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 __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
857 struct nfsd4_conn *conn;
860 conn = alloc_conn(rqstp, dir);
862 return nfserr_jukebox;
863 nfsd4_hash_conn(conn, ses);
864 ret = nfsd4_register_conn(conn);
866 /* oops; xprt is already down: */
867 nfsd4_conn_lost(&conn->cn_xpt_user);
868 if (ses->se_client->cl_cb_state == NFSD4_CB_DOWN &&
869 dir & NFS4_CDFC4_BACK) {
870 /* callback channel may be back up */
871 nfsd4_probe_callback(ses->se_client);
876 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
878 u32 dir = NFS4_CDFC4_FORE;
880 if (ses->se_flags & SESSION4_BACK_CHAN)
881 dir |= NFS4_CDFC4_BACK;
883 return nfsd4_new_conn(rqstp, ses, dir);
886 /* must be called under client_lock */
887 static void nfsd4_del_conns(struct nfsd4_session *s)
889 struct nfs4_client *clp = s->se_client;
890 struct nfsd4_conn *c;
892 spin_lock(&clp->cl_lock);
893 while (!list_empty(&s->se_conns)) {
894 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
895 list_del_init(&c->cn_persession);
896 spin_unlock(&clp->cl_lock);
898 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
901 spin_lock(&clp->cl_lock);
903 spin_unlock(&clp->cl_lock);
906 static void free_session(struct kref *kref)
908 struct nfsd4_session *ses;
911 lockdep_assert_held(&client_lock);
912 ses = container_of(kref, struct nfsd4_session, se_ref);
913 nfsd4_del_conns(ses);
914 spin_lock(&nfsd_drc_lock);
915 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
916 nfsd_drc_mem_used -= mem;
917 spin_unlock(&nfsd_drc_lock);
918 free_session_slots(ses);
922 void nfsd4_put_session(struct nfsd4_session *ses)
924 spin_lock(&client_lock);
925 nfsd4_put_session_locked(ses);
926 spin_unlock(&client_lock);
929 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
931 struct nfsd4_session *new;
932 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
933 int numslots, slotsize;
938 * Note decreasing slot size below client's request may
939 * make it difficult for client to function correctly, whereas
940 * decreasing the number of slots will (just?) affect
941 * performance. When short on memory we therefore prefer to
942 * decrease number of slots instead of their size.
944 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
945 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
949 new = alloc_session(slotsize, numslots);
951 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
954 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
956 new->se_client = clp;
959 INIT_LIST_HEAD(&new->se_conns);
961 new->se_cb_seq_nr = 1;
962 new->se_flags = cses->flags;
963 new->se_cb_prog = cses->callback_prog;
964 kref_init(&new->se_ref);
965 idx = hash_sessionid(&new->se_sessionid);
966 spin_lock(&client_lock);
967 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
968 spin_lock(&clp->cl_lock);
969 list_add(&new->se_perclnt, &clp->cl_sessions);
970 spin_unlock(&clp->cl_lock);
971 spin_unlock(&client_lock);
973 status = nfsd4_new_conn_from_crses(rqstp, new);
974 /* whoops: benny points out, status is ignored! (err, or bogus) */
976 spin_lock(&client_lock);
977 free_session(&new->se_ref);
978 spin_unlock(&client_lock);
981 if (cses->flags & SESSION4_BACK_CHAN) {
982 struct sockaddr *sa = svc_addr(rqstp);
984 * This is a little silly; with sessions there's no real
985 * use for the callback address. Use the peer address
986 * as a reasonable default for now, but consider fixing
987 * the rpc client not to require an address in the
990 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
991 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
993 nfsd4_probe_callback(clp);
997 /* caller must hold client_lock */
998 static struct nfsd4_session *
999 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
1001 struct nfsd4_session *elem;
1004 dump_sessionid(__func__, sessionid);
1005 idx = hash_sessionid(sessionid);
1006 /* Search in the appropriate list */
1007 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
1008 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1009 NFS4_MAX_SESSIONID_LEN)) {
1014 dprintk("%s: session not found\n", __func__);
1018 /* caller must hold client_lock */
1020 unhash_session(struct nfsd4_session *ses)
1022 list_del(&ses->se_hash);
1023 spin_lock(&ses->se_client->cl_lock);
1024 list_del(&ses->se_perclnt);
1025 spin_unlock(&ses->se_client->cl_lock);
1028 /* must be called under the client_lock */
1030 renew_client_locked(struct nfs4_client *clp)
1032 if (is_client_expired(clp)) {
1034 printk("%s: client (clientid %08x/%08x) already expired\n",
1036 clp->cl_clientid.cl_boot,
1037 clp->cl_clientid.cl_id);
1041 dprintk("renewing client (clientid %08x/%08x)\n",
1042 clp->cl_clientid.cl_boot,
1043 clp->cl_clientid.cl_id);
1044 list_move_tail(&clp->cl_lru, &client_lru);
1045 clp->cl_time = get_seconds();
1049 renew_client(struct nfs4_client *clp)
1051 spin_lock(&client_lock);
1052 renew_client_locked(clp);
1053 spin_unlock(&client_lock);
1056 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1058 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1060 if (clid->cl_boot == nn->boot_time)
1062 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1063 clid->cl_boot, clid->cl_id, nn->boot_time);
1068 * XXX Should we use a slab cache ?
1069 * This type of memory management is somewhat inefficient, but we use it
1070 * anyway since SETCLIENTID is not a common operation.
1072 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1074 struct nfs4_client *clp;
1076 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1079 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1080 if (clp->cl_name.data == NULL) {
1084 clp->cl_name.len = name.len;
1089 free_client(struct nfs4_client *clp)
1091 lockdep_assert_held(&client_lock);
1092 while (!list_empty(&clp->cl_sessions)) {
1093 struct nfsd4_session *ses;
1094 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1096 list_del(&ses->se_perclnt);
1097 nfsd4_put_session_locked(ses);
1099 free_svc_cred(&clp->cl_cred);
1100 kfree(clp->cl_name.data);
1105 release_session_client(struct nfsd4_session *session)
1107 struct nfs4_client *clp = session->se_client;
1109 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1111 if (is_client_expired(clp)) {
1113 session->se_client = NULL;
1115 renew_client_locked(clp);
1116 spin_unlock(&client_lock);
1119 /* must be called under the client_lock */
1121 unhash_client_locked(struct nfs4_client *clp)
1123 struct nfsd4_session *ses;
1125 mark_client_expired(clp);
1126 list_del(&clp->cl_lru);
1127 spin_lock(&clp->cl_lock);
1128 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1129 list_del_init(&ses->se_hash);
1130 spin_unlock(&clp->cl_lock);
1134 expire_client(struct nfs4_client *clp)
1136 struct nfs4_openowner *oo;
1137 struct nfs4_delegation *dp;
1138 struct list_head reaplist;
1140 INIT_LIST_HEAD(&reaplist);
1141 spin_lock(&recall_lock);
1142 while (!list_empty(&clp->cl_delegations)) {
1143 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1144 list_del_init(&dp->dl_perclnt);
1145 list_move(&dp->dl_recall_lru, &reaplist);
1147 spin_unlock(&recall_lock);
1148 while (!list_empty(&reaplist)) {
1149 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1150 unhash_delegation(dp);
1152 while (!list_empty(&clp->cl_openowners)) {
1153 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1154 release_openowner(oo);
1156 nfsd4_shutdown_callback(clp);
1157 if (clp->cl_cb_conn.cb_xprt)
1158 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1159 list_del(&clp->cl_idhash);
1160 list_del(&clp->cl_strhash);
1161 spin_lock(&client_lock);
1162 unhash_client_locked(clp);
1163 if (atomic_read(&clp->cl_refcount) == 0)
1165 spin_unlock(&client_lock);
1168 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1170 memcpy(target->cl_verifier.data, source->data,
1171 sizeof(target->cl_verifier.data));
1174 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1176 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1177 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1180 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1182 if (source->cr_principal) {
1183 target->cr_principal =
1184 kstrdup(source->cr_principal, GFP_KERNEL);
1185 if (target->cr_principal == NULL)
