2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/clnt.h>
46 #include "current_stateid.h"
47 #include "fault_inject.h"
51 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 time_t nfsd4_lease = 90; /* default lease time */
55 time_t nfsd4_grace = 90;
57 #define all_ones {{~0,~0},~0}
58 static const stateid_t one_stateid = {
60 .si_opaque = all_ones,
62 static const stateid_t zero_stateid = {
65 static const stateid_t currentstateid = {
69 static u64 current_sessionid = 1;
71 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
72 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
73 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
75 /* forward declarations */
76 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
80 /* Currently used for almost all code touching nfsv4 state: */
81 static DEFINE_MUTEX(client_mutex);
84 * Currently used for the del_recall_lru and file hash table. In an
85 * effort to decrease the scope of the client_mutex, this spinlock may
86 * eventually cover more:
88 static DEFINE_SPINLOCK(recall_lock);
90 static struct kmem_cache *openowner_slab = NULL;
91 static struct kmem_cache *lockowner_slab = NULL;
92 static struct kmem_cache *file_slab = NULL;
93 static struct kmem_cache *stateid_slab = NULL;
94 static struct kmem_cache *deleg_slab = NULL;
99 mutex_lock(&client_mutex);
102 static void free_session(struct kref *);
104 /* Must be called under the client_lock */
105 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
107 kref_put(&ses->se_ref, free_session);
110 static void nfsd4_get_session(struct nfsd4_session *ses)
112 kref_get(&ses->se_ref);
116 nfs4_unlock_state(void)
118 mutex_unlock(&client_mutex);
122 opaque_hashval(const void *ptr, int nbytes)
124 unsigned char *cptr = (unsigned char *) ptr;
134 static struct list_head del_recall_lru;
136 static void nfsd4_free_file(struct nfs4_file *f)
138 kmem_cache_free(file_slab, f);
142 put_nfs4_file(struct nfs4_file *fi)
144 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
145 list_del(&fi->fi_hash);
146 spin_unlock(&recall_lock);
153 get_nfs4_file(struct nfs4_file *fi)
155 atomic_inc(&fi->fi_ref);
158 static int num_delegations;
159 unsigned int max_delegations;
162 * Open owner state (share locks)
165 /* hash tables for lock and open owners */
166 #define OWNER_HASH_BITS 8
167 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
168 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
170 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
174 ret = opaque_hashval(ownername->data, ownername->len);
176 return ret & OWNER_HASH_MASK;
179 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
181 /* hash table for nfs4_file */
182 #define FILE_HASH_BITS 8
183 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
185 static unsigned int file_hashval(struct inode *ino)
187 /* XXX: why are we hashing on inode pointer, anyway? */
188 return hash_ptr(ino, FILE_HASH_BITS);
191 static struct list_head file_hashtbl[FILE_HASH_SIZE];
193 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
195 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
196 atomic_inc(&fp->fi_access[oflag]);
199 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
201 if (oflag == O_RDWR) {
202 __nfs4_file_get_access(fp, O_RDONLY);
203 __nfs4_file_get_access(fp, O_WRONLY);
205 __nfs4_file_get_access(fp, oflag);
208 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
210 if (fp->fi_fds[oflag]) {
211 fput(fp->fi_fds[oflag]);
212 fp->fi_fds[oflag] = NULL;
216 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
218 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
219 nfs4_file_put_fd(fp, oflag);
221 * It's also safe to get rid of the RDWR open *if*
222 * we no longer have need of the other kind of access
223 * or if we already have the other kind of open:
225 if (fp->fi_fds[1-oflag]
226 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
227 nfs4_file_put_fd(fp, O_RDWR);
231 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
233 if (oflag == O_RDWR) {
234 __nfs4_file_put_access(fp, O_RDONLY);
235 __nfs4_file_put_access(fp, O_WRONLY);
237 __nfs4_file_put_access(fp, oflag);
240 static inline int get_new_stid(struct nfs4_stid *stid)
242 static int min_stateid = 0;
243 struct idr *stateids = &stid->sc_client->cl_stateids;
247 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
249 * Note: the necessary preallocation was done in
250 * nfs4_alloc_stateid(). The idr code caps the number of
251 * preallocations that can exist at a time, but the state lock
252 * prevents anyone from using ours before we get here:
256 * It shouldn't be a problem to reuse an opaque stateid value.
257 * I don't think it is for 4.1. But with 4.0 I worry that, for
258 * example, a stray write retransmission could be accepted by
259 * the server when it should have been rejected. Therefore,
260 * adopt a trick from the sctp code to attempt to maximize the
261 * amount of time until an id is reused, by ensuring they always
262 * "increase" (mod INT_MAX):
265 min_stateid = new_stid+1;
266 if (min_stateid == INT_MAX)
271 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
273 stateid_t *s = &stid->sc_stateid;
276 stid->sc_type = type;
277 stid->sc_client = cl;
278 s->si_opaque.so_clid = cl->cl_clientid;
279 new_id = get_new_stid(stid);
280 s->si_opaque.so_id = (u32)new_id;
281 /* Will be incremented before return to client: */
282 s->si_generation = 0;
285 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
287 struct idr *stateids = &cl->cl_stateids;
289 if (!idr_pre_get(stateids, GFP_KERNEL))
292 * Note: if we fail here (or any time between now and the time
293 * we actually get the new idr), we won't need to undo the idr
294 * preallocation, since the idr code caps the number of
295 * preallocated entries.
297 return kmem_cache_alloc(slab, GFP_KERNEL);
300 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
302 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
305 static struct nfs4_delegation *
306 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
308 struct nfs4_delegation *dp;
309 struct nfs4_file *fp = stp->st_file;
311 dprintk("NFSD alloc_init_deleg\n");
313 * Major work on the lease subsystem (for example, to support
314 * calbacks on stat) will be required before we can support
315 * write delegations properly.
317 if (type != NFS4_OPEN_DELEGATE_READ)
319 if (fp->fi_had_conflict)
321 if (num_delegations > max_delegations)
323 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
326 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
328 * delegation seqid's are never incremented. The 4.1 special
329 * meaning of seqid 0 isn't meaningful, really, but let's avoid
330 * 0 anyway just for consistency and use 1:
332 dp->dl_stid.sc_stateid.si_generation = 1;
334 INIT_LIST_HEAD(&dp->dl_perfile);
335 INIT_LIST_HEAD(&dp->dl_perclnt);
336 INIT_LIST_HEAD(&dp->dl_recall_lru);
340 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
342 atomic_set(&dp->dl_count, 1);
343 nfsd4_init_callback(&dp->dl_recall);
348 nfs4_put_delegation(struct nfs4_delegation *dp)
350 if (atomic_dec_and_test(&dp->dl_count)) {
351 dprintk("NFSD: freeing dp %p\n",dp);
352 put_nfs4_file(dp->dl_file);
353 kmem_cache_free(deleg_slab, dp);
358 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
360 if (atomic_dec_and_test(&fp->fi_delegees)) {
361 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
363 fput(fp->fi_deleg_file);
364 fp->fi_deleg_file = NULL;
368 static void unhash_stid(struct nfs4_stid *s)
370 struct idr *stateids = &s->sc_client->cl_stateids;
372 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
375 /* Called under the state lock. */
377 unhash_delegation(struct nfs4_delegation *dp)
379 unhash_stid(&dp->dl_stid);
380 list_del_init(&dp->dl_perclnt);
381 spin_lock(&recall_lock);
382 list_del_init(&dp->dl_perfile);
383 list_del_init(&dp->dl_recall_lru);
384 spin_unlock(&recall_lock);
385 nfs4_put_deleg_lease(dp->dl_file);
386 nfs4_put_delegation(dp);
393 /* client_lock protects the client lru list and session hash table */
394 static DEFINE_SPINLOCK(client_lock);
396 static unsigned int clientid_hashval(u32 id)
398 return id & CLIENT_HASH_MASK;
401 static unsigned int clientstr_hashval(const char *name)
403 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
407 * unconf_id_hashtbl[] and unconf_name_tree hold unconfirmed
410 * client_lru holds client queue ordered by nfs4_client.cl_time
413 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
414 * for last close replay.
416 * All of the above fields are protected by the client_mutex.
418 static struct rb_root unconf_name_tree;
419 static struct list_head client_lru;
420 static struct list_head close_lru;
423 * We store the NONE, READ, WRITE, and BOTH bits separately in the
424 * st_{access,deny}_bmap field of the stateid, in order to track not
425 * only what share bits are currently in force, but also what
426 * combinations of share bits previous opens have used. This allows us
427 * to enforce the recommendation of rfc 3530 14.2.19 that the server
428 * return an error if the client attempt to downgrade to a combination
429 * of share bits not explicable by closing some of its previous opens.
431 * XXX: This enforcement is actually incomplete, since we don't keep
432 * track of access/deny bit combinations; so, e.g., we allow:
434 * OPEN allow read, deny write
435 * OPEN allow both, deny none
436 * DOWNGRADE allow read, deny none
438 * which we should reject.
441 bmap_to_share_mode(unsigned long bmap) {
443 unsigned int access = 0;
445 for (i = 1; i < 4; i++) {
446 if (test_bit(i, &bmap))
453 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
454 unsigned int access, deny;
456 access = bmap_to_share_mode(stp->st_access_bmap);
457 deny = bmap_to_share_mode(stp->st_deny_bmap);
458 if ((access & open->op_share_deny) || (deny & open->op_share_access))
463 /* set share access for a given stateid */
465 set_access(u32 access, struct nfs4_ol_stateid *stp)
467 __set_bit(access, &stp->st_access_bmap);
470 /* clear share access for a given stateid */
472 clear_access(u32 access, struct nfs4_ol_stateid *stp)
474 __clear_bit(access, &stp->st_access_bmap);
477 /* test whether a given stateid has access */
479 test_access(u32 access, struct nfs4_ol_stateid *stp)
481 return test_bit(access, &stp->st_access_bmap);
484 /* set share deny for a given stateid */
486 set_deny(u32 access, struct nfs4_ol_stateid *stp)
488 __set_bit(access, &stp->st_deny_bmap);
491 /* clear share deny for a given stateid */
493 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
495 __clear_bit(access, &stp->st_deny_bmap);
498 /* test whether a given stateid is denying specific access */
500 test_deny(u32 access, struct nfs4_ol_stateid *stp)
502 return test_bit(access, &stp->st_deny_bmap);
505 static int nfs4_access_to_omode(u32 access)
507 switch (access & NFS4_SHARE_ACCESS_BOTH) {
508 case NFS4_SHARE_ACCESS_READ:
510 case NFS4_SHARE_ACCESS_WRITE:
512 case NFS4_SHARE_ACCESS_BOTH:
518 /* release all access and file references for a given stateid */
520 release_all_access(struct nfs4_ol_stateid *stp)
524 for (i = 1; i < 4; i++) {
525 if (test_access(i, stp))
526 nfs4_file_put_access(stp->st_file,
527 nfs4_access_to_omode(i));
528 clear_access(i, stp);
532 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
534 list_del(&stp->st_perfile);
535 list_del(&stp->st_perstateowner);
538 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
540 release_all_access(stp);
541 put_nfs4_file(stp->st_file);
545 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
547 kmem_cache_free(stateid_slab, stp);
550 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
554 unhash_generic_stateid(stp);
555 unhash_stid(&stp->st_stid);
556 file = find_any_file(stp->st_file);
558 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
559 close_generic_stateid(stp);
560 free_generic_stateid(stp);
563 static void unhash_lockowner(struct nfs4_lockowner *lo)
565 struct nfs4_ol_stateid *stp;
567 list_del(&lo->lo_owner.so_strhash);
568 list_del(&lo->lo_perstateid);
569 list_del(&lo->lo_owner_ino_hash);
570 while (!list_empty(&lo->lo_owner.so_stateids)) {
571 stp = list_first_entry(&lo->lo_owner.so_stateids,
572 struct nfs4_ol_stateid, st_perstateowner);
573 release_lock_stateid(stp);
577 static void release_lockowner(struct nfs4_lockowner *lo)
579 unhash_lockowner(lo);
580 nfs4_free_lockowner(lo);
584 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
586 struct nfs4_lockowner *lo;
588 while (!list_empty(&open_stp->st_lockowners)) {
589 lo = list_entry(open_stp->st_lockowners.next,
590 struct nfs4_lockowner, lo_perstateid);
591 release_lockowner(lo);
595 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
597 unhash_generic_stateid(stp);
598 release_stateid_lockowners(stp);
599 close_generic_stateid(stp);
602 static void release_open_stateid(struct nfs4_ol_stateid *stp)
604 unhash_open_stateid(stp);
605 unhash_stid(&stp->st_stid);
606 free_generic_stateid(stp);
609 static void unhash_openowner(struct nfs4_openowner *oo)
611 struct nfs4_ol_stateid *stp;
613 list_del(&oo->oo_owner.so_strhash);
614 list_del(&oo->oo_perclient);
615 while (!list_empty(&oo->oo_owner.so_stateids)) {
616 stp = list_first_entry(&oo->oo_owner.so_stateids,
617 struct nfs4_ol_stateid, st_perstateowner);
618 release_open_stateid(stp);
622 static void release_last_closed_stateid(struct nfs4_openowner *oo)
624 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
627 unhash_stid(&s->st_stid);
628 free_generic_stateid(s);
629 oo->oo_last_closed_stid = NULL;
633 static void release_openowner(struct nfs4_openowner *oo)
635 unhash_openowner(oo);
636 list_del(&oo->oo_close_lru);
637 release_last_closed_stateid(oo);
638 nfs4_free_openowner(oo);
641 #define SESSION_HASH_SIZE 512
642 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
645 hash_sessionid(struct nfs4_sessionid *sessionid)
647 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
649 return sid->sequence % SESSION_HASH_SIZE;
654 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
656 u32 *ptr = (u32 *)(&sessionid->data[0]);
657 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
661 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
668 gen_sessionid(struct nfsd4_session *ses)
670 struct nfs4_client *clp = ses->se_client;
671 struct nfsd4_sessionid *sid;
673 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
674 sid->clientid = clp->cl_clientid;
675 sid->sequence = current_sessionid++;
680 * The protocol defines ca_maxresponssize_cached to include the size of
681 * the rpc header, but all we need to cache is the data starting after
682 * the end of the initial SEQUENCE operation--the rest we regenerate
683 * each time. Therefore we can advertise a ca_maxresponssize_cached
684 * value that is the number of bytes in our cache plus a few additional
685 * bytes. In order to stay on the safe side, and not promise more than
686 * we can cache, those additional bytes must be the minimum possible: 24
687 * bytes of rpc header (xid through accept state, with AUTH_NULL
688 * verifier), 12 for the compound header (with zero-length tag), and 44
689 * for the SEQUENCE op response:
691 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
694 free_session_slots(struct nfsd4_session *ses)
698 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
699 kfree(ses->se_slots[i]);
703 * We don't actually need to cache the rpc and session headers, so we
704 * can allocate a little less for each slot:
706 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
708 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
711 static int nfsd4_sanitize_slot_size(u32 size)
713 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
714 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
720 * XXX: If we run out of reserved DRC memory we could (up to a point)
721 * re-negotiate active sessions and reduce their slot usage to make
722 * room for new connections. For now we just fail the create session.
