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/addr.h>
46 #include "current_stateid.h"
50 #define NFSDDBG_FACILITY NFSDDBG_PROC
52 #define all_ones {{~0,~0},~0}
53 static const stateid_t one_stateid = {
55 .si_opaque = all_ones,
57 static const stateid_t zero_stateid = {
60 static const stateid_t currentstateid = {
64 static u64 current_sessionid = 1;
66 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
67 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
68 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
70 /* forward declarations */
71 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
75 /* Currently used for almost all code touching nfsv4 state: */
76 static DEFINE_MUTEX(client_mutex);
79 * Currently used for the del_recall_lru and file hash table. In an
80 * effort to decrease the scope of the client_mutex, this spinlock may
81 * eventually cover more:
83 static DEFINE_SPINLOCK(recall_lock);
85 static struct kmem_cache *openowner_slab = NULL;
86 static struct kmem_cache *lockowner_slab = NULL;
87 static struct kmem_cache *file_slab = NULL;
88 static struct kmem_cache *stateid_slab = NULL;
89 static struct kmem_cache *deleg_slab = NULL;
94 mutex_lock(&client_mutex);
97 static void free_session(struct kref *);
99 /* Must be called under the client_lock */
100 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
102 kref_put(&ses->se_ref, free_session);
105 static void nfsd4_get_session(struct nfsd4_session *ses)
107 kref_get(&ses->se_ref);
111 nfs4_unlock_state(void)
113 mutex_unlock(&client_mutex);
117 opaque_hashval(const void *ptr, int nbytes)
119 unsigned char *cptr = (unsigned char *) ptr;
129 static struct list_head del_recall_lru;
131 static void nfsd4_free_file(struct nfs4_file *f)
133 kmem_cache_free(file_slab, f);
137 put_nfs4_file(struct nfs4_file *fi)
139 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
140 list_del(&fi->fi_hash);
141 spin_unlock(&recall_lock);
148 get_nfs4_file(struct nfs4_file *fi)
150 atomic_inc(&fi->fi_ref);
153 static int num_delegations;
154 unsigned long max_delegations;
157 * Open owner state (share locks)
160 /* hash tables for lock and open owners */
161 #define OWNER_HASH_BITS 8
162 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
163 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
165 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
169 ret = opaque_hashval(ownername->data, ownername->len);
171 return ret & OWNER_HASH_MASK;
174 /* hash table for nfs4_file */
175 #define FILE_HASH_BITS 8
176 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
178 static unsigned int file_hashval(struct inode *ino)
180 /* XXX: why are we hashing on inode pointer, anyway? */
181 return hash_ptr(ino, FILE_HASH_BITS);
184 static struct list_head file_hashtbl[FILE_HASH_SIZE];
186 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
188 WARN_ON_ONCE(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
189 atomic_inc(&fp->fi_access[oflag]);
192 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
194 if (oflag == O_RDWR) {
195 __nfs4_file_get_access(fp, O_RDONLY);
196 __nfs4_file_get_access(fp, O_WRONLY);
198 __nfs4_file_get_access(fp, oflag);
201 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
203 if (fp->fi_fds[oflag]) {
204 fput(fp->fi_fds[oflag]);
205 fp->fi_fds[oflag] = NULL;
209 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
211 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
212 nfs4_file_put_fd(fp, oflag);
214 * It's also safe to get rid of the RDWR open *if*
215 * we no longer have need of the other kind of access
216 * or if we already have the other kind of open:
218 if (fp->fi_fds[1-oflag]
219 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
220 nfs4_file_put_fd(fp, O_RDWR);
224 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
226 if (oflag == O_RDWR) {
227 __nfs4_file_put_access(fp, O_RDONLY);
228 __nfs4_file_put_access(fp, O_WRONLY);
230 __nfs4_file_put_access(fp, oflag);
233 static inline int get_new_stid(struct nfs4_stid *stid)
235 static int min_stateid = 0;
236 struct idr *stateids = &stid->sc_client->cl_stateids;
240 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
242 * Note: the necessary preallocation was done in
243 * nfs4_alloc_stateid(). The idr code caps the number of
244 * preallocations that can exist at a time, but the state lock
245 * prevents anyone from using ours before we get here:
249 * It shouldn't be a problem to reuse an opaque stateid value.
250 * I don't think it is for 4.1. But with 4.0 I worry that, for
251 * example, a stray write retransmission could be accepted by
252 * the server when it should have been rejected. Therefore,
253 * adopt a trick from the sctp code to attempt to maximize the
254 * amount of time until an id is reused, by ensuring they always
255 * "increase" (mod INT_MAX):
258 min_stateid = new_stid+1;
259 if (min_stateid == INT_MAX)
264 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
267 struct idr *stateids = &cl->cl_stateids;
268 static int min_stateid = 0;
269 struct nfs4_stid *stid;
272 stid = kmem_cache_alloc(slab, GFP_KERNEL);
276 if (!idr_pre_get(stateids, GFP_KERNEL))
278 if (idr_get_new_above(stateids, stid, min_stateid, &new_id))
280 stid->sc_client = cl;
282 stid->sc_stateid.si_opaque.so_id = new_id;
283 stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
284 /* Will be incremented before return to client: */
285 stid->sc_stateid.si_generation = 0;
288 * It shouldn't be a problem to reuse an opaque stateid value.
289 * I don't think it is for 4.1. But with 4.0 I worry that, for
290 * example, a stray write retransmission could be accepted by
291 * the server when it should have been rejected. Therefore,
292 * adopt a trick from the sctp code to attempt to maximize the
293 * amount of time until an id is reused, by ensuring they always
294 * "increase" (mod INT_MAX):
297 min_stateid = new_id+1;
298 if (min_stateid == INT_MAX)
306 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
308 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
311 static struct nfs4_delegation *
312 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
314 struct nfs4_delegation *dp;
315 struct nfs4_file *fp = stp->st_file;
317 dprintk("NFSD alloc_init_deleg\n");
319 * Major work on the lease subsystem (for example, to support
320 * calbacks on stat) will be required before we can support
321 * write delegations properly.
323 if (type != NFS4_OPEN_DELEGATE_READ)
325 if (fp->fi_had_conflict)
327 if (num_delegations > max_delegations)
329 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
332 dp->dl_stid.sc_type = NFS4_DELEG_STID;
334 * delegation seqid's are never incremented. The 4.1 special
335 * meaning of seqid 0 isn't meaningful, really, but let's avoid
336 * 0 anyway just for consistency and use 1:
338 dp->dl_stid.sc_stateid.si_generation = 1;
340 INIT_LIST_HEAD(&dp->dl_perfile);
341 INIT_LIST_HEAD(&dp->dl_perclnt);
342 INIT_LIST_HEAD(&dp->dl_recall_lru);
346 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
348 atomic_set(&dp->dl_count, 1);
349 nfsd4_init_callback(&dp->dl_recall);
353 static void free_stid(struct nfs4_stid *s, struct kmem_cache *slab)
355 struct idr *stateids = &s->sc_client->cl_stateids;
357 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
358 kmem_cache_free(slab, s);
362 nfs4_put_delegation(struct nfs4_delegation *dp)
364 if (atomic_dec_and_test(&dp->dl_count)) {
365 dprintk("NFSD: freeing dp %p\n",dp);
366 put_nfs4_file(dp->dl_file);
367 free_stid(&dp->dl_stid, deleg_slab);
372 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
374 if (atomic_dec_and_test(&fp->fi_delegees)) {
375 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
377 fput(fp->fi_deleg_file);
378 fp->fi_deleg_file = NULL;
382 static void unhash_stid(struct nfs4_stid *s)
387 /* Called under the state lock. */
389 unhash_delegation(struct nfs4_delegation *dp)
391 unhash_stid(&dp->dl_stid);
392 list_del_init(&dp->dl_perclnt);
393 spin_lock(&recall_lock);
394 list_del_init(&dp->dl_perfile);
395 list_del_init(&dp->dl_recall_lru);
396 spin_unlock(&recall_lock);
397 nfs4_put_deleg_lease(dp->dl_file);
398 nfs4_put_delegation(dp);
405 static unsigned int clientid_hashval(u32 id)
407 return id & CLIENT_HASH_MASK;
410 static unsigned int clientstr_hashval(const char *name)
412 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
416 * We store the NONE, READ, WRITE, and BOTH bits separately in the
417 * st_{access,deny}_bmap field of the stateid, in order to track not
418 * only what share bits are currently in force, but also what
419 * combinations of share bits previous opens have used. This allows us
420 * to enforce the recommendation of rfc 3530 14.2.19 that the server
421 * return an error if the client attempt to downgrade to a combination
422 * of share bits not explicable by closing some of its previous opens.
424 * XXX: This enforcement is actually incomplete, since we don't keep
425 * track of access/deny bit combinations; so, e.g., we allow:
427 * OPEN allow read, deny write
428 * OPEN allow both, deny none
429 * DOWNGRADE allow read, deny none
431 * which we should reject.
434 bmap_to_share_mode(unsigned long bmap) {
436 unsigned int access = 0;
438 for (i = 1; i < 4; i++) {
439 if (test_bit(i, &bmap))
446 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
447 unsigned int access, deny;
449 access = bmap_to_share_mode(stp->st_access_bmap);
450 deny = bmap_to_share_mode(stp->st_deny_bmap);
451 if ((access & open->op_share_deny) || (deny & open->op_share_access))
456 /* set share access for a given stateid */
458 set_access(u32 access, struct nfs4_ol_stateid *stp)
460 __set_bit(access, &stp->st_access_bmap);
463 /* clear share access for a given stateid */
465 clear_access(u32 access, struct nfs4_ol_stateid *stp)
467 __clear_bit(access, &stp->st_access_bmap);
470 /* test whether a given stateid has access */
472 test_access(u32 access, struct nfs4_ol_stateid *stp)
474 return test_bit(access, &stp->st_access_bmap);
477 /* set share deny for a given stateid */
479 set_deny(u32 access, struct nfs4_ol_stateid *stp)
481 __set_bit(access, &stp->st_deny_bmap);
484 /* clear share deny for a given stateid */
486 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
488 __clear_bit(access, &stp->st_deny_bmap);
491 /* test whether a given stateid is denying specific access */
493 test_deny(u32 access, struct nfs4_ol_stateid *stp)
495 return test_bit(access, &stp->st_deny_bmap);
498 static int nfs4_access_to_omode(u32 access)
500 switch (access & NFS4_SHARE_ACCESS_BOTH) {
501 case NFS4_SHARE_ACCESS_READ:
503 case NFS4_SHARE_ACCESS_WRITE:
505 case NFS4_SHARE_ACCESS_BOTH:
512 /* release all access and file references for a given stateid */
514 release_all_access(struct nfs4_ol_stateid *stp)
518 for (i = 1; i < 4; i++) {
519 if (test_access(i, stp))
520 nfs4_file_put_access(stp->st_file,
521 nfs4_access_to_omode(i));
522 clear_access(i, stp);
526 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
528 list_del(&stp->st_perfile);
529 list_del(&stp->st_perstateowner);
532 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
534 release_all_access(stp);
535 put_nfs4_file(stp->st_file);
539 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
541 free_stid(&stp->st_stid, stateid_slab);
544 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
548 unhash_generic_stateid(stp);
549 unhash_stid(&stp->st_stid);
550 file = find_any_file(stp->st_file);
552 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
553 close_generic_stateid(stp);
554 free_generic_stateid(stp);
557 static void unhash_lockowner(struct nfs4_lockowner *lo)
559 struct nfs4_ol_stateid *stp;
561 list_del(&lo->lo_owner.so_strhash);
562 list_del(&lo->lo_perstateid);
563 list_del(&lo->lo_owner_ino_hash);
564 while (!list_empty(&lo->lo_owner.so_stateids)) {
565 stp = list_first_entry(&lo->lo_owner.so_stateids,
566 struct nfs4_ol_stateid, st_perstateowner);
567 release_lock_stateid(stp);
571 static void release_lockowner(struct nfs4_lockowner *lo)
573 unhash_lockowner(lo);
574 nfs4_free_lockowner(lo);
578 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
580 struct nfs4_lockowner *lo;
582 while (!list_empty(&open_stp->st_lockowners)) {
583 lo = list_entry(open_stp->st_lockowners.next,
584 struct nfs4_lockowner, lo_perstateid);
585 release_lockowner(lo);
589 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
591 unhash_generic_stateid(stp);
592 release_stateid_lockowners(stp);
593 close_generic_stateid(stp);
596 static void release_open_stateid(struct nfs4_ol_stateid *stp)
598 unhash_open_stateid(stp);
599 unhash_stid(&stp->st_stid);
600 free_generic_stateid(stp);
603 static void unhash_openowner(struct nfs4_openowner *oo)
605 struct nfs4_ol_stateid *stp;
607 list_del(&oo->oo_owner.so_strhash);
608 list_del(&oo->oo_perclient);
609 while (!list_empty(&oo->oo_owner.so_stateids)) {
610 stp = list_first_entry(&oo->oo_owner.so_stateids,
611 struct nfs4_ol_stateid, st_perstateowner);
612 release_open_stateid(stp);
616 static void release_last_closed_stateid(struct nfs4_openowner *oo)
618 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
621 unhash_stid(&s->st_stid);
622 free_generic_stateid(s);
623 oo->oo_last_closed_stid = NULL;
627 static void release_openowner(struct nfs4_openowner *oo)
629 unhash_openowner(oo);
630 list_del(&oo->oo_close_lru);
631 release_last_closed_stateid(oo);
632 nfs4_free_openowner(oo);
636 hash_sessionid(struct nfs4_sessionid *sessionid)
638 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
640 return sid->sequence % SESSION_HASH_SIZE;
645 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
647 u32 *ptr = (u32 *)(&sessionid->data[0]);
648 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
652 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
659 gen_sessionid(struct nfsd4_session *ses)
661 struct nfs4_client *clp = ses->se_client;
662 struct nfsd4_sessionid *sid;
664 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
665 sid->clientid = clp->cl_clientid;
666 sid->sequence = current_sessionid++;
671 * The protocol defines ca_maxresponssize_cached to include the size of
672 * the rpc header, but all we need to cache is the data starting after
673 * the end of the initial SEQUENCE operation--the rest we regenerate
674 * each time. Therefore we can advertise a ca_maxresponssize_cached
675 * value that is the number of bytes in our cache plus a few additional
676 * bytes. In order to stay on the safe side, and not promise more than
677 * we can cache, those additional bytes must be the minimum possible: 24
678 * bytes of rpc header (xid through accept state, with AUTH_NULL
679 * verifier), 12 for the compound header (with zero-length tag), and 44
680 * for the SEQUENCE op response:
682 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
685 free_session_slots(struct nfsd4_session *ses)
689 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
690 kfree(ses->se_slots[i]);
694 * We don't actually need to cache the rpc and session headers, so we
695 * can allocate a little less for each slot:
697 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
699 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
702 static int nfsd4_sanitize_slot_size(u32 size)
704 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
705 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
711 * XXX: If we run out of reserved DRC memory we could (up to a point)
712 * re-negotiate active sessions and reduce their slot usage to make
713 * room for new connections. For now we just fail the create session.
