessions DRC could not cache compound\n", __func__);
1503 * Encode the replay sequence operation from the slot values.
1504 * If cachethis is FALSE encode the uncached rep error on the next
1505 * operation which sets resp->p and increments resp->opcnt for
1506 * nfs4svc_encode_compoundres.
1510 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1511 struct nfsd4_compoundres *resp)
1513 struct nfsd4_op *op;
1514 struct nfsd4_slot *slot = resp->cstate.slot;
1516 /* Encode the replayed sequence operation */
1517 op = &args->ops[resp->opcnt - 1];
1518 nfsd4_encode_operation(resp, op);
1520 /* Return nfserr_retry_uncached_rep in next operation. */
1521 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1522 op = &args->ops[resp->opcnt++];
1523 op->status = nfserr_retry_uncached_rep;
1524 nfsd4_encode_operation(resp, op);
1530 * The sequence operation is not cached because we can use the slot and
1534 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1535 struct nfsd4_sequence *seq)
1537 struct nfsd4_slot *slot = resp->cstate.slot;
1540 dprintk("--> %s slot %p\n", __func__, slot);
1542 /* Either returns 0 or nfserr_retry_uncached */
1543 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1544 if (status == nfserr_retry_uncached_rep)
1547 /* The sequence operation has been encoded, cstate->datap set. */
1548 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1550 resp->opcnt = slot->sl_opcnt;
1551 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1552 status = slot->sl_status;
1558 * Set the exchange_id flags returned by the server.
1561 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1563 /* pNFS is not supported */
1564 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1566 /* Referrals are supported, Migration is not. */
1567 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1569 /* set the wire flags to return to client. */
1570 clid->flags = new->cl_exchange_flags;
1573 static bool client_has_state(struct nfs4_client *clp)
1576 * Note clp->cl_openowners check isn't quite right: there's no
1577 * need to count owners without stateid's.
1579 * Also note we should probably be using this in 4.0 case too.
1581 return !list_empty(&clp->cl_openowners)
1582 || !list_empty(&clp->cl_delegations)
1583 || !list_empty(&clp->cl_sessions);
1587 nfsd4_exchange_id(struct svc_rqst *rqstp,
1588 struct nfsd4_compound_state *cstate,
1589 struct nfsd4_exchange_id *exid)
1591 struct nfs4_client *unconf, *conf, *new;
1593 char addr_str[INET6_ADDRSTRLEN];
1594 nfs4_verifier verf = exid->verifier;
1595 struct sockaddr *sa = svc_addr(rqstp);
1596 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1597 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1599 rpc_ntop(sa, addr_str, sizeof(addr_str));
1600 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1601 "ip_addr=%s flags %x, spa_how %d\n",
1602 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1603 addr_str, exid->flags, exid->spa_how);
1605 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1606 return nfserr_inval;
1608 /* Currently only support SP4_NONE */
1609 switch (exid->spa_how) {
1612 default: /* checked by xdr code */
1616 return nfserr_serverfault; /* no excuse :-/ */
1619 /* Cases below refer to rfc 5661 section 18.35.4: */
1621 conf = find_confirmed_client_by_name(&exid->clname, nn);
1623 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1624 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1627 if (!clp_used_exchangeid(conf)) { /* buggy client */
1628 status = nfserr_inval;
1631 if (!creds_match) { /* case 9 */
1632 status = nfserr_perm;
1635 if (!verfs_match) { /* case 8 */
1636 status = nfserr_not_same;
1640 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1644 if (!creds_match) { /* case 3 */
1645 if (client_has_state(conf)) {
1646 status = nfserr_clid_inuse;
1649 expire_client(conf);
1652 if (verfs_match) { /* case 2 */
1653 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1657 /* case 5, client reboot */
1661 if (update) { /* case 7 */
1662 status = nfserr_noent;
1666 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1667 if (unconf) /* case 4, possible retry or client restart */
1668 expire_client(unconf);
1670 /* case 1 (normal case) */
1672 new = create_client(exid->clname, rqstp, &verf);
1674 status = nfserr_jukebox;
1677 new->cl_minorversion = 1;
1680 add_to_unconfirmed(new);
1682 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1683 exid->clientid.cl_id = new->cl_clientid.cl_id;
1685 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1686 nfsd4_set_ex_flags(new, exid);
1688 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1689 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1693 nfs4_unlock_state();
1698 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1700 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1703 /* The slot is in use, and no response has been sent. */
1705 if (seqid == slot_seqid)
1706 return nfserr_jukebox;
1708 return nfserr_seq_misordered;
1710 /* Note unsigned 32-bit arithmetic handles wraparound: */
1711 if (likely(seqid == slot_seqid + 1))
1713 if (seqid == slot_seqid)
1714 return nfserr_replay_cache;
1715 return nfserr_seq_misordered;
1719 * Cache the create session result into the create session single DRC
1720 * slot cache by saving the xdr structure. sl_seqid has been set.
1721 * Do this for solo or embedded create session operations.
1724 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1725 struct nfsd4_clid_slot *slot, __be32 nfserr)
1727 slot->sl_status = nfserr;
1728 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1732 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1733 struct nfsd4_clid_slot *slot)
1735 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1736 return slot->sl_status;
1739 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1740 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1741 1 + /* MIN tag is length with zero, only length */ \
1742 3 + /* version, opcount, opcode */ \
1743 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1744 /* seqid, slotID, slotID, cache */ \
1745 4 ) * sizeof(__be32))
1747 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1748 2 + /* verifier: AUTH_NULL, length 0 */\
1750 1 + /* MIN tag is length with zero, only length */ \
1751 3 + /* opcount, opcode, opstatus*/ \
1752 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1753 /* seqid, slotID, slotID, slotID, status */ \
1754 5 ) * sizeof(__be32))
1756 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1758 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1759 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1763 nfsd4_create_session(struct svc_rqst *rqstp,
1764 struct nfsd4_compound_state *cstate,
1765 struct nfsd4_create_session *cr_ses)
1767 struct sockaddr *sa = svc_addr(rqstp);
1768 struct nfs4_client *conf, *unconf;
1769 struct nfsd4_session *new;
1770 struct nfsd4_conn *conn;
1771 struct nfsd4_clid_slot *cs_slot = NULL;
1773 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1775 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1776 return nfserr_inval;
1777 if (check_forechannel_attrs(cr_ses->fore_channel))
1778 return nfserr_toosmall;
1779 new = alloc_session(&cr_ses->fore_channel);
1781 return nfserr_jukebox;
1782 status = nfserr_jukebox;
1783 conn = alloc_conn_from_crses(rqstp, cr_ses);
1785 goto out_free_session;
1788 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1789 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1792 cs_slot = &conf->cl_cs_slot;
1793 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1794 if (status == nfserr_replay_cache) {
1795 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1797 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1798 status = nfserr_seq_misordered;
1801 } else if (unconf) {
1802 struct nfs4_client *old;
1803 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1804 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1805 status = nfserr_clid_inuse;
1808 cs_slot = &unconf->cl_cs_slot;
1809 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1811 /* an unconfirmed replay returns misordered */
1812 status = nfserr_seq_misordered;
1815 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1818 move_to_confirmed(unconf);
1821 status = nfserr_stale_clientid;
1826 * We do not support RDMA or persistent sessions
1828 cr_ses->flags &= ~SESSION4_PERSIST;
1829 cr_ses->flags &= ~SESSION4_RDMA;
1831 init_session(rqstp, new, conf, cr_ses);
1832 nfsd4_init_conn(rqstp, conn, new);
1834 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1835 NFS4_MAX_SESSIONID_LEN);
1836 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1837 sizeof(struct nfsd4_channel_attrs));
1838 cs_slot->sl_seqid++;
1839 cr_ses->seqid = cs_slot->sl_seqid;
1841 /* cache solo and embedded create sessions under the state lock */
1842 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1844 nfs4_unlock_state();
1845 dprintk("%s returns %d\n", __func__, ntohl(status));
1850 __free_session(new);
1854 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1856 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1857 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1859 return argp->opcnt == resp->opcnt;
1862 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1865 case NFS4_CDFC4_FORE:
1866 case NFS4_CDFC4_BACK:
1868 case NFS4_CDFC4_FORE_OR_BOTH:
1869 case NFS4_CDFC4_BACK_OR_BOTH:
1870 *dir = NFS4_CDFC4_BOTH;
1873 return nfserr_inval;
1876 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1878 struct nfsd4_session *session = cstate->session;
1879 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1881 spin_lock(&nn->client_lock);
1882 session->se_cb_prog = bc->bc_cb_program;
1883 session->se_cb_sec = bc->bc_cb_sec;
1884 spin_unlock(&nn->client_lock);
1886 nfsd4_probe_callback(session->se_client);
1891 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1892 struct nfsd4_compound_state *cstate,
1893 struct nfsd4_bind_conn_to_session *bcts)
1896 struct nfsd4_conn *conn;
1897 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1899 if (!nfsd4_last_compound_op(rqstp))
1900 return nfserr_not_only_op;
1901 spin_lock(&nn->client_lock);
1902 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
1903 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1904 * client_lock iself: */
1905 if (cstate->session) {
1906 nfsd4_get_session(cstate->session);
1907 atomic_inc(&cstate->session->se_client->cl_refcount);
1909 spin_unlock(&nn->client_lock);
1910 if (!cstate->session)
1911 return nfserr_badsession;
1913 status = nfsd4_map_bcts_dir(&bcts->dir);
1916 conn = alloc_conn(rqstp, bcts->dir);
1918 return nfserr_jukebox;
1919 nfsd4_init_conn(rqstp, conn, cstate->session);
1923 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1927 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1931 nfsd4_destroy_session(struct svc_rqst *r,
1932 struct nfsd4_compound_state *cstate,
1933 struct nfsd4_destroy_session *sessionid)
1935 struct nfsd4_session *ses;
1936 __be32 status = nfserr_badsession;
1937 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
1940 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1941 * - Should we return nfserr_back_chan_busy if waiting for
1942 * callbacks on to-be-destroyed session?
