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 (conn->cn_flags & NFS4_CDFC4_BACK) {
865 /* callback channel may be back up */
866 nfsd4_probe_callback(ses->se_client);
870 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
872 u32 dir = NFS4_CDFC4_FORE;
874 if (cses->flags & SESSION4_BACK_CHAN)
875 dir |= NFS4_CDFC4_BACK;
876 return alloc_conn(rqstp, dir);
879 /* must be called under client_lock */
880 static void nfsd4_del_conns(struct nfsd4_session *s)
882 struct nfs4_client *clp = s->se_client;
883 struct nfsd4_conn *c;
885 spin_lock(&clp->cl_lock);
886 while (!list_empty(&s->se_conns)) {
887 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
888 list_del_init(&c->cn_persession);
889 spin_unlock(&clp->cl_lock);
891 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
894 spin_lock(&clp->cl_lock);
896 spin_unlock(&clp->cl_lock);
899 static void __free_session(struct nfsd4_session *ses)
901 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
902 free_session_slots(ses);
906 static void free_session(struct kref *kref)
908 struct nfsd4_session *ses;
910 lockdep_assert_held(&client_lock);
911 ses = container_of(kref, struct nfsd4_session, se_ref);
912 nfsd4_del_conns(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_session(struct nfsd4_channel_attrs *fchan)
925 struct nfsd4_session *new;
926 int numslots, slotsize;
928 * Note decreasing slot size below client's request may
929 * make it difficult for client to function correctly, whereas
930 * decreasing the number of slots will (just?) affect
931 * performance. When short on memory we therefore prefer to
932 * decrease number of slots instead of their size.
934 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
935 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
939 new = __alloc_session(slotsize, numslots);
941 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
944 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
948 void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
952 new->se_client = clp;
955 INIT_LIST_HEAD(&new->se_conns);
957 new->se_cb_seq_nr = 1;
958 new->se_flags = cses->flags;
959 new->se_cb_prog = cses->callback_prog;
960 kref_init(&new->se_ref);
961 idx = hash_sessionid(&new->se_sessionid);
962 spin_lock(&client_lock);
963 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
964 spin_lock(&clp->cl_lock);
965 list_add(&new->se_perclnt, &clp->cl_sessions);
966 spin_unlock(&clp->cl_lock);
967 spin_unlock(&client_lock);
969 if (cses->flags & SESSION4_BACK_CHAN) {
970 struct sockaddr *sa = svc_addr(rqstp);
972 * This is a little silly; with sessions there's no real
973 * use for the callback address. Use the peer address
974 * as a reasonable default for now, but consider fixing
975 * the rpc client not to require an address in the
978 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
979 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
983 /* caller must hold client_lock */
984 static struct nfsd4_session *
985 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
987 struct nfsd4_session *elem;
990 dump_sessionid(__func__, sessionid);
991 idx = hash_sessionid(sessionid);
992 /* Search in the appropriate list */
993 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
994 if (!memcmp(elem->se_sessionid.data, sessionid->data,
995 NFS4_MAX_SESSIONID_LEN)) {
1000 dprintk("%s: session not found\n", __func__);
1004 /* caller must hold client_lock */
1006 unhash_session(struct nfsd4_session *ses)
1008 list_del(&ses->se_hash);
1009 spin_lock(&ses->se_client->cl_lock);
1010 list_del(&ses->se_perclnt);
1011 spin_unlock(&ses->se_client->cl_lock);
1014 /* must be called under the client_lock */
1016 renew_client_locked(struct nfs4_client *clp)
1018 if (is_client_expired(clp)) {
1020 printk("%s: client (clientid %08x/%08x) already expired\n",
1022 clp->cl_clientid.cl_boot,
1023 clp->cl_clientid.cl_id);
1027 dprintk("renewing client (clientid %08x/%08x)\n",
1028 clp->cl_clientid.cl_boot,
1029 clp->cl_clientid.cl_id);
1030 list_move_tail(&clp->cl_lru, &client_lru);
1031 clp->cl_time = get_seconds();
1035 renew_client(struct nfs4_client *clp)
1037 spin_lock(&client_lock);
1038 renew_client_locked(clp);
1039 spin_unlock(&client_lock);
1042 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1044 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1046 if (clid->cl_boot == nn->boot_time)
1048 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1049 clid->cl_boot, clid->cl_id, nn->boot_time);
1054 * XXX Should we use a slab cache ?
1055 * This type of memory management is somewhat inefficient, but we use it
1056 * anyway since SETCLIENTID is not a common operation.
1058 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1060 struct nfs4_client *clp;
1062 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1065 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1066 if (clp->cl_name.data == NULL) {
1070 clp->cl_name.len = name.len;
1075 free_client(struct nfs4_client *clp)
1077 lockdep_assert_held(&client_lock);
1078 while (!list_empty(&clp->cl_sessions)) {
1079 struct nfsd4_session *ses;
1080 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1082 list_del(&ses->se_perclnt);
1083 nfsd4_put_session_locked(ses);
1085 free_svc_cred(&clp->cl_cred);
1086 kfree(clp->cl_name.data);
1091 release_session_client(struct nfsd4_session *session)
1093 struct nfs4_client *clp = session->se_client;
1095 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1097 if (is_client_expired(clp)) {
1099 session->se_client = NULL;
1101 renew_client_locked(clp);
1102 spin_unlock(&client_lock);
1105 /* must be called under the client_lock */
1107 unhash_client_locked(struct nfs4_client *clp)
1109 struct nfsd4_session *ses;
1111 mark_client_expired(clp);
1112 list_del(&clp->cl_lru);
1113 spin_lock(&clp->cl_lock);
1114 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1115 list_del_init(&ses->se_hash);
1116 spin_unlock(&clp->cl_lock);
1120 destroy_client(struct nfs4_client *clp)
1122 struct nfs4_openowner *oo;
1123 struct nfs4_delegation *dp;
1124 struct list_head reaplist;
1126 INIT_LIST_HEAD(&reaplist);
1127 spin_lock(&recall_lock);
1128 while (!list_empty(&clp->cl_delegations)) {
1129 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1130 list_del_init(&dp->dl_perclnt);
1131 list_move(&dp->dl_recall_lru, &reaplist);
1133 spin_unlock(&recall_lock);
1134 while (!list_empty(&reaplist)) {
1135 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1136 unhash_delegation(dp);
1138 while (!list_empty(&clp->cl_openowners)) {
1139 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1140 release_openowner(oo);
1142 nfsd4_shutdown_callback(clp);
1143 if (clp->cl_cb_conn.cb_xprt)
1144 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1145 list_del(&clp->cl_idhash);
1146 list_del(&clp->cl_strhash);
1147 spin_lock(&client_lock);
1148 unhash_client_locked(clp);
1149 if (atomic_read(&clp->cl_refcount) == 0)
1151 spin_unlock(&client_lock);
1154 static void expire_client(struct nfs4_client *clp)
1156 nfsd4_client_record_remove(clp);
1157 destroy_client(clp);
1160 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1162 memcpy(target->cl_verifier.data, source->data,
1163 sizeof(target->cl_verifier.data));
1166 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1168 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1169 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1172 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1174 if (source->cr_principal) {
1175 target->cr_principal =
1176 kstrdup(source->cr_principal, GFP_KERNEL);
1177 if (target->cr_principal == NULL)
1180 target->cr_principal = NULL;
1181 target->cr_flavor = source->cr_flavor;
1182 target->cr_uid = source->cr_uid;
1183 target->cr_gid = source->cr_gid;
1184 target->cr_group_info = source->cr_group_info;
1185 get_group_info(target->cr_group_info);
1189 static int same_name(const char *n1, const char *n2)
1191 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1195 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1197 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1201 same_clid(clientid_t *cl1, clientid_t *cl2)
1203 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1206 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1210 if (g1->ngroups != g2->ngroups)
1212 for (i=0; i<g1->ngroups; i++)
1213 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1219 * RFC 3530 language requires clid_inuse be returned when the
1220 * "principal" associated with a requests differs from that previously
1221 * used. We use uid, gid's, and gss principal string as our best
1222 * approximation. We also don't want to allow non-gss use of a client
1223 * established using gss: in theory cr_principal should catch that
1224 * change, but in practice cr_principal can be null even in the gss case
1225 * since gssd doesn't always pass down a principal string.
