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
53 #define all_ones {{~0,~0},~0}
54 static const stateid_t one_stateid = {
56 .si_opaque = all_ones,
58 static const stateid_t zero_stateid = {
61 static const stateid_t currentstateid = {
65 static u64 current_sessionid = 1;
67 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
68 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
69 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
71 /* forward declarations */
72 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
76 /* Currently used for almost all code touching nfsv4 state: */
77 static DEFINE_MUTEX(client_mutex);
80 * Currently used for the del_recall_lru and file hash table. In an
81 * effort to decrease the scope of the client_mutex, this spinlock may
82 * eventually cover more:
84 static DEFINE_SPINLOCK(recall_lock);
86 static struct kmem_cache *openowner_slab = NULL;
87 static struct kmem_cache *lockowner_slab = NULL;
88 static struct kmem_cache *file_slab = NULL;
89 static struct kmem_cache *stateid_slab = NULL;
90 static struct kmem_cache *deleg_slab = NULL;
95 mutex_lock(&client_mutex);
98 static void free_session(struct kref *);
100 /* Must be called under the client_lock */
101 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
103 kref_put(&ses->se_ref, free_session);
106 static void nfsd4_get_session(struct nfsd4_session *ses)
108 kref_get(&ses->se_ref);
112 nfs4_unlock_state(void)
114 mutex_unlock(&client_mutex);
118 opaque_hashval(const void *ptr, int nbytes)
120 unsigned char *cptr = (unsigned char *) ptr;
130 static struct list_head del_recall_lru;
132 static void nfsd4_free_file(struct nfs4_file *f)
134 kmem_cache_free(file_slab, f);
138 put_nfs4_file(struct nfs4_file *fi)
140 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
141 list_del(&fi->fi_hash);
142 spin_unlock(&recall_lock);
149 get_nfs4_file(struct nfs4_file *fi)
151 atomic_inc(&fi->fi_ref);
154 static int num_delegations;
155 unsigned int max_delegations;
158 * Open owner state (share locks)
161 /* hash tables for lock and open owners */
162 #define OWNER_HASH_BITS 8
163 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
164 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
166 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
170 ret = opaque_hashval(ownername->data, ownername->len);
172 return ret & OWNER_HASH_MASK;
175 /* hash table for nfs4_file */
176 #define FILE_HASH_BITS 8
177 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
179 static unsigned int file_hashval(struct inode *ino)
181 /* XXX: why are we hashing on inode pointer, anyway? */
182 return hash_ptr(ino, FILE_HASH_BITS);
185 static struct list_head file_hashtbl[FILE_HASH_SIZE];
187 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
189 WARN_ON_ONCE(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
190 atomic_inc(&fp->fi_access[oflag]);
193 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
195 if (oflag == O_RDWR) {
196 __nfs4_file_get_access(fp, O_RDONLY);
197 __nfs4_file_get_access(fp, O_WRONLY);
199 __nfs4_file_get_access(fp, oflag);
202 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
204 if (fp->fi_fds[oflag]) {
205 fput(fp->fi_fds[oflag]);
206 fp->fi_fds[oflag] = NULL;
210 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
212 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
213 nfs4_file_put_fd(fp, oflag);
215 * It's also safe to get rid of the RDWR open *if*
216 * we no longer have need of the other kind of access
217 * or if we already have the other kind of open:
219 if (fp->fi_fds[1-oflag]
220 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
221 nfs4_file_put_fd(fp, O_RDWR);
225 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
227 if (oflag == O_RDWR) {
228 __nfs4_file_put_access(fp, O_RDONLY);
229 __nfs4_file_put_access(fp, O_WRONLY);
231 __nfs4_file_put_access(fp, oflag);
234 static inline int get_new_stid(struct nfs4_stid *stid)
236 static int min_stateid = 0;
237 struct idr *stateids = &stid->sc_client->cl_stateids;
241 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
243 * Note: the necessary preallocation was done in
244 * nfs4_alloc_stateid(). The idr code caps the number of
245 * preallocations that can exist at a time, but the state lock
246 * prevents anyone from using ours before we get here:
250 * It shouldn't be a problem to reuse an opaque stateid value.
251 * I don't think it is for 4.1. But with 4.0 I worry that, for
252 * example, a stray write retransmission could be accepted by
253 * the server when it should have been rejected. Therefore,
254 * adopt a trick from the sctp code to attempt to maximize the
255 * amount of time until an id is reused, by ensuring they always
256 * "increase" (mod INT_MAX):
259 min_stateid = new_stid+1;
260 if (min_stateid == INT_MAX)
265 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
267 stateid_t *s = &stid->sc_stateid;
270 stid->sc_type = type;
271 stid->sc_client = cl;
272 s->si_opaque.so_clid = cl->cl_clientid;
273 new_id = get_new_stid(stid);
274 s->si_opaque.so_id = (u32)new_id;
275 /* Will be incremented before return to client: */
276 s->si_generation = 0;
279 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
281 struct idr *stateids = &cl->cl_stateids;
283 if (!idr_pre_get(stateids, GFP_KERNEL))
286 * Note: if we fail here (or any time between now and the time
287 * we actually get the new idr), we won't need to undo the idr
288 * preallocation, since the idr code caps the number of
289 * preallocated entries.
291 return kmem_cache_alloc(slab, GFP_KERNEL);
294 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
296 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
299 static struct nfs4_delegation *
300 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
302 struct nfs4_delegation *dp;
303 struct nfs4_file *fp = stp->st_file;
305 dprintk("NFSD alloc_init_deleg\n");
307 * Major work on the lease subsystem (for example, to support
308 * calbacks on stat) will be required before we can support
309 * write delegations properly.
311 if (type != NFS4_OPEN_DELEGATE_READ)
313 if (fp->fi_had_conflict)
315 if (num_delegations > max_delegations)
317 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
320 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
322 * delegation seqid's are never incremented. The 4.1 special
323 * meaning of seqid 0 isn't meaningful, really, but let's avoid
324 * 0 anyway just for consistency and use 1:
326 dp->dl_stid.sc_stateid.si_generation = 1;
328 INIT_LIST_HEAD(&dp->dl_perfile);
329 INIT_LIST_HEAD(&dp->dl_perclnt);
330 INIT_LIST_HEAD(&dp->dl_recall_lru);
334 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
336 atomic_set(&dp->dl_count, 1);
337 nfsd4_init_callback(&dp->dl_recall);
342 nfs4_put_delegation(struct nfs4_delegation *dp)
344 if (atomic_dec_and_test(&dp->dl_count)) {
345 dprintk("NFSD: freeing dp %p\n",dp);
346 put_nfs4_file(dp->dl_file);
347 kmem_cache_free(deleg_slab, dp);
352 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
354 if (atomic_dec_and_test(&fp->fi_delegees)) {
355 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
357 fput(fp->fi_deleg_file);
358 fp->fi_deleg_file = NULL;
362 static void unhash_stid(struct nfs4_stid *s)
364 struct idr *stateids = &s->sc_client->cl_stateids;
366 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
369 /* Called under the state lock. */
371 unhash_delegation(struct nfs4_delegation *dp)
373 unhash_stid(&dp->dl_stid);
374 list_del_init(&dp->dl_perclnt);
375 spin_lock(&recall_lock);
376 list_del_init(&dp->dl_perfile);
377 list_del_init(&dp->dl_recall_lru);
378 spin_unlock(&recall_lock);
379 nfs4_put_deleg_lease(dp->dl_file);
380 nfs4_put_delegation(dp);
387 static unsigned int clientid_hashval(u32 id)
389 return id & CLIENT_HASH_MASK;
392 static unsigned int clientstr_hashval(const char *name)
394 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
398 * We store the NONE, READ, WRITE, and BOTH bits separately in the
399 * st_{access,deny}_bmap field of the stateid, in order to track not
400 * only what share bits are currently in force, but also what
401 * combinations of share bits previous opens have used. This allows us
402 * to enforce the recommendation of rfc 3530 14.2.19 that the server
403 * return an error if the client attempt to downgrade to a combination
404 * of share bits not explicable by closing some of its previous opens.
406 * XXX: This enforcement is actually incomplete, since we don't keep
407 * track of access/deny bit combinations; so, e.g., we allow:
409 * OPEN allow read, deny write
410 * OPEN allow both, deny none
411 * DOWNGRADE allow read, deny none
413 * which we should reject.
416 bmap_to_share_mode(unsigned long bmap) {
418 unsigned int access = 0;
420 for (i = 1; i < 4; i++) {
421 if (test_bit(i, &bmap))
428 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
429 unsigned int access, deny;
431 access = bmap_to_share_mode(stp->st_access_bmap);
432 deny = bmap_to_share_mode(stp->st_deny_bmap);
433 if ((access & open->op_share_deny) || (deny & open->op_share_access))
438 /* set share access for a given stateid */
440 set_access(u32 access, struct nfs4_ol_stateid *stp)
442 __set_bit(access, &stp->st_access_bmap);
445 /* clear share access for a given stateid */
447 clear_access(u32 access, struct nfs4_ol_stateid *stp)
449 __clear_bit(access, &stp->st_access_bmap);
452 /* test whether a given stateid has access */
454 test_access(u32 access, struct nfs4_ol_stateid *stp)
456 return test_bit(access, &stp->st_access_bmap);
459 /* set share deny for a given stateid */
461 set_deny(u32 access, struct nfs4_ol_stateid *stp)
463 __set_bit(access, &stp->st_deny_bmap);
466 /* clear share deny for a given stateid */
468 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
470 __clear_bit(access, &stp->st_deny_bmap);
473 /* test whether a given stateid is denying specific access */
475 test_deny(u32 access, struct nfs4_ol_stateid *stp)
477 return test_bit(access, &stp->st_deny_bmap);
480 static int nfs4_access_to_omode(u32 access)
482 switch (access & NFS4_SHARE_ACCESS_BOTH) {
483 case NFS4_SHARE_ACCESS_READ:
485 case NFS4_SHARE_ACCESS_WRITE:
487 case NFS4_SHARE_ACCESS_BOTH:
494 /* release all access and file references for a given stateid */
496 release_all_access(struct nfs4_ol_stateid *stp)
500 for (i = 1; i < 4; i++) {
501 if (test_access(i, stp))
502 nfs4_file_put_access(stp->st_file,
503 nfs4_access_to_omode(i));
504 clear_access(i, stp);
508 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
510 list_del(&stp->st_perfile);
511 list_del(&stp->st_perstateowner);
514 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
516 release_all_access(stp);
517 put_nfs4_file(stp->st_file);
521 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
523 kmem_cache_free(stateid_slab, stp);
526 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
530 unhash_generic_stateid(stp);
531 unhash_stid(&stp->st_stid);
532 file = find_any_file(stp->st_file);
534 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
535 close_generic_stateid(stp);
536 free_generic_stateid(stp);
539 static void unhash_lockowner(struct nfs4_lockowner *lo)
541 struct nfs4_ol_stateid *stp;
543 list_del(&lo->lo_owner.so_strhash);
544 list_del(&lo->lo_perstateid);
545 list_del(&lo->lo_owner_ino_hash);
546 while (!list_empty(&lo->lo_owner.so_stateids)) {
547 stp = list_first_entry(&lo->lo_owner.so_stateids,
548 struct nfs4_ol_stateid, st_perstateowner);
549 release_lock_stateid(stp);
553 static void release_lockowner(struct nfs4_lockowner *lo)
555 unhash_lockowner(lo);
556 nfs4_free_lockowner(lo);
560 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
562 struct nfs4_lockowner *lo;
564 while (!list_empty(&open_stp->st_lockowners)) {
565 lo = list_entry(open_stp->st_lockowners.next,
566 struct nfs4_lockowner, lo_perstateid);
567 release_lockowner(lo);
571 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
573 unhash_generic_stateid(stp);
574 release_stateid_lockowners(stp);
575 close_generic_stateid(stp);
578 static void release_open_stateid(struct nfs4_ol_stateid *stp)
580 unhash_open_stateid(stp);
581 unhash_stid(&stp->st_stid);
582 free_generic_stateid(stp);
585 static void unhash_openowner(struct nfs4_openowner *oo)
587 struct nfs4_ol_stateid *stp;
589 list_del(&oo->oo_owner.so_strhash);
590 list_del(&oo->oo_perclient);
591 while (!list_empty(&oo->oo_owner.so_stateids)) {
592 stp = list_first_entry(&oo->oo_owner.so_stateids,
593 struct nfs4_ol_stateid, st_perstateowner);
594 release_open_stateid(stp);
598 static void release_last_closed_stateid(struct nfs4_openowner *oo)
600 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
603 unhash_stid(&s->st_stid);
604 free_generic_stateid(s);
605 oo->oo_last_closed_stid = NULL;
609 static void release_openowner(struct nfs4_openowner *oo)
611 unhash_openowner(oo);
612 list_del(&oo->oo_close_lru);
613 release_last_closed_stateid(oo);
614 nfs4_free_openowner(oo);
618 hash_sessionid(struct nfs4_sessionid *sessionid)
620 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
622 return sid->sequence % SESSION_HASH_SIZE;
627 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
629 u32 *ptr = (u32 *)(&sessionid->data[0]);
630 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
634 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
641 gen_sessionid(struct nfsd4_session *ses)
643 struct nfs4_client *clp = ses->se_client;
644 struct nfsd4_sessionid *sid;
646 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
647 sid->clientid = clp->cl_clientid;
648 sid->sequence = current_sessionid++;
653 * The protocol defines ca_maxresponssize_cached to include the size of
654 * the rpc header, but all we need to cache is the data starting after
655 * the end of the initial SEQUENCE operation--the rest we regenerate
656 * each time. Therefore we can advertise a ca_maxresponssize_cached
657 * value that is the number of bytes in our cache plus a few additional
658 * bytes. In order to stay on the safe side, and not promise more than
659 * we can cache, those additional bytes must be the minimum possible: 24
660 * bytes of rpc header (xid through accept state, with AUTH_NULL
661 * verifier), 12 for the compound header (with zero-length tag), and 44
662 * for the SEQUENCE op response:
664 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
667 free_session_slots(struct nfsd4_session *ses)
671 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
672 kfree(ses->se_slots[i]);
676 * We don't actually need to cache the rpc and session headers, so we
677 * can allocate a little less for each slot:
679 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
681 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
684 static int nfsd4_sanitize_slot_size(u32 size)
686 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
687 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
693 * XXX: If we run out of reserved DRC memory we could (up to a point)
694 * re-negotiate active sessions and reduce their slot usage to make
695 * room for new connections. For now we just fail the create session.