1943 * - Do we need to clear any callback info from previous session?
1946 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1947 if (!nfsd4_last_compound_op(r))
1948 return nfserr_not_only_op;
1950 dump_sessionid(__func__, &sessionid->sessionid);
1951 spin_lock(&nn->client_lock);
1952 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
1954 spin_unlock(&nn->client_lock);
1958 unhash_session(ses);
1959 spin_unlock(&nn->client_lock);
1962 nfsd4_probe_callback_sync(ses->se_client);
1963 nfs4_unlock_state();
1965 spin_lock(&nn->client_lock);
1966 nfsd4_del_conns(ses);
1967 nfsd4_put_session_locked(ses);
1968 spin_unlock(&nn->client_lock);
1971 dprintk("%s returns %d\n", __func__, ntohl(status));
1975 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1977 struct nfsd4_conn *c;
1979 list_for_each_entry(c, &s->se_conns, cn_persession) {
1980 if (c->cn_xprt == xpt) {
1987 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1989 struct nfs4_client *clp = ses->se_client;
1990 struct nfsd4_conn *c;
1993 spin_lock(&clp->cl_lock);
1994 c = __nfsd4_find_conn(new->cn_xprt, ses);
1996 spin_unlock(&clp->cl_lock);
2000 __nfsd4_hash_conn(new, ses);
2001 spin_unlock(&clp->cl_lock);
2002 ret = nfsd4_register_conn(new);
2004 /* oops; xprt is already down: */
2005 nfsd4_conn_lost(&new->cn_xpt_user);
2009 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2011 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2013 return args->opcnt > session->se_fchannel.maxops;
2016 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2017 struct nfsd4_session *session)
2019 struct xdr_buf *xb = &rqstp->rq_arg;
2021 return xb->len > session->se_fchannel.maxreq_sz;
2025 nfsd4_sequence(struct svc_rqst *rqstp,
2026 struct nfsd4_compound_state *cstate,
2027 struct nfsd4_sequence *seq)
2029 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2030 struct nfsd4_session *session;
2031 struct nfsd4_slot *slot;
2032 struct nfsd4_conn *conn;
2034 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2036 if (resp->opcnt != 1)
2037 return nfserr_sequence_pos;
2040 * Will be either used or freed by nfsd4_sequence_check_conn
2043 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2045 return nfserr_jukebox;
2047 spin_lock(&nn->client_lock);
2048 status = nfserr_badsession;
2049 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2053 status = nfserr_too_many_ops;
2054 if (nfsd4_session_too_many_ops(rqstp, session))
2057 status = nfserr_req_too_big;
2058 if (nfsd4_request_too_big(rqstp, session))
2061 status = nfserr_badslot;
2062 if (seq->slotid >= session->se_fchannel.maxreqs)
2065 slot = session->se_slots[seq->slotid];
2066 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2068 /* We do not negotiate the number of slots yet, so set the
2069 * maxslots to the session maxreqs which is used to encode
2070 * sr_highest_slotid and the sr_target_slot id to maxslots */
2071 seq->maxslots = session->se_fchannel.maxreqs;
2073 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2074 slot->sl_flags & NFSD4_SLOT_INUSE);
2075 if (status == nfserr_replay_cache) {
2076 status = nfserr_seq_misordered;
2077 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2079 cstate->slot = slot;
2080 cstate->session = session;
2081 /* Return the cached reply status and set cstate->status
2082 * for nfsd4_proc_compound processing */
2083 status = nfsd4_replay_cache_entry(resp, seq);
2084 cstate->status = nfserr_replay_cache;
2090 nfsd4_sequence_check_conn(conn, session);
2093 /* Success! bump slot seqid */
2094 slot->sl_seqid = seq->seqid;
2095 slot->sl_flags |= NFSD4_SLOT_INUSE;
2097 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2099 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2101 cstate->slot = slot;
2102 cstate->session = session;
2105 /* Hold a session reference until done processing the compound. */
2106 if (cstate->session) {
2107 struct nfs4_client *clp = session->se_client;
2109 nfsd4_get_session(cstate->session);
2110 atomic_inc(&clp->cl_refcount);
2111 switch (clp->cl_cb_state) {
2113 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2115 case NFSD4_CB_FAULT:
2116 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2119 seq->status_flags = 0;
2123 spin_unlock(&nn->client_lock);
2124 dprintk("%s: return %d\n", __func__, ntohl(status));
2129 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2131 struct nfs4_client *conf, *unconf, *clp;
2133 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2136 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2137 conf = find_confirmed_client(&dc->clientid, true, nn);
2142 if (!is_client_expired(conf) && client_has_state(conf)) {
2143 status = nfserr_clientid_busy;
2147 /* rfc5661 18.50.3 */
2148 if (cstate->session && conf == cstate->session->se_client) {
2149 status = nfserr_clientid_busy;
2155 status = nfserr_stale_clientid;
2161 nfs4_unlock_state();
2162 dprintk("%s return %d\n", __func__, ntohl(status));
2167 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2171 if (rc->rca_one_fs) {
2172 if (!cstate->current_fh.fh_dentry)
2173 return nfserr_nofilehandle;
2175 * We don't take advantage of the rca_one_fs case.
2176 * That's OK, it's optional, we can safely ignore it.
2182 status = nfserr_complete_already;
2183 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2184 &cstate->session->se_client->cl_flags))
2187 status = nfserr_stale_clientid;
2188 if (is_client_expired(cstate->session->se_client))
2190 * The following error isn't really legal.
2191 * But we only get here if the client just explicitly
2192 * destroyed the client. Surely it no longer cares what
2193 * error it gets back on an operation for the dead
2199 nfsd4_client_record_create(cstate->session->se_client);
2201 nfs4_unlock_state();
2206 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2207 struct nfsd4_setclientid *setclid)
2209 struct xdr_netobj clname = setclid->se_name;
2210 nfs4_verifier clverifier = setclid->se_verf;
2211 struct nfs4_client *conf, *unconf, *new;
2213 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2215 /* Cases below refer to rfc 3530 section 14.2.33: */
2217 conf = find_confirmed_client_by_name(&clname, nn);
2220 status = nfserr_clid_inuse;
2221 if (clp_used_exchangeid(conf))
2223 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2224 char addr_str[INET6_ADDRSTRLEN];
2225 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2227 dprintk("NFSD: setclientid: string in use by client "
2228 "at %s\n", addr_str);
2232 unconf = find_unconfirmed_client_by_name(&clname, nn);
2234 expire_client(unconf);
2235 status = nfserr_jukebox;
2236 new = create_client(clname, rqstp, &clverifier);
2239 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2240 /* case 1: probable callback update */
2241 copy_clid(new, conf);
2242 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2244 new->cl_minorversion = 0;
2245 gen_callback(new, setclid, rqstp);
2246 add_to_unconfirmed(new);
2247 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2248 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2249 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2252 nfs4_unlock_state();
2258 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2259 struct nfsd4_compound_state *cstate,
2260 struct nfsd4_setclientid_confirm *setclientid_confirm)
2262 struct nfs4_client *conf, *unconf;
2263 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2264 clientid_t * clid = &setclientid_confirm->sc_clientid;
2266 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2268 if (STALE_CLIENTID(clid, nn))
2269 return nfserr_stale_clientid;
2272 conf = find_confirmed_client(clid, false, nn);
2273 unconf = find_unconfirmed_client(clid, false, nn);
2275 * We try hard to give out unique clientid's, so if we get an
2276 * attempt to confirm the same clientid with a different cred,
2277 * there's a bug somewhere. Let's charitably assume it's our
2280 status = nfserr_serverfault;
2281 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2283 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2285 /* cases below refer to rfc 3530 section 14.2.34: */
2286 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2287 if (conf && !unconf) /* case 2: probable retransmit */
2289 else /* case 4: client hasn't noticed we rebooted yet? */
2290 status = nfserr_stale_clientid;
2294 if (conf) { /* case 1: callback update */
2295 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2296 nfsd4_probe_callback(conf);
2297 expire_client(unconf);
2298 } else { /* case 3: normal case; new or rebooted client */
2299 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2301 expire_client(conf);
2302 move_to_confirmed(unconf);
2303 nfsd4_probe_callback(unconf);
2306 nfs4_unlock_state();
2310 static struct nfs4_file *nfsd4_alloc_file(void)
2312 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2315 /* OPEN Share state helper functions */
2316 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2318 unsigned int hashval = file_hashval(ino);
2320 atomic_set(&fp->fi_ref, 1);
2321 INIT_LIST_HEAD(&fp->fi_hash);
2322 INIT_LIST_HEAD(&fp->fi_stateids);
2323 INIT_LIST_HEAD(&fp->fi_delegations);
2324 fp->fi_inode = igrab(ino);