724 static int nfsd4_get_drc_mem(int slotsize, u32 num)
728 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
730 spin_lock(&nfsd_drc_lock);
731 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
732 nfsd_drc_max_mem - nfsd_drc_mem_used);
733 num = min_t(int, num, avail / slotsize);
734 nfsd_drc_mem_used += num * slotsize;
735 spin_unlock(&nfsd_drc_lock);
740 static void nfsd4_put_drc_mem(int slotsize, int num)
742 spin_lock(&nfsd_drc_lock);
743 nfsd_drc_mem_used -= slotsize * num;
744 spin_unlock(&nfsd_drc_lock);
747 static struct nfsd4_session *__alloc_session(int slotsize, int numslots)
749 struct nfsd4_session *new;
752 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
753 + sizeof(struct nfsd4_session) > PAGE_SIZE);
754 mem = numslots * sizeof(struct nfsd4_slot *);
756 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
759 /* allocate each struct nfsd4_slot and data cache in one piece */
760 for (i = 0; i < numslots; i++) {
761 mem = sizeof(struct nfsd4_slot) + slotsize;
762 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
763 if (!new->se_slots[i])
769 kfree(new->se_slots[i]);
774 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
776 u32 maxrpc = nfsd_serv->sv_max_mesg;
778 new->maxreqs = numslots;
779 new->maxresp_cached = min_t(u32, req->maxresp_cached,
780 slotsize + NFSD_MIN_HDR_SEQ_SZ);
781 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
782 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
783 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
786 static void free_conn(struct nfsd4_conn *c)
788 svc_xprt_put(c->cn_xprt);
792 static void nfsd4_conn_lost(struct svc_xpt_user *u)
794 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
795 struct nfs4_client *clp = c->cn_session->se_client;
797 spin_lock(&clp->cl_lock);
798 if (!list_empty(&c->cn_persession)) {
799 list_del(&c->cn_persession);
802 spin_unlock(&clp->cl_lock);
803 nfsd4_probe_callback(clp);
806 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
808 struct nfsd4_conn *conn;
810 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
813 svc_xprt_get(rqstp->rq_xprt);
814 conn->cn_xprt = rqstp->rq_xprt;
815 conn->cn_flags = flags;
816 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
820 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
822 conn->cn_session = ses;
823 list_add(&conn->cn_persession, &ses->se_conns);
826 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
828 struct nfs4_client *clp = ses->se_client;
830 spin_lock(&clp->cl_lock);
831 __nfsd4_hash_conn(conn, ses);
832 spin_unlock(&clp->cl_lock);
835 static int nfsd4_register_conn(struct nfsd4_conn *conn)
837 conn->cn_xpt_user.callback = nfsd4_conn_lost;
838 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
841 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
845 nfsd4_hash_conn(conn, ses);
846 ret = nfsd4_register_conn(conn);
848 /* oops; xprt is already down: */
849 nfsd4_conn_lost(&conn->cn_xpt_user);
850 if (conn->cn_flags & NFS4_CDFC4_BACK) {
851 /* callback channel may be back up */
852 nfsd4_probe_callback(ses->se_client);
856 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
858 u32 dir = NFS4_CDFC4_FORE;
860 if (cses->flags & SESSION4_BACK_CHAN)
861 dir |= NFS4_CDFC4_BACK;
862 return alloc_conn(rqstp, dir);
865 /* must be called under client_lock */
866 static void nfsd4_del_conns(struct nfsd4_session *s)
868 struct nfs4_client *clp = s->se_client;
869 struct nfsd4_conn *c;
871 spin_lock(&clp->cl_lock);
872 while (!list_empty(&s->se_conns)) {
873 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
874 list_del_init(&c->cn_persession);
875 spin_unlock(&clp->cl_lock);
877 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
880 spin_lock(&clp->cl_lock);
882 spin_unlock(&clp->cl_lock);
885 static void __free_session(struct nfsd4_session *ses)
887 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
888 free_session_slots(ses);
892 static void free_session(struct kref *kref)
894 struct nfsd4_session *ses;
896 lockdep_assert_held(&client_lock);
897 ses = container_of(kref, struct nfsd4_session, se_ref);
898 nfsd4_del_conns(ses);
902 void nfsd4_put_session(struct nfsd4_session *ses)
904 spin_lock(&client_lock);
905 nfsd4_put_session_locked(ses);
906 spin_unlock(&client_lock);
909 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fchan)
911 struct nfsd4_session *new;
912 int numslots, slotsize;
914 * Note decreasing slot size below client's request may
915 * make it difficult for client to function correctly, whereas
916 * decreasing the number of slots will (just?) affect
917 * performance. When short on memory we therefore prefer to
918 * decrease number of slots instead of their size.
920 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
921 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
925 new = __alloc_session(slotsize, numslots);
927 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
930 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
934 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
938 new->se_client = clp;
941 INIT_LIST_HEAD(&new->se_conns);
943 new->se_cb_seq_nr = 1;
944 new->se_flags = cses->flags;
945 new->se_cb_prog = cses->callback_prog;
946 new->se_cb_sec = cses->cb_sec;
947 kref_init(&new->se_ref);
948 idx = hash_sessionid(&new->se_sessionid);
949 spin_lock(&client_lock);
950 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
951 spin_lock(&clp->cl_lock);
952 list_add(&new->se_perclnt, &clp->cl_sessions);
953 spin_unlock(&clp->cl_lock);
954 spin_unlock(&client_lock);
956 if (cses->flags & SESSION4_BACK_CHAN) {
957 struct sockaddr *sa = svc_addr(rqstp);
959 * This is a little silly; with sessions there's no real
960 * use for the callback address. Use the peer address
961 * as a reasonable default for now, but consider fixing
962 * the rpc client not to require an address in the
965 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
966 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
970 /* caller must hold client_lock */
971 static struct nfsd4_session *
972 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
974 struct nfsd4_session *elem;
977 dump_sessionid(__func__, sessionid);
978 idx = hash_sessionid(sessionid);
979 /* Search in the appropriate list */
980 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
981 if (!memcmp(elem->se_sessionid.data, sessionid->data,
982 NFS4_MAX_SESSIONID_LEN)) {
987 dprintk("%s: session not found\n", __func__);
991 /* caller must hold client_lock */
993 unhash_session(struct nfsd4_session *ses)
995 list_del(&ses->se_hash);
996 spin_lock(&ses->se_client->cl_lock);
997 list_del(&ses->se_perclnt);
998 spin_unlock(&ses->se_client->cl_lock);
1001 /* must be called under the client_lock */
1003 renew_client_locked(struct nfs4_client *clp)
1005 if (is_client_expired(clp)) {
1007 printk("%s: client (clientid %08x/%08x) already expired\n",
1009 clp->cl_clientid.cl_boot,
1010 clp->cl_clientid.cl_id);
1014 dprintk("renewing client (clientid %08x/%08x)\n",
1015 clp->cl_clientid.cl_boot,
1016 clp->cl_clientid.cl_id);
1017 list_move_tail(&clp->cl_lru, &client_lru);
1018 clp->cl_time = get_seconds();
1022 renew_client(struct nfs4_client *clp)
1024 spin_lock(&client_lock);
1025 renew_client_locked(clp);
1026 spin_unlock(&client_lock);
1029 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1031 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1033 if (clid->cl_boot == nn->boot_time)
1035 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1036 clid->cl_boot, clid->cl_id, nn->boot_time);
1041 * XXX Should we use a slab cache ?
1042 * This type of memory management is somewhat inefficient, but we use it
1043 * anyway since SETCLIENTID is not a common operation.
1045 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1047 struct nfs4_client *clp;
1049 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1052 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1053 if (clp->cl_name.data == NULL) {
1057 clp->cl_name.len = name.len;
1062 free_client(struct nfs4_client *clp)
1064 lockdep_assert_held(&client_lock);
1065 while (!list_empty(&clp->cl_sessions)) {
1066 struct nfsd4_session *ses;
1067 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1069 list_del(&ses->se_perclnt);
1070 nfsd4_put_session_locked(ses);
1072 free_svc_cred(&clp->cl_cred);
1073 kfree(clp->cl_name.data);
1078 release_session_client(struct nfsd4_session *session)
1080 struct nfs4_client *clp = session->se_client;
1082 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1084 if (is_client_expired(clp)) {
1086 session->se_client = NULL;
1088 renew_client_locked(clp);
1089 spin_unlock(&client_lock);
1092 /* must be called under the client_lock */
1094 unhash_client_locked(struct nfs4_client *clp)
1096 struct nfsd4_session *ses;
1098 mark_client_expired(clp);
1099 list_del(&clp->cl_lru);
1100 spin_lock(&clp->cl_lock);
1101 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1102 list_del_init(&ses->se_hash);
1103 spin_unlock(&clp->cl_lock);
1107 destroy_client(struct nfs4_client *clp)
1109 struct nfs4_openowner *oo;
1110 struct nfs4_delegation *dp;
1111 struct list_head reaplist;
1112 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1114 INIT_LIST_HEAD(&reaplist);
1115 spin_lock(&recall_lock);
1116 while (!list_empty(&clp->cl_delegations)) {
1117 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1118 list_del_init(&dp->dl_perclnt);
1119 list_move(&dp->dl_recall_lru, &reaplist);
1121 spin_unlock(&recall_lock);
1122 while (!list_empty(&reaplist)) {
1123 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1124 unhash_delegation(dp);
1126 while (!list_empty(&clp->cl_openowners)) {
1127 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1128 release_openowner(oo);
1130 nfsd4_shutdown_callback(clp);
1131 if (clp->cl_cb_conn.cb_xprt)
1132 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1133 list_del(&clp->cl_idhash);
1134 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1135 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1137 rb_erase(&clp->cl_namenode, &unconf_name_tree);
1138 spin_lock(&client_lock);
1139 unhash_client_locked(clp);
1140 if (atomic_read(&clp->cl_refcount) == 0)
1142 spin_unlock(&client_lock);
1145 static void expire_client(struct nfs4_client *clp)
1147 nfsd4_client_record_remove(clp);
1148 destroy_client(clp);
1151 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1153 memcpy(target->cl_verifier.data, source->data,
1154 sizeof(target->cl_verifier.data));
1157 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1159 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1160 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1163 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1165 if (source->cr_principal) {
1166 target->cr_principal =
1167 kstrdup(source->cr_principal, GFP_KERNEL);
1168 if (target->cr_principal == NULL)
1171 target->cr_principal = NULL;
1172 target->cr_flavor = source->cr_flavor;
1173 target->cr_uid = source->cr_uid;
1174 target->cr_gid = source->cr_gid;
1175 target->cr_group_info = source->cr_group_info;
1176 get_group_info(target->cr_group_info);
1181 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1185 res = o1->len - o2->len;
1188 return (long long)memcmp(o1->data, o2->data, o1->len);
1191 static int same_name(const char *n1, const char *n2)
1193 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1197 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1199 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1203 same_clid(clientid_t *cl1, clientid_t *cl2)
1205 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1208 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1212 if (g1->ngroups != g2->ngroups)
1214 for (i=0; i<g1->ngroups; i++)
1215 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1221 * RFC 3530 language requires clid_inuse be returned when the
1222 * "principal" associated with a requests differs from that previously
1223 * used. We use uid, gid's, and gss principal string as our best
1224 * approximation. We also don't want to allow non-gss use of a client
1225 * established using gss: in theory cr_principal should catch that
1226 * change, but in practice cr_principal can be null even in the gss case
1227 * since gssd doesn't always pass down a principal string.