697 static int nfsd4_get_drc_mem(int slotsize, u32 num)
701 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
703 spin_lock(&nfsd_drc_lock);
704 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
705 nfsd_drc_max_mem - nfsd_drc_mem_used);
706 num = min_t(int, num, avail / slotsize);
707 nfsd_drc_mem_used += num * slotsize;
708 spin_unlock(&nfsd_drc_lock);
713 static void nfsd4_put_drc_mem(int slotsize, int num)
715 spin_lock(&nfsd_drc_lock);
716 nfsd_drc_mem_used -= slotsize * num;
717 spin_unlock(&nfsd_drc_lock);
720 static struct nfsd4_session *__alloc_session(int slotsize, int numslots)
722 struct nfsd4_session *new;
725 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
726 + sizeof(struct nfsd4_session) > PAGE_SIZE);
727 mem = numslots * sizeof(struct nfsd4_slot *);
729 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
732 /* allocate each struct nfsd4_slot and data cache in one piece */
733 for (i = 0; i < numslots; i++) {
734 mem = sizeof(struct nfsd4_slot) + slotsize;
735 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
736 if (!new->se_slots[i])
742 kfree(new->se_slots[i]);
747 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
749 u32 maxrpc = nfsd_serv->sv_max_mesg;
751 new->maxreqs = numslots;
752 new->maxresp_cached = min_t(u32, req->maxresp_cached,
753 slotsize + NFSD_MIN_HDR_SEQ_SZ);
754 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
755 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
756 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
759 static void free_conn(struct nfsd4_conn *c)
761 svc_xprt_put(c->cn_xprt);
765 static void nfsd4_conn_lost(struct svc_xpt_user *u)
767 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
768 struct nfs4_client *clp = c->cn_session->se_client;
770 spin_lock(&clp->cl_lock);
771 if (!list_empty(&c->cn_persession)) {
772 list_del(&c->cn_persession);
775 spin_unlock(&clp->cl_lock);
776 nfsd4_probe_callback(clp);
779 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
781 struct nfsd4_conn *conn;
783 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
786 svc_xprt_get(rqstp->rq_xprt);
787 conn->cn_xprt = rqstp->rq_xprt;
788 conn->cn_flags = flags;
789 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
793 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
795 conn->cn_session = ses;
796 list_add(&conn->cn_persession, &ses->se_conns);
799 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
801 struct nfs4_client *clp = ses->se_client;
803 spin_lock(&clp->cl_lock);
804 __nfsd4_hash_conn(conn, ses);
805 spin_unlock(&clp->cl_lock);
808 static int nfsd4_register_conn(struct nfsd4_conn *conn)
810 conn->cn_xpt_user.callback = nfsd4_conn_lost;
811 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
814 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
818 nfsd4_hash_conn(conn, ses);
819 ret = nfsd4_register_conn(conn);
821 /* oops; xprt is already down: */
822 nfsd4_conn_lost(&conn->cn_xpt_user);
823 if (conn->cn_flags & NFS4_CDFC4_BACK) {
824 /* callback channel may be back up */
825 nfsd4_probe_callback(ses->se_client);
829 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
831 u32 dir = NFS4_CDFC4_FORE;
833 if (cses->flags & SESSION4_BACK_CHAN)
834 dir |= NFS4_CDFC4_BACK;
835 return alloc_conn(rqstp, dir);
838 /* must be called under client_lock */
839 static void nfsd4_del_conns(struct nfsd4_session *s)
841 struct nfs4_client *clp = s->se_client;
842 struct nfsd4_conn *c;
844 spin_lock(&clp->cl_lock);
845 while (!list_empty(&s->se_conns)) {
846 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
847 list_del_init(&c->cn_persession);
848 spin_unlock(&clp->cl_lock);
850 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
853 spin_lock(&clp->cl_lock);
855 spin_unlock(&clp->cl_lock);
858 static void __free_session(struct nfsd4_session *ses)
860 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
861 free_session_slots(ses);
865 static void free_session(struct kref *kref)
867 struct nfsd4_session *ses;
870 ses = container_of(kref, struct nfsd4_session, se_ref);
871 nn = net_generic(ses->se_client->net, nfsd_net_id);
873 lockdep_assert_held(&nn->client_lock);
874 nfsd4_del_conns(ses);
878 void nfsd4_put_session(struct nfsd4_session *ses)
880 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
882 spin_lock(&nn->client_lock);
883 nfsd4_put_session_locked(ses);
884 spin_unlock(&nn->client_lock);
887 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fchan)
889 struct nfsd4_session *new;
890 int numslots, slotsize;
892 * Note decreasing slot size below client's request may
893 * make it difficult for client to function correctly, whereas
894 * decreasing the number of slots will (just?) affect
895 * performance. When short on memory we therefore prefer to
896 * decrease number of slots instead of their size.
898 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
899 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
903 new = __alloc_session(slotsize, numslots);
905 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
908 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
912 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
915 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
917 new->se_client = clp;
920 INIT_LIST_HEAD(&new->se_conns);
922 new->se_cb_seq_nr = 1;
923 new->se_flags = cses->flags;
924 new->se_cb_prog = cses->callback_prog;
925 new->se_cb_sec = cses->cb_sec;
926 kref_init(&new->se_ref);
927 idx = hash_sessionid(&new->se_sessionid);
928 spin_lock(&nn->client_lock);
929 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
930 spin_lock(&clp->cl_lock);
931 list_add(&new->se_perclnt, &clp->cl_sessions);
932 spin_unlock(&clp->cl_lock);
933 spin_unlock(&nn->client_lock);
935 if (cses->flags & SESSION4_BACK_CHAN) {
936 struct sockaddr *sa = svc_addr(rqstp);
938 * This is a little silly; with sessions there's no real
939 * use for the callback address. Use the peer address
940 * as a reasonable default for now, but consider fixing
941 * the rpc client not to require an address in the
944 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
945 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
949 /* caller must hold client_lock */
950 static struct nfsd4_session *
951 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
953 struct nfsd4_session *elem;
955 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
957 dump_sessionid(__func__, sessionid);
958 idx = hash_sessionid(sessionid);
959 /* Search in the appropriate list */
960 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
961 if (!memcmp(elem->se_sessionid.data, sessionid->data,
962 NFS4_MAX_SESSIONID_LEN)) {
967 dprintk("%s: session not found\n", __func__);
971 /* caller must hold client_lock */
973 unhash_session(struct nfsd4_session *ses)
975 list_del(&ses->se_hash);
976 spin_lock(&ses->se_client->cl_lock);
977 list_del(&ses->se_perclnt);
978 spin_unlock(&ses->se_client->cl_lock);
981 /* must be called under the client_lock */
983 renew_client_locked(struct nfs4_client *clp)
985 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
987 if (is_client_expired(clp)) {
989 printk("%s: client (clientid %08x/%08x) already expired\n",
991 clp->cl_clientid.cl_boot,
992 clp->cl_clientid.cl_id);
996 dprintk("renewing client (clientid %08x/%08x)\n",
997 clp->cl_clientid.cl_boot,
998 clp->cl_clientid.cl_id);
999 list_move_tail(&clp->cl_lru, &nn->client_lru);
1000 clp->cl_time = get_seconds();
1004 renew_client(struct nfs4_client *clp)
1006 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1008 spin_lock(&nn->client_lock);
1009 renew_client_locked(clp);
1010 spin_unlock(&nn->client_lock);
1013 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1015 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1017 if (clid->cl_boot == nn->boot_time)
1019 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1020 clid->cl_boot, clid->cl_id, nn->boot_time);
1025 * XXX Should we use a slab cache ?
1026 * This type of memory management is somewhat inefficient, but we use it
1027 * anyway since SETCLIENTID is not a common operation.
1029 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1031 struct nfs4_client *clp;
1033 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1036 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1037 if (clp->cl_name.data == NULL) {
1041 clp->cl_name.len = name.len;
1046 free_client(struct nfs4_client *clp)
1048 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1050 lockdep_assert_held(&nn->client_lock);
1051 while (!list_empty(&clp->cl_sessions)) {
1052 struct nfsd4_session *ses;
1053 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1055 list_del(&ses->se_perclnt);
1056 nfsd4_put_session_locked(ses);
1058 free_svc_cred(&clp->cl_cred);
1059 kfree(clp->cl_name.data);
1064 release_session_client(struct nfsd4_session *session)
1066 struct nfs4_client *clp = session->se_client;
1067 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1069 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
1071 if (is_client_expired(clp)) {
1073 session->se_client = NULL;
1075 renew_client_locked(clp);
1076 spin_unlock(&nn->client_lock);
1079 /* must be called under the client_lock */
1081 unhash_client_locked(struct nfs4_client *clp)
1083 struct nfsd4_session *ses;
1085 mark_client_expired(clp);
1086 list_del(&clp->cl_lru);
1087 spin_lock(&clp->cl_lock);
1088 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1089 list_del_init(&ses->se_hash);
1090 spin_unlock(&clp->cl_lock);
1094 destroy_client(struct nfs4_client *clp)
1096 struct nfs4_openowner *oo;
1097 struct nfs4_delegation *dp;
1098 struct list_head reaplist;
1099 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1101 INIT_LIST_HEAD(&reaplist);
1102 spin_lock(&recall_lock);
1103 while (!list_empty(&clp->cl_delegations)) {
1104 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1105 list_del_init(&dp->dl_perclnt);
1106 list_move(&dp->dl_recall_lru, &reaplist);
1108 spin_unlock(&recall_lock);
1109 while (!list_empty(&reaplist)) {
1110 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1111 unhash_delegation(dp);
1113 while (!list_empty(&clp->cl_openowners)) {
1114 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1115 release_openowner(oo);
1117 nfsd4_shutdown_callback(clp);
1118 if (clp->cl_cb_conn.cb_xprt)
1119 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1120 list_del(&clp->cl_idhash);
1121 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1122 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1124 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1125 spin_lock(&nn->client_lock);
1126 unhash_client_locked(clp);
1127 if (atomic_read(&clp->cl_refcount) == 0)
1129 spin_unlock(&nn->client_lock);
1132 static void expire_client(struct nfs4_client *clp)
1134 nfsd4_client_record_remove(clp);
1135 destroy_client(clp);
1138 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1140 memcpy(target->cl_verifier.data, source->data,
1141 sizeof(target->cl_verifier.data));
1144 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1146 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1147 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1150 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1152 if (source->cr_principal) {
1153 target->cr_principal =
1154 kstrdup(source->cr_principal, GFP_KERNEL);
1155 if (target->cr_principal == NULL)
1158 target->cr_principal = NULL;
1159 target->cr_flavor = source->cr_flavor;
1160 target->cr_uid = source->cr_uid;
1161 target->cr_gid = source->cr_gid;
1162 target->cr_group_info = source->cr_group_info;
1163 get_group_info(target->cr_group_info);
1168 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1172 res = o1->len - o2->len;
1175 return (long long)memcmp(o1->data, o2->data, o1->len);
1178 static int same_name(const char *n1, const char *n2)
1180 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1184 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1186 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1190 same_clid(clientid_t *cl1, clientid_t *cl2)
1192 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1195 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1199 if (g1->ngroups != g2->ngroups)
1201 for (i=0; i<g1->ngroups; i++)
1202 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1208 * RFC 3530 language requires clid_inuse be returned when the
1209 * "principal" associated with a requests differs from that previously
1210 * used. We use uid, gid's, and gss principal string as our best
1211 * approximation. We also don't want to allow non-gss use of a client
1212 * established using gss: in theory cr_principal should catch that
1213 * change, but in practice cr_principal can be null even in the gss case
1214 * since gssd doesn't always pass down a principal string.