2325 fp->fi_had_conflict = false;
2326 fp->fi_lease = NULL;
2327 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2328 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2329 spin_lock(&recall_lock);
2330 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2331 spin_unlock(&recall_lock);
2335 nfsd4_free_slab(struct kmem_cache **slab)
2339 kmem_cache_destroy(*slab);
2344 nfsd4_free_slabs(void)
2346 nfsd4_free_slab(&openowner_slab);
2347 nfsd4_free_slab(&lockowner_slab);
2348 nfsd4_free_slab(&file_slab);
2349 nfsd4_free_slab(&stateid_slab);
2350 nfsd4_free_slab(&deleg_slab);
2354 nfsd4_init_slabs(void)
2356 openowner_slab = kmem_cache_create("nfsd4_openowners",
2357 sizeof(struct nfs4_openowner), 0, 0, NULL);
2358 if (openowner_slab == NULL)
2360 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2361 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2362 if (lockowner_slab == NULL)
2364 file_slab = kmem_cache_create("nfsd4_files",
2365 sizeof(struct nfs4_file), 0, 0, NULL);
2366 if (file_slab == NULL)
2368 stateid_slab = kmem_cache_create("nfsd4_stateids",
2369 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2370 if (stateid_slab == NULL)
2372 deleg_slab = kmem_cache_create("nfsd4_delegations",
2373 sizeof(struct nfs4_delegation), 0, 0, NULL);
2374 if (deleg_slab == NULL)
2379 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2383 void nfs4_free_openowner(struct nfs4_openowner *oo)
2385 kfree(oo->oo_owner.so_owner.data);
2386 kmem_cache_free(openowner_slab, oo);
2389 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2391 kfree(lo->lo_owner.so_owner.data);
2392 kmem_cache_free(lockowner_slab, lo);
2395 static void init_nfs4_replay(struct nfs4_replay *rp)
2397 rp->rp_status = nfserr_serverfault;
2399 rp->rp_buf = rp->rp_ibuf;
2402 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2404 struct nfs4_stateowner *sop;
2406 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2410 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2411 if (!sop->so_owner.data) {
2412 kmem_cache_free(slab, sop);
2415 sop->so_owner.len = owner->len;
2417 INIT_LIST_HEAD(&sop->so_stateids);
2418 sop->so_client = clp;
2419 init_nfs4_replay(&sop->so_replay);
2423 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2425 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2427 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2428 list_add(&oo->oo_perclient, &clp->cl_openowners);
2431 static struct nfs4_openowner *
2432 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2433 struct nfs4_openowner *oo;
2435 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2438 oo->oo_owner.so_is_open_owner = 1;
2439 oo->oo_owner.so_seqid = open->op_seqid;
2440 oo->oo_flags = NFS4_OO_NEW;
2442 oo->oo_last_closed_stid = NULL;
2443 INIT_LIST_HEAD(&oo->oo_close_lru);
2444 hash_openowner(oo, clp, strhashval);
2448 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2449 struct nfs4_openowner *oo = open->op_openowner;
2450 struct nfs4_client *clp = oo->oo_owner.so_client;
2452 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2453 INIT_LIST_HEAD(&stp->st_lockowners);
2454 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2455 list_add(&stp->st_perfile, &fp->fi_stateids);
2456 stp->st_stateowner = &oo->oo_owner;
2459 stp->st_access_bmap = 0;
2460 stp->st_deny_bmap = 0;
2461 set_access(open->op_share_access, stp);
2462 set_deny(open->op_share_deny, stp);
2463 stp->st_openstp = NULL;
2467 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2469 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2471 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2473 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2474 oo->oo_time = get_seconds();
2478 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2481 return (sop->so_owner.len == owner->len) &&
2482 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2483 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2486 static struct nfs4_openowner *
2487 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2488 bool sessions, struct nfsd_net *nn)
2490 struct nfs4_stateowner *so;
2491 struct nfs4_openowner *oo;
2492 struct nfs4_client *clp;
2494 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2495 if (!so->so_is_open_owner)
2497 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2499 clp = oo->oo_owner.so_client;
2500 if ((bool)clp->cl_minorversion != sessions)
2502 renew_client(oo->oo_owner.so_client);
2509 /* search file_hashtbl[] for file */
2510 static struct nfs4_file *
2511 find_file(struct inode *ino)
2513 unsigned int hashval = file_hashval(ino);
2514 struct nfs4_file *fp;
2516 spin_lock(&recall_lock);
2517 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2518 if (fp->fi_inode == ino) {
2520 spin_unlock(&recall_lock);
2524 spin_unlock(&recall_lock);
2529 * Called to check deny when READ with all zero stateid or
2530 * WRITE with all zero or all one stateid
2533 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2535 struct inode *ino = current_fh->fh_dentry->d_inode;
2536 struct nfs4_file *fp;
2537 struct nfs4_ol_stateid *stp;
2540 dprintk("NFSD: nfs4_share_conflict\n");
2542 fp = find_file(ino);
2545 ret = nfserr_locked;
2546 /* Search for conflicting share reservations */
2547 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2548 if (test_deny(deny_type, stp) ||
2549 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2558 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2560 /* We're assuming the state code never drops its reference
2561 * without first removing the lease. Since we're in this lease
2562 * callback (and since the lease code is serialized by the kernel
2563 * lock) we know the server hasn't removed the lease yet, we know
2564 * it's safe to take a reference: */
2565 atomic_inc(&dp->dl_count);
2567 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2569 /* only place dl_time is set. protected by lock_flocks*/
2570 dp->dl_time = get_seconds();
2572 nfsd4_cb_recall(dp);
2575 /* Called from break_lease() with lock_flocks() held. */
2576 static void nfsd_break_deleg_cb(struct file_lock *fl)
2578 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2579 struct nfs4_delegation *dp;
2582 WARN(1, "(%p)->fl_owner NULL\n", fl);
2585 if (fp->fi_had_conflict) {
2586 WARN(1, "duplicate break on %p\n", fp);
2590 * We don't want the locks code to timeout the lease for us;
2591 * we'll remove it ourself if a delegation isn't returned
2594 fl->fl_break_time = 0;
2596 spin_lock(&recall_lock);
2597 fp->fi_had_conflict = true;
2598 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2599 nfsd_break_one_deleg(dp);
2600 spin_unlock(&recall_lock);
2604 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2607 return lease_modify(onlist, arg);
2612 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2613 .lm_break = nfsd_break_deleg_cb,
2614 .lm_change = nfsd_change_deleg_cb,
2617 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2619 if (nfsd4_has_session(cstate))
2621 if (seqid == so->so_seqid - 1)
2622 return nfserr_replay_me;
2623 if (seqid == so->so_seqid)
2625 return nfserr_bad_seqid;
2629 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2630 struct nfsd4_open *open, struct nfsd_net *nn)
2632 clientid_t *clientid = &open->op_clientid;
2633 struct nfs4_client *clp = NULL;
2634 unsigned int strhashval;
2635 struct nfs4_openowner *oo = NULL;
2638 if (STALE_CLIENTID(&open->op_clientid, nn))
2639 return nfserr_stale_clientid;
2641 * In case we need it later, after we've already created the
2642 * file and don't want to risk a further failure:
2644 open->op_file = nfsd4_alloc_file();
2645 if (open->op_file == NULL)
2646 return nfserr_jukebox;
2648 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2649 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2650 open->op_openowner = oo;
2652 clp = find_confirmed_client(clientid, cstate->minorversion,
2655 return nfserr_expired;
2658 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2659 /* Replace unconfirmed owners without checking for replay. */
2660 clp = oo->oo_owner.so_client;
2661 release_openowner(oo);
2662 open->op_openowner = NULL;
2665 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2668 clp = oo->oo_owner.so_client;
2671 oo = alloc_init_open_stateowner(strhashval, clp, open);
2673 return nfserr_jukebox;
2674 open->op_openowner = oo;
2676 open->op_stp = nfs4_alloc_stateid(clp);
2678 return nfserr_jukebox;
2682 static inline __be32
2683 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2685 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2686 return nfserr_openmode;
2691 static int share_access_to_flags(u32 share_access)
2693 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2696 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2698 struct nfs4_stid *ret;
2700 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2703 return delegstateid(ret);
2706 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2708 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2709 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2713 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2714 struct nfs4_delegation **dp)
2717 __be32 status = nfserr_bad_stateid;