1229 static bool is_gss_cred(struct svc_cred *cr)
1231 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1232 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1237 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1239 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1240 || (cr1->cr_uid != cr2->cr_uid)
1241 || (cr1->cr_gid != cr2->cr_gid)
1242 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1244 if (cr1->cr_principal == cr2->cr_principal)
1246 if (!cr1->cr_principal || !cr2->cr_principal)
1248 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1251 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1253 static u32 current_clientid = 1;
1255 clp->cl_clientid.cl_boot = nn->boot_time;
1256 clp->cl_clientid.cl_id = current_clientid++;
1259 static void gen_confirm(struct nfs4_client *clp)
1264 verf[0] = (__be32)get_seconds();
1265 verf[1] = (__be32)i++;
1266 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1269 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1271 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1274 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1276 struct nfs4_stid *s;
1278 s = find_stateid(cl, t);
1281 if (typemask & s->sc_type)
1286 static struct nfs4_client *create_client(struct xdr_netobj name,
1287 struct svc_rqst *rqstp, nfs4_verifier *verf)
1289 struct nfs4_client *clp;
1290 struct sockaddr *sa = svc_addr(rqstp);
1292 struct net *net = SVC_NET(rqstp);
1294 clp = alloc_client(name);
1298 INIT_LIST_HEAD(&clp->cl_sessions);
1299 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1301 spin_lock(&client_lock);
1303 spin_unlock(&client_lock);
1306 idr_init(&clp->cl_stateids);
1307 atomic_set(&clp->cl_refcount, 0);
1308 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1309 INIT_LIST_HEAD(&clp->cl_idhash);
1310 INIT_LIST_HEAD(&clp->cl_openowners);
1311 INIT_LIST_HEAD(&clp->cl_delegations);
1312 INIT_LIST_HEAD(&clp->cl_lru);
1313 INIT_LIST_HEAD(&clp->cl_callbacks);
1314 spin_lock_init(&clp->cl_lock);
1315 nfsd4_init_callback(&clp->cl_cb_null);
1316 clp->cl_time = get_seconds();
1317 clear_bit(0, &clp->cl_cb_slot_busy);
1318 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1319 copy_verf(clp, verf);
1320 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1322 clp->cl_cb_session = NULL;
1328 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1330 struct rb_node **new = &(root->rb_node), *parent = NULL;
1331 struct nfs4_client *clp;
1334 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1337 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1338 new = &((*new)->rb_left);
1340 new = &((*new)->rb_right);
1343 rb_link_node(&new_clp->cl_namenode, parent, new);
1344 rb_insert_color(&new_clp->cl_namenode, root);
1347 static struct nfs4_client *
1348 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1351 struct rb_node *node = root->rb_node;
1352 struct nfs4_client *clp;
1355 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1356 cmp = compare_blob(&clp->cl_name, name);
1358 node = node->rb_left;
1360 node = node->rb_right;
1368 add_to_unconfirmed(struct nfs4_client *clp)
1370 unsigned int idhashval;
1371 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1373 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1374 add_clp_to_name_tree(clp, &unconf_name_tree);
1375 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1376 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1381 move_to_confirmed(struct nfs4_client *clp)
1383 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1384 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1386 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1387 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1388 rb_erase(&clp->cl_namenode, &unconf_name_tree);
1389 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1390 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1394 static struct nfs4_client *
1395 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1397 struct nfs4_client *clp;
1398 unsigned int idhashval = clientid_hashval(clid->cl_id);
1400 list_for_each_entry(clp, &nn->conf_id_hashtbl[idhashval], cl_idhash) {
1401 if (same_clid(&clp->cl_clientid, clid)) {
1402 if ((bool)clp->cl_minorversion != sessions)
1411 static struct nfs4_client *
1412 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1414 struct nfs4_client *clp;
1415 unsigned int idhashval = clientid_hashval(clid->cl_id);
1417 list_for_each_entry(clp, &nn->unconf_id_hashtbl[idhashval], cl_idhash) {
1418 if (same_clid(&clp->cl_clientid, clid)) {
1419 if ((bool)clp->cl_minorversion != sessions)
1427 static bool clp_used_exchangeid(struct nfs4_client *clp)
1429 return clp->cl_exchange_flags != 0;
1432 static struct nfs4_client *
1433 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1435 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1438 static struct nfs4_client *
1439 find_unconfirmed_client_by_name(struct xdr_netobj *name)
1441 return find_clp_in_name_tree(name, &unconf_name_tree);
1445 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1447 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1448 struct sockaddr *sa = svc_addr(rqstp);
1449 u32 scopeid = rpc_get_scope_id(sa);
1450 unsigned short expected_family;
1452 /* Currently, we only support tcp and tcp6 for the callback channel */
1453 if (se->se_callback_netid_len == 3 &&
1454 !memcmp(se->se_callback_netid_val, "tcp", 3))
1455 expected_family = AF_INET;
1456 else if (se->se_callback_netid_len == 4 &&
1457 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1458 expected_family = AF_INET6;
1462 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1463 se->se_callback_addr_len,
1464 (struct sockaddr *)&conn->cb_addr,
1465 sizeof(conn->cb_addr));
1467 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1470 if (conn->cb_addr.ss_family == AF_INET6)
1471 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1473 conn->cb_prog = se->se_callback_prog;
1474 conn->cb_ident = se->se_callback_ident;
1475 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1478 conn->cb_addr.ss_family = AF_UNSPEC;
1479 conn->cb_addrlen = 0;
1480 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1481 "will not receive delegations\n",
1482 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1488 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1491 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1493 struct nfsd4_slot *slot = resp->cstate.slot;
1496 dprintk("--> %s slot %p\n", __func__, slot);
1498 slot->sl_opcnt = resp->opcnt;
1499 slot->sl_status = resp->cstate.status;
1501 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1502 if (nfsd4_not_cached(resp)) {
1503 slot->sl_datalen = 0;
1506 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1507 base = (char *)resp->cstate.datap -
1508 (char *)resp->xbuf->head[0].iov_base;
1509 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1511 WARN("%s: sessions DRC could not cache compound\n", __func__);
1516 * Encode the replay sequence operation from the slot values.
1517 * If cachethis is FALSE encode the uncached rep error on the next
1518 * operation which sets resp->p and increments resp->opcnt for
1519 * nfs4svc_encode_compoundres.
1523 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1524 struct nfsd4_compoundres *resp)
1526 struct nfsd4_op *op;
1527 struct nfsd4_slot *slot = resp->cstate.slot;
1529 /* Encode the replayed sequence operation */
1530 op = &args->ops[resp->opcnt - 1];
1531 nfsd4_encode_operation(resp, op);
1533 /* Return nfserr_retry_uncached_rep in next operation. */
1534 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1535 op = &args->ops[resp->opcnt++];
1536 op->status = nfserr_retry_uncached_rep;
1537 nfsd4_encode_operation(resp, op);
1543 * The sequence operation is not cached because we can use the slot and
1547 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1548 struct nfsd4_sequence *seq)
1550 struct nfsd4_slot *slot = resp->cstate.slot;
1553 dprintk("--> %s slot %p\n", __func__, slot);
1555 /* Either returns 0 or nfserr_retry_uncached */
1556 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1557 if (status == nfserr_retry_uncached_rep)
1560 /* The sequence operation has been encoded, cstate->datap set. */
1561 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1563 resp->opcnt = slot->sl_opcnt;
1564 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1565 status = slot->sl_status;
1571 * Set the exchange_id flags returned by the server.
1574 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1576 /* pNFS is not supported */
1577 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1579 /* Referrals are supported, Migration is not. */
1580 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1582 /* set the wire flags to return to client. */
1583 clid->flags = new->cl_exchange_flags;
1586 static bool client_has_state(struct nfs4_client *clp)
1589 * Note clp->cl_openowners check isn't quite right: there's no
1590 * need to count owners without stateid's.
1592 * Also note we should probably be using this in 4.0 case too.
1594 return !list_empty(&clp->cl_openowners)
1595 || !list_empty(&clp->cl_delegations)
1596 || !list_empty(&clp->cl_sessions);
1600 nfsd4_exchange_id(struct svc_rqst *rqstp,
1601 struct nfsd4_compound_state *cstate,
1602 struct nfsd4_exchange_id *exid)
1604 struct nfs4_client *unconf, *conf, *new;
1606 char addr_str[INET6_ADDRSTRLEN];
1607 nfs4_verifier verf = exid->verifier;
1608 struct sockaddr *sa = svc_addr(rqstp);
1609 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1610 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1612 rpc_ntop(sa, addr_str, sizeof(addr_str));
1613 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1614 "ip_addr=%s flags %x, spa_how %d\n",
1615 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1616 addr_str, exid->flags, exid->spa_how);
1618 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1619 return nfserr_inval;
1621 /* Currently only support SP4_NONE */
1622 switch (exid->spa_how) {
1626 return nfserr_serverfault;
1628 BUG(); /* checked by xdr code */
1630 return nfserr_serverfault; /* no excuse :-/ */
1633 /* Cases below refer to rfc 5661 section 18.35.4: */
1635 conf = find_confirmed_client_by_name(&exid->clname, nn);
1637 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1638 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1641 if (!clp_used_exchangeid(conf)) { /* buggy client */
1642 status = nfserr_inval;
1645 if (!creds_match) { /* case 9 */
1646 status = nfserr_perm;
1649 if (!verfs_match) { /* case 8 */
1650 status = nfserr_not_same;
1654 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1658 if (!creds_match) { /* case 3 */
1659 if (client_has_state(conf)) {
1660 status = nfserr_clid_inuse;
1663 expire_client(conf);
1666 if (verfs_match) { /* case 2 */
1667 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1671 /* case 5, client reboot */
1675 if (update) { /* case 7 */
1676 status = nfserr_noent;
1680 unconf = find_unconfirmed_client_by_name(&exid->clname);
1681 if (unconf) /* case 4, possible retry or client restart */
1682 expire_client(unconf);
1684 /* case 1 (normal case) */
1686 new = create_client(exid->clname, rqstp, &verf);
1688 status = nfserr_jukebox;
1691 new->cl_minorversion = 1;
1694 add_to_unconfirmed(new);
1696 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1697 exid->clientid.cl_id = new->cl_clientid.cl_id;
1699 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1700 nfsd4_set_ex_flags(new, exid);
1702 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1703 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1707 nfs4_unlock_state();
1712 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1714 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1717 /* The slot is in use, and no response has been sent. */
1719 if (seqid == slot_seqid)
1720 return nfserr_jukebox;
1722 return nfserr_seq_misordered;
1724 /* Note unsigned 32-bit arithmetic handles wraparound: */
1725 if (likely(seqid == slot_seqid + 1))
1727 if (seqid == slot_seqid)
1728 return nfserr_replay_cache;
1729 return nfserr_seq_misordered;
1733 * Cache the create session result into the create session single DRC
1734 * slot cache by saving the xdr structure. sl_seqid has been set.
1735 * Do this for solo or embedded create session operations.