1227 static bool is_gss_cred(struct svc_cred *cr)
1229 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1230 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1235 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1237 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1238 || (cr1->cr_uid != cr2->cr_uid)
1239 || (cr1->cr_gid != cr2->cr_gid)
1240 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1242 if (cr1->cr_principal == cr2->cr_principal)
1244 if (!cr1->cr_principal || !cr2->cr_principal)
1246 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1249 static void gen_clid(struct nfs4_client *clp)
1251 static u32 current_clientid = 1;
1252 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
1254 clp->cl_clientid.cl_boot = nn->boot_time;
1255 clp->cl_clientid.cl_id = current_clientid++;
1258 static void gen_confirm(struct nfs4_client *clp)
1263 verf[0] = (__be32)get_seconds();
1264 verf[1] = (__be32)i++;
1265 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1268 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1270 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1273 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1275 struct nfs4_stid *s;
1277 s = find_stateid(cl, t);
1280 if (typemask & s->sc_type)
1285 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1286 struct svc_rqst *rqstp, nfs4_verifier *verf)
1288 struct nfs4_client *clp;
1289 struct sockaddr *sa = svc_addr(rqstp);
1292 clp = alloc_client(name);
1296 INIT_LIST_HEAD(&clp->cl_sessions);
1297 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1299 spin_lock(&client_lock);
1301 spin_unlock(&client_lock);
1304 idr_init(&clp->cl_stateids);
1305 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1306 atomic_set(&clp->cl_refcount, 0);
1307 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1308 INIT_LIST_HEAD(&clp->cl_idhash);
1309 INIT_LIST_HEAD(&clp->cl_strhash);
1310 INIT_LIST_HEAD(&clp->cl_openowners);
1311 INIT_LIST_HEAD(&clp->cl_delegations);
1312 INIT_LIST_HEAD(&clp->cl_lru);
1313 INIT_LIST_HEAD(&clp->cl_callbacks);
1314 spin_lock_init(&clp->cl_lock);
1315 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1316 clp->cl_time = get_seconds();
1317 clear_bit(0, &clp->cl_cb_slot_busy);
1318 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1319 copy_verf(clp, verf);
1320 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1322 clp->cl_cb_session = NULL;
1327 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1329 unsigned int idhashval;
1331 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1332 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1333 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1338 move_to_confirmed(struct nfs4_client *clp)
1340 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1341 unsigned int strhashval;
1343 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1344 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1345 strhashval = clientstr_hashval(clp->cl_recdir);
1346 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1350 static struct nfs4_client *
1351 find_confirmed_client(clientid_t *clid, bool sessions)
1353 struct nfs4_client *clp;
1354 unsigned int idhashval = clientid_hashval(clid->cl_id);
1356 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1357 if (same_clid(&clp->cl_clientid, clid)) {
1358 if ((bool)clp->cl_minorversion != sessions)
1367 static struct nfs4_client *
1368 find_unconfirmed_client(clientid_t *clid, bool sessions)
1370 struct nfs4_client *clp;
1371 unsigned int idhashval = clientid_hashval(clid->cl_id);
1373 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1374 if (same_clid(&clp->cl_clientid, clid)) {
1375 if ((bool)clp->cl_minorversion != sessions)
1383 static bool clp_used_exchangeid(struct nfs4_client *clp)
1385 return clp->cl_exchange_flags != 0;
1388 static struct nfs4_client *
1389 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1391 struct nfs4_client *clp;
1393 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1394 if (same_name(clp->cl_recdir, dname))
1400 static struct nfs4_client *
1401 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1403 struct nfs4_client *clp;
1405 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1406 if (same_name(clp->cl_recdir, dname))
1413 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1415 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1416 struct sockaddr *sa = svc_addr(rqstp);
1417 u32 scopeid = rpc_get_scope_id(sa);
1418 unsigned short expected_family;
1420 /* Currently, we only support tcp and tcp6 for the callback channel */
1421 if (se->se_callback_netid_len == 3 &&
1422 !memcmp(se->se_callback_netid_val, "tcp", 3))
1423 expected_family = AF_INET;
1424 else if (se->se_callback_netid_len == 4 &&
1425 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1426 expected_family = AF_INET6;
1430 conn->cb_addrlen = rpc_uaddr2sockaddr(&init_net, se->se_callback_addr_val,
1431 se->se_callback_addr_len,
1432 (struct sockaddr *)&conn->cb_addr,
1433 sizeof(conn->cb_addr));
1435 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1438 if (conn->cb_addr.ss_family == AF_INET6)
1439 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1441 conn->cb_prog = se->se_callback_prog;
1442 conn->cb_ident = se->se_callback_ident;
1443 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1446 conn->cb_addr.ss_family = AF_UNSPEC;
1447 conn->cb_addrlen = 0;
1448 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1449 "will not receive delegations\n",
1450 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1456 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1459 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1461 struct nfsd4_slot *slot = resp->cstate.slot;
1464 dprintk("--> %s slot %p\n", __func__, slot);
1466 slot->sl_opcnt = resp->opcnt;
1467 slot->sl_status = resp->cstate.status;
1469 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1470 if (nfsd4_not_cached(resp)) {
1471 slot->sl_datalen = 0;
1474 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1475 base = (char *)resp->cstate.datap -
1476 (char *)resp->xbuf->head[0].iov_base;
1477 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1479 WARN("%s: sessions DRC could not cache compound\n", __func__);
1484 * Encode the replay sequence operation from the slot values.
1485 * If cachethis is FALSE encode the uncached rep error on the next
1486 * operation which sets resp->p and increments resp->opcnt for
1487 * nfs4svc_encode_compoundres.
1491 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1492 struct nfsd4_compoundres *resp)
1494 struct nfsd4_op *op;
1495 struct nfsd4_slot *slot = resp->cstate.slot;
1497 /* Encode the replayed sequence operation */
1498 op = &args->ops[resp->opcnt - 1];
1499 nfsd4_encode_operation(resp, op);
1501 /* Return nfserr_retry_uncached_rep in next operation. */
1502 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1503 op = &args->ops[resp->opcnt++];
1504 op->status = nfserr_retry_uncached_rep;
1505 nfsd4_encode_operation(resp, op);
1511 * The sequence operation is not cached because we can use the slot and
1515 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1516 struct nfsd4_sequence *seq)
1518 struct nfsd4_slot *slot = resp->cstate.slot;
1521 dprintk("--> %s slot %p\n", __func__, slot);
1523 /* Either returns 0 or nfserr_retry_uncached */
1524 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1525 if (status == nfserr_retry_uncached_rep)
1528 /* The sequence operation has been encoded, cstate->datap set. */
1529 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1531 resp->opcnt = slot->sl_opcnt;
1532 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1533 status = slot->sl_status;
1539 * Set the exchange_id flags returned by the server.
1542 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1544 /* pNFS is not supported */
1545 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1547 /* Referrals are supported, Migration is not. */
1548 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1550 /* set the wire flags to return to client. */
1551 clid->flags = new->cl_exchange_flags;
1554 static bool client_has_state(struct nfs4_client *clp)
1557 * Note clp->cl_openowners check isn't quite right: there's no
1558 * need to count owners without stateid's.
1560 * Also note we should probably be using this in 4.0 case too.
1562 return !list_empty(&clp->cl_openowners)
1563 || !list_empty(&clp->cl_delegations)
1564 || !list_empty(&clp->cl_sessions);
1568 nfsd4_exchange_id(struct svc_rqst *rqstp,
1569 struct nfsd4_compound_state *cstate,
1570 struct nfsd4_exchange_id *exid)
1572 struct nfs4_client *unconf, *conf, *new;
1574 unsigned int strhashval;
1575 char dname[HEXDIR_LEN];
1576 char addr_str[INET6_ADDRSTRLEN];
1577 nfs4_verifier verf = exid->verifier;
1578 struct sockaddr *sa = svc_addr(rqstp);
1579 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1581 rpc_ntop(sa, addr_str, sizeof(addr_str));
1582 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1583 "ip_addr=%s flags %x, spa_how %d\n",
1584 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1585 addr_str, exid->flags, exid->spa_how);
1587 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1588 return nfserr_inval;
1590 /* Currently only support SP4_NONE */
1591 switch (exid->spa_how) {
1595 return nfserr_serverfault;
1597 BUG(); /* checked by xdr code */
1599 return nfserr_serverfault; /* no excuse :-/ */
1602 status = nfs4_make_rec_clidname(dname, &exid->clname);
1607 strhashval = clientstr_hashval(dname);
1609 /* Cases below refer to rfc 5661 section 18.35.4: */
1611 conf = find_confirmed_client_by_str(dname, strhashval);
1613 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1614 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1617 if (!clp_used_exchangeid(conf)) { /* buggy client */
1618 status = nfserr_inval;
1621 if (!creds_match) { /* case 9 */
1622 status = nfserr_perm;
1625 if (!verfs_match) { /* case 8 */
1626 status = nfserr_not_same;
1630 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1634 if (!creds_match) { /* case 3 */
1635 if (client_has_state(conf)) {
1636 status = nfserr_clid_inuse;
1639 expire_client(conf);
1642 if (verfs_match) { /* case 2 */
1643 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1647 /* case 5, client reboot */
1651 if (update) { /* case 7 */
1652 status = nfserr_noent;
1656 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1657 if (unconf) /* case 4, possible retry or client restart */
1658 expire_client(unconf);
1660 /* case 1 (normal case) */
1662 new = create_client(exid->clname, dname, rqstp, &verf);
1664 status = nfserr_jukebox;
1667 new->cl_minorversion = 1;
1670 add_to_unconfirmed(new, strhashval);
1672 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1673 exid->clientid.cl_id = new->cl_clientid.cl_id;
1675 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1676 nfsd4_set_ex_flags(new, exid);
1678 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1679 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1683 nfs4_unlock_state();
1688 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1690 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1693 /* The slot is in use, and no response has been sent. */
1695 if (seqid == slot_seqid)
1696 return nfserr_jukebox;
1698 return nfserr_seq_misordered;
1700 /* Note unsigned 32-bit arithmetic handles wraparound: */
1701 if (likely(seqid == slot_seqid + 1))
1703 if (seqid == slot_seqid)
1704 return nfserr_replay_cache;
1705 return nfserr_seq_misordered;
1709 * Cache the create session result into the create session single DRC
1710 * slot cache by saving the xdr structure. sl_seqid has been set.
1711 * Do this for solo or embedded create session operations.
1714 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1715 struct nfsd4_clid_slot *slot, __be32 nfserr)
1717 slot->sl_status = nfserr;
1718 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1722 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1723 struct nfsd4_clid_slot *slot)
1725 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1726 return slot->sl_status;
1729 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1730 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1731 1 + /* MIN tag is length with zero, only length */ \
1732 3 + /* version, opcount, opcode */ \
1733 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1734 /* seqid, slotID, slotID, cache */ \
1735 4 ) * sizeof(__be32))
1737 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1738 2 + /* verifier: AUTH_NULL, length 0 */\
1740 1 + /* MIN tag is length with zero, only length */ \
1741 3 + /* opcount, opcode, opstatus*/ \
1742 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1743 /* seqid, slotID, slotID, slotID, status */ \
1744 5 ) * sizeof(__be32))
1746 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1748 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1749 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1753 nfsd4_create_session(struct svc_rqst *rqstp,
1754 struct nfsd4_compound_state *cstate,
1755 struct nfsd4_create_session *cr_ses)
1757 struct sockaddr *sa = svc_addr(rqstp);
1758 struct nfs4_client *conf, *unconf;
1759 struct nfsd4_session *new;
1760 struct nfsd4_conn *conn;
1761 struct nfsd4_clid_slot *cs_slot = NULL;
1764 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1765 return nfserr_inval;
1766 if (check_forechannel_attrs(cr_ses->fore_channel))
1767 return nfserr_toosmall;
1768 new = alloc_session(&cr_ses->fore_channel);
1770 return nfserr_jukebox;
1771 status = nfserr_jukebox;
1772 conn = alloc_conn_from_crses(rqstp, cr_ses);
1774 goto out_free_session;
1777 unconf = find_unconfirmed_client(&cr_ses->clientid, true);
1778 conf = find_confirmed_client(&cr_ses->clientid, true);
1781 cs_slot = &conf->cl_cs_slot;
1782 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1783 if (status == nfserr_replay_cache) {
1784 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1786 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1787 status = nfserr_seq_misordered;
1790 } else if (unconf) {
1792 struct nfs4_client *old;
1793 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1794 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1795 status = nfserr_clid_inuse;
1798 cs_slot = &unconf->cl_cs_slot;
1799 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1801 /* an unconfirmed replay returns misordered */
1802 status = nfserr_seq_misordered;
1805 hash = clientstr_hashval(unconf->cl_recdir);
1806 old = find_confirmed_client_by_str(unconf->cl_recdir, hash);
1809 move_to_confirmed(unconf);
1812 status = nfserr_stale_clientid;
1817 * We do not support RDMA or persistent sessions
1819 cr_ses->flags &= ~SESSION4_PERSIST;
1820 cr_ses->flags &= ~SESSION4_RDMA;
1822 init_session(rqstp, new, conf, cr_ses);
1823 nfsd4_init_conn(rqstp, conn, new);
1825 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1826 NFS4_MAX_SESSIONID_LEN);
1827 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1828 sizeof(struct nfsd4_channel_attrs));
1829 cs_slot->sl_seqid++;
1830 cr_ses->seqid = cs_slot->sl_seqid;
1832 /* cache solo and embedded create sessions under the state lock */
1833 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1835 nfs4_unlock_state();
1836 dprintk("%s returns %d\n", __func__, ntohl(status));
1841 __free_session(new);
1845 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1847 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1848 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1850 return argp->opcnt == resp->opcnt;
1853 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1856 case NFS4_CDFC4_FORE:
1857 case NFS4_CDFC4_BACK:
1859 case NFS4_CDFC4_FORE_OR_BOTH:
1860 case NFS4_CDFC4_BACK_OR_BOTH:
1861 *dir = NFS4_CDFC4_BOTH;
1864 return nfserr_inval;
1867 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1868 struct nfsd4_compound_state *cstate,
1869 struct nfsd4_bind_conn_to_session *bcts)
1872 struct nfsd4_conn *conn;
1874 if (!nfsd4_last_compound_op(rqstp))
1875 return nfserr_not_only_op;
1876 spin_lock(&client_lock);
1877 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1878 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1879 * client_lock iself: */
1880 if (cstate->session) {
1881 nfsd4_get_session(cstate->session);
1882 atomic_inc(&cstate->session->se_client->cl_refcount);
1884 spin_unlock(&client_lock);
1885 if (!cstate->session)
1886 return nfserr_badsession;
1888 status = nfsd4_map_bcts_dir(&bcts->dir);
1891 conn = alloc_conn(rqstp, bcts->dir);
1893 return nfserr_jukebox;
1894 nfsd4_init_conn(rqstp, conn, cstate->session);
1898 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1902 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1906 nfsd4_destroy_session(struct svc_rqst *r,
1907 struct nfsd4_compound_state *cstate,
1908 struct nfsd4_destroy_session *sessionid)
1910 struct nfsd4_session *ses;
1911 __be32 status = nfserr_badsession;
1914 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1915 * - Should we return nfserr_back_chan_busy if waiting for
1916 * callbacks on to-be-destroyed session?