2719 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2722 flags = share_access_to_flags(open->op_share_access);
2723 status = nfs4_check_delegmode(*dp, flags);
2727 if (!nfsd4_is_deleg_cur(open))
2731 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2736 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2738 struct nfs4_ol_stateid *local;
2739 struct nfs4_openowner *oo = open->op_openowner;
2741 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2742 /* ignore lock owners */
2743 if (local->st_stateowner->so_is_open_owner == 0)
2745 /* remember if we have seen this open owner */
2746 if (local->st_stateowner == &oo->oo_owner)
2748 /* check for conflicting share reservations */
2749 if (!test_share(local, open))
2750 return nfserr_share_denied;
2755 static inline int nfs4_access_to_access(u32 nfs4_access)
2759 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2760 flags |= NFSD_MAY_READ;
2761 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2762 flags |= NFSD_MAY_WRITE;
2766 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2767 struct svc_fh *cur_fh, struct nfsd4_open *open)
2770 int oflag = nfs4_access_to_omode(open->op_share_access);
2771 int access = nfs4_access_to_access(open->op_share_access);
2773 if (!fp->fi_fds[oflag]) {
2774 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2775 &fp->fi_fds[oflag]);
2779 nfs4_file_get_access(fp, oflag);
2784 static inline __be32
2785 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2786 struct nfsd4_open *open)
2788 struct iattr iattr = {
2789 .ia_valid = ATTR_SIZE,
2792 if (!open->op_truncate)
2794 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2795 return nfserr_inval;
2796 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2800 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)
2802 u32 op_share_access = open->op_share_access;
2806 new_access = !test_access(op_share_access, stp);
2808 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2812 status = nfsd4_truncate(rqstp, cur_fh, open);
2815 int oflag = nfs4_access_to_omode(op_share_access);
2816 nfs4_file_put_access(fp, oflag);
2820 /* remember the open */
2821 set_access(op_share_access, stp);
2822 set_deny(open->op_share_deny, stp);
2829 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2831 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2834 /* Should we give out recallable state?: */
2835 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2837 if (clp->cl_cb_state == NFSD4_CB_UP)
2840 * In the sessions case, since we don't have to establish a
2841 * separate connection for callbacks, we assume it's OK
2842 * until we hear otherwise:
2844 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2847 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2849 struct file_lock *fl;
2851 fl = locks_alloc_lock();
2854 locks_init_lock(fl);
2855 fl->fl_lmops = &nfsd_lease_mng_ops;
2856 fl->fl_flags = FL_LEASE;
2857 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2858 fl->fl_end = OFFSET_MAX;
2859 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2860 fl->fl_pid = current->tgid;
2864 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2866 struct nfs4_file *fp = dp->dl_file;
2867 struct file_lock *fl;
2870 fl = nfs4_alloc_init_lease(dp, flag);
2873 fl->fl_file = find_readable_file(fp);
2874 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2875 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2877 list_del_init(&dp->dl_perclnt);
2878 locks_free_lock(fl);
2882 fp->fi_deleg_file = get_file(fl->fl_file);
2883 atomic_set(&fp->fi_delegees, 1);
2884 list_add(&dp->dl_perfile, &fp->fi_delegations);
2888 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2890 struct nfs4_file *fp = dp->dl_file;
2893 return nfs4_setlease(dp, flag);
2894 spin_lock(&recall_lock);
2895 if (fp->fi_had_conflict) {
2896 spin_unlock(&recall_lock);
2899 atomic_inc(&fp->fi_delegees);
2900 list_add(&dp->dl_perfile, &fp->fi_delegations);
2901 spin_unlock(&recall_lock);
2902 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2906 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2908 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2909 if (status == -EAGAIN)
2910 open->op_why_no_deleg = WND4_CONTENTION;
2912 open->op_why_no_deleg = WND4_RESOURCE;
2913 switch (open->op_deleg_want) {
2914 case NFS4_SHARE_WANT_READ_DELEG:
2915 case NFS4_SHARE_WANT_WRITE_DELEG:
2916 case NFS4_SHARE_WANT_ANY_DELEG:
2918 case NFS4_SHARE_WANT_CANCEL:
2919 open->op_why_no_deleg = WND4_CANCELLED;
2921 case NFS4_SHARE_WANT_NO_DELEG:
2928 * Attempt to hand out a delegation.
2931 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2932 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2934 struct nfs4_delegation *dp;
2935 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2937 int status = 0, flag = 0;
2939 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2940 flag = NFS4_OPEN_DELEGATE_NONE;
2941 open->op_recall = 0;
2942 switch (open->op_claim_type) {
2943 case NFS4_OPEN_CLAIM_PREVIOUS:
2945 open->op_recall = 1;
2946 flag = open->op_delegate_type;
2947 if (flag == NFS4_OPEN_DELEGATE_NONE)
2950 case NFS4_OPEN_CLAIM_NULL:
2951 /* Let's not give out any delegations till everyone's
2952 * had the chance to reclaim theirs.... */
2953 if (locks_in_grace(net))
2955 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2957 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2958 flag = NFS4_OPEN_DELEGATE_WRITE;
2960 flag = NFS4_OPEN_DELEGATE_READ;
2966 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2969 status = nfs4_set_delegation(dp, flag);
2973 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2975 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2976 STATEID_VAL(&dp->dl_stid.sc_stateid));
2978 open->op_delegate_type = flag;
2979 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2980 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2981 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2982 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2984 /* 4.1 client asking for a delegation? */
2985 if (open->op_deleg_want)
2986 nfsd4_open_deleg_none_ext(open, status);
2990 nfs4_put_delegation(dp);
2992 flag = NFS4_OPEN_DELEGATE_NONE;
2996 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2997 struct nfs4_delegation *dp)
2999 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3000 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3001 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3002 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3003 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3004 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3005 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3006 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3008 /* Otherwise the client must be confused wanting a delegation
3009 * it already has, therefore we don't return
3010 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3015 * called with nfs4_lock_state() held.
3018 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3020 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3021 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3022 struct nfs4_file *fp = NULL;
3023 struct inode *ino = current_fh->fh_dentry->d_inode;
3024 struct nfs4_ol_stateid *stp = NULL;
3025 struct nfs4_delegation *dp = NULL;
3029 * Lookup file; if found, lookup stateid and check open request,
3030 * and check for delegations in the process of being recalled.
3031 * If not found, create the nfs4_file struct
3033 fp = find_file(ino);
3035 if ((status = nfs4_check_open(fp, open, &stp)))
3037 status = nfs4_check_deleg(cl, fp, open, &dp);
3041 status = nfserr_bad_stateid;
3042 if (nfsd4_is_deleg_cur(open))
3044 status = nfserr_jukebox;
3046 open->op_file = NULL;
3047 nfsd4_init_file(fp, ino);
3051 * OPEN the file, or upgrade an existing OPEN.
3052 * If truncate fails, the OPEN fails.
3055 /* Stateid was found, this is an OPEN upgrade */
3056 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3060 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3063 status = nfsd4_truncate(rqstp, current_fh, open);
3067 open->op_stp = NULL;
3068 init_open_stateid(stp, fp, open);
3070 update_stateid(&stp->st_stid.sc_stateid);
3071 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3073 if (nfsd4_has_session(&resp->cstate)) {
3074 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3076 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3077 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3078 open->op_why_no_deleg = WND4_NOT_WANTED;
3084 * Attempt to hand out a delegation. No error return, because the
3085 * OPEN succeeds even if we fail.
3087 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3091 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3092 STATEID_VAL(&stp->st_stid.sc_stateid));
3094 /* 4.1 client trying to upgrade/downgrade delegation? */
3095 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3096 open->op_deleg_want)
3097 nfsd4_deleg_xgrade_none_ext(open, dp);
3101 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3102 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3104 * To finish the open response, we just need to set the rflags.
