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/sunrpc/svcauth_gss.h>
42 #include <linux/sunrpc/clnt.h>
46 #define NFSDDBG_FACILITY NFSDDBG_PROC
49 time_t nfsd4_lease = 90; /* default lease time */
50 time_t nfsd4_grace = 90;
51 static time_t boot_time;
52 static stateid_t zerostateid; /* bits all 0 */
53 static stateid_t onestateid; /* bits all 1 */
54 static u64 current_sessionid = 1;
56 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
57 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
59 /* forward declarations */
60 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
64 /* Currently used for almost all code touching nfsv4 state: */
65 static DEFINE_MUTEX(client_mutex);
68 * Currently used for the del_recall_lru and file hash table. In an
69 * effort to decrease the scope of the client_mutex, this spinlock may
70 * eventually cover more:
72 static DEFINE_SPINLOCK(recall_lock);
74 static struct kmem_cache *openowner_slab = NULL;
75 static struct kmem_cache *lockowner_slab = NULL;
76 static struct kmem_cache *file_slab = NULL;
77 static struct kmem_cache *stateid_slab = NULL;
78 static struct kmem_cache *deleg_slab = NULL;
83 mutex_lock(&client_mutex);
87 nfs4_unlock_state(void)
89 mutex_unlock(&client_mutex);
93 opaque_hashval(const void *ptr, int nbytes)
95 unsigned char *cptr = (unsigned char *) ptr;
105 static struct list_head del_recall_lru;
107 static void nfsd4_free_file(struct nfs4_file *f)
109 kmem_cache_free(file_slab, f);
113 put_nfs4_file(struct nfs4_file *fi)
115 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
116 list_del(&fi->fi_hash);
117 spin_unlock(&recall_lock);
124 get_nfs4_file(struct nfs4_file *fi)
126 atomic_inc(&fi->fi_ref);
129 static int num_delegations;
130 unsigned int max_delegations;
133 * Open owner state (share locks)
136 /* hash tables for open owners */
137 #define OPEN_OWNER_HASH_BITS 8
138 #define OPEN_OWNER_HASH_SIZE (1 << OPEN_OWNER_HASH_BITS)
139 #define OPEN_OWNER_HASH_MASK (OPEN_OWNER_HASH_SIZE - 1)
141 static unsigned int open_ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
145 ret = opaque_hashval(ownername->data, ownername->len);
147 return ret & OPEN_OWNER_HASH_MASK;
150 static struct list_head open_ownerstr_hashtbl[OPEN_OWNER_HASH_SIZE];
152 /* hash table for nfs4_file */
153 #define FILE_HASH_BITS 8
154 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
156 static unsigned int file_hashval(struct inode *ino)
158 /* XXX: why are we hashing on inode pointer, anyway? */
159 return hash_ptr(ino, FILE_HASH_BITS);
162 static struct list_head file_hashtbl[FILE_HASH_SIZE];
164 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
166 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
167 atomic_inc(&fp->fi_access[oflag]);
170 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
172 if (oflag == O_RDWR) {
173 __nfs4_file_get_access(fp, O_RDONLY);
174 __nfs4_file_get_access(fp, O_WRONLY);
176 __nfs4_file_get_access(fp, oflag);
179 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
181 if (fp->fi_fds[oflag]) {
182 fput(fp->fi_fds[oflag]);
183 fp->fi_fds[oflag] = NULL;
187 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
189 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
190 nfs4_file_put_fd(fp, oflag);
191 if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
192 nfs4_file_put_fd(fp, O_RDWR);
196 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
198 if (oflag == O_RDWR) {
199 __nfs4_file_put_access(fp, O_RDONLY);
200 __nfs4_file_put_access(fp, O_WRONLY);
202 __nfs4_file_put_access(fp, oflag);
205 static inline int get_new_stid(struct nfs4_stid *stid)
207 static int min_stateid = 0;
208 struct idr *stateids = &stid->sc_client->cl_stateids;
212 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
214 * Note: the necessary preallocation was done in
215 * nfs4_alloc_stateid(). The idr code caps the number of
216 * preallocations that can exist at a time, but the state lock
217 * prevents anyone from using ours before we get here:
221 * It shouldn't be a problem to reuse an opaque stateid value.
222 * I don't think it is for 4.1. But with 4.0 I worry that, for
223 * example, a stray write retransmission could be accepted by
224 * the server when it should have been rejected. Therefore,
225 * adopt a trick from the sctp code to attempt to maximize the
226 * amount of time until an id is reused, by ensuring they always
227 * "increase" (mod INT_MAX):
230 min_stateid = new_stid+1;
231 if (min_stateid == INT_MAX)
236 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
238 stateid_t *s = &stid->sc_stateid;
241 stid->sc_type = type;
242 stid->sc_client = cl;
243 s->si_opaque.so_clid = cl->cl_clientid;
244 new_id = get_new_stid(stid);
245 s->si_opaque.so_id = (u32)new_id;
246 /* Will be incremented before return to client: */
247 s->si_generation = 0;
250 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
252 struct idr *stateids = &cl->cl_stateids;
254 if (!idr_pre_get(stateids, GFP_KERNEL))
257 * Note: if we fail here (or any time between now and the time
258 * we actually get the new idr), we won't need to undo the idr
259 * preallocation, since the idr code caps the number of
260 * preallocated entries.
262 return kmem_cache_alloc(slab, GFP_KERNEL);
265 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
267 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
270 static struct nfs4_delegation *
271 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
273 struct nfs4_delegation *dp;
274 struct nfs4_file *fp = stp->st_file;
276 dprintk("NFSD alloc_init_deleg\n");
278 * Major work on the lease subsystem (for example, to support
279 * calbacks on stat) will be required before we can support
280 * write delegations properly.
282 if (type != NFS4_OPEN_DELEGATE_READ)
284 if (fp->fi_had_conflict)
286 if (num_delegations > max_delegations)
288 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
291 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
293 * delegation seqid's are never incremented. The 4.1 special
294 * meaning of seqid 0 isn't meaningful, really, but let's avoid
295 * 0 anyway just for consistency and use 1:
297 dp->dl_stid.sc_stateid.si_generation = 1;
299 INIT_LIST_HEAD(&dp->dl_perfile);
300 INIT_LIST_HEAD(&dp->dl_perclnt);
301 INIT_LIST_HEAD(&dp->dl_recall_lru);
305 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
307 atomic_set(&dp->dl_count, 1);
308 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
313 nfs4_put_delegation(struct nfs4_delegation *dp)
315 if (atomic_dec_and_test(&dp->dl_count)) {
316 dprintk("NFSD: freeing dp %p\n",dp);
317 put_nfs4_file(dp->dl_file);
318 kmem_cache_free(deleg_slab, dp);
323 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
325 if (atomic_dec_and_test(&fp->fi_delegees)) {
326 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
328 fput(fp->fi_deleg_file);
329 fp->fi_deleg_file = NULL;
333 static void unhash_stid(struct nfs4_stid *s)
335 struct idr *stateids = &s->sc_client->cl_stateids;
337 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
340 /* Called under the state lock. */
342 unhash_delegation(struct nfs4_delegation *dp)
344 unhash_stid(&dp->dl_stid);
345 list_del_init(&dp->dl_perclnt);
346 spin_lock(&recall_lock);
347 list_del_init(&dp->dl_perfile);
348 list_del_init(&dp->dl_recall_lru);
349 spin_unlock(&recall_lock);
350 nfs4_put_deleg_lease(dp->dl_file);
351 nfs4_put_delegation(dp);
358 /* client_lock protects the client lru list and session hash table */
359 static DEFINE_SPINLOCK(client_lock);
361 /* Hash tables for nfs4_clientid state */
362 #define CLIENT_HASH_BITS 4
363 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
364 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
366 static unsigned int clientid_hashval(u32 id)
368 return id & CLIENT_HASH_MASK;
371 static unsigned int clientstr_hashval(const char *name)
373 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
377 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
378 * used in reboot/reset lease grace period processing
380 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
381 * setclientid_confirmed info.
383 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
386 * client_lru holds client queue ordered by nfs4_client.cl_time
389 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
390 * for last close replay.
392 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
393 static int reclaim_str_hashtbl_size = 0;
394 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
395 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
396 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
397 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
398 static struct list_head client_lru;
399 static struct list_head close_lru;
402 * We store the NONE, READ, WRITE, and BOTH bits separately in the
403 * st_{access,deny}_bmap field of the stateid, in order to track not
404 * only what share bits are currently in force, but also what
405 * combinations of share bits previous opens have used. This allows us
406 * to enforce the recommendation of rfc 3530 14.2.19 that the server
407 * return an error if the client attempt to downgrade to a combination
408 * of share bits not explicable by closing some of its previous opens.
410 * XXX: This enforcement is actually incomplete, since we don't keep
411 * track of access/deny bit combinations; so, e.g., we allow:
413 * OPEN allow read, deny write
414 * OPEN allow both, deny none
415 * DOWNGRADE allow read, deny none
417 * which we should reject.
420 set_access(unsigned int *access, unsigned long bmap) {
424 for (i = 1; i < 4; i++) {
425 if (test_bit(i, &bmap))
431 set_deny(unsigned int *deny, unsigned long bmap) {
435 for (i = 0; i < 4; i++) {
436 if (test_bit(i, &bmap))
442 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
443 unsigned int access, deny;
445 set_access(&access, stp->st_access_bmap);
446 set_deny(&deny, stp->st_deny_bmap);
447 if ((access & open->op_share_deny) || (deny & open->op_share_access))
452 static int nfs4_access_to_omode(u32 access)
454 switch (access & NFS4_SHARE_ACCESS_BOTH) {
455 case NFS4_SHARE_ACCESS_READ:
457 case NFS4_SHARE_ACCESS_WRITE:
459 case NFS4_SHARE_ACCESS_BOTH:
465 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
467 list_del(&stp->st_perfile);
468 list_del(&stp->st_perstateowner);
471 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
475 if (stp->st_access_bmap) {
476 for (i = 1; i < 4; i++) {
477 if (test_bit(i, &stp->st_access_bmap))
478 nfs4_file_put_access(stp->st_file,
479 nfs4_access_to_omode(i));
480 __clear_bit(i, &stp->st_access_bmap);
483 put_nfs4_file(stp->st_file);
487 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
489 kmem_cache_free(stateid_slab, stp);
492 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
496 unhash_generic_stateid(stp);
497 unhash_stid(&stp->st_stid);
498 file = find_any_file(stp->st_file);
500 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
501 close_generic_stateid(stp);
502 free_generic_stateid(stp);
505 static void unhash_lockowner(struct nfs4_lockowner *lo)
507 struct nfs4_ol_stateid *stp;
509 list_del(&lo->lo_owner.so_strhash);
510 list_del(&lo->lo_perstateid);
511 while (!list_empty(&lo->lo_owner.so_stateids)) {
512 stp = list_first_entry(&lo->lo_owner.so_stateids,
513 struct nfs4_ol_stateid, st_perstateowner);
514 release_lock_stateid(stp);
518 static void release_lockowner(struct nfs4_lockowner *lo)
520 unhash_lockowner(lo);
521 nfs4_free_lockowner(lo);
525 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
527 struct nfs4_lockowner *lo;
529 while (!list_empty(&open_stp->st_lockowners)) {
530 lo = list_entry(open_stp->st_lockowners.next,
531 struct nfs4_lockowner, lo_perstateid);
532 release_lockowner(lo);
536 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
538 unhash_generic_stateid(stp);
539 release_stateid_lockowners(stp);
540 close_generic_stateid(stp);
543 static void release_open_stateid(struct nfs4_ol_stateid *stp)
545 unhash_open_stateid(stp);
546 unhash_stid(&stp->st_stid);
547 free_generic_stateid(stp);
550 static void unhash_openowner(struct nfs4_openowner *oo)
552 struct nfs4_ol_stateid *stp;
554 list_del(&oo->oo_owner.so_strhash);
555 list_del(&oo->oo_perclient);
556 while (!list_empty(&oo->oo_owner.so_stateids)) {
557 stp = list_first_entry(&oo->oo_owner.so_stateids,
558 struct nfs4_ol_stateid, st_perstateowner);
559 release_open_stateid(stp);
563 static void release_last_closed_stateid(struct nfs4_openowner *oo)
565 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
568 unhash_stid(&s->st_stid);
569 free_generic_stateid(s);
570 oo->oo_last_closed_stid = NULL;
574 static void release_openowner(struct nfs4_openowner *oo)
576 unhash_openowner(oo);
577 list_del(&oo->oo_close_lru);
578 release_last_closed_stateid(oo);
579 nfs4_free_openowner(oo);
582 #define SESSION_HASH_SIZE 512
583 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
586 hash_sessionid(struct nfs4_sessionid *sessionid)
588 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
590 return sid->sequence % SESSION_HASH_SIZE;
594 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
596 u32 *ptr = (u32 *)(&sessionid->data[0]);
597 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
601 gen_sessionid(struct nfsd4_session *ses)
603 struct nfs4_client *clp = ses->se_client;
604 struct nfsd4_sessionid *sid;
606 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
607 sid->clientid = clp->cl_clientid;
608 sid->sequence = current_sessionid++;
613 * The protocol defines ca_maxresponssize_cached to include the size of
614 * the rpc header, but all we need to cache is the data starting after
615 * the end of the initial SEQUENCE operation--the rest we regenerate
616 * each time. Therefore we can advertise a ca_maxresponssize_cached
617 * value that is the number of bytes in our cache plus a few additional
618 * bytes. In order to stay on the safe side, and not promise more than
619 * we can cache, those additional bytes must be the minimum possible: 24
620 * bytes of rpc header (xid through accept state, with AUTH_NULL
621 * verifier), 12 for the compound header (with zero-length tag), and 44
622 * for the SEQUENCE op response:
624 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
627 free_session_slots(struct nfsd4_session *ses)
631 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
632 kfree(ses->se_slots[i]);
636 * We don't actually need to cache the rpc and session headers, so we
637 * can allocate a little less for each slot:
639 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
641 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
644 static int nfsd4_sanitize_slot_size(u32 size)
646 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
647 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
653 * XXX: If we run out of reserved DRC memory we could (up to a point)
654 * re-negotiate active sessions and reduce their slot usage to make
655 * rooom for new connections. For now we just fail the create session.