1188 target->cr_principal = NULL;
1189 target->cr_flavor = source->cr_flavor;
1190 target->cr_uid = source->cr_uid;
1191 target->cr_gid = source->cr_gid;
1192 target->cr_group_info = source->cr_group_info;
1193 get_group_info(target->cr_group_info);
1197 static int same_name(const char *n1, const char *n2)
1199 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1203 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1205 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1209 same_clid(clientid_t *cl1, clientid_t *cl2)
1211 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1214 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1218 if (g1->ngroups != g2->ngroups)
1220 for (i=0; i<g1->ngroups; i++)
1221 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1227 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1229 if ((cr1->cr_flavor != cr2->cr_flavor)
1230 || (cr1->cr_uid != cr2->cr_uid)
1231 || (cr1->cr_gid != cr2->cr_gid)
1232 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1234 if (cr1->cr_principal == cr2->cr_principal)
1236 if (!cr1->cr_principal || !cr2->cr_principal)
1238 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1241 static void gen_clid(struct nfs4_client *clp)
1243 static u32 current_clientid = 1;
1244 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
1246 clp->cl_clientid.cl_boot = nn->boot_time;
1247 clp->cl_clientid.cl_id = current_clientid++;
1250 static void gen_confirm(struct nfs4_client *clp)
1255 verf[0] = (__be32)get_seconds();
1256 verf[1] = (__be32)i++;
1257 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1260 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1262 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1265 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1267 struct nfs4_stid *s;
1269 s = find_stateid(cl, t);
1272 if (typemask & s->sc_type)
1277 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1278 struct svc_rqst *rqstp, nfs4_verifier *verf)
1280 struct nfs4_client *clp;
1281 struct sockaddr *sa = svc_addr(rqstp);
1284 clp = alloc_client(name);
1288 INIT_LIST_HEAD(&clp->cl_sessions);
1289 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1291 spin_lock(&client_lock);
1293 spin_unlock(&client_lock);
1296 idr_init(&clp->cl_stateids);
1297 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1298 atomic_set(&clp->cl_refcount, 0);
1299 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1300 INIT_LIST_HEAD(&clp->cl_idhash);
1301 INIT_LIST_HEAD(&clp->cl_strhash);
1302 INIT_LIST_HEAD(&clp->cl_openowners);
1303 INIT_LIST_HEAD(&clp->cl_delegations);
1304 INIT_LIST_HEAD(&clp->cl_lru);
1305 INIT_LIST_HEAD(&clp->cl_callbacks);
1306 spin_lock_init(&clp->cl_lock);
1307 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1308 clp->cl_time = get_seconds();
1309 clear_bit(0, &clp->cl_cb_slot_busy);
1310 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1311 copy_verf(clp, verf);
1312 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1314 clp->cl_cb_session = NULL;
1319 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1321 unsigned int idhashval;
1323 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1324 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1325 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1330 move_to_confirmed(struct nfs4_client *clp)
1332 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1333 unsigned int strhashval;
1335 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1336 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1337 strhashval = clientstr_hashval(clp->cl_recdir);
1338 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1342 static struct nfs4_client *
1343 find_confirmed_client(clientid_t *clid)
1345 struct nfs4_client *clp;
1346 unsigned int idhashval = clientid_hashval(clid->cl_id);
1348 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1349 if (same_clid(&clp->cl_clientid, clid)) {
1357 static struct nfs4_client *
1358 find_unconfirmed_client(clientid_t *clid)
1360 struct nfs4_client *clp;
1361 unsigned int idhashval = clientid_hashval(clid->cl_id);
1363 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1364 if (same_clid(&clp->cl_clientid, clid))
1370 static bool clp_used_exchangeid(struct nfs4_client *clp)
1372 return clp->cl_exchange_flags != 0;
1375 static struct nfs4_client *
1376 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1378 struct nfs4_client *clp;
1380 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1381 if (same_name(clp->cl_recdir, dname))
1387 static struct nfs4_client *
1388 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1390 struct nfs4_client *clp;
1392 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1393 if (same_name(clp->cl_recdir, dname))
1400 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1402 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1403 struct sockaddr *sa = svc_addr(rqstp);
1404 u32 scopeid = rpc_get_scope_id(sa);
1405 unsigned short expected_family;
1407 /* Currently, we only support tcp and tcp6 for the callback channel */
1408 if (se->se_callback_netid_len == 3 &&
1409 !memcmp(se->se_callback_netid_val, "tcp", 3))
1410 expected_family = AF_INET;
1411 else if (se->se_callback_netid_len == 4 &&
1412 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1413 expected_family = AF_INET6;
1417 conn->cb_addrlen = rpc_uaddr2sockaddr(&init_net, se->se_callback_addr_val,
1418 se->se_callback_addr_len,
1419 (struct sockaddr *)&conn->cb_addr,
1420 sizeof(conn->cb_addr));
1422 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1425 if (conn->cb_addr.ss_family == AF_INET6)
1426 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1428 conn->cb_prog = se->se_callback_prog;
1429 conn->cb_ident = se->se_callback_ident;
1430 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1433 conn->cb_addr.ss_family = AF_UNSPEC;
1434 conn->cb_addrlen = 0;
1435 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1436 "will not receive delegations\n",
1437 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1443 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1446 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1448 struct nfsd4_slot *slot = resp->cstate.slot;
1451 dprintk("--> %s slot %p\n", __func__, slot);
1453 slot->sl_opcnt = resp->opcnt;
1454 slot->sl_status = resp->cstate.status;
1456 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1457 if (nfsd4_not_cached(resp)) {
1458 slot->sl_datalen = 0;
1461 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1462 base = (char *)resp->cstate.datap -
1463 (char *)resp->xbuf->head[0].iov_base;
1464 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1466 WARN("%s: sessions DRC could not cache compound\n", __func__);
1471 * Encode the replay sequence operation from the slot values.
1472 * If cachethis is FALSE encode the uncached rep error on the next
1473 * operation which sets resp->p and increments resp->opcnt for
1474 * nfs4svc_encode_compoundres.
1478 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1479 struct nfsd4_compoundres *resp)
1481 struct nfsd4_op *op;
1482 struct nfsd4_slot *slot = resp->cstate.slot;
1484 /* Encode the replayed sequence operation */
1485 op = &args->ops[resp->opcnt - 1];
1486 nfsd4_encode_operation(resp, op);
1488 /* Return nfserr_retry_uncached_rep in next operation. */
1489 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1490 op = &args->ops[resp->opcnt++];
1491 op->status = nfserr_retry_uncached_rep;
1492 nfsd4_encode_operation(resp, op);
1498 * The sequence operation is not cached because we can use the slot and
1502 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1503 struct nfsd4_sequence *seq)
1505 struct nfsd4_slot *slot = resp->cstate.slot;
1508 dprintk("--> %s slot %p\n", __func__, slot);
1510 /* Either returns 0 or nfserr_retry_uncached */
1511 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1512 if (status == nfserr_retry_uncached_rep)
1515 /* The sequence operation has been encoded, cstate->datap set. */
1516 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1518 resp->opcnt = slot->sl_opcnt;
1519 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1520 status = slot->sl_status;
1526 * Set the exchange_id flags returned by the server.
