2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <kandros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.h>
39 #include <linux/swap.h>
40 #include <linux/pagemap.h>
41 #include <linux/ratelimit.h>
42 #include <linux/sunrpc/svcauth_gss.h>
43 #include <linux/sunrpc/clnt.h>
46 #include "current_stateid.h"
47 #include "fault_inject.h"
51 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 time_t nfsd4_lease = 90; /* default lease time */
55 time_t nfsd4_grace = 90;
57 #define all_ones {{~0,~0},~0}
58 static const stateid_t one_stateid = {
60 .si_opaque = all_ones,
62 static const stateid_t zero_stateid = {
65 static const stateid_t currentstateid = {
69 static u64 current_sessionid = 1;
71 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
72 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
73 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
75 /* forward declarations */
76 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
80 /* Currently used for almost all code touching nfsv4 state: */
81 static DEFINE_MUTEX(client_mutex);
84 * Currently used for the del_recall_lru and file hash table. In an
85 * effort to decrease the scope of the client_mutex, this spinlock may
86 * eventually cover more:
88 static DEFINE_SPINLOCK(recall_lock);
90 static struct kmem_cache *openowner_slab = NULL;
91 static struct kmem_cache *lockowner_slab = NULL;
92 static struct kmem_cache *file_slab = NULL;
93 static struct kmem_cache *stateid_slab = NULL;
94 static struct kmem_cache *deleg_slab = NULL;
99 mutex_lock(&client_mutex);
102 static void free_session(struct kref *);
104 /* Must be called under the client_lock */
105 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
107 kref_put(&ses->se_ref, free_session);
110 static void nfsd4_get_session(struct nfsd4_session *ses)
112 kref_get(&ses->se_ref);
116 nfs4_unlock_state(void)
118 mutex_unlock(&client_mutex);
122 opaque_hashval(const void *ptr, int nbytes)
124 unsigned char *cptr = (unsigned char *) ptr;
134 static struct list_head del_recall_lru;
136 static void nfsd4_free_file(struct nfs4_file *f)
138 kmem_cache_free(file_slab, f);
142 put_nfs4_file(struct nfs4_file *fi)
144 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
145 list_del(&fi->fi_hash);
146 spin_unlock(&recall_lock);
153 get_nfs4_file(struct nfs4_file *fi)
155 atomic_inc(&fi->fi_ref);
158 static int num_delegations;
159 unsigned int max_delegations;
162 * Open owner state (share locks)
165 /* hash tables for lock and open owners */
166 #define OWNER_HASH_BITS 8
167 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
168 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
170 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
174 ret = opaque_hashval(ownername->data, ownername->len);
176 return ret & OWNER_HASH_MASK;
179 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
181 /* hash table for nfs4_file */
182 #define FILE_HASH_BITS 8
183 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
185 static unsigned int file_hashval(struct inode *ino)
187 /* XXX: why are we hashing on inode pointer, anyway? */
188 return hash_ptr(ino, FILE_HASH_BITS);
191 static struct list_head file_hashtbl[FILE_HASH_SIZE];
193 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
195 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
196 atomic_inc(&fp->fi_access[oflag]);
199 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
201 if (oflag == O_RDWR) {
202 __nfs4_file_get_access(fp, O_RDONLY);
203 __nfs4_file_get_access(fp, O_WRONLY);
205 __nfs4_file_get_access(fp, oflag);
208 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
210 if (fp->fi_fds[oflag]) {
211 fput(fp->fi_fds[oflag]);
212 fp->fi_fds[oflag] = NULL;
216 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
218 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
219 nfs4_file_put_fd(fp, oflag);
221 * It's also safe to get rid of the RDWR open *if*
222 * we no longer have need of the other kind of access
223 * or if we already have the other kind of open:
225 if (fp->fi_fds[1-oflag]
226 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
227 nfs4_file_put_fd(fp, O_RDWR);
231 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
233 if (oflag == O_RDWR) {
234 __nfs4_file_put_access(fp, O_RDONLY);
235 __nfs4_file_put_access(fp, O_WRONLY);
237 __nfs4_file_put_access(fp, oflag);
240 static inline int get_new_stid(struct nfs4_stid *stid)
242 static int min_stateid = 0;
243 struct idr *stateids = &stid->sc_client->cl_stateids;
247 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
249 * Note: the necessary preallocation was done in
250 * nfs4_alloc_stateid(). The idr code caps the number of
251 * preallocations that can exist at a time, but the state lock
252 * prevents anyone from using ours before we get here:
256 * It shouldn't be a problem to reuse an opaque stateid value.
257 * I don't think it is for 4.1. But with 4.0 I worry that, for
258 * example, a stray write retransmission could be accepted by
259 * the server when it should have been rejected. Therefore,
260 * adopt a trick from the sctp code to attempt to maximize the
261 * amount of time until an id is reused, by ensuring they always
262 * "increase" (mod INT_MAX):
265 min_stateid = new_stid+1;
266 if (min_stateid == INT_MAX)
271 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
273 stateid_t *s = &stid->sc_stateid;
276 stid->sc_type = type;
277 stid->sc_client = cl;
278 s->si_opaque.so_clid = cl->cl_clientid;
279 new_id = get_new_stid(stid);
280 s->si_opaque.so_id = (u32)new_id;
281 /* Will be incremented before return to client: */
282 s->si_generation = 0;
285 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
287 struct idr *stateids = &cl->cl_stateids;
289 if (!idr_pre_get(stateids, GFP_KERNEL))
292 * Note: if we fail here (or any time between now and the time
293 * we actually get the new idr), we won't need to undo the idr
294 * preallocation, since the idr code caps the number of
295 * preallocated entries.
297 return kmem_cache_alloc(slab, GFP_KERNEL);
300 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
302 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
305 static struct nfs4_delegation *
306 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
308 struct nfs4_delegation *dp;
309 struct nfs4_file *fp = stp->st_file;
311 dprintk("NFSD alloc_init_deleg\n");
313 * Major work on the lease subsystem (for example, to support
314 * calbacks on stat) will be required before we can support
315 * write delegations properly.
317 if (type != NFS4_OPEN_DELEGATE_READ)
319 if (fp->fi_had_conflict)
321 if (num_delegations > max_delegations)
323 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
326 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
328 * delegation seqid's are never incremented. The 4.1 special
329 * meaning of seqid 0 isn't meaningful, really, but let's avoid
330 * 0 anyway just for consistency and use 1:
332 dp->dl_stid.sc_stateid.si_generation = 1;
334 INIT_LIST_HEAD(&dp->dl_perfile);
335 INIT_LIST_HEAD(&dp->dl_perclnt);
336 INIT_LIST_HEAD(&dp->dl_recall_lru);
340 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
342 atomic_set(&dp->dl_count, 1);
343 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
348 nfs4_put_delegation(struct nfs4_delegation *dp)
350 if (atomic_dec_and_test(&dp->dl_count)) {
351 dprintk("NFSD: freeing dp %p\n",dp);
352 put_nfs4_file(dp->dl_file);
353 kmem_cache_free(deleg_slab, dp);
358 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
360 if (atomic_dec_and_test(&fp->fi_delegees)) {
361 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
363 fput(fp->fi_deleg_file);
364 fp->fi_deleg_file = NULL;
368 static void unhash_stid(struct nfs4_stid *s)
370 struct idr *stateids = &s->sc_client->cl_stateids;
372 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
375 /* Called under the state lock. */
377 unhash_delegation(struct nfs4_delegation *dp)
379 unhash_stid(&dp->dl_stid);
380 list_del_init(&dp->dl_perclnt);
381 spin_lock(&recall_lock);
382 list_del_init(&dp->dl_perfile);
383 list_del_init(&dp->dl_recall_lru);
384 spin_unlock(&recall_lock);
385 nfs4_put_deleg_lease(dp->dl_file);
386 nfs4_put_delegation(dp);
393 /* client_lock protects the client lru list and session hash table */
394 static DEFINE_SPINLOCK(client_lock);
396 /* Hash tables for nfs4_clientid state */
397 #define CLIENT_HASH_BITS 4
398 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
399 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
401 static unsigned int clientid_hashval(u32 id)
403 return id & CLIENT_HASH_MASK;
406 static unsigned int clientstr_hashval(const char *name)
408 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
412 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
413 * used in reboot/reset lease grace period processing
415 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
416 * setclientid_confirmed info.
418 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
421 * client_lru holds client queue ordered by nfs4_client.cl_time
424 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
425 * for last close replay.
427 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
428 static int reclaim_str_hashtbl_size = 0;
429 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
430 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
431 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
432 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
433 static struct list_head client_lru;
434 static struct list_head close_lru;
437 * We store the NONE, READ, WRITE, and BOTH bits separately in the
438 * st_{access,deny}_bmap field of the stateid, in order to track not
439 * only what share bits are currently in force, but also what
440 * combinations of share bits previous opens have used. This allows us
441 * to enforce the recommendation of rfc 3530 14.2.19 that the server
442 * return an error if the client attempt to downgrade to a combination
443 * of share bits not explicable by closing some of its previous opens.
445 * XXX: This enforcement is actually incomplete, since we don't keep
446 * track of access/deny bit combinations; so, e.g., we allow:
448 * OPEN allow read, deny write
449 * OPEN allow both, deny none
450 * DOWNGRADE allow read, deny none
452 * which we should reject.
455 bmap_to_share_mode(unsigned long bmap) {
457 unsigned int access = 0;
459 for (i = 1; i < 4; i++) {
460 if (test_bit(i, &bmap))
467 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
468 unsigned int access, deny;
470 access = bmap_to_share_mode(stp->st_access_bmap);
471 deny = bmap_to_share_mode(stp->st_deny_bmap);
472 if ((access & open->op_share_deny) || (deny & open->op_share_access))
477 /* set share access for a given stateid */
479 set_access(u32 access, struct nfs4_ol_stateid *stp)
481 __set_bit(access, &stp->st_access_bmap);
484 /* clear share access for a given stateid */
486 clear_access(u32 access, struct nfs4_ol_stateid *stp)
488 __clear_bit(access, &stp->st_access_bmap);
491 /* test whether a given stateid has access */
493 test_access(u32 access, struct nfs4_ol_stateid *stp)
495 return test_bit(access, &stp->st_access_bmap);
498 /* set share deny for a given stateid */
500 set_deny(u32 access, struct nfs4_ol_stateid *stp)
502 __set_bit(access, &stp->st_deny_bmap);
505 /* clear share deny for a given stateid */
507 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
509 __clear_bit(access, &stp->st_deny_bmap);
512 /* test whether a given stateid is denying specific access */
514 test_deny(u32 access, struct nfs4_ol_stateid *stp)
516 return test_bit(access, &stp->st_deny_bmap);
519 static int nfs4_access_to_omode(u32 access)
521 switch (access & NFS4_SHARE_ACCESS_BOTH) {
522 case NFS4_SHARE_ACCESS_READ:
524 case NFS4_SHARE_ACCESS_WRITE:
526 case NFS4_SHARE_ACCESS_BOTH:
532 /* release all access and file references for a given stateid */
534 release_all_access(struct nfs4_ol_stateid *stp)
538 for (i = 1; i < 4; i++) {
539 if (test_access(i, stp))
540 nfs4_file_put_access(stp->st_file,
541 nfs4_access_to_omode(i));
542 clear_access(i, stp);
546 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
548 list_del(&stp->st_perfile);
549 list_del(&stp->st_perstateowner);
552 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
554 release_all_access(stp);
555 put_nfs4_file(stp->st_file);
559 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
561 kmem_cache_free(stateid_slab, stp);
564 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
568 unhash_generic_stateid(stp);
569 unhash_stid(&stp->st_stid);
570 file = find_any_file(stp->st_file);
572 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
573 close_generic_stateid(stp);
574 free_generic_stateid(stp);
577 static void unhash_lockowner(struct nfs4_lockowner *lo)
579 struct nfs4_ol_stateid *stp;
581 list_del(&lo->lo_owner.so_strhash);
582 list_del(&lo->lo_perstateid);
583 list_del(&lo->lo_owner_ino_hash);
584 while (!list_empty(&lo->lo_owner.so_stateids)) {
585 stp = list_first_entry(&lo->lo_owner.so_stateids,
586 struct nfs4_ol_stateid, st_perstateowner);
587 release_lock_stateid(stp);
591 static void release_lockowner(struct nfs4_lockowner *lo)
593 unhash_lockowner(lo);
594 nfs4_free_lockowner(lo);
598 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
600 struct nfs4_lockowner *lo;
602 while (!list_empty(&open_stp->st_lockowners)) {
603 lo = list_entry(open_stp->st_lockowners.next,
604 struct nfs4_lockowner, lo_perstateid);
605 release_lockowner(lo);
609 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
611 unhash_generic_stateid(stp);
612 release_stateid_lockowners(stp);
613 close_generic_stateid(stp);
616 static void release_open_stateid(struct nfs4_ol_stateid *stp)
618 unhash_open_stateid(stp);
619 unhash_stid(&stp->st_stid);
620 free_generic_stateid(stp);
623 static void unhash_openowner(struct nfs4_openowner *oo)
625 struct nfs4_ol_stateid *stp;
627 list_del(&oo->oo_owner.so_strhash);
628 list_del(&oo->oo_perclient);
629 while (!list_empty(&oo->oo_owner.so_stateids)) {
630 stp = list_first_entry(&oo->oo_owner.so_stateids,
631 struct nfs4_ol_stateid, st_perstateowner);
632 release_open_stateid(stp);
636 static void release_last_closed_stateid(struct nfs4_openowner *oo)
638 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
641 unhash_stid(&s->st_stid);
642 free_generic_stateid(s);
643 oo->oo_last_closed_stid = NULL;
647 static void release_openowner(struct nfs4_openowner *oo)
649 unhash_openowner(oo);
650 list_del(&oo->oo_close_lru);
651 release_last_closed_stateid(oo);
652 nfs4_free_openowner(oo);
655 #define SESSION_HASH_SIZE 512
656 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
659 hash_sessionid(struct nfs4_sessionid *sessionid)
661 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
663 return sid->sequence % SESSION_HASH_SIZE;
668 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
670 u32 *ptr = (u32 *)(&sessionid->data[0]);
671 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
675 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
682 gen_sessionid(struct nfsd4_session *ses)
684 struct nfs4_client *clp = ses->se_client;
685 struct nfsd4_sessionid *sid;
687 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
688 sid->clientid = clp->cl_clientid;
689 sid->sequence = current_sessionid++;
694 * The protocol defines ca_maxresponssize_cached to include the size of
695 * the rpc header, but all we need to cache is the data starting after
696 * the end of the initial SEQUENCE operation--the rest we regenerate
697 * each time. Therefore we can advertise a ca_maxresponssize_cached
698 * value that is the number of bytes in our cache plus a few additional
699 * bytes. In order to stay on the safe side, and not promise more than
700 * we can cache, those additional bytes must be the minimum possible: 24
701 * bytes of rpc header (xid through accept state, with AUTH_NULL
702 * verifier), 12 for the compound header (with zero-length tag), and 44
703 * for the SEQUENCE op response:
705 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
708 free_session_slots(struct nfsd4_session *ses)
712 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
713 kfree(ses->se_slots[i]);
717 * We don't actually need to cache the rpc and session headers, so we
718 * can allocate a little less for each slot:
720 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
722 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
725 static int nfsd4_sanitize_slot_size(u32 size)
727 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
728 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
734 * XXX: If we run out of reserved DRC memory we could (up to a point)
735 * re-negotiate active sessions and reduce their slot usage to make
736 * room for new connections. For now we just fail the create session.