1917 * - Do we need to clear any callback info from previous session?
1920 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1921 if (!nfsd4_last_compound_op(r))
1922 return nfserr_not_only_op;
1924 dump_sessionid(__func__, &sessionid->sessionid);
1925 spin_lock(&client_lock);
1926 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1928 spin_unlock(&client_lock);
1932 unhash_session(ses);
1933 spin_unlock(&client_lock);
1936 nfsd4_probe_callback_sync(ses->se_client);
1937 nfs4_unlock_state();
1939 spin_lock(&client_lock);
1940 nfsd4_del_conns(ses);
1941 nfsd4_put_session_locked(ses);
1942 spin_unlock(&client_lock);
1945 dprintk("%s returns %d\n", __func__, ntohl(status));
1949 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1951 struct nfsd4_conn *c;
1953 list_for_each_entry(c, &s->se_conns, cn_persession) {
1954 if (c->cn_xprt == xpt) {
1961 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1963 struct nfs4_client *clp = ses->se_client;
1964 struct nfsd4_conn *c;
1967 spin_lock(&clp->cl_lock);
1968 c = __nfsd4_find_conn(new->cn_xprt, ses);
1970 spin_unlock(&clp->cl_lock);
1974 __nfsd4_hash_conn(new, ses);
1975 spin_unlock(&clp->cl_lock);
1976 ret = nfsd4_register_conn(new);
1978 /* oops; xprt is already down: */
1979 nfsd4_conn_lost(&new->cn_xpt_user);
1983 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1985 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1987 return args->opcnt > session->se_fchannel.maxops;
1990 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1991 struct nfsd4_session *session)
1993 struct xdr_buf *xb = &rqstp->rq_arg;
1995 return xb->len > session->se_fchannel.maxreq_sz;
1999 nfsd4_sequence(struct svc_rqst *rqstp,
2000 struct nfsd4_compound_state *cstate,
2001 struct nfsd4_sequence *seq)
2003 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2004 struct nfsd4_session *session;
2005 struct nfsd4_slot *slot;
2006 struct nfsd4_conn *conn;
2009 if (resp->opcnt != 1)
2010 return nfserr_sequence_pos;
2013 * Will be either used or freed by nfsd4_sequence_check_conn
2016 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2018 return nfserr_jukebox;
2020 spin_lock(&client_lock);
2021 status = nfserr_badsession;
2022 session = find_in_sessionid_hashtbl(&seq->sessionid);
2026 status = nfserr_too_many_ops;
2027 if (nfsd4_session_too_many_ops(rqstp, session))
2030 status = nfserr_req_too_big;
2031 if (nfsd4_request_too_big(rqstp, session))
2034 status = nfserr_badslot;
2035 if (seq->slotid >= session->se_fchannel.maxreqs)
2038 slot = session->se_slots[seq->slotid];
2039 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2041 /* We do not negotiate the number of slots yet, so set the
2042 * maxslots to the session maxreqs which is used to encode
2043 * sr_highest_slotid and the sr_target_slot id to maxslots */
2044 seq->maxslots = session->se_fchannel.maxreqs;
2046 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2047 slot->sl_flags & NFSD4_SLOT_INUSE);
2048 if (status == nfserr_replay_cache) {
2049 status = nfserr_seq_misordered;
2050 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2052 cstate->slot = slot;
2053 cstate->session = session;
2054 /* Return the cached reply status and set cstate->status
2055 * for nfsd4_proc_compound processing */
2056 status = nfsd4_replay_cache_entry(resp, seq);
2057 cstate->status = nfserr_replay_cache;
2063 nfsd4_sequence_check_conn(conn, session);
2066 /* Success! bump slot seqid */
2067 slot->sl_seqid = seq->seqid;
2068 slot->sl_flags |= NFSD4_SLOT_INUSE;
2070 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2072 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2074 cstate->slot = slot;
2075 cstate->session = session;
2078 /* Hold a session reference until done processing the compound. */
2079 if (cstate->session) {
2080 struct nfs4_client *clp = session->se_client;
2082 nfsd4_get_session(cstate->session);
2083 atomic_inc(&clp->cl_refcount);
2084 switch (clp->cl_cb_state) {
2086 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2088 case NFSD4_CB_FAULT:
2089 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2092 seq->status_flags = 0;
2096 spin_unlock(&client_lock);
2097 dprintk("%s: return %d\n", __func__, ntohl(status));
2102 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2104 struct nfs4_client *conf, *unconf, *clp;
2108 unconf = find_unconfirmed_client(&dc->clientid, true);
2109 conf = find_confirmed_client(&dc->clientid, true);
2114 if (!is_client_expired(conf) && client_has_state(conf)) {
2115 status = nfserr_clientid_busy;
2119 /* rfc5661 18.50.3 */
2120 if (cstate->session && conf == cstate->session->se_client) {
2121 status = nfserr_clientid_busy;
2127 status = nfserr_stale_clientid;
2133 nfs4_unlock_state();
2134 dprintk("%s return %d\n", __func__, ntohl(status));
2139 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2143 if (rc->rca_one_fs) {
2144 if (!cstate->current_fh.fh_dentry)
2145 return nfserr_nofilehandle;
2147 * We don't take advantage of the rca_one_fs case.
2148 * That's OK, it's optional, we can safely ignore it.
2154 status = nfserr_complete_already;
2155 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2156 &cstate->session->se_client->cl_flags))
2159 status = nfserr_stale_clientid;
2160 if (is_client_expired(cstate->session->se_client))
2162 * The following error isn't really legal.
2163 * But we only get here if the client just explicitly
2164 * destroyed the client. Surely it no longer cares what
2165 * error it gets back on an operation for the dead
2171 nfsd4_client_record_create(cstate->session->se_client);
2173 nfs4_unlock_state();
2178 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2179 struct nfsd4_setclientid *setclid)
2181 struct xdr_netobj clname = setclid->se_name;
2182 nfs4_verifier clverifier = setclid->se_verf;
2183 unsigned int strhashval;
2184 struct nfs4_client *conf, *unconf, *new;
2186 char dname[HEXDIR_LEN];
2188 status = nfs4_make_rec_clidname(dname, &clname);
2192 strhashval = clientstr_hashval(dname);
2194 /* Cases below refer to rfc 3530 section 14.2.33: */
2196 conf = find_confirmed_client_by_str(dname, strhashval);
2199 status = nfserr_clid_inuse;
2200 if (clp_used_exchangeid(conf))
2202 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2203 char addr_str[INET6_ADDRSTRLEN];
2204 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2206 dprintk("NFSD: setclientid: string in use by client "
2207 "at %s\n", addr_str);
2211 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2213 expire_client(unconf);
2214 status = nfserr_jukebox;
2215 new = create_client(clname, dname, rqstp, &clverifier);
2218 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2219 /* case 1: probable callback update */
2220 copy_clid(new, conf);
2221 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2223 new->cl_minorversion = 0;
2224 gen_callback(new, setclid, rqstp);
2225 add_to_unconfirmed(new, strhashval);
2226 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2227 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2228 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2231 nfs4_unlock_state();
2237 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2238 struct nfsd4_compound_state *cstate,
2239 struct nfsd4_setclientid_confirm *setclientid_confirm)
2241 struct nfs4_client *conf, *unconf;
2242 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2243 clientid_t * clid = &setclientid_confirm->sc_clientid;
2245 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2247 if (STALE_CLIENTID(clid, nn))
2248 return nfserr_stale_clientid;
2251 conf = find_confirmed_client(clid, false);
2252 unconf = find_unconfirmed_client(clid, false);
2254 * We try hard to give out unique clientid's, so if we get an
2255 * attempt to confirm the same clientid with a different cred,
2256 * there's a bug somewhere. Let's charitably assume it's our
2259 status = nfserr_serverfault;
2260 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2262 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2264 /* cases below refer to rfc 3530 section 14.2.34: */
2265 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2266 if (conf && !unconf) /* case 2: probable retransmit */
2268 else /* case 4: client hasn't noticed we rebooted yet? */
2269 status = nfserr_stale_clientid;
2273 if (conf) { /* case 1: callback update */
2274 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2275 nfsd4_probe_callback(conf);
2276 expire_client(unconf);
2277 } else { /* case 3: normal case; new or rebooted client */
2278 unsigned int hash = clientstr_hashval(unconf->cl_recdir);
2280 conf = find_confirmed_client_by_str(unconf->cl_recdir, hash);
2282 expire_client(conf);
2283 move_to_confirmed(unconf);
2284 nfsd4_probe_callback(unconf);
2287 nfs4_unlock_state();
2291 static struct nfs4_file *nfsd4_alloc_file(void)
2293 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2296 /* OPEN Share state helper functions */
2297 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2299 unsigned int hashval = file_hashval(ino);
2301 atomic_set(&fp->fi_ref, 1);
2302 INIT_LIST_HEAD(&fp->fi_hash);
2303 INIT_LIST_HEAD(&fp->fi_stateids);
2304 INIT_LIST_HEAD(&fp->fi_delegations);
2305 fp->fi_inode = igrab(ino);
2306 fp->fi_had_conflict = false;
2307 fp->fi_lease = NULL;
2308 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2309 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2310 spin_lock(&recall_lock);
2311 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2312 spin_unlock(&recall_lock);
2316 nfsd4_free_slab(struct kmem_cache **slab)
2320 kmem_cache_destroy(*slab);
2325 nfsd4_free_slabs(void)
2327 nfsd4_free_slab(&openowner_slab);
2328 nfsd4_free_slab(&lockowner_slab);
2329 nfsd4_free_slab(&file_slab);
2330 nfsd4_free_slab(&stateid_slab);
2331 nfsd4_free_slab(&deleg_slab);
2335 nfsd4_init_slabs(void)
2337 openowner_slab = kmem_cache_create("nfsd4_openowners",
2338 sizeof(struct nfs4_openowner), 0, 0, NULL);
2339 if (openowner_slab == NULL)
2341 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2342 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2343 if (lockowner_slab == NULL)
2345 file_slab = kmem_cache_create("nfsd4_files",
2346 sizeof(struct nfs4_file), 0, 0, NULL);
2347 if (file_slab == NULL)
2349 stateid_slab = kmem_cache_create("nfsd4_stateids",
2350 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2351 if (stateid_slab == NULL)
2353 deleg_slab = kmem_cache_create("nfsd4_delegations",
2354 sizeof(struct nfs4_delegation), 0, 0, NULL);
2355 if (deleg_slab == NULL)
2360 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2364 void nfs4_free_openowner(struct nfs4_openowner *oo)
2366 kfree(oo->oo_owner.so_owner.data);
2367 kmem_cache_free(openowner_slab, oo);
2370 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2372 kfree(lo->lo_owner.so_owner.data);
2373 kmem_cache_free(lockowner_slab, lo);
2376 static void init_nfs4_replay(struct nfs4_replay *rp)
2378 rp->rp_status = nfserr_serverfault;
2380 rp->rp_buf = rp->rp_ibuf;
2383 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2385 struct nfs4_stateowner *sop;
2387 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2391 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2392 if (!sop->so_owner.data) {
2393 kmem_cache_free(slab, sop);
2396 sop->so_owner.len = owner->len;
2398 INIT_LIST_HEAD(&sop->so_stateids);
2399 sop->so_client = clp;
2400 init_nfs4_replay(&sop->so_replay);