3106 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3107 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3108 !nfsd4_has_session(&resp->cstate))
3109 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3114 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3116 if (open->op_openowner) {
3117 struct nfs4_openowner *oo = open->op_openowner;
3119 if (!list_empty(&oo->oo_owner.so_stateids))
3120 list_del_init(&oo->oo_close_lru);
3121 if (oo->oo_flags & NFS4_OO_NEW) {
3123 release_openowner(oo);
3124 open->op_openowner = NULL;
3126 oo->oo_flags &= ~NFS4_OO_NEW;
3130 nfsd4_free_file(open->op_file);
3132 free_generic_stateid(open->op_stp);
3135 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3137 struct nfs4_client *found;
3139 if (STALE_CLIENTID(clid, nn))
3140 return nfserr_stale_clientid;
3141 found = find_confirmed_client(clid, session, nn);
3144 return found ? nfs_ok : nfserr_expired;
3148 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3151 struct nfs4_client *clp;
3153 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3156 dprintk("process_renew(%08x/%08x): starting\n",
3157 clid->cl_boot, clid->cl_id);
3158 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3161 status = nfserr_cb_path_down;
3162 if (!list_empty(&clp->cl_delegations)
3163 && clp->cl_cb_state != NFSD4_CB_UP)
3167 nfs4_unlock_state();
3172 nfsd4_end_grace(struct nfsd_net *nn)
3174 /* do nothing if grace period already ended */
3175 if (nn->grace_ended)
3178 dprintk("NFSD: end of grace period\n");
3179 nn->grace_ended = true;
3180 nfsd4_record_grace_done(nn, nn->boot_time);
3181 locks_end_grace(&nn->nfsd4_manager);
3183 * Now that every NFSv4 client has had the chance to recover and
3184 * to see the (possibly new, possibly shorter) lease time, we
3185 * can safely set the next grace time to the current lease time:
3187 nn->nfsd4_grace = nn->nfsd4_lease;
3191 nfs4_laundromat(struct nfsd_net *nn)
3193 struct nfs4_client *clp;
3194 struct nfs4_openowner *oo;
3195 struct nfs4_delegation *dp;
3196 struct list_head *pos, *next, reaplist;
3197 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3198 time_t t, clientid_val = nn->nfsd4_lease;
3199 time_t u, test_val = nn->nfsd4_lease;
3203 dprintk("NFSD: laundromat service - starting\n");
3204 nfsd4_end_grace(nn);
3205 INIT_LIST_HEAD(&reaplist);
3206 spin_lock(&nn->client_lock);
3207 list_for_each_safe(pos, next, &nn->client_lru) {
3208 clp = list_entry(pos, struct nfs4_client, cl_lru);
3209 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3210 t = clp->cl_time - cutoff;
3211 if (clientid_val > t)
3215 if (atomic_read(&clp->cl_refcount)) {
3216 dprintk("NFSD: client in use (clientid %08x)\n",
3217 clp->cl_clientid.cl_id);
3220 unhash_client_locked(clp);
3221 list_add(&clp->cl_lru, &reaplist);
3223 spin_unlock(&nn->client_lock);
3224 list_for_each_safe(pos, next, &reaplist) {
3225 clp = list_entry(pos, struct nfs4_client, cl_lru);
3226 dprintk("NFSD: purging unused client (clientid %08x)\n",
3227 clp->cl_clientid.cl_id);
3230 spin_lock(&recall_lock);
3231 list_for_each_safe(pos, next, &del_recall_lru) {
3232 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3233 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3235 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3236 u = dp->dl_time - cutoff;
3241 list_move(&dp->dl_recall_lru, &reaplist);
3243 spin_unlock(&recall_lock);
3244 list_for_each_safe(pos, next, &reaplist) {
3245 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3246 unhash_delegation(dp);
3248 test_val = nn->nfsd4_lease;
3249 list_for_each_safe(pos, next, &nn->close_lru) {
3250 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3251 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3252 u = oo->oo_time - cutoff;
3257 release_openowner(oo);
3259 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3260 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3261 nfs4_unlock_state();
3262 return clientid_val;
3265 static struct workqueue_struct *laundry_wq;
3266 static void laundromat_main(struct work_struct *);
3269 laundromat_main(struct work_struct *laundry)
3272 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3274 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3277 t = nfs4_laundromat(nn);
3278 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3279 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3282 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3284 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3285 return nfserr_bad_stateid;
3290 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3292 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3294 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3295 STATEID_VAL(stateid));
3300 access_permit_read(struct nfs4_ol_stateid *stp)
3302 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3303 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3304 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3308 access_permit_write(struct nfs4_ol_stateid *stp)
3310 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3311 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3315 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3317 __be32 status = nfserr_openmode;
3319 /* For lock stateid's, we test the parent open, not the lock: */
3320 if (stp->st_openstp)
3321 stp = stp->st_openstp;
3322 if ((flags & WR_STATE) && !access_permit_write(stp))
3324 if ((flags & RD_STATE) && !access_permit_read(stp))
3331 static inline __be32
3332 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3334 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3336 else if (locks_in_grace(net)) {
3337 /* Answer in remaining cases depends on existence of
3338 * conflicting state; so we must wait out the grace period. */
3339 return nfserr_grace;
3340 } else if (flags & WR_STATE)
3341 return nfs4_share_conflict(current_fh,
3342 NFS4_SHARE_DENY_WRITE);
3343 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3344 return nfs4_share_conflict(current_fh,
3345 NFS4_SHARE_DENY_READ);
3349 * Allow READ/WRITE during grace period on recovered state only for files
3350 * that are not able to provide mandatory locking.
3353 grace_disallows_io(struct net *net, struct inode *inode)
3355 return locks_in_grace(net) && mandatory_lock(inode);
3358 /* Returns true iff a is later than b: */
3359 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3361 return (s32)a->si_generation - (s32)b->si_generation > 0;
3364 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3367 * When sessions are used the stateid generation number is ignored
3370 if (has_session && in->si_generation == 0)
3373 if (in->si_generation == ref->si_generation)
3376 /* If the client sends us a stateid from the future, it's buggy: */
3377 if (stateid_generation_after(in, ref))
3378 return nfserr_bad_stateid;
3380 * However, we could see a stateid from the past, even from a
3381 * non-buggy client. For example, if the client sends a lock
3382 * while some IO is outstanding, the lock may bump si_generation
3383 * while the IO is still in flight. The client could avoid that
3384 * situation by waiting for responses on all the IO requests,
3385 * but better performance may result in retrying IO that
3386 * receives an old_stateid error if requests are rarely
3387 * reordered in flight:
3389 return nfserr_old_stateid;
3392 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3394 struct nfs4_stid *s;
3395 struct nfs4_ol_stateid *ols;
3398 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3399 return nfserr_bad_stateid;
3400 /* Client debugging aid. */
3401 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3402 char addr_str[INET6_ADDRSTRLEN];
3403 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3405 pr_warn_ratelimited("NFSD: client %s testing state ID "
3406 "with incorrect client ID\n", addr_str);
3407 return nfserr_bad_stateid;
3409 s = find_stateid(cl, stateid);
3411 return nfserr_bad_stateid;
3412 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3415 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3417 ols = openlockstateid(s);
3418 if (ols->st_stateowner->so_is_open_owner
3419 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3420 return nfserr_bad_stateid;
3424 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3425 struct nfs4_stid **s, bool sessions,
3426 struct nfsd_net *nn)
3428 struct nfs4_client *cl;
3430 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3431 return nfserr_bad_stateid;
3432 if (STALE_STATEID(stateid, nn))
3433 return nfserr_stale_stateid;
3434 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions, nn);
3436 return nfserr_expired;
3437 *s = find_stateid_by_type(cl, stateid, typemask);
3439 return nfserr_bad_stateid;
3445 * Checks for stateid operations
3448 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3449 stateid_t *stateid, int flags, struct file **filpp)
3451 struct nfs4_stid *s;
3452 struct nfs4_ol_stateid *stp = NULL;
3453 struct nfs4_delegation *dp = NULL;
3454 struct svc_fh *current_fh = &cstate->current_fh;
3455 struct inode *ino = current_fh->fh_dentry->d_inode;
3456 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3462 if (grace_disallows_io(net, ino))
3463 return nfserr_grace;
3465 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3466 return check_special_stateids(net, current_fh, stateid, flags);
3468 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3469 &s, cstate->minorversion, nn);
3472 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3475 switch (s->sc_type) {
3476 case NFS4_DELEG_STID:
3477 dp = delegstateid(s);
3478 status = nfs4_check_delegmode(dp, flags);
3482 *filpp = dp->dl_file->fi_deleg_file;
3485 status = nfserr_serverfault;
3490 case NFS4_OPEN_STID:
3491 case NFS4_LOCK_STID:
3492 stp = openlockstateid(s);
3493 status = nfs4_check_fh(current_fh, stp);
3496 if (stp->st_stateowner->so_is_open_owner
3497 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3499 status = nfs4_check_openmode(stp, flags);
3503 if (flags & RD_STATE)
3504 *filpp = find_readable_file(stp->st_file);
3506 *filpp = find_writeable_file(stp->st_file);
3510 return nfserr_bad_stateid;
3518 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3520 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3521 return nfserr_locks_held;
3522 release_lock_stateid(stp);
3527 * Test if the stateid is valid
3530 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3531 struct nfsd4_test_stateid *test_stateid)
3533 struct nfsd4_test_stateid_id *stateid;
3534 struct nfs4_client *cl = cstate->session->se_client;
3537 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3538 stateid->ts_id_status =
3539 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3540 nfs4_unlock_state();
3546 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3547 struct nfsd4_free_stateid *free_stateid)
3549 stateid_t *stateid = &free_stateid->fr_stateid;
3550 struct nfs4_stid *s;
3551 struct nfs4_client *cl = cstate->session->se_client;
3552 __be32 ret = nfserr_bad_stateid;
3555 s = find_stateid(cl, stateid);
3558 switch (s->sc_type) {
3559 case NFS4_DELEG_STID:
3560 ret = nfserr_locks_held;
3562 case NFS4_OPEN_STID:
3563 case NFS4_LOCK_STID:
3564 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3567 if (s->sc_type == NFS4_LOCK_STID)
3568 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3570 ret = nfserr_locks_held;
3573 ret = nfserr_bad_stateid;
3576 nfs4_unlock_state();
3583 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3584 RD_STATE : WR_STATE;
3587 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3589 struct svc_fh *current_fh = &cstate->current_fh;
3590 struct nfs4_stateowner *sop = stp->st_stateowner;
3593 status = nfsd4_check_seqid(cstate, sop, seqid);
3596 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3598 * "Closed" stateid's exist *only* to return
3599 * nfserr_replay_me from the previous step.
3601 return nfserr_bad_stateid;
3602 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3605 return nfs4_check_fh(current_fh, stp);
3609 * Checks for sequence id mutating operations.