715 static int nfsd4_get_drc_mem(int slotsize, u32 num)
719 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
721 spin_lock(&nfsd_drc_lock);
722 avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
723 nfsd_drc_max_mem - nfsd_drc_mem_used);
724 num = min_t(int, num, avail / slotsize);
725 nfsd_drc_mem_used += num * slotsize;
726 spin_unlock(&nfsd_drc_lock);
731 static void nfsd4_put_drc_mem(int slotsize, int num)
733 spin_lock(&nfsd_drc_lock);
734 nfsd_drc_mem_used -= slotsize * num;
735 spin_unlock(&nfsd_drc_lock);
738 static struct nfsd4_session *__alloc_session(int slotsize, int numslots)
740 struct nfsd4_session *new;
743 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
744 + sizeof(struct nfsd4_session) > PAGE_SIZE);
745 mem = numslots * sizeof(struct nfsd4_slot *);
747 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
750 /* allocate each struct nfsd4_slot and data cache in one piece */
751 for (i = 0; i < numslots; i++) {
752 mem = sizeof(struct nfsd4_slot) + slotsize;
753 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
754 if (!new->se_slots[i])
760 kfree(new->se_slots[i]);
765 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new,
766 struct nfsd4_channel_attrs *req,
767 int numslots, int slotsize,
770 u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
772 new->maxreqs = numslots;
773 new->maxresp_cached = min_t(u32, req->maxresp_cached,
774 slotsize + NFSD_MIN_HDR_SEQ_SZ);
775 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
776 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
777 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
780 static void free_conn(struct nfsd4_conn *c)
782 svc_xprt_put(c->cn_xprt);
786 static void nfsd4_conn_lost(struct svc_xpt_user *u)
788 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
789 struct nfs4_client *clp = c->cn_session->se_client;
791 spin_lock(&clp->cl_lock);
792 if (!list_empty(&c->cn_persession)) {
793 list_del(&c->cn_persession);
796 spin_unlock(&clp->cl_lock);
797 nfsd4_probe_callback(clp);
800 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
802 struct nfsd4_conn *conn;
804 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
807 svc_xprt_get(rqstp->rq_xprt);
808 conn->cn_xprt = rqstp->rq_xprt;
809 conn->cn_flags = flags;
810 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
814 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
816 conn->cn_session = ses;
817 list_add(&conn->cn_persession, &ses->se_conns);
820 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
822 struct nfs4_client *clp = ses->se_client;
824 spin_lock(&clp->cl_lock);
825 __nfsd4_hash_conn(conn, ses);
826 spin_unlock(&clp->cl_lock);
829 static int nfsd4_register_conn(struct nfsd4_conn *conn)
831 conn->cn_xpt_user.callback = nfsd4_conn_lost;
832 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
835 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
839 nfsd4_hash_conn(conn, ses);
840 ret = nfsd4_register_conn(conn);
842 /* oops; xprt is already down: */
843 nfsd4_conn_lost(&conn->cn_xpt_user);
844 if (conn->cn_flags & NFS4_CDFC4_BACK) {
845 /* callback channel may be back up */
846 nfsd4_probe_callback(ses->se_client);
850 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
852 u32 dir = NFS4_CDFC4_FORE;
854 if (cses->flags & SESSION4_BACK_CHAN)
855 dir |= NFS4_CDFC4_BACK;
856 return alloc_conn(rqstp, dir);
859 /* must be called under client_lock */
860 static void nfsd4_del_conns(struct nfsd4_session *s)
862 struct nfs4_client *clp = s->se_client;
863 struct nfsd4_conn *c;
865 spin_lock(&clp->cl_lock);
866 while (!list_empty(&s->se_conns)) {
867 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
868 list_del_init(&c->cn_persession);
869 spin_unlock(&clp->cl_lock);
871 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
874 spin_lock(&clp->cl_lock);
876 spin_unlock(&clp->cl_lock);
879 static void __free_session(struct nfsd4_session *ses)
881 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
882 free_session_slots(ses);
886 static void free_session(struct kref *kref)
888 struct nfsd4_session *ses;
891 ses = container_of(kref, struct nfsd4_session, se_ref);
892 nn = net_generic(ses->se_client->net, nfsd_net_id);
894 lockdep_assert_held(&nn->client_lock);
895 nfsd4_del_conns(ses);
899 void nfsd4_put_session(struct nfsd4_session *ses)
901 struct nfsd_net *nn = net_generic(ses->se_client->net, nfsd_net_id);
903 spin_lock(&nn->client_lock);
904 nfsd4_put_session_locked(ses);
905 spin_unlock(&nn->client_lock);
908 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fchan,
911 struct nfsd4_session *new;
912 int numslots, slotsize;
914 * Note decreasing slot size below client's request may
915 * make it difficult for client to function correctly, whereas
916 * decreasing the number of slots will (just?) affect
917 * performance. When short on memory we therefore prefer to
918 * decrease number of slots instead of their size.
920 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
921 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
925 new = __alloc_session(slotsize, numslots);
927 nfsd4_put_drc_mem(slotsize, numslots);
930 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize, nn);
934 static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
937 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
939 new->se_client = clp;
942 INIT_LIST_HEAD(&new->se_conns);
944 new->se_cb_seq_nr = 1;
945 new->se_flags = cses->flags;
946 new->se_cb_prog = cses->callback_prog;
947 new->se_cb_sec = cses->cb_sec;
948 kref_init(&new->se_ref);
949 idx = hash_sessionid(&new->se_sessionid);
950 spin_lock(&nn->client_lock);
951 list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
952 spin_lock(&clp->cl_lock);
953 list_add(&new->se_perclnt, &clp->cl_sessions);
954 spin_unlock(&clp->cl_lock);
955 spin_unlock(&nn->client_lock);
957 if (cses->flags & SESSION4_BACK_CHAN) {
958 struct sockaddr *sa = svc_addr(rqstp);
960 * This is a little silly; with sessions there's no real
961 * use for the callback address. Use the peer address
962 * as a reasonable default for now, but consider fixing
963 * the rpc client not to require an address in the
966 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
967 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
971 /* caller must hold client_lock */
972 static struct nfsd4_session *
973 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
975 struct nfsd4_session *elem;
977 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
979 dump_sessionid(__func__, sessionid);
980 idx = hash_sessionid(sessionid);
981 /* Search in the appropriate list */
982 list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
983 if (!memcmp(elem->se_sessionid.data, sessionid->data,
984 NFS4_MAX_SESSIONID_LEN)) {
989 dprintk("%s: session not found\n", __func__);
993 /* caller must hold client_lock */
995 unhash_session(struct nfsd4_session *ses)
997 list_del(&ses->se_hash);
998 spin_lock(&ses->se_client->cl_lock);
999 list_del(&ses->se_perclnt);
1000 spin_unlock(&ses->se_client->cl_lock);
1003 /* must be called under the client_lock */
1005 renew_client_locked(struct nfs4_client *clp)
1007 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1009 if (is_client_expired(clp)) {
1011 printk("%s: client (clientid %08x/%08x) already expired\n",
1013 clp->cl_clientid.cl_boot,
1014 clp->cl_clientid.cl_id);
1018 dprintk("renewing client (clientid %08x/%08x)\n",
1019 clp->cl_clientid.cl_boot,
1020 clp->cl_clientid.cl_id);
1021 list_move_tail(&clp->cl_lru, &nn->client_lru);
1022 clp->cl_time = get_seconds();
1026 renew_client(struct nfs4_client *clp)
1028 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1030 spin_lock(&nn->client_lock);
1031 renew_client_locked(clp);
1032 spin_unlock(&nn->client_lock);
1035 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1037 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1039 if (clid->cl_boot == nn->boot_time)
1041 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1042 clid->cl_boot, clid->cl_id, nn->boot_time);
1047 * XXX Should we use a slab cache ?
1048 * This type of memory management is somewhat inefficient, but we use it
1049 * anyway since SETCLIENTID is not a common operation.
1051 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1053 struct nfs4_client *clp;
1055 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1058 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1059 if (clp->cl_name.data == NULL) {
1063 clp->cl_name.len = name.len;
1068 free_client(struct nfs4_client *clp)
1070 struct nfsd_net __maybe_unused *nn = net_generic(clp->net, nfsd_net_id);
1072 lockdep_assert_held(&nn->client_lock);
1073 while (!list_empty(&clp->cl_sessions)) {
1074 struct nfsd4_session *ses;
1075 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1077 list_del(&ses->se_perclnt);
1078 nfsd4_put_session_locked(ses);
1080 free_svc_cred(&clp->cl_cred);
1081 kfree(clp->cl_name.data);
1082 idr_destroy(&clp->cl_stateids);
1087 release_session_client(struct nfsd4_session *session)
1089 struct nfs4_client *clp = session->se_client;
1090 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1092 if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
1094 if (is_client_expired(clp)) {
1096 session->se_client = NULL;
1098 renew_client_locked(clp);
1099 spin_unlock(&nn->client_lock);
1102 /* must be called under the client_lock */
1104 unhash_client_locked(struct nfs4_client *clp)
1106 struct nfsd4_session *ses;
1108 mark_client_expired(clp);
1109 list_del(&clp->cl_lru);
1110 spin_lock(&clp->cl_lock);
1111 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1112 list_del_init(&ses->se_hash);
1113 spin_unlock(&clp->cl_lock);
1117 destroy_client(struct nfs4_client *clp)
1119 struct nfs4_openowner *oo;
1120 struct nfs4_delegation *dp;
1121 struct list_head reaplist;
1122 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1124 INIT_LIST_HEAD(&reaplist);
1125 spin_lock(&recall_lock);
1126 while (!list_empty(&clp->cl_delegations)) {
1127 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1128 list_del_init(&dp->dl_perclnt);
1129 list_move(&dp->dl_recall_lru, &reaplist);
1131 spin_unlock(&recall_lock);
1132 while (!list_empty(&reaplist)) {
1133 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1134 unhash_delegation(dp);
1136 while (!list_empty(&clp->cl_openowners)) {
1137 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1138 release_openowner(oo);
1140 nfsd4_shutdown_callback(clp);
1141 if (clp->cl_cb_conn.cb_xprt)
1142 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1143 list_del(&clp->cl_idhash);
1144 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1145 rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1147 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1148 spin_lock(&nn->client_lock);
1149 unhash_client_locked(clp);
1150 if (atomic_read(&clp->cl_refcount) == 0)
1152 spin_unlock(&nn->client_lock);
1155 static void expire_client(struct nfs4_client *clp)
1157 nfsd4_client_record_remove(clp);
1158 destroy_client(clp);
1161 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1163 memcpy(target->cl_verifier.data, source->data,
1164 sizeof(target->cl_verifier.data));
1167 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1169 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1170 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1173 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1175 if (source->cr_principal) {
1176 target->cr_principal =
1177 kstrdup(source->cr_principal, GFP_KERNEL);
1178 if (target->cr_principal == NULL)
1181 target->cr_principal = NULL;
1182 target->cr_flavor = source->cr_flavor;
1183 target->cr_uid = source->cr_uid;
1184 target->cr_gid = source->cr_gid;
1185 target->cr_group_info = source->cr_group_info;
1186 get_group_info(target->cr_group_info);
1191 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1195 res = o1->len - o2->len;
1198 return (long long)memcmp(o1->data, o2->data, o1->len);
1201 static int same_name(const char *n1, const char *n2)
1203 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1207 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1209 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1213 same_clid(clientid_t *cl1, clientid_t *cl2)
1215 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1218 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1222 if (g1->ngroups != g2->ngroups)
1224 for (i=0; i<g1->ngroups; i++)
1225 if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1231 * RFC 3530 language requires clid_inuse be returned when the
1232 * "principal" associated with a requests differs from that previously
1233 * used. We use uid, gid's, and gss principal string as our best
1234 * approximation. We also don't want to allow non-gss use of a client
1235 * established using gss: in theory cr_principal should catch that
1236 * change, but in practice cr_principal can be null even in the gss case
1237 * since gssd doesn't always pass down a principal string.