2404 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2406 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2407 list_add(&oo->oo_perclient, &clp->cl_openowners);
2410 static struct nfs4_openowner *
2411 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2412 struct nfs4_openowner *oo;
2414 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2417 oo->oo_owner.so_is_open_owner = 1;
2418 oo->oo_owner.so_seqid = open->op_seqid;
2419 oo->oo_flags = NFS4_OO_NEW;
2421 oo->oo_last_closed_stid = NULL;
2422 INIT_LIST_HEAD(&oo->oo_close_lru);
2423 hash_openowner(oo, clp, strhashval);
2427 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2428 struct nfs4_openowner *oo = open->op_openowner;
2429 struct nfs4_client *clp = oo->oo_owner.so_client;
2431 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2432 INIT_LIST_HEAD(&stp->st_lockowners);
2433 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2434 list_add(&stp->st_perfile, &fp->fi_stateids);
2435 stp->st_stateowner = &oo->oo_owner;
2438 stp->st_access_bmap = 0;
2439 stp->st_deny_bmap = 0;
2440 set_access(open->op_share_access, stp);
2441 set_deny(open->op_share_deny, stp);
2442 stp->st_openstp = NULL;
2446 move_to_close_lru(struct nfs4_openowner *oo)
2448 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2450 list_move_tail(&oo->oo_close_lru, &close_lru);
2451 oo->oo_time = get_seconds();
2455 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2458 return (sop->so_owner.len == owner->len) &&
2459 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2460 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2463 static struct nfs4_openowner *
2464 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open, bool sessions)
2466 struct nfs4_stateowner *so;
2467 struct nfs4_openowner *oo;
2468 struct nfs4_client *clp;
2470 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2471 if (!so->so_is_open_owner)
2473 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2475 clp = oo->oo_owner.so_client;
2476 if ((bool)clp->cl_minorversion != sessions)
2478 renew_client(oo->oo_owner.so_client);
2485 /* search file_hashtbl[] for file */
2486 static struct nfs4_file *
2487 find_file(struct inode *ino)
2489 unsigned int hashval = file_hashval(ino);
2490 struct nfs4_file *fp;
2492 spin_lock(&recall_lock);
2493 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2494 if (fp->fi_inode == ino) {
2496 spin_unlock(&recall_lock);
2500 spin_unlock(&recall_lock);
2505 * Called to check deny when READ with all zero stateid or
2506 * WRITE with all zero or all one stateid
2509 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2511 struct inode *ino = current_fh->fh_dentry->d_inode;
2512 struct nfs4_file *fp;
2513 struct nfs4_ol_stateid *stp;
2516 dprintk("NFSD: nfs4_share_conflict\n");
2518 fp = find_file(ino);
2521 ret = nfserr_locked;
2522 /* Search for conflicting share reservations */
2523 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2524 if (test_deny(deny_type, stp) ||
2525 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2534 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2536 /* We're assuming the state code never drops its reference
2537 * without first removing the lease. Since we're in this lease
2538 * callback (and since the lease code is serialized by the kernel
2539 * lock) we know the server hasn't removed the lease yet, we know
2540 * it's safe to take a reference: */
2541 atomic_inc(&dp->dl_count);
2543 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2545 /* only place dl_time is set. protected by lock_flocks*/
2546 dp->dl_time = get_seconds();
2548 nfsd4_cb_recall(dp);
2551 /* Called from break_lease() with lock_flocks() held. */
2552 static void nfsd_break_deleg_cb(struct file_lock *fl)
2554 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2555 struct nfs4_delegation *dp;
2558 WARN(1, "(%p)->fl_owner NULL\n", fl);
2561 if (fp->fi_had_conflict) {
2562 WARN(1, "duplicate break on %p\n", fp);
2566 * We don't want the locks code to timeout the lease for us;
2567 * we'll remove it ourself if a delegation isn't returned
2570 fl->fl_break_time = 0;
2572 spin_lock(&recall_lock);
2573 fp->fi_had_conflict = true;
2574 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2575 nfsd_break_one_deleg(dp);
2576 spin_unlock(&recall_lock);
2580 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2583 return lease_modify(onlist, arg);
2588 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2589 .lm_break = nfsd_break_deleg_cb,
2590 .lm_change = nfsd_change_deleg_cb,
2593 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2595 if (nfsd4_has_session(cstate))
2597 if (seqid == so->so_seqid - 1)
2598 return nfserr_replay_me;
2599 if (seqid == so->so_seqid)
2601 return nfserr_bad_seqid;
2605 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2606 struct nfsd4_open *open)
2608 clientid_t *clientid = &open->op_clientid;
2609 struct nfs4_client *clp = NULL;
2610 unsigned int strhashval;
2611 struct nfs4_openowner *oo = NULL;
2613 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2615 if (STALE_CLIENTID(&open->op_clientid, nn))
2616 return nfserr_stale_clientid;
2618 * In case we need it later, after we've already created the
2619 * file and don't want to risk a further failure:
2621 open->op_file = nfsd4_alloc_file();
2622 if (open->op_file == NULL)
2623 return nfserr_jukebox;
2625 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2626 oo = find_openstateowner_str(strhashval, open, cstate->minorversion);
2627 open->op_openowner = oo;
2629 clp = find_confirmed_client(clientid, cstate->minorversion);
2631 return nfserr_expired;
2634 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2635 /* Replace unconfirmed owners without checking for replay. */
2636 clp = oo->oo_owner.so_client;
2637 release_openowner(oo);
2638 open->op_openowner = NULL;
2641 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2644 clp = oo->oo_owner.so_client;
2647 oo = alloc_init_open_stateowner(strhashval, clp, open);
2649 return nfserr_jukebox;
2650 open->op_openowner = oo;
2652 open->op_stp = nfs4_alloc_stateid(clp);
2654 return nfserr_jukebox;
2658 static inline __be32
2659 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2661 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2662 return nfserr_openmode;
2667 static int share_access_to_flags(u32 share_access)
2669 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2672 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2674 struct nfs4_stid *ret;
2676 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2679 return delegstateid(ret);
2682 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2684 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2685 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2689 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2690 struct nfs4_delegation **dp)
2693 __be32 status = nfserr_bad_stateid;
2695 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2698 flags = share_access_to_flags(open->op_share_access);
2699 status = nfs4_check_delegmode(*dp, flags);
2703 if (!nfsd4_is_deleg_cur(open))
2707 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2712 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2714 struct nfs4_ol_stateid *local;
2715 struct nfs4_openowner *oo = open->op_openowner;
2717 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2718 /* ignore lock owners */
2719 if (local->st_stateowner->so_is_open_owner == 0)
2721 /* remember if we have seen this open owner */
2722 if (local->st_stateowner == &oo->oo_owner)
2724 /* check for conflicting share reservations */
2725 if (!test_share(local, open))
2726 return nfserr_share_denied;
2731 static inline int nfs4_access_to_access(u32 nfs4_access)
2735 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2736 flags |= NFSD_MAY_READ;
2737 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2738 flags |= NFSD_MAY_WRITE;
2742 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2743 struct svc_fh *cur_fh, struct nfsd4_open *open)
2746 int oflag = nfs4_access_to_omode(open->op_share_access);
2747 int access = nfs4_access_to_access(open->op_share_access);
2749 if (!fp->fi_fds[oflag]) {
2750 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2751 &fp->fi_fds[oflag]);
2755 nfs4_file_get_access(fp, oflag);
2760 static inline __be32
2761 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2762 struct nfsd4_open *open)
2764 struct iattr iattr = {
2765 .ia_valid = ATTR_SIZE,
2768 if (!open->op_truncate)
2770 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2771 return nfserr_inval;
2772 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2776 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)
2778 u32 op_share_access = open->op_share_access;
2782 new_access = !test_access(op_share_access, stp);
2784 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2788 status = nfsd4_truncate(rqstp, cur_fh, open);
2791 int oflag = nfs4_access_to_omode(op_share_access);
2792 nfs4_file_put_access(fp, oflag);
2796 /* remember the open */
2797 set_access(op_share_access, stp);
2798 set_deny(open->op_share_deny, stp);
2805 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2807 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2810 /* Should we give out recallable state?: */
2811 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2813 if (clp->cl_cb_state == NFSD4_CB_UP)
2816 * In the sessions case, since we don't have to establish a
2817 * separate connection for callbacks, we assume it's OK
2818 * until we hear otherwise:
2820 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2823 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2825 struct file_lock *fl;
2827 fl = locks_alloc_lock();
2830 locks_init_lock(fl);
2831 fl->fl_lmops = &nfsd_lease_mng_ops;
2832 fl->fl_flags = FL_LEASE;
2833 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2834 fl->fl_end = OFFSET_MAX;
2835 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2836 fl->fl_pid = current->tgid;
2840 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2842 struct nfs4_file *fp = dp->dl_file;
2843 struct file_lock *fl;
2846 fl = nfs4_alloc_init_lease(dp, flag);
2849 fl->fl_file = find_readable_file(fp);
2850 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2851 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2853 list_del_init(&dp->dl_perclnt);
2854 locks_free_lock(fl);
2858 fp->fi_deleg_file = get_file(fl->fl_file);
2859 atomic_set(&fp->fi_delegees, 1);
2860 list_add(&dp->dl_perfile, &fp->fi_delegations);
2864 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2866 struct nfs4_file *fp = dp->dl_file;
2869 return nfs4_setlease(dp, flag);
2870 spin_lock(&recall_lock);
2871 if (fp->fi_had_conflict) {
2872 spin_unlock(&recall_lock);
2875 atomic_inc(&fp->fi_delegees);
2876 list_add(&dp->dl_perfile, &fp->fi_delegations);
2877 spin_unlock(&recall_lock);
2878 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2882 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2884 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2885 if (status == -EAGAIN)
2886 open->op_why_no_deleg = WND4_CONTENTION;
2888 open->op_why_no_deleg = WND4_RESOURCE;
2889 switch (open->op_deleg_want) {
2890 case NFS4_SHARE_WANT_READ_DELEG:
2891 case NFS4_SHARE_WANT_WRITE_DELEG:
2892 case NFS4_SHARE_WANT_ANY_DELEG:
2894 case NFS4_SHARE_WANT_CANCEL:
2895 open->op_why_no_deleg = WND4_CANCELLED;
2897 case NFS4_SHARE_WANT_NO_DELEG:
2898 BUG(); /* not supposed to get here */
2904 * Attempt to hand out a delegation.