3612 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3613 stateid_t *stateid, char typemask,
3614 struct nfs4_ol_stateid **stpp,
3615 struct nfsd_net *nn)
3618 struct nfs4_stid *s;
3620 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3621 seqid, STATEID_VAL(stateid));
3624 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3625 cstate->minorversion, nn);
3628 *stpp = openlockstateid(s);
3629 cstate->replay_owner = (*stpp)->st_stateowner;
3631 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3634 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3635 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3638 struct nfs4_openowner *oo;
3640 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3641 NFS4_OPEN_STID, stpp, nn);
3644 oo = openowner((*stpp)->st_stateowner);
3645 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3646 return nfserr_bad_stateid;
3651 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3652 struct nfsd4_open_confirm *oc)
3655 struct nfs4_openowner *oo;
3656 struct nfs4_ol_stateid *stp;
3657 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3659 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3660 (int)cstate->current_fh.fh_dentry->d_name.len,
3661 cstate->current_fh.fh_dentry->d_name.name);
3663 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3669 status = nfs4_preprocess_seqid_op(cstate,
3670 oc->oc_seqid, &oc->oc_req_stateid,
3671 NFS4_OPEN_STID, &stp, nn);
3674 oo = openowner(stp->st_stateowner);
3675 status = nfserr_bad_stateid;
3676 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3678 oo->oo_flags |= NFS4_OO_CONFIRMED;
3679 update_stateid(&stp->st_stid.sc_stateid);
3680 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3681 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3682 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3684 nfsd4_client_record_create(oo->oo_owner.so_client);
3687 if (!cstate->replay_owner)
3688 nfs4_unlock_state();
3692 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3694 if (!test_access(access, stp))
3696 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3697 clear_access(access, stp);
3700 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3702 switch (to_access) {
3703 case NFS4_SHARE_ACCESS_READ:
3704 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3705 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3707 case NFS4_SHARE_ACCESS_WRITE:
3708 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3709 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3711 case NFS4_SHARE_ACCESS_BOTH:
3719 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3722 for (i = 0; i < 4; i++) {
3723 if ((i & deny) != i)
3729 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3730 struct nfsd4_compound_state *cstate,
3731 struct nfsd4_open_downgrade *od)
3734 struct nfs4_ol_stateid *stp;
3735 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3737 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3738 (int)cstate->current_fh.fh_dentry->d_name.len,
3739 cstate->current_fh.fh_dentry->d_name.name);
3741 /* We don't yet support WANT bits: */
3742 if (od->od_deleg_want)
3743 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3747 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3748 &od->od_stateid, &stp, nn);
3751 status = nfserr_inval;
3752 if (!test_access(od->od_share_access, stp)) {
3753 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3754 stp->st_access_bmap, od->od_share_access);
3757 if (!test_deny(od->od_share_deny, stp)) {
3758 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3759 stp->st_deny_bmap, od->od_share_deny);
3762 nfs4_stateid_downgrade(stp, od->od_share_access);
3764 reset_union_bmap_deny(od->od_share_deny, stp);
3766 update_stateid(&stp->st_stid.sc_stateid);
3767 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3770 if (!cstate->replay_owner)
3771 nfs4_unlock_state();
3775 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3777 struct nfs4_openowner *oo;
3778 struct nfs4_ol_stateid *s;
3780 if (!so->so_is_open_owner)
3783 s = oo->oo_last_closed_stid;
3786 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3787 /* Release the last_closed_stid on the next seqid bump: */
3788 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3791 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3792 release_last_closed_stateid(oo);
3795 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3797 unhash_open_stateid(s);
3798 s->st_stid.sc_type = NFS4_CLOSED_STID;
3802 * nfs4_unlock_state() called after encode
3805 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3806 struct nfsd4_close *close)
3809 struct nfs4_openowner *oo;
3810 struct nfs4_ol_stateid *stp;
3811 struct net *net = SVC_NET(rqstp);
3812 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3814 dprintk("NFSD: nfsd4_close on file %.*s\n",
3815 (int)cstate->current_fh.fh_dentry->d_name.len,
3816 cstate->current_fh.fh_dentry->d_name.name);
3819 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3821 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3825 oo = openowner(stp->st_stateowner);
3827 update_stateid(&stp->st_stid.sc_stateid);
3828 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3830 nfsd4_close_open_stateid(stp);
3831 release_last_closed_stateid(oo);
3832 oo->oo_last_closed_stid = stp;
3834 if (list_empty(&oo->oo_owner.so_stateids)) {
3835 if (cstate->minorversion) {
3836 release_openowner(oo);
3837 cstate->replay_owner = NULL;
3840 * In the 4.0 case we need to keep the owners around a
3841 * little while to handle CLOSE replay.
3843 if (list_empty(&oo->oo_owner.so_stateids))
3844 move_to_close_lru(oo, SVC_NET(rqstp));
3848 if (!cstate->replay_owner)
3849 nfs4_unlock_state();
3854 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3855 struct nfsd4_delegreturn *dr)
3857 struct nfs4_delegation *dp;
3858 stateid_t *stateid = &dr->dr_stateid;
3859 struct nfs4_stid *s;
3861 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3863 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3867 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
3868 cstate->minorversion, nn);
3871 dp = delegstateid(s);
3872 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3876 unhash_delegation(dp);
3878 nfs4_unlock_state();
3884 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3886 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3889 end_offset(u64 start, u64 len)
3894 return end >= start ? end: NFS4_MAX_UINT64;
3897 /* last octet in a range */
3899 last_byte_offset(u64 start, u64 len)
3905 return end > start ? end - 1: NFS4_MAX_UINT64;
3908 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3910 return (file_hashval(inode) + cl_id
3911 + opaque_hashval(ownername->data, ownername->len))
3912 & LOCKOWNER_INO_HASH_MASK;
3916 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3917 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3918 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3919 * locking, this prevents us from being completely protocol-compliant. The
3920 * real solution to this problem is to start using unsigned file offsets in
3921 * the VFS, but this is a very deep change!
3924 nfs4_transform_lock_offset(struct file_lock *lock)
3926 if (lock->fl_start < 0)
3927 lock->fl_start = OFFSET_MAX;
3928 if (lock->fl_end < 0)
3929 lock->fl_end = OFFSET_MAX;
3932 /* Hack!: For now, we're defining this just so we can use a pointer to it
3933 * as a unique cookie to identify our (NFSv4's) posix locks. */
3934 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3938 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3940 struct nfs4_lockowner *lo;
3942 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3943 lo = (struct nfs4_lockowner *) fl->fl_owner;
3944 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3945 lo->lo_owner.so_owner.len, GFP_KERNEL);
3946 if (!deny->ld_owner.data)
3947 /* We just don't care that much */
3949 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3950 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3953 deny->ld_owner.len = 0;
3954 deny->ld_owner.data = NULL;
3955 deny->ld_clientid.cl_boot = 0;
3956 deny->ld_clientid.cl_id = 0;
3958 deny->ld_start = fl->fl_start;
3959 deny->ld_length = NFS4_MAX_UINT64;
3960 if (fl->fl_end != NFS4_MAX_UINT64)
3961 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3962 deny->ld_type = NFS4_READ_LT;
3963 if (fl->fl_type != F_RDLCK)
3964 deny->ld_type = NFS4_WRITE_LT;
3967 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3969 struct nfs4_ol_stateid *lst;
3971 if (!same_owner_str(&lo->lo_owner, owner, clid))
3973 lst = list_first_entry(&lo->lo_owner.so_stateids,
3974 struct nfs4_ol_stateid, st_perstateowner);
3975 return lst->st_file->fi_inode == inode;
3978 static struct nfs4_lockowner *
3979 find_lockowner_str(struct inode *inode, clientid_t *clid,
3980 struct xdr_netobj *owner, struct nfsd_net *nn)
3982 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3983 struct nfs4_lockowner *lo;
3985 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3986 if (same_lockowner_ino(lo, inode, clid, owner))
3992 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3994 struct inode *inode = open_stp->st_file->fi_inode;
3995 unsigned int inohash = lockowner_ino_hashval(inode,
3996 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3997 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
3999 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4000 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
4001 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4005 * Alloc a lock owner structure.