1529 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1531 /* pNFS is not supported */
1532 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1534 /* Referrals are supported, Migration is not. */
1535 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1537 /* set the wire flags to return to client. */
1538 clid->flags = new->cl_exchange_flags;
1541 static bool client_has_state(struct nfs4_client *clp)
1544 * Note clp->cl_openowners check isn't quite right: there's no
1545 * need to count owners without stateid's.
1547 * Also note we should probably be using this in 4.0 case too.
1549 return !list_empty(&clp->cl_openowners)
1550 || !list_empty(&clp->cl_delegations)
1551 || !list_empty(&clp->cl_sessions);
1555 nfsd4_exchange_id(struct svc_rqst *rqstp,
1556 struct nfsd4_compound_state *cstate,
1557 struct nfsd4_exchange_id *exid)
1559 struct nfs4_client *unconf, *conf, *new;
1561 unsigned int strhashval;
1562 char dname[HEXDIR_LEN];
1563 char addr_str[INET6_ADDRSTRLEN];
1564 nfs4_verifier verf = exid->verifier;
1565 struct sockaddr *sa = svc_addr(rqstp);
1566 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1568 rpc_ntop(sa, addr_str, sizeof(addr_str));
1569 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1570 "ip_addr=%s flags %x, spa_how %d\n",
1571 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1572 addr_str, exid->flags, exid->spa_how);
1574 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1575 return nfserr_inval;
1577 /* Currently only support SP4_NONE */
1578 switch (exid->spa_how) {
1582 return nfserr_serverfault;
1584 BUG(); /* checked by xdr code */
1586 return nfserr_serverfault; /* no excuse :-/ */
1589 status = nfs4_make_rec_clidname(dname, &exid->clname);
1594 strhashval = clientstr_hashval(dname);
1596 /* Cases below refer to rfc 5661 section 18.35.4: */
1598 conf = find_confirmed_client_by_str(dname, strhashval);
1600 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1601 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1604 if (!clp_used_exchangeid(conf)) { /* buggy client */
1605 status = nfserr_inval;
1608 if (!creds_match) { /* case 9 */
1609 status = nfserr_perm;
1612 if (!verfs_match) { /* case 8 */
1613 status = nfserr_not_same;
1617 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1621 if (!creds_match) { /* case 3 */
1622 if (client_has_state(conf)) {
1623 status = nfserr_clid_inuse;
1626 expire_client(conf);
1629 if (verfs_match) { /* case 2 */
1630 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1634 /* case 5, client reboot */
1638 if (update) { /* case 7 */
1639 status = nfserr_noent;
1643 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1644 if (unconf) /* case 4, possible retry or client restart */
1645 expire_client(unconf);
1647 /* case 1 (normal case) */
1649 new = create_client(exid->clname, dname, rqstp, &verf);
1651 status = nfserr_jukebox;
1656 add_to_unconfirmed(new, strhashval);
1658 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1659 exid->clientid.cl_id = new->cl_clientid.cl_id;
1661 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1662 nfsd4_set_ex_flags(new, exid);
1664 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1665 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1669 nfs4_unlock_state();
1674 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1676 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1679 /* The slot is in use, and no response has been sent. */
1681 if (seqid == slot_seqid)
1682 return nfserr_jukebox;
1684 return nfserr_seq_misordered;
1686 /* Note unsigned 32-bit arithmetic handles wraparound: */
1687 if (likely(seqid == slot_seqid + 1))
1689 if (seqid == slot_seqid)
1690 return nfserr_replay_cache;
1691 return nfserr_seq_misordered;
1695 * Cache the create session result into the create session single DRC
1696 * slot cache by saving the xdr structure. sl_seqid has been set.
1697 * Do this for solo or embedded create session operations.
1700 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1701 struct nfsd4_clid_slot *slot, __be32 nfserr)
1703 slot->sl_status = nfserr;
1704 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1708 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1709 struct nfsd4_clid_slot *slot)
1711 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1712 return slot->sl_status;
1715 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1716 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1717 1 + /* MIN tag is length with zero, only length */ \
1718 3 + /* version, opcount, opcode */ \
1719 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1720 /* seqid, slotID, slotID, cache */ \
1721 4 ) * sizeof(__be32))
1723 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1724 2 + /* verifier: AUTH_NULL, length 0 */\
1726 1 + /* MIN tag is length with zero, only length */ \
1727 3 + /* opcount, opcode, opstatus*/ \
1728 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1729 /* seqid, slotID, slotID, slotID, status */ \
1730 5 ) * sizeof(__be32))
1732 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1734 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1735 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1739 nfsd4_create_session(struct svc_rqst *rqstp,
1740 struct nfsd4_compound_state *cstate,
1741 struct nfsd4_create_session *cr_ses)
1743 struct sockaddr *sa = svc_addr(rqstp);
1744 struct nfs4_client *conf, *unconf;
1745 struct nfsd4_session *new;
1746 struct nfsd4_clid_slot *cs_slot = NULL;
1747 bool confirm_me = false;
1750 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1751 return nfserr_inval;
1754 unconf = find_unconfirmed_client(&cr_ses->clientid);
1755 conf = find_confirmed_client(&cr_ses->clientid);
1758 cs_slot = &conf->cl_cs_slot;
1759 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1760 if (status == nfserr_replay_cache) {
1761 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1763 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1764 status = nfserr_seq_misordered;
1767 } else if (unconf) {
1768 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1769 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1770 status = nfserr_clid_inuse;
1773 cs_slot = &unconf->cl_cs_slot;
1774 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1776 /* an unconfirmed replay returns misordered */
1777 status = nfserr_seq_misordered;
1783 status = nfserr_stale_clientid;
1788 * XXX: we should probably set this at creation time, and check
1789 * for consistent minorversion use throughout:
1791 conf->cl_minorversion = 1;
1793 * We do not support RDMA or persistent sessions
1795 cr_ses->flags &= ~SESSION4_PERSIST;
1796 cr_ses->flags &= ~SESSION4_RDMA;
1798 status = nfserr_toosmall;
1799 if (check_forechannel_attrs(cr_ses->fore_channel))
1802 status = nfserr_jukebox;
1803 new = alloc_init_session(rqstp, conf, cr_ses);
1807 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1808 NFS4_MAX_SESSIONID_LEN);
1809 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1810 sizeof(struct nfsd4_channel_attrs));
1811 cs_slot->sl_seqid++;
1812 cr_ses->seqid = cs_slot->sl_seqid;
1814 /* cache solo and embedded create sessions under the state lock */
1815 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1817 unsigned int hash = clientstr_hashval(unconf->cl_recdir);
1818 struct nfs4_client *old =
1819 find_confirmed_client_by_str(conf->cl_recdir, hash);
1822 move_to_confirmed(conf);
1825 nfs4_unlock_state();
1826 dprintk("%s returns %d\n", __func__, ntohl(status));
1830 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1832 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1833 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1835 return argp->opcnt == resp->opcnt;
1838 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1841 case NFS4_CDFC4_FORE:
1842 case NFS4_CDFC4_BACK:
1844 case NFS4_CDFC4_FORE_OR_BOTH:
1845 case NFS4_CDFC4_BACK_OR_BOTH:
1846 *dir = NFS4_CDFC4_BOTH;
1849 return nfserr_inval;
1852 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1853 struct nfsd4_compound_state *cstate,
1854 struct nfsd4_bind_conn_to_session *bcts)
1858 if (!nfsd4_last_compound_op(rqstp))
1859 return nfserr_not_only_op;
1860 spin_lock(&client_lock);
1861 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1862 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1863 * client_lock iself: */
1864 if (cstate->session) {
1865 nfsd4_get_session(cstate->session);
1866 atomic_inc(&cstate->session->se_client->cl_refcount);
1868 spin_unlock(&client_lock);
1869 if (!cstate->session)
1870 return nfserr_badsession;
1872 status = nfsd4_map_bcts_dir(&bcts->dir);
1874 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1878 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1882 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1886 nfsd4_destroy_session(struct svc_rqst *r,
1887 struct nfsd4_compound_state *cstate,
1888 struct nfsd4_destroy_session *sessionid)
1890 struct nfsd4_session *ses;
1891 __be32 status = nfserr_badsession;
1894 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1895 * - Should we return nfserr_back_chan_busy if waiting for
1896 * callbacks on to-be-destroyed session?