1239 static bool is_gss_cred(struct svc_cred *cr)
1241 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1242 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1247 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1249 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1250 || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1251 || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1252 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1254 if (cr1->cr_principal == cr2->cr_principal)
1256 if (!cr1->cr_principal || !cr2->cr_principal)
1258 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1261 static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1263 static u32 current_clientid = 1;
1265 clp->cl_clientid.cl_boot = nn->boot_time;
1266 clp->cl_clientid.cl_id = current_clientid++;
1269 static void gen_confirm(struct nfs4_client *clp)
1274 verf[0] = (__be32)get_seconds();
1275 verf[1] = (__be32)i++;
1276 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1279 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1281 struct nfs4_stid *ret;
1283 ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1284 if (!ret || !ret->sc_type)
1289 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1291 struct nfs4_stid *s;
1293 s = find_stateid(cl, t);
1296 if (typemask & s->sc_type)
1301 static struct nfs4_client *create_client(struct xdr_netobj name,
1302 struct svc_rqst *rqstp, nfs4_verifier *verf)
1304 struct nfs4_client *clp;
1305 struct sockaddr *sa = svc_addr(rqstp);
1307 struct net *net = SVC_NET(rqstp);
1308 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1310 clp = alloc_client(name);
1314 INIT_LIST_HEAD(&clp->cl_sessions);
1315 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1317 spin_lock(&nn->client_lock);
1319 spin_unlock(&nn->client_lock);
1322 idr_init(&clp->cl_stateids);
1323 atomic_set(&clp->cl_refcount, 0);
1324 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1325 INIT_LIST_HEAD(&clp->cl_idhash);
1326 INIT_LIST_HEAD(&clp->cl_openowners);
1327 INIT_LIST_HEAD(&clp->cl_delegations);
1328 INIT_LIST_HEAD(&clp->cl_lru);
1329 INIT_LIST_HEAD(&clp->cl_callbacks);
1330 spin_lock_init(&clp->cl_lock);
1331 nfsd4_init_callback(&clp->cl_cb_null);
1332 clp->cl_time = get_seconds();
1333 clear_bit(0, &clp->cl_cb_slot_busy);
1334 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1335 copy_verf(clp, verf);
1336 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1338 clp->cl_cb_session = NULL;
1344 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1346 struct rb_node **new = &(root->rb_node), *parent = NULL;
1347 struct nfs4_client *clp;
1350 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1353 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1354 new = &((*new)->rb_left);
1356 new = &((*new)->rb_right);
1359 rb_link_node(&new_clp->cl_namenode, parent, new);
1360 rb_insert_color(&new_clp->cl_namenode, root);
1363 static struct nfs4_client *
1364 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1367 struct rb_node *node = root->rb_node;
1368 struct nfs4_client *clp;
1371 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1372 cmp = compare_blob(&clp->cl_name, name);
1374 node = node->rb_left;
1376 node = node->rb_right;
1384 add_to_unconfirmed(struct nfs4_client *clp)
1386 unsigned int idhashval;
1387 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1389 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1390 add_clp_to_name_tree(clp, &nn->unconf_name_tree);
1391 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1392 list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
1397 move_to_confirmed(struct nfs4_client *clp)
1399 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1400 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1402 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1403 list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
1404 rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1405 add_clp_to_name_tree(clp, &nn->conf_name_tree);
1406 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1410 static struct nfs4_client *
1411 find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1413 struct nfs4_client *clp;
1414 unsigned int idhashval = clientid_hashval(clid->cl_id);
1416 list_for_each_entry(clp, &nn->conf_id_hashtbl[idhashval], cl_idhash) {
1417 if (same_clid(&clp->cl_clientid, clid)) {
1418 if ((bool)clp->cl_minorversion != sessions)
1427 static struct nfs4_client *
1428 find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
1430 struct nfs4_client *clp;
1431 unsigned int idhashval = clientid_hashval(clid->cl_id);
1433 list_for_each_entry(clp, &nn->unconf_id_hashtbl[idhashval], cl_idhash) {
1434 if (same_clid(&clp->cl_clientid, clid)) {
1435 if ((bool)clp->cl_minorversion != sessions)
1443 static bool clp_used_exchangeid(struct nfs4_client *clp)
1445 return clp->cl_exchange_flags != 0;
1448 static struct nfs4_client *
1449 find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1451 return find_clp_in_name_tree(name, &nn->conf_name_tree);
1454 static struct nfs4_client *
1455 find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
1457 return find_clp_in_name_tree(name, &nn->unconf_name_tree);
1461 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1463 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1464 struct sockaddr *sa = svc_addr(rqstp);
1465 u32 scopeid = rpc_get_scope_id(sa);
1466 unsigned short expected_family;
1468 /* Currently, we only support tcp and tcp6 for the callback channel */
1469 if (se->se_callback_netid_len == 3 &&
1470 !memcmp(se->se_callback_netid_val, "tcp", 3))
1471 expected_family = AF_INET;
1472 else if (se->se_callback_netid_len == 4 &&
1473 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1474 expected_family = AF_INET6;
1478 conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
1479 se->se_callback_addr_len,
1480 (struct sockaddr *)&conn->cb_addr,
1481 sizeof(conn->cb_addr));
1483 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1486 if (conn->cb_addr.ss_family == AF_INET6)
1487 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1489 conn->cb_prog = se->se_callback_prog;
1490 conn->cb_ident = se->se_callback_ident;
1491 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1494 conn->cb_addr.ss_family = AF_UNSPEC;
1495 conn->cb_addrlen = 0;
1496 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1497 "will not receive delegations\n",
1498 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1504 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1507 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1509 struct nfsd4_slot *slot = resp->cstate.slot;
1512 dprintk("--> %s slot %p\n", __func__, slot);
1514 slot->sl_opcnt = resp->opcnt;
1515 slot->sl_status = resp->cstate.status;
1517 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1518 if (nfsd4_not_cached(resp)) {
1519 slot->sl_datalen = 0;
1522 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1523 base = (char *)resp->cstate.datap -
1524 (char *)resp->xbuf->head[0].iov_base;
1525 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1527 WARN("%s: sessions DRC could not cache compound\n", __func__);
1532 * Encode the replay sequence operation from the slot values.
1533 * If cachethis is FALSE encode the uncached rep error on the next
1534 * operation which sets resp->p and increments resp->opcnt for
1535 * nfs4svc_encode_compoundres.
1539 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1540 struct nfsd4_compoundres *resp)
1542 struct nfsd4_op *op;
1543 struct nfsd4_slot *slot = resp->cstate.slot;
1545 /* Encode the replayed sequence operation */
1546 op = &args->ops[resp->opcnt - 1];
1547 nfsd4_encode_operation(resp, op);
1549 /* Return nfserr_retry_uncached_rep in next operation. */
1550 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1551 op = &args->ops[resp->opcnt++];
1552 op->status = nfserr_retry_uncached_rep;
1553 nfsd4_encode_operation(resp, op);
1559 * The sequence operation is not cached because we can use the slot and
1563 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1564 struct nfsd4_sequence *seq)
1566 struct nfsd4_slot *slot = resp->cstate.slot;
1569 dprintk("--> %s slot %p\n", __func__, slot);
1571 /* Either returns 0 or nfserr_retry_uncached */
1572 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1573 if (status == nfserr_retry_uncached_rep)
1576 /* The sequence operation has been encoded, cstate->datap set. */
1577 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1579 resp->opcnt = slot->sl_opcnt;
1580 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1581 status = slot->sl_status;
1587 * Set the exchange_id flags returned by the server.
1590 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1592 /* pNFS is not supported */
1593 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1595 /* Referrals are supported, Migration is not. */
1596 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1598 /* set the wire flags to return to client. */
1599 clid->flags = new->cl_exchange_flags;
1602 static bool client_has_state(struct nfs4_client *clp)
1605 * Note clp->cl_openowners check isn't quite right: there's no
1606 * need to count owners without stateid's.
1608 * Also note we should probably be using this in 4.0 case too.
1610 return !list_empty(&clp->cl_openowners)
1611 || !list_empty(&clp->cl_delegations)
1612 || !list_empty(&clp->cl_sessions);
1616 nfsd4_exchange_id(struct svc_rqst *rqstp,
1617 struct nfsd4_compound_state *cstate,
1618 struct nfsd4_exchange_id *exid)
1620 struct nfs4_client *unconf, *conf, *new;
1622 char addr_str[INET6_ADDRSTRLEN];
1623 nfs4_verifier verf = exid->verifier;
1624 struct sockaddr *sa = svc_addr(rqstp);
1625 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1626 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1628 rpc_ntop(sa, addr_str, sizeof(addr_str));
1629 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1630 "ip_addr=%s flags %x, spa_how %d\n",
1631 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1632 addr_str, exid->flags, exid->spa_how);
1634 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1635 return nfserr_inval;
1637 /* Currently only support SP4_NONE */
1638 switch (exid->spa_how) {
1641 default: /* checked by xdr code */
1645 return nfserr_serverfault; /* no excuse :-/ */
1648 /* Cases below refer to rfc 5661 section 18.35.4: */
1650 conf = find_confirmed_client_by_name(&exid->clname, nn);
1652 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1653 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1656 if (!clp_used_exchangeid(conf)) { /* buggy client */
1657 status = nfserr_inval;
1660 if (!creds_match) { /* case 9 */
1661 status = nfserr_perm;
1664 if (!verfs_match) { /* case 8 */
1665 status = nfserr_not_same;
1669 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1673 if (!creds_match) { /* case 3 */
1674 if (client_has_state(conf)) {
1675 status = nfserr_clid_inuse;
1678 expire_client(conf);
1681 if (verfs_match) { /* case 2 */
1682 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1686 /* case 5, client reboot */
1690 if (update) { /* case 7 */
1691 status = nfserr_noent;
1695 unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
1696 if (unconf) /* case 4, possible retry or client restart */
1697 expire_client(unconf);
1699 /* case 1 (normal case) */
1701 new = create_client(exid->clname, rqstp, &verf);
1703 status = nfserr_jukebox;
1706 new->cl_minorversion = 1;
1709 add_to_unconfirmed(new);
1711 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1712 exid->clientid.cl_id = new->cl_clientid.cl_id;
1714 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1715 nfsd4_set_ex_flags(new, exid);
1717 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1718 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1722 nfs4_unlock_state();
1727 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1729 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1732 /* The slot is in use, and no response has been sent. */
1734 if (seqid == slot_seqid)
1735 return nfserr_jukebox;
1737 return nfserr_seq_misordered;
1739 /* Note unsigned 32-bit arithmetic handles wraparound: */
1740 if (likely(seqid == slot_seqid + 1))
1742 if (seqid == slot_seqid)
1743 return nfserr_replay_cache;
1744 return nfserr_seq_misordered;
1748 * Cache the create session result into the create session single DRC
1749 * slot cache by saving the xdr structure. sl_seqid has been set.
1750 * Do this for solo or embedded create session operations.
1753 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1754 struct nfsd4_clid_slot *slot, __be32 nfserr)
1756 slot->sl_status = nfserr;
1757 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1761 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1762 struct nfsd4_clid_slot *slot)
1764 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1765 return slot->sl_status;
1768 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1769 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1770 1 + /* MIN tag is length with zero, only length */ \
1771 3 + /* version, opcount, opcode */ \
1772 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1773 /* seqid, slotID, slotID, cache */ \
1774 4 ) * sizeof(__be32))
1776 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1777 2 + /* verifier: AUTH_NULL, length 0 */\
1779 1 + /* MIN tag is length with zero, only length */ \
1780 3 + /* opcount, opcode, opstatus*/ \
1781 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1782 /* seqid, slotID, slotID, slotID, status */ \
1783 5 ) * sizeof(__be32))
1785 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1787 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1788 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1792 nfsd4_create_session(struct svc_rqst *rqstp,
1793 struct nfsd4_compound_state *cstate,
1794 struct nfsd4_create_session *cr_ses)
1796 struct sockaddr *sa = svc_addr(rqstp);
1797 struct nfs4_client *conf, *unconf;
1798 struct nfsd4_session *new;
1799 struct nfsd4_conn *conn;
1800 struct nfsd4_clid_slot *cs_slot = NULL;
1802 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1804 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1805 return nfserr_inval;
1806 if (check_forechannel_attrs(cr_ses->fore_channel))
1807 return nfserr_toosmall;
1808 new = alloc_session(&cr_ses->fore_channel, nn);
1810 return nfserr_jukebox;
1811 status = nfserr_jukebox;
1812 conn = alloc_conn_from_crses(rqstp, cr_ses);
1814 goto out_free_session;
1817 unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
1818 conf = find_confirmed_client(&cr_ses->clientid, true, nn);
1821 cs_slot = &conf->cl_cs_slot;
1822 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1823 if (status == nfserr_replay_cache) {
1824 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1826 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1827 status = nfserr_seq_misordered;
1830 } else if (unconf) {
1831 struct nfs4_client *old;
1832 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1833 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1834 status = nfserr_clid_inuse;
1837 cs_slot = &unconf->cl_cs_slot;
1838 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1840 /* an unconfirmed replay returns misordered */
1841 status = nfserr_seq_misordered;
1844 old = find_confirmed_client_by_name(&unconf->cl_name, nn);
1847 move_to_confirmed(unconf);
1850 status = nfserr_stale_clientid;
1855 * We do not support RDMA or persistent sessions
1857 cr_ses->flags &= ~SESSION4_PERSIST;
1858 cr_ses->flags &= ~SESSION4_RDMA;
1860 init_session(rqstp, new, conf, cr_ses);
1861 nfsd4_init_conn(rqstp, conn, new);
1863 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1864 NFS4_MAX_SESSIONID_LEN);
1865 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1866 sizeof(struct nfsd4_channel_attrs));
1867 cs_slot->sl_seqid++;
1868 cr_ses->seqid = cs_slot->sl_seqid;
1870 /* cache solo and embedded create sessions under the state lock */
1871 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1872 nfs4_unlock_state();
1874 dprintk("%s returns %d\n", __func__, ntohl(status));
1877 nfs4_unlock_state();
1880 __free_session(new);
1884 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1887 case NFS4_CDFC4_FORE:
1888 case NFS4_CDFC4_BACK:
1890 case NFS4_CDFC4_FORE_OR_BOTH:
1891 case NFS4_CDFC4_BACK_OR_BOTH:
1892 *dir = NFS4_CDFC4_BOTH;
1895 return nfserr_inval;
1898 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1900 struct nfsd4_session *session = cstate->session;
1901 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1903 spin_lock(&nn->client_lock);
1904 session->se_cb_prog = bc->bc_cb_program;
1905 session->se_cb_sec = bc->bc_cb_sec;
1906 spin_unlock(&nn->client_lock);
1908 nfsd4_probe_callback(session->se_client);
1913 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1914 struct nfsd4_compound_state *cstate,
1915 struct nfsd4_bind_conn_to_session *bcts)
1918 struct nfsd4_conn *conn;
1919 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1921 if (!nfsd4_last_compound_op(rqstp))
1922 return nfserr_not_only_op;
1923 spin_lock(&nn->client_lock);
1924 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid, SVC_NET(rqstp));
1925 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1926 * client_lock iself: */
1927 if (cstate->session) {
1928 nfsd4_get_session(cstate->session);
1929 atomic_inc(&cstate->session->se_client->cl_refcount);
1931 spin_unlock(&nn->client_lock);
1932 if (!cstate->session)
1933 return nfserr_badsession;
1935 status = nfsd4_map_bcts_dir(&bcts->dir);
1938 conn = alloc_conn(rqstp, bcts->dir);
1940 return nfserr_jukebox;
1941 nfsd4_init_conn(rqstp, conn, cstate->session);
1945 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1949 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1953 nfsd4_destroy_session(struct svc_rqst *r,
1954 struct nfsd4_compound_state *cstate,
1955 struct nfsd4_destroy_session *sessionid)
1957 struct nfsd4_session *ses;
1958 __be32 status = nfserr_badsession;
1959 struct nfsd_net *nn = net_generic(SVC_NET(r), nfsd_net_id);
1962 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1963 * - Should we return nfserr_back_chan_busy if waiting for
1964 * callbacks on to-be-destroyed session?