1216 static bool is_gss_cred(struct svc_cred *cr)
1218 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1219 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1224 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1226 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1227 || (cr1->cr_uid != cr2->cr_uid)
1228 || (cr1->cr_gid != cr2->cr_gid)
1229 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1231 if (cr1->cr_principal == cr2->cr_principal)
1233 if (!cr1->cr_principal || !cr2->cr_principal)
1235 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1238 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1240 static u32 current_clientid = 1;
1242 clp->cl_clientid.cl_boot = nn->boot_time;
1243 clp->cl_clientid.cl_id = current_clientid++;
1246 static void gen_confirm(struct nfs4_client *clp)
1251 verf[0] = (__be32)get_seconds();
1252 verf[1] = (__be32)i++;
1253 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1256 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1258 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1261 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1263 struct nfs4_stid *s;
1265 s = find_stateid(cl, t);
1268 if (typemask & s->sc_type)
1273 static struct nfs4_client *create_client(struct xdr_netobj name,
1274 struct svc_rqst *rqstp, nfs4_verifier *verf)
1276 struct nfs4_client *clp;
1277 struct sockaddr *sa = svc_addr(rqstp);
1279 struct net *net = SVC_NET(rqstp);
1280 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1282 clp = alloc_client(name);
1286 INIT_LIST_HEAD(&clp->cl_sessions);
1287 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1289 spin_lock(&nn->client_lock);
1291 spin_unlock(&nn->client_lock);
1294 idr_init(&clp->cl_stateids);
1295 atomic_set(&clp->cl_refcount, 0);
1296 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1297 INIT_LIST_HEAD(&clp->cl_idhash);
1298 INIT_LIST_HEAD(&clp->cl_openowners);
1299 INIT_LIST_HEAD(&clp->cl_delegations);
1300 INIT_LIST_HEAD(&clp->cl_lru);
1301 INIT_LIST_HEAD(&clp->cl_callbacks);
1302 spin_lock_init(&clp->cl_lock);
1303 nfsd4_init_callback(&clp->cl_cb_null);
1304 clp->cl_time = get_seconds();
1305 clear_bit(0, &clp->cl_cb_slot_busy);
1306 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1307 copy_verf(clp, verf);
1308 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1310 clp->cl_cb_session = NULL;
1316 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1318 struct rb_node **new = &(root->rb_node), *parent = NULL;
1319 struct nfs4_client *clp;
1322 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1325 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1326 new = &((*new)->rb_left);
1328 new = &((*new)->rb_right);
1331 rb_link_node(&new_clp->cl_namenode, parent, new);
1332 rb_insert_color(&new_clp->cl_namenode, root);
1335 static struct nfs4_client *
1336 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1339 struct rb_node *node = root->rb_node;
1340 struct nfs4_client *clp;
1343 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1344 cmp = compare_blob(&clp->cl_name, name);
1346 node = node->rb_left;
1348 node = node->rb_right;
1356 add_to_unconfirmed(struct nfs4_client *clp)
1358 unsigned int idhashval;
1359 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1361 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1362 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1363 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1364 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1369 move_to_confirmed(struct nfs4_client *clp)
1371 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1372 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1374 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1375 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1376 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1377 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1378 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1382 static struct nfs4_client *
1383 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1385 struct nfs4_client *clp;
1386 unsigned int idhashval = clientid_hashval(clid->cl_id);
1388 list_for_each_entry(clp, &nn->conf_id_hashtbl[idhashval], cl_idhash) {
1389 if (same_clid(&clp->cl_clientid, clid)) {
1390 if ((bool)clp->cl_minorversion != sessions)
1399 static struct nfs4_client *
1400 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1402 struct nfs4_client *clp;
1403 unsigned int idhashval = clientid_hashval(clid->cl_id);
1405 list_for_each_entry(clp, &nn->unconf_id_hashtbl[idhashval], cl_idhash) {
1406 if (same_clid(&clp->cl_clientid, clid)) {
1407 if ((bool)clp->cl_minorversion != sessions)
1415 static bool clp_used_exchangeid(struct nfs4_client *clp)
1417 return clp->cl_exchange_flags != 0;
1420 static struct nfs4_client *
1421 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1423 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1426 static struct nfs4_client *
1427 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1429 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1433 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1435 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1436 struct sockaddr *sa = svc_addr(rqstp);
1437 u32 scopeid = rpc_get_scope_id(sa);
1438 unsigned short expected_family;
1440 /* Currently, we only support tcp and tcp6 for the callback channel */
1441 if (se->se_callback_netid_len == 3 &&
1442 !memcmp(se->se_callback_netid_val, "tcp", 3))
1443 expected_family = AF_INET;
1444 else if (se->se_callback_netid_len == 4 &&
1445 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1446 expected_family = AF_INET6;
1450 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1451 se->se_callback_addr_len,
1452 (struct sockaddr *)&conn->cb_addr,
1453 sizeof(conn->cb_addr));
1455 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1458 if (conn->cb_addr.ss_family == AF_INET6)
1459 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1461 conn->cb_prog = se->se_callback_prog;
1462 conn->cb_ident = se->se_callback_ident;
1463 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1466 conn->cb_addr.ss_family = AF_UNSPEC;
1467 conn->cb_addrlen = 0;
1468 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1469 "will not receive delegations\n",
1470 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1476 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1479 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1481 struct nfsd4_slot *slot = resp->cstate.slot;
1484 dprintk("--> %s slot %p\n", __func__, slot);
1486 slot->sl_opcnt = resp->opcnt;
1487 slot->sl_status = resp->cstate.status;
1489 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1490 if (nfsd4_not_cached(resp)) {
1491 slot->sl_datalen = 0;
1494 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1495 base = (char *)resp->cstate.datap -
1496 (char *)resp->xbuf->head[0].iov_base;
1497 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1499 WARN("%s: sessions DRC could not cache compound\n", __func__);
1504 * Encode the replay sequence operation from the slot values.
1505 * If cachethis is FALSE encode the uncached rep error on the next
1506 * operation which sets resp->p and increments resp->opcnt for
1507 * nfs4svc_encode_compoundres.
1511 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1512 struct nfsd4_compoundres *resp)
1514 struct nfsd4_op *op;
1515 struct nfsd4_slot *slot = resp->cstate.slot;
1517 /* Encode the replayed sequence operation */
1518 op = &args->ops[resp->opcnt - 1];
1519 nfsd4_encode_operation(resp, op);
1521 /* Return nfserr_retry_uncached_rep in next operation. */
1522 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1523 op = &args->ops[resp->opcnt++];
1524 op->status = nfserr_retry_uncached_rep;
1525 nfsd4_encode_operation(resp, op);
1531 * The sequence operation is not cached because we can use the slot and
1535 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1536 struct nfsd4_sequence *seq)
1538 struct nfsd4_slot *slot = resp->cstate.slot;
1541 dprintk("--> %s slot %p\n", __func__, slot);
1543 /* Either returns 0 or nfserr_retry_uncached */
1544 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1545 if (status == nfserr_retry_uncached_rep)
1548 /* The sequence operation has been encoded, cstate->datap set. */
1549 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1551 resp->opcnt = slot->sl_opcnt;
1552 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1553 status = slot->sl_status;
1559 * Set the exchange_id flags returned by the server.
1562 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1564 /* pNFS is not supported */
1565 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1567 /* Referrals are supported, Migration is not. */
1568 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1570 /* set the wire flags to return to client. */
1571 clid->flags = new->cl_exchange_flags;
1574 static bool client_has_state(struct nfs4_client *clp)
1577 * Note clp->cl_openowners check isn't quite right: there's no
1578 * need to count owners without stateid's.
1580 * Also note we should probably be using this in 4.0 case too.
1582 return !list_empty(&clp->cl_openowners)
1583 || !list_empty(&clp->cl_delegations)
1584 || !list_empty(&clp->cl_sessions);
1588 nfsd4_exchange_id(struct svc_rqst *rqstp,
1589 struct nfsd4_compound_state *cstate,
1590 struct nfsd4_exchange_id *exid)
1592 struct nfs4_client *unconf, *conf, *new;
1594 char addr_str[INET6_ADDRSTRLEN];
1595 nfs4_verifier verf = exid->verifier;
1596 struct sockaddr *sa = svc_addr(rqstp);
1597 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1598 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1600 rpc_ntop(sa, addr_str, sizeof(addr_str));
1601 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1602 "ip_addr=%s flags %x, spa_how %d\n",
1603 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1604 addr_str, exid->flags, exid->spa_how);
1606 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1607 return nfserr_inval;
1609 /* Currently only support SP4_NONE */
1610 switch (exid->spa_how) {
1613 default: /* checked by xdr code */
1617 return nfserr_serverfault; /* no excuse :-/ */
1620 /* Cases below refer to rfc 5661 section 18.35.4: */
1622 conf = find_confirmed_client_by_name(&exid->clname, nn);
1624 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1625 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1628 if (!clp_used_exchangeid(conf)) { /* buggy client */
1629 status = nfserr_inval;
1632 if (!creds_match) { /* case 9 */
1633 status = nfserr_perm;
1636 if (!verfs_match) { /* case 8 */
1637 status = nfserr_not_same;
1641 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1645 if (!creds_match) { /* case 3 */
1646 if (client_has_state(conf)) {
1647 status = nfserr_clid_inuse;
1650 expire_client(conf);
1653 if (verfs_match) { /* case 2 */
1654 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1658 /* case 5, client reboot */
1662 if (update) { /* case 7 */
1663 status = nfserr_noent;
1667 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1668 if (unconf) /* case 4, possible retry or client restart */
1669 expire_client(unconf);
1671 /* case 1 (normal case) */
1673 new = create_client(exid->clname, rqstp, &verf);
1675 status = nfserr_jukebox;
1678 new->cl_minorversion = 1;
1681 add_to_unconfirmed(new);
1683 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1684 exid->clientid.cl_id = new->cl_clientid.cl_id;
1686 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1687 nfsd4_set_ex_flags(new, exid);
1689 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1690 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1694 nfs4_unlock_state();
1699 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1701 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1704 /* The slot is in use, and no response has been sent. */
1706 if (seqid == slot_seqid)
1707 return nfserr_jukebox;
1709 return nfserr_seq_misordered;
1711 /* Note unsigned 32-bit arithmetic handles wraparound: */
1712 if (likely(seqid == slot_seqid + 1))
1714 if (seqid == slot_seqid)
1715 return nfserr_replay_cache;
1716 return nfserr_seq_misordered;
1720 * Cache the create session result into the create session single DRC
1721 * slot cache by saving the xdr structure. sl_seqid has been set.
1722 * Do this for solo or embedded create session operations.
1725 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1726 struct nfsd4_clid_slot *slot, __be32 nfserr)
1728 slot->sl_status = nfserr;
1729 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1733 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1734 struct nfsd4_clid_slot *slot)
1736 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1737 return slot->sl_status;
1740 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1741 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1742 1 + /* MIN tag is length with zero, only length */ \
1743 3 + /* version, opcount, opcode */ \
1744 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1745 /* seqid, slotID, slotID, cache */ \
1746 4 ) * sizeof(__be32))
1748 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1749 2 + /* verifier: AUTH_NULL, length 0 */\
1751 1 + /* MIN tag is length with zero, only length */ \
1752 3 + /* opcount, opcode, opstatus*/ \
1753 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1754 /* seqid, slotID, slotID, slotID, status */ \
1755 5 ) * sizeof(__be32))
1757 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1759 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1760 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1764 nfsd4_create_session(struct svc_rqst *rqstp,
1765 struct nfsd4_compound_state *cstate,
1766 struct nfsd4_create_session *cr_ses)
1768 struct sockaddr *sa = svc_addr(rqstp);
1769 struct nfs4_client *conf, *unconf;
1770 struct nfsd4_session *new;
1771 struct nfsd4_conn *conn;
1772 struct nfsd4_clid_slot *cs_slot = NULL;
1774 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1776 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1777 return nfserr_inval;
1778 if (check_forechannel_attrs(cr_ses->fore_channel))
1779 return nfserr_toosmall;
1780 new = alloc_session(&cr_ses->fore_channel);
1782 return nfserr_jukebox;
1783 status = nfserr_jukebox;
1784 conn = alloc_conn_from_crses(rqstp, cr_ses);
1786 goto out_free_session;
1789 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1790 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1793 cs_slot = &conf->cl_cs_slot;
1794 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1795 if (status == nfserr_replay_cache) {
1796 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1798 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1799 status = nfserr_seq_misordered;
1802 } else if (unconf) {
1803 struct nfs4_client *old;
1804 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1805 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1806 status = nfserr_clid_inuse;
1809 cs_slot = &unconf->cl_cs_slot;
1810 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1812 /* an unconfirmed replay returns misordered */
1813 status = nfserr_seq_misordered;
1816 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1819 move_to_confirmed(unconf);
1822 status = nfserr_stale_clientid;
1827 * We do not support RDMA or persistent sessions
1829 cr_ses->flags &= ~SESSION4_PERSIST;
1830 cr_ses->flags &= ~SESSION4_RDMA;
1832 init_session(rqstp, new, conf, cr_ses);
1833 nfsd4_init_conn(rqstp, conn, new);
1835 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1836 NFS4_MAX_SESSIONID_LEN);
1837 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1838 sizeof(struct nfsd4_channel_attrs));
1839 cs_slot->sl_seqid++;
1840 cr_ses->seqid = cs_slot->sl_seqid;
1842 /* cache solo and embedded create sessions under the state lock */
1843 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1845 nfs4_unlock_state();
1846 dprintk("%s returns %d\n", __func__, ntohl(status));
1851 __free_session(new);
1855 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1857 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1858 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1860 return argp->opcnt == resp->opcnt;
1863 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1866 case NFS4_CDFC4_FORE:
1867 case NFS4_CDFC4_BACK:
1869 case NFS4_CDFC4_FORE_OR_BOTH:
1870 case NFS4_CDFC4_BACK_OR_BOTH:
1871 *dir = NFS4_CDFC4_BOTH;
1874 return nfserr_inval;
1877 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1879 struct nfsd4_session *session = cstate->session;
1880 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1882 spin_lock(&nn->client_lock);
1883 session->se_cb_prog = bc->bc_cb_program;
1884 session->se_cb_sec = bc->bc_cb_sec;
1885 spin_unlock(&nn->client_lock);
1887 nfsd4_probe_callback(session->se_client);
1892 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1893 struct nfsd4_compound_state *cstate,
1894 struct nfsd4_bind_conn_to_session *bcts)
1897 struct nfsd4_conn *conn;
1898 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1900 if (!nfsd4_last_compound_op(rqstp))
1901 return nfserr_not_only_op;
1902 spin_lock(&nn->client_lock);
1903 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
1904 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1905 * client_lock iself: */
1906 if (cstate->session) {
1907 nfsd4_get_session(cstate->session);
1908 atomic_inc(&cstate->session->se_client->cl_refcount);
1910 spin_unlock(&nn->client_lock);
1911 if (!cstate->session)
1912 return nfserr_badsession;
1914 status = nfsd4_map_bcts_dir(&bcts->dir);
1917 conn = alloc_conn(rqstp, bcts->dir);
1919 return nfserr_jukebox;
1920 nfsd4_init_conn(rqstp, conn, cstate->session);
1924 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1928 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1932 nfsd4_destroy_session(struct svc_rqst *r,
1933 struct nfsd4_compound_state *cstate,
1934 struct nfsd4_destroy_session *sessionid)
1936 struct nfsd4_session *ses;
1937 __be32 status = nfserr_badsession;
1938 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
1941 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1942 * - Should we return nfserr_back_chan_busy if waiting for
1943 * callbacks on to-be-destroyed session?
