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 /* hash table for nfs4_file */
180 #define FILE_HASH_BITS 8
181 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
183 static unsigned int file_hashval(struct inode *ino)
185 /* XXX: why are we hashing on inode pointer, anyway? */
186 return hash_ptr(ino, FILE_HASH_BITS);
189 static struct list_head file_hashtbl[FILE_HASH_SIZE];
191 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
193 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
194 atomic_inc(&fp->fi_access[oflag]);
197 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
199 if (oflag == O_RDWR) {
200 __nfs4_file_get_access(fp, O_RDONLY);
201 __nfs4_file_get_access(fp, O_WRONLY);
203 __nfs4_file_get_access(fp, oflag);
206 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
208 if (fp->fi_fds[oflag]) {
209 fput(fp->fi_fds[oflag]);
210 fp->fi_fds[oflag] = NULL;
214 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
216 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
217 nfs4_file_put_fd(fp, oflag);
219 * It's also safe to get rid of the RDWR open *if*
220 * we no longer have need of the other kind of access
221 * or if we already have the other kind of open:
223 if (fp->fi_fds[1-oflag]
224 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
225 nfs4_file_put_fd(fp, O_RDWR);
229 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
231 if (oflag == O_RDWR) {
232 __nfs4_file_put_access(fp, O_RDONLY);
233 __nfs4_file_put_access(fp, O_WRONLY);
235 __nfs4_file_put_access(fp, oflag);
238 static inline int get_new_stid(struct nfs4_stid *stid)
240 static int min_stateid = 0;
241 struct idr *stateids = &stid->sc_client->cl_stateids;
245 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
247 * Note: the necessary preallocation was done in
248 * nfs4_alloc_stateid(). The idr code caps the number of
249 * preallocations that can exist at a time, but the state lock
250 * prevents anyone from using ours before we get here:
254 * It shouldn't be a problem to reuse an opaque stateid value.
255 * I don't think it is for 4.1. But with 4.0 I worry that, for
256 * example, a stray write retransmission could be accepted by
257 * the server when it should have been rejected. Therefore,
258 * adopt a trick from the sctp code to attempt to maximize the
259 * amount of time until an id is reused, by ensuring they always
260 * "increase" (mod INT_MAX):
263 min_stateid = new_stid+1;
264 if (min_stateid == INT_MAX)
269 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
271 stateid_t *s = &stid->sc_stateid;
274 stid->sc_type = type;
275 stid->sc_client = cl;
276 s->si_opaque.so_clid = cl->cl_clientid;
277 new_id = get_new_stid(stid);
278 s->si_opaque.so_id = (u32)new_id;
279 /* Will be incremented before return to client: */
280 s->si_generation = 0;
283 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
285 struct idr *stateids = &cl->cl_stateids;
287 if (!idr_pre_get(stateids, GFP_KERNEL))
290 * Note: if we fail here (or any time between now and the time
291 * we actually get the new idr), we won't need to undo the idr
292 * preallocation, since the idr code caps the number of
293 * preallocated entries.
295 return kmem_cache_alloc(slab, GFP_KERNEL);
298 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
300 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
303 static struct nfs4_delegation *
304 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
306 struct nfs4_delegation *dp;
307 struct nfs4_file *fp = stp->st_file;
309 dprintk("NFSD alloc_init_deleg\n");
311 * Major work on the lease subsystem (for example, to support
312 * calbacks on stat) will be required before we can support
313 * write delegations properly.
315 if (type != NFS4_OPEN_DELEGATE_READ)
317 if (fp->fi_had_conflict)
319 if (num_delegations > max_delegations)
321 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
324 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
326 * delegation seqid's are never incremented. The 4.1 special
327 * meaning of seqid 0 isn't meaningful, really, but let's avoid
328 * 0 anyway just for consistency and use 1:
330 dp->dl_stid.sc_stateid.si_generation = 1;
332 INIT_LIST_HEAD(&dp->dl_perfile);
333 INIT_LIST_HEAD(&dp->dl_perclnt);
334 INIT_LIST_HEAD(&dp->dl_recall_lru);
338 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
340 atomic_set(&dp->dl_count, 1);
341 nfsd4_init_callback(&dp->dl_recall);
346 nfs4_put_delegation(struct nfs4_delegation *dp)
348 if (atomic_dec_and_test(&dp->dl_count)) {
349 dprintk("NFSD: freeing dp %p\n",dp);
350 put_nfs4_file(dp->dl_file);
351 kmem_cache_free(deleg_slab, dp);
356 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
358 if (atomic_dec_and_test(&fp->fi_delegees)) {
359 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
361 fput(fp->fi_deleg_file);
362 fp->fi_deleg_file = NULL;
366 static void unhash_stid(struct nfs4_stid *s)
368 struct idr *stateids = &s->sc_client->cl_stateids;
370 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
373 /* Called under the state lock. */
375 unhash_delegation(struct nfs4_delegation *dp)
377 unhash_stid(&dp->dl_stid);
378 list_del_init(&dp->dl_perclnt);
379 spin_lock(&recall_lock);
380 list_del_init(&dp->dl_perfile);
381 list_del_init(&dp->dl_recall_lru);
382 spin_unlock(&recall_lock);
383 nfs4_put_deleg_lease(dp->dl_file);
384 nfs4_put_delegation(dp);
391 /* client_lock protects the client lru list and session hash table */
392 static DEFINE_SPINLOCK(client_lock);
394 static unsigned int clientid_hashval(u32 id)
396 return id & CLIENT_HASH_MASK;
399 static unsigned int clientstr_hashval(const char *name)
401 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
405 * We store the NONE, READ, WRITE, and BOTH bits separately in the
406 * st_{access,deny}_bmap field of the stateid, in order to track not
407 * only what share bits are currently in force, but also what
408 * combinations of share bits previous opens have used. This allows us
409 * to enforce the recommendation of rfc 3530 14.2.19 that the server
410 * return an error if the client attempt to downgrade to a combination
411 * of share bits not explicable by closing some of its previous opens.
413 * XXX: This enforcement is actually incomplete, since we don't keep
414 * track of access/deny bit combinations; so, e.g., we allow:
416 * OPEN allow read, deny write
417 * OPEN allow both, deny none
418 * DOWNGRADE allow read, deny none
420 * which we should reject.
423 bmap_to_share_mode(unsigned long bmap) {
425 unsigned int access = 0;
427 for (i = 1; i < 4; i++) {
428 if (test_bit(i, &bmap))
435 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
436 unsigned int access, deny;
438 access = bmap_to_share_mode(stp->st_access_bmap);
439 deny = bmap_to_share_mode(stp->st_deny_bmap);
440 if ((access & open->op_share_deny) || (deny & open->op_share_access))
445 /* set share access for a given stateid */
447 set_access(u32 access, struct nfs4_ol_stateid *stp)
449 __set_bit(access, &stp->st_access_bmap);
452 /* clear share access for a given stateid */
454 clear_access(u32 access, struct nfs4_ol_stateid *stp)
456 __clear_bit(access, &stp->st_access_bmap);
459 /* test whether a given stateid has access */
461 test_access(u32 access, struct nfs4_ol_stateid *stp)
463 return test_bit(access, &stp->st_access_bmap);
466 /* set share deny for a given stateid */
468 set_deny(u32 access, struct nfs4_ol_stateid *stp)
470 __set_bit(access, &stp->st_deny_bmap);
473 /* clear share deny for a given stateid */
475 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
477 __clear_bit(access, &stp->st_deny_bmap);
480 /* test whether a given stateid is denying specific access */
482 test_deny(u32 access, struct nfs4_ol_stateid *stp)
484 return test_bit(access, &stp->st_deny_bmap);
487 static int nfs4_access_to_omode(u32 access)
489 switch (access & NFS4_SHARE_ACCESS_BOTH) {
490 case NFS4_SHARE_ACCESS_READ:
492 case NFS4_SHARE_ACCESS_WRITE:
494 case NFS4_SHARE_ACCESS_BOTH:
500 /* release all access and file references for a given stateid */
502 release_all_access(struct nfs4_ol_stateid *stp)
506 for (i = 1; i < 4; i++) {
507 if (test_access(i, stp))
508 nfs4_file_put_access(stp->st_file,
509 nfs4_access_to_omode(i));
510 clear_access(i, stp);
514 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
516 list_del(&stp->st_perfile);
517 list_del(&stp->st_perstateowner);
520 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
522 release_all_access(stp);
523 put_nfs4_file(stp->st_file);
527 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
529 kmem_cache_free(stateid_slab, stp);
532 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
536 unhash_generic_stateid(stp);
537 unhash_stid(&stp->st_stid);
538 file = find_any_file(stp->st_file);
540 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
541 close_generic_stateid(stp);
542 free_generic_stateid(stp);
545 static void unhash_lockowner(struct nfs4_lockowner *lo)
547 struct nfs4_ol_stateid *stp;
549 list_del(&lo->lo_owner.so_strhash);
550 list_del(&lo->lo_perstateid);
551 list_del(&lo->lo_owner_ino_hash);
552 while (!list_empty(&lo->lo_owner.so_stateids)) {
553 stp = list_first_entry(&lo->lo_owner.so_stateids,
554 struct nfs4_ol_stateid, st_perstateowner);
555 release_lock_stateid(stp);
559 static void release_lockowner(struct nfs4_lockowner *lo)
561 unhash_lockowner(lo);
562 nfs4_free_lockowner(lo);
566 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
568 struct nfs4_lockowner *lo;
570 while (!list_empty(&open_stp->st_lockowners)) {
571 lo = list_entry(open_stp->st_lockowners.next,
572 struct nfs4_lockowner, lo_perstateid);
573 release_lockowner(lo);
577 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
579 unhash_generic_stateid(stp);
580 release_stateid_lockowners(stp);
581 close_generic_stateid(stp);
584 static void release_open_stateid(struct nfs4_ol_stateid *stp)
586 unhash_open_stateid(stp);
587 unhash_stid(&stp->st_stid);
588 free_generic_stateid(stp);
591 static void unhash_openowner(struct nfs4_openowner *oo)
593 struct nfs4_ol_stateid *stp;
595 list_del(&oo->oo_owner.so_strhash);
596 list_del(&oo->oo_perclient);
597 while (!list_empty(&oo->oo_owner.so_stateids)) {
598 stp = list_first_entry(&oo->oo_owner.so_stateids,
599 struct nfs4_ol_stateid, st_perstateowner);
600 release_open_stateid(stp);
604 static void release_last_closed_stateid(struct nfs4_openowner *oo)
606 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
609 unhash_stid(&s->st_stid);
610 free_generic_stateid(s);
611 oo->oo_last_closed_stid = NULL;
615 static void release_openowner(struct nfs4_openowner *oo)
617 unhash_openowner(oo);
618 list_del(&oo->oo_close_lru);
619 release_last_closed_stateid(oo);
620 nfs4_free_openowner(oo);
624 hash_sessionid(struct nfs4_sessionid *sessionid)
626 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
628 return sid->sequence % SESSION_HASH_SIZE;
633 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
635 u32 *ptr = (u32 *)(&sessionid->data[0]);
636 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
640 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
647 gen_sessionid(struct nfsd4_session *ses)
649 struct nfs4_client *clp = ses->se_client;
650 struct nfsd4_sessionid *sid;
652 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
653 sid->clientid = clp->cl_clientid;
654 sid->sequence = current_sessionid++;
659 * The protocol defines ca_maxresponssize_cached to include the size of
660 * the rpc header, but all we need to cache is the data starting after
661 * the end of the initial SEQUENCE operation--the rest we regenerate
662 * each time. Therefore we can advertise a ca_maxresponssize_cached
663 * value that is the number of bytes in our cache plus a few additional
664 * bytes. In order to stay on the safe side, and not promise more than
665 * we can cache, those additional bytes must be the minimum possible: 24
666 * bytes of rpc header (xid through accept state, with AUTH_NULL
667 * verifier), 12 for the compound header (with zero-length tag), and 44
668 * for the SEQUENCE op response:
670 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
673 free_session_slots(struct nfsd4_session *ses)
677 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
678 kfree(ses->se_slots[i]);
682 * We don't actually need to cache the rpc and session headers, so we
683 * can allocate a little less for each slot:
685 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
687 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
690 static int nfsd4_sanitize_slot_size(u32 size)
692 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
693 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
699 * XXX: If we run out of reserved DRC memory we could (up to a point)
700 * re-negotiate active sessions and reduce their slot usage to make
701 * room for new connections. For now we just fail the create session.