738 static int nfsd4_get_drc_mem(int slotsize, u32 num)
742 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
744 spin_lock(&nfsd_drc_lock);
745 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
746 nfsd_drc_max_mem - nfsd_drc_mem_used);
747 num = min_t(int, num, avail / slotsize);
748 nfsd_drc_mem_used += num * slotsize;
749 spin_unlock(&nfsd_drc_lock);
754 static void nfsd4_put_drc_mem(int slotsize, int num)
756 spin_lock(&nfsd_drc_lock);
757 nfsd_drc_mem_used -= slotsize * num;
758 spin_unlock(&nfsd_drc_lock);
761 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
763 struct nfsd4_session *new;
766 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
767 + sizeof(struct nfsd4_session) > PAGE_SIZE);
768 mem = numslots * sizeof(struct nfsd4_slot *);
770 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
773 /* allocate each struct nfsd4_slot and data cache in one piece */
774 for (i = 0; i < numslots; i++) {
775 mem = sizeof(struct nfsd4_slot) + slotsize;
776 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
777 if (!new->se_slots[i])
783 kfree(new->se_slots[i]);
788 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
790 u32 maxrpc = nfsd_serv->sv_max_mesg;
792 new->maxreqs = numslots;
793 new->maxresp_cached = min_t(u32, req->maxresp_cached,
794 slotsize + NFSD_MIN_HDR_SEQ_SZ);
795 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
796 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
797 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
800 static void free_conn(struct nfsd4_conn *c)
802 svc_xprt_put(c->cn_xprt);
806 static void nfsd4_conn_lost(struct svc_xpt_user *u)
808 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
809 struct nfs4_client *clp = c->cn_session->se_client;
811 spin_lock(&clp->cl_lock);
812 if (!list_empty(&c->cn_persession)) {
813 list_del(&c->cn_persession);
816 spin_unlock(&clp->cl_lock);
817 nfsd4_probe_callback(clp);
820 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
822 struct nfsd4_conn *conn;
824 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
827 svc_xprt_get(rqstp->rq_xprt);
828 conn->cn_xprt = rqstp->rq_xprt;
829 conn->cn_flags = flags;
830 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
834 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
836 conn->cn_session = ses;
837 list_add(&conn->cn_persession, &ses->se_conns);
840 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
842 struct nfs4_client *clp = ses->se_client;
844 spin_lock(&clp->cl_lock);
845 __nfsd4_hash_conn(conn, ses);
846 spin_unlock(&clp->cl_lock);
849 static int nfsd4_register_conn(struct nfsd4_conn *conn)
851 conn->cn_xpt_user.callback = nfsd4_conn_lost;
852 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
855 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
859 nfsd4_hash_conn(conn, ses);
860 ret = nfsd4_register_conn(conn);
862 /* oops; xprt is already down: */
863 nfsd4_conn_lost(&conn->cn_xpt_user);
864 if (ses->se_client->cl_cb_state == NFSD4_CB_DOWN &&
865 conn->cn_flags & NFS4_CDFC4_BACK) {
866 /* callback channel may be back up */
867 nfsd4_probe_callback(ses->se_client);
871 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
873 u32 dir = NFS4_CDFC4_FORE;
875 if (cses->flags & SESSION4_BACK_CHAN)
876 dir |= NFS4_CDFC4_BACK;
877 return alloc_conn(rqstp, dir);
880 /* must be called under client_lock */
881 static void nfsd4_del_conns(struct nfsd4_session *s)
883 struct nfs4_client *clp = s->se_client;
884 struct nfsd4_conn *c;
886 spin_lock(&clp->cl_lock);
887 while (!list_empty(&s->se_conns)) {
888 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
889 list_del_init(&c->cn_persession);
890 spin_unlock(&clp->cl_lock);
892 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
895 spin_lock(&clp->cl_lock);
897 spin_unlock(&clp->cl_lock);
900 static void free_session(struct kref *kref)
902 struct nfsd4_session *ses;
905 lockdep_assert_held(&client_lock);
906 ses = container_of(kref, struct nfsd4_session, se_ref);
907 nfsd4_del_conns(ses);
908 spin_lock(&nfsd_drc_lock);
909 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
910 nfsd_drc_mem_used -= mem;
911 spin_unlock(&nfsd_drc_lock);
912 free_session_slots(ses);
916 void nfsd4_put_session(struct nfsd4_session *ses)
918 spin_lock(&client_lock);
919 nfsd4_put_session_locked(ses);
920 spin_unlock(&client_lock);
923 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
925 struct nfsd4_session *new;
926 struct nfsd4_conn *conn;
927 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
928 int numslots, slotsize;
932 * Note decreasing slot size below client's request may
933 * make it difficult for client to function correctly, whereas
934 * decreasing the number of slots will (just?) affect
935 * performance. When short on memory we therefore prefer to
936 * decrease number of slots instead of their size.
938 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
939 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
943 new = alloc_session(slotsize, numslots);
945 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
948 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
950 new->se_client = clp;
953 INIT_LIST_HEAD(&new->se_conns);
955 new->se_cb_seq_nr = 1;
956 new->se_flags = cses->flags;
957 new->se_cb_prog = cses->callback_prog;
958 kref_init(&new->se_ref);
959 idx = hash_sessionid(&new->se_sessionid);
960 spin_lock(&client_lock);
961 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
962 spin_lock(&clp->cl_lock);
963 list_add(&new->se_perclnt, &clp->cl_sessions);
964 spin_unlock(&clp->cl_lock);
965 spin_unlock(&client_lock);
967 conn = alloc_conn_from_crses(rqstp, cses);
969 spin_lock(&client_lock);
970 free_session(&new->se_ref);
971 spin_unlock(&client_lock);
974 nfsd4_init_conn(rqstp, conn, new);
975 if (cses->flags & SESSION4_BACK_CHAN) {
976 struct sockaddr *sa = svc_addr(rqstp);
978 * This is a little silly; with sessions there's no real
979 * use for the callback address. Use the peer address
980 * as a reasonable default for now, but consider fixing
981 * the rpc client not to require an address in the
984 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
985 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
987 nfsd4_probe_callback(clp);
991 /* caller must hold client_lock */
992 static struct nfsd4_session *
993 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
995 struct nfsd4_session *elem;
998 dump_sessionid(__func__, sessionid);
999 idx = hash_sessionid(sessionid);
1000 /* Search in the appropriate list */
1001 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
1002 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1003 NFS4_MAX_SESSIONID_LEN)) {
1008 dprintk("%s: session not found\n", __func__);
1012 /* caller must hold client_lock */
1014 unhash_session(struct nfsd4_session *ses)
1016 list_del(&ses->se_hash);
1017 spin_lock(&ses->se_client->cl_lock);
1018 list_del(&ses->se_perclnt);
1019 spin_unlock(&ses->se_client->cl_lock);
1022 /* must be called under the client_lock */
1024 renew_client_locked(struct nfs4_client *clp)
1026 if (is_client_expired(clp)) {
1028 printk("%s: client (clientid %08x/%08x) already expired\n",
1030 clp->cl_clientid.cl_boot,
1031 clp->cl_clientid.cl_id);
1035 dprintk("renewing client (clientid %08x/%08x)\n",
1036 clp->cl_clientid.cl_boot,
1037 clp->cl_clientid.cl_id);
1038 list_move_tail(&clp->cl_lru, &client_lru);
1039 clp->cl_time = get_seconds();
1043 renew_client(struct nfs4_client *clp)
1045 spin_lock(&client_lock);
1046 renew_client_locked(clp);
1047 spin_unlock(&client_lock);
1050 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1052 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1054 if (clid->cl_boot == nn->boot_time)
1056 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1057 clid->cl_boot, clid->cl_id, nn->boot_time);
1062 * XXX Should we use a slab cache ?
1063 * This type of memory management is somewhat inefficient, but we use it
1064 * anyway since SETCLIENTID is not a common operation.
1066 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1068 struct nfs4_client *clp;
1070 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1073 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1074 if (clp->cl_name.data == NULL) {
1078 clp->cl_name.len = name.len;
1083 free_client(struct nfs4_client *clp)
1085 lockdep_assert_held(&client_lock);
1086 while (!list_empty(&clp->cl_sessions)) {
1087 struct nfsd4_session *ses;
1088 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1090 list_del(&ses->se_perclnt);
1091 nfsd4_put_session_locked(ses);
1093 free_svc_cred(&clp->cl_cred);
1094 kfree(clp->cl_name.data);
1099 release_session_client(struct nfsd4_session *session)
1101 struct nfs4_client *clp = session->se_client;
1103 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1105 if (is_client_expired(clp)) {
1107 session->se_client = NULL;
1109 renew_client_locked(clp);
1110 spin_unlock(&client_lock);
1113 /* must be called under the client_lock */
1115 unhash_client_locked(struct nfs4_client *clp)
1117 struct nfsd4_session *ses;
1119 mark_client_expired(clp);
1120 list_del(&clp->cl_lru);
1121 spin_lock(&clp->cl_lock);
1122 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1123 list_del_init(&ses->se_hash);
1124 spin_unlock(&clp->cl_lock);
1128 expire_client(struct nfs4_client *clp)
1130 struct nfs4_openowner *oo;
1131 struct nfs4_delegation *dp;
1132 struct list_head reaplist;
1134 INIT_LIST_HEAD(&reaplist);
1135 spin_lock(&recall_lock);
1136 while (!list_empty(&clp->cl_delegations)) {
1137 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1138 list_del_init(&dp->dl_perclnt);
1139 list_move(&dp->dl_recall_lru, &reaplist);
1141 spin_unlock(&recall_lock);
1142 while (!list_empty(&reaplist)) {
1143 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1144 unhash_delegation(dp);
1146 while (!list_empty(&clp->cl_openowners)) {
1147 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1148 release_openowner(oo);
1150 nfsd4_shutdown_callback(clp);
1151 if (clp->cl_cb_conn.cb_xprt)
1152 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1153 list_del(&clp->cl_idhash);
1154 list_del(&clp->cl_strhash);
1155 spin_lock(&client_lock);
1156 unhash_client_locked(clp);
1157 if (atomic_read(&clp->cl_refcount) == 0)
1159 spin_unlock(&client_lock);
1162 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1164 memcpy(target->cl_verifier.data, source->data,
1165 sizeof(target->cl_verifier.data));
1168 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1170 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1171 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1174 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1176 if (source->cr_principal) {
1177 target->cr_principal =
1178 kstrdup(source->cr_principal, GFP_KERNEL);
1179 if (target->cr_principal == NULL)
1182 target->cr_principal = NULL;
1183 target->cr_flavor = source->cr_flavor;
1184 target->cr_uid = source->cr_uid;
1185 target->cr_gid = source->cr_gid;
1186 target->cr_group_info = source->cr_group_info;
1187 get_group_info(target->cr_group_info);
1191 static int same_name(const char *n1, const char *n2)
1193 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1197 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1199 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1203 same_clid(clientid_t *cl1, clientid_t *cl2)
1205 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1208 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1212 if (g1->ngroups != g2->ngroups)
1214 for (i=0; i<g1->ngroups; i++)
1215 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1221 * RFC 3530 language requires clid_inuse be returned when the
1222 * "principal" associated with a requests differs from that previously
1223 * used. We use uid, gid's, and gss principal string as our best
1224 * approximation. We also don't want to allow non-gss use of a client
1225 * established using gss: in theory cr_principal should catch that
1226 * change, but in practice cr_principal can be null even in the gss case
1227 * since gssd doesn't always pass down a principal string.