1897 * - Do we need to clear any callback info from previous session?
1900 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1901 if (!nfsd4_last_compound_op(r))
1902 return nfserr_not_only_op;
1904 dump_sessionid(__func__, &sessionid->sessionid);
1905 spin_lock(&client_lock);
1906 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1908 spin_unlock(&client_lock);
1912 unhash_session(ses);
1913 spin_unlock(&client_lock);
1916 nfsd4_probe_callback_sync(ses->se_client);
1917 nfs4_unlock_state();
1919 spin_lock(&client_lock);
1920 nfsd4_del_conns(ses);
1921 nfsd4_put_session_locked(ses);
1922 spin_unlock(&client_lock);
1925 dprintk("%s returns %d\n", __func__, ntohl(status));
1929 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1931 struct nfsd4_conn *c;
1933 list_for_each_entry(c, &s->se_conns, cn_persession) {
1934 if (c->cn_xprt == xpt) {
1941 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1943 struct nfs4_client *clp = ses->se_client;
1944 struct nfsd4_conn *c;
1947 spin_lock(&clp->cl_lock);
1948 c = __nfsd4_find_conn(new->cn_xprt, ses);
1950 spin_unlock(&clp->cl_lock);
1954 __nfsd4_hash_conn(new, ses);
1955 spin_unlock(&clp->cl_lock);
1956 ret = nfsd4_register_conn(new);
1958 /* oops; xprt is already down: */
1959 nfsd4_conn_lost(&new->cn_xpt_user);
1963 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1965 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1967 return args->opcnt > session->se_fchannel.maxops;
1970 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1971 struct nfsd4_session *session)
1973 struct xdr_buf *xb = &rqstp->rq_arg;
1975 return xb->len > session->se_fchannel.maxreq_sz;
1979 nfsd4_sequence(struct svc_rqst *rqstp,
1980 struct nfsd4_compound_state *cstate,
1981 struct nfsd4_sequence *seq)
1983 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1984 struct nfsd4_session *session;
1985 struct nfsd4_slot *slot;
1986 struct nfsd4_conn *conn;
1989 if (resp->opcnt != 1)
1990 return nfserr_sequence_pos;
1993 * Will be either used or freed by nfsd4_sequence_check_conn
1996 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1998 return nfserr_jukebox;
2000 spin_lock(&client_lock);
2001 status = nfserr_badsession;
2002 session = find_in_sessionid_hashtbl(&seq->sessionid);
2006 status = nfserr_too_many_ops;
2007 if (nfsd4_session_too_many_ops(rqstp, session))
2010 status = nfserr_req_too_big;
2011 if (nfsd4_request_too_big(rqstp, session))
2014 status = nfserr_badslot;
2015 if (seq->slotid >= session->se_fchannel.maxreqs)
2018 slot = session->se_slots[seq->slotid];
2019 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2021 /* We do not negotiate the number of slots yet, so set the
2022 * maxslots to the session maxreqs which is used to encode
2023 * sr_highest_slotid and the sr_target_slot id to maxslots */
2024 seq->maxslots = session->se_fchannel.maxreqs;
2026 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2027 slot->sl_flags & NFSD4_SLOT_INUSE);
2028 if (status == nfserr_replay_cache) {
2029 status = nfserr_seq_misordered;
2030 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2032 cstate->slot = slot;
2033 cstate->session = session;
2034 /* Return the cached reply status and set cstate->status
2035 * for nfsd4_proc_compound processing */
2036 status = nfsd4_replay_cache_entry(resp, seq);
2037 cstate->status = nfserr_replay_cache;
2043 nfsd4_sequence_check_conn(conn, session);
2046 /* Success! bump slot seqid */
2047 slot->sl_seqid = seq->seqid;
2048 slot->sl_flags |= NFSD4_SLOT_INUSE;
2050 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2052 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2054 cstate->slot = slot;
2055 cstate->session = session;
2058 /* Hold a session reference until done processing the compound. */
2059 if (cstate->session) {
2060 struct nfs4_client *clp = session->se_client;
2062 nfsd4_get_session(cstate->session);
2063 atomic_inc(&clp->cl_refcount);
2064 switch (clp->cl_cb_state) {
2066 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2068 case NFSD4_CB_FAULT:
2069 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2072 seq->status_flags = 0;
2076 spin_unlock(&client_lock);
2077 dprintk("%s: return %d\n", __func__, ntohl(status));
2082 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2084 struct nfs4_client *conf, *unconf, *clp;
2088 unconf = find_unconfirmed_client(&dc->clientid);
2089 conf = find_confirmed_client(&dc->clientid);
2094 if (!is_client_expired(conf) && client_has_state(conf)) {
2095 status = nfserr_clientid_busy;
2099 /* rfc5661 18.50.3 */
2100 if (cstate->session && conf == cstate->session->se_client) {
2101 status = nfserr_clientid_busy;
2107 status = nfserr_stale_clientid;
2113 nfs4_unlock_state();
2114 dprintk("%s return %d\n", __func__, ntohl(status));
2119 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2123 if (rc->rca_one_fs) {
2124 if (!cstate->current_fh.fh_dentry)
2125 return nfserr_nofilehandle;
2127 * We don't take advantage of the rca_one_fs case.
2128 * That's OK, it's optional, we can safely ignore it.
2134 status = nfserr_complete_already;
2135 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2136 &cstate->session->se_client->cl_flags))
2139 status = nfserr_stale_clientid;
2140 if (is_client_expired(cstate->session->se_client))
2142 * The following error isn't really legal.