703 static int nfsd4_get_drc_mem(int slotsize, u32 num)
707 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
709 spin_lock(&nfsd_drc_lock);
710 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
711 nfsd_drc_max_mem - nfsd_drc_mem_used);
712 num = min_t(int, num, avail / slotsize);
713 nfsd_drc_mem_used += num * slotsize;
714 spin_unlock(&nfsd_drc_lock);
719 static void nfsd4_put_drc_mem(int slotsize, int num)
721 spin_lock(&nfsd_drc_lock);
722 nfsd_drc_mem_used -= slotsize * num;
723 spin_unlock(&nfsd_drc_lock);
726 static struct nfsd4_session *__alloc_session(int slotsize, int numslots)
728 struct nfsd4_session *new;
731 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
732 + sizeof(struct nfsd4_session) > PAGE_SIZE);
733 mem = numslots * sizeof(struct nfsd4_slot *);
735 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
738 /* allocate each struct nfsd4_slot and data cache in one piece */
739 for (i = 0; i < numslots; i++) {
740 mem = sizeof(struct nfsd4_slot) + slotsize;
741 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
742 if (!new->se_slots[i])
748 kfree(new->se_slots[i]);
753 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
755 u32 maxrpc = nfsd_serv->sv_max_mesg;
757 new->maxreqs = numslots;
758 new->maxresp_cached = min_t(u32, req->maxresp_cached,
759 slotsize + NFSD_MIN_HDR_SEQ_SZ);
760 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
761 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
762 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
765 static void free_conn(struct nfsd4_conn *c)
767 svc_xprt_put(c->cn_xprt);
771 static void nfsd4_conn_lost(struct svc_xpt_user *u)
773 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
774 struct nfs4_client *clp = c->cn_session->se_client;
776 spin_lock(&clp->cl_lock);
777 if (!list_empty(&c->cn_persession)) {
778 list_del(&c->cn_persession);
781 spin_unlock(&clp->cl_lock);
782 nfsd4_probe_callback(clp);
785 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
787 struct nfsd4_conn *conn;
789 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
792 svc_xprt_get(rqstp->rq_xprt);
793 conn->cn_xprt = rqstp->rq_xprt;
794 conn->cn_flags = flags;
795 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
799 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
801 conn->cn_session = ses;
802 list_add(&conn->cn_persession, &ses->se_conns);
805 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
807 struct nfs4_client *clp = ses->se_client;
809 spin_lock(&clp->cl_lock);
810 __nfsd4_hash_conn(conn, ses);
811 spin_unlock(&clp->cl_lock);
814 static int nfsd4_register_conn(struct nfsd4_conn *conn)
816 conn->cn_xpt_user.callback = nfsd4_conn_lost;
817 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
820 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
824 nfsd4_hash_conn(conn, ses);
825 ret = nfsd4_register_conn(conn);
827 /* oops; xprt is already down: */
828 nfsd4_conn_lost(&conn->cn_xpt_user);
829 if (conn->cn_flags & NFS4_CDFC4_BACK) {
830 /* callback channel may be back up */
831 nfsd4_probe_callback(ses->se_client);
835 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
837 u32 dir = NFS4_CDFC4_FORE;
839 if (cses->flags & SESSION4_BACK_CHAN)
840 dir |= NFS4_CDFC4_BACK;
841 return alloc_conn(rqstp, dir);
844 /* must be called under client_lock */
845 static void nfsd4_del_conns(struct nfsd4_session *s)
847 struct nfs4_client *clp = s->se_client;
848 struct nfsd4_conn *c;
850 spin_lock(&clp->cl_lock);
851 while (!list_empty(&s->se_conns)) {
852 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
853 list_del_init(&c->cn_persession);
854 spin_unlock(&clp->cl_lock);
856 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
859 spin_lock(&clp->cl_lock);
861 spin_unlock(&clp->cl_lock);
864 static void __free_session(struct nfsd4_session *ses)
866 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
867 free_session_slots(ses);
871 static void free_session(struct kref *kref)
873 struct nfsd4_session *ses;
875 lockdep_assert_held(&client_lock);
876 ses = container_of(kref, struct nfsd4_session, se_ref);
877 nfsd4_del_conns(ses);
881 void nfsd4_put_session(struct nfsd4_session *ses)
883 spin_lock(&client_lock);
884 nfsd4_put_session_locked(ses);
885 spin_unlock(&client_lock);
888 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fchan)
890 struct nfsd4_session *new;
891 int numslots, slotsize;
893 * Note decreasing slot size below client's request may
894 * make it difficult for client to function correctly, whereas
895 * decreasing the number of slots will (just?) affect
896 * performance. When short on memory we therefore prefer to
897 * decrease number of slots instead of their size.
899 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
900 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
904 new = __alloc_session(slotsize, numslots);
906 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
909 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
913 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
916 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
918 new->se_client = clp;
921 INIT_LIST_HEAD(&new->se_conns);
923 new->se_cb_seq_nr = 1;
924 new->se_flags = cses->flags;
925 new->se_cb_prog = cses->callback_prog;
926 new->se_cb_sec = cses->cb_sec;
927 kref_init(&new->se_ref);
928 idx = hash_sessionid(&new->se_sessionid);
929 spin_lock(&client_lock);
930 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
931 spin_lock(&clp->cl_lock);
932 list_add(&new->se_perclnt, &clp->cl_sessions);
933 spin_unlock(&clp->cl_lock);
934 spin_unlock(&client_lock);
936 if (cses->flags & SESSION4_BACK_CHAN) {
937 struct sockaddr *sa = svc_addr(rqstp);
939 * This is a little silly; with sessions there's no real
940 * use for the callback address. Use the peer address
941 * as a reasonable default for now, but consider fixing
942 * the rpc client not to require an address in the
945 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
946 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
950 /* caller must hold client_lock */
951 static struct nfsd4_session *
952 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
954 struct nfsd4_session *elem;
956 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
958 dump_sessionid(__func__, sessionid);
959 idx = hash_sessionid(sessionid);
960 /* Search in the appropriate list */
961 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
962 if (!memcmp(elem->se_sessionid.data, sessionid->data,
963 NFS4_MAX_SESSIONID_LEN)) {
968 dprintk("%s: session not found\n", __func__);
972 /* caller must hold client_lock */
974 unhash_session(struct nfsd4_session *ses)
976 list_del(&ses->se_hash);
977 spin_lock(&ses->se_client->cl_lock);
978 list_del(&ses->se_perclnt);
979 spin_unlock(&ses->se_client->cl_lock);
982 /* must be called under the client_lock */
984 renew_client_locked(struct nfs4_client *clp)
986 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
988 if (is_client_expired(clp)) {
990 printk("%s: client (clientid %08x/%08x) already expired\n",
992 clp->cl_clientid.cl_boot,
993 clp->cl_clientid.cl_id);
997 dprintk("renewing client (clientid %08x/%08x)\n",
998 clp->cl_clientid.cl_boot,
999 clp->cl_clientid.cl_id);
1000 list_move_tail(&clp->cl_lru, &nn->client_lru);
1001 clp->cl_time = get_seconds();
1005 renew_client(struct nfs4_client *clp)
1007 spin_lock(&client_lock);
1008 renew_client_locked(clp);
1009 spin_unlock(&client_lock);
1012 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1014 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1016 if (clid->cl_boot == nn->boot_time)
1018 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1019 clid->cl_boot, clid->cl_id, nn->boot_time);
1024 * XXX Should we use a slab cache ?
1025 * This type of memory management is somewhat inefficient, but we use it
1026 * anyway since SETCLIENTID is not a common operation.
1028 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1030 struct nfs4_client *clp;
1032 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1035 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1036 if (clp->cl_name.data == NULL) {
1040 clp->cl_name.len = name.len;
1045 free_client(struct nfs4_client *clp)
1047 lockdep_assert_held(&client_lock);
1048 while (!list_empty(&clp->cl_sessions)) {
1049 struct nfsd4_session *ses;
1050 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1052 list_del(&ses->se_perclnt);
1053 nfsd4_put_session_locked(ses);
1055 free_svc_cred(&clp->cl_cred);
1056 kfree(clp->cl_name.data);
1061 release_session_client(struct nfsd4_session *session)
1063 struct nfs4_client *clp = session->se_client;
1065 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1067 if (is_client_expired(clp)) {
1069 session->se_client = NULL;
1071 renew_client_locked(clp);
1072 spin_unlock(&client_lock);
1075 /* must be called under the client_lock */
1077 unhash_client_locked(struct nfs4_client *clp)
1079 struct nfsd4_session *ses;
1081 mark_client_expired(clp);
1082 list_del(&clp->cl_lru);
1083 spin_lock(&clp->cl_lock);
1084 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1085 list_del_init(&ses->se_hash);
1086 spin_unlock(&clp->cl_lock);
1090 destroy_client(struct nfs4_client *clp)
1092 struct nfs4_openowner *oo;
1093 struct nfs4_delegation *dp;
1094 struct list_head reaplist;
1095 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1097 INIT_LIST_HEAD(&reaplist);
1098 spin_lock(&recall_lock);
1099 while (!list_empty(&clp->cl_delegations)) {
1100 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1101 list_del_init(&dp->dl_perclnt);
1102 list_move(&dp->dl_recall_lru, &reaplist);
1104 spin_unlock(&recall_lock);
1105 while (!list_empty(&reaplist)) {
1106 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1107 unhash_delegation(dp);
1109 while (!list_empty(&clp->cl_openowners)) {
1110 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1111 release_openowner(oo);
1113 nfsd4_shutdown_callback(clp);
1114 if (clp->cl_cb_conn.cb_xprt)
1115 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1116 list_del(&clp->cl_idhash);
1117 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1118 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1120 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1121 spin_lock(&client_lock);
1122 unhash_client_locked(clp);
1123 if (atomic_read(&clp->cl_refcount) == 0)
1125 spin_unlock(&client_lock);
1128 static void expire_client(struct nfs4_client *clp)
1130 nfsd4_client_record_remove(clp);
1131 destroy_client(clp);
1134 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1136 memcpy(target->cl_verifier.data, source->data,
1137 sizeof(target->cl_verifier.data));
1140 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1142 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1143 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1146 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1148 if (source->cr_principal) {
1149 target->cr_principal =
1150 kstrdup(source->cr_principal, GFP_KERNEL);
1151 if (target->cr_principal == NULL)
1154 target->cr_principal = NULL;
1155 target->cr_flavor = source->cr_flavor;
1156 target->cr_uid = source->cr_uid;
1157 target->cr_gid = source->cr_gid;
1158 target->cr_group_info = source->cr_group_info;
1159 get_group_info(target->cr_group_info);
1164 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1168 res = o1->len - o2->len;
1171 return (long long)memcmp(o1->data, o2->data, o1->len);
1174 static int same_name(const char *n1, const char *n2)
1176 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1180 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1182 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1186 same_clid(clientid_t *cl1, clientid_t *cl2)
1188 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1191 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1195 if (g1->ngroups != g2->ngroups)
1197 for (i=0; i<g1->ngroups; i++)
1198 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1204 * RFC 3530 language requires clid_inuse be returned when the
1205 * "principal" associated with a requests differs from that previously
1206 * used. We use uid, gid's, and gss principal string as our best
1207 * approximation. We also don't want to allow non-gss use of a client
1208 * established using gss: in theory cr_principal should catch that
1209 * change, but in practice cr_principal can be null even in the gss case
1210 * since gssd doesn't always pass down a principal string.