1738 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1739 struct nfsd4_clid_slot *slot, __be32 nfserr)
1741 slot->sl_status = nfserr;
1742 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1746 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1747 struct nfsd4_clid_slot *slot)
1749 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1750 return slot->sl_status;
1753 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1754 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1755 1 + /* MIN tag is length with zero, only length */ \
1756 3 + /* version, opcount, opcode */ \
1757 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1758 /* seqid, slotID, slotID, cache */ \
1759 4 ) * sizeof(__be32))
1761 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1762 2 + /* verifier: AUTH_NULL, length 0 */\
1764 1 + /* MIN tag is length with zero, only length */ \
1765 3 + /* opcount, opcode, opstatus*/ \
1766 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1767 /* seqid, slotID, slotID, slotID, status */ \
1768 5 ) * sizeof(__be32))
1770 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1772 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1773 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1777 nfsd4_create_session(struct svc_rqst *rqstp,
1778 struct nfsd4_compound_state *cstate,
1779 struct nfsd4_create_session *cr_ses)
1781 struct sockaddr *sa = svc_addr(rqstp);
1782 struct nfs4_client *conf, *unconf;
1783 struct nfsd4_session *new;
1784 struct nfsd4_conn *conn;
1785 struct nfsd4_clid_slot *cs_slot = NULL;
1787 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1789 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1790 return nfserr_inval;
1791 if (check_forechannel_attrs(cr_ses->fore_channel))
1792 return nfserr_toosmall;
1793 new = alloc_session(&cr_ses->fore_channel);
1795 return nfserr_jukebox;
1796 status = nfserr_jukebox;
1797 conn = alloc_conn_from_crses(rqstp, cr_ses);
1799 goto out_free_session;
1802 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1803 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1806 cs_slot = &conf->cl_cs_slot;
1807 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1808 if (status == nfserr_replay_cache) {
1809 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1811 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1812 status = nfserr_seq_misordered;
1815 } else if (unconf) {
1816 struct nfs4_client *old;
1817 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1818 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1819 status = nfserr_clid_inuse;
1822 cs_slot = &unconf->cl_cs_slot;
1823 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1825 /* an unconfirmed replay returns misordered */
1826 status = nfserr_seq_misordered;
1829 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1832 move_to_confirmed(unconf);
1835 status = nfserr_stale_clientid;
1840 * We do not support RDMA or persistent sessions
1842 cr_ses->flags &= ~SESSION4_PERSIST;
1843 cr_ses->flags &= ~SESSION4_RDMA;
1845 init_session(rqstp, new, conf, cr_ses);
1846 nfsd4_init_conn(rqstp, conn, new);
1848 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1849 NFS4_MAX_SESSIONID_LEN);
1850 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1851 sizeof(struct nfsd4_channel_attrs));
1852 cs_slot->sl_seqid++;
1853 cr_ses->seqid = cs_slot->sl_seqid;
1855 /* cache solo and embedded create sessions under the state lock */
1856 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1858 nfs4_unlock_state();
1859 dprintk("%s returns %d\n", __func__, ntohl(status));
1864 __free_session(new);
1868 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1870 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1871 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1873 return argp->opcnt == resp->opcnt;
1876 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1879 case NFS4_CDFC4_FORE:
1880 case NFS4_CDFC4_BACK:
1882 case NFS4_CDFC4_FORE_OR_BOTH:
1883 case NFS4_CDFC4_BACK_OR_BOTH:
1884 *dir = NFS4_CDFC4_BOTH;
1887 return nfserr_inval;
1890 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1892 struct nfsd4_session *session = cstate->session;
1894 spin_lock(&client_lock);
1895 session->se_cb_prog = bc->bc_cb_program;
1896 session->se_cb_sec = bc->bc_cb_sec;
1897 spin_unlock(&client_lock);
1899 nfsd4_probe_callback(session->se_client);
1904 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1905 struct nfsd4_compound_state *cstate,
1906 struct nfsd4_bind_conn_to_session *bcts)
1909 struct nfsd4_conn *conn;
1911 if (!nfsd4_last_compound_op(rqstp))
1912 return nfserr_not_only_op;
1913 spin_lock(&client_lock);
1914 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1915 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1916 * client_lock iself: */
1917 if (cstate->session) {
1918 nfsd4_get_session(cstate->session);
1919 atomic_inc(&cstate->session->se_client->cl_refcount);
1921 spin_unlock(&client_lock);
1922 if (!cstate->session)
1923 return nfserr_badsession;
1925 status = nfsd4_map_bcts_dir(&bcts->dir);
1928 conn = alloc_conn(rqstp, bcts->dir);
1930 return nfserr_jukebox;
1931 nfsd4_init_conn(rqstp, conn, cstate->session);
1935 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1939 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1943 nfsd4_destroy_session(struct svc_rqst *r,
1944 struct nfsd4_compound_state *cstate,
1945 struct nfsd4_destroy_session *sessionid)
1947 struct nfsd4_session *ses;
1948 __be32 status = nfserr_badsession;
1951 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1952 * - Should we return nfserr_back_chan_busy if waiting for
1953 * callbacks on to-be-destroyed session?
1954 * - Do we need to clear any callback info from previous session?
1957 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1958 if (!nfsd4_last_compound_op(r))
1959 return nfserr_not_only_op;
1961 dump_sessionid(__func__, &sessionid->sessionid);
1962 spin_lock(&client_lock);
1963 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1965 spin_unlock(&client_lock);
1969 unhash_session(ses);
1970 spin_unlock(&client_lock);
1973 nfsd4_probe_callback_sync(ses->se_client);
1974 nfs4_unlock_state();
1976 spin_lock(&client_lock);
1977 nfsd4_del_conns(ses);
1978 nfsd4_put_session_locked(ses);
1979 spin_unlock(&client_lock);
1982 dprintk("%s returns %d\n", __func__, ntohl(status));
1986 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1988 struct nfsd4_conn *c;
1990 list_for_each_entry(c, &s->se_conns, cn_persession) {
1991 if (c->cn_xprt == xpt) {
1998 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2000 struct nfs4_client *clp = ses->se_client;
2001 struct nfsd4_conn *c;
2004 spin_lock(&clp->cl_lock);
2005 c = __nfsd4_find_conn(new->cn_xprt, ses);
2007 spin_unlock(&clp->cl_lock);
2011 __nfsd4_hash_conn(new, ses);
2012 spin_unlock(&clp->cl_lock);
2013 ret = nfsd4_register_conn(new);
2015 /* oops; xprt is already down: */
2016 nfsd4_conn_lost(&new->cn_xpt_user);
2020 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2022 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2024 return args->opcnt > session->se_fchannel.maxops;
2027 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2028 struct nfsd4_session *session)
2030 struct xdr_buf *xb = &rqstp->rq_arg;
2032 return xb->len > session->se_fchannel.maxreq_sz;
2036 nfsd4_sequence(struct svc_rqst *rqstp,
2037 struct nfsd4_compound_state *cstate,
2038 struct nfsd4_sequence *seq)
2040 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2041 struct nfsd4_session *session;
2042 struct nfsd4_slot *slot;
2043 struct nfsd4_conn *conn;
2046 if (resp->opcnt != 1)
2047 return nfserr_sequence_pos;
2050 * Will be either used or freed by nfsd4_sequence_check_conn
2053 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2055 return nfserr_jukebox;
2057 spin_lock(&client_lock);
2058 status = nfserr_badsession;
2059 session = find_in_sessionid_hashtbl(&seq->sessionid);
2063 status = nfserr_too_many_ops;
2064 if (nfsd4_session_too_many_ops(rqstp, session))
2067 status = nfserr_req_too_big;
2068 if (nfsd4_request_too_big(rqstp, session))
2071 status = nfserr_badslot;
2072 if (seq->slotid >= session->se_fchannel.maxreqs)
2075 slot = session->se_slots[seq->slotid];
2076 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2078 /* We do not negotiate the number of slots yet, so set the
2079 * maxslots to the session maxreqs which is used to encode
2080 * sr_highest_slotid and the sr_target_slot id to maxslots */
2081 seq->maxslots = session->se_fchannel.maxreqs;
2083 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2084 slot->sl_flags & NFSD4_SLOT_INUSE);
2085 if (status == nfserr_replay_cache) {
2086 status = nfserr_seq_misordered;
2087 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2089 cstate->slot = slot;
2090 cstate->session = session;
2091 /* Return the cached reply status and set cstate->status
2092 * for nfsd4_proc_compound processing */
2093 status = nfsd4_replay_cache_entry(resp, seq);
2094 cstate->status = nfserr_replay_cache;
2100 nfsd4_sequence_check_conn(conn, session);
2103 /* Success! bump slot seqid */
2104 slot->sl_seqid = seq->seqid;
2105 slot->sl_flags |= NFSD4_SLOT_INUSE;
2107 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2109 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2111 cstate->slot = slot;
2112 cstate->session = session;
2115 /* Hold a session reference until done processing the compound. */
2116 if (cstate->session) {
2117 struct nfs4_client *clp = session->se_client;
2119 nfsd4_get_session(cstate->session);
2120 atomic_inc(&clp->cl_refcount);
2121 switch (clp->cl_cb_state) {
2123 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2125 case NFSD4_CB_FAULT:
2126 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2129 seq->status_flags = 0;
2133 spin_unlock(&client_lock);
2134 dprintk("%s: return %d\n", __func__, ntohl(status));
2139 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2141 struct nfs4_client *conf, *unconf, *clp;
2143 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2146 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2147 conf = find_confirmed_client(&dc->clientid, true, nn);
2152 if (!is_client_expired(conf) && client_has_state(conf)) {
2153 status = nfserr_clientid_busy;
2157 /* rfc5661 18.50.3 */
2158 if (cstate->session && conf == cstate->session->se_client) {
2159 status = nfserr_clientid_busy;
2165 status = nfserr_stale_clientid;
2171 nfs4_unlock_state();
2172 dprintk("%s return %d\n", __func__, ntohl(status));
2177 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2181 if (rc->rca_one_fs) {
2182 if (!cstate->current_fh.fh_dentry)
2183 return nfserr_nofilehandle;
2185 * We don't take advantage of the rca_one_fs case.
2186 * That's OK, it's optional, we can safely ignore it.
2192 status = nfserr_complete_already;
2193 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2194 &cstate->session->se_client->cl_flags))
2197 status = nfserr_stale_clientid;
2198 if (is_client_expired(cstate->session->se_client))
2200 * The following error isn't really legal.
2201 * But we only get here if the client just explicitly
2202 * destroyed the client. Surely it no longer cares what
2203 * error it gets back on an operation for the dead
2209 nfsd4_client_record_create(cstate->session->se_client);
2211 nfs4_unlock_state();
2216 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2217 struct nfsd4_setclientid *setclid)
2219 struct xdr_netobj clname = setclid->se_name;
2220 nfs4_verifier clverifier = setclid->se_verf;
2221 struct nfs4_client *conf, *unconf, *new;
2223 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2225 /* Cases below refer to rfc 3530 section 14.2.33: */
2227 conf = find_confirmed_client_by_name(&clname, nn);
2230 status = nfserr_clid_inuse;
2231 if (clp_used_exchangeid(conf))
2233 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2234 char addr_str[INET6_ADDRSTRLEN];
2235 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2237 dprintk("NFSD: setclientid: string in use by client "
2238 "at %s\n", addr_str);
2242 unconf = find_unconfirmed_client_by_name(&clname);
2244 expire_client(unconf);
2245 status = nfserr_jukebox;
2246 new = create_client(clname, rqstp, &clverifier);
2249 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2250 /* case 1: probable callback update */
2251 copy_clid(new, conf);
2252 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2254 new->cl_minorversion = 0;
2255 gen_callback(new, setclid, rqstp);
2256 add_to_unconfirmed(new);
2257 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2258 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2259 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2262 nfs4_unlock_state();
2268 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2269 struct nfsd4_compound_state *cstate,
2270 struct nfsd4_setclientid_confirm *setclientid_confirm)
2272 struct nfs4_client *conf, *unconf;
2273 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2274 clientid_t * clid = &setclientid_confirm->sc_clientid;
2276 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2278 if (STALE_CLIENTID(clid, nn))
2279 return nfserr_stale_clientid;
2282 conf = find_confirmed_client(clid, false, nn);
2283 unconf = find_unconfirmed_client(clid, false, nn);
2285 * We try hard to give out unique clientid's, so if we get an
2286 * attempt to confirm the same clientid with a different cred,
2287 * there's a bug somewhere. Let's charitably assume it's our
2290 status = nfserr_serverfault;
2291 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2293 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2295 /* cases below refer to rfc 3530 section 14.2.34: */
2296 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2297 if (conf && !unconf) /* case 2: probable retransmit */
2299 else /* case 4: client hasn't noticed we rebooted yet? */
2300 status = nfserr_stale_clientid;
2304 if (conf) { /* case 1: callback update */
2305 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2306 nfsd4_probe_callback(conf);
2307 expire_client(unconf);
2308 } else { /* case 3: normal case; new or rebooted client */
2309 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2311 expire_client(conf);
2312 move_to_confirmed(unconf);
2313 nfsd4_probe_callback(unconf);
2316 nfs4_unlock_state();
2320 static struct nfs4_file *nfsd4_alloc_file(void)
2322 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2325 /* OPEN Share state helper functions */
2326 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2328 unsigned int hashval = file_hashval(ino);
2330 atomic_set(&fp->fi_ref, 1);
2331 INIT_LIST_HEAD(&fp->fi_hash);
2332 INIT_LIST_HEAD(&fp->fi_stateids);
2333 INIT_LIST_HEAD(&fp->fi_delegations);
2334 fp->fi_inode = igrab(ino);
2335 fp->fi_had_conflict = false;
2336 fp->fi_lease = NULL;
2337 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2338 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2339 spin_lock(&recall_lock);
2340 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2341 spin_unlock(&recall_lock);
2345 nfsd4_free_slab(struct kmem_cache **slab)
2349 kmem_cache_destroy(*slab);
2354 nfsd4_free_slabs(void)
2356 nfsd4_free_slab(&openowner_slab);
2357 nfsd4_free_slab(&lockowner_slab);
2358 nfsd4_free_slab(&file_slab);
2359 nfsd4_free_slab(&stateid_slab);
2360 nfsd4_free_slab(&deleg_slab);
2364 nfsd4_init_slabs(void)
2366 openowner_slab = kmem_cache_create("nfsd4_openowners",
2367 sizeof(struct nfs4_openowner), 0, 0, NULL);
2368 if (openowner_slab == NULL)
2370 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2371 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2372 if (lockowner_slab == NULL)
2374 file_slab = kmem_cache_create("nfsd4_files",
2375 sizeof(struct nfs4_file), 0, 0, NULL);
2376 if (file_slab == NULL)
2378 stateid_slab = kmem_cache_create("nfsd4_stateids",