2143 * But we only get here if the client just explicitly
2144 * destroyed the client. Surely it no longer cares what
2145 * error it gets back on an operation for the dead
2151 nfsd4_client_record_create(cstate->session->se_client);
2153 nfs4_unlock_state();
2158 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2159 struct nfsd4_setclientid *setclid)
2161 struct xdr_netobj clname = setclid->se_name;
2162 nfs4_verifier clverifier = setclid->se_verf;
2163 unsigned int strhashval;
2164 struct nfs4_client *conf, *unconf, *new;
2166 char dname[HEXDIR_LEN];
2168 status = nfs4_make_rec_clidname(dname, &clname);
2172 strhashval = clientstr_hashval(dname);
2174 /* Cases below refer to rfc 3530 section 14.2.33: */
2176 conf = find_confirmed_client_by_str(dname, strhashval);
2179 status = nfserr_clid_inuse;
2180 if (clp_used_exchangeid(conf))
2182 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2183 char addr_str[INET6_ADDRSTRLEN];
2184 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2186 dprintk("NFSD: setclientid: string in use by client "
2187 "at %s\n", addr_str);
2191 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2193 expire_client(unconf);
2194 status = nfserr_jukebox;
2195 new = create_client(clname, dname, rqstp, &clverifier);
2198 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2199 /* case 1: probable callback update */
2200 copy_clid(new, conf);
2201 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2204 * XXX: we should probably set this at creation time, and check
2205 * for consistent minorversion use throughout:
2207 new->cl_minorversion = 0;
2208 gen_callback(new, setclid, rqstp);
2209 add_to_unconfirmed(new, strhashval);
2210 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2211 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2212 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2215 nfs4_unlock_state();
2221 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2222 struct nfsd4_compound_state *cstate,
2223 struct nfsd4_setclientid_confirm *setclientid_confirm)
2225 struct nfs4_client *conf, *unconf;
2226 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2227 clientid_t * clid = &setclientid_confirm->sc_clientid;
2229 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2231 if (STALE_CLIENTID(clid, nn))
2232 return nfserr_stale_clientid;
2235 conf = find_confirmed_client(clid);
2236 unconf = find_unconfirmed_client(clid);
2238 * We try hard to give out unique clientid's, so if we get an
2239 * attempt to confirm the same clientid with a different cred,
2240 * there's a bug somewhere. Let's charitably assume it's our
2243 status = nfserr_serverfault;
2244 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2246 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2248 /* cases below refer to rfc 3530 section 14.2.34: */
2249 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2250 if (conf && !unconf) /* case 2: probable retransmit */
2252 else /* case 4: client hasn't noticed we rebooted yet? */
2253 status = nfserr_stale_clientid;
2257 if (conf) { /* case 1: callback update */
2258 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2259 nfsd4_probe_callback(conf);
2260 expire_client(unconf);
2261 } else { /* case 3: normal case; new or rebooted client */
2262 unsigned int hash = clientstr_hashval(unconf->cl_recdir);
2264 conf = find_confirmed_client_by_str(unconf->cl_recdir, hash);
2266 nfsd4_client_record_remove(conf);
2267 expire_client(conf);
2269 move_to_confirmed(unconf);
2270 nfsd4_probe_callback(unconf);
2273 nfs4_unlock_state();
2277 static struct nfs4_file *nfsd4_alloc_file(void)
2279 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2282 /* OPEN Share state helper functions */
2283 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2285 unsigned int hashval = file_hashval(ino);
2287 atomic_set(&fp->fi_ref, 1);
2288 INIT_LIST_HEAD(&fp->fi_hash);
2289 INIT_LIST_HEAD(&fp->fi_stateids);
2290 INIT_LIST_HEAD(&fp->fi_delegations);
2291 fp->fi_inode = igrab(ino);
2292 fp->fi_had_conflict = false;
2293 fp->fi_lease = NULL;
2294 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2295 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2296 spin_lock(&recall_lock);
2297 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2298 spin_unlock(&recall_lock);
2302 nfsd4_free_slab(struct kmem_cache **slab)
2306 kmem_cache_destroy(*slab);
2311 nfsd4_free_slabs(void)
2313 nfsd4_free_slab(&openowner_slab);
2314 nfsd4_free_slab(&lockowner_slab);
2315 nfsd4_free_slab(&file_slab);
2316 nfsd4_free_slab(&stateid_slab);
2317 nfsd4_free_slab(&deleg_slab);
2321 nfsd4_init_slabs(void)
2323 openowner_slab = kmem_cache_create("nfsd4_openowners",
2324 sizeof(struct nfs4_openowner), 0, 0, NULL);
2325 if (openowner_slab == NULL)
2327 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2328 sizeof(struct nfs4_openowner), 0, 0, NULL);
2329 if (lockowner_slab == NULL)
2331 file_slab = kmem_cache_create("nfsd4_files",
2332 sizeof(struct nfs4_file), 0, 0, NULL);
2333 if (file_slab == NULL)
2335 stateid_slab = kmem_cache_create("nfsd4_stateids",
2336 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2337 if (stateid_slab == NULL)
2339 deleg_slab = kmem_cache_create("nfsd4_delegations",
2340 sizeof(struct nfs4_delegation), 0, 0, NULL);
2341 if (deleg_slab == NULL)
2346 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2350 void nfs4_free_openowner(struct nfs4_openowner *oo)
2352 kfree(oo->oo_owner.so_owner.data);
2353 kmem_cache_free(openowner_slab, oo);
2356 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2358 kfree(lo->lo_owner.so_owner.data);
2359 kmem_cache_free(lockowner_slab, lo);
2362 static void init_nfs4_replay(struct nfs4_replay *rp)
2364 rp->rp_status = nfserr_serverfault;
2366 rp->rp_buf = rp->rp_ibuf;
2369 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2371 struct nfs4_stateowner *sop;
2373 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2377 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2378 if (!sop->so_owner.data) {
2379 kmem_cache_free(slab, sop);
2382 sop->so_owner.len = owner->len;
2384 INIT_LIST_HEAD(&sop->so_stateids);
2385 sop->so_client = clp;
2386 init_nfs4_replay(&sop->so_replay);
2390 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2392 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2393 list_add(&oo->oo_perclient, &clp->cl_openowners);
2396 static struct nfs4_openowner *
2397 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2398 struct nfs4_openowner *oo;
2400 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2403 oo->oo_owner.so_is_open_owner = 1;
2404 oo->oo_owner.so_seqid = open->op_seqid;
2405 oo->oo_flags = NFS4_OO_NEW;
2407 oo->oo_last_closed_stid = NULL;
2408 INIT_LIST_HEAD(&oo->oo_close_lru);
2409 hash_openowner(oo, clp, strhashval);
2413 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2414 struct nfs4_openowner *oo = open->op_openowner;
2415 struct nfs4_client *clp = oo->oo_owner.so_client;
2417 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2418 INIT_LIST_HEAD(&stp->st_lockowners);
2419 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2420 list_add(&stp->st_perfile, &fp->fi_stateids);
2421 stp->st_stateowner = &oo->oo_owner;
2424 stp->st_access_bmap = 0;
2425 stp->st_deny_bmap = 0;
2426 set_access(open->op_share_access, stp);
2427 set_deny(open->op_share_deny, stp);
2428 stp->st_openstp = NULL;
2432 move_to_close_lru(struct nfs4_openowner *oo)
2434 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2436 list_move_tail(&oo->oo_close_lru, &close_lru);
2437 oo->oo_time = get_seconds();
2441 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2444 return (sop->so_owner.len == owner->len) &&
2445 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2446 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2449 static struct nfs4_openowner *
2450 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2452 struct nfs4_stateowner *so;
2453 struct nfs4_openowner *oo;
2455 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2456 if (!so->so_is_open_owner)
2458 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2460 renew_client(oo->oo_owner.so_client);
2467 /* search file_hashtbl[] for file */
2468 static struct nfs4_file *
2469 find_file(struct inode *ino)
2471 unsigned int hashval = file_hashval(ino);
2472 struct nfs4_file *fp;
2474 spin_lock(&recall_lock);
2475 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2476 if (fp->fi_inode == ino) {
2478 spin_unlock(&recall_lock);
2482 spin_unlock(&recall_lock);
2487 * Called to check deny when READ with all zero stateid or
2488 * WRITE with all zero or all one stateid
2491 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2493 struct inode *ino = current_fh->fh_dentry->d_inode;
2494 struct nfs4_file *fp;
2495 struct nfs4_ol_stateid *stp;