1229 static bool is_gss_cred(struct svc_cred *cr)
1231 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1232 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1237 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1239 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1240 || (cr1->cr_uid != cr2->cr_uid)
1241 || (cr1->cr_gid != cr2->cr_gid)
1242 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1244 if (cr1->cr_principal == cr2->cr_principal)
1246 if (!cr1->cr_principal || !cr2->cr_principal)
1248 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1251 static void gen_clid(struct nfs4_client *clp)
1253 static u32 current_clientid = 1;
1254 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
1256 clp->cl_clientid.cl_boot = nn->boot_time;
1257 clp->cl_clientid.cl_id = current_clientid++;
1260 static void gen_confirm(struct nfs4_client *clp)
1265 verf[0] = (__be32)get_seconds();
1266 verf[1] = (__be32)i++;
1267 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1270 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1272 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1275 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1277 struct nfs4_stid *s;
1279 s = find_stateid(cl, t);
1282 if (typemask & s->sc_type)
1287 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1288 struct svc_rqst *rqstp, nfs4_verifier *verf)
1290 struct nfs4_client *clp;
1291 struct sockaddr *sa = svc_addr(rqstp);
1294 clp = alloc_client(name);
1298 INIT_LIST_HEAD(&clp->cl_sessions);
1299 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1301 spin_lock(&client_lock);
1303 spin_unlock(&client_lock);
1306 idr_init(&clp->cl_stateids);
1307 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1308 atomic_set(&clp->cl_refcount, 0);
1309 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1310 INIT_LIST_HEAD(&clp->cl_idhash);
1311 INIT_LIST_HEAD(&clp->cl_strhash);
1312 INIT_LIST_HEAD(&clp->cl_openowners);
1313 INIT_LIST_HEAD(&clp->cl_delegations);
1314 INIT_LIST_HEAD(&clp->cl_lru);
1315 INIT_LIST_HEAD(&clp->cl_callbacks);
1316 spin_lock_init(&clp->cl_lock);
1317 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1318 clp->cl_time = get_seconds();
1319 clear_bit(0, &clp->cl_cb_slot_busy);
1320 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1321 copy_verf(clp, verf);
1322 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1324 clp->cl_cb_session = NULL;
1329 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1331 unsigned int idhashval;
1333 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1334 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1335 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1340 move_to_confirmed(struct nfs4_client *clp)
1342 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1343 unsigned int strhashval;
1345 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1346 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1347 strhashval = clientstr_hashval(clp->cl_recdir);
1348 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1352 static struct nfs4_client *
1353 find_confirmed_client(clientid_t *clid, bool sessions)
1355 struct nfs4_client *clp;
1356 unsigned int idhashval = clientid_hashval(clid->cl_id);
1358 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1359 if (same_clid(&clp->cl_clientid, clid)) {
1360 if ((bool)clp->cl_minorversion != sessions)
1369 static struct nfs4_client *
1370 find_unconfirmed_client(clientid_t *clid, bool sessions)
1372 struct nfs4_client *clp;
1373 unsigned int idhashval = clientid_hashval(clid->cl_id);
1375 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1376 if (same_clid(&clp->cl_clientid, clid)) {
1377 if ((bool)clp->cl_minorversion != sessions)
1385 static bool clp_used_exchangeid(struct nfs4_client *clp)
1387 return clp->cl_exchange_flags != 0;
1390 static struct nfs4_client *
1391 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1393 struct nfs4_client *clp;
1395 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1396 if (same_name(clp->cl_recdir, dname))
1402 static struct nfs4_client *
1403 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1405 struct nfs4_client *clp;
1407 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1408 if (same_name(clp->cl_recdir, dname))
1415 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1417 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1418 struct sockaddr *sa = svc_addr(rqstp);
1419 u32 scopeid = rpc_get_scope_id(sa);
1420 unsigned short expected_family;
1422 /* Currently, we only support tcp and tcp6 for the callback channel */
1423 if (se->se_callback_netid_len == 3 &&
1424 !memcmp(se->se_callback_netid_val, "tcp", 3))
1425 expected_family = AF_INET;
1426 else if (se->se_callback_netid_len == 4 &&
1427 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1428 expected_family = AF_INET6;
1432 conn->cb_addrlen = rpc_uaddr2sockaddr(&init_net, se->se_callback_addr_val,
1433 se->se_callback_addr_len,
1434 (struct sockaddr *)&conn->cb_addr,
1435 sizeof(conn->cb_addr));
1437 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1440 if (conn->cb_addr.ss_family == AF_INET6)
1441 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1443 conn->cb_prog = se->se_callback_prog;
1444 conn->cb_ident = se->se_callback_ident;
1445 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1448 conn->cb_addr.ss_family = AF_UNSPEC;
1449 conn->cb_addrlen = 0;
1450 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1451 "will not receive delegations\n",
1452 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1458 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1461 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1463 struct nfsd4_slot *slot = resp->cstate.slot;
1466 dprintk("--> %s slot %p\n", __func__, slot);
1468 slot->sl_opcnt = resp->opcnt;
1469 slot->sl_status = resp->cstate.status;
1471 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1472 if (nfsd4_not_cached(resp)) {
1473 slot->sl_datalen = 0;
1476 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1477 base = (char *)resp->cstate.datap -
1478 (char *)resp->xbuf->head[0].iov_base;
1479 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1481 WARN("%s: sessions DRC could not cache compound\n", __func__);
1486 * Encode the replay sequence operation from the slot values.
1487 * If cachethis is FALSE encode the uncached rep error on the next
1488 * operation which sets resp->p and increments resp->opcnt for
1489 * nfs4svc_encode_compoundres.
1493 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1494 struct nfsd4_compoundres *resp)
1496 struct nfsd4_op *op;
1497 struct nfsd4_slot *slot = resp->cstate.slot;
1499 /* Encode the replayed sequence operation */
1500 op = &args->ops[resp->opcnt - 1];
1501 nfsd4_encode_operation(resp, op);
1503 /* Return nfserr_retry_uncached_rep in next operation. */
1504 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1505 op = &args->ops[resp->opcnt++];
1506 op->status = nfserr_retry_uncached_rep;
1507 nfsd4_encode_operation(resp, op);
1513 * The sequence operation is not cached because we can use the slot and
1517 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1518 struct nfsd4_sequence *seq)
1520 struct nfsd4_slot *slot = resp->cstate.slot;
1523 dprintk("--> %s slot %p\n", __func__, slot);
1525 /* Either returns 0 or nfserr_retry_uncached */
1526 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1527 if (status == nfserr_retry_uncached_rep)
1530 /* The sequence operation has been encoded, cstate->datap set. */
1531 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1533 resp->opcnt = slot->sl_opcnt;
1534 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1535 status = slot->sl_status;
1541 * Set the exchange_id flags returned by the server.
1544 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1546 /* pNFS is not supported */
1547 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1549 /* Referrals are supported, Migration is not. */
1550 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1552 /* set the wire flags to return to client. */
1553 clid->flags = new->cl_exchange_flags;
1556 static bool client_has_state(struct nfs4_client *clp)
1559 * Note clp->cl_openowners check isn't quite right: there's no
1560 * need to count owners without stateid's.
1562 * Also note we should probably be using this in 4.0 case too.
1564 return !list_empty(&clp->cl_openowners)
1565 || !list_empty(&clp->cl_delegations)
1566 || !list_empty(&clp->cl_sessions);
1570 nfsd4_exchange_id(struct svc_rqst *rqstp,
1571 struct nfsd4_compound_state *cstate,
1572 struct nfsd4_exchange_id *exid)
1574 struct nfs4_client *unconf, *conf, *new;
1576 unsigned int strhashval;
1577 char dname[HEXDIR_LEN];
1578 char addr_str[INET6_ADDRSTRLEN];
1579 nfs4_verifier verf = exid->verifier;
1580 struct sockaddr *sa = svc_addr(rqstp);
1581 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1583 rpc_ntop(sa, addr_str, sizeof(addr_str));
1584 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1585 "ip_addr=%s flags %x, spa_how %d\n",
1586 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1587 addr_str, exid->flags, exid->spa_how);
1589 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1590 return nfserr_inval;
1592 /* Currently only support SP4_NONE */
1593 switch (exid->spa_how) {
1597 return nfserr_serverfault;
1599 BUG(); /* checked by xdr code */
1601 return nfserr_serverfault; /* no excuse :-/ */
1604 status = nfs4_make_rec_clidname(dname, &exid->clname);
1609 strhashval = clientstr_hashval(dname);
1611 /* Cases below refer to rfc 5661 section 18.35.4: */
1613 conf = find_confirmed_client_by_str(dname, strhashval);
1615 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1616 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1619 if (!clp_used_exchangeid(conf)) { /* buggy client */
1620 status = nfserr_inval;
1623 if (!creds_match) { /* case 9 */
1624 status = nfserr_perm;
1627 if (!verfs_match) { /* case 8 */
1628 status = nfserr_not_same;
1632 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1636 if (!creds_match) { /* case 3 */
1637 if (client_has_state(conf)) {
1638 status = nfserr_clid_inuse;
1641 expire_client(conf);
1644 if (verfs_match) { /* case 2 */
1645 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1649 /* case 5, client reboot */
1653 if (update) { /* case 7 */
1654 status = nfserr_noent;
1658 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1659 if (unconf) /* case 4, possible retry or client restart */
1660 expire_client(unconf);
1662 /* case 1 (normal case) */
1664 new = create_client(exid->clname, dname, rqstp, &verf);
1666 status = nfserr_jukebox;
1669 new->cl_minorversion = 1;
1672 add_to_unconfirmed(new, strhashval);
1674 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1675 exid->clientid.cl_id = new->cl_clientid.cl_id;
1677 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1678 nfsd4_set_ex_flags(new, exid);
1680 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1681 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1685 nfs4_unlock_state();
1690 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1692 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1695 /* The slot is in use, and no response has been sent. */
1697 if (seqid == slot_seqid)
1698 return nfserr_jukebox;
1700 return nfserr_seq_misordered;
1702 /* Note unsigned 32-bit arithmetic handles wraparound: */
1703 if (likely(seqid == slot_seqid + 1))
1705 if (seqid == slot_seqid)
1706 return nfserr_replay_cache;
1707 return nfserr_seq_misordered;
1711 * Cache the create session result into the create session single DRC
1712 * slot cache by saving the xdr structure. sl_seqid has been set.
1713 * Do this for solo or embedded create session operations.