657 static int nfsd4_get_drc_mem(int slotsize, u32 num)
661 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
663 spin_lock(&nfsd_drc_lock);
664 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
665 nfsd_drc_max_mem - nfsd_drc_mem_used);
666 num = min_t(int, num, avail / slotsize);
667 nfsd_drc_mem_used += num * slotsize;
668 spin_unlock(&nfsd_drc_lock);
673 static void nfsd4_put_drc_mem(int slotsize, int num)
675 spin_lock(&nfsd_drc_lock);
676 nfsd_drc_mem_used -= slotsize * num;
677 spin_unlock(&nfsd_drc_lock);
680 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
682 struct nfsd4_session *new;
685 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
686 + sizeof(struct nfsd4_session) > PAGE_SIZE);
687 mem = numslots * sizeof(struct nfsd4_slot *);
689 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
692 /* allocate each struct nfsd4_slot and data cache in one piece */
693 for (i = 0; i < numslots; i++) {
694 mem = sizeof(struct nfsd4_slot) + slotsize;
695 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
696 if (!new->se_slots[i])
702 kfree(new->se_slots[i]);
707 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
709 u32 maxrpc = nfsd_serv->sv_max_mesg;
711 new->maxreqs = numslots;
712 new->maxresp_cached = min_t(u32, req->maxresp_cached,
713 slotsize + NFSD_MIN_HDR_SEQ_SZ);
714 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
715 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
716 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
719 static void free_conn(struct nfsd4_conn *c)
721 svc_xprt_put(c->cn_xprt);
725 static void nfsd4_conn_lost(struct svc_xpt_user *u)
727 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
728 struct nfs4_client *clp = c->cn_session->se_client;
730 spin_lock(&clp->cl_lock);
731 if (!list_empty(&c->cn_persession)) {
732 list_del(&c->cn_persession);
735 spin_unlock(&clp->cl_lock);
736 nfsd4_probe_callback(clp);
739 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
741 struct nfsd4_conn *conn;
743 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
746 svc_xprt_get(rqstp->rq_xprt);
747 conn->cn_xprt = rqstp->rq_xprt;
748 conn->cn_flags = flags;
749 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
753 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
755 conn->cn_session = ses;
756 list_add(&conn->cn_persession, &ses->se_conns);
759 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
761 struct nfs4_client *clp = ses->se_client;
763 spin_lock(&clp->cl_lock);
764 __nfsd4_hash_conn(conn, ses);
765 spin_unlock(&clp->cl_lock);
768 static int nfsd4_register_conn(struct nfsd4_conn *conn)
770 conn->cn_xpt_user.callback = nfsd4_conn_lost;
771 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
774 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
776 struct nfsd4_conn *conn;
779 conn = alloc_conn(rqstp, dir);
781 return nfserr_jukebox;
782 nfsd4_hash_conn(conn, ses);
783 ret = nfsd4_register_conn(conn);
785 /* oops; xprt is already down: */
786 nfsd4_conn_lost(&conn->cn_xpt_user);
790 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
792 u32 dir = NFS4_CDFC4_FORE;
794 if (ses->se_flags & SESSION4_BACK_CHAN)
795 dir |= NFS4_CDFC4_BACK;
797 return nfsd4_new_conn(rqstp, ses, dir);
800 /* must be called under client_lock */
801 static void nfsd4_del_conns(struct nfsd4_session *s)
803 struct nfs4_client *clp = s->se_client;
804 struct nfsd4_conn *c;
806 spin_lock(&clp->cl_lock);
807 while (!list_empty(&s->se_conns)) {
808 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
809 list_del_init(&c->cn_persession);
810 spin_unlock(&clp->cl_lock);
812 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
815 spin_lock(&clp->cl_lock);
817 spin_unlock(&clp->cl_lock);
820 void free_session(struct kref *kref)
822 struct nfsd4_session *ses;
825 ses = container_of(kref, struct nfsd4_session, se_ref);
826 nfsd4_del_conns(ses);
827 spin_lock(&nfsd_drc_lock);
828 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
829 nfsd_drc_mem_used -= mem;
830 spin_unlock(&nfsd_drc_lock);
831 free_session_slots(ses);
835 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
837 struct nfsd4_session *new;
838 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
839 int numslots, slotsize;
844 * Note decreasing slot size below client's request may
845 * make it difficult for client to function correctly, whereas
846 * decreasing the number of slots will (just?) affect
847 * performance. When short on memory we therefore prefer to
848 * decrease number of slots instead of their size.
850 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
851 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
855 new = alloc_session(slotsize, numslots);
857 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
860 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
862 new->se_client = clp;
865 INIT_LIST_HEAD(&new->se_conns);
867 new->se_cb_seq_nr = 1;
868 new->se_flags = cses->flags;
869 new->se_cb_prog = cses->callback_prog;
870 kref_init(&new->se_ref);
871 idx = hash_sessionid(&new->se_sessionid);
872 spin_lock(&client_lock);
873 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
874 spin_lock(&clp->cl_lock);
875 list_add(&new->se_perclnt, &clp->cl_sessions);
876 spin_unlock(&clp->cl_lock);
877 spin_unlock(&client_lock);
879 status = nfsd4_new_conn_from_crses(rqstp, new);
880 /* whoops: benny points out, status is ignored! (err, or bogus) */
882 free_session(&new->se_ref);
885 if (cses->flags & SESSION4_BACK_CHAN) {
886 struct sockaddr *sa = svc_addr(rqstp);
888 * This is a little silly; with sessions there's no real
889 * use for the callback address. Use the peer address
890 * as a reasonable default for now, but consider fixing
891 * the rpc client not to require an address in the
894 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
895 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
897 nfsd4_probe_callback(clp);
901 /* caller must hold client_lock */
902 static struct nfsd4_session *
903 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
905 struct nfsd4_session *elem;
908 dump_sessionid(__func__, sessionid);
909 idx = hash_sessionid(sessionid);
910 /* Search in the appropriate list */
911 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
912 if (!memcmp(elem->se_sessionid.data, sessionid->data,
913 NFS4_MAX_SESSIONID_LEN)) {
918 dprintk("%s: session not found\n", __func__);
922 /* caller must hold client_lock */
924 unhash_session(struct nfsd4_session *ses)
926 list_del(&ses->se_hash);
927 spin_lock(&ses->se_client->cl_lock);
928 list_del(&ses->se_perclnt);
929 spin_unlock(&ses->se_client->cl_lock);
932 /* must be called under the client_lock */
934 renew_client_locked(struct nfs4_client *clp)
936 if (is_client_expired(clp)) {
937 dprintk("%s: client (clientid %08x/%08x) already expired\n",
939 clp->cl_clientid.cl_boot,
940 clp->cl_clientid.cl_id);
944 dprintk("renewing client (clientid %08x/%08x)\n",
945 clp->cl_clientid.cl_boot,
946 clp->cl_clientid.cl_id);
947 list_move_tail(&clp->cl_lru, &client_lru);
948 clp->cl_time = get_seconds();
952 renew_client(struct nfs4_client *clp)
954 spin_lock(&client_lock);
955 renew_client_locked(clp);
956 spin_unlock(&client_lock);
959 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
961 STALE_CLIENTID(clientid_t *clid)
963 if (clid->cl_boot == boot_time)
965 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
966 clid->cl_boot, clid->cl_id, boot_time);
971 * XXX Should we use a slab cache ?
972 * This type of memory management is somewhat inefficient, but we use it
973 * anyway since SETCLIENTID is not a common operation.