2498 dprintk("NFSD: nfs4_share_conflict\n");
2500 fp = find_file(ino);
2503 ret = nfserr_locked;
2504 /* Search for conflicting share reservations */
2505 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2506 if (test_deny(deny_type, stp) ||
2507 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2516 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2518 /* We're assuming the state code never drops its reference
2519 * without first removing the lease. Since we're in this lease
2520 * callback (and since the lease code is serialized by the kernel
2521 * lock) we know the server hasn't removed the lease yet, we know
2522 * it's safe to take a reference: */
2523 atomic_inc(&dp->dl_count);
2525 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2527 /* only place dl_time is set. protected by lock_flocks*/
2528 dp->dl_time = get_seconds();
2530 nfsd4_cb_recall(dp);
2533 /* Called from break_lease() with lock_flocks() held. */
2534 static void nfsd_break_deleg_cb(struct file_lock *fl)
2536 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2537 struct nfs4_delegation *dp;
2540 /* We assume break_lease is only called once per lease: */
2541 BUG_ON(fp->fi_had_conflict);
2543 * We don't want the locks code to timeout the lease for us;
2544 * we'll remove it ourself if a delegation isn't returned
2547 fl->fl_break_time = 0;
2549 spin_lock(&recall_lock);
2550 fp->fi_had_conflict = true;
2551 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2552 nfsd_break_one_deleg(dp);
2553 spin_unlock(&recall_lock);
2557 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2560 return lease_modify(onlist, arg);
2565 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2566 .lm_break = nfsd_break_deleg_cb,
2567 .lm_change = nfsd_change_deleg_cb,
2570 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2572 if (nfsd4_has_session(cstate))
2574 if (seqid == so->so_seqid - 1)
2575 return nfserr_replay_me;
2576 if (seqid == so->so_seqid)
2578 return nfserr_bad_seqid;
2582 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2583 struct nfsd4_open *open)
2585 clientid_t *clientid = &open->op_clientid;
2586 struct nfs4_client *clp = NULL;
2587 unsigned int strhashval;
2588 struct nfs4_openowner *oo = NULL;
2590 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2592 if (STALE_CLIENTID(&open->op_clientid, nn))
2593 return nfserr_stale_clientid;
2595 * In case we need it later, after we've already created the
2596 * file and don't want to risk a further failure:
2598 open->op_file = nfsd4_alloc_file();
2599 if (open->op_file == NULL)
2600 return nfserr_jukebox;
2602 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2603 oo = find_openstateowner_str(strhashval, open);
2604 open->op_openowner = oo;
2606 clp = find_confirmed_client(clientid);
2608 return nfserr_expired;
2611 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2612 /* Replace unconfirmed owners without checking for replay. */
2613 clp = oo->oo_owner.so_client;
2614 release_openowner(oo);
2615 open->op_openowner = NULL;
2618 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2621 clp = oo->oo_owner.so_client;
2624 oo = alloc_init_open_stateowner(strhashval, clp, open);
2626 return nfserr_jukebox;
2627 open->op_openowner = oo;
2629 open->op_stp = nfs4_alloc_stateid(clp);
2631 return nfserr_jukebox;
2635 static inline __be32
2636 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2638 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2639 return nfserr_openmode;
2644 static int share_access_to_flags(u32 share_access)
2646 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2649 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2651 struct nfs4_stid *ret;
2653 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2656 return delegstateid(ret);
2659 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2661 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2662 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2666 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2667 struct nfs4_delegation **dp)
2670 __be32 status = nfserr_bad_stateid;
2672 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2675 flags = share_access_to_flags(open->op_share_access);
2676 status = nfs4_check_delegmode(*dp, flags);
2680 if (!nfsd4_is_deleg_cur(open))
2684 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2689 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2691 struct nfs4_ol_stateid *local;
2692 struct nfs4_openowner *oo = open->op_openowner;
2694 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2695 /* ignore lock owners */
2696 if (local->st_stateowner->so_is_open_owner == 0)
2698 /* remember if we have seen this open owner */
2699 if (local->st_stateowner == &oo->oo_owner)
2701 /* check for conflicting share reservations */
2702 if (!test_share(local, open))
2703 return nfserr_share_denied;
2708 static void nfs4_free_stateid(struct nfs4_ol_stateid *s)
2710 kmem_cache_free(stateid_slab, s);
2713 static inline int nfs4_access_to_access(u32 nfs4_access)
2717 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2718 flags |= NFSD_MAY_READ;
2719 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2720 flags |= NFSD_MAY_WRITE;
2724 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2725 struct svc_fh *cur_fh, struct nfsd4_open *open)
2728 int oflag = nfs4_access_to_omode(open->op_share_access);
2729 int access = nfs4_access_to_access(open->op_share_access);
2731 if (!fp->fi_fds[oflag]) {
2732 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2733 &fp->fi_fds[oflag]);
2737 nfs4_file_get_access(fp, oflag);
2742 static inline __be32
2743 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2744 struct nfsd4_open *open)
2746 struct iattr iattr = {
2747 .ia_valid = ATTR_SIZE,
2750 if (!open->op_truncate)
2752 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2753 return nfserr_inval;
2754 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2758 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)
2760 u32 op_share_access = open->op_share_access;
2764 new_access = !test_access(op_share_access, stp);
2766 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2770 status = nfsd4_truncate(rqstp, cur_fh, open);
2773 int oflag = nfs4_access_to_omode(op_share_access);
2774 nfs4_file_put_access(fp, oflag);
2778 /* remember the open */
2779 set_access(op_share_access, stp);
2780 set_deny(open->op_share_deny, stp);
2787 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2789 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2792 /* Should we give out recallable state?: */
2793 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2795 if (clp->cl_cb_state == NFSD4_CB_UP)
2798 * In the sessions case, since we don't have to establish a
2799 * separate connection for callbacks, we assume it's OK
2800 * until we hear otherwise:
2802 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2805 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2807 struct file_lock *fl;
2809 fl = locks_alloc_lock();
2812 locks_init_lock(fl);
2813 fl->fl_lmops = &nfsd_lease_mng_ops;
2814 fl->fl_flags = FL_LEASE;
2815 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2816 fl->fl_end = OFFSET_MAX;
2817 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2818 fl->fl_pid = current->tgid;
2822 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2824 struct nfs4_file *fp = dp->dl_file;
2825 struct file_lock *fl;
2828 fl = nfs4_alloc_init_lease(dp, flag);
2831 fl->fl_file = find_readable_file(fp);
2832 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2833 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2835 list_del_init(&dp->dl_perclnt);
2836 locks_free_lock(fl);
2840 fp->fi_deleg_file = get_file(fl->fl_file);
2841 atomic_set(&fp->fi_delegees, 1);
2842 list_add(&dp->dl_perfile, &fp->fi_delegations);
2846 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2848 struct nfs4_file *fp = dp->dl_file;
2851 return nfs4_setlease(dp, flag);
2852 spin_lock(&recall_lock);
2853 if (fp->fi_had_conflict) {
2854 spin_unlock(&recall_lock);
2857 atomic_inc(&fp->fi_delegees);
2858 list_add(&dp->dl_perfile, &fp->fi_delegations);
2859 spin_unlock(&recall_lock);
2860 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2864 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2866 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2867 if (status == -EAGAIN)
2868 open->op_why_no_deleg = WND4_CONTENTION;
2870 open->op_why_no_deleg = WND4_RESOURCE;
2871 switch (open->op_deleg_want) {
2872 case NFS4_SHARE_WANT_READ_DELEG:
2873 case NFS4_SHARE_WANT_WRITE_DELEG:
2874 case NFS4_SHARE_WANT_ANY_DELEG:
2876 case NFS4_SHARE_WANT_CANCEL:
2877 open->op_why_no_deleg = WND4_CANCELLED;
2879 case NFS4_SHARE_WANT_NO_DELEG:
2880 BUG(); /* not supposed to get here */
2886 * Attempt to hand out a delegation.