2379 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2380 if (stateid_slab == NULL)
2382 deleg_slab = kmem_cache_create("nfsd4_delegations",
2383 sizeof(struct nfs4_delegation), 0, 0, NULL);
2384 if (deleg_slab == NULL)
2389 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2393 void nfs4_free_openowner(struct nfs4_openowner *oo)
2395 kfree(oo->oo_owner.so_owner.data);
2396 kmem_cache_free(openowner_slab, oo);
2399 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2401 kfree(lo->lo_owner.so_owner.data);
2402 kmem_cache_free(lockowner_slab, lo);
2405 static void init_nfs4_replay(struct nfs4_replay *rp)
2407 rp->rp_status = nfserr_serverfault;
2409 rp->rp_buf = rp->rp_ibuf;
2412 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2414 struct nfs4_stateowner *sop;
2416 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2420 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2421 if (!sop->so_owner.data) {
2422 kmem_cache_free(slab, sop);
2425 sop->so_owner.len = owner->len;
2427 INIT_LIST_HEAD(&sop->so_stateids);
2428 sop->so_client = clp;
2429 init_nfs4_replay(&sop->so_replay);
2433 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2435 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2436 list_add(&oo->oo_perclient, &clp->cl_openowners);
2439 static struct nfs4_openowner *
2440 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2441 struct nfs4_openowner *oo;
2443 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2446 oo->oo_owner.so_is_open_owner = 1;
2447 oo->oo_owner.so_seqid = open->op_seqid;
2448 oo->oo_flags = NFS4_OO_NEW;
2450 oo->oo_last_closed_stid = NULL;
2451 INIT_LIST_HEAD(&oo->oo_close_lru);
2452 hash_openowner(oo, clp, strhashval);
2456 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2457 struct nfs4_openowner *oo = open->op_openowner;
2458 struct nfs4_client *clp = oo->oo_owner.so_client;
2460 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2461 INIT_LIST_HEAD(&stp->st_lockowners);
2462 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2463 list_add(&stp->st_perfile, &fp->fi_stateids);
2464 stp->st_stateowner = &oo->oo_owner;
2467 stp->st_access_bmap = 0;
2468 stp->st_deny_bmap = 0;
2469 set_access(open->op_share_access, stp);
2470 set_deny(open->op_share_deny, stp);
2471 stp->st_openstp = NULL;
2475 move_to_close_lru(struct nfs4_openowner *oo)
2477 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2479 list_move_tail(&oo->oo_close_lru, &close_lru);
2480 oo->oo_time = get_seconds();
2484 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2487 return (sop->so_owner.len == owner->len) &&
2488 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2489 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2492 static struct nfs4_openowner *
2493 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open, bool sessions)
2495 struct nfs4_stateowner *so;
2496 struct nfs4_openowner *oo;
2497 struct nfs4_client *clp;
2499 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2500 if (!so->so_is_open_owner)
2502 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2504 clp = oo->oo_owner.so_client;
2505 if ((bool)clp->cl_minorversion != sessions)
2507 renew_client(oo->oo_owner.so_client);
2514 /* search file_hashtbl[] for file */
2515 static struct nfs4_file *
2516 find_file(struct inode *ino)
2518 unsigned int hashval = file_hashval(ino);
2519 struct nfs4_file *fp;
2521 spin_lock(&recall_lock);
2522 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2523 if (fp->fi_inode == ino) {
2525 spin_unlock(&recall_lock);
2529 spin_unlock(&recall_lock);
2534 * Called to check deny when READ with all zero stateid or
2535 * WRITE with all zero or all one stateid
2538 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2540 struct inode *ino = current_fh->fh_dentry->d_inode;
2541 struct nfs4_file *fp;
2542 struct nfs4_ol_stateid *stp;
2545 dprintk("NFSD: nfs4_share_conflict\n");
2547 fp = find_file(ino);
2550 ret = nfserr_locked;
2551 /* Search for conflicting share reservations */
2552 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2553 if (test_deny(deny_type, stp) ||
2554 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2563 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2565 /* We're assuming the state code never drops its reference
2566 * without first removing the lease. Since we're in this lease
2567 * callback (and since the lease code is serialized by the kernel
2568 * lock) we know the server hasn't removed the lease yet, we know
2569 * it's safe to take a reference: */
2570 atomic_inc(&dp->dl_count);
2572 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2574 /* only place dl_time is set. protected by lock_flocks*/
2575 dp->dl_time = get_seconds();
2577 nfsd4_cb_recall(dp);
2580 /* Called from break_lease() with lock_flocks() held. */
2581 static void nfsd_break_deleg_cb(struct file_lock *fl)
2583 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2584 struct nfs4_delegation *dp;
2587 WARN(1, "(%p)->fl_owner NULL\n", fl);
2590 if (fp->fi_had_conflict) {
2591 WARN(1, "duplicate break on %p\n", fp);
2595 * We don't want the locks code to timeout the lease for us;
2596 * we'll remove it ourself if a delegation isn't returned
2599 fl->fl_break_time = 0;
2601 spin_lock(&recall_lock);
2602 fp->fi_had_conflict = true;
2603 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2604 nfsd_break_one_deleg(dp);
2605 spin_unlock(&recall_lock);
2609 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2612 return lease_modify(onlist, arg);
2617 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2618 .lm_break = nfsd_break_deleg_cb,
2619 .lm_change = nfsd_change_deleg_cb,
2622 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2624 if (nfsd4_has_session(cstate))
2626 if (seqid == so->so_seqid - 1)
2627 return nfserr_replay_me;
2628 if (seqid == so->so_seqid)
2630 return nfserr_bad_seqid;
2634 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2635 struct nfsd4_open *open)
2637 clientid_t *clientid = &open->op_clientid;
2638 struct nfs4_client *clp = NULL;
2639 unsigned int strhashval;
2640 struct nfs4_openowner *oo = NULL;
2642 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2644 if (STALE_CLIENTID(&open->op_clientid, nn))
2645 return nfserr_stale_clientid;
2647 * In case we need it later, after we've already created the
2648 * file and don't want to risk a further failure:
2650 open->op_file = nfsd4_alloc_file();
2651 if (open->op_file == NULL)
2652 return nfserr_jukebox;
2654 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2655 oo = find_openstateowner_str(strhashval, open, cstate->minorversion);
2656 open->op_openowner = oo;
2658 clp = find_confirmed_client(clientid, cstate->minorversion,
2661 return nfserr_expired;
2664 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2665 /* Replace unconfirmed owners without checking for replay. */
2666 clp = oo->oo_owner.so_client;
2667 release_openowner(oo);
2668 open->op_openowner = NULL;
2671 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2674 clp = oo->oo_owner.so_client;
2677 oo = alloc_init_open_stateowner(strhashval, clp, open);
2679 return nfserr_jukebox;
2680 open->op_openowner = oo;
2682 open->op_stp = nfs4_alloc_stateid(clp);
2684 return nfserr_jukebox;
2688 static inline __be32
2689 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2691 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2692 return nfserr_openmode;
2697 static int share_access_to_flags(u32 share_access)
2699 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2702 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2704 struct nfs4_stid *ret;
2706 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2709 return delegstateid(ret);
2712 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2714 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2715 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2719 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2720 struct nfs4_delegation **dp)
2723 __be32 status = nfserr_bad_stateid;
2725 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2728 flags = share_access_to_flags(open->op_share_access);
2729 status = nfs4_check_delegmode(*dp, flags);
2733 if (!nfsd4_is_deleg_cur(open))
2737 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2742 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2744 struct nfs4_ol_stateid *local;
2745 struct nfs4_openowner *oo = open->op_openowner;
2747 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2748 /* ignore lock owners */
2749 if (local->st_stateowner->so_is_open_owner == 0)
2751 /* remember if we have seen this open owner */
2752 if (local->st_stateowner == &oo->oo_owner)
2754 /* check for conflicting share reservations */
2755 if (!test_share(local, open))
2756 return nfserr_share_denied;
2761 static inline int nfs4_access_to_access(u32 nfs4_access)
2765 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2766 flags |= NFSD_MAY_READ;
2767 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2768 flags |= NFSD_MAY_WRITE;
2772 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2773 struct svc_fh *cur_fh, struct nfsd4_open *open)
2776 int oflag = nfs4_access_to_omode(open->op_share_access);
2777 int access = nfs4_access_to_access(open->op_share_access);
2779 if (!fp->fi_fds[oflag]) {
2780 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2781 &fp->fi_fds[oflag]);
2785 nfs4_file_get_access(fp, oflag);
2790 static inline __be32
2791 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2792 struct nfsd4_open *open)
2794 struct iattr iattr = {
2795 .ia_valid = ATTR_SIZE,
2798 if (!open->op_truncate)
2800 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2801 return nfserr_inval;
2802 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2806 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)
2808 u32 op_share_access = open->op_share_access;
2812 new_access = !test_access(op_share_access, stp);
2814 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2818 status = nfsd4_truncate(rqstp, cur_fh, open);
2821 int oflag = nfs4_access_to_omode(op_share_access);
2822 nfs4_file_put_access(fp, oflag);
2826 /* remember the open */
2827 set_access(op_share_access, stp);
2828 set_deny(open->op_share_deny, stp);
2835 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2837 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2840 /* Should we give out recallable state?: */
2841 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2843 if (clp->cl_cb_state == NFSD4_CB_UP)
2846 * In the sessions case, since we don't have to establish a
2847 * separate connection for callbacks, we assume it's OK
2848 * until we hear otherwise:
2850 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2853 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2855 struct file_lock *fl;
2857 fl = locks_alloc_lock();
2860 locks_init_lock(fl);
2861 fl->fl_lmops = &nfsd_lease_mng_ops;
2862 fl->fl_flags = FL_LEASE;
2863 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2864 fl->fl_end = OFFSET_MAX;
2865 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2866 fl->fl_pid = current->tgid;
2870 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2872 struct nfs4_file *fp = dp->dl_file;
2873 struct file_lock *fl;
2876 fl = nfs4_alloc_init_lease(dp, flag);
2879 fl->fl_file = find_readable_file(fp);
2880 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2881 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2883 list_del_init(&dp->dl_perclnt);
2884 locks_free_lock(fl);
2888 fp->fi_deleg_file = get_file(fl->fl_file);
2889 atomic_set(&fp->fi_delegees, 1);
2890 list_add(&dp->dl_perfile, &fp->fi_delegations);
2894 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2896 struct nfs4_file *fp = dp->dl_file;
2899 return nfs4_setlease(dp, flag);
2900 spin_lock(&recall_lock);
2901 if (fp->fi_had_conflict) {
2902 spin_unlock(&recall_lock);
2905 atomic_inc(&fp->fi_delegees);
2906 list_add(&dp->dl_perfile, &fp->fi_delegations);
2907 spin_unlock(&recall_lock);
2908 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2912 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2914 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2915 if (status == -EAGAIN)
2916 open->op_why_no_deleg = WND4_CONTENTION;
2918 open->op_why_no_deleg = WND4_RESOURCE;
2919 switch (open->op_deleg_want) {
2920 case NFS4_SHARE_WANT_READ_DELEG:
2921 case NFS4_SHARE_WANT_WRITE_DELEG:
2922 case NFS4_SHARE_WANT_ANY_DELEG:
2924 case NFS4_SHARE_WANT_CANCEL:
2925 open->op_why_no_deleg = WND4_CANCELLED;
2927 case NFS4_SHARE_WANT_NO_DELEG:
2928 BUG(); /* not supposed to get here */
2934 * Attempt to hand out a delegation.
2937 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2938 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2940 struct nfs4_delegation *dp;
2941 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2943 int status = 0, flag = 0;
2945 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2946 flag = NFS4_OPEN_DELEGATE_NONE;
2947 open->op_recall = 0;
2948 switch (open->op_claim_type) {
2949 case NFS4_OPEN_CLAIM_PREVIOUS:
2951 open->op_recall = 1;
2952 flag = open->op_delegate_type;
2953 if (flag == NFS4_OPEN_DELEGATE_NONE)
2956 case NFS4_OPEN_CLAIM_NULL:
2957 /* Let's not give out any delegations till everyone's
2958 * had the chance to reclaim theirs.... */
2959 if (locks_in_grace(net))
2961 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2963 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2964 flag = NFS4_OPEN_DELEGATE_WRITE;
2966 flag = NFS4_OPEN_DELEGATE_READ;
2972 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2975 status = nfs4_set_delegation(dp, flag);
2979 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2981 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2982 STATEID_VAL(&dp->dl_stid.sc_stateid));
2984 open->op_delegate_type = flag;
2985 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2986 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2987 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2988 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2990 /* 4.1 client asking for a delegation? */
2991 if (open->op_deleg_want)
2992 nfsd4_open_deleg_none_ext(open, status);
2996 nfs4_put_delegation(dp);
2998 flag = NFS4_OPEN_DELEGATE_NONE;
3002 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3003 struct nfs4_delegation *dp)
3005 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3006 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3007 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3008 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3009 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3010 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3011 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3012 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3014 /* Otherwise the client must be confused wanting a delegation
3015 * it already has, therefore we don't return
3016 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3021 * called with nfs4_lock_state() held.
3024 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3026 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3027 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3028 struct nfs4_file *fp = NULL;
3029 struct inode *ino = current_fh->fh_dentry->d_inode;
3030 struct nfs4_ol_stateid *stp = NULL;
3031 struct nfs4_delegation *dp = NULL;
3035 * Lookup file; if found, lookup stateid and check open request,
3036 * and check for delegations in the process of being recalled.
3037 * If not found, create the nfs4_file struct
3039 fp = find_file(ino);
3041 if ((status = nfs4_check_open(fp, open, &stp)))
3043 status = nfs4_check_deleg(cl, fp, open, &dp);
3047 status = nfserr_bad_stateid;
3048 if (nfsd4_is_deleg_cur(open))
3050 status = nfserr_jukebox;
3052 open->op_file = NULL;
3053 nfsd4_init_file(fp, ino);
3057 * OPEN the file, or upgrade an existing OPEN.