1212 static bool is_gss_cred(struct svc_cred *cr)
1214 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1215 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1220 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1222 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1223 || (cr1->cr_uid != cr2->cr_uid)
1224 || (cr1->cr_gid != cr2->cr_gid)
1225 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1227 if (cr1->cr_principal == cr2->cr_principal)
1229 if (!cr1->cr_principal || !cr2->cr_principal)
1231 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1234 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1236 static u32 current_clientid = 1;
1238 clp->cl_clientid.cl_boot = nn->boot_time;
1239 clp->cl_clientid.cl_id = current_clientid++;
1242 static void gen_confirm(struct nfs4_client *clp)
1247 verf[0] = (__be32)get_seconds();
1248 verf[1] = (__be32)i++;
1249 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1252 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1254 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1257 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1259 struct nfs4_stid *s;
1261 s = find_stateid(cl, t);
1264 if (typemask & s->sc_type)
1269 static struct nfs4_client *create_client(struct xdr_netobj name,
1270 struct svc_rqst *rqstp, nfs4_verifier *verf)
1272 struct nfs4_client *clp;
1273 struct sockaddr *sa = svc_addr(rqstp);
1275 struct net *net = SVC_NET(rqstp);
1277 clp = alloc_client(name);
1281 INIT_LIST_HEAD(&clp->cl_sessions);
1282 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1284 spin_lock(&client_lock);
1286 spin_unlock(&client_lock);
1289 idr_init(&clp->cl_stateids);
1290 atomic_set(&clp->cl_refcount, 0);
1291 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1292 INIT_LIST_HEAD(&clp->cl_idhash);
1293 INIT_LIST_HEAD(&clp->cl_openowners);
1294 INIT_LIST_HEAD(&clp->cl_delegations);
1295 INIT_LIST_HEAD(&clp->cl_lru);
1296 INIT_LIST_HEAD(&clp->cl_callbacks);
1297 spin_lock_init(&clp->cl_lock);
1298 nfsd4_init_callback(&clp->cl_cb_null);
1299 clp->cl_time = get_seconds();
1300 clear_bit(0, &clp->cl_cb_slot_busy);
1301 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1302 copy_verf(clp, verf);
1303 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1305 clp->cl_cb_session = NULL;
1311 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1313 struct rb_node **new = &(root->rb_node), *parent = NULL;
1314 struct nfs4_client *clp;
1317 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1320 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1321 new = &((*new)->rb_left);
1323 new = &((*new)->rb_right);
1326 rb_link_node(&new_clp->cl_namenode, parent, new);
1327 rb_insert_color(&new_clp->cl_namenode, root);
1330 static struct nfs4_client *
1331 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1334 struct rb_node *node = root->rb_node;
1335 struct nfs4_client *clp;
1338 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1339 cmp = compare_blob(&clp->cl_name, name);
1341 node = node->rb_left;
1343 node = node->rb_right;
1351 add_to_unconfirmed(struct nfs4_client *clp)
1353 unsigned int idhashval;
1354 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1356 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1357 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1358 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1359 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1364 move_to_confirmed(struct nfs4_client *clp)
1366 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1367 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1369 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1370 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1371 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1372 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1373 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1377 static struct nfs4_client *
1378 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1380 struct nfs4_client *clp;
1381 unsigned int idhashval = clientid_hashval(clid->cl_id);
1383 list_for_each_entry(clp, &nn->conf_id_hashtbl[idhashval], cl_idhash) {
1384 if (same_clid(&clp->cl_clientid, clid)) {
1385 if ((bool)clp->cl_minorversion != sessions)
1394 static struct nfs4_client *
1395 find_unconfirmed_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->unconf_id_hashtbl[idhashval], cl_idhash) {
1401 if (same_clid(&clp->cl_clientid, clid)) {
1402 if ((bool)clp->cl_minorversion != sessions)
1410 static bool clp_used_exchangeid(struct nfs4_client *clp)
1412 return clp->cl_exchange_flags != 0;
1415 static struct nfs4_client *
1416 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1418 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1421 static struct nfs4_client *
1422 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1424 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1428 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1430 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1431 struct sockaddr *sa = svc_addr(rqstp);
1432 u32 scopeid = rpc_get_scope_id(sa);
1433 unsigned short expected_family;
1435 /* Currently, we only support tcp and tcp6 for the callback channel */
1436 if (se->se_callback_netid_len == 3 &&
1437 !memcmp(se->se_callback_netid_val, "tcp", 3))
1438 expected_family = AF_INET;
1439 else if (se->se_callback_netid_len == 4 &&
1440 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1441 expected_family = AF_INET6;
1445 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1446 se->se_callback_addr_len,
1447 (struct sockaddr *)&conn->cb_addr,
1448 sizeof(conn->cb_addr));
1450 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1453 if (conn->cb_addr.ss_family == AF_INET6)
1454 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1456 conn->cb_prog = se->se_callback_prog;
1457 conn->cb_ident = se->se_callback_ident;
1458 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1461 conn->cb_addr.ss_family = AF_UNSPEC;
1462 conn->cb_addrlen = 0;
1463 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1464 "will not receive delegations\n",
1465 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1471 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1474 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1476 struct nfsd4_slot *slot = resp->cstate.slot;
1479 dprintk("--> %s slot %p\n", __func__, slot);
1481 slot->sl_opcnt = resp->opcnt;
1482 slot->sl_status = resp->cstate.status;
1484 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1485 if (nfsd4_not_cached(resp)) {
1486 slot->sl_datalen = 0;
1489 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1490 base = (char *)resp->cstate.datap -
1491 (char *)resp->xbuf->head[0].iov_base;
1492 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1494 WARN("%s: sessions DRC could not cache compound\n", __func__);
1499 * Encode the replay sequence operation from the slot values.
1500 * If cachethis is FALSE encode the uncached rep error on the next
1501 * operation which sets resp->p and increments resp->opcnt for
1502 * nfs4svc_encode_compoundres.
1506 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1507 struct nfsd4_compoundres *resp)
1509 struct nfsd4_op *op;
1510 struct nfsd4_slot *slot = resp->cstate.slot;
1512 /* Encode the replayed sequence operation */
1513 op = &args->ops[resp->opcnt - 1];
1514 nfsd4_encode_operation(resp, op);
1516 /* Return nfserr_retry_uncached_rep in next operation. */
1517 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1518 op = &args->ops[resp->opcnt++];
1519 op->status = nfserr_retry_uncached_rep;
1520 nfsd4_encode_operation(resp, op);
1526 * The sequence operation is not cached because we can use the slot and
1530 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1531 struct nfsd4_sequence *seq)
1533 struct nfsd4_slot *slot = resp->cstate.slot;
1536 dprintk("--> %s slot %p\n", __func__, slot);
1538 /* Either returns 0 or nfserr_retry_uncached */
1539 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1540 if (status == nfserr_retry_uncached_rep)
1543 /* The sequence operation has been encoded, cstate->datap set. */
1544 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1546 resp->opcnt = slot->sl_opcnt;
1547 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1548 status = slot->sl_status;
1554 * Set the exchange_id flags returned by the server.
1557 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1559 /* pNFS is not supported */
1560 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1562 /* Referrals are supported, Migration is not. */
1563 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1565 /* set the wire flags to return to client. */
1566 clid->flags = new->cl_exchange_flags;
1569 static bool client_has_state(struct nfs4_client *clp)
1572 * Note clp->cl_openowners check isn't quite right: there's no
1573 * need to count owners without stateid's.
1575 * Also note we should probably be using this in 4.0 case too.
1577 return !list_empty(&clp->cl_openowners)
1578 || !list_empty(&clp->cl_delegations)
1579 || !list_empty(&clp->cl_sessions);
1583 nfsd4_exchange_id(struct svc_rqst *rqstp,
1584 struct nfsd4_compound_state *cstate,
1585 struct nfsd4_exchange_id *exid)
1587 struct nfs4_client *unconf, *conf, *new;
1589 char addr_str[INET6_ADDRSTRLEN];
1590 nfs4_verifier verf = exid->verifier;
1591 struct sockaddr *sa = svc_addr(rqstp);
1592 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1593 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1595 rpc_ntop(sa, addr_str, sizeof(addr_str));
1596 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1597 "ip_addr=%s flags %x, spa_how %d\n",
1598 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1599 addr_str, exid->flags, exid->spa_how);
1601 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1602 return nfserr_inval;
1604 /* Currently only support SP4_NONE */
1605 switch (exid->spa_how) {
1609 return nfserr_serverfault;
1611 BUG(); /* checked by xdr code */
1613 return nfserr_serverfault; /* no excuse :-/ */
1616 /* Cases below refer to rfc 5661 section 18.35.4: */
1618 conf = find_confirmed_client_by_name(&exid->clname, nn);
1620 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1621 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1624 if (!clp_used_exchangeid(conf)) { /* buggy client */
1625 status = nfserr_inval;
1628 if (!creds_match) { /* case 9 */
1629 status = nfserr_perm;
1632 if (!verfs_match) { /* case 8 */
1633 status = nfserr_not_same;
1637 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1641 if (!creds_match) { /* case 3 */
1642 if (client_has_state(conf)) {
1643 status = nfserr_clid_inuse;
1646 expire_client(conf);
1649 if (verfs_match) { /* case 2 */
1650 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1654 /* case 5, client reboot */
1658 if (update) { /* case 7 */
1659 status = nfserr_noent;
1663 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1664 if (unconf) /* case 4, possible retry or client restart */
1665 expire_client(unconf);
1667 /* case 1 (normal case) */
1669 new = create_client(exid->clname, rqstp, &verf);
1671 status = nfserr_jukebox;
1674 new->cl_minorversion = 1;
1677 add_to_unconfirmed(new);
1679 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1680 exid->clientid.cl_id = new->cl_clientid.cl_id;
1682 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1683 nfsd4_set_ex_flags(new, exid);
1685 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1686 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1690 nfs4_unlock_state();
1695 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1697 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1700 /* The slot is in use, and no response has been sent. */
1702 if (seqid == slot_seqid)
1703 return nfserr_jukebox;
1705 return nfserr_seq_misordered;
1707 /* Note unsigned 32-bit arithmetic handles wraparound: */
1708 if (likely(seqid == slot_seqid + 1))
1710 if (seqid == slot_seqid)
1711 return nfserr_replay_cache;
1712 return nfserr_seq_misordered;
1716 * Cache the create session result into the create session single DRC
1717 * slot cache by saving the xdr structure. sl_seqid has been set.
1718 * Do this for solo or embedded create session operations.
1721 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1722 struct nfsd4_clid_slot *slot, __be32 nfserr)
1724 slot->sl_status = nfserr;
1725 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1729 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1730 struct nfsd4_clid_slot *slot)
1732 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1733 return slot->sl_status;
1736 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1737 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1738 1 + /* MIN tag is length with zero, only length */ \
1739 3 + /* version, opcount, opcode */ \
1740 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1741 /* seqid, slotID, slotID, cache */ \
1742 4 ) * sizeof(__be32))
1744 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1745 2 + /* verifier: AUTH_NULL, length 0 */\
1747 1 + /* MIN tag is length with zero, only length */ \
1748 3 + /* opcount, opcode, opstatus*/ \
1749 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1750 /* seqid, slotID, slotID, slotID, status */ \
1751 5 ) * sizeof(__be32))
1753 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1755 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1756 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1760 nfsd4_create_session(struct svc_rqst *rqstp,
1761 struct nfsd4_compound_state *cstate,
1762 struct nfsd4_create_session *cr_ses)
1764 struct sockaddr *sa = svc_addr(rqstp);
1765 struct nfs4_client *conf, *unconf;
1766 struct nfsd4_session *new;
1767 struct nfsd4_conn *conn;
1768 struct nfsd4_clid_slot *cs_slot = NULL;
1770 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1772 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1773 return nfserr_inval;
1774 if (check_forechannel_attrs(cr_ses->fore_channel))
1775 return nfserr_toosmall;
1776 new = alloc_session(&cr_ses->fore_channel);
1778 return nfserr_jukebox;
1779 status = nfserr_jukebox;
1780 conn = alloc_conn_from_crses(rqstp, cr_ses);
1782 goto out_free_session;
1785 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1786 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1789 cs_slot = &conf->cl_cs_slot;
1790 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1791 if (status == nfserr_replay_cache) {
1792 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1794 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1795 status = nfserr_seq_misordered;
1798 } else if (unconf) {
1799 struct nfs4_client *old;
1800 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1801 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1802 status = nfserr_clid_inuse;
1805 cs_slot = &unconf->cl_cs_slot;
1806 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1808 /* an unconfirmed replay returns misordered */
1809 status = nfserr_seq_misordered;
1812 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1815 move_to_confirmed(unconf);
1818 status = nfserr_stale_clientid;
1823 * We do not support RDMA or persistent sessions
1825 cr_ses->flags &= ~SESSION4_PERSIST;
1826 cr_ses->flags &= ~SESSION4_RDMA;
1828 init_session(rqstp, new, conf, cr_ses);
1829 nfsd4_init_conn(rqstp, conn, new);
1831 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1832 NFS4_MAX_SESSIONID_LEN);
1833 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1834 sizeof(struct nfsd4_channel_attrs));
1835 cs_slot->sl_seqid++;
1836 cr_ses->seqid = cs_slot->sl_seqid;
1838 /* cache solo and embedded create sessions under the state lock */
1839 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1841 nfs4_unlock_state();
1842 dprintk("%s returns %d\n", __func__, ntohl(status));
1847 __free_session(new);
1851 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1853 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1854 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1856 return argp->opcnt == resp->opcnt;
1859 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1862 case NFS4_CDFC4_FORE:
1863 case NFS4_CDFC4_BACK:
1865 case NFS4_CDFC4_FORE_OR_BOTH:
1866 case NFS4_CDFC4_BACK_OR_BOTH:
1867 *dir = NFS4_CDFC4_BOTH;
1870 return nfserr_inval;
1873 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1875 struct nfsd4_session *session = cstate->session;
1877 spin_lock(&client_lock);
1878 session->se_cb_prog = bc->bc_cb_program;
1879 session->se_cb_sec = bc->bc_cb_sec;
1880 spin_unlock(&client_lock);
1882 nfsd4_probe_callback(session->se_client);
1887 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1888 struct nfsd4_compound_state *cstate,
1889 struct nfsd4_bind_conn_to_session *bcts)
1892 struct nfsd4_conn *conn;
1894 if (!nfsd4_last_compound_op(rqstp))
1895 return nfserr_not_only_op;
1896 spin_lock(&client_lock);
1897 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
1898 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1899 * client_lock iself: */
1900 if (cstate->session) {
1901 nfsd4_get_session(cstate->session);
1902 atomic_inc(&cstate->session->se_client->cl_refcount);
1904 spin_unlock(&client_lock);
1905 if (!cstate->session)
1906 return nfserr_badsession;
1908 status = nfsd4_map_bcts_dir(&bcts->dir);
1911 conn = alloc_conn(rqstp, bcts->dir);
1913 return nfserr_jukebox;
1914 nfsd4_init_conn(rqstp, conn, cstate->session);
1918 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1922 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1926 nfsd4_destroy_session(struct svc_rqst *r,
1927 struct nfsd4_compound_state *cstate,
1928 struct nfsd4_destroy_session *sessionid)
1930 struct nfsd4_session *ses;
1931 __be32 status = nfserr_badsession;
1934 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1935 * - Should we return nfserr_back_chan_busy if waiting for
1936 * callbacks on to-be-destroyed session?