975 static struct nfs4_client *alloc_client(struct xdr_netobj name)
977 struct nfs4_client *clp;
979 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
982 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
983 if (clp->cl_name.data == NULL) {
987 memcpy(clp->cl_name.data, name.data, name.len);
988 clp->cl_name.len = name.len;
989 INIT_LIST_HEAD(&clp->cl_sessions);
990 idr_init(&clp->cl_stateids);
991 atomic_set(&clp->cl_refcount, 0);
992 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
993 INIT_LIST_HEAD(&clp->cl_idhash);
994 INIT_LIST_HEAD(&clp->cl_strhash);
995 INIT_LIST_HEAD(&clp->cl_openowners);
996 INIT_LIST_HEAD(&clp->cl_delegations);
997 INIT_LIST_HEAD(&clp->cl_lru);
998 INIT_LIST_HEAD(&clp->cl_callbacks);
999 spin_lock_init(&clp->cl_lock);
1000 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1005 free_client(struct nfs4_client *clp)
1007 while (!list_empty(&clp->cl_sessions)) {
1008 struct nfsd4_session *ses;
1009 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1011 list_del(&ses->se_perclnt);
1012 nfsd4_put_session(ses);
1014 rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1015 if (clp->cl_cred.cr_group_info)
1016 put_group_info(clp->cl_cred.cr_group_info);
1017 kfree(clp->cl_principal);
1018 kfree(clp->cl_name.data);
1019 idr_remove_all(&clp->cl_stateids);
1020 idr_destroy(&clp->cl_stateids);
1025 release_session_client(struct nfsd4_session *session)
1027 struct nfs4_client *clp = session->se_client;
1029 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1031 if (is_client_expired(clp)) {
1033 session->se_client = NULL;
1035 renew_client_locked(clp);
1036 spin_unlock(&client_lock);
1039 /* must be called under the client_lock */
1041 unhash_client_locked(struct nfs4_client *clp)
1043 struct nfsd4_session *ses;
1045 mark_client_expired(clp);
1046 list_del(&clp->cl_lru);
1047 spin_lock(&clp->cl_lock);
1048 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1049 list_del_init(&ses->se_hash);
1050 spin_unlock(&clp->cl_lock);
1054 expire_client(struct nfs4_client *clp)
1056 struct nfs4_openowner *oo;
1057 struct nfs4_delegation *dp;
1058 struct list_head reaplist;
1060 INIT_LIST_HEAD(&reaplist);
1061 spin_lock(&recall_lock);
1062 while (!list_empty(&clp->cl_delegations)) {
1063 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1064 list_del_init(&dp->dl_perclnt);
1065 list_move(&dp->dl_recall_lru, &reaplist);
1067 spin_unlock(&recall_lock);
1068 while (!list_empty(&reaplist)) {
1069 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1070 list_del_init(&dp->dl_recall_lru);
1071 unhash_delegation(dp);
1073 while (!list_empty(&clp->cl_openowners)) {
1074 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1075 release_openowner(oo);
1077 nfsd4_shutdown_callback(clp);
1078 if (clp->cl_cb_conn.cb_xprt)
1079 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1080 list_del(&clp->cl_idhash);
1081 list_del(&clp->cl_strhash);
1082 spin_lock(&client_lock);
1083 unhash_client_locked(clp);
1084 if (atomic_read(&clp->cl_refcount) == 0)
1086 spin_unlock(&client_lock);
1089 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1091 memcpy(target->cl_verifier.data, source->data,
1092 sizeof(target->cl_verifier.data));
1095 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1097 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1098 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1101 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
1103 target->cr_uid = source->cr_uid;
1104 target->cr_gid = source->cr_gid;
1105 target->cr_group_info = source->cr_group_info;
1106 get_group_info(target->cr_group_info);
1109 static int same_name(const char *n1, const char *n2)
1111 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1115 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1117 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1121 same_clid(clientid_t *cl1, clientid_t *cl2)
1123 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1126 /* XXX what about NGROUP */
1128 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1130 return cr1->cr_uid == cr2->cr_uid;
1133 static void gen_clid(struct nfs4_client *clp)
1135 static u32 current_clientid = 1;
1137 clp->cl_clientid.cl_boot = boot_time;
1138 clp->cl_clientid.cl_id = current_clientid++;
1141 static void gen_confirm(struct nfs4_client *clp)
1146 p = (u32 *)clp->cl_confirm.data;
1147 *p++ = get_seconds();
1151 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1153 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1156 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1158 struct nfs4_stid *s;
1160 s = find_stateid(cl, t);
1163 if (typemask & s->sc_type)
1168 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1169 struct svc_rqst *rqstp, nfs4_verifier *verf)
1171 struct nfs4_client *clp;
1172 struct sockaddr *sa = svc_addr(rqstp);
1175 clp = alloc_client(name);
1180 princ = svc_gss_principal(rqstp);
1182 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
1183 if (clp->cl_principal == NULL) {
1189 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1190 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1191 clp->cl_time = get_seconds();
1192 clear_bit(0, &clp->cl_cb_slot_busy);
1193 copy_verf(clp, verf);
1194 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1195 clp->cl_flavor = rqstp->rq_flavor;
1196 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1198 clp->cl_cb_session = NULL;
1203 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1205 unsigned int idhashval;
1207 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1208 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1209 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1214 move_to_confirmed(struct nfs4_client *clp)
1216 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1217 unsigned int strhashval;
1219 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1220 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1221 strhashval = clientstr_hashval(clp->cl_recdir);
1222 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1226 static struct nfs4_client *
1227 find_confirmed_client(clientid_t *clid)
1229 struct nfs4_client *clp;
1230 unsigned int idhashval = clientid_hashval(clid->cl_id);
1232 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1233 if (same_clid(&clp->cl_clientid, clid)) {
1241 static struct nfs4_client *
1242 find_unconfirmed_client(clientid_t *clid)
1244 struct nfs4_client *clp;
1245 unsigned int idhashval = clientid_hashval(clid->cl_id);
1247 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1248 if (same_clid(&clp->cl_clientid, clid))
1254 static bool clp_used_exchangeid(struct nfs4_client *clp)
1256 return clp->cl_exchange_flags != 0;
1259 static struct nfs4_client *
1260 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1262 struct nfs4_client *clp;
1264 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1265 if (same_name(clp->cl_recdir, dname))
1271 static struct nfs4_client *
1272 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1274 struct nfs4_client *clp;
1276 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1277 if (same_name(clp->cl_recdir, dname))
1284 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1286 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1287 struct sockaddr *sa = svc_addr(rqstp);
1288 u32 scopeid = rpc_get_scope_id(sa);
1289 unsigned short expected_family;
1291 /* Currently, we only support tcp and tcp6 for the callback channel */
1292 if (se->se_callback_netid_len == 3 &&
1293 !memcmp(se->se_callback_netid_val, "tcp", 3))
1294 expected_family = AF_INET;
1295 else if (se->se_callback_netid_len == 4 &&
1296 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1297 expected_family = AF_INET6;
1301 conn->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1302 se->se_callback_addr_len,
1303 (struct sockaddr *)&conn->cb_addr,
1304 sizeof(conn->cb_addr));
1306 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1309 if (conn->cb_addr.ss_family == AF_INET6)
1310 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1312 conn->cb_prog = se->se_callback_prog;
1313 conn->cb_ident = se->se_callback_ident;
1314 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1317 conn->cb_addr.ss_family = AF_UNSPEC;
1318 conn->cb_addrlen = 0;
1319 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1320 "will not receive delegations\n",
1321 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1327 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1330 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1332 struct nfsd4_slot *slot = resp->cstate.slot;
1335 dprintk("--> %s slot %p\n", __func__, slot);
1337 slot->sl_opcnt = resp->opcnt;
1338 slot->sl_status = resp->cstate.status;
1340 if (nfsd4_not_cached(resp)) {
1341 slot->sl_datalen = 0;
1344 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1345 base = (char *)resp->cstate.datap -
1346 (char *)resp->xbuf->head[0].iov_base;
1347 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1349 WARN("%s: sessions DRC could not cache compound\n", __func__);
1354 * Encode the replay sequence operation from the slot values.
1355 * If cachethis is FALSE encode the uncached rep error on the next
1356 * operation which sets resp->p and increments resp->opcnt for
1357 * nfs4svc_encode_compoundres.
1361 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1362 struct nfsd4_compoundres *resp)
1364 struct nfsd4_op *op;
1365 struct nfsd4_slot *slot = resp->cstate.slot;
1367 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1368 resp->opcnt, resp->cstate.slot->sl_cachethis);
1370 /* Encode the replayed sequence operation */
1371 op = &args->ops[resp->opcnt - 1];
1372 nfsd4_encode_operation(resp, op);
1374 /* Return nfserr_retry_uncached_rep in next operation. */
1375 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
1376 op = &args->ops[resp->opcnt++];
1377 op->status = nfserr_retry_uncached_rep;
1378 nfsd4_encode_operation(resp, op);
1384 * The sequence operation is not cached because we can use the slot and
1388 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1389 struct nfsd4_sequence *seq)
1391 struct nfsd4_slot *slot = resp->cstate.slot;
1394 dprintk("--> %s slot %p\n", __func__, slot);
1396 /* Either returns 0 or nfserr_retry_uncached */
1397 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1398 if (status == nfserr_retry_uncached_rep)
1401 /* The sequence operation has been encoded, cstate->datap set. */
1402 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1404 resp->opcnt = slot->sl_opcnt;
1405 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1406 status = slot->sl_status;
1412 * Set the exchange_id flags returned by the server.
1415 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1417 /* pNFS is not supported */
1418 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1420 /* Referrals are supported, Migration is not. */
1421 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1423 /* set the wire flags to return to client. */
1424 clid->flags = new->cl_exchange_flags;
1428 nfsd4_exchange_id(struct svc_rqst *rqstp,
1429 struct nfsd4_compound_state *cstate,
1430 struct nfsd4_exchange_id *exid)
1432 struct nfs4_client *unconf, *conf, *new;
1434 unsigned int strhashval;
1435 char dname[HEXDIR_LEN];
1436 char addr_str[INET6_ADDRSTRLEN];
1437 nfs4_verifier verf = exid->verifier;
1438 struct sockaddr *sa = svc_addr(rqstp);
1440 rpc_ntop(sa, addr_str, sizeof(addr_str));
1441 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1442 "ip_addr=%s flags %x, spa_how %d\n",
1443 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1444 addr_str, exid->flags, exid->spa_how);
1446 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1447 return nfserr_inval;
1449 /* Currently only support SP4_NONE */
1450 switch (exid->spa_how) {
1454 return nfserr_serverfault;
1456 BUG(); /* checked by xdr code */
1458 return nfserr_serverfault; /* no excuse :-/ */
1461 status = nfs4_make_rec_clidname(dname, &exid->clname);
1466 strhashval = clientstr_hashval(dname);
1471 conf = find_confirmed_client_by_str(dname, strhashval);
1473 if (!clp_used_exchangeid(conf)) {
1474 status = nfserr_clid_inuse; /* XXX: ? */
1477 if (!same_verf(&verf, &conf->cl_verifier)) {
1478 /* 18.35.4 case 8 */
1479 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1480 status = nfserr_not_same;
1483 /* Client reboot: destroy old state */
1484 expire_client(conf);
1487 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1488 /* 18.35.4 case 9 */
1489 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1490 status = nfserr_perm;
1493 expire_client(conf);
1497 * Set bit when the owner id and verifier map to an already
1498 * confirmed client id (18.35.3).
1500 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1503 * Falling into 18.35.4 case 2, possible router replay.
1504 * Leave confirmed record intact and return same result.
1506 copy_verf(conf, &verf);
1511 /* 18.35.4 case 7 */
1512 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1513 status = nfserr_noent;
1517 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1520 * Possible retry or client restart. Per 18.35.4 case 4,
1521 * a new unconfirmed record should be generated regardless
1522 * of whether any properties have changed.
1524 expire_client(unconf);
1529 new = create_client(exid->clname, dname, rqstp, &verf);
1531 status = nfserr_jukebox;
1536 add_to_unconfirmed(new, strhashval);
1538 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1539 exid->clientid.cl_id = new->cl_clientid.cl_id;
1542 nfsd4_set_ex_flags(new, exid);
1544 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1545 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1549 nfs4_unlock_state();
1551 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1556 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1558 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1561 /* The slot is in use, and no response has been sent. */
1563 if (seqid == slot_seqid)
1564 return nfserr_jukebox;
1566 return nfserr_seq_misordered;
1569 if (likely(seqid == slot_seqid + 1))
1572 if (seqid == slot_seqid)
1573 return nfserr_replay_cache;
1575 if (seqid == 1 && (slot_seqid + 1) == 0)
1577 /* Misordered replay or misordered new request */
1578 return nfserr_seq_misordered;
1582 * Cache the create session result into the create session single DRC
1583 * slot cache by saving the xdr structure. sl_seqid has been set.
1584 * Do this for solo or embedded create session operations.
1587 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1588 struct nfsd4_clid_slot *slot, int nfserr)
1590 slot->sl_status = nfserr;
1591 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1595 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1596 struct nfsd4_clid_slot *slot)
1598 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1599 return slot->sl_status;
1602 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1603 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1604 1 + /* MIN tag is length with zero, only length */ \
1605 3 + /* version, opcount, opcode */ \
1606 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1607 /* seqid, slotID, slotID, cache */ \
1608 4 ) * sizeof(__be32))
1610 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1611 2 + /* verifier: AUTH_NULL, length 0 */\
1613 1 + /* MIN tag is length with zero, only length */ \
1614 3 + /* opcount, opcode, opstatus*/ \
1615 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1616 /* seqid, slotID, slotID, slotID, status */ \
1617 5 ) * sizeof(__be32))
1619 static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1621 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1622 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1626 nfsd4_create_session(struct svc_rqst *rqstp,
1627 struct nfsd4_compound_state *cstate,
1628 struct nfsd4_create_session *cr_ses)
1630 struct sockaddr *sa = svc_addr(rqstp);
1631 struct nfs4_client *conf, *unconf;
1632 struct nfsd4_session *new;
1633 struct nfsd4_clid_slot *cs_slot = NULL;
1634 bool confirm_me = false;
1637 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1638 return nfserr_inval;
1641 unconf = find_unconfirmed_client(&cr_ses->clientid);
1642 conf = find_confirmed_client(&cr_ses->clientid);
1645 cs_slot = &conf->cl_cs_slot;
1646 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1647 if (status == nfserr_replay_cache) {
1648 dprintk("Got a create_session replay! seqid= %d\n",
1650 /* Return the cached reply status */
1651 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1653 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1654 status = nfserr_seq_misordered;
1655 dprintk("Sequence misordered!\n");
1656 dprintk("Expected seqid= %d but got seqid= %d\n",
1657 cs_slot->sl_seqid, cr_ses->seqid);
1660 } else if (unconf) {
1661 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1662 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1663 status = nfserr_clid_inuse;
1667 cs_slot = &unconf->cl_cs_slot;
1668 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1670 /* an unconfirmed replay returns misordered */
1671 status = nfserr_seq_misordered;
1678 status = nfserr_stale_clientid;
1683 * XXX: we should probably set this at creation time, and check
1684 * for consistent minorversion use throughout:
1686 conf->cl_minorversion = 1;
1688 * We do not support RDMA or persistent sessions
1690 cr_ses->flags &= ~SESSION4_PERSIST;
1691 cr_ses->flags &= ~SESSION4_RDMA;
1693 status = nfserr_toosmall;
1694 if (check_forechannel_attrs(cr_ses->fore_channel))
1697 status = nfserr_jukebox;
1698 new = alloc_init_session(rqstp, conf, cr_ses);
1702 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1703 NFS4_MAX_SESSIONID_LEN);
1704 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1705 sizeof(struct nfsd4_channel_attrs));
1706 cs_slot->sl_seqid++;
1707 cr_ses->seqid = cs_slot->sl_seqid;
1709 /* cache solo and embedded create sessions under the state lock */
1710 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1712 move_to_confirmed(conf);
1714 nfs4_unlock_state();
1715 dprintk("%s returns %d\n", __func__, ntohl(status));
1719 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1721 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1722 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1724 return argp->opcnt == resp->opcnt;
1727 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1730 case NFS4_CDFC4_FORE:
1731 case NFS4_CDFC4_BACK:
1733 case NFS4_CDFC4_FORE_OR_BOTH:
1734 case NFS4_CDFC4_BACK_OR_BOTH:
1735 *dir = NFS4_CDFC4_BOTH;
1738 return nfserr_inval;
1741 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1742 struct nfsd4_compound_state *cstate,
1743 struct nfsd4_bind_conn_to_session *bcts)
1747 if (!nfsd4_last_compound_op(rqstp))
1748 return nfserr_not_only_op;
1749 spin_lock(&client_lock);
1750 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1751 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1752 * client_lock iself: */
1753 if (cstate->session) {
1754 nfsd4_get_session(cstate->session);
1755 atomic_inc(&cstate->session->se_client->cl_refcount);
1757 spin_unlock(&client_lock);
1758 if (!cstate->session)
1759 return nfserr_badsession;
1761 status = nfsd4_map_bcts_dir(&bcts->dir);
1763 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1767 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1771 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1775 nfsd4_destroy_session(struct svc_rqst *r,
1776 struct nfsd4_compound_state *cstate,
1777 struct nfsd4_destroy_session *sessionid)
1779 struct nfsd4_session *ses;
1780 u32 status = nfserr_badsession;
1783 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1784 * - Should we return nfserr_back_chan_busy if waiting for
1785 * callbacks on to-be-destroyed session?