4006 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4009 * strhashval = ownerstr_hashval
4012 static struct nfs4_lockowner *
4013 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4014 struct nfs4_lockowner *lo;
4016 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4019 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4020 lo->lo_owner.so_is_open_owner = 0;
4021 /* It is the openowner seqid that will be incremented in encode in the
4022 * case of new lockowners; so increment the lock seqid manually: */
4023 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4024 hash_lockowner(lo, strhashval, clp, open_stp);
4028 static struct nfs4_ol_stateid *
4029 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4031 struct nfs4_ol_stateid *stp;
4032 struct nfs4_client *clp = lo->lo_owner.so_client;
4034 stp = nfs4_alloc_stateid(clp);
4037 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
4038 list_add(&stp->st_perfile, &fp->fi_stateids);
4039 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4040 stp->st_stateowner = &lo->lo_owner;
4043 stp->st_access_bmap = 0;
4044 stp->st_deny_bmap = open_stp->st_deny_bmap;
4045 stp->st_openstp = open_stp;
4050 check_lock_length(u64 offset, u64 length)
4052 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4053 LOFF_OVERFLOW(offset, length)));
4056 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4058 struct nfs4_file *fp = lock_stp->st_file;
4059 int oflag = nfs4_access_to_omode(access);
4061 if (test_access(access, lock_stp))
4063 nfs4_file_get_access(fp, oflag);
4064 set_access(access, lock_stp);
4067 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)
4069 struct nfs4_file *fi = ost->st_file;
4070 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4071 struct nfs4_client *cl = oo->oo_owner.so_client;
4072 struct nfs4_lockowner *lo;
4073 unsigned int strhashval;
4074 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4076 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4077 &lock->v.new.owner, nn);
4079 if (!cstate->minorversion)
4080 return nfserr_bad_seqid;
4081 /* XXX: a lockowner always has exactly one stateid: */
4082 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4083 struct nfs4_ol_stateid, st_perstateowner);
4086 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4087 &lock->v.new.owner);
4088 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4090 return nfserr_jukebox;
4091 *lst = alloc_init_lock_stateid(lo, fi, ost);
4093 release_lockowner(lo);
4094 return nfserr_jukebox;
4104 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4105 struct nfsd4_lock *lock)
4107 struct nfs4_openowner *open_sop = NULL;
4108 struct nfs4_lockowner *lock_sop = NULL;
4109 struct nfs4_ol_stateid *lock_stp;
4110 struct file *filp = NULL;
4111 struct file_lock *file_lock = NULL;
4112 struct file_lock *conflock = NULL;
4114 bool new_state = false;
4117 struct net *net = SVC_NET(rqstp);
4118 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4120 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4121 (long long) lock->lk_offset,
4122 (long long) lock->lk_length);
4124 if (check_lock_length(lock->lk_offset, lock->lk_length))
4125 return nfserr_inval;
4127 if ((status = fh_verify(rqstp, &cstate->current_fh,
4128 S_IFREG, NFSD_MAY_LOCK))) {
4129 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4135 if (lock->lk_is_new) {
4136 struct nfs4_ol_stateid *open_stp = NULL;
4138 if (nfsd4_has_session(cstate))
4139 /* See rfc 5661 18.10.3: given clientid is ignored: */
4140 memcpy(&lock->v.new.clientid,
4141 &cstate->session->se_client->cl_clientid,
4142 sizeof(clientid_t));
4144 status = nfserr_stale_clientid;
4145 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4148 /* validate and update open stateid and open seqid */
4149 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4150 lock->lk_new_open_seqid,
4151 &lock->lk_new_open_stateid,
4155 open_sop = openowner(open_stp->st_stateowner);
4156 status = nfserr_bad_stateid;
4157 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4158 &lock->v.new.clientid))
4160 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4161 &lock_stp, &new_state);
4163 status = nfs4_preprocess_seqid_op(cstate,
4164 lock->lk_old_lock_seqid,
4165 &lock->lk_old_lock_stateid,
4166 NFS4_LOCK_STID, &lock_stp, nn);
4169 lock_sop = lockowner(lock_stp->st_stateowner);
4171 lkflg = setlkflg(lock->lk_type);
4172 status = nfs4_check_openmode(lock_stp, lkflg);
4176 status = nfserr_grace;
4177 if (locks_in_grace(net) && !lock->lk_reclaim)
4179 status = nfserr_no_grace;
4180 if (!locks_in_grace(net) && lock->lk_reclaim)
4183 file_lock = locks_alloc_lock();
4185 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4186 status = nfserr_jukebox;
4190 locks_init_lock(file_lock);
4191 switch (lock->lk_type) {
4194 filp = find_readable_file(lock_stp->st_file);
4196 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4197 file_lock->fl_type = F_RDLCK;
4200 case NFS4_WRITEW_LT:
4201 filp = find_writeable_file(lock_stp->st_file);
4203 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4204 file_lock->fl_type = F_WRLCK;
4207 status = nfserr_inval;
4211 status = nfserr_openmode;
4214 file_lock->fl_owner = (fl_owner_t)lock_sop;
4215 file_lock->fl_pid = current->tgid;
4216 file_lock->fl_file = filp;
4217 file_lock->fl_flags = FL_POSIX;
4218 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4219 file_lock->fl_start = lock->lk_offset;
4220 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4221 nfs4_transform_lock_offset(file_lock);
4223 conflock = locks_alloc_lock();
4225 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4226 status = nfserr_jukebox;
4230 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4232 case 0: /* success! */
4233 update_stateid(&lock_stp->st_stid.sc_stateid);
4234 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4238 case (EAGAIN): /* conflock holds conflicting lock */
4239 status = nfserr_denied;
4240 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4241 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4244 status = nfserr_deadlock;
4247 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4248 status = nfserrno(err);
4252 if (status && new_state)
4253 release_lockowner(lock_sop);
4254 if (!cstate->replay_owner)
4255 nfs4_unlock_state();
4257 locks_free_lock(file_lock);
4259 locks_free_lock(conflock);
4264 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4265 * so we do a temporary open here just to get an open file to pass to
4266 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4269 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4272 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4274 err = nfserrno(vfs_test_lock(file, lock));
4284 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4285 struct nfsd4_lockt *lockt)
4287 struct inode *inode;
4288 struct file_lock *file_lock = NULL;
4289 struct nfs4_lockowner *lo;
4291 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4293 if (locks_in_grace(SVC_NET(rqstp)))
4294 return nfserr_grace;
4296 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4297 return nfserr_inval;
4301 if (!nfsd4_has_session(cstate)) {
4302 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4307 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4310 inode = cstate->current_fh.fh_dentry->d_inode;
4311 file_lock = locks_alloc_lock();
4313 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4314 status = nfserr_jukebox;
4317 locks_init_lock(file_lock);
4318 switch (lockt->lt_type) {
4321 file_lock->fl_type = F_RDLCK;
4324 case NFS4_WRITEW_LT:
4325 file_lock->fl_type = F_WRLCK;
4328 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4329 status = nfserr_inval;
4333 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4335 file_lock->fl_owner = (fl_owner_t)lo;
4336 file_lock->fl_pid = current->tgid;
4337 file_lock->fl_flags = FL_POSIX;
4339 file_lock->fl_start = lockt->lt_offset;
4340 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4342 nfs4_transform_lock_offset(file_lock);
4344 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4348 if (file_lock->fl_type != F_UNLCK) {
4349 status = nfserr_denied;
4350 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4353 nfs4_unlock_state();
4355 locks_free_lock(file_lock);
4360 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4361 struct nfsd4_locku *locku)
4363 struct nfs4_ol_stateid *stp;
4364 struct file *filp = NULL;
4365 struct file_lock *file_lock = NULL;
4368 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4370 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4371 (long long) locku->lu_offset,
4372 (long long) locku->lu_length);
4374 if (check_lock_length(locku->lu_offset, locku->lu_length))
4375 return nfserr_inval;
4379 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4380 &locku->lu_stateid, NFS4_LOCK_STID,
4384 filp = find_any_file(stp->st_file);
4386 status = nfserr_lock_range;
4389 file_lock = locks_alloc_lock();
4391 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4392 status = nfserr_jukebox;
4395 locks_init_lock(file_lock);
4396 file_lock->fl_type = F_UNLCK;
4397 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4398 file_lock->fl_pid = current->tgid;
4399 file_lock->fl_file = filp;
4400 file_lock->fl_flags = FL_POSIX;
4401 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4402 file_lock->fl_start = locku->lu_offset;
4404 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4406 nfs4_transform_lock_offset(file_lock);
4409 * Try to unlock the file in the VFS.
4411 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4413 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4417 * OK, unlock succeeded; the only thing left to do is update the stateid.
4419 update_stateid(&stp->st_stid.sc_stateid);
4420 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4423 if (!cstate->replay_owner)
4424 nfs4_unlock_state();
4426 locks_free_lock(file_lock);
4430 status = nfserrno(err);
4436 * 1: locks held by lockowner
4437 * 0: no locks held by lockowner
4440 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4442 struct file_lock **flpp;
4443 struct inode *inode = filp->fi_inode;
4447 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4448 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4459 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4460 struct nfsd4_compound_state *cstate,
4461 struct nfsd4_release_lockowner *rlockowner)
4463 clientid_t *clid = &rlockowner->rl_clientid;
4464 struct nfs4_stateowner *sop;
4465 struct nfs4_lockowner *lo;
4466 struct nfs4_ol_stateid *stp;
4467 struct xdr_netobj *owner = &rlockowner->rl_owner;
4468 struct list_head matches;
4469 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4471 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4473 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4474 clid->cl_boot, clid->cl_id);
4478 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4482 status = nfserr_locks_held;
4483 INIT_LIST_HEAD(&matches);
4485 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4486 if (sop->so_is_open_owner)
4488 if (!same_owner_str(sop, owner, clid))
4490 list_for_each_entry(stp, &sop->so_stateids,
4492 lo = lockowner(sop);
4493 if (check_for_locks(stp->st_file, lo))
4495 list_add(&lo->lo_list, &matches);
4498 /* Clients probably won't expect us to return with some (but not all)
4499 * of the lockowner state released; so don't release any until all
4500 * have been checked. */
4502 while (!list_empty(&matches)) {
4503 lo = list_entry(matches.next, struct nfs4_lockowner,
4505 /* unhash_stateowner deletes so_perclient only
4506 * for openowners. */
4507 list_del(&lo->lo_list);
4508 release_lockowner(lo);
4511 nfs4_unlock_state();
4515 static inline struct nfs4_client_reclaim *
4518 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4522 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4524 struct nfs4_client_reclaim *crp;
4526 crp = nfsd4_find_reclaim_client(name, nn);
4527 return (crp && crp->cr_clp);
4531 * failure => all reset bets are off, nfserr_no_grace...
4533 struct nfs4_client_reclaim *
4534 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4536 unsigned int strhashval;
4537 struct nfs4_client_reclaim *crp;
4539 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4540 crp = alloc_reclaim();
4542 strhashval = clientstr_hashval(name);
4543 INIT_LIST_HEAD(&crp->cr_strhash);
4544 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4545 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4547 nn->reclaim_str_hashtbl_size++;
4553 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4555 list_del(&crp->cr_strhash);
4557 nn->reclaim_str_hashtbl_size--;
4561 nfs4_release_reclaim(struct nfsd_net *nn)
4563 struct nfs4_client_reclaim *crp = NULL;
4566 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4567 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4568 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4569 struct nfs4_client_reclaim, cr_strhash);
4570 nfs4_remove_reclaim_record(crp, nn);
4573 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4577 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4578 struct nfs4_client_reclaim *
4579 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4581 unsigned int strhashval;
4582 struct nfs4_client_reclaim *crp = NULL;
4584 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4586 strhashval = clientstr_hashval(recdir);
4587 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4588 if (same_name(crp->cr_recdir, recdir)) {
4596 * Called from OPEN. Look for clientid in reclaim list.