2889 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2890 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2892 struct nfs4_delegation *dp;
2893 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2895 int status = 0, flag = 0;
2897 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2898 flag = NFS4_OPEN_DELEGATE_NONE;
2899 open->op_recall = 0;
2900 switch (open->op_claim_type) {
2901 case NFS4_OPEN_CLAIM_PREVIOUS:
2903 open->op_recall = 1;
2904 flag = open->op_delegate_type;
2905 if (flag == NFS4_OPEN_DELEGATE_NONE)
2908 case NFS4_OPEN_CLAIM_NULL:
2909 /* Let's not give out any delegations till everyone's
2910 * had the chance to reclaim theirs.... */
2911 if (locks_in_grace(net))
2913 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2915 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2916 flag = NFS4_OPEN_DELEGATE_WRITE;
2918 flag = NFS4_OPEN_DELEGATE_READ;
2924 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2927 status = nfs4_set_delegation(dp, flag);
2931 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2933 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2934 STATEID_VAL(&dp->dl_stid.sc_stateid));
2936 open->op_delegate_type = flag;
2937 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2938 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2939 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2940 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2942 /* 4.1 client asking for a delegation? */
2943 if (open->op_deleg_want)
2944 nfsd4_open_deleg_none_ext(open, status);
2948 nfs4_put_delegation(dp);
2950 flag = NFS4_OPEN_DELEGATE_NONE;
2954 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2955 struct nfs4_delegation *dp)
2957 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2958 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2959 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2960 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2961 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2962 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2963 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2964 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
2966 /* Otherwise the client must be confused wanting a delegation
2967 * it already has, therefore we don't return
2968 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
2973 * called with nfs4_lock_state() held.
2976 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2978 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2979 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
2980 struct nfs4_file *fp = NULL;
2981 struct inode *ino = current_fh->fh_dentry->d_inode;
2982 struct nfs4_ol_stateid *stp = NULL;
2983 struct nfs4_delegation *dp = NULL;
2987 * Lookup file; if found, lookup stateid and check open request,
2988 * and check for delegations in the process of being recalled.
2989 * If not found, create the nfs4_file struct
2991 fp = find_file(ino);
2993 if ((status = nfs4_check_open(fp, open, &stp)))
2995 status = nfs4_check_deleg(cl, fp, open, &dp);
2999 status = nfserr_bad_stateid;
3000 if (nfsd4_is_deleg_cur(open))
3002 status = nfserr_jukebox;
3004 open->op_file = NULL;
3005 nfsd4_init_file(fp, ino);
3009 * OPEN the file, or upgrade an existing OPEN.
3010 * If truncate fails, the OPEN fails.
3013 /* Stateid was found, this is an OPEN upgrade */
3014 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3018 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3021 status = nfsd4_truncate(rqstp, current_fh, open);
3025 open->op_stp = NULL;
3026 init_open_stateid(stp, fp, open);
3028 update_stateid(&stp->st_stid.sc_stateid);
3029 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3031 if (nfsd4_has_session(&resp->cstate)) {
3032 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3034 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3035 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3036 open->op_why_no_deleg = WND4_NOT_WANTED;
3042 * Attempt to hand out a delegation. No error return, because the
3043 * OPEN succeeds even if we fail.
3045 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3049 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3050 STATEID_VAL(&stp->st_stid.sc_stateid));
3052 /* 4.1 client trying to upgrade/downgrade delegation? */
3053 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3054 open->op_deleg_want)
3055 nfsd4_deleg_xgrade_none_ext(open, dp);
3059 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3060 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3062 * To finish the open response, we just need to set the rflags.
3064 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3065 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3066 !nfsd4_has_session(&resp->cstate))
3067 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3072 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3074 if (open->op_openowner) {
3075 struct nfs4_openowner *oo = open->op_openowner;
3077 if (!list_empty(&oo->oo_owner.so_stateids))
3078 list_del_init(&oo->oo_close_lru);
3079 if (oo->oo_flags & NFS4_OO_NEW) {
3081 release_openowner(oo);
3082 open->op_openowner = NULL;
3084 oo->oo_flags &= ~NFS4_OO_NEW;
3088 nfsd4_free_file(open->op_file);
3090 nfs4_free_stateid(open->op_stp);
3094 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3097 struct nfs4_client *clp;
3099 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3102 dprintk("process_renew(%08x/%08x): starting\n",
3103 clid->cl_boot, clid->cl_id);
3104 status = nfserr_stale_clientid;
3105 if (STALE_CLIENTID(clid, nn))
3107 clp = find_confirmed_client(clid);
3108 status = nfserr_expired;
3110 /* We assume the client took too long to RENEW. */
3111 dprintk("nfsd4_renew: clientid not found!\n");
3114 status = nfserr_cb_path_down;
3115 if (!list_empty(&clp->cl_delegations)
3116 && clp->cl_cb_state != NFSD4_CB_UP)
3120 nfs4_unlock_state();
3125 nfsd4_end_grace(struct net *net)
3127 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3129 /* do nothing if grace period already ended */
3130 if (nn->grace_ended)
3133 dprintk("NFSD: end of grace period\n");
3134 nn->grace_ended = true;
3135 nfsd4_record_grace_done(net, nn->boot_time);
3136 locks_end_grace(&nn->nfsd4_manager);
3138 * Now that every NFSv4 client has had the chance to recover and
3139 * to see the (possibly new, possibly shorter) lease time, we
3140 * can safely set the next grace time to the current lease time:
3142 nfsd4_grace = nfsd4_lease;
3146 nfs4_laundromat(void)
3148 struct nfs4_client *clp;
3149 struct nfs4_openowner *oo;
3150 struct nfs4_delegation *dp;
3151 struct list_head *pos, *next, reaplist;
3152 time_t cutoff = get_seconds() - nfsd4_lease;
3153 time_t t, clientid_val = nfsd4_lease;
3154 time_t u, test_val = nfsd4_lease;
3158 dprintk("NFSD: laundromat service - starting\n");
3159 nfsd4_end_grace(&init_net);
3160 INIT_LIST_HEAD(&reaplist);
3161 spin_lock(&client_lock);
3162 list_for_each_safe(pos, next, &client_lru) {
3163 clp = list_entry(pos, struct nfs4_client, cl_lru);
3164 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3165 t = clp->cl_time - cutoff;
3166 if (clientid_val > t)
3170 if (atomic_read(&clp->cl_refcount)) {
3171 dprintk("NFSD: client in use (clientid %08x)\n",
3172 clp->cl_clientid.cl_id);
3175 unhash_client_locked(clp);
3176 list_add(&clp->cl_lru, &reaplist);
3178 spin_unlock(&client_lock);
3179 list_for_each_safe(pos, next, &reaplist) {
3180 clp = list_entry(pos, struct nfs4_client, cl_lru);
3181 dprintk("NFSD: purging unused client (clientid %08x)\n",
3182 clp->cl_clientid.cl_id);
3183 nfsd4_client_record_remove(clp);
3186 spin_lock(&recall_lock);
3187 list_for_each_safe(pos, next, &del_recall_lru) {
3188 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3189 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3190 u = dp->dl_time - cutoff;
3195 list_move(&dp->dl_recall_lru, &reaplist);
3197 spin_unlock(&recall_lock);
3198 list_for_each_safe(pos, next, &reaplist) {
3199 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3200 unhash_delegation(dp);
3202 test_val = nfsd4_lease;
3203 list_for_each_safe(pos, next, &close_lru) {
3204 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3205 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3206 u = oo->oo_time - cutoff;
3211 release_openowner(oo);
3213 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3214 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3215 nfs4_unlock_state();
3216 return clientid_val;
3219 static struct workqueue_struct *laundry_wq;
3220 static void laundromat_main(struct work_struct *);
3221 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3224 laundromat_main(struct work_struct *not_used)
3228 t = nfs4_laundromat();
3229 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3230 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3233 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3235 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3236 return nfserr_bad_stateid;
3241 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3243 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3245 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3246 STATEID_VAL(stateid));
3251 access_permit_read(struct nfs4_ol_stateid *stp)
3253 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3254 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3255 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3259 access_permit_write(struct nfs4_ol_stateid *stp)
3261 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3262 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3266 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3268 __be32 status = nfserr_openmode;
3270 /* For lock stateid's, we test the parent open, not the lock: */
3271 if (stp->st_openstp)
3272 stp = stp->st_openstp;
3273 if ((flags & WR_STATE) && !access_permit_write(stp))
3275 if ((flags & RD_STATE) && !access_permit_read(stp))
3282 static inline __be32
3283 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3285 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3287 else if (locks_in_grace(net)) {
3288 /* Answer in remaining cases depends on existence of
3289 * conflicting state; so we must wait out the grace period. */
3290 return nfserr_grace;
3291 } else if (flags & WR_STATE)
3292 return nfs4_share_conflict(current_fh,
3293 NFS4_SHARE_DENY_WRITE);
3294 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3295 return nfs4_share_conflict(current_fh,
3296 NFS4_SHARE_DENY_READ);
3300 * Allow READ/WRITE during grace period on recovered state only for files
3301 * that are not able to provide mandatory locking.