1716 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1717 struct nfsd4_clid_slot *slot, __be32 nfserr)
1719 slot->sl_status = nfserr;
1720 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1724 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1725 struct nfsd4_clid_slot *slot)
1727 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1728 return slot->sl_status;
1731 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1732 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1733 1 + /* MIN tag is length with zero, only length */ \
1734 3 + /* version, opcount, opcode */ \
1735 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1736 /* seqid, slotID, slotID, cache */ \
1737 4 ) * sizeof(__be32))
1739 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1740 2 + /* verifier: AUTH_NULL, length 0 */\
1742 1 + /* MIN tag is length with zero, only length */ \
1743 3 + /* opcount, opcode, opstatus*/ \
1744 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1745 /* seqid, slotID, slotID, slotID, status */ \
1746 5 ) * sizeof(__be32))
1748 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1750 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1751 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1755 nfsd4_create_session(struct svc_rqst *rqstp,
1756 struct nfsd4_compound_state *cstate,
1757 struct nfsd4_create_session *cr_ses)
1759 struct sockaddr *sa = svc_addr(rqstp);
1760 struct nfs4_client *conf, *unconf;
1761 struct nfsd4_session *new;
1762 struct nfsd4_clid_slot *cs_slot = NULL;
1763 bool confirm_me = false;
1766 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1767 return nfserr_inval;
1768 if (check_forechannel_attrs(cr_ses->fore_channel))
1769 return nfserr_toosmall;
1772 unconf = find_unconfirmed_client(&cr_ses->clientid, true);
1773 conf = find_confirmed_client(&cr_ses->clientid, true);
1776 cs_slot = &conf->cl_cs_slot;
1777 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1778 if (status == nfserr_replay_cache) {
1779 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1781 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1782 status = nfserr_seq_misordered;
1785 } else if (unconf) {
1786 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1787 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1788 status = nfserr_clid_inuse;
1791 cs_slot = &unconf->cl_cs_slot;
1792 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1794 /* an unconfirmed replay returns misordered */
1795 status = nfserr_seq_misordered;
1801 status = nfserr_stale_clientid;
1805 * We do not support RDMA or persistent sessions
1807 cr_ses->flags &= ~SESSION4_PERSIST;
1808 cr_ses->flags &= ~SESSION4_RDMA;
1810 status = nfserr_jukebox;
1811 new = alloc_init_session(rqstp, conf, cr_ses);
1815 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1816 NFS4_MAX_SESSIONID_LEN);
1817 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1818 sizeof(struct nfsd4_channel_attrs));
1819 cs_slot->sl_seqid++;
1820 cr_ses->seqid = cs_slot->sl_seqid;
1822 /* cache solo and embedded create sessions under the state lock */
1823 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1825 unsigned int hash = clientstr_hashval(unconf->cl_recdir);
1826 struct nfs4_client *old =
1827 find_confirmed_client_by_str(conf->cl_recdir, hash);
1830 move_to_confirmed(conf);
1833 nfs4_unlock_state();
1834 dprintk("%s returns %d\n", __func__, ntohl(status));
1838 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1840 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1841 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1843 return argp->opcnt == resp->opcnt;
1846 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1849 case NFS4_CDFC4_FORE:
1850 case NFS4_CDFC4_BACK:
1852 case NFS4_CDFC4_FORE_OR_BOTH:
1853 case NFS4_CDFC4_BACK_OR_BOTH:
1854 *dir = NFS4_CDFC4_BOTH;
1857 return nfserr_inval;
1860 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1861 struct nfsd4_compound_state *cstate,
1862 struct nfsd4_bind_conn_to_session *bcts)
1865 struct nfsd4_conn *conn;
1867 if (!nfsd4_last_compound_op(rqstp))
1868 return nfserr_not_only_op;
1869 spin_lock(&client_lock);
1870 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1871 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1872 * client_lock iself: */
1873 if (cstate->session) {
1874 nfsd4_get_session(cstate->session);
1875 atomic_inc(&cstate->session->se_client->cl_refcount);
1877 spin_unlock(&client_lock);
1878 if (!cstate->session)
1879 return nfserr_badsession;
1881 status = nfsd4_map_bcts_dir(&bcts->dir);
1884 conn = alloc_conn(rqstp, bcts->dir);
1886 return nfserr_jukebox;
1887 nfsd4_init_conn(rqstp, conn, cstate->session);
1891 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1895 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1899 nfsd4_destroy_session(struct svc_rqst *r,
1900 struct nfsd4_compound_state *cstate,
1901 struct nfsd4_destroy_session *sessionid)
1903 struct nfsd4_session *ses;
1904 __be32 status = nfserr_badsession;
1907 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1908 * - Should we return nfserr_back_chan_busy if waiting for
1909 * callbacks on to-be-destroyed session?
1910 * - Do we need to clear any callback info from previous session?
1913 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1914 if (!nfsd4_last_compound_op(r))
1915 return nfserr_not_only_op;
1917 dump_sessionid(__func__, &sessionid->sessionid);
1918 spin_lock(&client_lock);
1919 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1921 spin_unlock(&client_lock);
1925 unhash_session(ses);
1926 spin_unlock(&client_lock);
1929 nfsd4_probe_callback_sync(ses->se_client);
1930 nfs4_unlock_state();
1932 spin_lock(&client_lock);
1933 nfsd4_del_conns(ses);
1934 nfsd4_put_session_locked(ses);
1935 spin_unlock(&client_lock);
1938 dprintk("%s returns %d\n", __func__, ntohl(status));
1942 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1944 struct nfsd4_conn *c;
1946 list_for_each_entry(c, &s->se_conns, cn_persession) {
1947 if (c->cn_xprt == xpt) {
1954 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1956 struct nfs4_client *clp = ses->se_client;
1957 struct nfsd4_conn *c;
1960 spin_lock(&clp->cl_lock);
1961 c = __nfsd4_find_conn(new->cn_xprt, ses);
1963 spin_unlock(&clp->cl_lock);
1967 __nfsd4_hash_conn(new, ses);
1968 spin_unlock(&clp->cl_lock);
1969 ret = nfsd4_register_conn(new);
1971 /* oops; xprt is already down: */
1972 nfsd4_conn_lost(&new->cn_xpt_user);
1976 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1978 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1980 return args->opcnt > session->se_fchannel.maxops;
1983 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1984 struct nfsd4_session *session)
1986 struct xdr_buf *xb = &rqstp->rq_arg;
1988 return xb->len > session->se_fchannel.maxreq_sz;
1992 nfsd4_sequence(struct svc_rqst *rqstp,
1993 struct nfsd4_compound_state *cstate,
1994 struct nfsd4_sequence *seq)
1996 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1997 struct nfsd4_session *session;
1998 struct nfsd4_slot *slot;
1999 struct nfsd4_conn *conn;
2002 if (resp->opcnt != 1)
2003 return nfserr_sequence_pos;
2006 * Will be either used or freed by nfsd4_sequence_check_conn
2009 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2011 return nfserr_jukebox;
2013 spin_lock(&client_lock);
2014 status = nfserr_badsession;
2015 session = find_in_sessionid_hashtbl(&seq->sessionid);
2019 status = nfserr_too_many_ops;
2020 if (nfsd4_session_too_many_ops(rqstp, session))
2023 status = nfserr_req_too_big;
2024 if (nfsd4_request_too_big(rqstp, session))
2027 status = nfserr_badslot;
2028 if (seq->slotid >= session->se_fchannel.maxreqs)
2031 slot = session->se_slots[seq->slotid];
2032 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2034 /* We do not negotiate the number of slots yet, so set the
2035 * maxslots to the session maxreqs which is used to encode
2036 * sr_highest_slotid and the sr_target_slot id to maxslots */
2037 seq->maxslots = session->se_fchannel.maxreqs;
2039 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2040 slot->sl_flags & NFSD4_SLOT_INUSE);
2041 if (status == nfserr_replay_cache) {
2042 status = nfserr_seq_misordered;
2043 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2045 cstate->slot = slot;
2046 cstate->session = session;
2047 /* Return the cached reply status and set cstate->status
2048 * for nfsd4_proc_compound processing */
2049 status = nfsd4_replay_cache_entry(resp, seq);
2050 cstate->status = nfserr_replay_cache;
2056 nfsd4_sequence_check_conn(conn, session);
2059 /* Success! bump slot seqid */
2060 slot->sl_seqid = seq->seqid;
2061 slot->sl_flags |= NFSD4_SLOT_INUSE;
2063 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2065 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2067 cstate->slot = slot;
2068 cstate->session = session;
2071 /* Hold a session reference until done processing the compound. */
2072 if (cstate->session) {
2073 struct nfs4_client *clp = session->se_client;
2075 nfsd4_get_session(cstate->session);
2076 atomic_inc(&clp->cl_refcount);
2077 switch (clp->cl_cb_state) {
2079 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2081 case NFSD4_CB_FAULT:
2082 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2085 seq->status_flags = 0;
2089 spin_unlock(&client_lock);
2090 dprintk("%s: return %d\n", __func__, ntohl(status));
2095 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2097 struct nfs4_client *conf, *unconf, *clp;
2101 unconf = find_unconfirmed_client(&dc->clientid, true);
2102 conf = find_confirmed_client(&dc->clientid, true);
2107 if (!is_client_expired(conf) && client_has_state(conf)) {
2108 status = nfserr_clientid_busy;
2112 /* rfc5661 18.50.3 */
2113 if (cstate->session && conf == cstate->session->se_client) {
2114 status = nfserr_clientid_busy;
2120 status = nfserr_stale_clientid;
2126 nfs4_unlock_state();
2127 dprintk("%s return %d\n", __func__, ntohl(status));
2132 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2136 if (rc->rca_one_fs) {
2137 if (!cstate->current_fh.fh_dentry)
2138 return nfserr_nofilehandle;
2140 * We don't take advantage of the rca_one_fs case.
2141 * That's OK, it's optional, we can safely ignore it.
2147 status = nfserr_complete_already;
2148 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2149 &cstate->session->se_client->cl_flags))
2152 status = nfserr_stale_clientid;
2153 if (is_client_expired(cstate->session->se_client))
2155 * The following error isn't really legal.
2156 * But we only get here if the client just explicitly
2157 * destroyed the client. Surely it no longer cares what
2158 * error it gets back on an operation for the dead
2164 nfsd4_client_record_create(cstate->session->se_client);
2166 nfs4_unlock_state();
2171 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2172 struct nfsd4_setclientid *setclid)
2174 struct xdr_netobj clname = setclid->se_name;
2175 nfs4_verifier clverifier = setclid->se_verf;
2176 unsigned int strhashval;
2177 struct nfs4_client *conf, *unconf, *new;
2179 char dname[HEXDIR_LEN];
2181 status = nfs4_make_rec_clidname(dname, &clname);
2185 strhashval = clientstr_hashval(dname);
2187 /* Cases below refer to rfc 3530 section 14.2.33: */
2189 conf = find_confirmed_client_by_str(dname, strhashval);
2192 status = nfserr_clid_inuse;
2193 if (clp_used_exchangeid(conf))
2195 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2196 char addr_str[INET6_ADDRSTRLEN];
2197 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2199 dprintk("NFSD: setclientid: string in use by client "
2200 "at %s\n", addr_str);
2204 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2206 expire_client(unconf);
2207 status = nfserr_jukebox;
2208 new = create_client(clname, dname, rqstp, &clverifier);
2211 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2212 /* case 1: probable callback update */
2213 copy_clid(new, conf);
2214 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2216 new->cl_minorversion = 0;
2217 gen_callback(new, setclid, rqstp);
2218 add_to_unconfirmed(new, strhashval);
2219 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2220 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2221 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2224 nfs4_unlock_state();
2230 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2231 struct nfsd4_compound_state *cstate,
2232 struct nfsd4_setclientid_confirm *setclientid_confirm)
2234 struct nfs4_client *conf, *unconf;
2235 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2236 clientid_t * clid = &setclientid_confirm->sc_clientid;
2238 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2240 if (STALE_CLIENTID(clid, nn))
2241 return nfserr_stale_clientid;
2244 conf = find_confirmed_client(clid, false);
2245 unconf = find_unconfirmed_client(clid, false);
2247 * We try hard to give out unique clientid's, so if we get an
2248 * attempt to confirm the same clientid with a different cred,
2249 * there's a bug somewhere. Let's charitably assume it's our
2252 status = nfserr_serverfault;
2253 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2255 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2257 /* cases below refer to rfc 3530 section 14.2.34: */
2258 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2259 if (conf && !unconf) /* case 2: probable retransmit */
2261 else /* case 4: client hasn't noticed we rebooted yet? */
2262 status = nfserr_stale_clientid;
2266 if (conf) { /* case 1: callback update */
2267 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2268 nfsd4_probe_callback(conf);
2269 expire_client(unconf);
2270 } else { /* case 3: normal case; new or rebooted client */
2271 unsigned int hash = clientstr_hashval(unconf->cl_recdir);
2273 conf = find_confirmed_client_by_str(unconf->cl_recdir, hash);
2275 nfsd4_client_record_remove(conf);
2276 expire_client(conf);
2278 move_to_confirmed(unconf);
2279 nfsd4_probe_callback(unconf);
2282 nfs4_unlock_state();
2286 static struct nfs4_file *nfsd4_alloc_file(void)
2288 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2291 /* OPEN Share state helper functions */
2292 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2294 unsigned int hashval = file_hashval(ino);
2296 atomic_set(&fp->fi_ref, 1);
2297 INIT_LIST_HEAD(&fp->fi_hash);
2298 INIT_LIST_HEAD(&fp->fi_stateids);
2299 INIT_LIST_HEAD(&fp->fi_delegations);
2300 fp->fi_inode = igrab(ino);
2301 fp->fi_had_conflict = false;
2302 fp->fi_lease = NULL;
2303 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2304 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2305 spin_lock(&recall_lock);
2306 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2307 spin_unlock(&recall_lock);
2311 nfsd4_free_slab(struct kmem_cache **slab)
2315 kmem_cache_destroy(*slab);
2320 nfsd4_free_slabs(void)
2322 nfsd4_free_slab(&openowner_slab);
2323 nfsd4_free_slab(&lockowner_slab);
2324 nfsd4_free_slab(&file_slab);
2325 nfsd4_free_slab(&stateid_slab);
2326 nfsd4_free_slab(&deleg_slab);
2330 nfsd4_init_slabs(void)
2332 openowner_slab = kmem_cache_create("nfsd4_openowners",
2333 sizeof(struct nfs4_openowner), 0, 0, NULL);
2334 if (openowner_slab == NULL)
2336 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2337 sizeof(struct nfs4_openowner), 0, 0, NULL);
2338 if (lockowner_slab == NULL)
2340 file_slab = kmem_cache_create("nfsd4_files",
2341 sizeof(struct nfs4_file), 0, 0, NULL);
2342 if (file_slab == NULL)
2344 stateid_slab = kmem_cache_create("nfsd4_stateids",