3058 * If truncate fails, the OPEN fails.
3061 /* Stateid was found, this is an OPEN upgrade */
3062 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3066 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3069 status = nfsd4_truncate(rqstp, current_fh, open);
3073 open->op_stp = NULL;
3074 init_open_stateid(stp, fp, open);
3076 update_stateid(&stp->st_stid.sc_stateid);
3077 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3079 if (nfsd4_has_session(&resp->cstate)) {
3080 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3082 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3083 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3084 open->op_why_no_deleg = WND4_NOT_WANTED;
3090 * Attempt to hand out a delegation. No error return, because the
3091 * OPEN succeeds even if we fail.
3093 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3097 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3098 STATEID_VAL(&stp->st_stid.sc_stateid));
3100 /* 4.1 client trying to upgrade/downgrade delegation? */
3101 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3102 open->op_deleg_want)
3103 nfsd4_deleg_xgrade_none_ext(open, dp);
3107 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3108 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3110 * To finish the open response, we just need to set the rflags.
3112 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3113 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3114 !nfsd4_has_session(&resp->cstate))
3115 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3120 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3122 if (open->op_openowner) {
3123 struct nfs4_openowner *oo = open->op_openowner;
3125 if (!list_empty(&oo->oo_owner.so_stateids))
3126 list_del_init(&oo->oo_close_lru);
3127 if (oo->oo_flags & NFS4_OO_NEW) {
3129 release_openowner(oo);
3130 open->op_openowner = NULL;
3132 oo->oo_flags &= ~NFS4_OO_NEW;
3136 nfsd4_free_file(open->op_file);
3138 free_generic_stateid(open->op_stp);
3142 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3145 struct nfs4_client *clp;
3147 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3150 dprintk("process_renew(%08x/%08x): starting\n",
3151 clid->cl_boot, clid->cl_id);
3152 status = nfserr_stale_clientid;
3153 if (STALE_CLIENTID(clid, nn))
3155 clp = find_confirmed_client(clid, cstate->minorversion, nn);
3156 status = nfserr_expired;
3158 /* We assume the client took too long to RENEW. */
3159 dprintk("nfsd4_renew: clientid not found!\n");
3162 status = nfserr_cb_path_down;
3163 if (!list_empty(&clp->cl_delegations)
3164 && clp->cl_cb_state != NFSD4_CB_UP)
3168 nfs4_unlock_state();
3173 nfsd4_end_grace(struct net *net)
3175 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3177 /* do nothing if grace period already ended */
3178 if (nn->grace_ended)
3181 dprintk("NFSD: end of grace period\n");
3182 nn->grace_ended = true;
3183 nfsd4_record_grace_done(net, nn->boot_time);
3184 locks_end_grace(&nn->nfsd4_manager);
3186 * Now that every NFSv4 client has had the chance to recover and
3187 * to see the (possibly new, possibly shorter) lease time, we
3188 * can safely set the next grace time to the current lease time:
3190 nfsd4_grace = nfsd4_lease;
3194 nfs4_laundromat(void)
3196 struct nfs4_client *clp;
3197 struct nfs4_openowner *oo;
3198 struct nfs4_delegation *dp;
3199 struct list_head *pos, *next, reaplist;
3200 time_t cutoff = get_seconds() - nfsd4_lease;
3201 time_t t, clientid_val = nfsd4_lease;
3202 time_t u, test_val = nfsd4_lease;
3206 dprintk("NFSD: laundromat service - starting\n");
3207 nfsd4_end_grace(&init_net);
3208 INIT_LIST_HEAD(&reaplist);
3209 spin_lock(&client_lock);
3210 list_for_each_safe(pos, next, &client_lru) {
3211 clp = list_entry(pos, struct nfs4_client, cl_lru);
3212 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3213 t = clp->cl_time - cutoff;
3214 if (clientid_val > t)
3218 if (atomic_read(&clp->cl_refcount)) {
3219 dprintk("NFSD: client in use (clientid %08x)\n",
3220 clp->cl_clientid.cl_id);
3223 unhash_client_locked(clp);
3224 list_add(&clp->cl_lru, &reaplist);
3226 spin_unlock(&client_lock);
3227 list_for_each_safe(pos, next, &reaplist) {
3228 clp = list_entry(pos, struct nfs4_client, cl_lru);
3229 dprintk("NFSD: purging unused client (clientid %08x)\n",
3230 clp->cl_clientid.cl_id);
3233 spin_lock(&recall_lock);
3234 list_for_each_safe(pos, next, &del_recall_lru) {
3235 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3236 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3237 u = dp->dl_time - cutoff;
3242 list_move(&dp->dl_recall_lru, &reaplist);
3244 spin_unlock(&recall_lock);
3245 list_for_each_safe(pos, next, &reaplist) {
3246 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3247 unhash_delegation(dp);
3249 test_val = nfsd4_lease;
3250 list_for_each_safe(pos, next, &close_lru) {
3251 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3252 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3253 u = oo->oo_time - cutoff;
3258 release_openowner(oo);
3260 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3261 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3262 nfs4_unlock_state();
3263 return clientid_val;
3266 static struct workqueue_struct *laundry_wq;
3267 static void laundromat_main(struct work_struct *);
3268 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3271 laundromat_main(struct work_struct *not_used)
3275 t = nfs4_laundromat();
3276 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3277 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3280 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3282 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3283 return nfserr_bad_stateid;
3288 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3290 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3292 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3293 STATEID_VAL(stateid));
3298 access_permit_read(struct nfs4_ol_stateid *stp)
3300 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3301 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3302 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3306 access_permit_write(struct nfs4_ol_stateid *stp)
3308 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3309 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3313 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3315 __be32 status = nfserr_openmode;
3317 /* For lock stateid's, we test the parent open, not the lock: */
3318 if (stp->st_openstp)
3319 stp = stp->st_openstp;
3320 if ((flags & WR_STATE) && !access_permit_write(stp))
3322 if ((flags & RD_STATE) && !access_permit_read(stp))
3329 static inline __be32
3330 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3332 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3334 else if (locks_in_grace(net)) {
3335 /* Answer in remaining cases depends on existence of
3336 * conflicting state; so we must wait out the grace period. */
3337 return nfserr_grace;
3338 } else if (flags & WR_STATE)
3339 return nfs4_share_conflict(current_fh,
3340 NFS4_SHARE_DENY_WRITE);
3341 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3342 return nfs4_share_conflict(current_fh,
3343 NFS4_SHARE_DENY_READ);
3347 * Allow READ/WRITE during grace period on recovered state only for files
3348 * that are not able to provide mandatory locking.
3351 grace_disallows_io(struct net *net, struct inode *inode)
3353 return locks_in_grace(net) && mandatory_lock(inode);
3356 /* Returns true iff a is later than b: */
3357 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3359 return (s32)a->si_generation - (s32)b->si_generation > 0;
3362 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3365 * When sessions are used the stateid generation number is ignored
3368 if (has_session && in->si_generation == 0)
3371 if (in->si_generation == ref->si_generation)
3374 /* If the client sends us a stateid from the future, it's buggy: */
3375 if (stateid_generation_after(in, ref))
3376 return nfserr_bad_stateid;
3378 * However, we could see a stateid from the past, even from a
3379 * non-buggy client. For example, if the client sends a lock
3380 * while some IO is outstanding, the lock may bump si_generation
3381 * while the IO is still in flight. The client could avoid that
3382 * situation by waiting for responses on all the IO requests,
3383 * but better performance may result in retrying IO that
3384 * receives an old_stateid error if requests are rarely
3385 * reordered in flight:
3387 return nfserr_old_stateid;
3390 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3392 struct nfs4_stid *s;
3393 struct nfs4_ol_stateid *ols;
3396 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3397 return nfserr_bad_stateid;
3398 /* Client debugging aid. */
3399 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3400 char addr_str[INET6_ADDRSTRLEN];
3401 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3403 pr_warn_ratelimited("NFSD: client %s testing state ID "
3404 "with incorrect client ID\n", addr_str);
3405 return nfserr_bad_stateid;
3407 s = find_stateid(cl, stateid);
3409 return nfserr_bad_stateid;
3410 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3413 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3415 ols = openlockstateid(s);
3416 if (ols->st_stateowner->so_is_open_owner
3417 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3418 return nfserr_bad_stateid;
3422 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s, bool sessions)
3424 struct nfs4_client *cl;
3425 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3427 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3428 return nfserr_bad_stateid;
3429 if (STALE_STATEID(stateid, nn))
3430 return nfserr_stale_stateid;
3431 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions, nn);
3433 return nfserr_expired;
3434 *s = find_stateid_by_type(cl, stateid, typemask);
3436 return nfserr_bad_stateid;
3442 * Checks for stateid operations
3445 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3446 stateid_t *stateid, int flags, struct file **filpp)
3448 struct nfs4_stid *s;
3449 struct nfs4_ol_stateid *stp = NULL;
3450 struct nfs4_delegation *dp = NULL;
3451 struct svc_fh *current_fh = &cstate->current_fh;
3452 struct inode *ino = current_fh->fh_dentry->d_inode;
3458 if (grace_disallows_io(net, ino))
3459 return nfserr_grace;
3461 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3462 return check_special_stateids(net, current_fh, stateid, flags);
3464 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s, cstate->minorversion);
3467 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3470 switch (s->sc_type) {
3471 case NFS4_DELEG_STID:
3472 dp = delegstateid(s);
3473 status = nfs4_check_delegmode(dp, flags);
3477 *filpp = dp->dl_file->fi_deleg_file;
3481 case NFS4_OPEN_STID:
3482 case NFS4_LOCK_STID:
3483 stp = openlockstateid(s);
3484 status = nfs4_check_fh(current_fh, stp);
3487 if (stp->st_stateowner->so_is_open_owner
3488 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3490 status = nfs4_check_openmode(stp, flags);
3494 if (flags & RD_STATE)
3495 *filpp = find_readable_file(stp->st_file);
3497 *filpp = find_writeable_file(stp->st_file);
3501 return nfserr_bad_stateid;
3509 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3511 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3512 return nfserr_locks_held;
3513 release_lock_stateid(stp);
3518 * Test if the stateid is valid
3521 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3522 struct nfsd4_test_stateid *test_stateid)
3524 struct nfsd4_test_stateid_id *stateid;
3525 struct nfs4_client *cl = cstate->session->se_client;
3528 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3529 stateid->ts_id_status =
3530 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3531 nfs4_unlock_state();
3537 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3538 struct nfsd4_free_stateid *free_stateid)
3540 stateid_t *stateid = &free_stateid->fr_stateid;
3541 struct nfs4_stid *s;
3542 struct nfs4_client *cl = cstate->session->se_client;
3543 __be32 ret = nfserr_bad_stateid;
3546 s = find_stateid(cl, stateid);
3549 switch (s->sc_type) {
3550 case NFS4_DELEG_STID:
3551 ret = nfserr_locks_held;
3553 case NFS4_OPEN_STID:
3554 case NFS4_LOCK_STID:
3555 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3558 if (s->sc_type == NFS4_LOCK_STID)
3559 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3561 ret = nfserr_locks_held;
3564 ret = nfserr_bad_stateid;
3567 nfs4_unlock_state();
3574 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3575 RD_STATE : WR_STATE;
3578 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3580 struct svc_fh *current_fh = &cstate->current_fh;
3581 struct nfs4_stateowner *sop = stp->st_stateowner;
3584 status = nfsd4_check_seqid(cstate, sop, seqid);
3587 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3589 * "Closed" stateid's exist *only* to return
3590 * nfserr_replay_me from the previous step.
3592 return nfserr_bad_stateid;
3593 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3596 return nfs4_check_fh(current_fh, stp);
3600 * Checks for sequence id mutating operations.