2907 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2908 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2910 struct nfs4_delegation *dp;
2911 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2913 int status = 0, flag = 0;
2915 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2916 flag = NFS4_OPEN_DELEGATE_NONE;
2917 open->op_recall = 0;
2918 switch (open->op_claim_type) {
2919 case NFS4_OPEN_CLAIM_PREVIOUS:
2921 open->op_recall = 1;
2922 flag = open->op_delegate_type;
2923 if (flag == NFS4_OPEN_DELEGATE_NONE)
2926 case NFS4_OPEN_CLAIM_NULL:
2927 /* Let's not give out any delegations till everyone's
2928 * had the chance to reclaim theirs.... */
2929 if (locks_in_grace(net))
2931 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2933 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2934 flag = NFS4_OPEN_DELEGATE_WRITE;
2936 flag = NFS4_OPEN_DELEGATE_READ;
2942 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2945 status = nfs4_set_delegation(dp, flag);
2949 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2951 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2952 STATEID_VAL(&dp->dl_stid.sc_stateid));
2954 open->op_delegate_type = flag;
2955 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2956 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2957 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2958 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2960 /* 4.1 client asking for a delegation? */
2961 if (open->op_deleg_want)
2962 nfsd4_open_deleg_none_ext(open, status);
2966 nfs4_put_delegation(dp);
2968 flag = NFS4_OPEN_DELEGATE_NONE;
2972 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2973 struct nfs4_delegation *dp)
2975 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2976 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2977 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2978 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2979 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2980 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2981 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2982 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
2984 /* Otherwise the client must be confused wanting a delegation
2985 * it already has, therefore we don't return
2986 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
2991 * called with nfs4_lock_state() held.
2994 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2996 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2997 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
2998 struct nfs4_file *fp = NULL;
2999 struct inode *ino = current_fh->fh_dentry->d_inode;
3000 struct nfs4_ol_stateid *stp = NULL;
3001 struct nfs4_delegation *dp = NULL;
3005 * Lookup file; if found, lookup stateid and check open request,
3006 * and check for delegations in the process of being recalled.
3007 * If not found, create the nfs4_file struct
3009 fp = find_file(ino);
3011 if ((status = nfs4_check_open(fp, open, &stp)))
3013 status = nfs4_check_deleg(cl, fp, open, &dp);
3017 status = nfserr_bad_stateid;
3018 if (nfsd4_is_deleg_cur(open))
3020 status = nfserr_jukebox;
3022 open->op_file = NULL;
3023 nfsd4_init_file(fp, ino);
3027 * OPEN the file, or upgrade an existing OPEN.
3028 * If truncate fails, the OPEN fails.
3031 /* Stateid was found, this is an OPEN upgrade */
3032 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3036 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3039 status = nfsd4_truncate(rqstp, current_fh, open);
3043 open->op_stp = NULL;
3044 init_open_stateid(stp, fp, open);
3046 update_stateid(&stp->st_stid.sc_stateid);
3047 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3049 if (nfsd4_has_session(&resp->cstate)) {
3050 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3052 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3053 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3054 open->op_why_no_deleg = WND4_NOT_WANTED;
3060 * Attempt to hand out a delegation. No error return, because the
3061 * OPEN succeeds even if we fail.
3063 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3067 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3068 STATEID_VAL(&stp->st_stid.sc_stateid));
3070 /* 4.1 client trying to upgrade/downgrade delegation? */
3071 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3072 open->op_deleg_want)
3073 nfsd4_deleg_xgrade_none_ext(open, dp);
3077 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3078 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3080 * To finish the open response, we just need to set the rflags.
3082 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3083 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3084 !nfsd4_has_session(&resp->cstate))
3085 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3090 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3092 if (open->op_openowner) {
3093 struct nfs4_openowner *oo = open->op_openowner;
3095 if (!list_empty(&oo->oo_owner.so_stateids))
3096 list_del_init(&oo->oo_close_lru);
3097 if (oo->oo_flags & NFS4_OO_NEW) {
3099 release_openowner(oo);
3100 open->op_openowner = NULL;
3102 oo->oo_flags &= ~NFS4_OO_NEW;
3106 nfsd4_free_file(open->op_file);
3108 free_generic_stateid(open->op_stp);
3112 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3115 struct nfs4_client *clp;
3117 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3120 dprintk("process_renew(%08x/%08x): starting\n",
3121 clid->cl_boot, clid->cl_id);
3122 status = nfserr_stale_clientid;
3123 if (STALE_CLIENTID(clid, nn))
3125 clp = find_confirmed_client(clid, cstate->minorversion);
3126 status = nfserr_expired;
3128 /* We assume the client took too long to RENEW. */
3129 dprintk("nfsd4_renew: clientid not found!\n");
3132 status = nfserr_cb_path_down;
3133 if (!list_empty(&clp->cl_delegations)
3134 && clp->cl_cb_state != NFSD4_CB_UP)
3138 nfs4_unlock_state();
3143 nfsd4_end_grace(struct net *net)
3145 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3147 /* do nothing if grace period already ended */
3148 if (nn->grace_ended)
3151 dprintk("NFSD: end of grace period\n");
3152 nn->grace_ended = true;
3153 nfsd4_record_grace_done(net, nn->boot_time);
3154 locks_end_grace(&nn->nfsd4_manager);
3156 * Now that every NFSv4 client has had the chance to recover and
3157 * to see the (possibly new, possibly shorter) lease time, we
3158 * can safely set the next grace time to the current lease time:
3160 nfsd4_grace = nfsd4_lease;
3164 nfs4_laundromat(void)
3166 struct nfs4_client *clp;
3167 struct nfs4_openowner *oo;
3168 struct nfs4_delegation *dp;
3169 struct list_head *pos, *next, reaplist;
3170 time_t cutoff = get_seconds() - nfsd4_lease;
3171 time_t t, clientid_val = nfsd4_lease;
3172 time_t u, test_val = nfsd4_lease;
3176 dprintk("NFSD: laundromat service - starting\n");
3177 nfsd4_end_grace(&init_net);
3178 INIT_LIST_HEAD(&reaplist);
3179 spin_lock(&client_lock);
3180 list_for_each_safe(pos, next, &client_lru) {
3181 clp = list_entry(pos, struct nfs4_client, cl_lru);
3182 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3183 t = clp->cl_time - cutoff;
3184 if (clientid_val > t)
3188 if (atomic_read(&clp->cl_refcount)) {
3189 dprintk("NFSD: client in use (clientid %08x)\n",
3190 clp->cl_clientid.cl_id);
3193 unhash_client_locked(clp);
3194 list_add(&clp->cl_lru, &reaplist);
3196 spin_unlock(&client_lock);
3197 list_for_each_safe(pos, next, &reaplist) {
3198 clp = list_entry(pos, struct nfs4_client, cl_lru);
3199 dprintk("NFSD: purging unused client (clientid %08x)\n",
3200 clp->cl_clientid.cl_id);
3203 spin_lock(&recall_lock);
3204 list_for_each_safe(pos, next, &del_recall_lru) {
3205 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3206 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3207 u = dp->dl_time - cutoff;
3212 list_move(&dp->dl_recall_lru, &reaplist);
3214 spin_unlock(&recall_lock);
3215 list_for_each_safe(pos, next, &reaplist) {
3216 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3217 unhash_delegation(dp);
3219 test_val = nfsd4_lease;
3220 list_for_each_safe(pos, next, &close_lru) {
3221 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3222 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3223 u = oo->oo_time - cutoff;
3228 release_openowner(oo);
3230 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3231 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3232 nfs4_unlock_state();
3233 return clientid_val;
3236 static struct workqueue_struct *laundry_wq;
3237 static void laundromat_main(struct work_struct *);
3238 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3241 laundromat_main(struct work_struct *not_used)
3245 t = nfs4_laundromat();
3246 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3247 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3250 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3252 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3253 return nfserr_bad_stateid;
3258 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3260 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3262 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3263 STATEID_VAL(stateid));
3268 access_permit_read(struct nfs4_ol_stateid *stp)
3270 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3271 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3272 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3276 access_permit_write(struct nfs4_ol_stateid *stp)
3278 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3279 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3283 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3285 __be32 status = nfserr_openmode;
3287 /* For lock stateid's, we test the parent open, not the lock: */
3288 if (stp->st_openstp)
3289 stp = stp->st_openstp;
3290 if ((flags & WR_STATE) && !access_permit_write(stp))
3292 if ((flags & RD_STATE) && !access_permit_read(stp))
3299 static inline __be32
3300 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3302 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3304 else if (locks_in_grace(net)) {
3305 /* Answer in remaining cases depends on existence of
3306 * conflicting state; so we must wait out the grace period. */
3307 return nfserr_grace;
3308 } else if (flags & WR_STATE)
3309 return nfs4_share_conflict(current_fh,
3310 NFS4_SHARE_DENY_WRITE);
3311 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3312 return nfs4_share_conflict(current_fh,
3313 NFS4_SHARE_DENY_READ);
3317 * Allow READ/WRITE during grace period on recovered state only for files
3318 * that are not able to provide mandatory locking.