1965 * - Do we need to clear any callback info from previous session?
1968 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1969 if (!nfsd4_last_compound_op(r))
1970 return nfserr_not_only_op;
1972 dump_sessionid(__func__, &sessionid->sessionid);
1973 spin_lock(&nn->client_lock);
1974 ses = find_in_sessionid_hashtbl(&sessionid->sessionid, SVC_NET(r));
1976 spin_unlock(&nn->client_lock);
1980 unhash_session(ses);
1981 spin_unlock(&nn->client_lock);
1984 nfsd4_probe_callback_sync(ses->se_client);
1985 nfs4_unlock_state();
1987 spin_lock(&nn->client_lock);
1988 nfsd4_del_conns(ses);
1989 nfsd4_put_session_locked(ses);
1990 spin_unlock(&nn->client_lock);
1993 dprintk("%s returns %d\n", __func__, ntohl(status));
1997 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1999 struct nfsd4_conn *c;
2001 list_for_each_entry(c, &s->se_conns, cn_persession) {
2002 if (c->cn_xprt == xpt) {
2009 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2011 struct nfs4_client *clp = ses->se_client;
2012 struct nfsd4_conn *c;
2015 spin_lock(&clp->cl_lock);
2016 c = __nfsd4_find_conn(new->cn_xprt, ses);
2018 spin_unlock(&clp->cl_lock);
2022 __nfsd4_hash_conn(new, ses);
2023 spin_unlock(&clp->cl_lock);
2024 ret = nfsd4_register_conn(new);
2026 /* oops; xprt is already down: */
2027 nfsd4_conn_lost(&new->cn_xpt_user);
2031 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2033 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2035 return args->opcnt > session->se_fchannel.maxops;
2038 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2039 struct nfsd4_session *session)
2041 struct xdr_buf *xb = &rqstp->rq_arg;
2043 return xb->len > session->se_fchannel.maxreq_sz;
2047 nfsd4_sequence(struct svc_rqst *rqstp,
2048 struct nfsd4_compound_state *cstate,
2049 struct nfsd4_sequence *seq)
2051 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2052 struct nfsd4_session *session;
2053 struct nfsd4_slot *slot;
2054 struct nfsd4_conn *conn;
2056 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2058 if (resp->opcnt != 1)
2059 return nfserr_sequence_pos;
2062 * Will be either used or freed by nfsd4_sequence_check_conn
2065 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2067 return nfserr_jukebox;
2069 spin_lock(&nn->client_lock);
2070 status = nfserr_badsession;
2071 session = find_in_sessionid_hashtbl(&seq->sessionid, SVC_NET(rqstp));
2075 status = nfserr_too_many_ops;
2076 if (nfsd4_session_too_many_ops(rqstp, session))
2079 status = nfserr_req_too_big;
2080 if (nfsd4_request_too_big(rqstp, session))
2083 status = nfserr_badslot;
2084 if (seq->slotid >= session->se_fchannel.maxreqs)
2087 slot = session->se_slots[seq->slotid];
2088 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2090 /* We do not negotiate the number of slots yet, so set the
2091 * maxslots to the session maxreqs which is used to encode
2092 * sr_highest_slotid and the sr_target_slot id to maxslots */
2093 seq->maxslots = session->se_fchannel.maxreqs;
2095 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2096 slot->sl_flags & NFSD4_SLOT_INUSE);
2097 if (status == nfserr_replay_cache) {
2098 status = nfserr_seq_misordered;
2099 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2101 cstate->slot = slot;
2102 cstate->session = session;
2103 /* Return the cached reply status and set cstate->status
2104 * for nfsd4_proc_compound processing */
2105 status = nfsd4_replay_cache_entry(resp, seq);
2106 cstate->status = nfserr_replay_cache;
2112 nfsd4_sequence_check_conn(conn, session);
2115 /* Success! bump slot seqid */
2116 slot->sl_seqid = seq->seqid;
2117 slot->sl_flags |= NFSD4_SLOT_INUSE;
2119 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2121 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2123 cstate->slot = slot;
2124 cstate->session = session;
2127 /* Hold a session reference until done processing the compound. */
2128 if (cstate->session) {
2129 struct nfs4_client *clp = session->se_client;
2131 nfsd4_get_session(cstate->session);
2132 atomic_inc(&clp->cl_refcount);
2133 switch (clp->cl_cb_state) {
2135 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2137 case NFSD4_CB_FAULT:
2138 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2141 seq->status_flags = 0;
2145 spin_unlock(&nn->client_lock);
2146 dprintk("%s: return %d\n", __func__, ntohl(status));
2151 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2153 struct nfs4_client *conf, *unconf, *clp;
2155 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2158 unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2159 conf = find_confirmed_client(&dc->clientid, true, nn);
2164 if (!is_client_expired(conf) && client_has_state(conf)) {
2165 status = nfserr_clientid_busy;
2169 /* rfc5661 18.50.3 */
2170 if (cstate->session && conf == cstate->session->se_client) {
2171 status = nfserr_clientid_busy;
2177 status = nfserr_stale_clientid;
2183 nfs4_unlock_state();
2184 dprintk("%s return %d\n", __func__, ntohl(status));
2189 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2193 if (rc->rca_one_fs) {
2194 if (!cstate->current_fh.fh_dentry)
2195 return nfserr_nofilehandle;
2197 * We don't take advantage of the rca_one_fs case.
2198 * That's OK, it's optional, we can safely ignore it.
2204 status = nfserr_complete_already;
2205 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2206 &cstate->session->se_client->cl_flags))
2209 status = nfserr_stale_clientid;
2210 if (is_client_expired(cstate->session->se_client))
2212 * The following error isn't really legal.
2213 * But we only get here if the client just explicitly
2214 * destroyed the client. Surely it no longer cares what
2215 * error it gets back on an operation for the dead
2221 nfsd4_client_record_create(cstate->session->se_client);
2223 nfs4_unlock_state();
2228 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2229 struct nfsd4_setclientid *setclid)
2231 struct xdr_netobj clname = setclid->se_name;
2232 nfs4_verifier clverifier = setclid->se_verf;
2233 struct nfs4_client *conf, *unconf, *new;
2235 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2237 /* Cases below refer to rfc 3530 section 14.2.33: */
2239 conf = find_confirmed_client_by_name(&clname, nn);
2242 status = nfserr_clid_inuse;
2243 if (clp_used_exchangeid(conf))
2245 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2246 char addr_str[INET6_ADDRSTRLEN];
2247 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2249 dprintk("NFSD: setclientid: string in use by client "
2250 "at %s\n", addr_str);
2254 unconf = find_unconfirmed_client_by_name(&clname, nn);
2256 expire_client(unconf);
2257 status = nfserr_jukebox;
2258 new = create_client(clname, rqstp, &clverifier);
2261 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2262 /* case 1: probable callback update */
2263 copy_clid(new, conf);
2264 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2266 new->cl_minorversion = 0;
2267 gen_callback(new, setclid, rqstp);
2268 add_to_unconfirmed(new);
2269 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2270 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2271 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2274 nfs4_unlock_state();
2280 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2281 struct nfsd4_compound_state *cstate,
2282 struct nfsd4_setclientid_confirm *setclientid_confirm)
2284 struct nfs4_client *conf, *unconf;
2285 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2286 clientid_t * clid = &setclientid_confirm->sc_clientid;
2288 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2290 if (STALE_CLIENTID(clid, nn))
2291 return nfserr_stale_clientid;
2294 conf = find_confirmed_client(clid, false, nn);
2295 unconf = find_unconfirmed_client(clid, false, nn);
2297 * We try hard to give out unique clientid's, so if we get an
2298 * attempt to confirm the same clientid with a different cred,
2299 * there's a bug somewhere. Let's charitably assume it's our
2302 status = nfserr_serverfault;
2303 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2305 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2307 /* cases below refer to rfc 3530 section 14.2.34: */
2308 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2309 if (conf && !unconf) /* case 2: probable retransmit */
2311 else /* case 4: client hasn't noticed we rebooted yet? */
2312 status = nfserr_stale_clientid;
2316 if (conf) { /* case 1: callback update */
2317 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2318 nfsd4_probe_callback(conf);
2319 expire_client(unconf);
2320 } else { /* case 3: normal case; new or rebooted client */
2321 conf = find_confirmed_client_by_name(&unconf->cl_name, nn);
2323 expire_client(conf);
2324 move_to_confirmed(unconf);
2325 nfsd4_probe_callback(unconf);
2328 nfs4_unlock_state();
2332 static struct nfs4_file *nfsd4_alloc_file(void)
2334 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2337 /* OPEN Share state helper functions */
2338 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2340 unsigned int hashval = file_hashval(ino);
2342 atomic_set(&fp->fi_ref, 1);
2343 INIT_LIST_HEAD(&fp->fi_hash);
2344 INIT_LIST_HEAD(&fp->fi_stateids);
2345 INIT_LIST_HEAD(&fp->fi_delegations);
2346 fp->fi_inode = igrab(ino);
2347 fp->fi_had_conflict = false;
2348 fp->fi_lease = NULL;
2349 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2350 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2351 spin_lock(&recall_lock);
2352 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2353 spin_unlock(&recall_lock);
2357 nfsd4_free_slab(struct kmem_cache **slab)
2361 kmem_cache_destroy(*slab);
2366 nfsd4_free_slabs(void)
2368 nfsd4_free_slab(&openowner_slab);
2369 nfsd4_free_slab(&lockowner_slab);
2370 nfsd4_free_slab(&file_slab);
2371 nfsd4_free_slab(&stateid_slab);
2372 nfsd4_free_slab(&deleg_slab);
2376 nfsd4_init_slabs(void)
2378 openowner_slab = kmem_cache_create("nfsd4_openowners",
2379 sizeof(struct nfs4_openowner), 0, 0, NULL);
2380 if (openowner_slab == NULL)
2382 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2383 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2384 if (lockowner_slab == NULL)
2386 file_slab = kmem_cache_create("nfsd4_files",
2387 sizeof(struct nfs4_file), 0, 0, NULL);
2388 if (file_slab == NULL)
2390 stateid_slab = kmem_cache_create("nfsd4_stateids",
2391 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2392 if (stateid_slab == NULL)
2394 deleg_slab = kmem_cache_create("nfsd4_delegations",
2395 sizeof(struct nfs4_delegation), 0, 0, NULL);
2396 if (deleg_slab == NULL)
2401 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2405 void nfs4_free_openowner(struct nfs4_openowner *oo)
2407 kfree(oo->oo_owner.so_owner.data);
2408 kmem_cache_free(openowner_slab, oo);
2411 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2413 kfree(lo->lo_owner.so_owner.data);
2414 kmem_cache_free(lockowner_slab, lo);
2417 static void init_nfs4_replay(struct nfs4_replay *rp)
2419 rp->rp_status = nfserr_serverfault;
2421 rp->rp_buf = rp->rp_ibuf;
2424 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2426 struct nfs4_stateowner *sop;
2428 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2432 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2433 if (!sop->so_owner.data) {
2434 kmem_cache_free(slab, sop);
2437 sop->so_owner.len = owner->len;
2439 INIT_LIST_HEAD(&sop->so_stateids);
2440 sop->so_client = clp;
2441 init_nfs4_replay(&sop->so_replay);
2445 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2447 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2449 list_add(&oo->oo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
2450 list_add(&oo->oo_perclient, &clp->cl_openowners);
2453 static struct nfs4_openowner *
2454 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2455 struct nfs4_openowner *oo;
2457 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2460 oo->oo_owner.so_is_open_owner = 1;
2461 oo->oo_owner.so_seqid = open->op_seqid;
2462 oo->oo_flags = NFS4_OO_NEW;
2464 oo->oo_last_closed_stid = NULL;
2465 INIT_LIST_HEAD(&oo->oo_close_lru);
2466 hash_openowner(oo, clp, strhashval);
2470 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2471 struct nfs4_openowner *oo = open->op_openowner;
2473 stp->st_stid.sc_type = NFS4_OPEN_STID;
2474 INIT_LIST_HEAD(&stp->st_lockowners);
2475 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2476 list_add(&stp->st_perfile, &fp->fi_stateids);
2477 stp->st_stateowner = &oo->oo_owner;
2480 stp->st_access_bmap = 0;
2481 stp->st_deny_bmap = 0;
2482 set_access(open->op_share_access, stp);
2483 set_deny(open->op_share_deny, stp);
2484 stp->st_openstp = NULL;
2488 move_to_close_lru(struct nfs4_openowner *oo, struct net *net)
2490 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2492 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2494 list_move_tail(&oo->oo_close_lru, &nn->close_lru);
2495 oo->oo_time = get_seconds();
2499 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2502 return (sop->so_owner.len == owner->len) &&
2503 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2504 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2507 static struct nfs4_openowner *
2508 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
2509 bool sessions, struct nfsd_net *nn)
2511 struct nfs4_stateowner *so;
2512 struct nfs4_openowner *oo;
2513 struct nfs4_client *clp;
2515 list_for_each_entry(so, &nn->ownerstr_hashtbl[hashval], so_strhash) {
2516 if (!so->so_is_open_owner)
2518 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2520 clp = oo->oo_owner.so_client;
2521 if ((bool)clp->cl_minorversion != sessions)
2523 renew_client(oo->oo_owner.so_client);
2530 /* search file_hashtbl[] for file */
2531 static struct nfs4_file *
2532 find_file(struct inode *ino)
2534 unsigned int hashval = file_hashval(ino);
2535 struct nfs4_file *fp;