1944 * - Do we need to clear any callback info from previous session?
1947 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1948 if (!nfsd4_last_compound_op(r))
1949 return nfserr_not_only_op;
1951 dump_sessionid(__func__, &sessionid->sessionid);
1952 spin_lock(&nn->client_lock);
1953 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
1955 spin_unlock(&nn->client_lock);
1959 unhash_session(ses);
1960 spin_unlock(&nn->client_lock);
1963 nfsd4_probe_callback_sync(ses->se_client);
1964 nfs4_unlock_state();
1966 spin_lock(&nn->client_lock);
1967 nfsd4_del_conns(ses);
1968 nfsd4_put_session_locked(ses);
1969 spin_unlock(&nn->client_lock);
1972 dprintk("%s returns %d\n", __func__, ntohl(status));
1976 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1978 struct nfsd4_conn *c;
1980 list_for_each_entry(c, &s->se_conns, cn_persession) {
1981 if (c->cn_xprt == xpt) {
1988 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1990 struct nfs4_client *clp = ses->se_client;
1991 struct nfsd4_conn *c;
1994 spin_lock(&clp->cl_lock);
1995 c = __nfsd4_find_conn(new->cn_xprt, ses);
1997 spin_unlock(&clp->cl_lock);
2001 __nfsd4_hash_conn(new, ses);
2002 spin_unlock(&clp->cl_lock);
2003 ret = nfsd4_register_conn(new);
2005 /* oops; xprt is already down: */
2006 nfsd4_conn_lost(&new->cn_xpt_user);
2010 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2012 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2014 return args->opcnt > session->se_fchannel.maxops;
2017 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2018 struct nfsd4_session *session)
2020 struct xdr_buf *xb = &rqstp->rq_arg;
2022 return xb->len > session->se_fchannel.maxreq_sz;
2026 nfsd4_sequence(struct svc_rqst *rqstp,
2027 struct nfsd4_compound_state *cstate,
2028 struct nfsd4_sequence *seq)
2030 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2031 struct nfsd4_session *session;
2032 struct nfsd4_slot *slot;
2033 struct nfsd4_conn *conn;
2035 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2037 if (resp->opcnt != 1)
2038 return nfserr_sequence_pos;
2041 * Will be either used or freed by nfsd4_sequence_check_conn
2044 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2046 return nfserr_jukebox;
2048 spin_lock(&nn->client_lock);
2049 status = nfserr_badsession;
2050 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2054 status = nfserr_too_many_ops;
2055 if (nfsd4_session_too_many_ops(rqstp, session))
2058 status = nfserr_req_too_big;
2059 if (nfsd4_request_too_big(rqstp, session))
2062 status = nfserr_badslot;
2063 if (seq->slotid >= session->se_fchannel.maxreqs)
2066 slot = session->se_slots[seq->slotid];
2067 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2069 /* We do not negotiate the number of slots yet, so set the
2070 * maxslots to the session maxreqs which is used to encode
2071 * sr_highest_slotid and the sr_target_slot id to maxslots */
2072 seq->maxslots = session->se_fchannel.maxreqs;
2074 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2075 slot->sl_flags & NFSD4_SLOT_INUSE);
2076 if (status == nfserr_replay_cache) {
2077 status = nfserr_seq_misordered;
2078 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2080 cstate->slot = slot;
2081 cstate->session = session;
2082 /* Return the cached reply status and set cstate->status
2083 * for nfsd4_proc_compound processing */
2084 status = nfsd4_replay_cache_entry(resp, seq);
2085 cstate->status = nfserr_replay_cache;
2091 nfsd4_sequence_check_conn(conn, session);
2094 /* Success! bump slot seqid */
2095 slot->sl_seqid = seq->seqid;
2096 slot->sl_flags |= NFSD4_SLOT_INUSE;
2098 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2100 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2102 cstate->slot = slot;
2103 cstate->session = session;
2106 /* Hold a session reference until done processing the compound. */
2107 if (cstate->session) {
2108 struct nfs4_client *clp = session->se_client;
2110 nfsd4_get_session(cstate->session);
2111 atomic_inc(&clp->cl_refcount);
2112 switch (clp->cl_cb_state) {
2114 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2116 case NFSD4_CB_FAULT:
2117 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2120 seq->status_flags = 0;
2124 spin_unlock(&nn->client_lock);
2125 dprintk("%s: return %d\n", __func__, ntohl(status));
2130 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2132 struct nfs4_client *conf, *unconf, *clp;
2134 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2137 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2138 conf = find_confirmed_client(&dc->clientid, true, nn);
2143 if (!is_client_expired(conf) && client_has_state(conf)) {
2144 status = nfserr_clientid_busy;
2148 /* rfc5661 18.50.3 */
2149 if (cstate->session && conf == cstate->session->se_client) {
2150 status = nfserr_clientid_busy;
2156 status = nfserr_stale_clientid;
2162 nfs4_unlock_state();
2163 dprintk("%s return %d\n", __func__, ntohl(status));
2168 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2172 if (rc->rca_one_fs) {
2173 if (!cstate->current_fh.fh_dentry)
2174 return nfserr_nofilehandle;
2176 * We don't take advantage of the rca_one_fs case.
2177 * That's OK, it's optional, we can safely ignore it.
2183 status = nfserr_complete_already;
2184 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2185 &cstate->session->se_client->cl_flags))
2188 status = nfserr_stale_clientid;
2189 if (is_client_expired(cstate->session->se_client))
2191 * The following error isn't really legal.
2192 * But we only get here if the client just explicitly
2193 * destroyed the client. Surely it no longer cares what
2194 * error it gets back on an operation for the dead
2200 nfsd4_client_record_create(cstate->session->se_client);
2202 nfs4_unlock_state();
2207 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2208 struct nfsd4_setclientid *setclid)
2210 struct xdr_netobj clname = setclid->se_name;
2211 nfs4_verifier clverifier = setclid->se_verf;
2212 struct nfs4_client *conf, *unconf, *new;
2214 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2216 /* Cases below refer to rfc 3530 section 14.2.33: */
2218 conf = find_confirmed_client_by_name(&clname, nn);
2221 status = nfserr_clid_inuse;
2222 if (clp_used_exchangeid(conf))
2224 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2225 char addr_str[INET6_ADDRSTRLEN];
2226 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2228 dprintk("NFSD: setclientid: string in use by client "
2229 "at %s\n", addr_str);
2233 unconf = find_unconfirmed_client_by_name(&clname, nn);
2235 expire_client(unconf);
2236 status = nfserr_jukebox;
2237 new = create_client(clname, rqstp, &clverifier);
2240 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2241 /* case 1: probable callback update */
2242 copy_clid(new, conf);
2243 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2245 new->cl_minorversion = 0;
2246 gen_callback(new, setclid, rqstp);
2247 add_to_unconfirmed(new);
2248 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2249 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2250 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2253 nfs4_unlock_state();
2259 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2260 struct nfsd4_compound_state *cstate,
2261 struct nfsd4_setclientid_confirm *setclientid_confirm)
2263 struct nfs4_client *conf, *unconf;
2264 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2265 clientid_t * clid = &setclientid_confirm->sc_clientid;
2267 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2269 if (STALE_CLIENTID(clid, nn))
2270 return nfserr_stale_clientid;
2273 conf = find_confirmed_client(clid, false, nn);
2274 unconf = find_unconfirmed_client(clid, false, nn);
2276 * We try hard to give out unique clientid's, so if we get an
2277 * attempt to confirm the same clientid with a different cred,
2278 * there's a bug somewhere. Let's charitably assume it's our
2281 status = nfserr_serverfault;
2282 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2284 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2286 /* cases below refer to rfc 3530 section 14.2.34: */
2287 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2288 if (conf && !unconf) /* case 2: probable retransmit */
2290 else /* case 4: client hasn't noticed we rebooted yet? */
2291 status = nfserr_stale_clientid;
2295 if (conf) { /* case 1: callback update */
2296 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2297 nfsd4_probe_callback(conf);
2298 expire_client(unconf);
2299 } else { /* case 3: normal case; new or rebooted client */
2300 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2302 expire_client(conf);
2303 move_to_confirmed(unconf);
2304 nfsd4_probe_callback(unconf);
2307 nfs4_unlock_state();
2311 static struct nfs4_file *nfsd4_alloc_file(void)
2313 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2316 /* OPEN Share state helper functions */
2317 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2319 unsigned int hashval = file_hashval(ino);
2321 atomic_set(&fp->fi_ref, 1);
2322 INIT_LIST_HEAD(&fp->fi_hash);
2323 INIT_LIST_HEAD(&fp->fi_stateids);
2324 INIT_LIST_HEAD(&fp->fi_delegations);
2325 fp->fi_inode = igrab(ino);
2326 fp->fi_had_conflict = false;
2327 fp->fi_lease = NULL;
2328 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2329 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2330 spin_lock(&recall_lock);
2331 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2332 spin_unlock(&recall_lock);
2336 nfsd4_free_slab(struct kmem_cache **slab)
2340 kmem_cache_destroy(*slab);
2345 nfsd4_free_slabs(void)
2347 nfsd4_free_slab(&openowner_slab);
2348 nfsd4_free_slab(&lockowner_slab);
2349 nfsd4_free_slab(&file_slab);
2350 nfsd4_free_slab(&stateid_slab);
2351 nfsd4_free_slab(&deleg_slab);
2355 nfsd4_init_slabs(void)
2357 openowner_slab = kmem_cache_create("nfsd4_openowners",
2358 sizeof(struct nfs4_openowner), 0, 0, NULL);
2359 if (openowner_slab == NULL)
2361 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2362 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2363 if (lockowner_slab == NULL)
2365 file_slab = kmem_cache_create("nfsd4_files",
2366 sizeof(struct nfs4_file), 0, 0, NULL);
2367 if (file_slab == NULL)
2369 stateid_slab = kmem_cache_create("nfsd4_stateids",
2370 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2371 if (stateid_slab == NULL)
2373 deleg_slab = kmem_cache_create("nfsd4_delegations",
2374 sizeof(struct nfs4_delegation), 0, 0, NULL);
2375 if (deleg_slab == NULL)
2380 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2384 void nfs4_free_openowner(struct nfs4_openowner *oo)
2386 kfree(oo->oo_owner.so_owner.data);
2387 kmem_cache_free(openowner_slab, oo);
2390 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2392 kfree(lo->lo_owner.so_owner.data);
2393 kmem_cache_free(lockowner_slab, lo);
2396 static void init_nfs4_replay(struct nfs4_replay *rp)
2398 rp->rp_status = nfserr_serverfault;
2400 rp->rp_buf = rp->rp_ibuf;
2403 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2405 struct nfs4_stateowner *sop;
2407 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2411 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2412 if (!sop->so_owner.data) {
2413 kmem_cache_free(slab, sop);
2416 sop->so_owner.len = owner->len;
2418 INIT_LIST_HEAD(&sop->so_stateids);
2419 sop->so_client = clp;
2420 init_nfs4_replay(&sop->so_replay);
2424 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2426 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2428 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2429 list_add(&oo->oo_perclient, &clp->cl_openowners);
2432 static struct nfs4_openowner *
2433 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2434 struct nfs4_openowner *oo;
2436 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2439 oo->oo_owner.so_is_open_owner = 1;
2440 oo->oo_owner.so_seqid = open->op_seqid;
2441 oo->oo_flags = NFS4_OO_NEW;
2443 oo->oo_last_closed_stid = NULL;
2444 INIT_LIST_HEAD(&oo->oo_close_lru);
2445 hash_openowner(oo, clp, strhashval);
2449 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2450 struct nfs4_openowner *oo = open->op_openowner;
2451 struct nfs4_client *clp = oo->oo_owner.so_client;
2453 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2454 INIT_LIST_HEAD(&stp->st_lockowners);
2455 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2456 list_add(&stp->st_perfile, &fp->fi_stateids);
2457 stp->st_stateowner = &oo->oo_owner;
2460 stp->st_access_bmap = 0;
2461 stp->st_deny_bmap = 0;
2462 set_access(open->op_share_access, stp);
2463 set_deny(open->op_share_deny, stp);
2464 stp->st_openstp = NULL;
2468 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2470 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2472 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2474 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2475 oo->oo_time = get_seconds();
2479 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2482 return (sop->so_owner.len == owner->len) &&
2483 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2484 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2487 static struct nfs4_openowner *
2488 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2489 bool sessions, struct nfsd_net *nn)
2491 struct nfs4_stateowner *so;
2492 struct nfs4_openowner *oo;
2493 struct nfs4_client *clp;
2495 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2496 if (!so->so_is_open_owner)
2498 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2500 clp = oo->oo_owner.so_client;
2501 if ((bool)clp->cl_minorversion != sessions)
2503 renew_client(oo->oo_owner.so_client);
2510 /* search file_hashtbl[] for file */
2511 static struct nfs4_file *
2512 find_file(struct inode *ino)
2514 unsigned int hashval = file_hashval(ino);