2345 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2346 if (stateid_slab == NULL)
2348 deleg_slab = kmem_cache_create("nfsd4_delegations",
2349 sizeof(struct nfs4_delegation), 0, 0, NULL);
2350 if (deleg_slab == NULL)
2355 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2359 void nfs4_free_openowner(struct nfs4_openowner *oo)
2361 kfree(oo->oo_owner.so_owner.data);
2362 kmem_cache_free(openowner_slab, oo);
2365 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2367 kfree(lo->lo_owner.so_owner.data);
2368 kmem_cache_free(lockowner_slab, lo);
2371 static void init_nfs4_replay(struct nfs4_replay *rp)
2373 rp->rp_status = nfserr_serverfault;
2375 rp->rp_buf = rp->rp_ibuf;
2378 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2380 struct nfs4_stateowner *sop;
2382 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2386 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2387 if (!sop->so_owner.data) {
2388 kmem_cache_free(slab, sop);
2391 sop->so_owner.len = owner->len;
2393 INIT_LIST_HEAD(&sop->so_stateids);
2394 sop->so_client = clp;
2395 init_nfs4_replay(&sop->so_replay);
2399 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2401 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2402 list_add(&oo->oo_perclient, &clp->cl_openowners);
2405 static struct nfs4_openowner *
2406 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2407 struct nfs4_openowner *oo;
2409 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2412 oo->oo_owner.so_is_open_owner = 1;
2413 oo->oo_owner.so_seqid = open->op_seqid;
2414 oo->oo_flags = NFS4_OO_NEW;
2416 oo->oo_last_closed_stid = NULL;
2417 INIT_LIST_HEAD(&oo->oo_close_lru);
2418 hash_openowner(oo, clp, strhashval);
2422 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2423 struct nfs4_openowner *oo = open->op_openowner;
2424 struct nfs4_client *clp = oo->oo_owner.so_client;
2426 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2427 INIT_LIST_HEAD(&stp->st_lockowners);
2428 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2429 list_add(&stp->st_perfile, &fp->fi_stateids);
2430 stp->st_stateowner = &oo->oo_owner;
2433 stp->st_access_bmap = 0;
2434 stp->st_deny_bmap = 0;
2435 set_access(open->op_share_access, stp);
2436 set_deny(open->op_share_deny, stp);
2437 stp->st_openstp = NULL;
2441 move_to_close_lru(struct nfs4_openowner *oo)
2443 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2445 list_move_tail(&oo->oo_close_lru, &close_lru);
2446 oo->oo_time = get_seconds();
2450 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2453 return (sop->so_owner.len == owner->len) &&
2454 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2455 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2458 static struct nfs4_openowner *
2459 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open, bool sessions)
2461 struct nfs4_stateowner *so;
2462 struct nfs4_openowner *oo;
2463 struct nfs4_client *clp;
2465 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2466 if (!so->so_is_open_owner)
2468 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2470 clp = oo->oo_owner.so_client;
2471 if ((bool)clp->cl_minorversion != sessions)
2473 renew_client(oo->oo_owner.so_client);
2480 /* search file_hashtbl[] for file */
2481 static struct nfs4_file *
2482 find_file(struct inode *ino)
2484 unsigned int hashval = file_hashval(ino);
2485 struct nfs4_file *fp;
2487 spin_lock(&recall_lock);
2488 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2489 if (fp->fi_inode == ino) {
2491 spin_unlock(&recall_lock);
2495 spin_unlock(&recall_lock);
2500 * Called to check deny when READ with all zero stateid or
2501 * WRITE with all zero or all one stateid
2504 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2506 struct inode *ino = current_fh->fh_dentry->d_inode;
2507 struct nfs4_file *fp;
2508 struct nfs4_ol_stateid *stp;
2511 dprintk("NFSD: nfs4_share_conflict\n");
2513 fp = find_file(ino);
2516 ret = nfserr_locked;
2517 /* Search for conflicting share reservations */
2518 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2519 if (test_deny(deny_type, stp) ||
2520 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2529 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2531 /* We're assuming the state code never drops its reference
2532 * without first removing the lease. Since we're in this lease
2533 * callback (and since the lease code is serialized by the kernel
2534 * lock) we know the server hasn't removed the lease yet, we know
2535 * it's safe to take a reference: */
2536 atomic_inc(&dp->dl_count);
2538 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2540 /* only place dl_time is set. protected by lock_flocks*/
2541 dp->dl_time = get_seconds();
2543 nfsd4_cb_recall(dp);
2546 /* Called from break_lease() with lock_flocks() held. */
2547 static void nfsd_break_deleg_cb(struct file_lock *fl)
2549 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2550 struct nfs4_delegation *dp;
2553 /* We assume break_lease is only called once per lease: */
2554 BUG_ON(fp->fi_had_conflict);
2556 * We don't want the locks code to timeout the lease for us;
2557 * we'll remove it ourself if a delegation isn't returned
2560 fl->fl_break_time = 0;
2562 spin_lock(&recall_lock);
2563 fp->fi_had_conflict = true;
2564 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2565 nfsd_break_one_deleg(dp);
2566 spin_unlock(&recall_lock);
2570 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2573 return lease_modify(onlist, arg);
2578 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2579 .lm_break = nfsd_break_deleg_cb,
2580 .lm_change = nfsd_change_deleg_cb,
2583 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2585 if (nfsd4_has_session(cstate))
2587 if (seqid == so->so_seqid - 1)
2588 return nfserr_replay_me;
2589 if (seqid == so->so_seqid)
2591 return nfserr_bad_seqid;
2595 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2596 struct nfsd4_open *open)
2598 clientid_t *clientid = &open->op_clientid;
2599 struct nfs4_client *clp = NULL;
2600 unsigned int strhashval;
2601 struct nfs4_openowner *oo = NULL;
2603 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2605 if (STALE_CLIENTID(&open->op_clientid, nn))
2606 return nfserr_stale_clientid;
2608 * In case we need it later, after we've already created the
2609 * file and don't want to risk a further failure:
2611 open->op_file = nfsd4_alloc_file();
2612 if (open->op_file == NULL)
2613 return nfserr_jukebox;
2615 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2616 oo = find_openstateowner_str(strhashval, open, cstate->minorversion);
2617 open->op_openowner = oo;
2619 clp = find_confirmed_client(clientid, cstate->minorversion);
2621 return nfserr_expired;
2624 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2625 /* Replace unconfirmed owners without checking for replay. */
2626 clp = oo->oo_owner.so_client;
2627 release_openowner(oo);
2628 open->op_openowner = NULL;
2631 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2634 clp = oo->oo_owner.so_client;
2637 oo = alloc_init_open_stateowner(strhashval, clp, open);
2639 return nfserr_jukebox;
2640 open->op_openowner = oo;
2642 open->op_stp = nfs4_alloc_stateid(clp);
2644 return nfserr_jukebox;
2648 static inline __be32
2649 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2651 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2652 return nfserr_openmode;
2657 static int share_access_to_flags(u32 share_access)
2659 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2662 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2664 struct nfs4_stid *ret;
2666 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2669 return delegstateid(ret);
2672 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2674 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2675 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2679 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2680 struct nfs4_delegation **dp)
2683 __be32 status = nfserr_bad_stateid;
2685 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2688 flags = share_access_to_flags(open->op_share_access);
2689 status = nfs4_check_delegmode(*dp, flags);
2693 if (!nfsd4_is_deleg_cur(open))
2697 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2702 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2704 struct nfs4_ol_stateid *local;
2705 struct nfs4_openowner *oo = open->op_openowner;
2707 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2708 /* ignore lock owners */
2709 if (local->st_stateowner->so_is_open_owner == 0)
2711 /* remember if we have seen this open owner */
2712 if (local->st_stateowner == &oo->oo_owner)
2714 /* check for conflicting share reservations */
2715 if (!test_share(local, open))
2716 return nfserr_share_denied;
2721 static inline int nfs4_access_to_access(u32 nfs4_access)
2725 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2726 flags |= NFSD_MAY_READ;
2727 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2728 flags |= NFSD_MAY_WRITE;
2732 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2733 struct svc_fh *cur_fh, struct nfsd4_open *open)
2736 int oflag = nfs4_access_to_omode(open->op_share_access);
2737 int access = nfs4_access_to_access(open->op_share_access);
2739 if (!fp->fi_fds[oflag]) {
2740 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2741 &fp->fi_fds[oflag]);
2745 nfs4_file_get_access(fp, oflag);
2750 static inline __be32
2751 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2752 struct nfsd4_open *open)
2754 struct iattr iattr = {
2755 .ia_valid = ATTR_SIZE,
2758 if (!open->op_truncate)
2760 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2761 return nfserr_inval;
2762 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2766 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)
2768 u32 op_share_access = open->op_share_access;
2772 new_access = !test_access(op_share_access, stp);
2774 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2778 status = nfsd4_truncate(rqstp, cur_fh, open);
2781 int oflag = nfs4_access_to_omode(op_share_access);
2782 nfs4_file_put_access(fp, oflag);
2786 /* remember the open */
2787 set_access(op_share_access, stp);
2788 set_deny(open->op_share_deny, stp);
2795 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2797 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2800 /* Should we give out recallable state?: */
2801 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2803 if (clp->cl_cb_state == NFSD4_CB_UP)
2806 * In the sessions case, since we don't have to establish a
2807 * separate connection for callbacks, we assume it's OK
2808 * until we hear otherwise:
2810 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2813 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2815 struct file_lock *fl;
2817 fl = locks_alloc_lock();
2820 locks_init_lock(fl);
2821 fl->fl_lmops = &nfsd_lease_mng_ops;
2822 fl->fl_flags = FL_LEASE;
2823 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2824 fl->fl_end = OFFSET_MAX;
2825 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2826 fl->fl_pid = current->tgid;
2830 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2832 struct nfs4_file *fp = dp->dl_file;
2833 struct file_lock *fl;
2836 fl = nfs4_alloc_init_lease(dp, flag);
2839 fl->fl_file = find_readable_file(fp);
2840 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2841 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2843 list_del_init(&dp->dl_perclnt);
2844 locks_free_lock(fl);
2848 fp->fi_deleg_file = fl->fl_file;
2849 get_file(fp->fi_deleg_file);
2850 atomic_set(&fp->fi_delegees, 1);
2851 list_add(&dp->dl_perfile, &fp->fi_delegations);
2855 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2857 struct nfs4_file *fp = dp->dl_file;
2860 return nfs4_setlease(dp, flag);
2861 spin_lock(&recall_lock);
2862 if (fp->fi_had_conflict) {
2863 spin_unlock(&recall_lock);
2866 atomic_inc(&fp->fi_delegees);
2867 list_add(&dp->dl_perfile, &fp->fi_delegations);
2868 spin_unlock(&recall_lock);
2869 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2873 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2875 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2876 if (status == -EAGAIN)
2877 open->op_why_no_deleg = WND4_CONTENTION;
2879 open->op_why_no_deleg = WND4_RESOURCE;
2880 switch (open->op_deleg_want) {
2881 case NFS4_SHARE_WANT_READ_DELEG:
2882 case NFS4_SHARE_WANT_WRITE_DELEG:
2883 case NFS4_SHARE_WANT_ANY_DELEG:
2885 case NFS4_SHARE_WANT_CANCEL:
2886 open->op_why_no_deleg = WND4_CANCELLED;
2888 case NFS4_SHARE_WANT_NO_DELEG:
2889 BUG(); /* not supposed to get here */
2895 * Attempt to hand out a delegation.
2898 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2899 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2901 struct nfs4_delegation *dp;
2902 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2904 int status = 0, flag = 0;
2906 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2907 flag = NFS4_OPEN_DELEGATE_NONE;
2908 open->op_recall = 0;
2909 switch (open->op_claim_type) {
2910 case NFS4_OPEN_CLAIM_PREVIOUS:
2912 open->op_recall = 1;
2913 flag = open->op_delegate_type;
2914 if (flag == NFS4_OPEN_DELEGATE_NONE)
2917 case NFS4_OPEN_CLAIM_NULL:
2918 /* Let's not give out any delegations till everyone's
2919 * had the chance to reclaim theirs.... */
2920 if (locks_in_grace(net))
2922 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2924 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2925 flag = NFS4_OPEN_DELEGATE_WRITE;
2927 flag = NFS4_OPEN_DELEGATE_READ;
2933 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2936 status = nfs4_set_delegation(dp, flag);
2940 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2942 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2943 STATEID_VAL(&dp->dl_stid.sc_stateid));
2945 open->op_delegate_type = flag;
2946 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2947 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2948 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2949 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2951 /* 4.1 client asking for a delegation? */
2952 if (open->op_deleg_want)
2953 nfsd4_open_deleg_none_ext(open, status);
2957 nfs4_put_delegation(dp);
2959 flag = NFS4_OPEN_DELEGATE_NONE;
2963 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2964 struct nfs4_delegation *dp)
2966 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2967 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2968 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2969 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2970 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2971 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2972 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2973 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
2975 /* Otherwise the client must be confused wanting a delegation
2976 * it already has, therefore we don't return
2977 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
2982 * called with nfs4_lock_state() held.
2985 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2987 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2988 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
2989 struct nfs4_file *fp = NULL;
2990 struct inode *ino = current_fh->fh_dentry->d_inode;
2991 struct nfs4_ol_stateid *stp = NULL;
2992 struct nfs4_delegation *dp = NULL;
2996 * Lookup file; if found, lookup stateid and check open request,
2997 * and check for delegations in the process of being recalled.