1786 * - Do we need to clear any callback info from previous session?
1789 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1790 if (!nfsd4_last_compound_op(r))
1791 return nfserr_not_only_op;
1793 dump_sessionid(__func__, &sessionid->sessionid);
1794 spin_lock(&client_lock);
1795 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1797 spin_unlock(&client_lock);
1801 unhash_session(ses);
1802 spin_unlock(&client_lock);
1805 nfsd4_probe_callback_sync(ses->se_client);
1806 nfs4_unlock_state();
1808 nfsd4_del_conns(ses);
1810 nfsd4_put_session(ses);
1813 dprintk("%s returns %d\n", __func__, ntohl(status));
1817 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1819 struct nfsd4_conn *c;
1821 list_for_each_entry(c, &s->se_conns, cn_persession) {
1822 if (c->cn_xprt == xpt) {
1829 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1831 struct nfs4_client *clp = ses->se_client;
1832 struct nfsd4_conn *c;
1835 spin_lock(&clp->cl_lock);
1836 c = __nfsd4_find_conn(new->cn_xprt, ses);
1838 spin_unlock(&clp->cl_lock);
1842 __nfsd4_hash_conn(new, ses);
1843 spin_unlock(&clp->cl_lock);
1844 ret = nfsd4_register_conn(new);
1846 /* oops; xprt is already down: */
1847 nfsd4_conn_lost(&new->cn_xpt_user);
1851 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1853 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1855 return args->opcnt > session->se_fchannel.maxops;
1858 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1859 struct nfsd4_session *session)
1861 struct xdr_buf *xb = &rqstp->rq_arg;
1863 return xb->len > session->se_fchannel.maxreq_sz;
1867 nfsd4_sequence(struct svc_rqst *rqstp,
1868 struct nfsd4_compound_state *cstate,
1869 struct nfsd4_sequence *seq)
1871 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1872 struct nfsd4_session *session;
1873 struct nfsd4_slot *slot;
1874 struct nfsd4_conn *conn;
1877 if (resp->opcnt != 1)
1878 return nfserr_sequence_pos;
1881 * Will be either used or freed by nfsd4_sequence_check_conn
1884 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1886 return nfserr_jukebox;
1888 spin_lock(&client_lock);
1889 status = nfserr_badsession;
1890 session = find_in_sessionid_hashtbl(&seq->sessionid);
1894 status = nfserr_too_many_ops;
1895 if (nfsd4_session_too_many_ops(rqstp, session))
1898 status = nfserr_req_too_big;
1899 if (nfsd4_request_too_big(rqstp, session))
1902 status = nfserr_badslot;
1903 if (seq->slotid >= session->se_fchannel.maxreqs)
1906 slot = session->se_slots[seq->slotid];
1907 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1909 /* We do not negotiate the number of slots yet, so set the
1910 * maxslots to the session maxreqs which is used to encode
1911 * sr_highest_slotid and the sr_target_slot id to maxslots */
1912 seq->maxslots = session->se_fchannel.maxreqs;
1914 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
1915 if (status == nfserr_replay_cache) {
1916 cstate->slot = slot;
1917 cstate->session = session;
1918 /* Return the cached reply status and set cstate->status
1919 * for nfsd4_proc_compound processing */
1920 status = nfsd4_replay_cache_entry(resp, seq);
1921 cstate->status = nfserr_replay_cache;
1927 nfsd4_sequence_check_conn(conn, session);
1930 /* Success! bump slot seqid */
1931 slot->sl_inuse = true;
1932 slot->sl_seqid = seq->seqid;
1933 slot->sl_cachethis = seq->cachethis;
1935 cstate->slot = slot;
1936 cstate->session = session;
1939 /* Hold a session reference until done processing the compound. */
1940 if (cstate->session) {
1941 struct nfs4_client *clp = session->se_client;
1943 nfsd4_get_session(cstate->session);
1944 atomic_inc(&clp->cl_refcount);
1945 switch (clp->cl_cb_state) {
1947 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
1949 case NFSD4_CB_FAULT:
1950 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
1953 seq->status_flags = 0;
1957 spin_unlock(&client_lock);
1958 dprintk("%s: return %d\n", __func__, ntohl(status));
1962 static inline bool has_resources(struct nfs4_client *clp)
1964 return !list_empty(&clp->cl_openowners)
1965 || !list_empty(&clp->cl_delegations)
1966 || !list_empty(&clp->cl_sessions);
1970 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
1972 struct nfs4_client *conf, *unconf, *clp;
1976 unconf = find_unconfirmed_client(&dc->clientid);
1977 conf = find_confirmed_client(&dc->clientid);
1982 if (!is_client_expired(conf) && has_resources(conf)) {
1983 status = nfserr_clientid_busy;
1987 /* rfc5661 18.50.3 */
1988 if (cstate->session && conf == cstate->session->se_client) {
1989 status = nfserr_clientid_busy;
1995 status = nfserr_stale_clientid;
2001 nfs4_unlock_state();
2002 dprintk("%s return %d\n", __func__, ntohl(status));
2007 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2011 if (rc->rca_one_fs) {
2012 if (!cstate->current_fh.fh_dentry)
2013 return nfserr_nofilehandle;
2015 * We don't take advantage of the rca_one_fs case.
2016 * That's OK, it's optional, we can safely ignore it.
2022 status = nfserr_complete_already;
2023 if (cstate->session->se_client->cl_firststate)
2026 status = nfserr_stale_clientid;
2027 if (is_client_expired(cstate->session->se_client))
2029 * The following error isn't really legal.
2030 * But we only get here if the client just explicitly
2031 * destroyed the client. Surely it no longer cares what
2032 * error it gets back on an operation for the dead
2038 nfsd4_create_clid_dir(cstate->session->se_client);
2040 nfs4_unlock_state();
2045 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2046 struct nfsd4_setclientid *setclid)
2048 struct xdr_netobj clname = setclid->se_name;
2049 nfs4_verifier clverifier = setclid->se_verf;
2050 unsigned int strhashval;
2051 struct nfs4_client *conf, *unconf, *new;
2053 char dname[HEXDIR_LEN];
2055 status = nfs4_make_rec_clidname(dname, &clname);
2060 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2061 * We get here on a DRC miss.
2064 strhashval = clientstr_hashval(dname);
2067 conf = find_confirmed_client_by_str(dname, strhashval);
2069 /* RFC 3530 14.2.33 CASE 0: */
2070 status = nfserr_clid_inuse;
2071 if (clp_used_exchangeid(conf))
2073 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2074 char addr_str[INET6_ADDRSTRLEN];
2075 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2077 dprintk("NFSD: setclientid: string in use by client "
2078 "at %s\n", addr_str);
2083 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2084 * has a description of SETCLIENTID request processing consisting
2085 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2087 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2088 status = nfserr_jukebox;
2091 * RFC 3530 14.2.33 CASE 4:
2092 * placed first, because it is the normal case
2095 expire_client(unconf);
2096 new = create_client(clname, dname, rqstp, &clverifier);
2100 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
2102 * RFC 3530 14.2.33 CASE 1:
2103 * probable callback update
2106 /* Note this is removing unconfirmed {*x***},
2107 * which is stronger than RFC recommended {vxc**}.
2108 * This has the advantage that there is at most
2109 * one {*x***} in either list at any time.
2111 expire_client(unconf);
2113 new = create_client(clname, dname, rqstp, &clverifier);
2116 copy_clid(new, conf);
2117 } else if (!unconf) {
2119 * RFC 3530 14.2.33 CASE 2:
2120 * probable client reboot; state will be removed if
2123 new = create_client(clname, dname, rqstp, &clverifier);
2129 * RFC 3530 14.2.33 CASE 3:
2130 * probable client reboot; state will be removed if
2133 expire_client(unconf);
2134 new = create_client(clname, dname, rqstp, &clverifier);
2140 * XXX: we should probably set this at creation time, and check
2141 * for consistent minorversion use throughout:
2143 new->cl_minorversion = 0;
2144 gen_callback(new, setclid, rqstp);
2145 add_to_unconfirmed(new, strhashval);
2146 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2147 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2148 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2151 nfs4_unlock_state();
2157 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2158 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2159 * bullets, labeled as CASE1 - CASE4 below.
2162 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2163 struct nfsd4_compound_state *cstate,
2164 struct nfsd4_setclientid_confirm *setclientid_confirm)
2166 struct sockaddr *sa = svc_addr(rqstp);
2167 struct nfs4_client *conf, *unconf;
2168 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2169 clientid_t * clid = &setclientid_confirm->sc_clientid;
2172 if (STALE_CLIENTID(clid))
2173 return nfserr_stale_clientid;
2175 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2176 * We get here on a DRC miss.
2181 conf = find_confirmed_client(clid);
2182 unconf = find_unconfirmed_client(clid);
2184 status = nfserr_clid_inuse;
2185 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
2187 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
2191 * section 14.2.34 of RFC 3530 has a description of
2192 * SETCLIENTID_CONFIRM request processing consisting
2193 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2195 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
2197 * RFC 3530 14.2.34 CASE 1:
2200 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
2201 status = nfserr_clid_inuse;
2203 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2204 nfsd4_probe_callback(conf);
2205 expire_client(unconf);
2209 } else if (conf && !unconf) {
2211 * RFC 3530 14.2.34 CASE 2:
2212 * probable retransmitted request; play it safe and
2215 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
2216 status = nfserr_clid_inuse;
2219 } else if (!conf && unconf
2220 && same_verf(&unconf->cl_confirm, &confirm)) {
2222 * RFC 3530 14.2.34 CASE 3:
2223 * Normal case; new or rebooted client:
2225 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
2226 status = nfserr_clid_inuse;
2229 clientstr_hashval(unconf->cl_recdir);
2230 conf = find_confirmed_client_by_str(unconf->cl_recdir,
2233 nfsd4_remove_clid_dir(conf);
2234 expire_client(conf);
2236 move_to_confirmed(unconf);
2238 nfsd4_probe_callback(conf);
2241 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
2242 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
2245 * RFC 3530 14.2.34 CASE 4:
2246 * Client probably hasn't noticed that we rebooted yet.