2537 spin_lock(&recall_lock);
2538 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2539 if (fp->fi_inode == ino) {
2541 spin_unlock(&recall_lock);
2545 spin_unlock(&recall_lock);
2550 * Called to check deny when READ with all zero stateid or
2551 * WRITE with all zero or all one stateid
2554 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2556 struct inode *ino = current_fh->fh_dentry->d_inode;
2557 struct nfs4_file *fp;
2558 struct nfs4_ol_stateid *stp;
2561 dprintk("NFSD: nfs4_share_conflict\n");
2563 fp = find_file(ino);
2566 ret = nfserr_locked;
2567 /* Search for conflicting share reservations */
2568 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2569 if (test_deny(deny_type, stp) ||
2570 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2579 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2581 /* We're assuming the state code never drops its reference
2582 * without first removing the lease. Since we're in this lease
2583 * callback (and since the lease code is serialized by the kernel
2584 * lock) we know the server hasn't removed the lease yet, we know
2585 * it's safe to take a reference: */
2586 atomic_inc(&dp->dl_count);
2588 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2590 /* only place dl_time is set. protected by lock_flocks*/
2591 dp->dl_time = get_seconds();
2593 nfsd4_cb_recall(dp);
2596 /* Called from break_lease() with lock_flocks() held. */
2597 static void nfsd_break_deleg_cb(struct file_lock *fl)
2599 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2600 struct nfs4_delegation *dp;
2603 WARN(1, "(%p)->fl_owner NULL\n", fl);
2606 if (fp->fi_had_conflict) {
2607 WARN(1, "duplicate break on %p\n", fp);
2611 * We don't want the locks code to timeout the lease for us;
2612 * we'll remove it ourself if a delegation isn't returned
2615 fl->fl_break_time = 0;
2617 spin_lock(&recall_lock);
2618 fp->fi_had_conflict = true;
2619 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2620 nfsd_break_one_deleg(dp);
2621 spin_unlock(&recall_lock);
2625 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2628 return lease_modify(onlist, arg);
2633 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2634 .lm_break = nfsd_break_deleg_cb,
2635 .lm_change = nfsd_change_deleg_cb,
2638 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2640 if (nfsd4_has_session(cstate))
2642 if (seqid == so->so_seqid - 1)
2643 return nfserr_replay_me;
2644 if (seqid == so->so_seqid)
2646 return nfserr_bad_seqid;
2650 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2651 struct nfsd4_open *open, struct nfsd_net *nn)
2653 clientid_t *clientid = &open->op_clientid;
2654 struct nfs4_client *clp = NULL;
2655 unsigned int strhashval;
2656 struct nfs4_openowner *oo = NULL;
2659 if (STALE_CLIENTID(&open->op_clientid, nn))
2660 return nfserr_stale_clientid;
2662 * In case we need it later, after we've already created the
2663 * file and don't want to risk a further failure:
2665 open->op_file = nfsd4_alloc_file();
2666 if (open->op_file == NULL)
2667 return nfserr_jukebox;
2669 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2670 oo = find_openstateowner_str(strhashval, open, cstate->minorversion, nn);
2671 open->op_openowner = oo;
2673 clp = find_confirmed_client(clientid, cstate->minorversion,
2676 return nfserr_expired;
2679 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2680 /* Replace unconfirmed owners without checking for replay. */
2681 clp = oo->oo_owner.so_client;
2682 release_openowner(oo);
2683 open->op_openowner = NULL;
2686 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2689 clp = oo->oo_owner.so_client;
2692 oo = alloc_init_open_stateowner(strhashval, clp, open);
2694 return nfserr_jukebox;
2695 open->op_openowner = oo;
2697 open->op_stp = nfs4_alloc_stateid(clp);
2699 return nfserr_jukebox;
2703 static inline __be32
2704 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2706 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2707 return nfserr_openmode;
2712 static int share_access_to_flags(u32 share_access)
2714 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2717 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2719 struct nfs4_stid *ret;
2721 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2724 return delegstateid(ret);
2727 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2729 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2730 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2734 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2735 struct nfs4_delegation **dp)
2738 __be32 status = nfserr_bad_stateid;
2740 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2743 flags = share_access_to_flags(open->op_share_access);
2744 status = nfs4_check_delegmode(*dp, flags);
2748 if (!nfsd4_is_deleg_cur(open))
2752 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2757 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2759 struct nfs4_ol_stateid *local;
2760 struct nfs4_openowner *oo = open->op_openowner;
2762 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2763 /* ignore lock owners */
2764 if (local->st_stateowner->so_is_open_owner == 0)
2766 /* remember if we have seen this open owner */
2767 if (local->st_stateowner == &oo->oo_owner)
2769 /* check for conflicting share reservations */
2770 if (!test_share(local, open))
2771 return nfserr_share_denied;
2776 static inline int nfs4_access_to_access(u32 nfs4_access)
2780 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2781 flags |= NFSD_MAY_READ;
2782 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2783 flags |= NFSD_MAY_WRITE;
2787 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2788 struct svc_fh *cur_fh, struct nfsd4_open *open)
2791 int oflag = nfs4_access_to_omode(open->op_share_access);
2792 int access = nfs4_access_to_access(open->op_share_access);
2794 if (!fp->fi_fds[oflag]) {
2795 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2796 &fp->fi_fds[oflag]);
2800 nfs4_file_get_access(fp, oflag);
2805 static inline __be32
2806 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2807 struct nfsd4_open *open)
2809 struct iattr iattr = {
2810 .ia_valid = ATTR_SIZE,
2813 if (!open->op_truncate)
2815 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2816 return nfserr_inval;
2817 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2821 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)
2823 u32 op_share_access = open->op_share_access;
2827 new_access = !test_access(op_share_access, stp);
2829 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2833 status = nfsd4_truncate(rqstp, cur_fh, open);
2836 int oflag = nfs4_access_to_omode(op_share_access);
2837 nfs4_file_put_access(fp, oflag);
2841 /* remember the open */
2842 set_access(op_share_access, stp);
2843 set_deny(open->op_share_deny, stp);
2850 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2852 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2855 /* Should we give out recallable state?: */
2856 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2858 if (clp->cl_cb_state == NFSD4_CB_UP)
2861 * In the sessions case, since we don't have to establish a
2862 * separate connection for callbacks, we assume it's OK
2863 * until we hear otherwise:
2865 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2868 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2870 struct file_lock *fl;
2872 fl = locks_alloc_lock();
2875 locks_init_lock(fl);
2876 fl->fl_lmops = &nfsd_lease_mng_ops;
2877 fl->fl_flags = FL_LEASE;
2878 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2879 fl->fl_end = OFFSET_MAX;
2880 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2881 fl->fl_pid = current->tgid;
2885 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2887 struct nfs4_file *fp = dp->dl_file;
2888 struct file_lock *fl;
2891 fl = nfs4_alloc_init_lease(dp, flag);
2894 fl->fl_file = find_readable_file(fp);
2895 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2896 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2898 list_del_init(&dp->dl_perclnt);
2899 locks_free_lock(fl);
2903 fp->fi_deleg_file = get_file(fl->fl_file);
2904 atomic_set(&fp->fi_delegees, 1);
2905 list_add(&dp->dl_perfile, &fp->fi_delegations);
2909 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2911 struct nfs4_file *fp = dp->dl_file;
2914 return nfs4_setlease(dp, flag);
2915 spin_lock(&recall_lock);
2916 if (fp->fi_had_conflict) {
2917 spin_unlock(&recall_lock);
2920 atomic_inc(&fp->fi_delegees);
2921 list_add(&dp->dl_perfile, &fp->fi_delegations);
2922 spin_unlock(&recall_lock);
2923 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2927 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2929 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2930 if (status == -EAGAIN)
2931 open->op_why_no_deleg = WND4_CONTENTION;
2933 open->op_why_no_deleg = WND4_RESOURCE;
2934 switch (open->op_deleg_want) {
2935 case NFS4_SHARE_WANT_READ_DELEG:
2936 case NFS4_SHARE_WANT_WRITE_DELEG:
2937 case NFS4_SHARE_WANT_ANY_DELEG:
2939 case NFS4_SHARE_WANT_CANCEL:
2940 open->op_why_no_deleg = WND4_CANCELLED;
2942 case NFS4_SHARE_WANT_NO_DELEG:
2949 * Attempt to hand out a delegation.
2952 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2953 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2955 struct nfs4_delegation *dp;
2956 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2958 int status = 0, flag = 0;
2960 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2961 flag = NFS4_OPEN_DELEGATE_NONE;
2962 open->op_recall = 0;
2963 switch (open->op_claim_type) {
2964 case NFS4_OPEN_CLAIM_PREVIOUS:
2966 open->op_recall = 1;
2967 flag = open->op_delegate_type;
2968 if (flag == NFS4_OPEN_DELEGATE_NONE)
2971 case NFS4_OPEN_CLAIM_NULL:
2972 /* Let's not give out any delegations till everyone's
2973 * had the chance to reclaim theirs.... */
2974 if (locks_in_grace(net))
2976 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2978 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2979 flag = NFS4_OPEN_DELEGATE_WRITE;
2981 flag = NFS4_OPEN_DELEGATE_READ;
2987 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2990 status = nfs4_set_delegation(dp, flag);
2994 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2996 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2997 STATEID_VAL(&dp->dl_stid.sc_stateid));
2999 open->op_delegate_type = flag;
3000 if (flag == NFS4_OPEN_DELEGATE_NONE) {
3001 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3002 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
3003 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3005 /* 4.1 client asking for a delegation? */
3006 if (open->op_deleg_want)
3007 nfsd4_open_deleg_none_ext(open, status);
3011 unhash_stid(&dp->dl_stid);
3012 nfs4_put_delegation(dp);
3014 flag = NFS4_OPEN_DELEGATE_NONE;
3018 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3019 struct nfs4_delegation *dp)
3021 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3022 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3023 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3024 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3025 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3026 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3027 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3028 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3030 /* Otherwise the client must be confused wanting a delegation
3031 * it already has, therefore we don't return
3032 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3037 * called with nfs4_lock_state() held.
3040 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3042 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3043 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3044 struct nfs4_file *fp = NULL;
3045 struct inode *ino = current_fh->fh_dentry->d_inode;
3046 struct nfs4_ol_stateid *stp = NULL;
3047 struct nfs4_delegation *dp = NULL;
3051 * Lookup file; if found, lookup stateid and check open request,
3052 * and check for delegations in the process of being recalled.
3053 * If not found, create the nfs4_file struct
3055 fp = find_file(ino);
3057 if ((status = nfs4_check_open(fp, open, &stp)))
3059 status = nfs4_check_deleg(cl, fp, open, &dp);
3063 status = nfserr_bad_stateid;
3064 if (nfsd4_is_deleg_cur(open))
3066 status = nfserr_jukebox;
3068 open->op_file = NULL;
3069 nfsd4_init_file(fp, ino);
3073 * OPEN the file, or upgrade an existing OPEN.
3074 * If truncate fails, the OPEN fails.
3077 /* Stateid was found, this is an OPEN upgrade */
3078 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3082 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3085 status = nfsd4_truncate(rqstp, current_fh, open);
3089 open->op_stp = NULL;
3090 init_open_stateid(stp, fp, open);
3092 update_stateid(&stp->st_stid.sc_stateid);
3093 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3095 if (nfsd4_has_session(&resp->cstate)) {
3096 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3098 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3099 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3100 open->op_why_no_deleg = WND4_NOT_WANTED;
3106 * Attempt to hand out a delegation. No error return, because the
3107 * OPEN succeeds even if we fail.
3109 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3113 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3114 STATEID_VAL(&stp->st_stid.sc_stateid));
3116 /* 4.1 client trying to upgrade/downgrade delegation? */
3117 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3118 open->op_deleg_want)
3119 nfsd4_deleg_xgrade_none_ext(open, dp);
3123 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3124 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3126 * To finish the open response, we just need to set the rflags.