2515 struct nfs4_file *fp;
2517 spin_lock(&recall_lock);
2518 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2519 if (fp->fi_inode == ino) {
2521 spin_unlock(&recall_lock);
2525 spin_unlock(&recall_lock);
2530 * Called to check deny when READ with all zero stateid or
2531 * WRITE with all zero or all one stateid
2534 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2536 struct inode *ino = current_fh->fh_dentry->d_inode;
2537 struct nfs4_file *fp;
2538 struct nfs4_ol_stateid *stp;
2541 dprintk("NFSD: nfs4_share_conflict\n");
2543 fp = find_file(ino);
2546 ret = nfserr_locked;
2547 /* Search for conflicting share reservations */
2548 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2549 if (test_deny(deny_type, stp) ||
2550 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2559 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2561 /* We're assuming the state code never drops its reference
2562 * without first removing the lease. Since we're in this lease
2563 * callback (and since the lease code is serialized by the kernel
2564 * lock) we know the server hasn't removed the lease yet, we know
2565 * it's safe to take a reference: */
2566 atomic_inc(&dp->dl_count);
2568 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2570 /* only place dl_time is set. protected by lock_flocks*/
2571 dp->dl_time = get_seconds();
2573 nfsd4_cb_recall(dp);
2576 /* Called from break_lease() with lock_flocks() held. */
2577 static void nfsd_break_deleg_cb(struct file_lock *fl)
2579 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2580 struct nfs4_delegation *dp;
2583 WARN(1, "(%p)->fl_owner NULL\n", fl);
2586 if (fp->fi_had_conflict) {
2587 WARN(1, "duplicate break on %p\n", fp);
2591 * We don't want the locks code to timeout the lease for us;
2592 * we'll remove it ourself if a delegation isn't returned
2595 fl->fl_break_time = 0;
2597 spin_lock(&recall_lock);
2598 fp->fi_had_conflict = true;
2599 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2600 nfsd_break_one_deleg(dp);
2601 spin_unlock(&recall_lock);
2605 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2608 return lease_modify(onlist, arg);
2613 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2614 .lm_break = nfsd_break_deleg_cb,
2615 .lm_change = nfsd_change_deleg_cb,
2618 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2620 if (nfsd4_has_session(cstate))
2622 if (seqid == so->so_seqid - 1)
2623 return nfserr_replay_me;
2624 if (seqid == so->so_seqid)
2626 return nfserr_bad_seqid;
2630 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2631 struct nfsd4_open *open, struct nfsd_net *nn)
2633 clientid_t *clientid = &open->op_clientid;
2634 struct nfs4_client *clp = NULL;
2635 unsigned int strhashval;
2636 struct nfs4_openowner *oo = NULL;
2639 if (STALE_CLIENTID(&open->op_clientid, nn))
2640 return nfserr_stale_clientid;
2642 * In case we need it later, after we've already created the
2643 * file and don't want to risk a further failure:
2645 open->op_file = nfsd4_alloc_file();
2646 if (open->op_file == NULL)
2647 return nfserr_jukebox;
2649 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2650 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2651 open->op_openowner = oo;
2653 clp = find_confirmed_client(clientid, cstate->minorversion,
2656 return nfserr_expired;
2659 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2660 /* Replace unconfirmed owners without checking for replay. */
2661 clp = oo->oo_owner.so_client;
2662 release_openowner(oo);
2663 open->op_openowner = NULL;
2666 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2669 clp = oo->oo_owner.so_client;
2672 oo = alloc_init_open_stateowner(strhashval, clp, open);
2674 return nfserr_jukebox;
2675 open->op_openowner = oo;
2677 open->op_stp = nfs4_alloc_stateid(clp);
2679 return nfserr_jukebox;
2683 static inline __be32
2684 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2686 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2687 return nfserr_openmode;
2692 static int share_access_to_flags(u32 share_access)
2694 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2697 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2699 struct nfs4_stid *ret;
2701 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2704 return delegstateid(ret);
2707 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2709 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2710 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2714 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2715 struct nfs4_delegation **dp)
2718 __be32 status = nfserr_bad_stateid;
2720 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2723 flags = share_access_to_flags(open->op_share_access);
2724 status = nfs4_check_delegmode(*dp, flags);
2728 if (!nfsd4_is_deleg_cur(open))
2732 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2737 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2739 struct nfs4_ol_stateid *local;
2740 struct nfs4_openowner *oo = open->op_openowner;
2742 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2743 /* ignore lock owners */
2744 if (local->st_stateowner->so_is_open_owner == 0)
2746 /* remember if we have seen this open owner */
2747 if (local->st_stateowner == &oo->oo_owner)
2749 /* check for conflicting share reservations */
2750 if (!test_share(local, open))
2751 return nfserr_share_denied;
2756 static inline int nfs4_access_to_access(u32 nfs4_access)
2760 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2761 flags |= NFSD_MAY_READ;
2762 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2763 flags |= NFSD_MAY_WRITE;
2767 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2768 struct svc_fh *cur_fh, struct nfsd4_open *open)
2771 int oflag = nfs4_access_to_omode(open->op_share_access);
2772 int access = nfs4_access_to_access(open->op_share_access);
2774 if (!fp->fi_fds[oflag]) {
2775 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2776 &fp->fi_fds[oflag]);
2780 nfs4_file_get_access(fp, oflag);
2785 static inline __be32
2786 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2787 struct nfsd4_open *open)
2789 struct iattr iattr = {
2790 .ia_valid = ATTR_SIZE,
2793 if (!open->op_truncate)
2795 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2796 return nfserr_inval;
2797 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2801 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)
2803 u32 op_share_access = open->op_share_access;
2807 new_access = !test_access(op_share_access, stp);
2809 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2813 status = nfsd4_truncate(rqstp, cur_fh, open);
2816 int oflag = nfs4_access_to_omode(op_share_access);
2817 nfs4_file_put_access(fp, oflag);
2821 /* remember the open */
2822 set_access(op_share_access, stp);
2823 set_deny(open->op_share_deny, stp);
2830 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2832 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2835 /* Should we give out recallable state?: */
2836 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2838 if (clp->cl_cb_state == NFSD4_CB_UP)
2841 * In the sessions case, since we don't have to establish a
2842 * separate connection for callbacks, we assume it's OK
2843 * until we hear otherwise:
2845 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2848 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2850 struct file_lock *fl;
2852 fl = locks_alloc_lock();
2855 locks_init_lock(fl);
2856 fl->fl_lmops = &nfsd_lease_mng_ops;
2857 fl->fl_flags = FL_LEASE;
2858 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2859 fl->fl_end = OFFSET_MAX;
2860 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2861 fl->fl_pid = current->tgid;
2865 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2867 struct nfs4_file *fp = dp->dl_file;
2868 struct file_lock *fl;
2871 fl = nfs4_alloc_init_lease(dp, flag);
2874 fl->fl_file = find_readable_file(fp);
2875 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2876 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2878 list_del_init(&dp->dl_perclnt);
2879 locks_free_lock(fl);
2883 fp->fi_deleg_file = get_file(fl->fl_file);
2884 atomic_set(&fp->fi_delegees, 1);
2885 list_add(&dp->dl_perfile, &fp->fi_delegations);
2889 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2891 struct nfs4_file *fp = dp->dl_file;
2894 return nfs4_setlease(dp, flag);
2895 spin_lock(&recall_lock);
2896 if (fp->fi_had_conflict) {
2897 spin_unlock(&recall_lock);
2900 atomic_inc(&fp->fi_delegees);
2901 list_add(&dp->dl_perfile, &fp->fi_delegations);
2902 spin_unlock(&recall_lock);
2903 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2907 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2909 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2910 if (status == -EAGAIN)
2911 open->op_why_no_deleg = WND4_CONTENTION;
2913 open->op_why_no_deleg = WND4_RESOURCE;
2914 switch (open->op_deleg_want) {
2915 case NFS4_SHARE_WANT_READ_DELEG:
2916 case NFS4_SHARE_WANT_WRITE_DELEG:
2917 case NFS4_SHARE_WANT_ANY_DELEG:
2919 case NFS4_SHARE_WANT_CANCEL:
2920 open->op_why_no_deleg = WND4_CANCELLED;
2922 case NFS4_SHARE_WANT_NO_DELEG:
2929 * Attempt to hand out a delegation.
2932 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2933 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2935 struct nfs4_delegation *dp;
2936 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2938 int status = 0, flag = 0;
2940 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2941 flag = NFS4_OPEN_DELEGATE_NONE;
2942 open->op_recall = 0;
2943 switch (open->op_claim_type) {
2944 case NFS4_OPEN_CLAIM_PREVIOUS:
2946 open->op_recall = 1;
2947 flag = open->op_delegate_type;
2948 if (flag == NFS4_OPEN_DELEGATE_NONE)
2951 case NFS4_OPEN_CLAIM_NULL:
2952 /* Let's not give out any delegations till everyone's
2953 * had the chance to reclaim theirs.... */
2954 if (locks_in_grace(net))
2956 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2958 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2959 flag = NFS4_OPEN_DELEGATE_WRITE;
2961 flag = NFS4_OPEN_DELEGATE_READ;
2967 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2970 status = nfs4_set_delegation(dp, flag);
2974 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2976 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2977 STATEID_VAL(&dp->dl_stid.sc_stateid));
2979 open->op_delegate_type = flag;
2980 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2981 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2982 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2983 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2985 /* 4.1 client asking for a delegation? */
2986 if (open->op_deleg_want)
2987 nfsd4_open_deleg_none_ext(open, status);
2991 nfs4_put_delegation(dp);
2993 flag = NFS4_OPEN_DELEGATE_NONE;
2997 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2998 struct nfs4_delegation *dp)
3000 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3001 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3002 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3003 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3004 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3005 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3006 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3007 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3009 /* Otherwise the client must be confused wanting a delegation
3010 * it already has, therefore we don't return
3011 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3016 * called with nfs4_lock_state() held.
3019 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3021 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3022 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3023 struct nfs4_file *fp = NULL;
3024 struct inode *ino = current_fh->fh_dentry->d_inode;
3025 struct nfs4_ol_stateid *stp = NULL;
3026 struct nfs4_delegation *dp = NULL;
3030 * Lookup file; if found, lookup stateid and check open request,
3031 * and check for delegations in the process of being recalled.
3032 * If not found, create the nfs4_file struct
3034 fp = find_file(ino);
3036 if ((status = nfs4_check_open(fp, open, &stp)))
3038 status = nfs4_check_deleg(cl, fp, open, &dp);
3042 status = nfserr_bad_stateid;
3043 if (nfsd4_is_deleg_cur(open))
3045 status = nfserr_jukebox;
3047 open->op_file = NULL;
3048 nfsd4_init_file(fp, ino);
3052 * OPEN the file, or upgrade an existing OPEN.
3053 * If truncate fails, the OPEN fails.
3056 /* Stateid was found, this is an OPEN upgrade */
3057 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3061 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3064 status = nfsd4_truncate(rqstp, current_fh, open);
3068 open->op_stp = NULL;
3069 init_open_stateid(stp, fp, open);
3071 update_stateid(&stp->st_stid.sc_stateid);
3072 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3074 if (nfsd4_has_session(&resp->cstate)) {
3075 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3077 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3078 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3079 open->op_why_no_deleg = WND4_NOT_WANTED;
3085 * Attempt to hand out a delegation. No error return, because the
3086 * OPEN succeeds even if we fail.
3088 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3092 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3093 STATEID_VAL(&stp->st_stid.sc_stateid));
3095 /* 4.1 client trying to upgrade/downgrade delegation? */
3096 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3097 open->op_deleg_want)
3098 nfsd4_deleg_xgrade_none_ext(open, dp);
3102 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3103 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3105 * To finish the open response, we just need to set the rflags.