2248 status = nfserr_stale_clientid;
2250 /* check that we have hit one of the cases...*/
2251 status = nfserr_clid_inuse;
2254 nfs4_unlock_state();
2258 static struct nfs4_file *nfsd4_alloc_file(void)
2260 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2263 /* OPEN Share state helper functions */
2264 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2266 unsigned int hashval = file_hashval(ino);
2268 atomic_set(&fp->fi_ref, 1);
2269 INIT_LIST_HEAD(&fp->fi_hash);
2270 INIT_LIST_HEAD(&fp->fi_stateids);
2271 INIT_LIST_HEAD(&fp->fi_delegations);
2272 fp->fi_inode = igrab(ino);
2273 fp->fi_had_conflict = false;
2274 fp->fi_lease = NULL;
2275 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2276 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2277 spin_lock(&recall_lock);
2278 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2279 spin_unlock(&recall_lock);
2283 nfsd4_free_slab(struct kmem_cache **slab)
2287 kmem_cache_destroy(*slab);
2292 nfsd4_free_slabs(void)
2294 nfsd4_free_slab(&openowner_slab);
2295 nfsd4_free_slab(&lockowner_slab);
2296 nfsd4_free_slab(&file_slab);
2297 nfsd4_free_slab(&stateid_slab);
2298 nfsd4_free_slab(&deleg_slab);
2302 nfsd4_init_slabs(void)
2304 openowner_slab = kmem_cache_create("nfsd4_openowners",
2305 sizeof(struct nfs4_openowner), 0, 0, NULL);
2306 if (openowner_slab == NULL)
2308 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2309 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2310 if (lockowner_slab == NULL)
2312 file_slab = kmem_cache_create("nfsd4_files",
2313 sizeof(struct nfs4_file), 0, 0, NULL);
2314 if (file_slab == NULL)
2316 stateid_slab = kmem_cache_create("nfsd4_stateids",
2317 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2318 if (stateid_slab == NULL)
2320 deleg_slab = kmem_cache_create("nfsd4_delegations",
2321 sizeof(struct nfs4_delegation), 0, 0, NULL);
2322 if (deleg_slab == NULL)
2327 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2331 void nfs4_free_openowner(struct nfs4_openowner *oo)
2333 kfree(oo->oo_owner.so_owner.data);
2334 kmem_cache_free(openowner_slab, oo);
2337 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2339 kfree(lo->lo_owner.so_owner.data);
2340 kmem_cache_free(lockowner_slab, lo);
2343 static void init_nfs4_replay(struct nfs4_replay *rp)
2345 rp->rp_status = nfserr_serverfault;
2347 rp->rp_buf = rp->rp_ibuf;
2350 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2352 struct nfs4_stateowner *sop;
2354 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2358 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2359 if (!sop->so_owner.data) {
2360 kmem_cache_free(slab, sop);
2363 sop->so_owner.len = owner->len;
2365 INIT_LIST_HEAD(&sop->so_stateids);
2366 sop->so_client = clp;
2367 init_nfs4_replay(&sop->so_replay);
2371 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2373 list_add(&oo->oo_owner.so_strhash, &open_ownerstr_hashtbl[strhashval]);
2374 list_add(&oo->oo_perclient, &clp->cl_openowners);
2377 static struct nfs4_openowner *
2378 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2379 struct nfs4_openowner *oo;
2381 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2384 oo->oo_owner.so_is_open_owner = 1;
2385 oo->oo_owner.so_seqid = open->op_seqid;
2386 oo->oo_flags = NFS4_OO_NEW;
2388 oo->oo_last_closed_stid = NULL;
2389 INIT_LIST_HEAD(&oo->oo_close_lru);
2390 hash_openowner(oo, clp, strhashval);
2394 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2395 struct nfs4_openowner *oo = open->op_openowner;
2396 struct nfs4_client *clp = oo->oo_owner.so_client;
2398 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2399 INIT_LIST_HEAD(&stp->st_lockowners);
2400 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2401 list_add(&stp->st_perfile, &fp->fi_stateids);
2402 stp->st_stateowner = &oo->oo_owner;
2405 stp->st_access_bmap = 0;
2406 stp->st_deny_bmap = 0;
2407 __set_bit(open->op_share_access, &stp->st_access_bmap);
2408 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2409 stp->st_openstp = NULL;
2413 move_to_close_lru(struct nfs4_openowner *oo)
2415 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2417 list_move_tail(&oo->oo_close_lru, &close_lru);
2418 oo->oo_time = get_seconds();
2422 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2425 return (sop->so_owner.len == owner->len) &&
2426 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2427 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2430 static struct nfs4_openowner *
2431 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2433 struct nfs4_stateowner *so;
2434 struct nfs4_openowner *oo;
2436 list_for_each_entry(so, &open_ownerstr_hashtbl[hashval], so_strhash) {
2437 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2439 renew_client(oo->oo_owner.so_client);
2446 /* search file_hashtbl[] for file */
2447 static struct nfs4_file *
2448 find_file(struct inode *ino)
2450 unsigned int hashval = file_hashval(ino);
2451 struct nfs4_file *fp;
2453 spin_lock(&recall_lock);
2454 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2455 if (fp->fi_inode == ino) {
2457 spin_unlock(&recall_lock);
2461 spin_unlock(&recall_lock);
2466 * Called to check deny when READ with all zero stateid or
2467 * WRITE with all zero or all one stateid
2470 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2472 struct inode *ino = current_fh->fh_dentry->d_inode;
2473 struct nfs4_file *fp;
2474 struct nfs4_ol_stateid *stp;
2477 dprintk("NFSD: nfs4_share_conflict\n");
2479 fp = find_file(ino);
2482 ret = nfserr_locked;
2483 /* Search for conflicting share reservations */
2484 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2485 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2486 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2495 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2497 /* We're assuming the state code never drops its reference
2498 * without first removing the lease. Since we're in this lease
2499 * callback (and since the lease code is serialized by the kernel
2500 * lock) we know the server hasn't removed the lease yet, we know
2501 * it's safe to take a reference: */
2502 atomic_inc(&dp->dl_count);
2504 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2506 /* only place dl_time is set. protected by lock_flocks*/
2507 dp->dl_time = get_seconds();
2509 nfsd4_cb_recall(dp);
2512 /* Called from break_lease() with lock_flocks() held. */
2513 static void nfsd_break_deleg_cb(struct file_lock *fl)
2515 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2516 struct nfs4_delegation *dp;
2519 /* We assume break_lease is only called once per lease: */
2520 BUG_ON(fp->fi_had_conflict);
2522 * We don't want the locks code to timeout the lease for us;
2523 * we'll remove it ourself if a delegation isn't returned
2526 fl->fl_break_time = 0;
2528 spin_lock(&recall_lock);
2529 fp->fi_had_conflict = true;
2530 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2531 nfsd_break_one_deleg(dp);
2532 spin_unlock(&recall_lock);
2536 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2539 return lease_modify(onlist, arg);
2544 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2545 .lm_break = nfsd_break_deleg_cb,
2546 .lm_change = nfsd_change_deleg_cb,
2549 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2551 if (nfsd4_has_session(cstate))
2553 if (seqid == so->so_seqid - 1)
2554 return nfserr_replay_me;
2555 if (seqid == so->so_seqid)
2557 return nfserr_bad_seqid;
2561 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2562 struct nfsd4_open *open)
2564 clientid_t *clientid = &open->op_clientid;
2565 struct nfs4_client *clp = NULL;
2566 unsigned int strhashval;
2567 struct nfs4_openowner *oo = NULL;
2570 if (STALE_CLIENTID(&open->op_clientid))
2571 return nfserr_stale_clientid;
2573 * In case we need it later, after we've already created the
2574 * file and don't want to risk a further failure:
2576 open->op_file = nfsd4_alloc_file();
2577 if (open->op_file == NULL)
2578 return nfserr_jukebox;
2580 strhashval = open_ownerstr_hashval(clientid->cl_id, &open->op_owner);
2581 oo = find_openstateowner_str(strhashval, open);
2582 open->op_openowner = oo;
2584 clp = find_confirmed_client(clientid);
2586 return nfserr_expired;
2589 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2590 /* Replace unconfirmed owners without checking for replay. */
2591 clp = oo->oo_owner.so_client;
2592 release_openowner(oo);
2593 open->op_openowner = NULL;
2596 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2599 clp = oo->oo_owner.so_client;
2602 oo = alloc_init_open_stateowner(strhashval, clp, open);
2604 return nfserr_jukebox;
2605 open->op_openowner = oo;
2607 open->op_stp = nfs4_alloc_stateid(clp);
2609 return nfserr_jukebox;
2613 static inline __be32
2614 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2616 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2617 return nfserr_openmode;
2622 static int share_access_to_flags(u32 share_access)
2624 share_access &= ~NFS4_SHARE_WANT_MASK;
2626 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2629 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2631 struct nfs4_stid *ret;
2633 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2636 return delegstateid(ret);
2639 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2641 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2642 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2646 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2647 struct nfs4_delegation **dp)
2650 __be32 status = nfserr_bad_stateid;
2652 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2655 flags = share_access_to_flags(open->op_share_access);
2656 status = nfs4_check_delegmode(*dp, flags);
2660 if (!nfsd4_is_deleg_cur(open))
2664 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2669 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2671 struct nfs4_ol_stateid *local;
2672 struct nfs4_openowner *oo = open->op_openowner;
2674 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2675 /* ignore lock owners */
2676 if (local->st_stateowner->so_is_open_owner == 0)
2678 /* remember if we have seen this open owner */
2679 if (local->st_stateowner == &oo->oo_owner)
2681 /* check for conflicting share reservations */
2682 if (!test_share(local, open))
2683 return nfserr_share_denied;
2688 static void nfs4_free_stateid(struct nfs4_ol_stateid *s)
2690 kmem_cache_free(stateid_slab, s);
2693 static inline int nfs4_access_to_access(u32 nfs4_access)
2697 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2698 flags |= NFSD_MAY_READ;
2699 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2700 flags |= NFSD_MAY_WRITE;
2704 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2705 struct svc_fh *cur_fh, struct nfsd4_open *open)
2708 int oflag = nfs4_access_to_omode(open->op_share_access);
2709 int access = nfs4_access_to_access(open->op_share_access);
2711 if (!fp->fi_fds[oflag]) {
2712 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2713 &fp->fi_fds[oflag]);
2717 nfs4_file_get_access(fp, oflag);
2722 static inline __be32
2723 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2724 struct nfsd4_open *open)
2726 struct iattr iattr = {
2727 .ia_valid = ATTR_SIZE,
2730 if (!open->op_truncate)
2732 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2733 return nfserr_inval;
2734 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2738 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)
2740 u32 op_share_access = open->op_share_access;
2744 new_access = !test_bit(op_share_access, &stp->st_access_bmap);
2746 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2750 status = nfsd4_truncate(rqstp, cur_fh, open);
2753 int oflag = nfs4_access_to_omode(op_share_access);
2754 nfs4_file_put_access(fp, oflag);
2758 /* remember the open */
2759 __set_bit(op_share_access, &stp->st_access_bmap);
2760 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2767 nfs4_set_claim_prev(struct nfsd4_open *open)
2769 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2770 open->op_openowner->oo_owner.so_client->cl_firststate = 1;
2773 /* Should we give out recallable state?: */
2774 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2776 if (clp->cl_cb_state == NFSD4_CB_UP)
2779 * In the sessions case, since we don't have to establish a
2780 * separate connection for callbacks, we assume it's OK
2781 * until we hear otherwise:
2783 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2786 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2788 struct file_lock *fl;
2790 fl = locks_alloc_lock();
2793 locks_init_lock(fl);
2794 fl->fl_lmops = &nfsd_lease_mng_ops;
2795 fl->fl_flags = FL_LEASE;
2796 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2797 fl->fl_end = OFFSET_MAX;
2798 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2799 fl->fl_pid = current->tgid;
2803 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2805 struct nfs4_file *fp = dp->dl_file;
2806 struct file_lock *fl;
2809 fl = nfs4_alloc_init_lease(dp, flag);
2812 fl->fl_file = find_readable_file(fp);
2813 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2814 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2816 list_del_init(&dp->dl_perclnt);
2817 locks_free_lock(fl);
2821 fp->fi_deleg_file = fl->fl_file;
2822 get_file(fp->fi_deleg_file);
2823 atomic_set(&fp->fi_delegees, 1);
2824 list_add(&dp->dl_perfile, &fp->fi_delegations);
2828 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2830 struct nfs4_file *fp = dp->dl_file;
2833 return nfs4_setlease(dp, flag);
2834 spin_lock(&recall_lock);
2835 if (fp->fi_had_conflict) {
2836 spin_unlock(&recall_lock);
2839 atomic_inc(&fp->fi_delegees);
2840 list_add(&dp->dl_perfile, &fp->fi_delegations);
2841 spin_unlock(&recall_lock);
2842 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2847 * Attempt to hand out a delegation.