4599 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4601 struct nfs4_client *clp;
4603 /* find clientid in conf_id_hashtbl */
4604 clp = find_confirmed_client(clid, sessions, nn);
4606 return nfserr_reclaim_bad;
4608 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4611 #ifdef CONFIG_NFSD_FAULT_INJECTION
4613 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4619 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4621 char buf[INET6_ADDRSTRLEN];
4622 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, 129);
4623 printk(KERN_INFO "NFS Client: %s\n", buf);
4627 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4630 char buf[INET6_ADDRSTRLEN];
4631 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, 129);
4632 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4635 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_lockowner *))
4637 struct nfs4_openowner *oop;
4638 struct nfs4_lockowner *lop, *lo_next;
4639 struct nfs4_ol_stateid *stp, *st_next;
4642 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4643 list_for_each_entry_safe(stp, st_next, &oop->oo_owner.so_stateids, st_perstateowner) {
4644 list_for_each_entry_safe(lop, lo_next, &stp->st_lockowners, lo_perstateid) {
4656 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
4658 return nfsd_foreach_client_lock(clp, max, release_lockowner);
4661 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
4663 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
4664 nfsd_print_count(clp, count, "locked files");
4668 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
4670 struct nfs4_openowner *oop, *next;
4673 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
4683 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
4685 return nfsd_foreach_client_open(clp, max, release_openowner);
4688 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
4690 u64 count = nfsd_foreach_client_open(clp, max, NULL);
4691 nfsd_print_count(clp, count, "open files");
4695 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
4696 struct list_head *victims)
4698 struct nfs4_delegation *dp, *next;
4701 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
4703 list_move(&dp->dl_recall_lru, victims);
4710 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
4712 struct nfs4_delegation *dp, *next;
4716 spin_lock(&recall_lock);
4717 count = nfsd_find_all_delegations(clp, max, &victims);
4718 spin_unlock(&recall_lock);
4720 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4721 unhash_delegation(dp);
4726 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
4728 struct nfs4_delegation *dp, *next;
4732 spin_lock(&recall_lock);
4733 count = nfsd_find_all_delegations(clp, max, &victims);
4734 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4735 nfsd_break_one_deleg(dp);
4736 spin_unlock(&recall_lock);
4741 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
4745 spin_lock(&recall_lock);
4746 count = nfsd_find_all_delegations(clp, max, NULL);
4747 spin_unlock(&recall_lock);
4749 nfsd_print_count(clp, count, "delegations");
4753 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
4755 struct nfs4_client *clp, *next;
4757 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4759 if (!nfsd_netns_ready(nn))
4762 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4763 count += func(clp, max - count);
4764 if ((max != 0) && (count >= max))
4771 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
4773 struct nfs4_client *clp;
4774 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4776 if (!nfsd_netns_ready(nn))
4779 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
4780 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
4786 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4788 /* initialization to perform at module load time: */
4791 nfs4_state_init(void)
4795 for (i = 0; i < FILE_HASH_SIZE; i++) {
4796 INIT_LIST_HEAD(&file_hashtbl[i]);
4798 INIT_LIST_HEAD(&del_recall_lru);
4802 * Since the lifetime of a delegation isn't limited to that of an open, a
4803 * client may quite reasonably hang on to a delegation as long as it has
4804 * the inode cached. This becomes an obvious problem the first time a
4805 * client's inode cache approaches the size of the server's total memory.
4807 * For now we avoid this problem by imposing a hard limit on the number
4808 * of delegations, which varies according to the server's memory size.
4811 set_max_delegations(void)
4814 * Allow at most 4 delegations per megabyte of RAM. Quick
4815 * estimates suggest that in the worst case (where every delegation
4816 * is for a different inode), a delegation could take about 1.5K,
4817 * giving a worst case usage of about 6% of memory.
4819 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4822 static int nfs4_state_create_net(struct net *net)
4824 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4827 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4828 CLIENT_HASH_SIZE, GFP_KERNEL);
4829 if (!nn->conf_id_hashtbl)
4831 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4832 CLIENT_HASH_SIZE, GFP_KERNEL);
4833 if (!nn->unconf_id_hashtbl)
4835 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
4836 OWNER_HASH_SIZE, GFP_KERNEL);
4837 if (!nn->ownerstr_hashtbl)
4839 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
4840 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
4841 if (!nn->lockowner_ino_hashtbl)
4842 goto err_lockowner_ino;
4843 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
4844 SESSION_HASH_SIZE, GFP_KERNEL);
4845 if (!nn->sessionid_hashtbl)
4848 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4849 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
4850 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
4852 for (i = 0; i < OWNER_HASH_SIZE; i++)
4853 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
4854 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4855 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
4856 for (i = 0; i < SESSION_HASH_SIZE; i++)
4857 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
4858 nn->conf_name_tree = RB_ROOT;
4859 nn->unconf_name_tree = RB_ROOT;
4860 INIT_LIST_HEAD(&nn->client_lru);
4861 INIT_LIST_HEAD(&nn->close_lru);
4862 spin_lock_init(&nn->client_lock);
4864 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
4870 kfree(nn->lockowner_ino_hashtbl);
4872 kfree(nn->ownerstr_hashtbl);
4874 kfree(nn->unconf_id_hashtbl);
4876 kfree(nn->conf_id_hashtbl);
4882 nfs4_state_destroy_net(struct net *net)
4885 struct nfs4_client *clp = NULL;
4886 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4887 struct rb_node *node, *tmp;
4889 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4890 while (!list_empty(&nn->conf_id_hashtbl[i])) {
4891 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4892 destroy_client(clp);
4896 node = rb_first(&nn->unconf_name_tree);
4897 while (node != NULL) {
4899 node = rb_next(tmp);
4900 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4901 rb_erase(tmp, &nn->unconf_name_tree);
4902 destroy_client(clp);
4905 kfree(nn->sessionid_hashtbl);
4906 kfree(nn->lockowner_ino_hashtbl);
4907 kfree(nn->ownerstr_hashtbl);
4908 kfree(nn->unconf_id_hashtbl);
4909 kfree(nn->conf_id_hashtbl);
4914 nfs4_state_start_net(struct net *net)
4916 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4920 * FIXME: For now, we hang most of the pernet global stuff off of
4921 * init_net until nfsd is fully containerized. Eventually, we'll
4922 * need to pass a net pointer into this function, take a reference
4923 * to that instead and then do most of the rest of this on a per-net
4926 if (net != &init_net)
4929 ret = nfs4_state_create_net(net);
4932 nfsd4_client_tracking_init(net);
4933 nn->boot_time = get_seconds();
4934 locks_start_grace(net, &nn->nfsd4_manager);
4935 nn->grace_ended = false;
4936 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
4937 nn->nfsd4_grace, net);
4938 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
4942 /* initialization to perform when the nfsd service is started: */
4945 nfs4_state_start(void)
4949 ret = set_callback_cred();
4952 laundry_wq = create_singlethread_workqueue("nfsd4");
4953 if (laundry_wq == NULL) {
4957 ret = nfsd4_create_callback_queue();
4959 goto out_free_laundry;
4961 set_max_delegations();
4966 destroy_workqueue(laundry_wq);
4971 /* should be called with the state lock held */
4973 nfs4_state_shutdown_net(struct net *net)
4975 struct nfs4_delegation *dp = NULL;
4976 struct list_head *pos, *next, reaplist;
4977 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4979 cancel_delayed_work_sync(&nn->laundromat_work);
4980 locks_end_grace(&nn->nfsd4_manager);
4982 INIT_LIST_HEAD(&reaplist);
4983 spin_lock(&recall_lock);
4984 list_for_each_safe(pos, next, &del_recall_lru) {
4985 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4986 if (dp->dl_stid.sc_client->net != net)
4988 list_move(&dp->dl_recall_lru, &reaplist);
4990 spin_unlock(&recall_lock);
4991 list_for_each_safe(pos, next, &reaplist) {
4992 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4993 unhash_delegation(dp);
4996 nfsd4_client_tracking_exit(net);
4997 nfs4_state_destroy_net(net);
5001 nfs4_state_shutdown(void)
5003 destroy_workqueue(laundry_wq);
5004 nfsd4_destroy_callback_queue();
5008 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5010 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5011 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5015 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5017 if (cstate->minorversion) {
5018 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5019 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5024 clear_current_stateid(struct nfsd4_compound_state *cstate)
5026 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5030 * functions to set current state id
5033 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5035 put_stateid(cstate, &odp->od_stateid);
5039 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5041 put_stateid(cstate, &open->op_stateid);
5045 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5047 put_stateid(cstate, &close->cl_stateid);
5051 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5053 put_stateid(cstate, &lock->lk_resp_stateid);
5057 * functions to consume current state id
5061 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5063 get_stateid(cstate, &odp->od_stateid);
5067 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5069 get_stateid(cstate, &drp->dr_stateid);
5073 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5075 get_stateid(cstate, &fsp->fr_stateid);
5079 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5081 get_stateid(cstate, &setattr->sa_stateid);
5085 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5087 get_stateid(cstate, &close->cl_stateid);
5091 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5093 get_stateid(cstate, &locku->lu_stateid);
5097 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5099 get_stateid(cstate, &read->rd_stateid);
5103 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5105 get_stateid(cstate, &write->wr_stateid);