3128 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3129 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3130 !nfsd4_has_session(&resp->cstate))
3131 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3136 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3138 if (open->op_openowner) {
3139 struct nfs4_openowner *oo = open->op_openowner;
3141 if (!list_empty(&oo->oo_owner.so_stateids))
3142 list_del_init(&oo->oo_close_lru);
3143 if (oo->oo_flags & NFS4_OO_NEW) {
3145 release_openowner(oo);
3146 open->op_openowner = NULL;
3148 oo->oo_flags &= ~NFS4_OO_NEW;
3152 nfsd4_free_file(open->op_file);
3154 free_generic_stateid(open->op_stp);
3157 static __be32 lookup_clientid(clientid_t *clid, bool session, struct nfsd_net *nn, struct nfs4_client **clp)
3159 struct nfs4_client *found;
3161 if (STALE_CLIENTID(clid, nn))
3162 return nfserr_stale_clientid;
3163 found = find_confirmed_client(clid, session, nn);
3166 return found ? nfs_ok : nfserr_expired;
3170 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3173 struct nfs4_client *clp;
3175 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3178 dprintk("process_renew(%08x/%08x): starting\n",
3179 clid->cl_boot, clid->cl_id);
3180 status = lookup_clientid(clid, cstate->minorversion, nn, &clp);
3183 status = nfserr_cb_path_down;
3184 if (!list_empty(&clp->cl_delegations)
3185 && clp->cl_cb_state != NFSD4_CB_UP)
3189 nfs4_unlock_state();
3194 nfsd4_end_grace(struct nfsd_net *nn)
3196 /* do nothing if grace period already ended */
3197 if (nn->grace_ended)
3200 dprintk("NFSD: end of grace period\n");
3201 nn->grace_ended = true;
3202 nfsd4_record_grace_done(nn, nn->boot_time);
3203 locks_end_grace(&nn->nfsd4_manager);
3205 * Now that every NFSv4 client has had the chance to recover and
3206 * to see the (possibly new, possibly shorter) lease time, we
3207 * can safely set the next grace time to the current lease time:
3209 nn->nfsd4_grace = nn->nfsd4_lease;
3213 nfs4_laundromat(struct nfsd_net *nn)
3215 struct nfs4_client *clp;
3216 struct nfs4_openowner *oo;
3217 struct nfs4_delegation *dp;
3218 struct list_head *pos, *next, reaplist;
3219 time_t cutoff = get_seconds() - nn->nfsd4_lease;
3220 time_t t, clientid_val = nn->nfsd4_lease;
3221 time_t u, test_val = nn->nfsd4_lease;
3225 dprintk("NFSD: laundromat service - starting\n");
3226 nfsd4_end_grace(nn);
3227 INIT_LIST_HEAD(&reaplist);
3228 spin_lock(&nn->client_lock);
3229 list_for_each_safe(pos, next, &nn->client_lru) {
3230 clp = list_entry(pos, struct nfs4_client, cl_lru);
3231 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3232 t = clp->cl_time - cutoff;
3233 if (clientid_val > t)
3237 if (atomic_read(&clp->cl_refcount)) {
3238 dprintk("NFSD: client in use (clientid %08x)\n",
3239 clp->cl_clientid.cl_id);
3242 unhash_client_locked(clp);
3243 list_add(&clp->cl_lru, &reaplist);
3245 spin_unlock(&nn->client_lock);
3246 list_for_each_safe(pos, next, &reaplist) {
3247 clp = list_entry(pos, struct nfs4_client, cl_lru);
3248 dprintk("NFSD: purging unused client (clientid %08x)\n",
3249 clp->cl_clientid.cl_id);
3252 spin_lock(&recall_lock);
3253 list_for_each_safe(pos, next, &del_recall_lru) {
3254 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3255 if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
3257 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3258 u = dp->dl_time - cutoff;
3263 list_move(&dp->dl_recall_lru, &reaplist);
3265 spin_unlock(&recall_lock);
3266 list_for_each_safe(pos, next, &reaplist) {
3267 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3268 unhash_delegation(dp);
3270 test_val = nn->nfsd4_lease;
3271 list_for_each_safe(pos, next, &nn->close_lru) {
3272 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3273 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3274 u = oo->oo_time - cutoff;
3279 release_openowner(oo);
3281 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3282 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3283 nfs4_unlock_state();
3284 return clientid_val;
3287 static struct workqueue_struct *laundry_wq;
3288 static void laundromat_main(struct work_struct *);
3291 laundromat_main(struct work_struct *laundry)
3294 struct delayed_work *dwork = container_of(laundry, struct delayed_work,
3296 struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
3299 t = nfs4_laundromat(nn);
3300 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3301 queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
3304 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3306 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3307 return nfserr_bad_stateid;
3312 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3314 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3316 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3317 STATEID_VAL(stateid));
3322 access_permit_read(struct nfs4_ol_stateid *stp)
3324 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3325 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3326 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3330 access_permit_write(struct nfs4_ol_stateid *stp)
3332 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3333 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3337 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3339 __be32 status = nfserr_openmode;
3341 /* For lock stateid's, we test the parent open, not the lock: */
3342 if (stp->st_openstp)
3343 stp = stp->st_openstp;
3344 if ((flags & WR_STATE) && !access_permit_write(stp))
3346 if ((flags & RD_STATE) && !access_permit_read(stp))
3353 static inline __be32
3354 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3356 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3358 else if (locks_in_grace(net)) {
3359 /* Answer in remaining cases depends on existence of
3360 * conflicting state; so we must wait out the grace period. */
3361 return nfserr_grace;
3362 } else if (flags & WR_STATE)
3363 return nfs4_share_conflict(current_fh,
3364 NFS4_SHARE_DENY_WRITE);
3365 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3366 return nfs4_share_conflict(current_fh,
3367 NFS4_SHARE_DENY_READ);
3371 * Allow READ/WRITE during grace period on recovered state only for files
3372 * that are not able to provide mandatory locking.
3375 grace_disallows_io(struct net *net, struct inode *inode)
3377 return locks_in_grace(net) && mandatory_lock(inode);
3380 /* Returns true iff a is later than b: */
3381 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3383 return (s32)a->si_generation - (s32)b->si_generation > 0;
3386 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3389 * When sessions are used the stateid generation number is ignored
3392 if (has_session && in->si_generation == 0)
3395 if (in->si_generation == ref->si_generation)
3398 /* If the client sends us a stateid from the future, it's buggy: */
3399 if (stateid_generation_after(in, ref))
3400 return nfserr_bad_stateid;
3402 * However, we could see a stateid from the past, even from a
3403 * non-buggy client. For example, if the client sends a lock
3404 * while some IO is outstanding, the lock may bump si_generation
3405 * while the IO is still in flight. The client could avoid that
3406 * situation by waiting for responses on all the IO requests,
3407 * but better performance may result in retrying IO that
3408 * receives an old_stateid error if requests are rarely
3409 * reordered in flight:
3411 return nfserr_old_stateid;
3414 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3416 struct nfs4_stid *s;
3417 struct nfs4_ol_stateid *ols;
3420 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3421 return nfserr_bad_stateid;
3422 /* Client debugging aid. */
3423 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3424 char addr_str[INET6_ADDRSTRLEN];
3425 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3427 pr_warn_ratelimited("NFSD: client %s testing state ID "
3428 "with incorrect client ID\n", addr_str);
3429 return nfserr_bad_stateid;
3431 s = find_stateid(cl, stateid);
3433 return nfserr_bad_stateid;
3434 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3437 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3439 ols = openlockstateid(s);
3440 if (ols->st_stateowner->so_is_open_owner
3441 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3442 return nfserr_bad_stateid;
3446 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask,
3447 struct nfs4_stid **s, bool sessions,
3448 struct nfsd_net *nn)
3450 struct nfs4_client *cl;
3452 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3453 return nfserr_bad_stateid;
3454 if (STALE_STATEID(stateid, nn))
3455 return nfserr_stale_stateid;
3456 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions, nn);
3458 return nfserr_expired;
3459 *s = find_stateid_by_type(cl, stateid, typemask);
3461 return nfserr_bad_stateid;
3467 * Checks for stateid operations
3470 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3471 stateid_t *stateid, int flags, struct file **filpp)
3473 struct nfs4_stid *s;
3474 struct nfs4_ol_stateid *stp = NULL;
3475 struct nfs4_delegation *dp = NULL;
3476 struct svc_fh *current_fh = &cstate->current_fh;
3477 struct inode *ino = current_fh->fh_dentry->d_inode;
3478 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3484 if (grace_disallows_io(net, ino))
3485 return nfserr_grace;
3487 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3488 return check_special_stateids(net, current_fh, stateid, flags);
3490 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
3491 &s, cstate->minorversion, nn);
3494 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3497 switch (s->sc_type) {
3498 case NFS4_DELEG_STID:
3499 dp = delegstateid(s);
3500 status = nfs4_check_delegmode(dp, flags);
3504 *filpp = dp->dl_file->fi_deleg_file;
3507 status = nfserr_serverfault;
3512 case NFS4_OPEN_STID:
3513 case NFS4_LOCK_STID:
3514 stp = openlockstateid(s);
3515 status = nfs4_check_fh(current_fh, stp);
3518 if (stp->st_stateowner->so_is_open_owner
3519 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3521 status = nfs4_check_openmode(stp, flags);
3525 if (flags & RD_STATE)
3526 *filpp = find_readable_file(stp->st_file);
3528 *filpp = find_writeable_file(stp->st_file);
3532 return nfserr_bad_stateid;
3540 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3542 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3543 return nfserr_locks_held;
3544 release_lock_stateid(stp);
3549 * Test if the stateid is valid
3552 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3553 struct nfsd4_test_stateid *test_stateid)
3555 struct nfsd4_test_stateid_id *stateid;
3556 struct nfs4_client *cl = cstate->session->se_client;
3559 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3560 stateid->ts_id_status =
3561 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3562 nfs4_unlock_state();
3568 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3569 struct nfsd4_free_stateid *free_stateid)
3571 stateid_t *stateid = &free_stateid->fr_stateid;
3572 struct nfs4_stid *s;
3573 struct nfs4_client *cl = cstate->session->se_client;
3574 __be32 ret = nfserr_bad_stateid;
3577 s = find_stateid(cl, stateid);
3580 switch (s->sc_type) {
3581 case NFS4_DELEG_STID:
3582 ret = nfserr_locks_held;
3584 case NFS4_OPEN_STID:
3585 case NFS4_LOCK_STID:
3586 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3589 if (s->sc_type == NFS4_LOCK_STID)
3590 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3592 ret = nfserr_locks_held;
3595 ret = nfserr_bad_stateid;
3598 nfs4_unlock_state();
3605 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3606 RD_STATE : WR_STATE;
3609 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3611 struct svc_fh *current_fh = &cstate->current_fh;
3612 struct nfs4_stateowner *sop = stp->st_stateowner;
3615 status = nfsd4_check_seqid(cstate, sop, seqid);
3618 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3620 * "Closed" stateid's exist *only* to return
3621 * nfserr_replay_me from the previous step.
3623 return nfserr_bad_stateid;
3624 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3627 return nfs4_check_fh(current_fh, stp);
3631 * Checks for sequence id mutating operations.
3634 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3635 stateid_t *stateid, char typemask,
3636 struct nfs4_ol_stateid **stpp,
3637 struct nfsd_net *nn)
3640 struct nfs4_stid *s;
3642 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3643 seqid, STATEID_VAL(stateid));
3646 status = nfsd4_lookup_stateid(stateid, typemask, &s,
3647 cstate->minorversion, nn);
3650 *stpp = openlockstateid(s);
3651 cstate->replay_owner = (*stpp)->st_stateowner;
3653 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3656 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3657 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
3660 struct nfs4_openowner *oo;
3662 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3663 NFS4_OPEN_STID, stpp, nn);
3666 oo = openowner((*stpp)->st_stateowner);
3667 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3668 return nfserr_bad_stateid;
3673 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3674 struct nfsd4_open_confirm *oc)
3677 struct nfs4_openowner *oo;
3678 struct nfs4_ol_stateid *stp;
3679 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3681 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3682 (int)cstate->current_fh.fh_dentry->d_name.len,
3683 cstate->current_fh.fh_dentry->d_name.name);
3685 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3691 status = nfs4_preprocess_seqid_op(cstate,
3692 oc->oc_seqid, &oc->oc_req_stateid,
3693 NFS4_OPEN_STID, &stp, nn);
3696 oo = openowner(stp->st_stateowner);
3697 status = nfserr_bad_stateid;
3698 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3700 oo->oo_flags |= NFS4_OO_CONFIRMED;
3701 update_stateid(&stp->st_stid.sc_stateid);
3702 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3703 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3704 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3706 nfsd4_client_record_create(oo->oo_owner.so_client);
3709 if (!cstate->replay_owner)
3710 nfs4_unlock_state();
3714 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3716 if (!test_access(access, stp))
3718 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3719 clear_access(access, stp);
3722 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3724 switch (to_access) {
3725 case NFS4_SHARE_ACCESS_READ:
3726 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3727 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3729 case NFS4_SHARE_ACCESS_WRITE:
3730 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3731 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3733 case NFS4_SHARE_ACCESS_BOTH:
3741 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3744 for (i = 0; i < 4; i++) {
3745 if ((i & deny) != i)
3751 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3752 struct nfsd4_compound_state *cstate,
3753 struct nfsd4_open_downgrade *od)
3756 struct nfs4_ol_stateid *stp;
3757 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3759 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3760 (int)cstate->current_fh.fh_dentry->d_name.len,
3761 cstate->current_fh.fh_dentry->d_name.name);
3763 /* We don't yet support WANT bits: */
3764 if (od->od_deleg_want)
3765 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3769 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3770 &od->od_stateid, &stp, nn);
3773 status = nfserr_inval;
3774 if (!test_access(od->od_share_access, stp)) {
3775 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3776 stp->st_access_bmap, od->od_share_access);
3779 if (!test_deny(od->od_share_deny, stp)) {
3780 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3781 stp->st_deny_bmap, od->od_share_deny);
3784 nfs4_stateid_downgrade(stp, od->od_share_access);
3786 reset_union_bmap_deny(od->od_share_deny, stp);
3788 update_stateid(&stp->st_stid.sc_stateid);
3789 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3792 if (!cstate->replay_owner)
3793 nfs4_unlock_state();
3797 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3799 struct nfs4_openowner *oo;
3800 struct nfs4_ol_stateid *s;
3802 if (!so->so_is_open_owner)
3805 s = oo->oo_last_closed_stid;
3808 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3809 /* Release the last_closed_stid on the next seqid bump: */
3810 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3813 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3814 release_last_closed_stateid(oo);
3817 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3819 unhash_open_stateid(s);
3820 s->st_stid.sc_type = NFS4_CLOSED_STID;
3824 * nfs4_unlock_state() called after encode
3827 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3828 struct nfsd4_close *close)
3831 struct nfs4_openowner *oo;
3832 struct nfs4_ol_stateid *stp;
3833 struct net *net = SVC_NET(rqstp);
3834 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3836 dprintk("NFSD: nfsd4_close on file %.*s\n",
3837 (int)cstate->current_fh.fh_dentry->d_name.len,
3838 cstate->current_fh.fh_dentry->d_name.name);
3841 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3843 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3847 oo = openowner(stp->st_stateowner);
3849 update_stateid(&stp->st_stid.sc_stateid);
3850 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3852 nfsd4_close_open_stateid(stp);
3853 release_last_closed_stateid(oo);
3854 oo->oo_last_closed_stid = stp;
3856 if (list_empty(&oo->oo_owner.so_stateids)) {
3857 if (cstate->minorversion) {
3858 release_openowner(oo);
3859 cstate->replay_owner = NULL;
3862 * In the 4.0 case we need to keep the owners around a
3863 * little while to handle CLOSE replay.