2850 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2852 struct nfs4_delegation *dp;
2853 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2855 int status, flag = 0;
2857 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2858 flag = NFS4_OPEN_DELEGATE_NONE;
2859 open->op_recall = 0;
2860 switch (open->op_claim_type) {
2861 case NFS4_OPEN_CLAIM_PREVIOUS:
2863 open->op_recall = 1;
2864 flag = open->op_delegate_type;
2865 if (flag == NFS4_OPEN_DELEGATE_NONE)
2868 case NFS4_OPEN_CLAIM_NULL:
2869 /* Let's not give out any delegations till everyone's
2870 * had the chance to reclaim theirs.... */
2871 if (locks_in_grace())
2873 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2875 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2876 flag = NFS4_OPEN_DELEGATE_WRITE;
2878 flag = NFS4_OPEN_DELEGATE_READ;
2884 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2887 status = nfs4_set_delegation(dp, flag);
2891 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2893 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2894 STATEID_VAL(&dp->dl_stid.sc_stateid));
2896 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2897 && flag == NFS4_OPEN_DELEGATE_NONE
2898 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2899 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2900 open->op_delegate_type = flag;
2903 nfs4_put_delegation(dp);
2905 flag = NFS4_OPEN_DELEGATE_NONE;
2910 * called with nfs4_lock_state() held.
2913 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2915 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2916 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
2917 struct nfs4_file *fp = NULL;
2918 struct inode *ino = current_fh->fh_dentry->d_inode;
2919 struct nfs4_ol_stateid *stp = NULL;
2920 struct nfs4_delegation *dp = NULL;
2924 * Lookup file; if found, lookup stateid and check open request,
2925 * and check for delegations in the process of being recalled.
2926 * If not found, create the nfs4_file struct
2928 fp = find_file(ino);
2930 if ((status = nfs4_check_open(fp, open, &stp)))
2932 status = nfs4_check_deleg(cl, fp, open, &dp);
2936 status = nfserr_bad_stateid;
2937 if (nfsd4_is_deleg_cur(open))
2939 status = nfserr_jukebox;
2941 open->op_file = NULL;
2942 nfsd4_init_file(fp, ino);
2946 * OPEN the file, or upgrade an existing OPEN.
2947 * If truncate fails, the OPEN fails.
2950 /* Stateid was found, this is an OPEN upgrade */
2951 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
2955 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
2959 open->op_stp = NULL;
2960 init_open_stateid(stp, fp, open);
2961 status = nfsd4_truncate(rqstp, current_fh, open);
2963 release_open_stateid(stp);
2967 update_stateid(&stp->st_stid.sc_stateid);
2968 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
2970 if (nfsd4_has_session(&resp->cstate))
2971 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2974 * Attempt to hand out a delegation. No error return, because the
2975 * OPEN succeeds even if we fail.
2977 nfs4_open_delegation(current_fh, open, stp);
2981 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2982 STATEID_VAL(&stp->st_stid.sc_stateid));
2986 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2987 nfs4_set_claim_prev(open);
2989 * To finish the open response, we just need to set the rflags.
2991 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2992 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
2993 !nfsd4_has_session(&resp->cstate))
2994 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2999 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3001 if (open->op_openowner) {
3002 struct nfs4_openowner *oo = open->op_openowner;
3004 if (!list_empty(&oo->oo_owner.so_stateids))
3005 list_del_init(&oo->oo_close_lru);
3006 if (oo->oo_flags & NFS4_OO_NEW) {
3008 release_openowner(oo);
3009 open->op_openowner = NULL;
3011 oo->oo_flags &= ~NFS4_OO_NEW;
3015 nfsd4_free_file(open->op_file);
3017 nfs4_free_stateid(open->op_stp);
3021 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3024 struct nfs4_client *clp;
3028 dprintk("process_renew(%08x/%08x): starting\n",
3029 clid->cl_boot, clid->cl_id);
3030 status = nfserr_stale_clientid;
3031 if (STALE_CLIENTID(clid))
3033 clp = find_confirmed_client(clid);
3034 status = nfserr_expired;
3036 /* We assume the client took too long to RENEW. */
3037 dprintk("nfsd4_renew: clientid not found!\n");
3040 status = nfserr_cb_path_down;
3041 if (!list_empty(&clp->cl_delegations)
3042 && clp->cl_cb_state != NFSD4_CB_UP)
3046 nfs4_unlock_state();
3050 static struct lock_manager nfsd4_manager = {
3054 nfsd4_end_grace(void)
3056 dprintk("NFSD: end of grace period\n");
3057 nfsd4_recdir_purge_old();
3058 locks_end_grace(&nfsd4_manager);
3060 * Now that every NFSv4 client has had the chance to recover and
3061 * to see the (possibly new, possibly shorter) lease time, we
3062 * can safely set the next grace time to the current lease time:
3064 nfsd4_grace = nfsd4_lease;
3068 nfs4_laundromat(void)
3070 struct nfs4_client *clp;
3071 struct nfs4_openowner *oo;
3072 struct nfs4_delegation *dp;
3073 struct list_head *pos, *next, reaplist;
3074 time_t cutoff = get_seconds() - nfsd4_lease;
3075 time_t t, clientid_val = nfsd4_lease;
3076 time_t u, test_val = nfsd4_lease;
3080 dprintk("NFSD: laundromat service - starting\n");
3081 if (locks_in_grace())
3083 INIT_LIST_HEAD(&reaplist);
3084 spin_lock(&client_lock);
3085 list_for_each_safe(pos, next, &client_lru) {
3086 clp = list_entry(pos, struct nfs4_client, cl_lru);
3087 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3088 t = clp->cl_time - cutoff;
3089 if (clientid_val > t)
3093 if (atomic_read(&clp->cl_refcount)) {
3094 dprintk("NFSD: client in use (clientid %08x)\n",
3095 clp->cl_clientid.cl_id);
3098 unhash_client_locked(clp);
3099 list_add(&clp->cl_lru, &reaplist);
3101 spin_unlock(&client_lock);
3102 list_for_each_safe(pos, next, &reaplist) {
3103 clp = list_entry(pos, struct nfs4_client, cl_lru);
3104 dprintk("NFSD: purging unused client (clientid %08x)\n",
3105 clp->cl_clientid.cl_id);
3106 nfsd4_remove_clid_dir(clp);
3109 spin_lock(&recall_lock);
3110 list_for_each_safe(pos, next, &del_recall_lru) {
3111 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3112 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3113 u = dp->dl_time - cutoff;
3118 list_move(&dp->dl_recall_lru, &reaplist);
3120 spin_unlock(&recall_lock);
3121 list_for_each_safe(pos, next, &reaplist) {
3122 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3123 list_del_init(&dp->dl_recall_lru);
3124 unhash_delegation(dp);
3126 test_val = nfsd4_lease;
3127 list_for_each_safe(pos, next, &close_lru) {
3128 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3129 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3130 u = oo->oo_time - cutoff;
3135 release_openowner(oo);
3137 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3138 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3139 nfs4_unlock_state();
3140 return clientid_val;
3143 static struct workqueue_struct *laundry_wq;
3144 static void laundromat_main(struct work_struct *);
3145 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3148 laundromat_main(struct work_struct *not_used)
3152 t = nfs4_laundromat();
3153 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3154 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3157 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3159 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3160 return nfserr_bad_stateid;
3165 STALE_STATEID(stateid_t *stateid)
3167 if (stateid->si_opaque.so_clid.cl_boot == boot_time)
3169 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3170 STATEID_VAL(stateid));
3175 access_permit_read(unsigned long access_bmap)
3177 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
3178 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
3179 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
3183 access_permit_write(unsigned long access_bmap)
3185 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
3186 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
3190 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3192 __be32 status = nfserr_openmode;
3194 /* For lock stateid's, we test the parent open, not the lock: */
3195 if (stp->st_openstp)
3196 stp = stp->st_openstp;
3197 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
3199 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
3206 static inline __be32
3207 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3209 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3211 else if (locks_in_grace()) {
3212 /* Answer in remaining cases depends on existence of
3213 * conflicting state; so we must wait out the grace period. */
3214 return nfserr_grace;
3215 } else if (flags & WR_STATE)
3216 return nfs4_share_conflict(current_fh,
3217 NFS4_SHARE_DENY_WRITE);
3218 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3219 return nfs4_share_conflict(current_fh,
3220 NFS4_SHARE_DENY_READ);
3224 * Allow READ/WRITE during grace period on recovered state only for files
3225 * that are not able to provide mandatory locking.
3228 grace_disallows_io(struct inode *inode)
3230 return locks_in_grace() && mandatory_lock(inode);
3233 /* Returns true iff a is later than b: */
3234 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3236 return (s32)a->si_generation - (s32)b->si_generation > 0;
3239 static int check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3242 * When sessions are used the stateid generation number is ignored
3245 if (has_session && in->si_generation == 0)
3248 if (in->si_generation == ref->si_generation)
3251 /* If the client sends us a stateid from the future, it's buggy: */
3252 if (stateid_generation_after(in, ref))
3253 return nfserr_bad_stateid;
3255 * However, we could see a stateid from the past, even from a
3256 * non-buggy client. For example, if the client sends a lock
3257 * while some IO is outstanding, the lock may bump si_generation
3258 * while the IO is still in flight. The client could avoid that
3259 * situation by waiting for responses on all the IO requests,
3260 * but better performance may result in retrying IO that
3261 * receives an old_stateid error if requests are rarely
3262 * reordered in flight:
3264 return nfserr_old_stateid;
3267 __be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3269 struct nfs4_stid *s;
3270 struct nfs4_ol_stateid *ols;
3273 if (STALE_STATEID(stateid))
3274 return nfserr_stale_stateid;
3276 s = find_stateid(cl, stateid);
3278 return nfserr_stale_stateid;
3279 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3282 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3284 ols = openlockstateid(s);
3285 if (ols->st_stateowner->so_is_open_owner
3286 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3287 return nfserr_bad_stateid;
3291 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
3293 struct nfs4_client *cl;
3295 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3296 return nfserr_bad_stateid;
3297 if (STALE_STATEID(stateid))
3298 return nfserr_stale_stateid;
3299 cl = find_confirmed_client(&stateid->si_opaque.so_clid);
3301 return nfserr_expired;
3302 *s = find_stateid_by_type(cl, stateid, typemask);
3304 return nfserr_bad_stateid;
3310 * Checks for stateid operations
3313 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3314 stateid_t *stateid, int flags, struct file **filpp)
3316 struct nfs4_stid *s;
3317 struct nfs4_ol_stateid *stp = NULL;
3318 struct nfs4_delegation *dp = NULL;
3319 struct svc_fh *current_fh = &cstate->current_fh;
3320 struct inode *ino = current_fh->fh_dentry->d_inode;
3326 if (grace_disallows_io(ino))
3327 return nfserr_grace;
3329 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3330 return check_special_stateids(current_fh, stateid, flags);
3332 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
3335 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3338 switch (s->sc_type) {
3339 case NFS4_DELEG_STID:
3340 dp = delegstateid(s);
3341 status = nfs4_check_delegmode(dp, flags);
3345 *filpp = dp->dl_file->fi_deleg_file;
3349 case NFS4_OPEN_STID:
3350 case NFS4_LOCK_STID:
3351 stp = openlockstateid(s);
3352 status = nfs4_check_fh(current_fh, stp);
3355 if (stp->st_stateowner->so_is_open_owner
3356 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3358 status = nfs4_check_openmode(stp, flags);
3362 if (flags & RD_STATE)
3363 *filpp = find_readable_file(stp->st_file);
3365 *filpp = find_writeable_file(stp->st_file);
3369 return nfserr_bad_stateid;
3377 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3379 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3380 return nfserr_locks_held;
3381 release_lock_stateid(stp);
3386 * Test if the stateid is valid
3389 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3390 struct nfsd4_test_stateid *test_stateid)
3392 /* real work is done during encoding */
3397 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3398 struct nfsd4_free_stateid *free_stateid)
3400 stateid_t *stateid = &free_stateid->fr_stateid;
3401 struct nfs4_stid *s;
3402 struct nfs4_client *cl = cstate->session->se_client;
3403 __be32 ret = nfserr_bad_stateid;
3406 s = find_stateid(cl, stateid);
3409 switch (s->sc_type) {
3410 case NFS4_DELEG_STID:
3411 ret = nfserr_locks_held;
3413 case NFS4_OPEN_STID:
3414 case NFS4_LOCK_STID:
3415 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3418 if (s->sc_type == NFS4_LOCK_STID)
3419 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3421 ret = nfserr_locks_held;
3424 ret = nfserr_bad_stateid;
3427 nfs4_unlock_state();
3434 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3435 RD_STATE : WR_STATE;
3438 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3440 struct svc_fh *current_fh = &cstate->current_fh;
3441 struct nfs4_stateowner *sop = stp->st_stateowner;
3444 status = nfsd4_check_seqid(cstate, sop, seqid);
3447 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3449 * "Closed" stateid's exist *only* to return
3450 * nfserr_replay_me from the previous step.