2998 * If not found, create the nfs4_file struct
3000 fp = find_file(ino);
3002 if ((status = nfs4_check_open(fp, open, &stp)))
3004 status = nfs4_check_deleg(cl, fp, open, &dp);
3008 status = nfserr_bad_stateid;
3009 if (nfsd4_is_deleg_cur(open))
3011 status = nfserr_jukebox;
3013 open->op_file = NULL;
3014 nfsd4_init_file(fp, ino);
3018 * OPEN the file, or upgrade an existing OPEN.
3019 * If truncate fails, the OPEN fails.
3022 /* Stateid was found, this is an OPEN upgrade */
3023 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3027 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3030 status = nfsd4_truncate(rqstp, current_fh, open);
3034 open->op_stp = NULL;
3035 init_open_stateid(stp, fp, open);
3037 update_stateid(&stp->st_stid.sc_stateid);
3038 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3040 if (nfsd4_has_session(&resp->cstate)) {
3041 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3043 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3044 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3045 open->op_why_no_deleg = WND4_NOT_WANTED;
3051 * Attempt to hand out a delegation. No error return, because the
3052 * OPEN succeeds even if we fail.
3054 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3058 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3059 STATEID_VAL(&stp->st_stid.sc_stateid));
3061 /* 4.1 client trying to upgrade/downgrade delegation? */
3062 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3063 open->op_deleg_want)
3064 nfsd4_deleg_xgrade_none_ext(open, dp);
3068 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3069 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3071 * To finish the open response, we just need to set the rflags.
3073 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3074 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3075 !nfsd4_has_session(&resp->cstate))
3076 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3081 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3083 if (open->op_openowner) {
3084 struct nfs4_openowner *oo = open->op_openowner;
3086 if (!list_empty(&oo->oo_owner.so_stateids))
3087 list_del_init(&oo->oo_close_lru);
3088 if (oo->oo_flags & NFS4_OO_NEW) {
3090 release_openowner(oo);
3091 open->op_openowner = NULL;
3093 oo->oo_flags &= ~NFS4_OO_NEW;
3097 nfsd4_free_file(open->op_file);
3099 free_generic_stateid(open->op_stp);
3103 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3106 struct nfs4_client *clp;
3108 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3111 dprintk("process_renew(%08x/%08x): starting\n",
3112 clid->cl_boot, clid->cl_id);
3113 status = nfserr_stale_clientid;
3114 if (STALE_CLIENTID(clid, nn))
3116 clp = find_confirmed_client(clid, cstate->minorversion);
3117 status = nfserr_expired;
3119 /* We assume the client took too long to RENEW. */
3120 dprintk("nfsd4_renew: clientid not found!\n");
3123 status = nfserr_cb_path_down;
3124 if (!list_empty(&clp->cl_delegations)
3125 && clp->cl_cb_state != NFSD4_CB_UP)
3129 nfs4_unlock_state();
3134 nfsd4_end_grace(struct net *net)
3136 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3138 /* do nothing if grace period already ended */
3139 if (nn->grace_ended)
3142 dprintk("NFSD: end of grace period\n");
3143 nn->grace_ended = true;
3144 nfsd4_record_grace_done(net, nn->boot_time);
3145 locks_end_grace(&nn->nfsd4_manager);
3147 * Now that every NFSv4 client has had the chance to recover and
3148 * to see the (possibly new, possibly shorter) lease time, we
3149 * can safely set the next grace time to the current lease time:
3151 nfsd4_grace = nfsd4_lease;
3155 nfs4_laundromat(void)
3157 struct nfs4_client *clp;
3158 struct nfs4_openowner *oo;
3159 struct nfs4_delegation *dp;
3160 struct list_head *pos, *next, reaplist;
3161 time_t cutoff = get_seconds() - nfsd4_lease;
3162 time_t t, clientid_val = nfsd4_lease;
3163 time_t u, test_val = nfsd4_lease;
3167 dprintk("NFSD: laundromat service - starting\n");
3168 nfsd4_end_grace(&init_net);
3169 INIT_LIST_HEAD(&reaplist);
3170 spin_lock(&client_lock);
3171 list_for_each_safe(pos, next, &client_lru) {
3172 clp = list_entry(pos, struct nfs4_client, cl_lru);
3173 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3174 t = clp->cl_time - cutoff;
3175 if (clientid_val > t)
3179 if (atomic_read(&clp->cl_refcount)) {
3180 dprintk("NFSD: client in use (clientid %08x)\n",
3181 clp->cl_clientid.cl_id);
3184 unhash_client_locked(clp);
3185 list_add(&clp->cl_lru, &reaplist);
3187 spin_unlock(&client_lock);
3188 list_for_each_safe(pos, next, &reaplist) {
3189 clp = list_entry(pos, struct nfs4_client, cl_lru);
3190 dprintk("NFSD: purging unused client (clientid %08x)\n",
3191 clp->cl_clientid.cl_id);
3192 nfsd4_client_record_remove(clp);
3195 spin_lock(&recall_lock);
3196 list_for_each_safe(pos, next, &del_recall_lru) {
3197 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3198 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3199 u = dp->dl_time - cutoff;
3204 list_move(&dp->dl_recall_lru, &reaplist);
3206 spin_unlock(&recall_lock);
3207 list_for_each_safe(pos, next, &reaplist) {
3208 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3209 unhash_delegation(dp);
3211 test_val = nfsd4_lease;
3212 list_for_each_safe(pos, next, &close_lru) {
3213 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3214 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3215 u = oo->oo_time - cutoff;
3220 release_openowner(oo);
3222 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3223 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3224 nfs4_unlock_state();
3225 return clientid_val;
3228 static struct workqueue_struct *laundry_wq;
3229 static void laundromat_main(struct work_struct *);
3230 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3233 laundromat_main(struct work_struct *not_used)
3237 t = nfs4_laundromat();
3238 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3239 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3242 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3244 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3245 return nfserr_bad_stateid;
3250 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3252 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3254 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3255 STATEID_VAL(stateid));
3260 access_permit_read(struct nfs4_ol_stateid *stp)
3262 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3263 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3264 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3268 access_permit_write(struct nfs4_ol_stateid *stp)
3270 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3271 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3275 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3277 __be32 status = nfserr_openmode;
3279 /* For lock stateid's, we test the parent open, not the lock: */
3280 if (stp->st_openstp)
3281 stp = stp->st_openstp;
3282 if ((flags & WR_STATE) && !access_permit_write(stp))
3284 if ((flags & RD_STATE) && !access_permit_read(stp))
3291 static inline __be32
3292 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3294 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3296 else if (locks_in_grace(net)) {
3297 /* Answer in remaining cases depends on existence of
3298 * conflicting state; so we must wait out the grace period. */
3299 return nfserr_grace;
3300 } else if (flags & WR_STATE)
3301 return nfs4_share_conflict(current_fh,
3302 NFS4_SHARE_DENY_WRITE);
3303 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3304 return nfs4_share_conflict(current_fh,
3305 NFS4_SHARE_DENY_READ);
3309 * Allow READ/WRITE during grace period on recovered state only for files
3310 * that are not able to provide mandatory locking.
3313 grace_disallows_io(struct net *net, struct inode *inode)
3315 return locks_in_grace(net) && mandatory_lock(inode);
3318 /* Returns true iff a is later than b: */
3319 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3321 return (s32)a->si_generation - (s32)b->si_generation > 0;
3324 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3327 * When sessions are used the stateid generation number is ignored
3330 if (has_session && in->si_generation == 0)
3333 if (in->si_generation == ref->si_generation)
3336 /* If the client sends us a stateid from the future, it's buggy: */
3337 if (stateid_generation_after(in, ref))
3338 return nfserr_bad_stateid;
3340 * However, we could see a stateid from the past, even from a
3341 * non-buggy client. For example, if the client sends a lock
3342 * while some IO is outstanding, the lock may bump si_generation
3343 * while the IO is still in flight. The client could avoid that
3344 * situation by waiting for responses on all the IO requests,
3345 * but better performance may result in retrying IO that
3346 * receives an old_stateid error if requests are rarely
3347 * reordered in flight:
3349 return nfserr_old_stateid;
3352 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3354 struct nfs4_stid *s;
3355 struct nfs4_ol_stateid *ols;
3358 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3359 return nfserr_bad_stateid;
3360 /* Client debugging aid. */
3361 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3362 char addr_str[INET6_ADDRSTRLEN];
3363 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3365 pr_warn_ratelimited("NFSD: client %s testing state ID "
3366 "with incorrect client ID\n", addr_str);
3367 return nfserr_bad_stateid;
3369 s = find_stateid(cl, stateid);
3371 return nfserr_bad_stateid;
3372 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3375 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3377 ols = openlockstateid(s);
3378 if (ols->st_stateowner->so_is_open_owner
3379 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3380 return nfserr_bad_stateid;
3384 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s, bool sessions)
3386 struct nfs4_client *cl;
3387 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3389 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3390 return nfserr_bad_stateid;
3391 if (STALE_STATEID(stateid, nn))
3392 return nfserr_stale_stateid;
3393 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions);
3395 return nfserr_expired;
3396 *s = find_stateid_by_type(cl, stateid, typemask);
3398 return nfserr_bad_stateid;
3404 * Checks for stateid operations
3407 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3408 stateid_t *stateid, int flags, struct file **filpp)
3410 struct nfs4_stid *s;
3411 struct nfs4_ol_stateid *stp = NULL;
3412 struct nfs4_delegation *dp = NULL;
3413 struct svc_fh *current_fh = &cstate->current_fh;
3414 struct inode *ino = current_fh->fh_dentry->d_inode;
3420 if (grace_disallows_io(net, ino))
3421 return nfserr_grace;
3423 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3424 return check_special_stateids(net, current_fh, stateid, flags);
3426 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s, cstate->minorversion);
3429 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3432 switch (s->sc_type) {
3433 case NFS4_DELEG_STID:
3434 dp = delegstateid(s);
3435 status = nfs4_check_delegmode(dp, flags);
3439 *filpp = dp->dl_file->fi_deleg_file;
3443 case NFS4_OPEN_STID:
3444 case NFS4_LOCK_STID:
3445 stp = openlockstateid(s);
3446 status = nfs4_check_fh(current_fh, stp);
3449 if (stp->st_stateowner->so_is_open_owner
3450 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3452 status = nfs4_check_openmode(stp, flags);
3456 if (flags & RD_STATE)
3457 *filpp = find_readable_file(stp->st_file);
3459 *filpp = find_writeable_file(stp->st_file);
3463 return nfserr_bad_stateid;
3471 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3473 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3474 return nfserr_locks_held;
3475 release_lock_stateid(stp);
3480 * Test if the stateid is valid
3483 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3484 struct nfsd4_test_stateid *test_stateid)
3486 struct nfsd4_test_stateid_id *stateid;
3487 struct nfs4_client *cl = cstate->session->se_client;
3490 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3491 stateid->ts_id_status =
3492 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3493 nfs4_unlock_state();
3499 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3500 struct nfsd4_free_stateid *free_stateid)
3502 stateid_t *stateid = &free_stateid->fr_stateid;
3503 struct nfs4_stid *s;
3504 struct nfs4_client *cl = cstate->session->se_client;
3505 __be32 ret = nfserr_bad_stateid;
3508 s = find_stateid(cl, stateid);
3511 switch (s->sc_type) {
3512 case NFS4_DELEG_STID:
3513 ret = nfserr_locks_held;
3515 case NFS4_OPEN_STID:
3516 case NFS4_LOCK_STID:
3517 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3520 if (s->sc_type == NFS4_LOCK_STID)
3521 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3523 ret = nfserr_locks_held;
3526 ret = nfserr_bad_stateid;
3529 nfs4_unlock_state();
3536 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3537 RD_STATE : WR_STATE;
3540 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3542 struct svc_fh *current_fh = &cstate->current_fh;
3543 struct nfs4_stateowner *sop = stp->st_stateowner;
3546 status = nfsd4_check_seqid(cstate, sop, seqid);
3549 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3551 * "Closed" stateid's exist *only* to return
3552 * nfserr_replay_me from the previous step.
3554 return nfserr_bad_stateid;
3555 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3558 return nfs4_check_fh(current_fh, stp);
3562 * Checks for sequence id mutating operations.