3321 grace_disallows_io(struct net *net, struct inode *inode)
3323 return locks_in_grace(net) && mandatory_lock(inode);
3326 /* Returns true iff a is later than b: */
3327 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3329 return (s32)a->si_generation - (s32)b->si_generation > 0;
3332 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3335 * When sessions are used the stateid generation number is ignored
3338 if (has_session && in->si_generation == 0)
3341 if (in->si_generation == ref->si_generation)
3344 /* If the client sends us a stateid from the future, it's buggy: */
3345 if (stateid_generation_after(in, ref))
3346 return nfserr_bad_stateid;
3348 * However, we could see a stateid from the past, even from a
3349 * non-buggy client. For example, if the client sends a lock
3350 * while some IO is outstanding, the lock may bump si_generation
3351 * while the IO is still in flight. The client could avoid that
3352 * situation by waiting for responses on all the IO requests,
3353 * but better performance may result in retrying IO that
3354 * receives an old_stateid error if requests are rarely
3355 * reordered in flight:
3357 return nfserr_old_stateid;
3360 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3362 struct nfs4_stid *s;
3363 struct nfs4_ol_stateid *ols;
3366 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3367 return nfserr_bad_stateid;
3368 /* Client debugging aid. */
3369 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3370 char addr_str[INET6_ADDRSTRLEN];
3371 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3373 pr_warn_ratelimited("NFSD: client %s testing state ID "
3374 "with incorrect client ID\n", addr_str);
3375 return nfserr_bad_stateid;
3377 s = find_stateid(cl, stateid);
3379 return nfserr_bad_stateid;
3380 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3383 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3385 ols = openlockstateid(s);
3386 if (ols->st_stateowner->so_is_open_owner
3387 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3388 return nfserr_bad_stateid;
3392 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s, bool sessions)
3394 struct nfs4_client *cl;
3395 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3397 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3398 return nfserr_bad_stateid;
3399 if (STALE_STATEID(stateid, nn))
3400 return nfserr_stale_stateid;
3401 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions);
3403 return nfserr_expired;
3404 *s = find_stateid_by_type(cl, stateid, typemask);
3406 return nfserr_bad_stateid;
3412 * Checks for stateid operations
3415 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3416 stateid_t *stateid, int flags, struct file **filpp)
3418 struct nfs4_stid *s;
3419 struct nfs4_ol_stateid *stp = NULL;
3420 struct nfs4_delegation *dp = NULL;
3421 struct svc_fh *current_fh = &cstate->current_fh;
3422 struct inode *ino = current_fh->fh_dentry->d_inode;
3428 if (grace_disallows_io(net, ino))
3429 return nfserr_grace;
3431 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3432 return check_special_stateids(net, current_fh, stateid, flags);
3434 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s, cstate->minorversion);
3437 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3440 switch (s->sc_type) {
3441 case NFS4_DELEG_STID:
3442 dp = delegstateid(s);
3443 status = nfs4_check_delegmode(dp, flags);
3447 *filpp = dp->dl_file->fi_deleg_file;
3451 case NFS4_OPEN_STID:
3452 case NFS4_LOCK_STID:
3453 stp = openlockstateid(s);
3454 status = nfs4_check_fh(current_fh, stp);
3457 if (stp->st_stateowner->so_is_open_owner
3458 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3460 status = nfs4_check_openmode(stp, flags);
3464 if (flags & RD_STATE)
3465 *filpp = find_readable_file(stp->st_file);
3467 *filpp = find_writeable_file(stp->st_file);
3471 return nfserr_bad_stateid;
3479 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3481 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3482 return nfserr_locks_held;
3483 release_lock_stateid(stp);
3488 * Test if the stateid is valid
3491 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3492 struct nfsd4_test_stateid *test_stateid)
3494 struct nfsd4_test_stateid_id *stateid;
3495 struct nfs4_client *cl = cstate->session->se_client;
3498 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3499 stateid->ts_id_status =
3500 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3501 nfs4_unlock_state();
3507 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3508 struct nfsd4_free_stateid *free_stateid)
3510 stateid_t *stateid = &free_stateid->fr_stateid;
3511 struct nfs4_stid *s;
3512 struct nfs4_client *cl = cstate->session->se_client;
3513 __be32 ret = nfserr_bad_stateid;
3516 s = find_stateid(cl, stateid);
3519 switch (s->sc_type) {
3520 case NFS4_DELEG_STID:
3521 ret = nfserr_locks_held;
3523 case NFS4_OPEN_STID:
3524 case NFS4_LOCK_STID:
3525 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3528 if (s->sc_type == NFS4_LOCK_STID)
3529 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3531 ret = nfserr_locks_held;
3534 ret = nfserr_bad_stateid;
3537 nfs4_unlock_state();
3544 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3545 RD_STATE : WR_STATE;
3548 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3550 struct svc_fh *current_fh = &cstate->current_fh;
3551 struct nfs4_stateowner *sop = stp->st_stateowner;
3554 status = nfsd4_check_seqid(cstate, sop, seqid);
3557 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3559 * "Closed" stateid's exist *only* to return
3560 * nfserr_replay_me from the previous step.
3562 return nfserr_bad_stateid;
3563 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3566 return nfs4_check_fh(current_fh, stp);
3570 * Checks for sequence id mutating operations.
3573 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3574 stateid_t *stateid, char typemask,
3575 struct nfs4_ol_stateid **stpp)
3578 struct nfs4_stid *s;
3580 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3581 seqid, STATEID_VAL(stateid));
3584 status = nfsd4_lookup_stateid(stateid, typemask, &s, cstate->minorversion);
3587 *stpp = openlockstateid(s);
3588 cstate->replay_owner = (*stpp)->st_stateowner;
3590 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3593 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3596 struct nfs4_openowner *oo;
3598 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3599 NFS4_OPEN_STID, stpp);
3602 oo = openowner((*stpp)->st_stateowner);
3603 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3604 return nfserr_bad_stateid;
3609 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3610 struct nfsd4_open_confirm *oc)
3613 struct nfs4_openowner *oo;
3614 struct nfs4_ol_stateid *stp;
3616 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3617 (int)cstate->current_fh.fh_dentry->d_name.len,
3618 cstate->current_fh.fh_dentry->d_name.name);
3620 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3626 status = nfs4_preprocess_seqid_op(cstate,
3627 oc->oc_seqid, &oc->oc_req_stateid,
3628 NFS4_OPEN_STID, &stp);
3631 oo = openowner(stp->st_stateowner);
3632 status = nfserr_bad_stateid;
3633 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3635 oo->oo_flags |= NFS4_OO_CONFIRMED;
3636 update_stateid(&stp->st_stid.sc_stateid);
3637 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3638 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3639 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3641 nfsd4_client_record_create(oo->oo_owner.so_client);
3644 if (!cstate->replay_owner)
3645 nfs4_unlock_state();
3649 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3651 if (!test_access(access, stp))
3653 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3654 clear_access(access, stp);
3657 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3659 switch (to_access) {
3660 case NFS4_SHARE_ACCESS_READ:
3661 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3662 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3664 case NFS4_SHARE_ACCESS_WRITE:
3665 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3666 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3668 case NFS4_SHARE_ACCESS_BOTH:
3676 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3679 for (i = 0; i < 4; i++) {
3680 if ((i & deny) != i)
3686 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3687 struct nfsd4_compound_state *cstate,
3688 struct nfsd4_open_downgrade *od)
3691 struct nfs4_ol_stateid *stp;
3693 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3694 (int)cstate->current_fh.fh_dentry->d_name.len,
3695 cstate->current_fh.fh_dentry->d_name.name);
3697 /* We don't yet support WANT bits: */
3698 if (od->od_deleg_want)
3699 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3703 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3704 &od->od_stateid, &stp);
3707 status = nfserr_inval;
3708 if (!test_access(od->od_share_access, stp)) {
3709 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3710 stp->st_access_bmap, od->od_share_access);
3713 if (!test_deny(od->od_share_deny, stp)) {
3714 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3715 stp->st_deny_bmap, od->od_share_deny);
3718 nfs4_stateid_downgrade(stp, od->od_share_access);
3720 reset_union_bmap_deny(od->od_share_deny, stp);
3722 update_stateid(&stp->st_stid.sc_stateid);
3723 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3726 if (!cstate->replay_owner)
3727 nfs4_unlock_state();
3731 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3733 struct nfs4_openowner *oo;
3734 struct nfs4_ol_stateid *s;
3736 if (!so->so_is_open_owner)
3739 s = oo->oo_last_closed_stid;
3742 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3743 /* Release the last_closed_stid on the next seqid bump: */
3744 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3747 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3748 release_last_closed_stateid(oo);
3751 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3753 unhash_open_stateid(s);
3754 s->st_stid.sc_type = NFS4_CLOSED_STID;
3758 * nfs4_unlock_state() called after encode
3761 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3762 struct nfsd4_close *close)
3765 struct nfs4_openowner *oo;
3766 struct nfs4_ol_stateid *stp;
3768 dprintk("NFSD: nfsd4_close on file %.*s\n",
3769 (int)cstate->current_fh.fh_dentry->d_name.len,
3770 cstate->current_fh.fh_dentry->d_name.name);
3773 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3775 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3779 oo = openowner(stp->st_stateowner);
3781 update_stateid(&stp->st_stid.sc_stateid);
3782 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3784 nfsd4_close_open_stateid(stp);
3785 release_last_closed_stateid(oo);
3786 oo->oo_last_closed_stid = stp;
3788 if (list_empty(&oo->oo_owner.so_stateids)) {
3789 if (cstate->minorversion) {
3790 release_openowner(oo);
3791 cstate->replay_owner = NULL;
3794 * In the 4.0 case we need to keep the owners around a
3795 * little while to handle CLOSE replay.
3797 if (list_empty(&oo->oo_owner.so_stateids))
3798 move_to_close_lru(oo);
3802 if (!cstate->replay_owner)
3803 nfs4_unlock_state();
3808 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3809 struct nfsd4_delegreturn *dr)
3811 struct nfs4_delegation *dp;
3812 stateid_t *stateid = &dr->dr_stateid;
3813 struct nfs4_stid *s;
3816 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3820 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s, cstate->minorversion);
3823 dp = delegstateid(s);
3824 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3828 unhash_delegation(dp);
3830 nfs4_unlock_state();
3836 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3838 #define LOCKOWNER_INO_HASH_BITS 8
3839 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3840 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3843 end_offset(u64 start, u64 len)
3848 return end >= start ? end: NFS4_MAX_UINT64;
3851 /* last octet in a range */
3853 last_byte_offset(u64 start, u64 len)
3859 return end > start ? end - 1: NFS4_MAX_UINT64;
3862 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3864 return (file_hashval(inode) + cl_id
3865 + opaque_hashval(ownername->data, ownername->len))
3866 & LOCKOWNER_INO_HASH_MASK;
3869 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3872 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3873 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3874 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3875 * locking, this prevents us from being completely protocol-compliant. The
3876 * real solution to this problem is to start using unsigned file offsets in
3877 * the VFS, but this is a very deep change!