3452 return nfserr_bad_stateid;
3453 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3456 return nfs4_check_fh(current_fh, stp);
3460 * Checks for sequence id mutating operations.
3463 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3464 stateid_t *stateid, char typemask,
3465 struct nfs4_ol_stateid **stpp)
3468 struct nfs4_stid *s;
3470 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3471 seqid, STATEID_VAL(stateid));
3474 status = nfsd4_lookup_stateid(stateid, typemask, &s);
3477 *stpp = openlockstateid(s);
3478 cstate->replay_owner = (*stpp)->st_stateowner;
3480 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3483 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3486 struct nfs4_openowner *oo;
3488 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3489 NFS4_OPEN_STID, stpp);
3492 oo = openowner((*stpp)->st_stateowner);
3493 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3494 return nfserr_bad_stateid;
3499 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3500 struct nfsd4_open_confirm *oc)
3503 struct nfs4_openowner *oo;
3504 struct nfs4_ol_stateid *stp;
3506 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3507 (int)cstate->current_fh.fh_dentry->d_name.len,
3508 cstate->current_fh.fh_dentry->d_name.name);
3510 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3516 status = nfs4_preprocess_seqid_op(cstate,
3517 oc->oc_seqid, &oc->oc_req_stateid,
3518 NFS4_OPEN_STID, &stp);
3521 oo = openowner(stp->st_stateowner);
3522 status = nfserr_bad_stateid;
3523 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3525 oo->oo_flags |= NFS4_OO_CONFIRMED;
3526 update_stateid(&stp->st_stid.sc_stateid);
3527 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3528 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3529 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3531 nfsd4_create_clid_dir(oo->oo_owner.so_client);
3534 if (!cstate->replay_owner)
3535 nfs4_unlock_state();
3539 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3541 if (!test_bit(access, &stp->st_access_bmap))
3543 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3544 __clear_bit(access, &stp->st_access_bmap);
3547 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3549 switch (to_access) {
3550 case NFS4_SHARE_ACCESS_READ:
3551 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3552 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3554 case NFS4_SHARE_ACCESS_WRITE:
3555 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3556 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3558 case NFS4_SHARE_ACCESS_BOTH:
3566 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3569 for (i = 0; i < 4; i++) {
3570 if ((i & deny) != i)
3571 __clear_bit(i, bmap);
3576 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3577 struct nfsd4_compound_state *cstate,
3578 struct nfsd4_open_downgrade *od)
3581 struct nfs4_ol_stateid *stp;
3583 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3584 (int)cstate->current_fh.fh_dentry->d_name.len,
3585 cstate->current_fh.fh_dentry->d_name.name);
3587 /* We don't yet support WANT bits: */
3588 od->od_share_access &= NFS4_SHARE_ACCESS_MASK;
3591 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3592 &od->od_stateid, &stp);
3595 status = nfserr_inval;
3596 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3597 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3598 stp->st_access_bmap, od->od_share_access);
3601 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3602 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3603 stp->st_deny_bmap, od->od_share_deny);
3606 nfs4_stateid_downgrade(stp, od->od_share_access);
3608 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3610 update_stateid(&stp->st_stid.sc_stateid);
3611 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3614 if (!cstate->replay_owner)
3615 nfs4_unlock_state();
3619 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3621 struct nfs4_openowner *oo;
3622 struct nfs4_ol_stateid *s;
3624 if (!so->so_is_open_owner)
3627 s = oo->oo_last_closed_stid;
3630 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3631 /* Release the last_closed_stid on the next seqid bump: */
3632 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3635 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3636 release_last_closed_stateid(oo);
3639 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3641 unhash_open_stateid(s);
3642 s->st_stid.sc_type = NFS4_CLOSED_STID;
3646 * nfs4_unlock_state() called after encode
3649 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3650 struct nfsd4_close *close)
3653 struct nfs4_openowner *oo;
3654 struct nfs4_ol_stateid *stp;
3656 dprintk("NFSD: nfsd4_close on file %.*s\n",
3657 (int)cstate->current_fh.fh_dentry->d_name.len,
3658 cstate->current_fh.fh_dentry->d_name.name);
3661 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3663 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3667 oo = openowner(stp->st_stateowner);
3669 update_stateid(&stp->st_stid.sc_stateid);
3670 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3672 nfsd4_close_open_stateid(stp);
3673 release_last_closed_stateid(oo);
3674 oo->oo_last_closed_stid = stp;
3676 /* place unused nfs4_stateowners on so_close_lru list to be
3677 * released by the laundromat service after the lease period
3678 * to enable us to handle CLOSE replay
3680 if (list_empty(&oo->oo_owner.so_stateids))
3681 move_to_close_lru(oo);
3683 if (!cstate->replay_owner)
3684 nfs4_unlock_state();
3689 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3690 struct nfsd4_delegreturn *dr)
3692 struct nfs4_delegation *dp;
3693 stateid_t *stateid = &dr->dr_stateid;
3694 struct nfs4_stid *s;
3695 struct inode *inode;
3698 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3700 inode = cstate->current_fh.fh_dentry->d_inode;
3703 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s);
3706 dp = delegstateid(s);
3707 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3711 unhash_delegation(dp);
3713 nfs4_unlock_state();
3720 * Lock owner state (byte-range locks)
3722 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3723 #define LOCK_HASH_BITS 8
3724 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3725 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3728 end_offset(u64 start, u64 len)
3733 return end >= start ? end: NFS4_MAX_UINT64;
3736 /* last octet in a range */
3738 last_byte_offset(u64 start, u64 len)
3744 return end > start ? end - 1: NFS4_MAX_UINT64;
3747 static inline unsigned int
3748 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3749 struct xdr_netobj *ownername)
3751 return (file_hashval(inode) + cl_id
3752 + opaque_hashval(ownername->data, ownername->len))
3756 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3759 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3760 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3761 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3762 * locking, this prevents us from being completely protocol-compliant. The
3763 * real solution to this problem is to start using unsigned file offsets in
3764 * the VFS, but this is a very deep change!
3767 nfs4_transform_lock_offset(struct file_lock *lock)
3769 if (lock->fl_start < 0)
3770 lock->fl_start = OFFSET_MAX;
3771 if (lock->fl_end < 0)
3772 lock->fl_end = OFFSET_MAX;
3775 /* Hack!: For now, we're defining this just so we can use a pointer to it
3776 * as a unique cookie to identify our (NFSv4's) posix locks. */
3777 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3781 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3783 struct nfs4_lockowner *lo;
3785 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3786 lo = (struct nfs4_lockowner *) fl->fl_owner;
3787 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3788 lo->lo_owner.so_owner.len, GFP_KERNEL);
3789 if (!deny->ld_owner.data)
3790 /* We just don't care that much */
3792 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3793 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3796 deny->ld_owner.len = 0;
3797 deny->ld_owner.data = NULL;
3798 deny->ld_clientid.cl_boot = 0;
3799 deny->ld_clientid.cl_id = 0;
3801 deny->ld_start = fl->fl_start;
3802 deny->ld_length = NFS4_MAX_UINT64;
3803 if (fl->fl_end != NFS4_MAX_UINT64)
3804 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3805 deny->ld_type = NFS4_READ_LT;
3806 if (fl->fl_type != F_RDLCK)
3807 deny->ld_type = NFS4_WRITE_LT;
3810 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3812 struct nfs4_ol_stateid *lst;
3814 if (!same_owner_str(&lo->lo_owner, owner, clid))
3816 lst = list_first_entry(&lo->lo_owner.so_stateids,
3817 struct nfs4_ol_stateid, st_perstateowner);
3818 return lst->st_file->fi_inode == inode;
3821 static struct nfs4_lockowner *
3822 find_lockowner_str(struct inode *inode, clientid_t *clid,
3823 struct xdr_netobj *owner)
3825 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3826 struct nfs4_lockowner *lo;
3827 struct nfs4_stateowner *op;
3829 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
3831 if (same_lockowner_ino(lo, inode, clid, owner))
3837 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3839 list_add(&lo->lo_owner.so_strhash, &lock_ownerstr_hashtbl[strhashval]);
3840 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3844 * Alloc a lock owner structure.
3845 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3848 * strhashval = lock_ownerstr_hashval
3851 static struct nfs4_lockowner *
3852 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3853 struct nfs4_lockowner *lo;
3855 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3858 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3859 lo->lo_owner.so_is_open_owner = 0;
3860 /* It is the openowner seqid that will be incremented in encode in the
3861 * case of new lockowners; so increment the lock seqid manually: */
3862 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3863 hash_lockowner(lo, strhashval, clp, open_stp);
3867 static struct nfs4_ol_stateid *
3868 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3870 struct nfs4_ol_stateid *stp;
3871 struct nfs4_client *clp = lo->lo_owner.so_client;
3873 stp = nfs4_alloc_stateid(clp);
3876 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3877 list_add(&stp->st_perfile, &fp->fi_stateids);
3878 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3879 stp->st_stateowner = &lo->lo_owner;
3882 stp->st_access_bmap = 0;
3883 stp->st_deny_bmap = open_stp->st_deny_bmap;
3884 stp->st_openstp = open_stp;
3889 check_lock_length(u64 offset, u64 length)
3891 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3892 LOFF_OVERFLOW(offset, length)));
3895 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
3897 struct nfs4_file *fp = lock_stp->st_file;
3898 int oflag = nfs4_access_to_omode(access);
3900 if (test_bit(access, &lock_stp->st_access_bmap))
3902 nfs4_file_get_access(fp, oflag);
3903 __set_bit(access, &lock_stp->st_access_bmap);
3910 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3911 struct nfsd4_lock *lock)
3913 struct nfs4_openowner *open_sop = NULL;
3914 struct nfs4_lockowner *lock_sop = NULL;
3915 struct nfs4_ol_stateid *lock_stp;
3916 struct nfs4_file *fp;
3917 struct file *filp = NULL;
3918 struct file_lock file_lock;
3919 struct file_lock conflock;
3921 unsigned int strhashval;
3925 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3926 (long long) lock->lk_offset,
3927 (long long) lock->lk_length);
3929 if (check_lock_length(lock->lk_offset, lock->lk_length))
3930 return nfserr_inval;
3932 if ((status = fh_verify(rqstp, &cstate->current_fh,
3933 S_IFREG, NFSD_MAY_LOCK))) {
3934 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3940 if (lock->lk_is_new) {
3942 * Client indicates that this is a new lockowner.