3107 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3108 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3109 !nfsd4_has_session(&resp->cstate))
3110 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3115 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3117 if (open->op_openowner) {
3118 struct nfs4_openowner *oo = open->op_openowner;
3120 if (!list_empty(&oo->oo_owner.so_stateids))
3121 list_del_init(&oo->oo_close_lru);
3122 if (oo->oo_flags & NFS4_OO_NEW) {
3124 release_openowner(oo);
3125 open->op_openowner = NULL;
3127 oo->oo_flags &= ~NFS4_OO_NEW;
3131 nfsd4_free_file(open->op_file);
3133 free_generic_stateid(open->op_stp);
3137 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3140 struct nfs4_client *clp;
3142 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3145 dprintk("process_renew(%08x/%08x): starting\n",
3146 clid->cl_boot, clid->cl_id);
3147 status = nfserr_stale_clientid;
3148 if (STALE_CLIENTID(clid, nn))
3150 clp = find_confirmed_client(clid, cstate->minorversion, nn);
3151 status = nfserr_expired;
3153 /* We assume the client took too long to RENEW. */
3154 dprintk("nfsd4_renew: clientid not found!\n");
3157 status = nfserr_cb_path_down;
3158 if (!list_empty(&clp->cl_delegations)
3159 && clp->cl_cb_state != NFSD4_CB_UP)
3163 nfs4_unlock_state();
3168 nfsd4_end_grace(struct nfsd_net *nn)
3170 /* do nothing if grace period already ended */
3171 if (nn->grace_ended)
3174 dprintk("NFSD: end of grace period\n");
3175 nn->grace_ended = true;
3176 nfsd4_record_grace_done(nn, nn->boot_time);
3177 locks_end_grace(&nn->nfsd4_manager);
3179 * Now that every NFSv4 client has had the chance to recover and
3180 * to see the (possibly new, possibly shorter) lease time, we
3181 * can safely set the next grace time to the current lease time:
3183 nn->nfsd4_grace = nn->nfsd4_lease;
3187 nfs4_laundromat(struct nfsd_net *nn)
3189 struct nfs4_client *clp;
3190 struct nfs4_openowner *oo;
3191 struct nfs4_delegation *dp;
3192 struct list_head *pos, *next, reaplist;
3193 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3194 time_t t, clientid_val = nn->nfsd4_lease;
3195 time_t u, test_val = nn->nfsd4_lease;
3199 dprintk("NFSD: laundromat service - starting\n");
3200 nfsd4_end_grace(nn);
3201 INIT_LIST_HEAD(&reaplist);
3202 spin_lock(&nn->client_lock);
3203 list_for_each_safe(pos, next, &nn->client_lru) {
3204 clp = list_entry(pos, struct nfs4_client, cl_lru);
3205 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3206 t = clp->cl_time - cutoff;
3207 if (clientid_val > t)
3211 if (atomic_read(&clp->cl_refcount)) {
3212 dprintk("NFSD: client in use (clientid %08x)\n",
3213 clp->cl_clientid.cl_id);
3216 unhash_client_locked(clp);
3217 list_add(&clp->cl_lru, &reaplist);
3219 spin_unlock(&nn->client_lock);
3220 list_for_each_safe(pos, next, &reaplist) {
3221 clp = list_entry(pos, struct nfs4_client, cl_lru);
3222 dprintk("NFSD: purging unused client (clientid %08x)\n",
3223 clp->cl_clientid.cl_id);
3226 spin_lock(&recall_lock);
3227 list_for_each_safe(pos, next, &del_recall_lru) {
3228 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3229 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3231 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3232 u = dp->dl_time - cutoff;
3237 list_move(&dp->dl_recall_lru, &reaplist);
3239 spin_unlock(&recall_lock);
3240 list_for_each_safe(pos, next, &reaplist) {
3241 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3242 unhash_delegation(dp);
3244 test_val = nn->nfsd4_lease;
3245 list_for_each_safe(pos, next, &nn->close_lru) {
3246 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3247 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3248 u = oo->oo_time - cutoff;
3253 release_openowner(oo);
3255 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3256 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3257 nfs4_unlock_state();
3258 return clientid_val;
3261 static struct workqueue_struct *laundry_wq;
3262 static void laundromat_main(struct work_struct *);
3265 laundromat_main(struct work_struct *laundry)
3268 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3270 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3273 t = nfs4_laundromat(nn);
3274 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3275 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3278 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3280 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3281 return nfserr_bad_stateid;
3286 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3288 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3290 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3291 STATEID_VAL(stateid));
3296 access_permit_read(struct nfs4_ol_stateid *stp)
3298 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3299 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3300 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3304 access_permit_write(struct nfs4_ol_stateid *stp)
3306 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3307 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3311 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3313 __be32 status = nfserr_openmode;
3315 /* For lock stateid's, we test the parent open, not the lock: */
3316 if (stp->st_openstp)
3317 stp = stp->st_openstp;
3318 if ((flags & WR_STATE) && !access_permit_write(stp))
3320 if ((flags & RD_STATE) && !access_permit_read(stp))
3327 static inline __be32
3328 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3330 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3332 else if (locks_in_grace(net)) {
3333 /* Answer in remaining cases depends on existence of
3334 * conflicting state; so we must wait out the grace period. */
3335 return nfserr_grace;
3336 } else if (flags & WR_STATE)
3337 return nfs4_share_conflict(current_fh,
3338 NFS4_SHARE_DENY_WRITE);
3339 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3340 return nfs4_share_conflict(current_fh,
3341 NFS4_SHARE_DENY_READ);
3345 * Allow READ/WRITE during grace period on recovered state only for files
3346 * that are not able to provide mandatory locking.
3349 grace_disallows_io(struct net *net, struct inode *inode)
3351 return locks_in_grace(net) && mandatory_lock(inode);
3354 /* Returns true iff a is later than b: */
3355 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3357 return (s32)a->si_generation - (s32)b->si_generation > 0;
3360 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3363 * When sessions are used the stateid generation number is ignored
3366 if (has_session && in->si_generation == 0)
3369 if (in->si_generation == ref->si_generation)
3372 /* If the client sends us a stateid from the future, it's buggy: */
3373 if (stateid_generation_after(in, ref))
3374 return nfserr_bad_stateid;
3376 * However, we could see a stateid from the past, even from a
3377 * non-buggy client. For example, if the client sends a lock
3378 * while some IO is outstanding, the lock may bump si_generation
3379 * while the IO is still in flight. The client could avoid that
3380 * situation by waiting for responses on all the IO requests,
3381 * but better performance may result in retrying IO that
3382 * receives an old_stateid error if requests are rarely
3383 * reordered in flight:
3385 return nfserr_old_stateid;
3388 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3390 struct nfs4_stid *s;
3391 struct nfs4_ol_stateid *ols;
3394 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3395 return nfserr_bad_stateid;
3396 /* Client debugging aid. */
3397 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3398 char addr_str[INET6_ADDRSTRLEN];
3399 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3401 pr_warn_ratelimited("NFSD: client %s testing state ID "
3402 "with incorrect client ID\n", addr_str);
3403 return nfserr_bad_stateid;
3405 s = find_stateid(cl, stateid);
3407 return nfserr_bad_stateid;
3408 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3411 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3413 ols = openlockstateid(s);
3414 if (ols->st_stateowner->so_is_open_owner
3415 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3416 return nfserr_bad_stateid;
3420 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3421 struct nfs4_stid **s, bool sessions,
3422 struct nfsd_net *nn)
3424 struct nfs4_client *cl;
3426 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3427 return nfserr_bad_stateid;
3428 if (STALE_STATEID(stateid, nn))
3429 return nfserr_stale_stateid;
3430 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions, nn);
3432 return nfserr_expired;
3433 *s = find_stateid_by_type(cl, stateid, typemask);
3435 return nfserr_bad_stateid;
3441 * Checks for stateid operations
3444 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3445 stateid_t *stateid, int flags, struct file **filpp)
3447 struct nfs4_stid *s;
3448 struct nfs4_ol_stateid *stp = NULL;
3449 struct nfs4_delegation *dp = NULL;
3450 struct svc_fh *current_fh = &cstate->current_fh;
3451 struct inode *ino = current_fh->fh_dentry->d_inode;
3452 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3458 if (grace_disallows_io(net, ino))
3459 return nfserr_grace;
3461 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3462 return check_special_stateids(net, current_fh, stateid, flags);
3464 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3465 &s, cstate->minorversion, nn);
3468 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3471 switch (s->sc_type) {
3472 case NFS4_DELEG_STID:
3473 dp = delegstateid(s);
3474 status = nfs4_check_delegmode(dp, flags);
3478 *filpp = dp->dl_file->fi_deleg_file;
3481 status = nfserr_serverfault;
3486 case NFS4_OPEN_STID:
3487 case NFS4_LOCK_STID:
3488 stp = openlockstateid(s);
3489 status = nfs4_check_fh(current_fh, stp);
3492 if (stp->st_stateowner->so_is_open_owner
3493 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3495 status = nfs4_check_openmode(stp, flags);
3499 if (flags & RD_STATE)
3500 *filpp = find_readable_file(stp->st_file);
3502 *filpp = find_writeable_file(stp->st_file);
3506 return nfserr_bad_stateid;
3514 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3516 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3517 return nfserr_locks_held;
3518 release_lock_stateid(stp);
3523 * Test if the stateid is valid
3526 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3527 struct nfsd4_test_stateid *test_stateid)
3529 struct nfsd4_test_stateid_id *stateid;
3530 struct nfs4_client *cl = cstate->session->se_client;
3533 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3534 stateid->ts_id_status =
3535 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3536 nfs4_unlock_state();
3542 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3543 struct nfsd4_free_stateid *free_stateid)
3545 stateid_t *stateid = &free_stateid->fr_stateid;
3546 struct nfs4_stid *s;
3547 struct nfs4_client *cl = cstate->session->se_client;
3548 __be32 ret = nfserr_bad_stateid;
3551 s = find_stateid(cl, stateid);
3554 switch (s->sc_type) {
3555 case NFS4_DELEG_STID:
3556 ret = nfserr_locks_held;
3558 case NFS4_OPEN_STID:
3559 case NFS4_LOCK_STID:
3560 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3563 if (s->sc_type == NFS4_LOCK_STID)
3564 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3566 ret = nfserr_locks_held;
3569 ret = nfserr_bad_stateid;
3572 nfs4_unlock_state();
3579 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3580 RD_STATE : WR_STATE;
3583 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3585 struct svc_fh *current_fh = &cstate->current_fh;
3586 struct nfs4_stateowner *sop = stp->st_stateowner;
3589 status = nfsd4_check_seqid(cstate, sop, seqid);
3592 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3594 * "Closed" stateid's exist *only* to return
3595 * nfserr_replay_me from the previous step.
3597 return nfserr_bad_stateid;
3598 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3601 return nfs4_check_fh(current_fh, stp);
3605 * Checks for sequence id mutating operations.
3608 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3609 stateid_t *stateid, char typemask,
3610 struct nfs4_ol_stateid **stpp,
3611 struct nfsd_net *nn)
3614 struct nfs4_stid *s;
3616 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3617 seqid, STATEID_VAL(stateid));
3620 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3621 cstate->minorversion, nn);
3624 *stpp = openlockstateid(s);
3625 cstate->replay_owner = (*stpp)->st_stateowner;
3627 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3630 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3631 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3634 struct nfs4_openowner *oo;
3636 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3637 NFS4_OPEN_STID, stpp, nn);
3640 oo = openowner((*stpp)->st_stateowner);
3641 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3642 return nfserr_bad_stateid;
3647 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3648 struct nfsd4_open_confirm *oc)
3651 struct nfs4_openowner *oo;
3652 struct nfs4_ol_stateid *stp;
3653 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3655 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3656 (int)cstate->current_fh.fh_dentry->d_name.len,
3657 cstate->current_fh.fh_dentry->d_name.name);
3659 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3665 status = nfs4_preprocess_seqid_op(cstate,
3666 oc->oc_seqid, &oc->oc_req_stateid,
3667 NFS4_OPEN_STID, &stp, nn);
3670 oo = openowner(stp->st_stateowner);
3671 status = nfserr_bad_stateid;
3672 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3674 oo->oo_flags |= NFS4_OO_CONFIRMED;
3675 update_stateid(&stp->st_stid.sc_stateid);
3676 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3677 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3678 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3680 nfsd4_client_record_create(oo->oo_owner.so_client);
3683 if (!cstate->replay_owner)
3684 nfs4_unlock_state();
3688 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3690 if (!test_access(access, stp))
3692 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3693 clear_access(access, stp);
3696 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3698 switch (to_access) {
3699 case NFS4_SHARE_ACCESS_READ:
3700 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3701 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3703 case NFS4_SHARE_ACCESS_WRITE:
3704 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3705 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3707 case NFS4_SHARE_ACCESS_BOTH:
3715 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3718 for (i = 0; i < 4; i++) {
3719 if ((i & deny) != i)
3725 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3726 struct nfsd4_compound_state *cstate,
3727 struct nfsd4_open_downgrade *od)
3730 struct nfs4_ol_stateid *stp;
3731 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3733 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3734 (int)cstate->current_fh.fh_dentry->d_name.len,
3735 cstate->current_fh.fh_dentry->d_name.name);
3737 /* We don't yet support WANT bits: */
3738 if (od->od_deleg_want)
3739 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3743 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3744 &od->od_stateid, &stp, nn);
3747 status = nfserr_inval;
3748 if (!test_access(od->od_share_access, stp)) {
3749 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3750 stp->st_access_bmap, od->od_share_access);
3753 if (!test_deny(od->od_share_deny, stp)) {
3754 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3755 stp->st_deny_bmap, od->od_share_deny);
3758 nfs4_stateid_downgrade(stp, od->od_share_access);
3760 reset_union_bmap_deny(od->od_share_deny, stp);
3762 update_stateid(&stp->st_stid.sc_stateid);
3763 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3766 if (!cstate->replay_owner)
3767 nfs4_unlock_state();
3771 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3773 struct nfs4_openowner *oo;
3774 struct nfs4_ol_stateid *s;
3776 if (!so->so_is_open_owner)
3779 s = oo->oo_last_closed_stid;
3782 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3783 /* Release the last_closed_stid on the next seqid bump: */
3784 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3787 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3788 release_last_closed_stateid(oo);
3791 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3793 unhash_open_stateid(s);
3794 s->st_stid.sc_type = NFS4_CLOSED_STID;
3798 * nfs4_unlock_state() called after encode
3801 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3802 struct nfsd4_close *close)
3805 struct nfs4_openowner *oo;
3806 struct nfs4_ol_stateid *stp;
3807 struct net *net = SVC_NET(rqstp);
3808 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3810 dprintk("NFSD: nfsd4_close on file %.*s\n",
3811 (int)cstate->current_fh.fh_dentry->d_name.len,
3812 cstate->current_fh.fh_dentry->d_name.name);
3815 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3817 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3821 oo = openowner(stp->st_stateowner);
3823 update_stateid(&stp->st_stid.sc_stateid);
3824 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3826 nfsd4_close_open_stateid(stp);
3827 release_last_closed_stateid(oo);
3828 oo->oo_last_closed_stid = stp;
3830 if (list_empty(&oo->oo_owner.so_stateids)) {
3831 if (cstate->minorversion) {
3832 release_openowner(oo);
3833 cstate->replay_owner = NULL;
3836 * In the 4.0 case we need to keep the owners around a
3837 * little while to handle CLOSE replay.