3865 if (list_empty(&oo->oo_owner.so_stateids))
3866 move_to_close_lru(oo, SVC_NET(rqstp));
3870 if (!cstate->replay_owner)
3871 nfs4_unlock_state();
3876 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3877 struct nfsd4_delegreturn *dr)
3879 struct nfs4_delegation *dp;
3880 stateid_t *stateid = &dr->dr_stateid;
3881 struct nfs4_stid *s;
3883 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3885 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3889 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s,
3890 cstate->minorversion, nn);
3893 dp = delegstateid(s);
3894 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3898 unhash_delegation(dp);
3900 nfs4_unlock_state();
3906 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3908 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3911 end_offset(u64 start, u64 len)
3916 return end >= start ? end: NFS4_MAX_UINT64;
3919 /* last octet in a range */
3921 last_byte_offset(u64 start, u64 len)
3927 return end > start ? end - 1: NFS4_MAX_UINT64;
3930 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3932 return (file_hashval(inode) + cl_id
3933 + opaque_hashval(ownername->data, ownername->len))
3934 & LOCKOWNER_INO_HASH_MASK;
3938 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3939 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3940 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3941 * locking, this prevents us from being completely protocol-compliant. The
3942 * real solution to this problem is to start using unsigned file offsets in
3943 * the VFS, but this is a very deep change!
3946 nfs4_transform_lock_offset(struct file_lock *lock)
3948 if (lock->fl_start < 0)
3949 lock->fl_start = OFFSET_MAX;
3950 if (lock->fl_end < 0)
3951 lock->fl_end = OFFSET_MAX;
3954 /* Hack!: For now, we're defining this just so we can use a pointer to it
3955 * as a unique cookie to identify our (NFSv4's) posix locks. */
3956 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3960 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3962 struct nfs4_lockowner *lo;
3964 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3965 lo = (struct nfs4_lockowner *) fl->fl_owner;
3966 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3967 lo->lo_owner.so_owner.len, GFP_KERNEL);
3968 if (!deny->ld_owner.data)
3969 /* We just don't care that much */
3971 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3972 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3975 deny->ld_owner.len = 0;
3976 deny->ld_owner.data = NULL;
3977 deny->ld_clientid.cl_boot = 0;
3978 deny->ld_clientid.cl_id = 0;
3980 deny->ld_start = fl->fl_start;
3981 deny->ld_length = NFS4_MAX_UINT64;
3982 if (fl->fl_end != NFS4_MAX_UINT64)
3983 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3984 deny->ld_type = NFS4_READ_LT;
3985 if (fl->fl_type != F_RDLCK)
3986 deny->ld_type = NFS4_WRITE_LT;
3989 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3991 struct nfs4_ol_stateid *lst;
3993 if (!same_owner_str(&lo->lo_owner, owner, clid))
3995 lst = list_first_entry(&lo->lo_owner.so_stateids,
3996 struct nfs4_ol_stateid, st_perstateowner);
3997 return lst->st_file->fi_inode == inode;
4000 static struct nfs4_lockowner *
4001 find_lockowner_str(struct inode *inode, clientid_t *clid,
4002 struct xdr_netobj *owner, struct nfsd_net *nn)
4004 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
4005 struct nfs4_lockowner *lo;
4007 list_for_each_entry(lo, &nn->lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
4008 if (same_lockowner_ino(lo, inode, clid, owner))
4014 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
4016 struct inode *inode = open_stp->st_file->fi_inode;
4017 unsigned int inohash = lockowner_ino_hashval(inode,
4018 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
4019 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
4021 list_add(&lo->lo_owner.so_strhash, &nn->ownerstr_hashtbl[strhashval]);
4022 list_add(&lo->lo_owner_ino_hash, &nn->lockowner_ino_hashtbl[inohash]);
4023 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4027 * Alloc a lock owner structure.
4028 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4031 * strhashval = ownerstr_hashval
4034 static struct nfs4_lockowner *
4035 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4036 struct nfs4_lockowner *lo;
4038 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4041 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4042 lo->lo_owner.so_is_open_owner = 0;
4043 /* It is the openowner seqid that will be incremented in encode in the
4044 * case of new lockowners; so increment the lock seqid manually: */
4045 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4046 hash_lockowner(lo, strhashval, clp, open_stp);
4050 static struct nfs4_ol_stateid *
4051 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4053 struct nfs4_ol_stateid *stp;
4054 struct nfs4_client *clp = lo->lo_owner.so_client;
4056 stp = nfs4_alloc_stateid(clp);
4059 stp->st_stid.sc_type = NFS4_LOCK_STID;
4060 list_add(&stp->st_perfile, &fp->fi_stateids);
4061 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4062 stp->st_stateowner = &lo->lo_owner;
4065 stp->st_access_bmap = 0;
4066 stp->st_deny_bmap = open_stp->st_deny_bmap;
4067 stp->st_openstp = open_stp;
4072 check_lock_length(u64 offset, u64 length)
4074 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4075 LOFF_OVERFLOW(offset, length)));
4078 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4080 struct nfs4_file *fp = lock_stp->st_file;
4081 int oflag = nfs4_access_to_omode(access);
4083 if (test_access(access, lock_stp))
4085 nfs4_file_get_access(fp, oflag);
4086 set_access(access, lock_stp);
4089 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)
4091 struct nfs4_file *fi = ost->st_file;
4092 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4093 struct nfs4_client *cl = oo->oo_owner.so_client;
4094 struct nfs4_lockowner *lo;
4095 unsigned int strhashval;
4096 struct nfsd_net *nn = net_generic(cl->net, nfsd_net_id);
4098 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid,
4099 &lock->v.new.owner, nn);
4101 if (!cstate->minorversion)
4102 return nfserr_bad_seqid;
4103 /* XXX: a lockowner always has exactly one stateid: */
4104 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4105 struct nfs4_ol_stateid, st_perstateowner);
4108 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4109 &lock->v.new.owner);
4110 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4112 return nfserr_jukebox;
4113 *lst = alloc_init_lock_stateid(lo, fi, ost);
4115 release_lockowner(lo);
4116 return nfserr_jukebox;
4126 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4127 struct nfsd4_lock *lock)
4129 struct nfs4_openowner *open_sop = NULL;
4130 struct nfs4_lockowner *lock_sop = NULL;
4131 struct nfs4_ol_stateid *lock_stp;
4132 struct file *filp = NULL;
4133 struct file_lock *file_lock = NULL;
4134 struct file_lock *conflock = NULL;
4136 bool new_state = false;
4139 struct net *net = SVC_NET(rqstp);
4140 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4142 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4143 (long long) lock->lk_offset,
4144 (long long) lock->lk_length);
4146 if (check_lock_length(lock->lk_offset, lock->lk_length))
4147 return nfserr_inval;
4149 if ((status = fh_verify(rqstp, &cstate->current_fh,
4150 S_IFREG, NFSD_MAY_LOCK))) {
4151 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4157 if (lock->lk_is_new) {
4158 struct nfs4_ol_stateid *open_stp = NULL;
4160 if (nfsd4_has_session(cstate))
4161 /* See rfc 5661 18.10.3: given clientid is ignored: */
4162 memcpy(&lock->v.new.clientid,
4163 &cstate->session->se_client->cl_clientid,
4164 sizeof(clientid_t));
4166 status = nfserr_stale_clientid;
4167 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4170 /* validate and update open stateid and open seqid */
4171 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4172 lock->lk_new_open_seqid,
4173 &lock->lk_new_open_stateid,
4177 open_sop = openowner(open_stp->st_stateowner);
4178 status = nfserr_bad_stateid;
4179 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4180 &lock->v.new.clientid))
4182 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4183 &lock_stp, &new_state);
4185 status = nfs4_preprocess_seqid_op(cstate,
4186 lock->lk_old_lock_seqid,
4187 &lock->lk_old_lock_stateid,
4188 NFS4_LOCK_STID, &lock_stp, nn);
4191 lock_sop = lockowner(lock_stp->st_stateowner);
4193 lkflg = setlkflg(lock->lk_type);
4194 status = nfs4_check_openmode(lock_stp, lkflg);
4198 status = nfserr_grace;
4199 if (locks_in_grace(net) && !lock->lk_reclaim)
4201 status = nfserr_no_grace;
4202 if (!locks_in_grace(net) && lock->lk_reclaim)
4205 file_lock = locks_alloc_lock();
4207 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4208 status = nfserr_jukebox;
4212 locks_init_lock(file_lock);
4213 switch (lock->lk_type) {
4216 filp = find_readable_file(lock_stp->st_file);
4218 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4219 file_lock->fl_type = F_RDLCK;
4222 case NFS4_WRITEW_LT:
4223 filp = find_writeable_file(lock_stp->st_file);
4225 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4226 file_lock->fl_type = F_WRLCK;
4229 status = nfserr_inval;
4233 status = nfserr_openmode;
4236 file_lock->fl_owner = (fl_owner_t)lock_sop;
4237 file_lock->fl_pid = current->tgid;
4238 file_lock->fl_file = filp;
4239 file_lock->fl_flags = FL_POSIX;
4240 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4241 file_lock->fl_start = lock->lk_offset;
4242 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4243 nfs4_transform_lock_offset(file_lock);
4245 conflock = locks_alloc_lock();
4247 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4248 status = nfserr_jukebox;
4252 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4254 case 0: /* success! */
4255 update_stateid(&lock_stp->st_stid.sc_stateid);
4256 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4260 case (EAGAIN): /* conflock holds conflicting lock */
4261 status = nfserr_denied;
4262 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4263 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4266 status = nfserr_deadlock;
4269 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4270 status = nfserrno(err);
4274 if (status && new_state)
4275 release_lockowner(lock_sop);
4276 if (!cstate->replay_owner)
4277 nfs4_unlock_state();
4279 locks_free_lock(file_lock);
4281 locks_free_lock(conflock);
4286 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4287 * so we do a temporary open here just to get an open file to pass to
4288 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4291 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4294 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4296 err = nfserrno(vfs_test_lock(file, lock));
4306 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4307 struct nfsd4_lockt *lockt)
4309 struct inode *inode;
4310 struct file_lock *file_lock = NULL;
4311 struct nfs4_lockowner *lo;
4313 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4315 if (locks_in_grace(SVC_NET(rqstp)))
4316 return nfserr_grace;
4318 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4319 return nfserr_inval;
4323 if (!nfsd4_has_session(cstate)) {
4324 status = lookup_clientid(&lockt->lt_clientid, false, nn, NULL);
4329 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4332 inode = cstate->current_fh.fh_dentry->d_inode;
4333 file_lock = locks_alloc_lock();
4335 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4336 status = nfserr_jukebox;
4339 locks_init_lock(file_lock);
4340 switch (lockt->lt_type) {
4343 file_lock->fl_type = F_RDLCK;
4346 case NFS4_WRITEW_LT:
4347 file_lock->fl_type = F_WRLCK;
4350 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4351 status = nfserr_inval;
4355 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner, nn);
4357 file_lock->fl_owner = (fl_owner_t)lo;
4358 file_lock->fl_pid = current->tgid;
4359 file_lock->fl_flags = FL_POSIX;
4361 file_lock->fl_start = lockt->lt_offset;
4362 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4364 nfs4_transform_lock_offset(file_lock);
4366 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4370 if (file_lock->fl_type != F_UNLCK) {
4371 status = nfserr_denied;
4372 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4375 nfs4_unlock_state();
4377 locks_free_lock(file_lock);
4382 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4383 struct nfsd4_locku *locku)
4385 struct nfs4_ol_stateid *stp;
4386 struct file *filp = NULL;
4387 struct file_lock *file_lock = NULL;
4390 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4392 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4393 (long long) locku->lu_offset,
4394 (long long) locku->lu_length);
4396 if (check_lock_length(locku->lu_offset, locku->lu_length))
4397 return nfserr_inval;
4401 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4402 &locku->lu_stateid, NFS4_LOCK_STID,
4406 filp = find_any_file(stp->st_file);
4408 status = nfserr_lock_range;
4411 file_lock = locks_alloc_lock();
4413 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4414 status = nfserr_jukebox;
4417 locks_init_lock(file_lock);
4418 file_lock->fl_type = F_UNLCK;
4419 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4420 file_lock->fl_pid = current->tgid;
4421 file_lock->fl_file = filp;
4422 file_lock->fl_flags = FL_POSIX;
4423 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4424 file_lock->fl_start = locku->lu_offset;
4426 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4428 nfs4_transform_lock_offset(file_lock);
4431 * Try to unlock the file in the VFS.
4433 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4435 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4439 * OK, unlock succeeded; the only thing left to do is update the stateid.