3603 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3604 stateid_t *stateid, char typemask,
3605 struct nfs4_ol_stateid **stpp)
3608 struct nfs4_stid *s;
3610 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3611 seqid, STATEID_VAL(stateid));
3614 status = nfsd4_lookup_stateid(stateid, typemask, &s, cstate->minorversion);
3617 *stpp = openlockstateid(s);
3618 cstate->replay_owner = (*stpp)->st_stateowner;
3620 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3623 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3626 struct nfs4_openowner *oo;
3628 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3629 NFS4_OPEN_STID, stpp);
3632 oo = openowner((*stpp)->st_stateowner);
3633 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3634 return nfserr_bad_stateid;
3639 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3640 struct nfsd4_open_confirm *oc)
3643 struct nfs4_openowner *oo;
3644 struct nfs4_ol_stateid *stp;
3646 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3647 (int)cstate->current_fh.fh_dentry->d_name.len,
3648 cstate->current_fh.fh_dentry->d_name.name);
3650 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3656 status = nfs4_preprocess_seqid_op(cstate,
3657 oc->oc_seqid, &oc->oc_req_stateid,
3658 NFS4_OPEN_STID, &stp);
3661 oo = openowner(stp->st_stateowner);
3662 status = nfserr_bad_stateid;
3663 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3665 oo->oo_flags |= NFS4_OO_CONFIRMED;
3666 update_stateid(&stp->st_stid.sc_stateid);
3667 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3668 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3669 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3671 nfsd4_client_record_create(oo->oo_owner.so_client);
3674 if (!cstate->replay_owner)
3675 nfs4_unlock_state();
3679 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3681 if (!test_access(access, stp))
3683 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3684 clear_access(access, stp);
3687 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3689 switch (to_access) {
3690 case NFS4_SHARE_ACCESS_READ:
3691 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3692 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3694 case NFS4_SHARE_ACCESS_WRITE:
3695 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3696 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3698 case NFS4_SHARE_ACCESS_BOTH:
3706 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3709 for (i = 0; i < 4; i++) {
3710 if ((i & deny) != i)
3716 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3717 struct nfsd4_compound_state *cstate,
3718 struct nfsd4_open_downgrade *od)
3721 struct nfs4_ol_stateid *stp;
3723 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3724 (int)cstate->current_fh.fh_dentry->d_name.len,
3725 cstate->current_fh.fh_dentry->d_name.name);
3727 /* We don't yet support WANT bits: */
3728 if (od->od_deleg_want)
3729 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3733 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3734 &od->od_stateid, &stp);
3737 status = nfserr_inval;
3738 if (!test_access(od->od_share_access, stp)) {
3739 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3740 stp->st_access_bmap, od->od_share_access);
3743 if (!test_deny(od->od_share_deny, stp)) {
3744 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3745 stp->st_deny_bmap, od->od_share_deny);
3748 nfs4_stateid_downgrade(stp, od->od_share_access);
3750 reset_union_bmap_deny(od->od_share_deny, stp);
3752 update_stateid(&stp->st_stid.sc_stateid);
3753 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3756 if (!cstate->replay_owner)
3757 nfs4_unlock_state();
3761 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3763 struct nfs4_openowner *oo;
3764 struct nfs4_ol_stateid *s;
3766 if (!so->so_is_open_owner)
3769 s = oo->oo_last_closed_stid;
3772 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3773 /* Release the last_closed_stid on the next seqid bump: */
3774 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3777 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3778 release_last_closed_stateid(oo);
3781 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3783 unhash_open_stateid(s);
3784 s->st_stid.sc_type = NFS4_CLOSED_STID;
3788 * nfs4_unlock_state() called after encode
3791 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3792 struct nfsd4_close *close)
3795 struct nfs4_openowner *oo;
3796 struct nfs4_ol_stateid *stp;
3798 dprintk("NFSD: nfsd4_close on file %.*s\n",
3799 (int)cstate->current_fh.fh_dentry->d_name.len,
3800 cstate->current_fh.fh_dentry->d_name.name);
3803 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3805 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3809 oo = openowner(stp->st_stateowner);
3811 update_stateid(&stp->st_stid.sc_stateid);
3812 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3814 nfsd4_close_open_stateid(stp);
3815 release_last_closed_stateid(oo);
3816 oo->oo_last_closed_stid = stp;
3818 if (list_empty(&oo->oo_owner.so_stateids)) {
3819 if (cstate->minorversion) {
3820 release_openowner(oo);
3821 cstate->replay_owner = NULL;
3824 * In the 4.0 case we need to keep the owners around a
3825 * little while to handle CLOSE replay.
3827 if (list_empty(&oo->oo_owner.so_stateids))
3828 move_to_close_lru(oo);
3832 if (!cstate->replay_owner)
3833 nfs4_unlock_state();
3838 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3839 struct nfsd4_delegreturn *dr)
3841 struct nfs4_delegation *dp;
3842 stateid_t *stateid = &dr->dr_stateid;
3843 struct nfs4_stid *s;
3846 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3850 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s, cstate->minorversion);
3853 dp = delegstateid(s);
3854 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3858 unhash_delegation(dp);
3860 nfs4_unlock_state();
3866 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3868 #define LOCKOWNER_INO_HASH_BITS 8
3869 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3870 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3873 end_offset(u64 start, u64 len)
3878 return end >= start ? end: NFS4_MAX_UINT64;
3881 /* last octet in a range */
3883 last_byte_offset(u64 start, u64 len)
3889 return end > start ? end - 1: NFS4_MAX_UINT64;
3892 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3894 return (file_hashval(inode) + cl_id
3895 + opaque_hashval(ownername->data, ownername->len))
3896 & LOCKOWNER_INO_HASH_MASK;
3899 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3902 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3903 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3904 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3905 * locking, this prevents us from being completely protocol-compliant. The
3906 * real solution to this problem is to start using unsigned file offsets in
3907 * the VFS, but this is a very deep change!
3910 nfs4_transform_lock_offset(struct file_lock *lock)
3912 if (lock->fl_start < 0)
3913 lock->fl_start = OFFSET_MAX;
3914 if (lock->fl_end < 0)
3915 lock->fl_end = OFFSET_MAX;
3918 /* Hack!: For now, we're defining this just so we can use a pointer to it
3919 * as a unique cookie to identify our (NFSv4's) posix locks. */
3920 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3924 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3926 struct nfs4_lockowner *lo;
3928 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3929 lo = (struct nfs4_lockowner *) fl->fl_owner;
3930 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3931 lo->lo_owner.so_owner.len, GFP_KERNEL);
3932 if (!deny->ld_owner.data)
3933 /* We just don't care that much */
3935 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3936 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3939 deny->ld_owner.len = 0;
3940 deny->ld_owner.data = NULL;
3941 deny->ld_clientid.cl_boot = 0;
3942 deny->ld_clientid.cl_id = 0;
3944 deny->ld_start = fl->fl_start;
3945 deny->ld_length = NFS4_MAX_UINT64;
3946 if (fl->fl_end != NFS4_MAX_UINT64)
3947 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3948 deny->ld_type = NFS4_READ_LT;
3949 if (fl->fl_type != F_RDLCK)
3950 deny->ld_type = NFS4_WRITE_LT;
3953 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3955 struct nfs4_ol_stateid *lst;
3957 if (!same_owner_str(&lo->lo_owner, owner, clid))
3959 lst = list_first_entry(&lo->lo_owner.so_stateids,
3960 struct nfs4_ol_stateid, st_perstateowner);
3961 return lst->st_file->fi_inode == inode;
3964 static struct nfs4_lockowner *
3965 find_lockowner_str(struct inode *inode, clientid_t *clid,
3966 struct xdr_netobj *owner)
3968 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3969 struct nfs4_lockowner *lo;
3971 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3972 if (same_lockowner_ino(lo, inode, clid, owner))
3978 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3980 struct inode *inode = open_stp->st_file->fi_inode;
3981 unsigned int inohash = lockowner_ino_hashval(inode,
3982 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3984 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3985 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3986 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3990 * Alloc a lock owner structure.
3991 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3994 * strhashval = ownerstr_hashval
3997 static struct nfs4_lockowner *
3998 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3999 struct nfs4_lockowner *lo;
4001 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4004 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4005 lo->lo_owner.so_is_open_owner = 0;
4006 /* It is the openowner seqid that will be incremented in encode in the
4007 * case of new lockowners; so increment the lock seqid manually: */
4008 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4009 hash_lockowner(lo, strhashval, clp, open_stp);
4013 static struct nfs4_ol_stateid *
4014 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4016 struct nfs4_ol_stateid *stp;
4017 struct nfs4_client *clp = lo->lo_owner.so_client;
4019 stp = nfs4_alloc_stateid(clp);
4022 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
4023 list_add(&stp->st_perfile, &fp->fi_stateids);
4024 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4025 stp->st_stateowner = &lo->lo_owner;
4028 stp->st_access_bmap = 0;
4029 stp->st_deny_bmap = open_stp->st_deny_bmap;
4030 stp->st_openstp = open_stp;
4035 check_lock_length(u64 offset, u64 length)
4037 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4038 LOFF_OVERFLOW(offset, length)));
4041 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4043 struct nfs4_file *fp = lock_stp->st_file;
4044 int oflag = nfs4_access_to_omode(access);
4046 if (test_access(access, lock_stp))
4048 nfs4_file_get_access(fp, oflag);
4049 set_access(access, lock_stp);
4052 static __be32 lookup_or_create_lock_state(struct nfsd4_compound_state *cstate, struct nfs4_ol_stateid *ost, struct nfsd4_lock *lock, struct nfs4_ol_stateid **lst, bool *new)
4054 struct nfs4_file *fi = ost->st_file;
4055 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4056 struct nfs4_client *cl = oo->oo_owner.so_client;
4057 struct nfs4_lockowner *lo;
4058 unsigned int strhashval;
4060 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4062 if (!cstate->minorversion)
4063 return nfserr_bad_seqid;
4064 /* XXX: a lockowner always has exactly one stateid: */
4065 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4066 struct nfs4_ol_stateid, st_perstateowner);
4069 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4070 &lock->v.new.owner);
4071 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4073 return nfserr_jukebox;
4074 *lst = alloc_init_lock_stateid(lo, fi, ost);
4076 release_lockowner(lo);
4077 return nfserr_jukebox;
4087 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4088 struct nfsd4_lock *lock)
4090 struct nfs4_openowner *open_sop = NULL;
4091 struct nfs4_lockowner *lock_sop = NULL;
4092 struct nfs4_ol_stateid *lock_stp;
4093 struct file *filp = NULL;
4094 struct file_lock *file_lock = NULL;
4095 struct file_lock *conflock = NULL;
4097 bool new_state = false;
4100 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4102 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4103 (long long) lock->lk_offset,
4104 (long long) lock->lk_length);
4106 if (check_lock_length(lock->lk_offset, lock->lk_length))
4107 return nfserr_inval;
4109 if ((status = fh_verify(rqstp, &cstate->current_fh,
4110 S_IFREG, NFSD_MAY_LOCK))) {
4111 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4117 if (lock->lk_is_new) {
4118 struct nfs4_ol_stateid *open_stp = NULL;
4120 if (nfsd4_has_session(cstate))
4121 /* See rfc 5661 18.10.3: given clientid is ignored: */
4122 memcpy(&lock->v.new.clientid,
4123 &cstate->session->se_client->cl_clientid,
4124 sizeof(clientid_t));
4126 status = nfserr_stale_clientid;
4127 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4130 /* validate and update open stateid and open seqid */
4131 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4132 lock->lk_new_open_seqid,
4133 &lock->lk_new_open_stateid,
4137 open_sop = openowner(open_stp->st_stateowner);
4138 status = nfserr_bad_stateid;
4139 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4140 &lock->v.new.clientid))
4142 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4143 &lock_stp, &new_state);
4145 status = nfs4_preprocess_seqid_op(cstate,
4146 lock->lk_old_lock_seqid,
4147 &lock->lk_old_lock_stateid,
4148 NFS4_LOCK_STID, &lock_stp);
4151 lock_sop = lockowner(lock_stp->st_stateowner);
4153 lkflg = setlkflg(lock->lk_type);
4154 status = nfs4_check_openmode(lock_stp, lkflg);
4158 status = nfserr_grace;
4159 if (locks_in_grace(SVC_NET(rqstp)) && !lock->lk_reclaim)
4161 status = nfserr_no_grace;
4162 if (!locks_in_grace(SVC_NET(rqstp)) && lock->lk_reclaim)
4165 file_lock = locks_alloc_lock();
4167 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4168 status = nfserr_jukebox;
4172 locks_init_lock(file_lock);
4173 switch (lock->lk_type) {
4176 filp = find_readable_file(lock_stp->st_file);
4178 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4179 file_lock->fl_type = F_RDLCK;
4182 case NFS4_WRITEW_LT:
4183 filp = find_writeable_file(lock_stp->st_file);
4185 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4186 file_lock->fl_type = F_WRLCK;
4189 status = nfserr_inval;
4193 status = nfserr_openmode;
4196 file_lock->fl_owner = (fl_owner_t)lock_sop;
4197 file_lock->fl_pid = current->tgid;
4198 file_lock->fl_file = filp;
4199 file_lock->fl_flags = FL_POSIX;
4200 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4201 file_lock->fl_start = lock->lk_offset;
4202 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4203 nfs4_transform_lock_offset(file_lock);
4205 conflock = locks_alloc_lock();
4207 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4208 status = nfserr_jukebox;
4212 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4214 case 0: /* success! */
4215 update_stateid(&lock_stp->st_stid.sc_stateid);
4216 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4220 case (EAGAIN): /* conflock holds conflicting lock */
4221 status = nfserr_denied;
4222 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4223 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4226 status = nfserr_deadlock;
4229 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4230 status = nfserrno(err);
4234 if (status && new_state)
4235 release_lockowner(lock_sop);
4236 if (!cstate->replay_owner)
4237 nfs4_unlock_state();
4239 locks_free_lock(file_lock);
4241 locks_free_lock(conflock);
4246 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4247 * so we do a temporary open here just to get an open file to pass to
4248 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4251 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4254 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4256 err = nfserrno(vfs_test_lock(file, lock));
4266 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4267 struct nfsd4_lockt *lockt)
4269 struct inode *inode;
4270 struct file_lock *file_lock = NULL;
4271 struct nfs4_lockowner *lo;
4273 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4275 if (locks_in_grace(SVC_NET(rqstp)))
4276 return nfserr_grace;
4278 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4279 return nfserr_inval;
4283 status = nfserr_stale_clientid;
4284 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid, nn))
4287 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4290 inode = cstate->current_fh.fh_dentry->d_inode;
4291 file_lock = locks_alloc_lock();
4293 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4294 status = nfserr_jukebox;
4297 locks_init_lock(file_lock);
4298 switch (lockt->lt_type) {
4301 file_lock->fl_type = F_RDLCK;
4304 case NFS4_WRITEW_LT:
4305 file_lock->fl_type = F_WRLCK;
4308 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4309 status = nfserr_inval;
4313 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4315 file_lock->fl_owner = (fl_owner_t)lo;
4316 file_lock->fl_pid = current->tgid;
4317 file_lock->fl_flags = FL_POSIX;
4319 file_lock->fl_start = lockt->lt_offset;
4320 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4322 nfs4_transform_lock_offset(file_lock);
4324 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4328 if (file_lock->fl_type != F_UNLCK) {
4329 status = nfserr_denied;
4330 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4333 nfs4_unlock_state();
4335 locks_free_lock(file_lock);
4340 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4341 struct nfsd4_locku *locku)
4343 struct nfs4_ol_stateid *stp;
4344 struct file *filp = NULL;
4345 struct file_lock *file_lock = NULL;
4349 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4350 (long long) locku->lu_offset,
4351 (long long) locku->lu_length);
4353 if (check_lock_length(locku->lu_offset, locku->lu_length))
4354 return nfserr_inval;
4358 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4359 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4362 filp = find_any_file(stp->st_file);
4364 status = nfserr_lock_range;
4367 file_lock = locks_alloc_lock();
4369 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4370 status = nfserr_jukebox;
4373 locks_init_lock(file_lock);
4374 file_lock->fl_type = F_UNLCK;
4375 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4376 file_lock->fl_pid = current->tgid;
4377 file_lock->fl_file = filp;
4378 file_lock->fl_flags = FL_POSIX;
4379 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4380 file_lock->fl_start = locku->lu_offset;
4382 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4384 nfs4_transform_lock_offset(file_lock);
4387 * Try to unlock the file in the VFS.