1937 * - Do we need to clear any callback info from previous session?
1940 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1941 if (!nfsd4_last_compound_op(r))
1942 return nfserr_not_only_op;
1944 dump_sessionid(__func__, &sessionid->sessionid);
1945 spin_lock(&client_lock);
1946 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
1948 spin_unlock(&client_lock);
1952 unhash_session(ses);
1953 spin_unlock(&client_lock);
1956 nfsd4_probe_callback_sync(ses->se_client);
1957 nfs4_unlock_state();
1959 spin_lock(&client_lock);
1960 nfsd4_del_conns(ses);
1961 nfsd4_put_session_locked(ses);
1962 spin_unlock(&client_lock);
1965 dprintk("%s returns %d\n", __func__, ntohl(status));
1969 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1971 struct nfsd4_conn *c;
1973 list_for_each_entry(c, &s->se_conns, cn_persession) {
1974 if (c->cn_xprt == xpt) {
1981 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1983 struct nfs4_client *clp = ses->se_client;
1984 struct nfsd4_conn *c;
1987 spin_lock(&clp->cl_lock);
1988 c = __nfsd4_find_conn(new->cn_xprt, ses);
1990 spin_unlock(&clp->cl_lock);
1994 __nfsd4_hash_conn(new, ses);
1995 spin_unlock(&clp->cl_lock);
1996 ret = nfsd4_register_conn(new);
1998 /* oops; xprt is already down: */
1999 nfsd4_conn_lost(&new->cn_xpt_user);
2003 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2005 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2007 return args->opcnt > session->se_fchannel.maxops;
2010 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2011 struct nfsd4_session *session)
2013 struct xdr_buf *xb = &rqstp->rq_arg;
2015 return xb->len > session->se_fchannel.maxreq_sz;
2019 nfsd4_sequence(struct svc_rqst *rqstp,
2020 struct nfsd4_compound_state *cstate,
2021 struct nfsd4_sequence *seq)
2023 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2024 struct nfsd4_session *session;
2025 struct nfsd4_slot *slot;
2026 struct nfsd4_conn *conn;
2029 if (resp->opcnt != 1)
2030 return nfserr_sequence_pos;
2033 * Will be either used or freed by nfsd4_sequence_check_conn
2036 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2038 return nfserr_jukebox;
2040 spin_lock(&client_lock);
2041 status = nfserr_badsession;
2042 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2046 status = nfserr_too_many_ops;
2047 if (nfsd4_session_too_many_ops(rqstp, session))
2050 status = nfserr_req_too_big;
2051 if (nfsd4_request_too_big(rqstp, session))
2054 status = nfserr_badslot;
2055 if (seq->slotid >= session->se_fchannel.maxreqs)
2058 slot = session->se_slots[seq->slotid];
2059 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2061 /* We do not negotiate the number of slots yet, so set the
2062 * maxslots to the session maxreqs which is used to encode
2063 * sr_highest_slotid and the sr_target_slot id to maxslots */
2064 seq->maxslots = session->se_fchannel.maxreqs;
2066 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2067 slot->sl_flags & NFSD4_SLOT_INUSE);
2068 if (status == nfserr_replay_cache) {
2069 status = nfserr_seq_misordered;
2070 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2072 cstate->slot = slot;
2073 cstate->session = session;
2074 /* Return the cached reply status and set cstate->status
2075 * for nfsd4_proc_compound processing */
2076 status = nfsd4_replay_cache_entry(resp, seq);
2077 cstate->status = nfserr_replay_cache;
2083 nfsd4_sequence_check_conn(conn, session);
2086 /* Success! bump slot seqid */
2087 slot->sl_seqid = seq->seqid;
2088 slot->sl_flags |= NFSD4_SLOT_INUSE;
2090 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2092 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2094 cstate->slot = slot;
2095 cstate->session = session;
2098 /* Hold a session reference until done processing the compound. */
2099 if (cstate->session) {
2100 struct nfs4_client *clp = session->se_client;
2102 nfsd4_get_session(cstate->session);
2103 atomic_inc(&clp->cl_refcount);
2104 switch (clp->cl_cb_state) {
2106 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2108 case NFSD4_CB_FAULT:
2109 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2112 seq->status_flags = 0;
2116 spin_unlock(&client_lock);
2117 dprintk("%s: return %d\n", __func__, ntohl(status));
2122 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2124 struct nfs4_client *conf, *unconf, *clp;
2126 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2129 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2130 conf = find_confirmed_client(&dc->clientid, true, nn);
2135 if (!is_client_expired(conf) && client_has_state(conf)) {
2136 status = nfserr_clientid_busy;
2140 /* rfc5661 18.50.3 */
2141 if (cstate->session && conf == cstate->session->se_client) {
2142 status = nfserr_clientid_busy;
2148 status = nfserr_stale_clientid;
2154 nfs4_unlock_state();
2155 dprintk("%s return %d\n", __func__, ntohl(status));
2160 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2164 if (rc->rca_one_fs) {
2165 if (!cstate->current_fh.fh_dentry)
2166 return nfserr_nofilehandle;
2168 * We don't take advantage of the rca_one_fs case.
2169 * That's OK, it's optional, we can safely ignore it.
2175 status = nfserr_complete_already;
2176 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2177 &cstate->session->se_client->cl_flags))
2180 status = nfserr_stale_clientid;
2181 if (is_client_expired(cstate->session->se_client))
2183 * The following error isn't really legal.
2184 * But we only get here if the client just explicitly
2185 * destroyed the client. Surely it no longer cares what
2186 * error it gets back on an operation for the dead
2192 nfsd4_client_record_create(cstate->session->se_client);
2194 nfs4_unlock_state();
2199 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2200 struct nfsd4_setclientid *setclid)
2202 struct xdr_netobj clname = setclid->se_name;
2203 nfs4_verifier clverifier = setclid->se_verf;
2204 struct nfs4_client *conf, *unconf, *new;
2206 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2208 /* Cases below refer to rfc 3530 section 14.2.33: */
2210 conf = find_confirmed_client_by_name(&clname, nn);
2213 status = nfserr_clid_inuse;
2214 if (clp_used_exchangeid(conf))
2216 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2217 char addr_str[INET6_ADDRSTRLEN];
2218 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2220 dprintk("NFSD: setclientid: string in use by client "
2221 "at %s\n", addr_str);
2225 unconf = find_unconfirmed_client_by_name(&clname, nn);
2227 expire_client(unconf);
2228 status = nfserr_jukebox;
2229 new = create_client(clname, rqstp, &clverifier);
2232 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2233 /* case 1: probable callback update */
2234 copy_clid(new, conf);
2235 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2237 new->cl_minorversion = 0;
2238 gen_callback(new, setclid, rqstp);
2239 add_to_unconfirmed(new);
2240 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2241 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2242 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2245 nfs4_unlock_state();
2251 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2252 struct nfsd4_compound_state *cstate,
2253 struct nfsd4_setclientid_confirm *setclientid_confirm)
2255 struct nfs4_client *conf, *unconf;
2256 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2257 clientid_t * clid = &setclientid_confirm->sc_clientid;
2259 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2261 if (STALE_CLIENTID(clid, nn))
2262 return nfserr_stale_clientid;
2265 conf = find_confirmed_client(clid, false, nn);
2266 unconf = find_unconfirmed_client(clid, false, nn);
2268 * We try hard to give out unique clientid's, so if we get an
2269 * attempt to confirm the same clientid with a different cred,
2270 * there's a bug somewhere. Let's charitably assume it's our
2273 status = nfserr_serverfault;
2274 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2276 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2278 /* cases below refer to rfc 3530 section 14.2.34: */
2279 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2280 if (conf && !unconf) /* case 2: probable retransmit */
2282 else /* case 4: client hasn't noticed we rebooted yet? */
2283 status = nfserr_stale_clientid;
2287 if (conf) { /* case 1: callback update */
2288 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2289 nfsd4_probe_callback(conf);
2290 expire_client(unconf);
2291 } else { /* case 3: normal case; new or rebooted client */
2292 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2294 expire_client(conf);
2295 move_to_confirmed(unconf);
2296 nfsd4_probe_callback(unconf);
2299 nfs4_unlock_state();
2303 static struct nfs4_file *nfsd4_alloc_file(void)
2305 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2308 /* OPEN Share state helper functions */
2309 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2311 unsigned int hashval = file_hashval(ino);
2313 atomic_set(&fp->fi_ref, 1);
2314 INIT_LIST_HEAD(&fp->fi_hash);
2315 INIT_LIST_HEAD(&fp->fi_stateids);
2316 INIT_LIST_HEAD(&fp->fi_delegations);
2317 fp->fi_inode = igrab(ino);
2318 fp->fi_had_conflict = false;
2319 fp->fi_lease = NULL;
2320 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2321 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2322 spin_lock(&recall_lock);
2323 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2324 spin_unlock(&recall_lock);
2328 nfsd4_free_slab(struct kmem_cache **slab)
2332 kmem_cache_destroy(*slab);
2337 nfsd4_free_slabs(void)
2339 nfsd4_free_slab(&openowner_slab);
2340 nfsd4_free_slab(&lockowner_slab);
2341 nfsd4_free_slab(&file_slab);
2342 nfsd4_free_slab(&stateid_slab);
2343 nfsd4_free_slab(&deleg_slab);
2347 nfsd4_init_slabs(void)
2349 openowner_slab = kmem_cache_create("nfsd4_openowners",
2350 sizeof(struct nfs4_openowner), 0, 0, NULL);
2351 if (openowner_slab == NULL)
2353 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2354 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2355 if (lockowner_slab == NULL)
2357 file_slab = kmem_cache_create("nfsd4_files",
2358 sizeof(struct nfs4_file), 0, 0, NULL);
2359 if (file_slab == NULL)
2361 stateid_slab = kmem_cache_create("nfsd4_stateids",
2362 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2363 if (stateid_slab == NULL)
2365 deleg_slab = kmem_cache_create("nfsd4_delegations",
2366 sizeof(struct nfs4_delegation), 0, 0, NULL);
2367 if (deleg_slab == NULL)
2372 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2376 void nfs4_free_openowner(struct nfs4_openowner *oo)
2378 kfree(oo->oo_owner.so_owner.data);
2379 kmem_cache_free(openowner_slab, oo);
2382 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2384 kfree(lo->lo_owner.so_owner.data);
2385 kmem_cache_free(lockowner_slab, lo);
2388 static void init_nfs4_replay(struct nfs4_replay *rp)
2390 rp->rp_status = nfserr_serverfault;
2392 rp->rp_buf = rp->rp_ibuf;
2395 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2397 struct nfs4_stateowner *sop;
2399 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2403 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2404 if (!sop->so_owner.data) {
2405 kmem_cache_free(slab, sop);
2408 sop->so_owner.len = owner->len;
2410 INIT_LIST_HEAD(&sop->so_stateids);
2411 sop->so_client = clp;
2412 init_nfs4_replay(&sop->so_replay);
2416 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2418 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2420 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2421 list_add(&oo->oo_perclient, &clp->cl_openowners);
2424 static struct nfs4_openowner *
2425 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2426 struct nfs4_openowner *oo;
2428 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2431 oo->oo_owner.so_is_open_owner = 1;
2432 oo->oo_owner.so_seqid = open->op_seqid;
2433 oo->oo_flags = NFS4_OO_NEW;
2435 oo->oo_last_closed_stid = NULL;
2436 INIT_LIST_HEAD(&oo->oo_close_lru);
2437 hash_openowner(oo, clp, strhashval);
2441 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2442 struct nfs4_openowner *oo = open->op_openowner;
2443 struct nfs4_client *clp = oo->oo_owner.so_client;
2445 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2446 INIT_LIST_HEAD(&stp->st_lockowners);
2447 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2448 list_add(&stp->st_perfile, &fp->fi_stateids);
2449 stp->st_stateowner = &oo->oo_owner;
2452 stp->st_access_bmap = 0;
2453 stp->st_deny_bmap = 0;
2454 set_access(open->op_share_access, stp);
2455 set_deny(open->op_share_deny, stp);
2456 stp->st_openstp = NULL;