4441 update_stateid(&stp->st_stid.sc_stateid);
4442 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4445 if (!cstate->replay_owner)
4446 nfs4_unlock_state();
4448 locks_free_lock(file_lock);
4452 status = nfserrno(err);
4458 * 1: locks held by lockowner
4459 * 0: no locks held by lockowner
4462 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4464 struct file_lock **flpp;
4465 struct inode *inode = filp->fi_inode;
4469 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4470 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4481 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4482 struct nfsd4_compound_state *cstate,
4483 struct nfsd4_release_lockowner *rlockowner)
4485 clientid_t *clid = &rlockowner->rl_clientid;
4486 struct nfs4_stateowner *sop;
4487 struct nfs4_lockowner *lo;
4488 struct nfs4_ol_stateid *stp;
4489 struct xdr_netobj *owner = &rlockowner->rl_owner;
4490 struct list_head matches;
4491 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4493 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4495 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4496 clid->cl_boot, clid->cl_id);
4500 status = lookup_clientid(clid, cstate->minorversion, nn, NULL);
4504 status = nfserr_locks_held;
4505 INIT_LIST_HEAD(&matches);
4507 list_for_each_entry(sop, &nn->ownerstr_hashtbl[hashval], so_strhash) {
4508 if (sop->so_is_open_owner)
4510 if (!same_owner_str(sop, owner, clid))
4512 list_for_each_entry(stp, &sop->so_stateids,
4514 lo = lockowner(sop);
4515 if (check_for_locks(stp->st_file, lo))
4517 list_add(&lo->lo_list, &matches);
4520 /* Clients probably won't expect us to return with some (but not all)
4521 * of the lockowner state released; so don't release any until all
4522 * have been checked. */
4524 while (!list_empty(&matches)) {
4525 lo = list_entry(matches.next, struct nfs4_lockowner,
4527 /* unhash_stateowner deletes so_perclient only
4528 * for openowners. */
4529 list_del(&lo->lo_list);
4530 release_lockowner(lo);
4533 nfs4_unlock_state();
4537 static inline struct nfs4_client_reclaim *
4540 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4544 nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
4546 struct nfs4_client_reclaim *crp;
4548 crp = nfsd4_find_reclaim_client(name, nn);
4549 return (crp && crp->cr_clp);
4553 * failure => all reset bets are off, nfserr_no_grace...
4555 struct nfs4_client_reclaim *
4556 nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
4558 unsigned int strhashval;
4559 struct nfs4_client_reclaim *crp;
4561 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4562 crp = alloc_reclaim();
4564 strhashval = clientstr_hashval(name);
4565 INIT_LIST_HEAD(&crp->cr_strhash);
4566 list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
4567 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4569 nn->reclaim_str_hashtbl_size++;
4575 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
4577 list_del(&crp->cr_strhash);
4579 nn->reclaim_str_hashtbl_size--;
4583 nfs4_release_reclaim(struct nfsd_net *nn)
4585 struct nfs4_client_reclaim *crp = NULL;
4588 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4589 while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
4590 crp = list_entry(nn->reclaim_str_hashtbl[i].next,
4591 struct nfs4_client_reclaim, cr_strhash);
4592 nfs4_remove_reclaim_record(crp, nn);
4595 WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
4599 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4600 struct nfs4_client_reclaim *
4601 nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
4603 unsigned int strhashval;
4604 struct nfs4_client_reclaim *crp = NULL;
4606 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4608 strhashval = clientstr_hashval(recdir);
4609 list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
4610 if (same_name(crp->cr_recdir, recdir)) {
4618 * Called from OPEN. Look for clientid in reclaim list.
4621 nfs4_check_open_reclaim(clientid_t *clid, bool sessions, struct nfsd_net *nn)
4623 struct nfs4_client *clp;
4625 /* find clientid in conf_id_hashtbl */
4626 clp = find_confirmed_client(clid, sessions, nn);
4628 return nfserr_reclaim_bad;
4630 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4633 #ifdef CONFIG_NFSD_FAULT_INJECTION
4635 u64 nfsd_forget_client(struct nfs4_client *clp, u64 max)
4641 u64 nfsd_print_client(struct nfs4_client *clp, u64 num)
4643 char buf[INET6_ADDRSTRLEN];
4644 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4645 printk(KERN_INFO "NFS Client: %s\n", buf);
4649 static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
4652 char buf[INET6_ADDRSTRLEN];
4653 rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
4654 printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
4657 static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_lockowner *))
4659 struct nfs4_openowner *oop;
4660 struct nfs4_lockowner *lop, *lo_next;
4661 struct nfs4_ol_stateid *stp, *st_next;
4664 list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
4665 list_for_each_entry_safe(stp, st_next, &oop->oo_owner.so_stateids, st_perstateowner) {
4666 list_for_each_entry_safe(lop, lo_next, &stp->st_lockowners, lo_perstateid) {
4678 u64 nfsd_forget_client_locks(struct nfs4_client *clp, u64 max)
4680 return nfsd_foreach_client_lock(clp, max, release_lockowner);
4683 u64 nfsd_print_client_locks(struct nfs4_client *clp, u64 max)
4685 u64 count = nfsd_foreach_client_lock(clp, max, NULL);
4686 nfsd_print_count(clp, count, "locked files");
4690 static u64 nfsd_foreach_client_open(struct nfs4_client *clp, u64 max, void (*func)(struct nfs4_openowner *))
4692 struct nfs4_openowner *oop, *next;
4695 list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
4705 u64 nfsd_forget_client_openowners(struct nfs4_client *clp, u64 max)
4707 return nfsd_foreach_client_open(clp, max, release_openowner);
4710 u64 nfsd_print_client_openowners(struct nfs4_client *clp, u64 max)
4712 u64 count = nfsd_foreach_client_open(clp, max, NULL);
4713 nfsd_print_count(clp, count, "open files");
4717 static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
4718 struct list_head *victims)
4720 struct nfs4_delegation *dp, *next;
4723 list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
4725 list_move(&dp->dl_recall_lru, victims);
4732 u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max)
4734 struct nfs4_delegation *dp, *next;
4738 spin_lock(&recall_lock);
4739 count = nfsd_find_all_delegations(clp, max, &victims);
4740 spin_unlock(&recall_lock);
4742 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4743 unhash_delegation(dp);
4748 u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max)
4750 struct nfs4_delegation *dp, *next;
4754 spin_lock(&recall_lock);
4755 count = nfsd_find_all_delegations(clp, max, &victims);
4756 list_for_each_entry_safe(dp, next, &victims, dl_recall_lru)
4757 nfsd_break_one_deleg(dp);
4758 spin_unlock(&recall_lock);
4763 u64 nfsd_print_client_delegations(struct nfs4_client *clp, u64 max)
4767 spin_lock(&recall_lock);
4768 count = nfsd_find_all_delegations(clp, max, NULL);
4769 spin_unlock(&recall_lock);
4771 nfsd_print_count(clp, count, "delegations");
4775 u64 nfsd_for_n_state(u64 max, u64 (*func)(struct nfs4_client *, u64))
4777 struct nfs4_client *clp, *next;
4779 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4781 if (!nfsd_netns_ready(nn))
4784 list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
4785 count += func(clp, max - count);
4786 if ((max != 0) && (count >= max))
4793 struct nfs4_client *nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
4795 struct nfs4_client *clp;
4796 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns, nfsd_net_id);
4798 if (!nfsd_netns_ready(nn))
4801 list_for_each_entry(clp, &nn->client_lru, cl_lru) {
4802 if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
4808 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4810 /* initialization to perform at module load time: */
4813 nfs4_state_init(void)
4817 for (i = 0; i < FILE_HASH_SIZE; i++) {
4818 INIT_LIST_HEAD(&file_hashtbl[i]);
4820 INIT_LIST_HEAD(&del_recall_lru);
4824 * Since the lifetime of a delegation isn't limited to that of an open, a
4825 * client may quite reasonably hang on to a delegation as long as it has
4826 * the inode cached. This becomes an obvious problem the first time a
4827 * client's inode cache approaches the size of the server's total memory.
4829 * For now we avoid this problem by imposing a hard limit on the number
4830 * of delegations, which varies according to the server's memory size.
4833 set_max_delegations(void)
4836 * Allow at most 4 delegations per megabyte of RAM. Quick
4837 * estimates suggest that in the worst case (where every delegation
4838 * is for a different inode), a delegation could take about 1.5K,
4839 * giving a worst case usage of about 6% of memory.
4841 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4844 static int nfs4_state_create_net(struct net *net)
4846 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4849 nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4850 CLIENT_HASH_SIZE, GFP_KERNEL);
4851 if (!nn->conf_id_hashtbl)
4853 nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
4854 CLIENT_HASH_SIZE, GFP_KERNEL);
4855 if (!nn->unconf_id_hashtbl)
4857 nn->ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
4858 OWNER_HASH_SIZE, GFP_KERNEL);
4859 if (!nn->ownerstr_hashtbl)
4861 nn->lockowner_ino_hashtbl = kmalloc(sizeof(struct list_head) *
4862 LOCKOWNER_INO_HASH_SIZE, GFP_KERNEL);
4863 if (!nn->lockowner_ino_hashtbl)
4864 goto err_lockowner_ino;
4865 nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
4866 SESSION_HASH_SIZE, GFP_KERNEL);
4867 if (!nn->sessionid_hashtbl)
4870 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4871 INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
4872 INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
4874 for (i = 0; i < OWNER_HASH_SIZE; i++)
4875 INIT_LIST_HEAD(&nn->ownerstr_hashtbl[i]);
4876 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4877 INIT_LIST_HEAD(&nn->lockowner_ino_hashtbl[i]);
4878 for (i = 0; i < SESSION_HASH_SIZE; i++)
4879 INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
4880 nn->conf_name_tree = RB_ROOT;
4881 nn->unconf_name_tree = RB_ROOT;
4882 INIT_LIST_HEAD(&nn->client_lru);
4883 INIT_LIST_HEAD(&nn->close_lru);
4884 spin_lock_init(&nn->client_lock);
4886 INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
4892 kfree(nn->lockowner_ino_hashtbl);
4894 kfree(nn->ownerstr_hashtbl);
4896 kfree(nn->unconf_id_hashtbl);
4898 kfree(nn->conf_id_hashtbl);
4904 nfs4_state_destroy_net(struct net *net)
4907 struct nfs4_client *clp = NULL;
4908 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4909 struct rb_node *node, *tmp;
4911 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4912 while (!list_empty(&nn->conf_id_hashtbl[i])) {
4913 clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4914 destroy_client(clp);
4918 node = rb_first(&nn->unconf_name_tree);
4919 while (node != NULL) {
4921 node = rb_next(tmp);
4922 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4923 rb_erase(tmp, &nn->unconf_name_tree);
4924 destroy_client(clp);
4927 kfree(nn->sessionid_hashtbl);
4928 kfree(nn->lockowner_ino_hashtbl);
4929 kfree(nn->ownerstr_hashtbl);
4930 kfree(nn->unconf_id_hashtbl);
4931 kfree(nn->conf_id_hashtbl);
4936 nfs4_state_start_net(struct net *net)
4938 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4941 ret = nfs4_state_create_net(net);
4944 nfsd4_client_tracking_init(net);
4945 nn->boot_time = get_seconds();
4946 locks_start_grace(net, &nn->nfsd4_manager);
4947 nn->grace_ended = false;
4948 printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
4949 nn->nfsd4_grace, net);
4950 queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
4954 /* initialization to perform when the nfsd service is started: */
4957 nfs4_state_start(void)
4961 ret = set_callback_cred();
4964 laundry_wq = create_singlethread_workqueue("nfsd4");
4965 if (laundry_wq == NULL) {
4969 ret = nfsd4_create_callback_queue();
4971 goto out_free_laundry;
4973 set_max_delegations();
4978 destroy_workqueue(laundry_wq);
4983 /* should be called with the state lock held */
4985 nfs4_state_shutdown_net(struct net *net)
4987 struct nfs4_delegation *dp = NULL;
4988 struct list_head *pos, *next, reaplist;
4989 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4991 cancel_delayed_work_sync(&nn->laundromat_work);
4992 locks_end_grace(&nn->nfsd4_manager);
4994 INIT_LIST_HEAD(&reaplist);
4995 spin_lock(&recall_lock);
4996 list_for_each_safe(pos, next, &del_recall_lru) {
4997 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4998 if (dp->dl_stid.sc_client->net != net)
5000 list_move(&dp->dl_recall_lru, &reaplist);
5002 spin_unlock(&recall_lock);
5003 list_for_each_safe(pos, next, &reaplist) {
5004 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5005 unhash_delegation(dp);
5008 nfsd4_client_tracking_exit(net);
5009 nfs4_state_destroy_net(net);
5013 nfs4_state_shutdown(void)
5015 destroy_workqueue(laundry_wq);
5016 nfsd4_destroy_callback_queue();
5020 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5022 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
5023 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
5027 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
5029 if (cstate->minorversion) {
5030 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
5031 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5036 clear_current_stateid(struct nfsd4_compound_state *cstate)
5038 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
5042 * functions to set current state id
5045 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5047 put_stateid(cstate, &odp->od_stateid);
5051 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
5053 put_stateid(cstate, &open->op_stateid);
5057 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5059 put_stateid(cstate, &close->cl_stateid);
5063 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
5065 put_stateid(cstate, &lock->lk_resp_stateid);
5069 * functions to consume current state id
5073 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
5075 get_stateid(cstate, &odp->od_stateid);
5079 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
5081 get_stateid(cstate, &drp->dr_stateid);
5085 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
5087 get_stateid(cstate, &fsp->fr_stateid);
5091 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
5093 get_stateid(cstate, &setattr->sa_stateid);
5097 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
5099 get_stateid(cstate, &close->cl_stateid);
5103 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
5105 get_stateid(cstate, &locku->lu_stateid);
5109 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
5111 get_stateid(cstate, &read->rd_stateid);
5115 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
5117 get_stateid(cstate, &write->wr_stateid);