3565 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3566 stateid_t *stateid, char typemask,
3567 struct nfs4_ol_stateid **stpp)
3570 struct nfs4_stid *s;
3572 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3573 seqid, STATEID_VAL(stateid));
3576 status = nfsd4_lookup_stateid(stateid, typemask, &s, cstate->minorversion);
3579 *stpp = openlockstateid(s);
3580 cstate->replay_owner = (*stpp)->st_stateowner;
3582 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3585 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3588 struct nfs4_openowner *oo;
3590 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3591 NFS4_OPEN_STID, stpp);
3594 oo = openowner((*stpp)->st_stateowner);
3595 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3596 return nfserr_bad_stateid;
3601 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3602 struct nfsd4_open_confirm *oc)
3605 struct nfs4_openowner *oo;
3606 struct nfs4_ol_stateid *stp;
3608 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3609 (int)cstate->current_fh.fh_dentry->d_name.len,
3610 cstate->current_fh.fh_dentry->d_name.name);
3612 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3618 status = nfs4_preprocess_seqid_op(cstate,
3619 oc->oc_seqid, &oc->oc_req_stateid,
3620 NFS4_OPEN_STID, &stp);
3623 oo = openowner(stp->st_stateowner);
3624 status = nfserr_bad_stateid;
3625 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3627 oo->oo_flags |= NFS4_OO_CONFIRMED;
3628 update_stateid(&stp->st_stid.sc_stateid);
3629 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3630 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3631 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3633 nfsd4_client_record_create(oo->oo_owner.so_client);
3636 if (!cstate->replay_owner)
3637 nfs4_unlock_state();
3641 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3643 if (!test_access(access, stp))
3645 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3646 clear_access(access, stp);
3649 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3651 switch (to_access) {
3652 case NFS4_SHARE_ACCESS_READ:
3653 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3654 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3656 case NFS4_SHARE_ACCESS_WRITE:
3657 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3658 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3660 case NFS4_SHARE_ACCESS_BOTH:
3668 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3671 for (i = 0; i < 4; i++) {
3672 if ((i & deny) != i)
3678 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3679 struct nfsd4_compound_state *cstate,
3680 struct nfsd4_open_downgrade *od)
3683 struct nfs4_ol_stateid *stp;
3685 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3686 (int)cstate->current_fh.fh_dentry->d_name.len,
3687 cstate->current_fh.fh_dentry->d_name.name);
3689 /* We don't yet support WANT bits: */
3690 if (od->od_deleg_want)
3691 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3695 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3696 &od->od_stateid, &stp);
3699 status = nfserr_inval;
3700 if (!test_access(od->od_share_access, stp)) {
3701 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3702 stp->st_access_bmap, od->od_share_access);
3705 if (!test_deny(od->od_share_deny, stp)) {
3706 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3707 stp->st_deny_bmap, od->od_share_deny);
3710 nfs4_stateid_downgrade(stp, od->od_share_access);
3712 reset_union_bmap_deny(od->od_share_deny, stp);
3714 update_stateid(&stp->st_stid.sc_stateid);
3715 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3718 if (!cstate->replay_owner)
3719 nfs4_unlock_state();
3723 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3725 struct nfs4_openowner *oo;
3726 struct nfs4_ol_stateid *s;
3728 if (!so->so_is_open_owner)
3731 s = oo->oo_last_closed_stid;
3734 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3735 /* Release the last_closed_stid on the next seqid bump: */
3736 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3739 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3740 release_last_closed_stateid(oo);
3743 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3745 unhash_open_stateid(s);
3746 s->st_stid.sc_type = NFS4_CLOSED_STID;
3750 * nfs4_unlock_state() called after encode
3753 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3754 struct nfsd4_close *close)
3757 struct nfs4_openowner *oo;
3758 struct nfs4_ol_stateid *stp;
3760 dprintk("NFSD: nfsd4_close on file %.*s\n",
3761 (int)cstate->current_fh.fh_dentry->d_name.len,
3762 cstate->current_fh.fh_dentry->d_name.name);
3765 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3767 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3771 oo = openowner(stp->st_stateowner);
3773 update_stateid(&stp->st_stid.sc_stateid);
3774 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3776 nfsd4_close_open_stateid(stp);
3777 release_last_closed_stateid(oo);
3778 oo->oo_last_closed_stid = stp;
3780 if (list_empty(&oo->oo_owner.so_stateids)) {
3781 if (cstate->minorversion) {
3782 release_openowner(oo);
3783 cstate->replay_owner = NULL;
3786 * In the 4.0 case we need to keep the owners around a
3787 * little while to handle CLOSE replay.
3789 if (list_empty(&oo->oo_owner.so_stateids))
3790 move_to_close_lru(oo);
3794 if (!cstate->replay_owner)
3795 nfs4_unlock_state();
3800 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3801 struct nfsd4_delegreturn *dr)
3803 struct nfs4_delegation *dp;
3804 stateid_t *stateid = &dr->dr_stateid;
3805 struct nfs4_stid *s;
3806 struct inode *inode;
3809 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3811 inode = cstate->current_fh.fh_dentry->d_inode;
3814 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s, cstate->minorversion);
3817 dp = delegstateid(s);
3818 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3822 unhash_delegation(dp);
3824 nfs4_unlock_state();
3830 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3832 #define LOCKOWNER_INO_HASH_BITS 8
3833 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3834 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3837 end_offset(u64 start, u64 len)
3842 return end >= start ? end: NFS4_MAX_UINT64;
3845 /* last octet in a range */
3847 last_byte_offset(u64 start, u64 len)
3853 return end > start ? end - 1: NFS4_MAX_UINT64;
3856 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3858 return (file_hashval(inode) + cl_id
3859 + opaque_hashval(ownername->data, ownername->len))
3860 & LOCKOWNER_INO_HASH_MASK;
3863 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3866 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3867 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3868 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3869 * locking, this prevents us from being completely protocol-compliant. The
3870 * real solution to this problem is to start using unsigned file offsets in
3871 * the VFS, but this is a very deep change!
3874 nfs4_transform_lock_offset(struct file_lock *lock)
3876 if (lock->fl_start < 0)
3877 lock->fl_start = OFFSET_MAX;
3878 if (lock->fl_end < 0)
3879 lock->fl_end = OFFSET_MAX;
3882 /* Hack!: For now, we're defining this just so we can use a pointer to it
3883 * as a unique cookie to identify our (NFSv4's) posix locks. */
3884 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3888 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3890 struct nfs4_lockowner *lo;
3892 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3893 lo = (struct nfs4_lockowner *) fl->fl_owner;
3894 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3895 lo->lo_owner.so_owner.len, GFP_KERNEL);
3896 if (!deny->ld_owner.data)
3897 /* We just don't care that much */
3899 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3900 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3903 deny->ld_owner.len = 0;
3904 deny->ld_owner.data = NULL;
3905 deny->ld_clientid.cl_boot = 0;
3906 deny->ld_clientid.cl_id = 0;
3908 deny->ld_start = fl->fl_start;
3909 deny->ld_length = NFS4_MAX_UINT64;
3910 if (fl->fl_end != NFS4_MAX_UINT64)
3911 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3912 deny->ld_type = NFS4_READ_LT;
3913 if (fl->fl_type != F_RDLCK)
3914 deny->ld_type = NFS4_WRITE_LT;
3917 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3919 struct nfs4_ol_stateid *lst;
3921 if (!same_owner_str(&lo->lo_owner, owner, clid))
3923 lst = list_first_entry(&lo->lo_owner.so_stateids,
3924 struct nfs4_ol_stateid, st_perstateowner);
3925 return lst->st_file->fi_inode == inode;
3928 static struct nfs4_lockowner *
3929 find_lockowner_str(struct inode *inode, clientid_t *clid,
3930 struct xdr_netobj *owner)
3932 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3933 struct nfs4_lockowner *lo;
3935 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3936 if (same_lockowner_ino(lo, inode, clid, owner))
3942 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3944 struct inode *inode = open_stp->st_file->fi_inode;
3945 unsigned int inohash = lockowner_ino_hashval(inode,
3946 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3948 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3949 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3950 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3954 * Alloc a lock owner structure.
3955 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3958 * strhashval = ownerstr_hashval
3961 static struct nfs4_lockowner *
3962 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3963 struct nfs4_lockowner *lo;
3965 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3968 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3969 lo->lo_owner.so_is_open_owner = 0;
3970 /* It is the openowner seqid that will be incremented in encode in the
3971 * case of new lockowners; so increment the lock seqid manually: */
3972 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3973 hash_lockowner(lo, strhashval, clp, open_stp);
3977 static struct nfs4_ol_stateid *
3978 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3980 struct nfs4_ol_stateid *stp;
3981 struct nfs4_client *clp = lo->lo_owner.so_client;
3983 stp = nfs4_alloc_stateid(clp);
3986 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3987 list_add(&stp->st_perfile, &fp->fi_stateids);
3988 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3989 stp->st_stateowner = &lo->lo_owner;
3992 stp->st_access_bmap = 0;
3993 stp->st_deny_bmap = open_stp->st_deny_bmap;
3994 stp->st_openstp = open_stp;
3999 check_lock_length(u64 offset, u64 length)
4001 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4002 LOFF_OVERFLOW(offset, length)));
4005 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4007 struct nfs4_file *fp = lock_stp->st_file;
4008 int oflag = nfs4_access_to_omode(access);
4010 if (test_access(access, lock_stp))
4012 nfs4_file_get_access(fp, oflag);
4013 set_access(access, lock_stp);
4016 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)
4018 struct nfs4_file *fi = ost->st_file;
4019 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4020 struct nfs4_client *cl = oo->oo_owner.so_client;
4021 struct nfs4_lockowner *lo;
4022 unsigned int strhashval;
4024 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4026 if (!cstate->minorversion)
4027 return nfserr_bad_seqid;
4028 /* XXX: a lockowner always has exactly one stateid: */
4029 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4030 struct nfs4_ol_stateid, st_perstateowner);
4033 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4034 &lock->v.new.owner);
4035 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4037 return nfserr_jukebox;
4038 *lst = alloc_init_lock_stateid(lo, fi, ost);
4040 release_lockowner(lo);
4041 return nfserr_jukebox;
4051 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4052 struct nfsd4_lock *lock)
4054 struct nfs4_openowner *open_sop = NULL;
4055 struct nfs4_lockowner *lock_sop = NULL;
4056 struct nfs4_ol_stateid *lock_stp;
4057 struct file *filp = NULL;
4058 struct file_lock *file_lock = NULL;
4059 struct file_lock *conflock = NULL;
4061 bool new_state = false;
4064 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4066 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4067 (long long) lock->lk_offset,
4068 (long long) lock->lk_length);
4070 if (check_lock_length(lock->lk_offset, lock->lk_length))
4071 return nfserr_inval;
4073 if ((status = fh_verify(rqstp, &cstate->current_fh,
4074 S_IFREG, NFSD_MAY_LOCK))) {
4075 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4081 if (lock->lk_is_new) {
4082 struct nfs4_ol_stateid *open_stp = NULL;
4084 if (nfsd4_has_session(cstate))
4085 /* See rfc 5661 18.10.3: given clientid is ignored: */
4086 memcpy(&lock->v.new.clientid,
4087 &cstate->session->se_client->cl_clientid,
4088 sizeof(clientid_t));
4090 status = nfserr_stale_clientid;
4091 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4094 /* validate and update open stateid and open seqid */
4095 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4096 lock->lk_new_open_seqid,
4097 &lock->lk_new_open_stateid,
4101 open_sop = openowner(open_stp->st_stateowner);
4102 status = nfserr_bad_stateid;
4103 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4104 &lock->v.new.clientid))
4106 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4107 &lock_stp, &new_state);
4109 status = nfs4_preprocess_seqid_op(cstate,
4110 lock->lk_old_lock_seqid,
4111 &lock->lk_old_lock_stateid,
4112 NFS4_LOCK_STID, &lock_stp);
4115 lock_sop = lockowner(lock_stp->st_stateowner);
4117 lkflg = setlkflg(lock->lk_type);
4118 status = nfs4_check_openmode(lock_stp, lkflg);
4122 status = nfserr_grace;
4123 if (locks_in_grace(SVC_NET(rqstp)) && !lock->lk_reclaim)
4125 status = nfserr_no_grace;
4126 if (!locks_in_grace(SVC_NET(rqstp)) && lock->lk_reclaim)
4129 file_lock = locks_alloc_lock();
4131 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4132 status = nfserr_jukebox;
4136 locks_init_lock(file_lock);
4137 switch (lock->lk_type) {
4140 filp = find_readable_file(lock_stp->st_file);
4142 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4143 file_lock->fl_type = F_RDLCK;
4146 case NFS4_WRITEW_LT:
4147 filp = find_writeable_file(lock_stp->st_file);
4149 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4150 file_lock->fl_type = F_WRLCK;
4153 status = nfserr_inval;
4157 status = nfserr_openmode;
4160 file_lock->fl_owner = (fl_owner_t)lock_sop;
4161 file_lock->fl_pid = current->tgid;
4162 file_lock->fl_file = filp;
4163 file_lock->fl_flags = FL_POSIX;
4164 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4165 file_lock->fl_start = lock->lk_offset;
4166 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4167 nfs4_transform_lock_offset(file_lock);
4169 conflock = locks_alloc_lock();
4171 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4172 status = nfserr_jukebox;
4176 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4178 case 0: /* success! */
4179 update_stateid(&lock_stp->st_stid.sc_stateid);
4180 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4184 case (EAGAIN): /* conflock holds conflicting lock */
4185 status = nfserr_denied;
4186 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4187 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4190 status = nfserr_deadlock;
4193 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4194 status = nfserrno(err);
4198 if (status && new_state)
4199 release_lockowner(lock_sop);
4200 if (!cstate->replay_owner)
4201 nfs4_unlock_state();
4203 locks_free_lock(file_lock);
4205 locks_free_lock(conflock);
4210 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4211 * so we do a temporary open here just to get an open file to pass to
4212 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4215 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4218 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4220 err = nfserrno(vfs_test_lock(file, lock));
4230 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4231 struct nfsd4_lockt *lockt)
4233 struct inode *inode;
4234 struct file_lock *file_lock = NULL;
4235 struct nfs4_lockowner *lo;
4237 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4239 if (locks_in_grace(SVC_NET(rqstp)))
4240 return nfserr_grace;
4242 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4243 return nfserr_inval;
4247 status = nfserr_stale_clientid;
4248 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid, nn))
4251 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4254 inode = cstate->current_fh.fh_dentry->d_inode;
4255 file_lock = locks_alloc_lock();
4257 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4258 status = nfserr_jukebox;
4261 locks_init_lock(file_lock);
4262 switch (lockt->lt_type) {
4265 file_lock->fl_type = F_RDLCK;
4268 case NFS4_WRITEW_LT:
4269 file_lock->fl_type = F_WRLCK;
4272 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4273 status = nfserr_inval;
4277 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4279 file_lock->fl_owner = (fl_owner_t)lo;
4280 file_lock->fl_pid = current->tgid;
4281 file_lock->fl_flags = FL_POSIX;
4283 file_lock->fl_start = lockt->lt_offset;
4284 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4286 nfs4_transform_lock_offset(file_lock);
4288 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4292 if (file_lock->fl_type != F_UNLCK) {
4293 status = nfserr_denied;
4294 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4297 nfs4_unlock_state();
4299 locks_free_lock(file_lock);
4304 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4305 struct nfsd4_locku *locku)
4307 struct nfs4_ol_stateid *stp;
4308 struct file *filp = NULL;
4309 struct file_lock *file_lock = NULL;
4313 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4314 (long long) locku->lu_offset,
4315 (long long) locku->lu_length);
4317 if (check_lock_length(locku->lu_offset, locku->lu_length))
4318 return nfserr_inval;
4322 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4323 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4326 filp = find_any_file(stp->st_file);
4328 status = nfserr_lock_range;
4331 file_lock = locks_alloc_lock();
4333 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4334 status = nfserr_jukebox;
4337 locks_init_lock(file_lock);
4338 file_lock->fl_type = F_UNLCK;
4339 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4340 file_lock->fl_pid = current->tgid;
4341 file_lock->fl_file = filp;
4342 file_lock->fl_flags = FL_POSIX;
4343 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4344 file_lock->fl_start = locku->lu_offset;
4346 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4348 nfs4_transform_lock_offset(file_lock);
4351 * Try to unlock the file in the VFS.