3880 nfs4_transform_lock_offset(struct file_lock *lock)
3882 if (lock->fl_start < 0)
3883 lock->fl_start = OFFSET_MAX;
3884 if (lock->fl_end < 0)
3885 lock->fl_end = OFFSET_MAX;
3888 /* Hack!: For now, we're defining this just so we can use a pointer to it
3889 * as a unique cookie to identify our (NFSv4's) posix locks. */
3890 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3894 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3896 struct nfs4_lockowner *lo;
3898 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3899 lo = (struct nfs4_lockowner *) fl->fl_owner;
3900 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3901 lo->lo_owner.so_owner.len, GFP_KERNEL);
3902 if (!deny->ld_owner.data)
3903 /* We just don't care that much */
3905 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3906 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3909 deny->ld_owner.len = 0;
3910 deny->ld_owner.data = NULL;
3911 deny->ld_clientid.cl_boot = 0;
3912 deny->ld_clientid.cl_id = 0;
3914 deny->ld_start = fl->fl_start;
3915 deny->ld_length = NFS4_MAX_UINT64;
3916 if (fl->fl_end != NFS4_MAX_UINT64)
3917 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3918 deny->ld_type = NFS4_READ_LT;
3919 if (fl->fl_type != F_RDLCK)
3920 deny->ld_type = NFS4_WRITE_LT;
3923 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3925 struct nfs4_ol_stateid *lst;
3927 if (!same_owner_str(&lo->lo_owner, owner, clid))
3929 lst = list_first_entry(&lo->lo_owner.so_stateids,
3930 struct nfs4_ol_stateid, st_perstateowner);
3931 return lst->st_file->fi_inode == inode;
3934 static struct nfs4_lockowner *
3935 find_lockowner_str(struct inode *inode, clientid_t *clid,
3936 struct xdr_netobj *owner)
3938 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3939 struct nfs4_lockowner *lo;
3941 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3942 if (same_lockowner_ino(lo, inode, clid, owner))
3948 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3950 struct inode *inode = open_stp->st_file->fi_inode;
3951 unsigned int inohash = lockowner_ino_hashval(inode,
3952 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3954 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3955 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3956 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3960 * Alloc a lock owner structure.
3961 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3964 * strhashval = ownerstr_hashval
3967 static struct nfs4_lockowner *
3968 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3969 struct nfs4_lockowner *lo;
3971 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3974 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3975 lo->lo_owner.so_is_open_owner = 0;
3976 /* It is the openowner seqid that will be incremented in encode in the
3977 * case of new lockowners; so increment the lock seqid manually: */
3978 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3979 hash_lockowner(lo, strhashval, clp, open_stp);
3983 static struct nfs4_ol_stateid *
3984 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3986 struct nfs4_ol_stateid *stp;
3987 struct nfs4_client *clp = lo->lo_owner.so_client;
3989 stp = nfs4_alloc_stateid(clp);
3992 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3993 list_add(&stp->st_perfile, &fp->fi_stateids);
3994 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3995 stp->st_stateowner = &lo->lo_owner;
3998 stp->st_access_bmap = 0;
3999 stp->st_deny_bmap = open_stp->st_deny_bmap;
4000 stp->st_openstp = open_stp;
4005 check_lock_length(u64 offset, u64 length)
4007 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4008 LOFF_OVERFLOW(offset, length)));
4011 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4013 struct nfs4_file *fp = lock_stp->st_file;
4014 int oflag = nfs4_access_to_omode(access);
4016 if (test_access(access, lock_stp))
4018 nfs4_file_get_access(fp, oflag);
4019 set_access(access, lock_stp);
4022 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)
4024 struct nfs4_file *fi = ost->st_file;
4025 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4026 struct nfs4_client *cl = oo->oo_owner.so_client;
4027 struct nfs4_lockowner *lo;
4028 unsigned int strhashval;
4030 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4032 if (!cstate->minorversion)
4033 return nfserr_bad_seqid;
4034 /* XXX: a lockowner always has exactly one stateid: */
4035 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4036 struct nfs4_ol_stateid, st_perstateowner);
4039 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4040 &lock->v.new.owner);
4041 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4043 return nfserr_jukebox;
4044 *lst = alloc_init_lock_stateid(lo, fi, ost);
4046 release_lockowner(lo);
4047 return nfserr_jukebox;
4057 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4058 struct nfsd4_lock *lock)
4060 struct nfs4_openowner *open_sop = NULL;
4061 struct nfs4_lockowner *lock_sop = NULL;
4062 struct nfs4_ol_stateid *lock_stp;
4063 struct file *filp = NULL;
4064 struct file_lock *file_lock = NULL;
4065 struct file_lock *conflock = NULL;
4067 bool new_state = false;
4070 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4072 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4073 (long long) lock->lk_offset,
4074 (long long) lock->lk_length);
4076 if (check_lock_length(lock->lk_offset, lock->lk_length))
4077 return nfserr_inval;
4079 if ((status = fh_verify(rqstp, &cstate->current_fh,
4080 S_IFREG, NFSD_MAY_LOCK))) {
4081 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4087 if (lock->lk_is_new) {
4088 struct nfs4_ol_stateid *open_stp = NULL;
4090 if (nfsd4_has_session(cstate))
4091 /* See rfc 5661 18.10.3: given clientid is ignored: */
4092 memcpy(&lock->v.new.clientid,
4093 &cstate->session->se_client->cl_clientid,
4094 sizeof(clientid_t));
4096 status = nfserr_stale_clientid;
4097 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4100 /* validate and update open stateid and open seqid */
4101 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4102 lock->lk_new_open_seqid,
4103 &lock->lk_new_open_stateid,
4107 open_sop = openowner(open_stp->st_stateowner);
4108 status = nfserr_bad_stateid;
4109 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4110 &lock->v.new.clientid))
4112 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4113 &lock_stp, &new_state);
4115 status = nfs4_preprocess_seqid_op(cstate,
4116 lock->lk_old_lock_seqid,
4117 &lock->lk_old_lock_stateid,
4118 NFS4_LOCK_STID, &lock_stp);
4121 lock_sop = lockowner(lock_stp->st_stateowner);
4123 lkflg = setlkflg(lock->lk_type);
4124 status = nfs4_check_openmode(lock_stp, lkflg);
4128 status = nfserr_grace;
4129 if (locks_in_grace(SVC_NET(rqstp)) && !lock->lk_reclaim)
4131 status = nfserr_no_grace;
4132 if (!locks_in_grace(SVC_NET(rqstp)) && lock->lk_reclaim)
4135 file_lock = locks_alloc_lock();
4137 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4138 status = nfserr_jukebox;
4142 locks_init_lock(file_lock);
4143 switch (lock->lk_type) {
4146 filp = find_readable_file(lock_stp->st_file);
4148 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4149 file_lock->fl_type = F_RDLCK;
4152 case NFS4_WRITEW_LT:
4153 filp = find_writeable_file(lock_stp->st_file);
4155 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4156 file_lock->fl_type = F_WRLCK;
4159 status = nfserr_inval;
4163 status = nfserr_openmode;
4166 file_lock->fl_owner = (fl_owner_t)lock_sop;
4167 file_lock->fl_pid = current->tgid;
4168 file_lock->fl_file = filp;
4169 file_lock->fl_flags = FL_POSIX;
4170 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4171 file_lock->fl_start = lock->lk_offset;
4172 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4173 nfs4_transform_lock_offset(file_lock);
4175 conflock = locks_alloc_lock();
4177 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4178 status = nfserr_jukebox;
4182 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4184 case 0: /* success! */
4185 update_stateid(&lock_stp->st_stid.sc_stateid);
4186 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4190 case (EAGAIN): /* conflock holds conflicting lock */
4191 status = nfserr_denied;
4192 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4193 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4196 status = nfserr_deadlock;
4199 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4200 status = nfserrno(err);
4204 if (status && new_state)
4205 release_lockowner(lock_sop);
4206 if (!cstate->replay_owner)
4207 nfs4_unlock_state();
4209 locks_free_lock(file_lock);
4211 locks_free_lock(conflock);
4216 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4217 * so we do a temporary open here just to get an open file to pass to
4218 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4221 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4224 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4226 err = nfserrno(vfs_test_lock(file, lock));
4236 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4237 struct nfsd4_lockt *lockt)
4239 struct inode *inode;
4240 struct file_lock *file_lock = NULL;
4241 struct nfs4_lockowner *lo;
4243 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4245 if (locks_in_grace(SVC_NET(rqstp)))
4246 return nfserr_grace;
4248 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4249 return nfserr_inval;
4253 status = nfserr_stale_clientid;
4254 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid, nn))
4257 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4260 inode = cstate->current_fh.fh_dentry->d_inode;
4261 file_lock = locks_alloc_lock();
4263 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4264 status = nfserr_jukebox;
4267 locks_init_lock(file_lock);
4268 switch (lockt->lt_type) {
4271 file_lock->fl_type = F_RDLCK;
4274 case NFS4_WRITEW_LT:
4275 file_lock->fl_type = F_WRLCK;
4278 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4279 status = nfserr_inval;
4283 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4285 file_lock->fl_owner = (fl_owner_t)lo;
4286 file_lock->fl_pid = current->tgid;
4287 file_lock->fl_flags = FL_POSIX;
4289 file_lock->fl_start = lockt->lt_offset;
4290 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4292 nfs4_transform_lock_offset(file_lock);
4294 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4298 if (file_lock->fl_type != F_UNLCK) {
4299 status = nfserr_denied;
4300 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4303 nfs4_unlock_state();
4305 locks_free_lock(file_lock);
4310 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4311 struct nfsd4_locku *locku)
4313 struct nfs4_ol_stateid *stp;
4314 struct file *filp = NULL;
4315 struct file_lock *file_lock = NULL;
4319 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4320 (long long) locku->lu_offset,
4321 (long long) locku->lu_length);
4323 if (check_lock_length(locku->lu_offset, locku->lu_length))
4324 return nfserr_inval;
4328 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4329 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4332 filp = find_any_file(stp->st_file);
4334 status = nfserr_lock_range;
4337 file_lock = locks_alloc_lock();
4339 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4340 status = nfserr_jukebox;
4343 locks_init_lock(file_lock);
4344 file_lock->fl_type = F_UNLCK;
4345 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4346 file_lock->fl_pid = current->tgid;
4347 file_lock->fl_file = filp;
4348 file_lock->fl_flags = FL_POSIX;
4349 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4350 file_lock->fl_start = locku->lu_offset;
4352 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4354 nfs4_transform_lock_offset(file_lock);
4357 * Try to unlock the file in the VFS.