4389 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4391 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4395 * OK, unlock succeeded; the only thing left to do is update the stateid.
4397 update_stateid(&stp->st_stid.sc_stateid);
4398 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4401 if (!cstate->replay_owner)
4402 nfs4_unlock_state();
4404 locks_free_lock(file_lock);
4408 status = nfserrno(err);
4414 * 1: locks held by lockowner
4415 * 0: no locks held by lockowner
4418 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4420 struct file_lock **flpp;
4421 struct inode *inode = filp->fi_inode;
4425 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4426 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4437 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4438 struct nfsd4_compound_state *cstate,
4439 struct nfsd4_release_lockowner *rlockowner)
4441 clientid_t *clid = &rlockowner->rl_clientid;
4442 struct nfs4_stateowner *sop;
4443 struct nfs4_lockowner *lo;
4444 struct nfs4_ol_stateid *stp;
4445 struct xdr_netobj *owner = &rlockowner->rl_owner;
4446 struct list_head matches;
4447 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4449 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4451 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4452 clid->cl_boot, clid->cl_id);
4454 /* XXX check for lease expiration */
4456 status = nfserr_stale_clientid;
4457 if (STALE_CLIENTID(clid, nn))
4462 status = nfserr_locks_held;
4463 INIT_LIST_HEAD(&matches);
4465 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4466 if (sop->so_is_open_owner)
4468 if (!same_owner_str(sop, owner, clid))
4470 list_for_each_entry(stp, &sop->so_stateids,
4472 lo = lockowner(sop);
4473 if (check_for_locks(stp->st_file, lo))
4475 list_add(&lo->lo_list, &matches);
4478 /* Clients probably won't expect us to return with some (but not all)
4479 * of the lockowner state released; so don't release any until all
4480 * have been checked. */
4482 while (!list_empty(&matches)) {
4483 lo = list_entry(matches.next, struct nfs4_lockowner,
4485 /* unhash_stateowner deletes so_perclient only
4486 * for openowners. */
4487 list_del(&lo->lo_list);
4488 release_lockowner(lo);
4491 nfs4_unlock_state();
4495 static inline struct nfs4_client_reclaim *
4498 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4502 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4504 struct nfs4_client_reclaim *crp;
4506 crp = nfsd4_find_reclaim_client(name, nn);
4507 return (crp && crp->cr_clp);
4511 * failure => all reset bets are off, nfserr_no_grace...
4513 struct nfs4_client_reclaim *
4514 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4516 unsigned int strhashval;
4517 struct nfs4_client_reclaim *crp;
4519 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4520 crp = alloc_reclaim();
4522 strhashval = clientstr_hashval(name);
4523 INIT_LIST_HEAD(&crp->cr_strhash);
4524 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4525 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4527 nn->reclaim_str_hashtbl_size++;
4533 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4535 list_del(&crp->cr_strhash);
4537 nn->reclaim_str_hashtbl_size--;
4541 nfs4_release_reclaim(struct nfsd_net *nn)
4543 struct nfs4_client_reclaim *crp = NULL;
4546 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4547 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4548 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4549 struct nfs4_client_reclaim, cr_strhash);
4550 nfs4_remove_reclaim_record(crp, nn);
4553 BUG_ON(nn->reclaim_str_hashtbl_size);
4557 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4558 struct nfs4_client_reclaim *
4559 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4561 unsigned int strhashval;
4562 struct nfs4_client_reclaim *crp = NULL;
4564 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4566 strhashval = clientstr_hashval(recdir);
4567 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4568 if (same_name(crp->cr_recdir, recdir)) {
4576 * Called from OPEN. Look for clientid in reclaim list.
4579 nfs4_check_open_reclaim(clientid_t *clid, bool sessions)
4581 struct nfs4_client *clp;
4582 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4584 /* find clientid in conf_id_hashtbl */
4585 clp = find_confirmed_client(clid, sessions, nn);
4587 return nfserr_reclaim_bad;
4589 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4592 #ifdef CONFIG_NFSD_FAULT_INJECTION
4594 void nfsd_forget_clients(u64 num)
4596 struct nfs4_client *clp, *next;
4600 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4605 nfs4_unlock_state();
4607 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4610 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4612 release_lockowner(lockowner(sop));
4615 static void release_openowner_sop(struct nfs4_stateowner *sop)
4617 release_openowner(openowner(sop));
4620 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4621 void (*release_sop)(struct nfs4_stateowner *))
4624 struct nfs4_stateowner *sop, *next;
4626 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4627 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4628 if (sop->so_is_open_owner != is_open_owner)
4638 void nfsd_forget_locks(u64 num)
4643 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4644 nfs4_unlock_state();
4646 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4649 void nfsd_forget_openowners(u64 num)
4654 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4655 nfs4_unlock_state();
4657 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4660 static int nfsd_process_n_delegations(u64 num, struct list_head *list)
4663 struct nfs4_file *fp, *fnext;
4664 struct nfs4_delegation *dp, *dnext;
4666 for (i = 0; i < FILE_HASH_SIZE; i++) {
4667 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4668 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4669 list_move(&dp->dl_recall_lru, list);
4679 void nfsd_forget_delegations(u64 num)
4683 struct nfs4_delegation *dp, *dnext;
4685 spin_lock(&recall_lock);
4686 count = nfsd_process_n_delegations(num, &victims);
4687 spin_unlock(&recall_lock);
4690 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
4691 unhash_delegation(dp);
4692 nfs4_unlock_state();
4694 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4697 void nfsd_recall_delegations(u64 num)
4701 struct nfs4_delegation *dp, *dnext;
4703 spin_lock(&recall_lock);
4704 count = nfsd_process_n_delegations(num, &victims);
4705 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
4706 list_del(&dp->dl_recall_lru);
4707 nfsd_break_one_deleg(dp);
4709 spin_unlock(&recall_lock);
4711 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4714 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4716 /* initialization to perform at module load time: */
4719 nfs4_state_init(void)
4723 unconf_name_tree = RB_ROOT;
4724 for (i = 0; i < SESSION_HASH_SIZE; i++)
4725 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4726 for (i = 0; i < FILE_HASH_SIZE; i++) {
4727 INIT_LIST_HEAD(&file_hashtbl[i]);
4729 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4730 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4732 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4733 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4734 INIT_LIST_HEAD(&close_lru);
4735 INIT_LIST_HEAD(&client_lru);
4736 INIT_LIST_HEAD(&del_recall_lru);
4740 * Since the lifetime of a delegation isn't limited to that of an open, a
4741 * client may quite reasonably hang on to a delegation as long as it has
4742 * the inode cached. This becomes an obvious problem the first time a
4743 * client's inode cache approaches the size of the server's total memory.
4745 * For now we avoid this problem by imposing a hard limit on the number
4746 * of delegations, which varies according to the server's memory size.
4749 set_max_delegations(void)
4752 * Allow at most 4 delegations per megabyte of RAM. Quick
4753 * estimates suggest that in the worst case (where every delegation
4754 * is for a different inode), a delegation could take about 1.5K,
4755 * giving a worst case usage of about 6% of memory.
4757 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4760 static int nfs4_state_start_net(struct net *net)
4762 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4765 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4766 CLIENT_HASH_SIZE, GFP_KERNEL);
4767 if (!nn->conf_id_hashtbl)
4769 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4770 CLIENT_HASH_SIZE, GFP_KERNEL);
4771 if (!nn->unconf_id_hashtbl)
4774 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4775 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
4776 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
4778 nn->conf_name_tree = RB_ROOT;
4783 kfree(nn->conf_id_hashtbl);
4789 __nfs4_state_shutdown_net(struct net *net)
4792 struct nfs4_client *clp = NULL;
4793 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4795 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4796 while (!list_empty(&nn->conf_id_hashtbl[i])) {
4797 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4798 destroy_client(clp);
4801 kfree(nn->unconf_id_hashtbl);
4802 kfree(nn->conf_id_hashtbl);
4805 /* initialization to perform when the nfsd service is started: */
4808 nfs4_state_start(void)
4810 struct net *net = &init_net;
4811 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4815 * FIXME: For now, we hang most of the pernet global stuff off of
4816 * init_net until nfsd is fully containerized. Eventually, we'll
4817 * need to pass a net pointer into this function, take a reference
4818 * to that instead and then do most of the rest of this on a per-net
4822 ret = nfs4_state_start_net(net);
4825 nfsd4_client_tracking_init(net);
4826 nn->boot_time = get_seconds();
4827 locks_start_grace(net, &nn->nfsd4_manager);
4828 nn->grace_ended = false;
4829 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4831 ret = set_callback_cred();
4836 laundry_wq = create_singlethread_workqueue("nfsd4");
4837 if (laundry_wq == NULL) {
4841 ret = nfsd4_create_callback_queue();
4843 goto out_free_laundry;
4844 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4845 set_max_delegations();
4848 destroy_workqueue(laundry_wq);
4850 nfsd4_client_tracking_exit(net);
4851 __nfs4_state_shutdown_net(net);
4856 /* should be called with the state lock held */
4858 __nfs4_state_shutdown(struct net *net)
4860 struct nfs4_client *clp = NULL;
4861 struct nfs4_delegation *dp = NULL;
4862 struct list_head *pos, *next, reaplist;
4863 struct rb_node *node, *tmp;
4865 __nfs4_state_shutdown_net(net);
4867 node = rb_first(&unconf_name_tree);
4868 while (node != NULL) {
4870 node = rb_next(tmp);
4871 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4872 rb_erase(tmp, &unconf_name_tree);
4873 destroy_client(clp);
4876 INIT_LIST_HEAD(&reaplist);
4877 spin_lock(&recall_lock);
4878 list_for_each_safe(pos, next, &del_recall_lru) {
4879 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4880 list_move(&dp->dl_recall_lru, &reaplist);
4882 spin_unlock(&recall_lock);
4883 list_for_each_safe(pos, next, &reaplist) {
4884 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4885 unhash_delegation(dp);
4888 nfsd4_client_tracking_exit(&init_net);
4893 nfs4_state_shutdown(void)
4895 struct net *net = &init_net;
4896 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4898 cancel_delayed_work_sync(&laundromat_work);
4899 destroy_workqueue(laundry_wq);
4900 locks_end_grace(&nn->nfsd4_manager);
4902 __nfs4_state_shutdown(net);
4903 nfs4_unlock_state();
4904 nfsd4_destroy_callback_queue();
4908 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4910 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4911 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4915 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4917 if (cstate->minorversion) {
4918 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4919 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4924 clear_current_stateid(struct nfsd4_compound_state *cstate)
4926 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4930 * functions to set current state id
4933 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4935 put_stateid(cstate, &odp->od_stateid);
4939 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4941 put_stateid(cstate, &open->op_stateid);
4945 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4947 put_stateid(cstate, &close->cl_stateid);
4951 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4953 put_stateid(cstate, &lock->lk_resp_stateid);
4957 * functions to consume current state id
4961 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4963 get_stateid(cstate, &odp->od_stateid);
4967 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4969 get_stateid(cstate, &drp->dr_stateid);
4973 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4975 get_stateid(cstate, &fsp->fr_stateid);
4979 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4981 get_stateid(cstate, &setattr->sa_stateid);
4985 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4987 get_stateid(cstate, &close->cl_stateid);
4991 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4993 get_stateid(cstate, &locku->lu_stateid);
4997 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4999 get_stateid(cstate, &read->rd_stateid);
5003 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5005 get_stateid(cstate, &write->wr_stateid);