4353 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4355 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4359 * OK, unlock succeeded; the only thing left to do is update the stateid.
4361 update_stateid(&stp->st_stid.sc_stateid);
4362 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4365 if (!cstate->replay_owner)
4366 nfs4_unlock_state();
4368 locks_free_lock(file_lock);
4372 status = nfserrno(err);
4378 * 1: locks held by lockowner
4379 * 0: no locks held by lockowner
4382 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4384 struct file_lock **flpp;
4385 struct inode *inode = filp->fi_inode;
4389 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4390 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4401 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4402 struct nfsd4_compound_state *cstate,
4403 struct nfsd4_release_lockowner *rlockowner)
4405 clientid_t *clid = &rlockowner->rl_clientid;
4406 struct nfs4_stateowner *sop;
4407 struct nfs4_lockowner *lo;
4408 struct nfs4_ol_stateid *stp;
4409 struct xdr_netobj *owner = &rlockowner->rl_owner;
4410 struct list_head matches;
4411 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4413 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4415 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4416 clid->cl_boot, clid->cl_id);
4418 /* XXX check for lease expiration */
4420 status = nfserr_stale_clientid;
4421 if (STALE_CLIENTID(clid, nn))
4426 status = nfserr_locks_held;
4427 INIT_LIST_HEAD(&matches);
4429 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4430 if (sop->so_is_open_owner)
4432 if (!same_owner_str(sop, owner, clid))
4434 list_for_each_entry(stp, &sop->so_stateids,
4436 lo = lockowner(sop);
4437 if (check_for_locks(stp->st_file, lo))
4439 list_add(&lo->lo_list, &matches);
4442 /* Clients probably won't expect us to return with some (but not all)
4443 * of the lockowner state released; so don't release any until all
4444 * have been checked. */
4446 while (!list_empty(&matches)) {
4447 lo = list_entry(matches.next, struct nfs4_lockowner,
4449 /* unhash_stateowner deletes so_perclient only
4450 * for openowners. */
4451 list_del(&lo->lo_list);
4452 release_lockowner(lo);
4455 nfs4_unlock_state();
4459 static inline struct nfs4_client_reclaim *
4462 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4466 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4468 unsigned int strhashval = clientstr_hashval(name);
4469 struct nfs4_client *clp;
4471 clp = find_confirmed_client_by_str(name, strhashval);
4474 return test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
4478 * failure => all reset bets are off, nfserr_no_grace...
4481 nfs4_client_to_reclaim(const char *name)
4483 unsigned int strhashval;
4484 struct nfs4_client_reclaim *crp = NULL;
4486 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4487 crp = alloc_reclaim();
4490 strhashval = clientstr_hashval(name);
4491 INIT_LIST_HEAD(&crp->cr_strhash);
4492 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4493 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4494 reclaim_str_hashtbl_size++;
4499 nfs4_release_reclaim(void)
4501 struct nfs4_client_reclaim *crp = NULL;
4504 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4505 while (!list_empty(&reclaim_str_hashtbl[i])) {
4506 crp = list_entry(reclaim_str_hashtbl[i].next,
4507 struct nfs4_client_reclaim, cr_strhash);
4508 list_del(&crp->cr_strhash);
4510 reclaim_str_hashtbl_size--;
4513 BUG_ON(reclaim_str_hashtbl_size);
4517 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4518 struct nfs4_client_reclaim *
4519 nfsd4_find_reclaim_client(struct nfs4_client *clp)
4521 unsigned int strhashval;
4522 struct nfs4_client_reclaim *crp = NULL;
4524 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4525 clp->cl_name.len, clp->cl_name.data,
4528 /* find clp->cl_name in reclaim_str_hashtbl */
4529 strhashval = clientstr_hashval(clp->cl_recdir);
4530 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4531 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4539 * Called from OPEN. Look for clientid in reclaim list.
4542 nfs4_check_open_reclaim(clientid_t *clid, bool sessions)
4544 struct nfs4_client *clp;
4546 /* find clientid in conf_id_hashtbl */
4547 clp = find_confirmed_client(clid, sessions);
4549 return nfserr_reclaim_bad;
4551 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4554 #ifdef CONFIG_NFSD_FAULT_INJECTION
4556 void nfsd_forget_clients(u64 num)
4558 struct nfs4_client *clp, *next;
4562 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4563 nfsd4_client_record_remove(clp);
4568 nfs4_unlock_state();
4570 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4573 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4575 release_lockowner(lockowner(sop));
4578 static void release_openowner_sop(struct nfs4_stateowner *sop)
4580 release_openowner(openowner(sop));
4583 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4584 void (*release_sop)(struct nfs4_stateowner *))
4587 struct nfs4_stateowner *sop, *next;
4589 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4590 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4591 if (sop->so_is_open_owner != is_open_owner)
4601 void nfsd_forget_locks(u64 num)
4606 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4607 nfs4_unlock_state();
4609 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4612 void nfsd_forget_openowners(u64 num)
4617 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4618 nfs4_unlock_state();
4620 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4623 static int nfsd_process_n_delegations(u64 num, struct list_head *list)
4626 struct nfs4_file *fp, *fnext;
4627 struct nfs4_delegation *dp, *dnext;
4629 for (i = 0; i < FILE_HASH_SIZE; i++) {
4630 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4631 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4632 list_move(&dp->dl_recall_lru, list);
4642 void nfsd_forget_delegations(u64 num)
4646 struct nfs4_delegation *dp, *dnext;
4648 spin_lock(&recall_lock);
4649 count = nfsd_process_n_delegations(num, &victims);
4650 spin_unlock(&recall_lock);
4653 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
4654 unhash_delegation(dp);
4655 nfs4_unlock_state();
4657 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4660 void nfsd_recall_delegations(u64 num)
4664 struct nfs4_delegation *dp, *dnext;
4666 spin_lock(&recall_lock);
4667 count = nfsd_process_n_delegations(num, &victims);
4668 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
4669 list_del(&dp->dl_recall_lru);
4670 nfsd_break_one_deleg(dp);
4672 spin_unlock(&recall_lock);
4674 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4677 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4679 /* initialization to perform at module load time: */
4682 nfs4_state_init(void)
4686 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4687 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4688 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4689 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4690 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4691 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4693 for (i = 0; i < SESSION_HASH_SIZE; i++)
4694 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4695 for (i = 0; i < FILE_HASH_SIZE; i++) {
4696 INIT_LIST_HEAD(&file_hashtbl[i]);
4698 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4699 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4701 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4702 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4703 INIT_LIST_HEAD(&close_lru);
4704 INIT_LIST_HEAD(&client_lru);
4705 INIT_LIST_HEAD(&del_recall_lru);
4706 reclaim_str_hashtbl_size = 0;
4710 * Since the lifetime of a delegation isn't limited to that of an open, a
4711 * client may quite reasonably hang on to a delegation as long as it has
4712 * the inode cached. This becomes an obvious problem the first time a
4713 * client's inode cache approaches the size of the server's total memory.
4715 * For now we avoid this problem by imposing a hard limit on the number
4716 * of delegations, which varies according to the server's memory size.
4719 set_max_delegations(void)
4722 * Allow at most 4 delegations per megabyte of RAM. Quick
4723 * estimates suggest that in the worst case (where every delegation
4724 * is for a different inode), a delegation could take about 1.5K,
4725 * giving a worst case usage of about 6% of memory.
4727 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4730 /* initialization to perform when the nfsd service is started: */
4733 nfs4_state_start(void)
4735 struct net *net = &init_net;
4736 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4740 * FIXME: For now, we hang most of the pernet global stuff off of
4741 * init_net until nfsd is fully containerized. Eventually, we'll
4742 * need to pass a net pointer into this function, take a reference
4743 * to that instead and then do most of the rest of this on a per-net
4747 nfsd4_client_tracking_init(net);
4748 nn->boot_time = get_seconds();
4749 locks_start_grace(net, &nn->nfsd4_manager);
4750 nn->grace_ended = false;
4751 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4753 ret = set_callback_cred();
4758 laundry_wq = create_singlethread_workqueue("nfsd4");
4759 if (laundry_wq == NULL) {
4763 ret = nfsd4_create_callback_queue();
4765 goto out_free_laundry;
4766 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4767 set_max_delegations();
4770 destroy_workqueue(laundry_wq);
4772 nfsd4_client_tracking_exit(net);
4778 __nfs4_state_shutdown(void)
4781 struct nfs4_client *clp = NULL;
4782 struct nfs4_delegation *dp = NULL;
4783 struct list_head *pos, *next, reaplist;
4785 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4786 while (!list_empty(&conf_id_hashtbl[i])) {
4787 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4790 while (!list_empty(&unconf_str_hashtbl[i])) {
4791 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4795 INIT_LIST_HEAD(&reaplist);
4796 spin_lock(&recall_lock);
4797 list_for_each_safe(pos, next, &del_recall_lru) {
4798 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4799 list_move(&dp->dl_recall_lru, &reaplist);
4801 spin_unlock(&recall_lock);
4802 list_for_each_safe(pos, next, &reaplist) {
4803 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4804 unhash_delegation(dp);
4807 nfsd4_client_tracking_exit(&init_net);
4812 nfs4_state_shutdown(void)
4814 struct net *net = &init_net;
4815 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4817 cancel_delayed_work_sync(&laundromat_work);
4818 destroy_workqueue(laundry_wq);
4819 locks_end_grace(&nn->nfsd4_manager);
4821 __nfs4_state_shutdown();
4822 nfs4_unlock_state();
4823 nfsd4_destroy_callback_queue();
4827 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4829 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4830 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4834 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4836 if (cstate->minorversion) {
4837 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4838 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4843 clear_current_stateid(struct nfsd4_compound_state *cstate)
4845 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4849 * functions to set current state id
4852 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4854 put_stateid(cstate, &odp->od_stateid);
4858 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4860 put_stateid(cstate, &open->op_stateid);
4864 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4866 put_stateid(cstate, &close->cl_stateid);
4870 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4872 put_stateid(cstate, &lock->lk_resp_stateid);
4876 * functions to consume current state id
4880 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4882 get_stateid(cstate, &odp->od_stateid);
4886 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4888 get_stateid(cstate, &drp->dr_stateid);
4892 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4894 get_stateid(cstate, &fsp->fr_stateid);
4898 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4900 get_stateid(cstate, &setattr->sa_stateid);
4904 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4906 get_stateid(cstate, &close->cl_stateid);
4910 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4912 get_stateid(cstate, &locku->lu_stateid);
4916 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4918 get_stateid(cstate, &read->rd_stateid);
4922 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4924 get_stateid(cstate, &write->wr_stateid);