2460 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2462 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2464 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2466 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2467 oo->oo_time = get_seconds();
2471 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2474 return (sop->so_owner.len == owner->len) &&
2475 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2476 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2479 static struct nfs4_openowner *
2480 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2481 bool sessions, struct nfsd_net *nn)
2483 struct nfs4_stateowner *so;
2484 struct nfs4_openowner *oo;
2485 struct nfs4_client *clp;
2487 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2488 if (!so->so_is_open_owner)
2490 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2492 clp = oo->oo_owner.so_client;
2493 if ((bool)clp->cl_minorversion != sessions)
2495 renew_client(oo->oo_owner.so_client);
2502 /* search file_hashtbl[] for file */
2503 static struct nfs4_file *
2504 find_file(struct inode *ino)
2506 unsigned int hashval = file_hashval(ino);
2507 struct nfs4_file *fp;
2509 spin_lock(&recall_lock);
2510 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2511 if (fp->fi_inode == ino) {
2513 spin_unlock(&recall_lock);
2517 spin_unlock(&recall_lock);
2522 * Called to check deny when READ with all zero stateid or
2523 * WRITE with all zero or all one stateid
2526 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2528 struct inode *ino = current_fh->fh_dentry->d_inode;
2529 struct nfs4_file *fp;
2530 struct nfs4_ol_stateid *stp;
2533 dprintk("NFSD: nfs4_share_conflict\n");
2535 fp = find_file(ino);
2538 ret = nfserr_locked;
2539 /* Search for conflicting share reservations */
2540 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2541 if (test_deny(deny_type, stp) ||
2542 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2551 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2553 /* We're assuming the state code never drops its reference
2554 * without first removing the lease. Since we're in this lease
2555 * callback (and since the lease code is serialized by the kernel
2556 * lock) we know the server hasn't removed the lease yet, we know
2557 * it's safe to take a reference: */
2558 atomic_inc(&dp->dl_count);
2560 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2562 /* only place dl_time is set. protected by lock_flocks*/
2563 dp->dl_time = get_seconds();
2565 nfsd4_cb_recall(dp);
2568 /* Called from break_lease() with lock_flocks() held. */
2569 static void nfsd_break_deleg_cb(struct file_lock *fl)
2571 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2572 struct nfs4_delegation *dp;
2575 WARN(1, "(%p)->fl_owner NULL\n", fl);
2578 if (fp->fi_had_conflict) {
2579 WARN(1, "duplicate break on %p\n", fp);
2583 * We don't want the locks code to timeout the lease for us;
2584 * we'll remove it ourself if a delegation isn't returned
2587 fl->fl_break_time = 0;
2589 spin_lock(&recall_lock);
2590 fp->fi_had_conflict = true;
2591 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2592 nfsd_break_one_deleg(dp);
2593 spin_unlock(&recall_lock);
2597 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2600 return lease_modify(onlist, arg);
2605 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2606 .lm_break = nfsd_break_deleg_cb,
2607 .lm_change = nfsd_change_deleg_cb,
2610 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2612 if (nfsd4_has_session(cstate))
2614 if (seqid == so->so_seqid - 1)
2615 return nfserr_replay_me;
2616 if (seqid == so->so_seqid)
2618 return nfserr_bad_seqid;
2622 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2623 struct nfsd4_open *open, struct nfsd_net *nn)
2625 clientid_t *clientid = &open->op_clientid;
2626 struct nfs4_client *clp = NULL;
2627 unsigned int strhashval;
2628 struct nfs4_openowner *oo = NULL;
2631 if (STALE_CLIENTID(&open->op_clientid, nn))
2632 return nfserr_stale_clientid;
2634 * In case we need it later, after we've already created the
2635 * file and don't want to risk a further failure:
2637 open->op_file = nfsd4_alloc_file();
2638 if (open->op_file == NULL)
2639 return nfserr_jukebox;
2641 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2642 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2643 open->op_openowner = oo;
2645 clp = find_confirmed_client(clientid, cstate->minorversion,
2648 return nfserr_expired;
2651 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2652 /* Replace unconfirmed owners without checking for replay. */
2653 clp = oo->oo_owner.so_client;
2654 release_openowner(oo);
2655 open->op_openowner = NULL;
2658 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2661 clp = oo->oo_owner.so_client;
2664 oo = alloc_init_open_stateowner(strhashval, clp, open);
2666 return nfserr_jukebox;
2667 open->op_openowner = oo;
2669 open->op_stp = nfs4_alloc_stateid(clp);
2671 return nfserr_jukebox;
2675 static inline __be32
2676 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2678 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2679 return nfserr_openmode;
2684 static int share_access_to_flags(u32 share_access)
2686 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2689 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2691 struct nfs4_stid *ret;
2693 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2696 return delegstateid(ret);
2699 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2701 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2702 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2706 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2707 struct nfs4_delegation **dp)
2710 __be32 status = nfserr_bad_stateid;
2712 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2715 flags = share_access_to_flags(open->op_share_access);
2716 status = nfs4_check_delegmode(*dp, flags);
2720 if (!nfsd4_is_deleg_cur(open))
2724 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2729 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2731 struct nfs4_ol_stateid *local;
2732 struct nfs4_openowner *oo = open->op_openowner;
2734 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2735 /* ignore lock owners */
2736 if (local->st_stateowner->so_is_open_owner == 0)
2738 /* remember if we have seen this open owner */
2739 if (local->st_stateowner == &oo->oo_owner)
2741 /* check for conflicting share reservations */
2742 if (!test_share(local, open))
2743 return nfserr_share_denied;
2748 static inline int nfs4_access_to_access(u32 nfs4_access)
2752 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2753 flags |= NFSD_MAY_READ;
2754 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2755 flags |= NFSD_MAY_WRITE;
2759 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2760 struct svc_fh *cur_fh, struct nfsd4_open *open)
2763 int oflag = nfs4_access_to_omode(open->op_share_access);
2764 int access = nfs4_access_to_access(open->op_share_access);
2766 if (!fp->fi_fds[oflag]) {
2767 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2768 &fp->fi_fds[oflag]);
2772 nfs4_file_get_access(fp, oflag);
2777 static inline __be32
2778 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2779 struct nfsd4_open *open)
2781 struct iattr iattr = {
2782 .ia_valid = ATTR_SIZE,
2785 if (!open->op_truncate)
2787 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2788 return nfserr_inval;
2789 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2793 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)
2795 u32 op_share_access = open->op_share_access;
2799 new_access = !test_access(op_share_access, stp);
2801 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2805 status = nfsd4_truncate(rqstp, cur_fh, open);
2808 int oflag = nfs4_access_to_omode(op_share_access);
2809 nfs4_file_put_access(fp, oflag);
2813 /* remember the open */
2814 set_access(op_share_access, stp);
2815 set_deny(open->op_share_deny, stp);
2822 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2824 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2827 /* Should we give out recallable state?: */
2828 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2830 if (clp->cl_cb_state == NFSD4_CB_UP)
2833 * In the sessions case, since we don't have to establish a
2834 * separate connection for callbacks, we assume it's OK
2835 * until we hear otherwise:
2837 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2840 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2842 struct file_lock *fl;
2844 fl = locks_alloc_lock();
2847 locks_init_lock(fl);
2848 fl->fl_lmops = &nfsd_lease_mng_ops;
2849 fl->fl_flags = FL_LEASE;
2850 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2851 fl->fl_end = OFFSET_MAX;
2852 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2853 fl->fl_pid = current->tgid;
2857 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2859 struct nfs4_file *fp = dp->dl_file;
2860 struct file_lock *fl;
2863 fl = nfs4_alloc_init_lease(dp, flag);
2866 fl->fl_file = find_readable_file(fp);
2867 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2868 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2870 list_del_init(&dp->dl_perclnt);
2871 locks_free_lock(fl);
2875 fp->fi_deleg_file = get_file(fl->fl_file);
2876 atomic_set(&fp->fi_delegees, 1);
2877 list_add(&dp->dl_perfile, &fp->fi_delegations);
2881 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2883 struct nfs4_file *fp = dp->dl_file;
2886 return nfs4_setlease(dp, flag);
2887 spin_lock(&recall_lock);
2888 if (fp->fi_had_conflict) {
2889 spin_unlock(&recall_lock);
2892 atomic_inc(&fp->fi_delegees);
2893 list_add(&dp->dl_perfile, &fp->fi_delegations);
2894 spin_unlock(&recall_lock);
2895 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2899 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2901 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2902 if (status == -EAGAIN)
2903 open->op_why_no_deleg = WND4_CONTENTION;
2905 open->op_why_no_deleg = WND4_RESOURCE;
2906 switch (open->op_deleg_want) {
2907 case NFS4_SHARE_WANT_READ_DELEG:
2908 case NFS4_SHARE_WANT_WRITE_DELEG:
2909 case NFS4_SHARE_WANT_ANY_DELEG:
2911 case NFS4_SHARE_WANT_CANCEL:
2912 open->op_why_no_deleg = WND4_CANCELLED;
2914 case NFS4_SHARE_WANT_NO_DELEG:
2915 BUG(); /* not supposed to get here */
2921 * Attempt to hand out a delegation.
2924 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2925 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2927 struct nfs4_delegation *dp;
2928 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2930 int status = 0, flag = 0;
2932 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2933 flag = NFS4_OPEN_DELEGATE_NONE;
2934 open->op_recall = 0;
2935 switch (open->op_claim_type) {
2936 case NFS4_OPEN_CLAIM_PREVIOUS:
2938 open->op_recall = 1;
2939 flag = open->op_delegate_type;
2940 if (flag == NFS4_OPEN_DELEGATE_NONE)
2943 case NFS4_OPEN_CLAIM_NULL:
2944 /* Let's not give out any delegations till everyone's
2945 * had the chance to reclaim theirs.... */
2946 if (locks_in_grace(net))
2948 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2950 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2951 flag = NFS4_OPEN_DELEGATE_WRITE;
2953 flag = NFS4_OPEN_DELEGATE_READ;
2959 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2962 status = nfs4_set_delegation(dp, flag);
2966 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2968 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2969 STATEID_VAL(&dp->dl_stid.sc_stateid));
2971 open->op_delegate_type = flag;
2972 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2973 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2974 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2975 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2977 /* 4.1 client asking for a delegation? */
2978 if (open->op_deleg_want)
2979 nfsd4_open_deleg_none_ext(open, status);
2983 nfs4_put_delegation(dp);
2985 flag = NFS4_OPEN_DELEGATE_NONE;
2989 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2990 struct nfs4_delegation *dp)
2992 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2993 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2994 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2995 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2996 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2997 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2998 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2999 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3001 /* Otherwise the client must be confused wanting a delegation
3002 * it already has, therefore we don't return
3003 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3008 * called with nfs4_lock_state() held.