3839 if (list_empty(&oo->oo_owner.so_stateids))
3840 move_to_close_lru(oo, SVC_NET(rqstp));
3844 if (!cstate->replay_owner)
3845 nfs4_unlock_state();
3850 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3851 struct nfsd4_delegreturn *dr)
3853 struct nfs4_delegation *dp;
3854 stateid_t *stateid = &dr->dr_stateid;
3855 struct nfs4_stid *s;
3857 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3859 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3863 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
3864 cstate->minorversion, nn);
3867 dp = delegstateid(s);
3868 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3872 unhash_delegation(dp);
3874 nfs4_unlock_state();
3880 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3882 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3885 end_offset(u64 start, u64 len)
3890 return end >= start ? end: NFS4_MAX_UINT64;
3893 /* last octet in a range */
3895 last_byte_offset(u64 start, u64 len)
3901 return end > start ? end - 1: NFS4_MAX_UINT64;
3904 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3906 return (file_hashval(inode) + cl_id
3907 + opaque_hashval(ownername->data, ownername->len))
3908 & LOCKOWNER_INO_HASH_MASK;
3912 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3913 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3914 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3915 * locking, this prevents us from being completely protocol-compliant. The
3916 * real solution to this problem is to start using unsigned file offsets in
3917 * the VFS, but this is a very deep change!
3920 nfs4_transform_lock_offset(struct file_lock *lock)
3922 if (lock->fl_start < 0)
3923 lock->fl_start = OFFSET_MAX;
3924 if (lock->fl_end < 0)
3925 lock->fl_end = OFFSET_MAX;
3928 /* Hack!: For now, we're defining this just so we can use a pointer to it
3929 * as a unique cookie to identify our (NFSv4's) posix locks. */
3930 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3934 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3936 struct nfs4_lockowner *lo;
3938 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3939 lo = (struct nfs4_lockowner *) fl->fl_owner;
3940 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3941 lo->lo_owner.so_owner.len, GFP_KERNEL);
3942 if (!deny->ld_owner.data)
3943 /* We just don't care that much */
3945 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3946 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3949 deny->ld_owner.len = 0;
3950 deny->ld_owner.data = NULL;
3951 deny->ld_clientid.cl_boot = 0;
3952 deny->ld_clientid.cl_id = 0;
3954 deny->ld_start = fl->fl_start;
3955 deny->ld_length = NFS4_MAX_UINT64;
3956 if (fl->fl_end != NFS4_MAX_UINT64)
3957 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3958 deny->ld_type = NFS4_READ_LT;
3959 if (fl->fl_type != F_RDLCK)
3960 deny->ld_type = NFS4_WRITE_LT;
3963 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3965 struct nfs4_ol_stateid *lst;
3967 if (!same_owner_str(&lo->lo_owner, owner, clid))
3969 lst = list_first_entry(&lo->lo_owner.so_stateids,
3970 struct nfs4_ol_stateid, st_perstateowner);
3971 return lst->st_file->fi_inode == inode;
3974 static struct nfs4_lockowner *
3975 find_lockowner_str(struct inode *inode, clientid_t *clid,
3976 struct xdr_netobj *owner, struct nfsd_net *nn)
3978 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3979 struct nfs4_lockowner *lo;
3981 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3982 if (same_lockowner_ino(lo, inode, clid, owner))
3988 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3990 struct inode *inode = open_stp->st_file->fi_inode;
3991 unsigned int inohash = lockowner_ino_hashval(inode,
3992 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3993 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
3995 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
3996 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
3997 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4001 * Alloc a lock owner structure.
4002 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4005 * strhashval = ownerstr_hashval
4008 static struct nfs4_lockowner *
4009 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4010 struct nfs4_lockowner *lo;
4012 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4015 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4016 lo->lo_owner.so_is_open_owner = 0;
4017 /* It is the openowner seqid that will be incremented in encode in the
4018 * case of new lockowners; so increment the lock seqid manually: */
4019 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4020 hash_lockowner(lo, strhashval, clp, open_stp);
4024 static struct nfs4_ol_stateid *
4025 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4027 struct nfs4_ol_stateid *stp;
4028 struct nfs4_client *clp = lo->lo_owner.so_client;
4030 stp = nfs4_alloc_stateid(clp);
4033 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
4034 list_add(&stp->st_perfile, &fp->fi_stateids);
4035 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4036 stp->st_stateowner = &lo->lo_owner;
4039 stp->st_access_bmap = 0;
4040 stp->st_deny_bmap = open_stp->st_deny_bmap;
4041 stp->st_openstp = open_stp;
4046 check_lock_length(u64 offset, u64 length)
4048 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4049 LOFF_OVERFLOW(offset, length)));
4052 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4054 struct nfs4_file *fp = lock_stp->st_file;
4055 int oflag = nfs4_access_to_omode(access);
4057 if (test_access(access, lock_stp))
4059 nfs4_file_get_access(fp, oflag);
4060 set_access(access, lock_stp);
4063 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)
4065 struct nfs4_file *fi = ost->st_file;
4066 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4067 struct nfs4_client *cl = oo->oo_owner.so_client;
4068 struct nfs4_lockowner *lo;
4069 unsigned int strhashval;
4070 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4072 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4073 &lock->v.new.owner, nn);
4075 if (!cstate->minorversion)
4076 return nfserr_bad_seqid;
4077 /* XXX: a lockowner always has exactly one stateid: */
4078 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4079 struct nfs4_ol_stateid, st_perstateowner);
4082 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4083 &lock->v.new.owner);
4084 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4086 return nfserr_jukebox;
4087 *lst = alloc_init_lock_stateid(lo, fi, ost);
4089 release_lockowner(lo);
4090 return nfserr_jukebox;
4100 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4101 struct nfsd4_lock *lock)
4103 struct nfs4_openowner *open_sop = NULL;
4104 struct nfs4_lockowner *lock_sop = NULL;
4105 struct nfs4_ol_stateid *lock_stp;
4106 struct file *filp = NULL;
4107 struct file_lock *file_lock = NULL;
4108 struct file_lock *conflock = NULL;
4110 bool new_state = false;
4113 struct net *net = SVC_NET(rqstp);
4114 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4116 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4117 (long long) lock->lk_offset,
4118 (long long) lock->lk_length);
4120 if (check_lock_length(lock->lk_offset, lock->lk_length))
4121 return nfserr_inval;
4123 if ((status = fh_verify(rqstp, &cstate->current_fh,
4124 S_IFREG, NFSD_MAY_LOCK))) {
4125 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4131 if (lock->lk_is_new) {
4132 struct nfs4_ol_stateid *open_stp = NULL;
4134 if (nfsd4_has_session(cstate))
4135 /* See rfc 5661 18.10.3: given clientid is ignored: */
4136 memcpy(&lock->v.new.clientid,
4137 &cstate->session->se_client->cl_clientid,
4138 sizeof(clientid_t));
4140 status = nfserr_stale_clientid;
4141 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4144 /* validate and update open stateid and open seqid */
4145 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4146 lock->lk_new_open_seqid,
4147 &lock->lk_new_open_stateid,
4151 open_sop = openowner(open_stp->st_stateowner);
4152 status = nfserr_bad_stateid;
4153 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4154 &lock->v.new.clientid))
4156 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4157 &lock_stp, &new_state);
4159 status = nfs4_preprocess_seqid_op(cstate,
4160 lock->lk_old_lock_seqid,
4161 &lock->lk_old_lock_stateid,
4162 NFS4_LOCK_STID, &lock_stp, nn);
4165 lock_sop = lockowner(lock_stp->st_stateowner);
4167 lkflg = setlkflg(lock->lk_type);
4168 status = nfs4_check_openmode(lock_stp, lkflg);
4172 status = nfserr_grace;
4173 if (locks_in_grace(net) && !lock->lk_reclaim)
4175 status = nfserr_no_grace;
4176 if (!locks_in_grace(net) && lock->lk_reclaim)
4179 file_lock = locks_alloc_lock();
4181 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4182 status = nfserr_jukebox;
4186 locks_init_lock(file_lock);
4187 switch (lock->lk_type) {
4190 filp = find_readable_file(lock_stp->st_file);
4192 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4193 file_lock->fl_type = F_RDLCK;
4196 case NFS4_WRITEW_LT:
4197 filp = find_writeable_file(lock_stp->st_file);
4199 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4200 file_lock->fl_type = F_WRLCK;
4203 status = nfserr_inval;
4207 status = nfserr_openmode;
4210 file_lock->fl_owner = (fl_owner_t)lock_sop;
4211 file_lock->fl_pid = current->tgid;
4212 file_lock->fl_file = filp;
4213 file_lock->fl_flags = FL_POSIX;
4214 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4215 file_lock->fl_start = lock->lk_offset;
4216 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4217 nfs4_transform_lock_offset(file_lock);
4219 conflock = locks_alloc_lock();
4221 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4222 status = nfserr_jukebox;
4226 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4228 case 0: /* success! */
4229 update_stateid(&lock_stp->st_stid.sc_stateid);
4230 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4234 case (EAGAIN): /* conflock holds conflicting lock */
4235 status = nfserr_denied;
4236 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4237 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4240 status = nfserr_deadlock;
4243 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4244 status = nfserrno(err);
4248 if (status && new_state)
4249 release_lockowner(lock_sop);
4250 if (!cstate->replay_owner)
4251 nfs4_unlock_state();
4253 locks_free_lock(file_lock);
4255 locks_free_lock(conflock);
4260 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4261 * so we do a temporary open here just to get an open file to pass to
4262 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4265 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4268 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4270 err = nfserrno(vfs_test_lock(file, lock));
4280 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4281 struct nfsd4_lockt *lockt)
4283 struct inode *inode;
4284 struct file_lock *file_lock = NULL;
4285 struct nfs4_lockowner *lo;
4287 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4289 if (locks_in_grace(SVC_NET(rqstp)))
4290 return nfserr_grace;
4292 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4293 return nfserr_inval;
4297 status = nfserr_stale_clientid;
4298 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid, nn))
4301 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4304 inode = cstate->current_fh.fh_dentry->d_inode;
4305 file_lock = locks_alloc_lock();
4307 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4308 status = nfserr_jukebox;
4311 locks_init_lock(file_lock);
4312 switch (lockt->lt_type) {
4315 file_lock->fl_type = F_RDLCK;
4318 case NFS4_WRITEW_LT:
4319 file_lock->fl_type = F_WRLCK;
4322 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4323 status = nfserr_inval;
4327 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4329 file_lock->fl_owner = (fl_owner_t)lo;
4330 file_lock->fl_pid = current->tgid;
4331 file_lock->fl_flags = FL_POSIX;
4333 file_lock->fl_start = lockt->lt_offset;
4334 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4336 nfs4_transform_lock_offset(file_lock);
4338 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4342 if (file_lock->fl_type != F_UNLCK) {
4343 status = nfserr_denied;
4344 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4347 nfs4_unlock_state();
4349 locks_free_lock(file_lock);
4354 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4355 struct nfsd4_locku *locku)
4357 struct nfs4_ol_stateid *stp;
4358 struct file *filp = NULL;
4359 struct file_lock *file_lock = NULL;
4362 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4364 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4365 (long long) locku->lu_offset,
4366 (long long) locku->lu_length);
4368 if (check_lock_length(locku->lu_offset, locku->lu_length))
4369 return nfserr_inval;
4373 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4374 &locku->lu_stateid, NFS4_LOCK_STID,
4378 filp = find_any_file(stp->st_file);
4380 status = nfserr_lock_range;
4383 file_lock = locks_alloc_lock();
4385 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4386 status = nfserr_jukebox;
4389 locks_init_lock(file_lock);
4390 file_lock->fl_type = F_UNLCK;
4391 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4392 file_lock->fl_pid = current->tgid;
4393 file_lock->fl_file = filp;
4394 file_lock->fl_flags = FL_POSIX;
4395 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4396 file_lock->fl_start = locku->lu_offset;
4398 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4400 nfs4_transform_lock_offset(file_lock);
4403 * Try to unlock the file in the VFS.