3943 * Use open owner and open stateid to create lock owner and
3946 struct nfs4_ol_stateid *open_stp = NULL;
3948 status = nfserr_stale_clientid;
3949 if (!nfsd4_has_session(cstate) &&
3950 STALE_CLIENTID(&lock->lk_new_clientid))
3953 /* validate and update open stateid and open seqid */
3954 status = nfs4_preprocess_confirmed_seqid_op(cstate,
3955 lock->lk_new_open_seqid,
3956 &lock->lk_new_open_stateid,
3960 open_sop = openowner(open_stp->st_stateowner);
3961 status = nfserr_bad_stateid;
3962 if (!nfsd4_has_session(cstate) &&
3963 !same_clid(&open_sop->oo_owner.so_client->cl_clientid,
3964 &lock->v.new.clientid))
3966 /* create lockowner and lock stateid */
3967 fp = open_stp->st_file;
3968 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3969 open_sop->oo_owner.so_client->cl_clientid.cl_id,
3970 &lock->v.new.owner);
3971 /* XXX: Do we need to check for duplicate stateowners on
3972 * the same file, or should they just be allowed (and
3973 * create new stateids)? */
3974 status = nfserr_jukebox;
3975 lock_sop = alloc_init_lock_stateowner(strhashval,
3976 open_sop->oo_owner.so_client, open_stp, lock);
3977 if (lock_sop == NULL)
3979 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
3980 if (lock_stp == NULL)
3983 /* lock (lock owner + lock stateid) already exists */
3984 status = nfs4_preprocess_seqid_op(cstate,
3985 lock->lk_old_lock_seqid,
3986 &lock->lk_old_lock_stateid,
3987 NFS4_LOCK_STID, &lock_stp);
3990 lock_sop = lockowner(lock_stp->st_stateowner);
3991 fp = lock_stp->st_file;
3993 /* lock_sop and lock_stp have been created or found */
3995 lkflg = setlkflg(lock->lk_type);
3996 status = nfs4_check_openmode(lock_stp, lkflg);
4000 status = nfserr_grace;
4001 if (locks_in_grace() && !lock->lk_reclaim)
4003 status = nfserr_no_grace;
4004 if (!locks_in_grace() && lock->lk_reclaim)
4007 locks_init_lock(&file_lock);
4008 switch (lock->lk_type) {
4011 filp = find_readable_file(lock_stp->st_file);
4013 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4014 file_lock.fl_type = F_RDLCK;
4017 case NFS4_WRITEW_LT:
4018 filp = find_writeable_file(lock_stp->st_file);
4020 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4021 file_lock.fl_type = F_WRLCK;
4024 status = nfserr_inval;
4028 status = nfserr_openmode;
4031 file_lock.fl_owner = (fl_owner_t)lock_sop;
4032 file_lock.fl_pid = current->tgid;
4033 file_lock.fl_file = filp;
4034 file_lock.fl_flags = FL_POSIX;
4035 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4037 file_lock.fl_start = lock->lk_offset;
4038 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4039 nfs4_transform_lock_offset(&file_lock);
4042 * Try to lock the file in the VFS.
4043 * Note: locks.c uses the BKL to protect the inode's lock list.
4046 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4048 case 0: /* success! */
4049 update_stateid(&lock_stp->st_stid.sc_stateid);
4050 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4054 case (EAGAIN): /* conflock holds conflicting lock */
4055 status = nfserr_denied;
4056 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4057 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4060 status = nfserr_deadlock;
4063 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4064 status = nfserrno(err);
4068 if (status && lock->lk_is_new && lock_sop)
4069 release_lockowner(lock_sop);
4070 if (!cstate->replay_owner)
4071 nfs4_unlock_state();
4076 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4077 * so we do a temporary open here just to get an open file to pass to
4078 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4081 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4084 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4086 err = nfserrno(vfs_test_lock(file, lock));
4096 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4097 struct nfsd4_lockt *lockt)
4099 struct inode *inode;
4100 struct file_lock file_lock;
4101 struct nfs4_lockowner *lo;
4104 if (locks_in_grace())
4105 return nfserr_grace;
4107 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4108 return nfserr_inval;
4112 status = nfserr_stale_clientid;
4113 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4116 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4119 inode = cstate->current_fh.fh_dentry->d_inode;
4120 locks_init_lock(&file_lock);
4121 switch (lockt->lt_type) {
4124 file_lock.fl_type = F_RDLCK;
4127 case NFS4_WRITEW_LT:
4128 file_lock.fl_type = F_WRLCK;
4131 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4132 status = nfserr_inval;
4136 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4138 file_lock.fl_owner = (fl_owner_t)lo;
4139 file_lock.fl_pid = current->tgid;
4140 file_lock.fl_flags = FL_POSIX;
4142 file_lock.fl_start = lockt->lt_offset;
4143 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4145 nfs4_transform_lock_offset(&file_lock);
4147 status = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4151 if (file_lock.fl_type != F_UNLCK) {
4152 status = nfserr_denied;
4153 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4156 nfs4_unlock_state();
4161 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4162 struct nfsd4_locku *locku)
4164 struct nfs4_ol_stateid *stp;
4165 struct file *filp = NULL;
4166 struct file_lock file_lock;
4170 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4171 (long long) locku->lu_offset,
4172 (long long) locku->lu_length);
4174 if (check_lock_length(locku->lu_offset, locku->lu_length))
4175 return nfserr_inval;
4179 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4180 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4183 filp = find_any_file(stp->st_file);
4185 status = nfserr_lock_range;
4189 locks_init_lock(&file_lock);
4190 file_lock.fl_type = F_UNLCK;
4191 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4192 file_lock.fl_pid = current->tgid;
4193 file_lock.fl_file = filp;
4194 file_lock.fl_flags = FL_POSIX;
4195 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4196 file_lock.fl_start = locku->lu_offset;
4198 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4199 nfs4_transform_lock_offset(&file_lock);
4202 * Try to unlock the file in the VFS.
4204 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4206 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4210 * OK, unlock succeeded; the only thing left to do is update the stateid.
4212 update_stateid(&stp->st_stid.sc_stateid);
4213 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4216 if (!cstate->replay_owner)
4217 nfs4_unlock_state();
4221 status = nfserrno(err);
4227 * 1: locks held by lockowner
4228 * 0: no locks held by lockowner
4231 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4233 struct file_lock **flpp;
4234 struct inode *inode = filp->fi_inode;
4238 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4239 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4250 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4251 struct nfsd4_compound_state *cstate,
4252 struct nfsd4_release_lockowner *rlockowner)
4254 clientid_t *clid = &rlockowner->rl_clientid;
4255 struct nfs4_stateowner *sop;
4256 struct nfs4_lockowner *lo;
4257 struct nfs4_ol_stateid *stp;
4258 struct xdr_netobj *owner = &rlockowner->rl_owner;
4259 struct list_head matches;
4263 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4264 clid->cl_boot, clid->cl_id);
4266 /* XXX check for lease expiration */
4268 status = nfserr_stale_clientid;
4269 if (STALE_CLIENTID(clid))
4274 status = nfserr_locks_held;
4275 /* XXX: we're doing a linear search through all the lockowners.
4276 * Yipes! For now we'll just hope clients aren't really using
4277 * release_lockowner much, but eventually we have to fix these
4278 * data structures. */
4279 INIT_LIST_HEAD(&matches);
4280 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4281 list_for_each_entry(sop, &lock_ownerstr_hashtbl[i], so_strhash) {
4282 if (!same_owner_str(sop, owner, clid))
4284 list_for_each_entry(stp, &sop->so_stateids,
4286 lo = lockowner(sop);
4287 if (check_for_locks(stp->st_file, lo))
4289 list_add(&lo->lo_list, &matches);
4293 /* Clients probably won't expect us to return with some (but not all)
4294 * of the lockowner state released; so don't release any until all
4295 * have been checked. */
4297 while (!list_empty(&matches)) {
4298 lo = list_entry(matches.next, struct nfs4_lockowner,
4300 /* unhash_stateowner deletes so_perclient only
4301 * for openowners. */
4302 list_del(&lo->lo_list);
4303 release_lockowner(lo);
4306 nfs4_unlock_state();
4310 static inline struct nfs4_client_reclaim *
4313 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4317 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4319 unsigned int strhashval = clientstr_hashval(name);
4320 struct nfs4_client *clp;
4322 clp = find_confirmed_client_by_str(name, strhashval);
4327 * failure => all reset bets are off, nfserr_no_grace...
4330 nfs4_client_to_reclaim(const char *name)
4332 unsigned int strhashval;
4333 struct nfs4_client_reclaim *crp = NULL;
4335 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4336 crp = alloc_reclaim();
4339 strhashval = clientstr_hashval(name);
4340 INIT_LIST_HEAD(&crp->cr_strhash);
4341 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4342 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4343 reclaim_str_hashtbl_size++;
4348 nfs4_release_reclaim(void)
4350 struct nfs4_client_reclaim *crp = NULL;
4353 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4354 while (!list_empty(&reclaim_str_hashtbl[i])) {
4355 crp = list_entry(reclaim_str_hashtbl[i].next,
4356 struct nfs4_client_reclaim, cr_strhash);
4357 list_del(&crp->cr_strhash);
4359 reclaim_str_hashtbl_size--;
4362 BUG_ON(reclaim_str_hashtbl_size);
4366 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4367 static struct nfs4_client_reclaim *
4368 nfs4_find_reclaim_client(clientid_t *clid)
4370 unsigned int strhashval;
4371 struct nfs4_client *clp;
4372 struct nfs4_client_reclaim *crp = NULL;
4375 /* find clientid in conf_id_hashtbl */
4376 clp = find_confirmed_client(clid);
4380 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4381 clp->cl_name.len, clp->cl_name.data,
4384 /* find clp->cl_name in reclaim_str_hashtbl */
4385 strhashval = clientstr_hashval(clp->cl_recdir);
4386 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4387 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4395 * Called from OPEN. Look for clientid in reclaim list.
4398 nfs4_check_open_reclaim(clientid_t *clid)
4400 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
4403 /* initialization to perform at module load time: */
4406 nfs4_state_init(void)
4410 status = nfsd4_init_slabs();
4413 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4414 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4415 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4416 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4417 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4418 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4420 for (i = 0; i < SESSION_HASH_SIZE; i++)
4421 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4422 for (i = 0; i < FILE_HASH_SIZE; i++) {
4423 INIT_LIST_HEAD(&file_hashtbl[i]);
4425 for (i = 0; i < OPEN_OWNER_HASH_SIZE; i++) {
4426 INIT_LIST_HEAD(&open_ownerstr_hashtbl[i]);
4428 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4429 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4431 memset(&onestateid, ~0, sizeof(stateid_t));
4432 INIT_LIST_HEAD(&close_lru);
4433 INIT_LIST_HEAD(&client_lru);
4434 INIT_LIST_HEAD(&del_recall_lru);
4435 reclaim_str_hashtbl_size = 0;
4440 nfsd4_load_reboot_recovery_data(void)
4445 nfsd4_init_recdir();
4446 status = nfsd4_recdir_load();
4447 nfs4_unlock_state();
4449 printk("NFSD: Failure reading reboot recovery data\n");
4453 * Since the lifetime of a delegation isn't limited to that of an open, a
4454 * client may quite reasonably hang on to a delegation as long as it has
4455 * the inode cached. This becomes an obvious problem the first time a
4456 * client's inode cache approaches the size of the server's total memory.
4458 * For now we avoid this problem by imposing a hard limit on the number
4459 * of delegations, which varies according to the server's memory size.
4462 set_max_delegations(void)
4465 * Allow at most 4 delegations per megabyte of RAM. Quick
4466 * estimates suggest that in the worst case (where every delegation
4467 * is for a different inode), a delegation could take about 1.5K,
4468 * giving a worst case usage of about 6% of memory.
4470 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4473 /* initialization to perform when the nfsd service is started: */
4476 __nfs4_state_start(void)
4480 boot_time = get_seconds();
4481 locks_start_grace(&nfsd4_manager);
4482 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4484 ret = set_callback_cred();
4487 laundry_wq = create_singlethread_workqueue("nfsd4");
4488 if (laundry_wq == NULL)
4490 ret = nfsd4_create_callback_queue();
4492 goto out_free_laundry;
4493 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4494 set_max_delegations();
4497 destroy_workqueue(laundry_wq);
4502 nfs4_state_start(void)
4504 nfsd4_load_reboot_recovery_data();
4505 return __nfs4_state_start();
4509 __nfs4_state_shutdown(void)
4512 struct nfs4_client *clp = NULL;
4513 struct nfs4_delegation *dp = NULL;
4514 struct list_head *pos, *next, reaplist;
4516 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4517 while (!list_empty(&conf_id_hashtbl[i])) {
4518 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4521 while (!list_empty(&unconf_str_hashtbl[i])) {
4522 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4526 INIT_LIST_HEAD(&reaplist);
4527 spin_lock(&recall_lock);
4528 list_for_each_safe(pos, next, &del_recall_lru) {
4529 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4530 list_move(&dp->dl_recall_lru, &reaplist);
4532 spin_unlock(&recall_lock);
4533 list_for_each_safe(pos, next, &reaplist) {
4534 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4535 list_del_init(&dp->dl_recall_lru);
4536 unhash_delegation(dp);
4539 nfsd4_shutdown_recdir();
4543 nfs4_state_shutdown(void)
4545 cancel_delayed_work_sync(&laundromat_work);
4546 destroy_workqueue(laundry_wq);
4547 locks_end_grace(&nfsd4_manager);
4549 nfs4_release_reclaim();
4550 __nfs4_state_shutdown();
4551 nfs4_unlock_state();
4552 nfsd4_destroy_callback_queue();