4359 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4361 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4365 * OK, unlock succeeded; the only thing left to do is update the stateid.
4367 update_stateid(&stp->st_stid.sc_stateid);
4368 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4371 if (!cstate->replay_owner)
4372 nfs4_unlock_state();
4374 locks_free_lock(file_lock);
4378 status = nfserrno(err);
4384 * 1: locks held by lockowner
4385 * 0: no locks held by lockowner
4388 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4390 struct file_lock **flpp;
4391 struct inode *inode = filp->fi_inode;
4395 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4396 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4407 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4408 struct nfsd4_compound_state *cstate,
4409 struct nfsd4_release_lockowner *rlockowner)
4411 clientid_t *clid = &rlockowner->rl_clientid;
4412 struct nfs4_stateowner *sop;
4413 struct nfs4_lockowner *lo;
4414 struct nfs4_ol_stateid *stp;
4415 struct xdr_netobj *owner = &rlockowner->rl_owner;
4416 struct list_head matches;
4417 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4419 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4421 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4422 clid->cl_boot, clid->cl_id);
4424 /* XXX check for lease expiration */
4426 status = nfserr_stale_clientid;
4427 if (STALE_CLIENTID(clid, nn))
4432 status = nfserr_locks_held;
4433 INIT_LIST_HEAD(&matches);
4435 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4436 if (sop->so_is_open_owner)
4438 if (!same_owner_str(sop, owner, clid))
4440 list_for_each_entry(stp, &sop->so_stateids,
4442 lo = lockowner(sop);
4443 if (check_for_locks(stp->st_file, lo))
4445 list_add(&lo->lo_list, &matches);
4448 /* Clients probably won't expect us to return with some (but not all)
4449 * of the lockowner state released; so don't release any until all
4450 * have been checked. */
4452 while (!list_empty(&matches)) {
4453 lo = list_entry(matches.next, struct nfs4_lockowner,
4455 /* unhash_stateowner deletes so_perclient only
4456 * for openowners. */
4457 list_del(&lo->lo_list);
4458 release_lockowner(lo);
4461 nfs4_unlock_state();
4465 static inline struct nfs4_client_reclaim *
4468 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4472 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4474 unsigned int strhashval = clientstr_hashval(name);
4475 struct nfs4_client *clp;
4477 clp = find_confirmed_client_by_str(name, strhashval);
4480 return test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
4484 * failure => all reset bets are off, nfserr_no_grace...
4487 nfs4_client_to_reclaim(const char *name)
4489 unsigned int strhashval;
4490 struct nfs4_client_reclaim *crp = NULL;
4492 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4493 crp = alloc_reclaim();
4496 strhashval = clientstr_hashval(name);
4497 INIT_LIST_HEAD(&crp->cr_strhash);
4498 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4499 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4500 reclaim_str_hashtbl_size++;
4505 nfs4_release_reclaim(void)
4507 struct nfs4_client_reclaim *crp = NULL;
4510 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4511 while (!list_empty(&reclaim_str_hashtbl[i])) {
4512 crp = list_entry(reclaim_str_hashtbl[i].next,
4513 struct nfs4_client_reclaim, cr_strhash);
4514 list_del(&crp->cr_strhash);
4516 reclaim_str_hashtbl_size--;
4519 BUG_ON(reclaim_str_hashtbl_size);
4523 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4524 struct nfs4_client_reclaim *
4525 nfsd4_find_reclaim_client(struct nfs4_client *clp)
4527 unsigned int strhashval;
4528 struct nfs4_client_reclaim *crp = NULL;
4530 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4531 clp->cl_name.len, clp->cl_name.data,
4534 /* find clp->cl_name in reclaim_str_hashtbl */
4535 strhashval = clientstr_hashval(clp->cl_recdir);
4536 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4537 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4545 * Called from OPEN. Look for clientid in reclaim list.
4548 nfs4_check_open_reclaim(clientid_t *clid, bool sessions)
4550 struct nfs4_client *clp;
4552 /* find clientid in conf_id_hashtbl */
4553 clp = find_confirmed_client(clid, sessions);
4555 return nfserr_reclaim_bad;
4557 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4560 #ifdef CONFIG_NFSD_FAULT_INJECTION
4562 void nfsd_forget_clients(u64 num)
4564 struct nfs4_client *clp, *next;
4568 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4573 nfs4_unlock_state();
4575 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4578 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4580 release_lockowner(lockowner(sop));
4583 static void release_openowner_sop(struct nfs4_stateowner *sop)
4585 release_openowner(openowner(sop));
4588 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4589 void (*release_sop)(struct nfs4_stateowner *))
4592 struct nfs4_stateowner *sop, *next;
4594 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4595 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4596 if (sop->so_is_open_owner != is_open_owner)
4606 void nfsd_forget_locks(u64 num)
4611 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4612 nfs4_unlock_state();
4614 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4617 void nfsd_forget_openowners(u64 num)
4622 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4623 nfs4_unlock_state();
4625 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4628 static int nfsd_process_n_delegations(u64 num, struct list_head *list)
4631 struct nfs4_file *fp, *fnext;
4632 struct nfs4_delegation *dp, *dnext;
4634 for (i = 0; i < FILE_HASH_SIZE; i++) {
4635 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4636 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4637 list_move(&dp->dl_recall_lru, list);
4647 void nfsd_forget_delegations(u64 num)
4651 struct nfs4_delegation *dp, *dnext;
4653 spin_lock(&recall_lock);
4654 count = nfsd_process_n_delegations(num, &victims);
4655 spin_unlock(&recall_lock);
4658 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
4659 unhash_delegation(dp);
4660 nfs4_unlock_state();
4662 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4665 void nfsd_recall_delegations(u64 num)
4669 struct nfs4_delegation *dp, *dnext;
4671 spin_lock(&recall_lock);
4672 count = nfsd_process_n_delegations(num, &victims);
4673 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
4674 list_del(&dp->dl_recall_lru);
4675 nfsd_break_one_deleg(dp);
4677 spin_unlock(&recall_lock);
4679 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4682 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4684 /* initialization to perform at module load time: */
4687 nfs4_state_init(void)
4691 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4692 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4693 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4694 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4695 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4696 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4698 for (i = 0; i < SESSION_HASH_SIZE; i++)
4699 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4700 for (i = 0; i < FILE_HASH_SIZE; i++) {
4701 INIT_LIST_HEAD(&file_hashtbl[i]);
4703 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4704 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4706 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4707 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4708 INIT_LIST_HEAD(&close_lru);
4709 INIT_LIST_HEAD(&client_lru);
4710 INIT_LIST_HEAD(&del_recall_lru);
4711 reclaim_str_hashtbl_size = 0;
4715 * Since the lifetime of a delegation isn't limited to that of an open, a
4716 * client may quite reasonably hang on to a delegation as long as it has
4717 * the inode cached. This becomes an obvious problem the first time a
4718 * client's inode cache approaches the size of the server's total memory.
4720 * For now we avoid this problem by imposing a hard limit on the number
4721 * of delegations, which varies according to the server's memory size.
4724 set_max_delegations(void)
4727 * Allow at most 4 delegations per megabyte of RAM. Quick
4728 * estimates suggest that in the worst case (where every delegation
4729 * is for a different inode), a delegation could take about 1.5K,
4730 * giving a worst case usage of about 6% of memory.
4732 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4735 /* initialization to perform when the nfsd service is started: */
4738 nfs4_state_start(void)
4740 struct net *net = &init_net;
4741 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4745 * FIXME: For now, we hang most of the pernet global stuff off of
4746 * init_net until nfsd is fully containerized. Eventually, we'll
4747 * need to pass a net pointer into this function, take a reference
4748 * to that instead and then do most of the rest of this on a per-net
4752 nfsd4_client_tracking_init(net);
4753 nn->boot_time = get_seconds();
4754 locks_start_grace(net, &nn->nfsd4_manager);
4755 nn->grace_ended = false;
4756 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4758 ret = set_callback_cred();
4763 laundry_wq = create_singlethread_workqueue("nfsd4");
4764 if (laundry_wq == NULL) {
4768 ret = nfsd4_create_callback_queue();
4770 goto out_free_laundry;
4771 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4772 set_max_delegations();
4775 destroy_workqueue(laundry_wq);
4777 nfsd4_client_tracking_exit(net);
4783 __nfs4_state_shutdown(void)
4786 struct nfs4_client *clp = NULL;
4787 struct nfs4_delegation *dp = NULL;
4788 struct list_head *pos, *next, reaplist;
4790 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4791 while (!list_empty(&conf_id_hashtbl[i])) {
4792 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4793 destroy_client(clp);
4795 while (!list_empty(&unconf_str_hashtbl[i])) {
4796 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4797 destroy_client(clp);
4800 INIT_LIST_HEAD(&reaplist);
4801 spin_lock(&recall_lock);
4802 list_for_each_safe(pos, next, &del_recall_lru) {
4803 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4804 list_move(&dp->dl_recall_lru, &reaplist);
4806 spin_unlock(&recall_lock);
4807 list_for_each_safe(pos, next, &reaplist) {
4808 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4809 unhash_delegation(dp);
4812 nfsd4_client_tracking_exit(&init_net);
4817 nfs4_state_shutdown(void)
4819 struct net *net = &init_net;
4820 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4822 cancel_delayed_work_sync(&laundromat_work);
4823 destroy_workqueue(laundry_wq);
4824 locks_end_grace(&nn->nfsd4_manager);
4826 __nfs4_state_shutdown();
4827 nfs4_unlock_state();
4828 nfsd4_destroy_callback_queue();
4832 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4834 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4835 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4839 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4841 if (cstate->minorversion) {
4842 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4843 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4848 clear_current_stateid(struct nfsd4_compound_state *cstate)
4850 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4854 * functions to set current state id
4857 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4859 put_stateid(cstate, &odp->od_stateid);
4863 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4865 put_stateid(cstate, &open->op_stateid);
4869 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4871 put_stateid(cstate, &close->cl_stateid);
4875 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4877 put_stateid(cstate, &lock->lk_resp_stateid);
4881 * functions to consume current state id
4885 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4887 get_stateid(cstate, &odp->od_stateid);
4891 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4893 get_stateid(cstate, &drp->dr_stateid);
4897 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4899 get_stateid(cstate, &fsp->fr_stateid);
4903 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4905 get_stateid(cstate, &setattr->sa_stateid);
4909 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4911 get_stateid(cstate, &close->cl_stateid);
4915 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4917 get_stateid(cstate, &locku->lu_stateid);
4921 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4923 get_stateid(cstate, &read->rd_stateid);
4927 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4929 get_stateid(cstate, &write->wr_stateid);