3011 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3013 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3014 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3015 struct nfs4_file *fp = NULL;
3016 struct inode *ino = current_fh->fh_dentry->d_inode;
3017 struct nfs4_ol_stateid *stp = NULL;
3018 struct nfs4_delegation *dp = NULL;
3022 * Lookup file; if found, lookup stateid and check open request,
3023 * and check for delegations in the process of being recalled.
3024 * If not found, create the nfs4_file struct
3026 fp = find_file(ino);
3028 if ((status = nfs4_check_open(fp, open, &stp)))
3030 status = nfs4_check_deleg(cl, fp, open, &dp);
3034 status = nfserr_bad_stateid;
3035 if (nfsd4_is_deleg_cur(open))
3037 status = nfserr_jukebox;
3039 open->op_file = NULL;
3040 nfsd4_init_file(fp, ino);
3044 * OPEN the file, or upgrade an existing OPEN.
3045 * If truncate fails, the OPEN fails.
3048 /* Stateid was found, this is an OPEN upgrade */
3049 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3053 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3056 status = nfsd4_truncate(rqstp, current_fh, open);
3060 open->op_stp = NULL;
3061 init_open_stateid(stp, fp, open);
3063 update_stateid(&stp->st_stid.sc_stateid);
3064 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3066 if (nfsd4_has_session(&resp->cstate)) {
3067 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3069 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3070 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3071 open->op_why_no_deleg = WND4_NOT_WANTED;
3077 * Attempt to hand out a delegation. No error return, because the
3078 * OPEN succeeds even if we fail.
3080 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3084 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3085 STATEID_VAL(&stp->st_stid.sc_stateid));
3087 /* 4.1 client trying to upgrade/downgrade delegation? */
3088 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3089 open->op_deleg_want)
3090 nfsd4_deleg_xgrade_none_ext(open, dp);
3094 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3095 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3097 * To finish the open response, we just need to set the rflags.
3099 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3100 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3101 !nfsd4_has_session(&resp->cstate))
3102 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3107 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3109 if (open->op_openowner) {
3110 struct nfs4_openowner *oo = open->op_openowner;
3112 if (!list_empty(&oo->oo_owner.so_stateids))
3113 list_del_init(&oo->oo_close_lru);
3114 if (oo->oo_flags & NFS4_OO_NEW) {
3116 release_openowner(oo);
3117 open->op_openowner = NULL;
3119 oo->oo_flags &= ~NFS4_OO_NEW;
3123 nfsd4_free_file(open->op_file);
3125 free_generic_stateid(open->op_stp);
3129 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3132 struct nfs4_client *clp;
3134 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3137 dprintk("process_renew(%08x/%08x): starting\n",
3138 clid->cl_boot, clid->cl_id);
3139 status = nfserr_stale_clientid;
3140 if (STALE_CLIENTID(clid, nn))
3142 clp = find_confirmed_client(clid, cstate->minorversion, nn);
3143 status = nfserr_expired;
3145 /* We assume the client took too long to RENEW. */
3146 dprintk("nfsd4_renew: clientid not found!\n");
3149 status = nfserr_cb_path_down;
3150 if (!list_empty(&clp->cl_delegations)
3151 && clp->cl_cb_state != NFSD4_CB_UP)
3155 nfs4_unlock_state();
3160 nfsd4_end_grace(struct net *net)
3162 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3164 /* do nothing if grace period already ended */
3165 if (nn->grace_ended)
3168 dprintk("NFSD: end of grace period\n");
3169 nn->grace_ended = true;
3170 nfsd4_record_grace_done(net, nn->boot_time);
3171 locks_end_grace(&nn->nfsd4_manager);
3173 * Now that every NFSv4 client has had the chance to recover and
3174 * to see the (possibly new, possibly shorter) lease time, we
3175 * can safely set the next grace time to the current lease time:
3177 nfsd4_grace = nfsd4_lease;
3181 nfs4_laundromat(void)
3183 struct nfs4_client *clp;
3184 struct nfs4_openowner *oo;
3185 struct nfs4_delegation *dp;
3186 struct list_head *pos, *next, reaplist;
3187 time_t cutoff = get_seconds() - nfsd4_lease;
3188 time_t t, clientid_val = nfsd4_lease;
3189 time_t u, test_val = nfsd4_lease;
3190 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3194 dprintk("NFSD: laundromat service - starting\n");
3195 nfsd4_end_grace(&init_net);
3196 INIT_LIST_HEAD(&reaplist);
3197 spin_lock(&client_lock);
3198 list_for_each_safe(pos, next, &nn->client_lru) {
3199 clp = list_entry(pos, struct nfs4_client, cl_lru);
3200 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3201 t = clp->cl_time - cutoff;
3202 if (clientid_val > t)
3206 if (atomic_read(&clp->cl_refcount)) {
3207 dprintk("NFSD: client in use (clientid %08x)\n",
3208 clp->cl_clientid.cl_id);
3211 unhash_client_locked(clp);
3212 list_add(&clp->cl_lru, &reaplist);
3214 spin_unlock(&client_lock);
3215 list_for_each_safe(pos, next, &reaplist) {
3216 clp = list_entry(pos, struct nfs4_client, cl_lru);
3217 dprintk("NFSD: purging unused client (clientid %08x)\n",
3218 clp->cl_clientid.cl_id);
3221 spin_lock(&recall_lock);
3222 list_for_each_safe(pos, next, &del_recall_lru) {
3223 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3224 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3225 u = dp->dl_time - cutoff;
3230 list_move(&dp->dl_recall_lru, &reaplist);
3232 spin_unlock(&recall_lock);
3233 list_for_each_safe(pos, next, &reaplist) {
3234 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3235 unhash_delegation(dp);
3237 test_val = nfsd4_lease;
3238 list_for_each_safe(pos, next, &nn->close_lru) {
3239 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3240 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3241 u = oo->oo_time - cutoff;
3246 release_openowner(oo);
3248 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3249 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3250 nfs4_unlock_state();
3251 return clientid_val;
3254 static struct workqueue_struct *laundry_wq;
3255 static void laundromat_main(struct work_struct *);
3256 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3259 laundromat_main(struct work_struct *not_used)
3263 t = nfs4_laundromat();
3264 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3265 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3268 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3270 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3271 return nfserr_bad_stateid;
3276 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3278 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3280 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3281 STATEID_VAL(stateid));
3286 access_permit_read(struct nfs4_ol_stateid *stp)
3288 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3289 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3290 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3294 access_permit_write(struct nfs4_ol_stateid *stp)
3296 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3297 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3301 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3303 __be32 status = nfserr_openmode;
3305 /* For lock stateid's, we test the parent open, not the lock: */
3306 if (stp->st_openstp)
3307 stp = stp->st_openstp;
3308 if ((flags & WR_STATE) && !access_permit_write(stp))
3310 if ((flags & RD_STATE) && !access_permit_read(stp))
3317 static inline __be32
3318 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3320 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3322 else if (locks_in_grace(net)) {
3323 /* Answer in remaining cases depends on existence of
3324 * conflicting state; so we must wait out the grace period. */
3325 return nfserr_grace;
3326 } else if (flags & WR_STATE)
3327 return nfs4_share_conflict(current_fh,
3328 NFS4_SHARE_DENY_WRITE);
3329 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3330 return nfs4_share_conflict(current_fh,
3331 NFS4_SHARE_DENY_READ);
3335 * Allow READ/WRITE during grace period on recovered state only for files
3336 * that are not able to provide mandatory locking.
3339 grace_disallows_io(struct net *net, struct inode *inode)
3341 return locks_in_grace(net) && mandatory_lock(inode);
3344 /* Returns true iff a is later than b: */
3345 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3347 return (s32)a->si_generation - (s32)b->si_generation > 0;
3350 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3353 * When sessions are used the stateid generation number is ignored
3356 if (has_session && in->si_generation == 0)
3359 if (in->si_generation == ref->si_generation)
3362 /* If the client sends us a stateid from the future, it's buggy: */
3363 if (stateid_generation_after(in, ref))
3364 return nfserr_bad_stateid;
3366 * However, we could see a stateid from the past, even from a
3367 * non-buggy client. For example, if the client sends a lock
3368 * while some IO is outstanding, the lock may bump si_generation
3369 * while the IO is still in flight. The client could avoid that
3370 * situation by waiting for responses on all the IO requests,
3371 * but better performance may result in retrying IO that
3372 * receives an old_stateid error if requests are rarely
3373 * reordered in flight:
3375 return nfserr_old_stateid;
3378 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3380 struct nfs4_stid *s;
3381 struct nfs4_ol_stateid *ols;
3384 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3385 return nfserr_bad_stateid;
3386 /* Client debugging aid. */
3387 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3388 char addr_str[INET6_ADDRSTRLEN];
3389 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3391 pr_warn_ratelimited("NFSD: client %s testing state ID "
3392 "with incorrect client ID\n", addr_str);
3393 return nfserr_bad_stateid;
3395 s = find_stateid(cl, stateid);
3397 return nfserr_bad_stateid;
3398 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3401 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3403 ols = openlockstateid(s);
3404 if (ols->st_stateowner->so_is_open_owner
3405 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3406 return nfserr_bad_stateid;
3410 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3411 struct nfs4_stid **s, bool sessions,
3412 struct nfsd_net *nn)
3414 struct nfs4_client *cl;
3416 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3417 return nfserr_bad_stateid;
3418 if (STALE_STATEID(stateid, nn))
3419 return nfserr_stale_stateid;
3420 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions, nn);
3422 return nfserr_expired;
3423 *s = find_stateid_by_type(cl, stateid, typemask);
3425 return nfserr_bad_stateid;
3431 * Checks for stateid operations
3434 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3435 stateid_t *stateid, int flags, struct file **filpp)
3437 struct nfs4_stid *s;
3438 struct nfs4_ol_stateid *stp = NULL;
3439 struct nfs4_delegation *dp = NULL;
3440 struct svc_fh *current_fh = &cstate->current_fh;
3441 struct inode *ino = current_fh->fh_dentry->d_inode;
3442 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3448 if (grace_disallows_io(net, ino))
3449 return nfserr_grace;
3451 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3452 return check_special_stateids(net, current_fh, stateid, flags);
3454 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3455 &s, cstate->minorversion, nn);
3458 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3461 switch (s->sc_type) {
3462 case NFS4_DELEG_STID:
3463 dp = delegstateid(s);
3464 status = nfs4_check_delegmode(dp, flags);
3468 *filpp = dp->dl_file->fi_deleg_file;
3472 case NFS4_OPEN_STID:
3473 case NFS4_LOCK_STID:
3474 stp = openlockstateid(s);
3475 status = nfs4_check_fh(current_fh, stp);
3478 if (stp->st_stateowner->so_is_open_owner
3479 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3481 status = nfs4_check_openmode(stp, flags);
3485 if (flags & RD_STATE)
3486 *filpp = find_readable_file(stp->st_file);
3488 *filpp = find_writeable_file(stp->st_file);
3492 return nfserr_bad_stateid;
3500 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3502 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3503 return nfserr_locks_held;
3504 release_lock_stateid(stp);
3509 * Test if the stateid is valid
3512 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3513 struct nfsd4_test_stateid *test_stateid)
3515 struct nfsd4_test_stateid_id *stateid;
3516 struct nfs4_client *cl = cstate->session->se_client;
3519 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3520 stateid->ts_id_status =
3521 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3522 nfs4_unlock_state();
3528 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3529 struct nfsd4_free_stateid *free_stateid)
3531 stateid_t *stateid = &free_stateid->fr_stateid;
3532 struct nfs4_stid *s;
3533 struct nfs4_client *cl = cstate->session->se_client;
3534 __be32 ret = nfserr_bad_stateid;
3537 s = find_stateid(cl, stateid);
3540 switch (s->sc_type) {
3541 case NFS4_DELEG_STID:
3542 ret = nfserr_locks_held;
3544 case NFS4_OPEN_STID:
3545 case NFS4_LOCK_STID:
3546 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3549 if (s->sc_type == NFS4_LOCK_STID)
3550 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3552 ret = nfserr_locks_held;
3555 ret = nfserr_bad_stateid;
3558 nfs4_unlock_state();
3565 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3566 RD_STATE : WR_STATE;
3569 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3571 struct svc_fh *current_fh = &cstate->current_fh;
3572 struct nfs4_stateowner *sop = stp->st_stateowner;
3575 status = nfsd4_check_seqid(cstate, sop, seqid);
3578 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3580 * "Closed" stateid's exist *only* to return
3581 * nfserr_replay_me from the previous step.