4405 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4407 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4411 * OK, unlock succeeded; the only thing left to do is update the stateid.
4413 update_stateid(&stp->st_stid.sc_stateid);
4414 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4417 if (!cstate->replay_owner)
4418 nfs4_unlock_state();
4420 locks_free_lock(file_lock);
4424 status = nfserrno(err);
4430 * 1: locks held by lockowner
4431 * 0: no locks held by lockowner
4434 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4436 struct file_lock **flpp;
4437 struct inode *inode = filp->fi_inode;
4441 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4442 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4453 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4454 struct nfsd4_compound_state *cstate,
4455 struct nfsd4_release_lockowner *rlockowner)
4457 clientid_t *clid = &rlockowner->rl_clientid;
4458 struct nfs4_stateowner *sop;
4459 struct nfs4_lockowner *lo;
4460 struct nfs4_ol_stateid *stp;
4461 struct xdr_netobj *owner = &rlockowner->rl_owner;
4462 struct list_head matches;
4463 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4465 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4467 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4468 clid->cl_boot, clid->cl_id);
4470 /* XXX check for lease expiration */
4472 status = nfserr_stale_clientid;
4473 if (STALE_CLIENTID(clid, nn))
4478 status = nfserr_locks_held;
4479 INIT_LIST_HEAD(&matches);
4481 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4482 if (sop->so_is_open_owner)
4484 if (!same_owner_str(sop, owner, clid))
4486 list_for_each_entry(stp, &sop->so_stateids,
4488 lo = lockowner(sop);
4489 if (check_for_locks(stp->st_file, lo))
4491 list_add(&lo->lo_list, &matches);
4494 /* Clients probably won't expect us to return with some (but not all)
4495 * of the lockowner state released; so don't release any until all
4496 * have been checked. */
4498 while (!list_empty(&matches)) {
4499 lo = list_entry(matches.next, struct nfs4_lockowner,
4501 /* unhash_stateowner deletes so_perclient only
4502 * for openowners. */
4503 list_del(&lo->lo_list);
4504 release_lockowner(lo);
4507 nfs4_unlock_state();
4511 static inline struct nfs4_client_reclaim *
4514 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4518 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4520 struct nfs4_client_reclaim *crp;
4522 crp = nfsd4_find_reclaim_client(name, nn);
4523 return (crp && crp->cr_clp);
4527 * failure => all reset bets are off, nfserr_no_grace...
4529 struct nfs4_client_reclaim *
4530 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4532 unsigned int strhashval;
4533 struct nfs4_client_reclaim *crp;
4535 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4536 crp = alloc_reclaim();
4538 strhashval = clientstr_hashval(name);
4539 INIT_LIST_HEAD(&crp->cr_strhash);
4540 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4541 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4543 nn->reclaim_str_hashtbl_size++;
4549 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4551 list_del(&crp->cr_strhash);
4553 nn->reclaim_str_hashtbl_size--;
4557 nfs4_release_reclaim(struct nfsd_net *nn)
4559 struct nfs4_client_reclaim *crp = NULL;
4562 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4563 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4564 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4565 struct nfs4_client_reclaim, cr_strhash);
4566 nfs4_remove_reclaim_record(crp, nn);
4569 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4573 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4574 struct nfs4_client_reclaim *
4575 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4577 unsigned int strhashval;
4578 struct nfs4_client_reclaim *crp = NULL;
4580 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4582 strhashval = clientstr_hashval(recdir);
4583 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4584 if (same_name(crp->cr_recdir, recdir)) {
4592 * Called from OPEN. Look for clientid in reclaim list.
4595 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4597 struct nfs4_client *clp;
4599 /* find clientid in conf_id_hashtbl */
4600 clp = find_confirmed_client(clid, sessions, nn);
4602 return nfserr_reclaim_bad;
4604 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4607 #ifdef CONFIG_NFSD_FAULT_INJECTION
4609 void nfsd_forget_clients(u64 num)
4611 struct nfs4_client *clp, *next;
4613 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4616 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4621 nfs4_unlock_state();
4623 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4626 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4628 release_lockowner(lockowner(sop));
4631 static void release_openowner_sop(struct nfs4_stateowner *sop)
4633 release_openowner(openowner(sop));
4636 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4637 void (*release_sop)(struct nfs4_stateowner *),
4638 struct nfsd_net *nn)
4641 struct nfs4_stateowner *sop, *next;
4643 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4644 list_for_each_entry_safe(sop, next, &nn->ownerstr_hashtbl[i], so_strhash) {
4645 if (sop->so_is_open_owner != is_open_owner)
4655 void nfsd_forget_locks(u64 num)
4658 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4661 count = nfsd_release_n_owners(num, false, release_lockowner_sop, nn);
4662 nfs4_unlock_state();
4664 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4667 void nfsd_forget_openowners(u64 num)
4670 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4673 count = nfsd_release_n_owners(num, true, release_openowner_sop, nn);
4674 nfs4_unlock_state();
4676 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4679 static int nfsd_process_n_delegations(u64 num, struct list_head *list)
4682 struct nfs4_file *fp, *fnext;
4683 struct nfs4_delegation *dp, *dnext;
4685 for (i = 0; i < FILE_HASH_SIZE; i++) {
4686 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4687 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4688 list_move(&dp->dl_recall_lru, list);
4698 void nfsd_forget_delegations(u64 num)
4702 struct nfs4_delegation *dp, *dnext;
4704 spin_lock(&recall_lock);
4705 count = nfsd_process_n_delegations(num, &victims);
4706 spin_unlock(&recall_lock);
4709 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
4710 unhash_delegation(dp);
4711 nfs4_unlock_state();
4713 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4716 void nfsd_recall_delegations(u64 num)
4720 struct nfs4_delegation *dp, *dnext;
4722 spin_lock(&recall_lock);
4723 count = nfsd_process_n_delegations(num, &victims);
4724 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
4725 list_del(&dp->dl_recall_lru);
4726 nfsd_break_one_deleg(dp);
4728 spin_unlock(&recall_lock);
4730 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4733 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4735 /* initialization to perform at module load time: */
4738 nfs4_state_init(void)
4742 for (i = 0; i < FILE_HASH_SIZE; i++) {
4743 INIT_LIST_HEAD(&file_hashtbl[i]);
4745 INIT_LIST_HEAD(&del_recall_lru);
4749 * Since the lifetime of a delegation isn't limited to that of an open, a
4750 * client may quite reasonably hang on to a delegation as long as it has
4751 * the inode cached. This becomes an obvious problem the first time a
4752 * client's inode cache approaches the size of the server's total memory.
4754 * For now we avoid this problem by imposing a hard limit on the number
4755 * of delegations, which varies according to the server's memory size.
4758 set_max_delegations(void)
4761 * Allow at most 4 delegations per megabyte of RAM. Quick
4762 * estimates suggest that in the worst case (where every delegation
4763 * is for a different inode), a delegation could take about 1.5K,
4764 * giving a worst case usage of about 6% of memory.
4766 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4769 static int nfs4_state_create_net(struct net *net)
4771 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4774 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4775 CLIENT_HASH_SIZE, GFP_KERNEL);
4776 if (!nn->conf_id_hashtbl)
4778 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4779 CLIENT_HASH_SIZE, GFP_KERNEL);
4780 if (!nn->unconf_id_hashtbl)
4782 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
4783 OWNER_HASH_SIZE, GFP_KERNEL);
4784 if (!nn->ownerstr_hashtbl)
4786 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
4787 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
4788 if (!nn->lockowner_ino_hashtbl)
4789 goto err_lockowner_ino;
4790 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
4791 SESSION_HASH_SIZE, GFP_KERNEL);
4792 if (!nn->sessionid_hashtbl)
4795 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4796 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
4797 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
4799 for (i = 0; i < OWNER_HASH_SIZE; i++)
4800 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
4801 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4802 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
4803 for (i = 0; i < SESSION_HASH_SIZE; i++)
4804 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
4805 nn->conf_name_tree = RB_ROOT;
4806 nn->unconf_name_tree = RB_ROOT;
4807 INIT_LIST_HEAD(&nn->client_lru);
4808 INIT_LIST_HEAD(&nn->close_lru);
4809 spin_lock_init(&nn->client_lock);
4811 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
4817 kfree(nn->lockowner_ino_hashtbl);
4819 kfree(nn->ownerstr_hashtbl);
4821 kfree(nn->unconf_id_hashtbl);
4823 kfree(nn->conf_id_hashtbl);
4829 nfs4_state_destroy_net(struct net *net)
4832 struct nfs4_client *clp = NULL;
4833 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4834 struct rb_node *node, *tmp;
4836 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4837 while (!list_empty(&nn->conf_id_hashtbl[i])) {
4838 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4839 destroy_client(clp);
4843 node = rb_first(&nn->unconf_name_tree);
4844 while (node != NULL) {
4846 node = rb_next(tmp);
4847 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4848 rb_erase(tmp, &nn->unconf_name_tree);
4849 destroy_client(clp);
4852 kfree(nn->sessionid_hashtbl);
4853 kfree(nn->lockowner_ino_hashtbl);
4854 kfree(nn->ownerstr_hashtbl);
4855 kfree(nn->unconf_id_hashtbl);
4856 kfree(nn->conf_id_hashtbl);
4861 nfs4_state_start_net(struct net *net)
4863 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4867 * FIXME: For now, we hang most of the pernet global stuff off of
4868 * init_net until nfsd is fully containerized. Eventually, we'll
4869 * need to pass a net pointer into this function, take a reference
4870 * to that instead and then do most of the rest of this on a per-net
4873 if (net != &init_net)
4876 ret = nfs4_state_create_net(net);
4879 nfsd4_client_tracking_init(net);
4880 nn->boot_time = get_seconds();
4881 locks_start_grace(net, &nn->nfsd4_manager);
4882 nn->grace_ended = false;
4883 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
4884 nn->nfsd4_grace, net);
4885 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
4889 /* initialization to perform when the nfsd service is started: */
4892 nfs4_state_start(void)
4896 ret = set_callback_cred();
4899 laundry_wq = create_singlethread_workqueue("nfsd4");
4900 if (laundry_wq == NULL) {
4904 ret = nfsd4_create_callback_queue();
4906 goto out_free_laundry;
4908 set_max_delegations();
4913 destroy_workqueue(laundry_wq);
4918 /* should be called with the state lock held */
4920 nfs4_state_shutdown_net(struct net *net)
4922 struct nfs4_delegation *dp = NULL;
4923 struct list_head *pos, *next, reaplist;
4924 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4926 cancel_delayed_work_sync(&nn->laundromat_work);
4927 locks_end_grace(&nn->nfsd4_manager);
4929 INIT_LIST_HEAD(&reaplist);
4930 spin_lock(&recall_lock);
4931 list_for_each_safe(pos, next, &del_recall_lru) {
4932 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4933 if (dp->dl_stid.sc_client->net != net)
4935 list_move(&dp->dl_recall_lru, &reaplist);
4937 spin_unlock(&recall_lock);
4938 list_for_each_safe(pos, next, &reaplist) {
4939 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4940 unhash_delegation(dp);
4943 nfsd4_client_tracking_exit(net);
4944 nfs4_state_destroy_net(net);
4948 nfs4_state_shutdown(void)
4950 destroy_workqueue(laundry_wq);
4951 nfsd4_destroy_callback_queue();
4955 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4957 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4958 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4962 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4964 if (cstate->minorversion) {
4965 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4966 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4971 clear_current_stateid(struct nfsd4_compound_state *cstate)
4973 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4977 * functions to set current state id
4980 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4982 put_stateid(cstate, &odp->od_stateid);
4986 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4988 put_stateid(cstate, &open->op_stateid);
4992 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4994 put_stateid(cstate, &close->cl_stateid);
4998 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5000 put_stateid(cstate, &lock->lk_resp_stateid);
5004 * functions to consume current state id
5008 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5010 get_stateid(cstate, &odp->od_stateid);
5014 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5016 get_stateid(cstate, &drp->dr_stateid);
5020 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5022 get_stateid(cstate, &fsp->fr_stateid);
5026 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5028 get_stateid(cstate, &setattr->sa_stateid);
5032 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5034 get_stateid(cstate, &close->cl_stateid);
5038 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5040 get_stateid(cstate, &locku->lu_stateid);
5044 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5046 get_stateid(cstate, &read->rd_stateid);
5050 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5052 get_stateid(cstate, &write->wr_stateid);