3583 return nfserr_bad_stateid;
3584 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3587 return nfs4_check_fh(current_fh, stp);
3591 * Checks for sequence id mutating operations.
3594 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3595 stateid_t *stateid, char typemask,
3596 struct nfs4_ol_stateid **stpp,
3597 struct nfsd_net *nn)
3600 struct nfs4_stid *s;
3602 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3603 seqid, STATEID_VAL(stateid));
3606 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3607 cstate->minorversion, nn);
3610 *stpp = openlockstateid(s);
3611 cstate->replay_owner = (*stpp)->st_stateowner;
3613 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3616 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3617 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3620 struct nfs4_openowner *oo;
3622 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3623 NFS4_OPEN_STID, stpp, nn);
3626 oo = openowner((*stpp)->st_stateowner);
3627 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3628 return nfserr_bad_stateid;
3633 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3634 struct nfsd4_open_confirm *oc)
3637 struct nfs4_openowner *oo;
3638 struct nfs4_ol_stateid *stp;
3639 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3641 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3642 (int)cstate->current_fh.fh_dentry->d_name.len,
3643 cstate->current_fh.fh_dentry->d_name.name);
3645 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3651 status = nfs4_preprocess_seqid_op(cstate,
3652 oc->oc_seqid, &oc->oc_req_stateid,
3653 NFS4_OPEN_STID, &stp, nn);
3656 oo = openowner(stp->st_stateowner);
3657 status = nfserr_bad_stateid;
3658 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3660 oo->oo_flags |= NFS4_OO_CONFIRMED;
3661 update_stateid(&stp->st_stid.sc_stateid);
3662 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3663 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3664 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3666 nfsd4_client_record_create(oo->oo_owner.so_client);
3669 if (!cstate->replay_owner)
3670 nfs4_unlock_state();
3674 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3676 if (!test_access(access, stp))
3678 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3679 clear_access(access, stp);
3682 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3684 switch (to_access) {
3685 case NFS4_SHARE_ACCESS_READ:
3686 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3687 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3689 case NFS4_SHARE_ACCESS_WRITE:
3690 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3691 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3693 case NFS4_SHARE_ACCESS_BOTH:
3701 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3704 for (i = 0; i < 4; i++) {
3705 if ((i & deny) != i)
3711 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3712 struct nfsd4_compound_state *cstate,
3713 struct nfsd4_open_downgrade *od)
3716 struct nfs4_ol_stateid *stp;
3717 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3719 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3720 (int)cstate->current_fh.fh_dentry->d_name.len,
3721 cstate->current_fh.fh_dentry->d_name.name);
3723 /* We don't yet support WANT bits: */
3724 if (od->od_deleg_want)
3725 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3729 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3730 &od->od_stateid, &stp, nn);
3733 status = nfserr_inval;
3734 if (!test_access(od->od_share_access, stp)) {
3735 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3736 stp->st_access_bmap, od->od_share_access);
3739 if (!test_deny(od->od_share_deny, stp)) {
3740 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3741 stp->st_deny_bmap, od->od_share_deny);
3744 nfs4_stateid_downgrade(stp, od->od_share_access);
3746 reset_union_bmap_deny(od->od_share_deny, stp);
3748 update_stateid(&stp->st_stid.sc_stateid);
3749 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3752 if (!cstate->replay_owner)
3753 nfs4_unlock_state();
3757 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3759 struct nfs4_openowner *oo;
3760 struct nfs4_ol_stateid *s;
3762 if (!so->so_is_open_owner)
3765 s = oo->oo_last_closed_stid;
3768 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3769 /* Release the last_closed_stid on the next seqid bump: */
3770 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3773 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3774 release_last_closed_stateid(oo);
3777 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3779 unhash_open_stateid(s);
3780 s->st_stid.sc_type = NFS4_CLOSED_STID;
3784 * nfs4_unlock_state() called after encode
3787 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3788 struct nfsd4_close *close)
3791 struct nfs4_openowner *oo;
3792 struct nfs4_ol_stateid *stp;
3793 struct net *net = SVC_NET(rqstp);
3794 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3796 dprintk("NFSD: nfsd4_close on file %.*s\n",
3797 (int)cstate->current_fh.fh_dentry->d_name.len,
3798 cstate->current_fh.fh_dentry->d_name.name);
3801 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3803 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3807 oo = openowner(stp->st_stateowner);
3809 update_stateid(&stp->st_stid.sc_stateid);
3810 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3812 nfsd4_close_open_stateid(stp);
3813 release_last_closed_stateid(oo);
3814 oo->oo_last_closed_stid = stp;
3816 if (list_empty(&oo->oo_owner.so_stateids)) {
3817 if (cstate->minorversion) {
3818 release_openowner(oo);
3819 cstate->replay_owner = NULL;
3822 * In the 4.0 case we need to keep the owners around a
3823 * little while to handle CLOSE replay.
3825 if (list_empty(&oo->oo_owner.so_stateids))
3826 move_to_close_lru(oo, SVC_NET(rqstp));
3830 if (!cstate->replay_owner)
3831 nfs4_unlock_state();
3836 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3837 struct nfsd4_delegreturn *dr)
3839 struct nfs4_delegation *dp;
3840 stateid_t *stateid = &dr->dr_stateid;
3841 struct nfs4_stid *s;
3843 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3845 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3849 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
3850 cstate->minorversion, nn);
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_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3871 end_offset(u64 start, u64 len)
3876 return end >= start ? end: NFS4_MAX_UINT64;
3879 /* last octet in a range */
3881 last_byte_offset(u64 start, u64 len)
3887 return end > start ? end - 1: NFS4_MAX_UINT64;
3890 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3892 return (file_hashval(inode) + cl_id
3893 + opaque_hashval(ownername->data, ownername->len))
3894 & LOCKOWNER_INO_HASH_MASK;
3898 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3899 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3900 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3901 * locking, this prevents us from being completely protocol-compliant. The
3902 * real solution to this problem is to start using unsigned file offsets in
3903 * the VFS, but this is a very deep change!
3906 nfs4_transform_lock_offset(struct file_lock *lock)
3908 if (lock->fl_start < 0)
3909 lock->fl_start = OFFSET_MAX;
3910 if (lock->fl_end < 0)
3911 lock->fl_end = OFFSET_MAX;
3914 /* Hack!: For now, we're defining this just so we can use a pointer to it
3915 * as a unique cookie to identify our (NFSv4's) posix locks. */
3916 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3920 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3922 struct nfs4_lockowner *lo;
3924 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3925 lo = (struct nfs4_lockowner *) fl->fl_owner;
3926 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3927 lo->lo_owner.so_owner.len, GFP_KERNEL);
3928 if (!deny->ld_owner.data)
3929 /* We just don't care that much */
3931 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3932 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3935 deny->ld_owner.len = 0;
3936 deny->ld_owner.data = NULL;
3937 deny->ld_clientid.cl_boot = 0;
3938 deny->ld_clientid.cl_id = 0;
3940 deny->ld_start = fl->fl_start;
3941 deny->ld_length = NFS4_MAX_UINT64;
3942 if (fl->fl_end != NFS4_MAX_UINT64)
3943 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3944 deny->ld_type = NFS4_READ_LT;
3945 if (fl->fl_type != F_RDLCK)
3946 deny->ld_type = NFS4_WRITE_LT;
3949 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3951 struct nfs4_ol_stateid *lst;
3953 if (!same_owner_str(&lo->lo_owner, owner, clid))
3955 lst = list_first_entry(&lo->lo_owner.so_stateids,
3956 struct nfs4_ol_stateid, st_perstateowner);
3957 return lst->st_file->fi_inode == inode;
3960 static struct nfs4_lockowner *
3961 find_lockowner_str(struct inode *inode, clientid_t *clid,
3962 struct xdr_netobj *owner, struct nfsd_net *nn)
3964 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3965 struct nfs4_lockowner *lo;
3967 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3968 if (same_lockowner_ino(lo, inode, clid, owner))
3974 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3976 struct inode *inode = open_stp->st_file->fi_inode;
3977 unsigned int inohash = lockowner_ino_hashval(inode,
3978 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3979 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
3981 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
3982 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
3983 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3987 * Alloc a lock owner structure.
3988 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3991 * strhashval = ownerstr_hashval
3994 static struct nfs4_lockowner *
3995 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3996 struct nfs4_lockowner *lo;
3998 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4001 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4002 lo->lo_owner.so_is_open_owner = 0;
4003 /* It is the openowner seqid that will be incremented in encode in the
4004 * case of new lockowners; so increment the lock seqid manually: */
4005 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4006 hash_lockowner(lo, strhashval, clp, open_stp);
4010 static struct nfs4_ol_stateid *
4011 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4013 struct nfs4_ol_stateid *stp;
4014 struct nfs4_client *clp = lo->lo_owner.so_client;
4016 stp = nfs4_alloc_stateid(clp);
4019 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
4020 list_add(&stp->st_perfile, &fp->fi_stateids);
4021 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4022 stp->st_stateowner = &lo->lo_owner;
4025 stp->st_access_bmap = 0;
4026 stp->st_deny_bmap = open_stp->st_deny_bmap;
4027 stp->st_openstp = open_stp;
4032 check_lock_length(u64 offset, u64 length)
4034 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4035 LOFF_OVERFLOW(offset, length)));
4038 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4040 struct nfs4_file *fp = lock_stp->st_file;
4041 int oflag = nfs4_access_to_omode(access);
4043 if (test_access(access, lock_stp))
4045 nfs4_file_get_access(fp, oflag);
4046 set_access(access, lock_stp);
4049 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)
4051 struct nfs4_file *fi = ost->st_file;
4052 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4053 struct nfs4_client *cl = oo->oo_owner.so_client;
4054 struct nfs4_lockowner *lo;
4055 unsigned int strhashval;
4056 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4058 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4059 &lock->v.new.owner, nn);
4061 if (!cstate->minorversion)
4062 return nfserr_bad_seqid;
4063 /* XXX: a lockowner always has exactly one stateid: */
4064 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4065 struct nfs4_ol_stateid, st_perstateowner);
4068 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4069 &lock->v.new.owner);
4070 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4072 return nfserr_jukebox;
4073 *lst = alloc_init_lock_stateid(lo, fi, ost);
4075 release_lockowner(lo);
4076 return nfserr_jukebox;
4086 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4087 struct nfsd4_lock *lock)
4089 struct nfs4_openowner *open_sop = NULL;
4090 struct nfs4_lockowner *lock_sop = NULL;
4091 struct nfs4_ol_stateid *lock_stp;
4092 struct file *filp = NULL;
4093 struct file_lock *file_lock = NULL;
4094 struct file_lock *conflock = NULL;
4096 bool new_state = false;
4099 struct net *net = SVC_NET(rqstp);
4100 struct nfsd_net *nn = net_generic(net, 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, nn);
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(net) && !lock->lk_reclaim)
4161 status = nfserr_no_grace;
4162 if (!locks_in_grace(net) && 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, nn);
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;
4348 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4350 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4351 (long long) locku->lu_offset,
4352 (long long) locku->lu_length);
4354 if (check_lock_length(locku->lu_offset, locku->lu_length))
4355 return nfserr_inval;
4359 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4360 &locku->lu_stateid, NFS4_LOCK_STID,
4364 filp = find_any_file(stp->st_file);
4366 status = nfserr_lock_range;
4369 file_lock = locks_alloc_lock();
4371 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4372 status = nfserr_jukebox;
4375 locks_init_lock(file_lock);
4376 file_lock->fl_type = F_UNLCK;
4377 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4378 file_lock->fl_pid = current->tgid;
4379 file_lock->fl_file = filp;
4380 file_lock->fl_flags = FL_POSIX;
4381 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4382 file_lock->fl_start = locku->lu_offset;
4384 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4386 nfs4_transform_lock_offset(file_lock);
4389 * Try to unlock the file in the VFS.
4391 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4393 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4397 * OK, unlock succeeded; the only thing left to do is update the stateid.
4399 update_stateid(&stp->st_stid.sc_stateid);
4400 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4403 if (!cstate->replay_owner)
4404 nfs4_unlock_state();
4406 locks_free_lock(file_lock);
4410 status = nfserrno(err);
4416 * 1: locks held by lockowner
4417 * 0: no locks held by lockowner
4420 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4422 struct file_lock **flpp;
4423 struct inode *inode = filp->fi_inode;
4427 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4428 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4439 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4440 struct nfsd4_compound_state *cstate,
4441 struct nfsd4_release_lockowner *rlockowner)
4443 clientid_t *clid = &rlockowner->rl_clientid;
4444 struct nfs4_stateowner *sop;
4445 struct nfs4_lockowner *lo;
4446 struct nfs4_ol_stateid *stp;
4447 struct xdr_netobj *owner = &rlockowner->rl_owner;
4448 struct list_head matches;
4449 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4451 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4453 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4454 clid->cl_boot, clid->cl_id);
4456 /* XXX check for lease expiration */
4458 status = nfserr_stale_clientid;
4459 if (STALE_CLIENTID(clid, nn))
4464 status = nfserr_locks_held;
4465 INIT_LIST_HEAD(&matches);
4467 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4468 if (sop->so_is_open_owner)
4470 if (!same_owner_str(sop, owner, clid))
4472 list_for_each_entry(stp, &sop->so_stateids,
4474 lo = lockowner(sop);
4475 if (check_for_locks(stp->st_file, lo))
4477 list_add(&lo->lo_list, &matches);
4480 /* Clients probably won't expect us to return with some (but not all)
4481 * of the lockowner state released; so don't release any until all
4482 * have been checked. */
4484 while (!list_empty(&matches)) {
4485 lo = list_entry(matches.next, struct nfs4_lockowner,
4487 /* unhash_stateowner deletes so_perclient only
4488 * for openowners. */
4489 list_del(&lo->lo_list);
4490 release_lockowner(lo);
4493 nfs4_unlock_state();
4497 static inline struct nfs4_client_reclaim *
4500 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4504 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4506 struct nfs4_client_reclaim *crp;
4508 crp = nfsd4_find_reclaim_client(name, nn);
4509 return (crp && crp->cr_clp);
4513 * failure => all reset bets are off, nfserr_no_grace...
4515 struct nfs4_client_reclaim *
4516 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4518 unsigned int strhashval;
4519 struct nfs4_client_reclaim *crp;
4521 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4522 crp = alloc_reclaim();
4524 strhashval = clientstr_hashval(name);
4525 INIT_LIST_HEAD(&crp->cr_strhash);
4526 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4527 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4529 nn->reclaim_str_hashtbl_size++;
4535 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4537 list_del(&crp->cr_strhash);
4539 nn->reclaim_str_hashtbl_size--;
4543 nfs4_release_reclaim(struct nfsd_net *nn)
4545 struct nfs4_client_reclaim *crp = NULL;
4548 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4549 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4550 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4551 struct nfs4_client_reclaim, cr_strhash);
4552 nfs4_remove_reclaim_record(crp, nn);
4555 BUG_ON(nn->reclaim_str_hashtbl_size);
4559 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4560 struct nfs4_client_reclaim *
4561 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4563 unsigned int strhashval;
4564 struct nfs4_client_reclaim *crp = NULL;
4566 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4568 strhashval = clientstr_hashval(recdir);
4569 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4570 if (same_name(crp->cr_recdir, recdir)) {
4578 * Called from OPEN. Look for clientid in reclaim list.
4581 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4583 struct nfs4_client *clp;
4585 /* find clientid in conf_id_hashtbl */
4586 clp = find_confirmed_client(clid, sessions, nn);
4588 return nfserr_reclaim_bad;
4590 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4593 #ifdef CONFIG_NFSD_FAULT_INJECTION
4595 void nfsd_forget_clients(u64 num)
4597 struct nfs4_client *clp, *next;
4599 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4602 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4607 nfs4_unlock_state();
4609 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4612 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4614 release_lockowner(lockowner(sop));
4617 static void release_openowner_sop(struct nfs4_stateowner *sop)
4619 release_openowner(openowner(sop));
4622 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4623 void (*release_sop)(struct nfs4_stateowner *),
4624 struct nfsd_net *nn)
4627 struct nfs4_stateowner *sop, *next;
4629 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4630 list_for_each_entry_safe(sop, next, &nn->ownerstr_hashtbl[i], so_strhash) {
4631 if (sop->so_is_open_owner != is_open_owner)
4641 void nfsd_forget_locks(u64 num)
4644 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4647 count = nfsd_release_n_owners(num, false, release_lockowner_sop, nn);
4648 nfs4_unlock_state();
4650 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4653 void nfsd_forget_openowners(u64 num)
4656 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4659 count = nfsd_release_n_owners(num, true, release_openowner_sop, nn);
4660 nfs4_unlock_state();
4662 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4665 static int nfsd_process_n_delegations(u64 num, struct list_head *list)
4668 struct nfs4_file *fp, *fnext;
4669 struct nfs4_delegation *dp, *dnext;
4671 for (i = 0; i < FILE_HASH_SIZE; i++) {
4672 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4673 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4674 list_move(&dp->dl_recall_lru, list);
4684 void nfsd_forget_delegations(u64 num)
4688 struct nfs4_delegation *dp, *dnext;
4690 spin_lock(&recall_lock);
4691 count = nfsd_process_n_delegations(num, &victims);
4692 spin_unlock(&recall_lock);
4695 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
4696 unhash_delegation(dp);
4697 nfs4_unlock_state();
4699 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4702 void nfsd_recall_delegations(u64 num)
4706 struct nfs4_delegation *dp, *dnext;
4708 spin_lock(&recall_lock);
4709 count = nfsd_process_n_delegations(num, &victims);
4710 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
4711 list_del(&dp->dl_recall_lru);
4712 nfsd_break_one_deleg(dp);
4714 spin_unlock(&recall_lock);
4716 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4719 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4721 /* initialization to perform at module load time: */
4724 nfs4_state_init(void)
4728 for (i = 0; i < FILE_HASH_SIZE; i++) {
4729 INIT_LIST_HEAD(&file_hashtbl[i]);
4731 INIT_LIST_HEAD(&del_recall_lru);
4735 * Since the lifetime of a delegation isn't limited to that of an open, a
4736 * client may quite reasonably hang on to a delegation as long as it has
4737 * the inode cached. This becomes an obvious problem the first time a
4738 * client's inode cache approaches the size of the server's total memory.
4740 * For now we avoid this problem by imposing a hard limit on the number
4741 * of delegations, which varies according to the server's memory size.
4744 set_max_delegations(void)
4747 * Allow at most 4 delegations per megabyte of RAM. Quick
4748 * estimates suggest that in the worst case (where every delegation
4749 * is for a different inode), a delegation could take about 1.5K,
4750 * giving a worst case usage of about 6% of memory.
4752 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4755 static int nfs4_state_start_net(struct net *net)
4757 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4760 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4761 CLIENT_HASH_SIZE, GFP_KERNEL);
4762 if (!nn->conf_id_hashtbl)
4764 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4765 CLIENT_HASH_SIZE, GFP_KERNEL);
4766 if (!nn->unconf_id_hashtbl)
4768 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
4769 OWNER_HASH_SIZE, GFP_KERNEL);
4770 if (!nn->ownerstr_hashtbl)
4772 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
4773 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
4774 if (!nn->lockowner_ino_hashtbl)
4775 goto err_lockowner_ino;
4776 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
4777 SESSION_HASH_SIZE, GFP_KERNEL);
4778 if (!nn->sessionid_hashtbl)
4781 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4782 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
4783 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
4785 for (i = 0; i < OWNER_HASH_SIZE; i++)
4786 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
4787 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4788 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
4789 for (i = 0; i < SESSION_HASH_SIZE; i++)
4790 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
4791 nn->conf_name_tree = RB_ROOT;
4792 nn->unconf_name_tree = RB_ROOT;
4793 INIT_LIST_HEAD(&nn->client_lru);
4794 INIT_LIST_HEAD(&nn->close_lru);
4799 kfree(nn->lockowner_ino_hashtbl);
4801 kfree(nn->ownerstr_hashtbl);
4803 kfree(nn->unconf_id_hashtbl);
4805 kfree(nn->conf_id_hashtbl);
4811 __nfs4_state_shutdown_net(struct net *net)
4814 struct nfs4_client *clp = NULL;
4815 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4816 struct rb_node *node, *tmp;
4818 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4819 while (!list_empty(&nn->conf_id_hashtbl[i])) {
4820 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4821 destroy_client(clp);
4825 node = rb_first(&nn->unconf_name_tree);
4826 while (node != NULL) {
4828 node = rb_next(tmp);
4829 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4830 rb_erase(tmp, &nn->unconf_name_tree);
4831 destroy_client(clp);
4834 kfree(nn->sessionid_hashtbl);
4835 kfree(nn->lockowner_ino_hashtbl);
4836 kfree(nn->ownerstr_hashtbl);
4837 kfree(nn->unconf_id_hashtbl);
4838 kfree(nn->conf_id_hashtbl);
4841 /* initialization to perform when the nfsd service is started: */
4844 nfs4_state_start(void)
4846 struct net *net = &init_net;
4847 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4851 * FIXME: For now, we hang most of the pernet global stuff off of
4852 * init_net until nfsd is fully containerized. Eventually, we'll
4853 * need to pass a net pointer into this function, take a reference
4854 * to that instead and then do most of the rest of this on a per-net
4858 ret = nfs4_state_start_net(net);
4861 nfsd4_client_tracking_init(net);
4862 nn->boot_time = get_seconds();
4863 locks_start_grace(net, &nn->nfsd4_manager);
4864 nn->grace_ended = false;
4865 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4867 ret = set_callback_cred();
4872 laundry_wq = create_singlethread_workqueue("nfsd4");
4873 if (laundry_wq == NULL) {
4877 ret = nfsd4_create_callback_queue();
4879 goto out_free_laundry;
4880 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4881 set_max_delegations();
4884 destroy_workqueue(laundry_wq);
4886 nfsd4_client_tracking_exit(net);
4887 __nfs4_state_shutdown_net(net);
4892 /* should be called with the state lock held */
4894 __nfs4_state_shutdown(struct net *net)
4896 struct nfs4_delegation *dp = NULL;
4897 struct list_head *pos, *next, reaplist;
4899 __nfs4_state_shutdown_net(net);
4901 INIT_LIST_HEAD(&reaplist);
4902 spin_lock(&recall_lock);
4903 list_for_each_safe(pos, next, &del_recall_lru) {
4904 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4905 list_move(&dp->dl_recall_lru, &reaplist);
4907 spin_unlock(&recall_lock);
4908 list_for_each_safe(pos, next, &reaplist) {
4909 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4910 unhash_delegation(dp);
4913 nfsd4_client_tracking_exit(net);
4918 nfs4_state_shutdown(void)
4920 struct net *net = &init_net;
4921 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4923 cancel_delayed_work_sync(&laundromat_work);
4924 destroy_workqueue(laundry_wq);
4925 locks_end_grace(&nn->nfsd4_manager);
4927 __nfs4_state_shutdown(net);
4928 nfs4_unlock_state();
4929 nfsd4_destroy_callback_queue();
4933 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4935 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4936 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4940 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4942 if (cstate->minorversion) {
4943 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4944 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4949 clear_current_stateid(struct nfsd4_compound_state *cstate)
4951 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4955 * functions to set current state id
4958 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4960 put_stateid(cstate, &odp->od_stateid);
4964 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4966 put_stateid(cstate, &open->op_stateid);
4970 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4972 put_stateid(cstate, &close->cl_stateid);
4976 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4978 put_stateid(cstate, &lock->lk_resp_stateid);
4982 * functions to consume current state id
4986 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4988 get_stateid(cstate, &odp->od_stateid);
4992 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4994 get_stateid(cstate, &drp->dr_stateid);
4998 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5000 get_stateid(cstate, &fsp->fr_stateid);
5004 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5006 get_stateid(cstate, &setattr->sa_stateid);
5010 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5012 get_stateid(cstate, &close->cl_stateid);
5016 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5018 get_stateid(cstate, &locku->lu_stateid);
5022 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5024 get_stateid(cstate, &read->rd_stateid);
5028 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5030 get_stateid(cstate, &write->wr_stateid);