3304 grace_disallows_io(struct net *net, struct inode *inode)
3306 return locks_in_grace(net) && mandatory_lock(inode);
3309 /* Returns true iff a is later than b: */
3310 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3312 return (s32)a->si_generation - (s32)b->si_generation > 0;
3315 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3318 * When sessions are used the stateid generation number is ignored
3321 if (has_session && in->si_generation == 0)
3324 if (in->si_generation == ref->si_generation)
3327 /* If the client sends us a stateid from the future, it's buggy: */
3328 if (stateid_generation_after(in, ref))
3329 return nfserr_bad_stateid;
3331 * However, we could see a stateid from the past, even from a
3332 * non-buggy client. For example, if the client sends a lock
3333 * while some IO is outstanding, the lock may bump si_generation
3334 * while the IO is still in flight. The client could avoid that
3335 * situation by waiting for responses on all the IO requests,
3336 * but better performance may result in retrying IO that
3337 * receives an old_stateid error if requests are rarely
3338 * reordered in flight:
3340 return nfserr_old_stateid;
3343 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3345 struct nfs4_stid *s;
3346 struct nfs4_ol_stateid *ols;
3349 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3350 return nfserr_bad_stateid;
3351 /* Client debugging aid. */
3352 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3353 char addr_str[INET6_ADDRSTRLEN];
3354 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3356 pr_warn_ratelimited("NFSD: client %s testing state ID "
3357 "with incorrect client ID\n", addr_str);
3358 return nfserr_bad_stateid;
3360 s = find_stateid(cl, stateid);
3362 return nfserr_bad_stateid;
3363 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3366 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3368 ols = openlockstateid(s);
3369 if (ols->st_stateowner->so_is_open_owner
3370 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3371 return nfserr_bad_stateid;
3375 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
3377 struct nfs4_client *cl;
3378 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3380 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3381 return nfserr_bad_stateid;
3382 if (STALE_STATEID(stateid, nn))
3383 return nfserr_stale_stateid;
3384 cl = find_confirmed_client(&stateid->si_opaque.so_clid);
3386 return nfserr_expired;
3387 *s = find_stateid_by_type(cl, stateid, typemask);
3389 return nfserr_bad_stateid;
3395 * Checks for stateid operations
3398 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3399 stateid_t *stateid, int flags, struct file **filpp)
3401 struct nfs4_stid *s;
3402 struct nfs4_ol_stateid *stp = NULL;
3403 struct nfs4_delegation *dp = NULL;
3404 struct svc_fh *current_fh = &cstate->current_fh;
3405 struct inode *ino = current_fh->fh_dentry->d_inode;
3411 if (grace_disallows_io(net, ino))
3412 return nfserr_grace;
3414 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3415 return check_special_stateids(net, current_fh, stateid, flags);
3417 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
3420 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3423 switch (s->sc_type) {
3424 case NFS4_DELEG_STID:
3425 dp = delegstateid(s);
3426 status = nfs4_check_delegmode(dp, flags);
3430 *filpp = dp->dl_file->fi_deleg_file;
3434 case NFS4_OPEN_STID:
3435 case NFS4_LOCK_STID:
3436 stp = openlockstateid(s);
3437 status = nfs4_check_fh(current_fh, stp);
3440 if (stp->st_stateowner->so_is_open_owner
3441 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3443 status = nfs4_check_openmode(stp, flags);
3447 if (flags & RD_STATE)
3448 *filpp = find_readable_file(stp->st_file);
3450 *filpp = find_writeable_file(stp->st_file);
3454 return nfserr_bad_stateid;
3462 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3464 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3465 return nfserr_locks_held;
3466 release_lock_stateid(stp);
3471 * Test if the stateid is valid
3474 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3475 struct nfsd4_test_stateid *test_stateid)
3477 struct nfsd4_test_stateid_id *stateid;
3478 struct nfs4_client *cl = cstate->session->se_client;
3481 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3482 stateid->ts_id_status =
3483 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3484 nfs4_unlock_state();
3490 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3491 struct nfsd4_free_stateid *free_stateid)
3493 stateid_t *stateid = &free_stateid->fr_stateid;
3494 struct nfs4_stid *s;
3495 struct nfs4_client *cl = cstate->session->se_client;
3496 __be32 ret = nfserr_bad_stateid;
3499 s = find_stateid(cl, stateid);
3502 switch (s->sc_type) {
3503 case NFS4_DELEG_STID:
3504 ret = nfserr_locks_held;
3506 case NFS4_OPEN_STID:
3507 case NFS4_LOCK_STID:
3508 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3511 if (s->sc_type == NFS4_LOCK_STID)
3512 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3514 ret = nfserr_locks_held;
3517 ret = nfserr_bad_stateid;
3520 nfs4_unlock_state();
3527 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3528 RD_STATE : WR_STATE;
3531 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3533 struct svc_fh *current_fh = &cstate->current_fh;
3534 struct nfs4_stateowner *sop = stp->st_stateowner;
3537 status = nfsd4_check_seqid(cstate, sop, seqid);
3540 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3542 * "Closed" stateid's exist *only* to return
3543 * nfserr_replay_me from the previous step.
3545 return nfserr_bad_stateid;
3546 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3549 return nfs4_check_fh(current_fh, stp);
3553 * Checks for sequence id mutating operations.
3556 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3557 stateid_t *stateid, char typemask,
3558 struct nfs4_ol_stateid **stpp)
3561 struct nfs4_stid *s;
3563 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3564 seqid, STATEID_VAL(stateid));
3567 status = nfsd4_lookup_stateid(stateid, typemask, &s);
3570 *stpp = openlockstateid(s);
3571 cstate->replay_owner = (*stpp)->st_stateowner;
3573 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3576 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3579 struct nfs4_openowner *oo;
3581 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3582 NFS4_OPEN_STID, stpp);
3585 oo = openowner((*stpp)->st_stateowner);
3586 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3587 return nfserr_bad_stateid;
3592 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3593 struct nfsd4_open_confirm *oc)
3596 struct nfs4_openowner *oo;
3597 struct nfs4_ol_stateid *stp;
3599 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3600 (int)cstate->current_fh.fh_dentry->d_name.len,
3601 cstate->current_fh.fh_dentry->d_name.name);
3603 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3609 status = nfs4_preprocess_seqid_op(cstate,
3610 oc->oc_seqid, &oc->oc_req_stateid,
3611 NFS4_OPEN_STID, &stp);
3614 oo = openowner(stp->st_stateowner);
3615 status = nfserr_bad_stateid;
3616 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3618 oo->oo_flags |= NFS4_OO_CONFIRMED;
3619 update_stateid(&stp->st_stid.sc_stateid);
3620 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3621 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3622 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3624 nfsd4_client_record_create(oo->oo_owner.so_client);
3627 if (!cstate->replay_owner)
3628 nfs4_unlock_state();
3632 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3634 if (!test_access(access, stp))
3636 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3637 clear_access(access, stp);
3640 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3642 switch (to_access) {
3643 case NFS4_SHARE_ACCESS_READ:
3644 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3645 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3647 case NFS4_SHARE_ACCESS_WRITE:
3648 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3649 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3651 case NFS4_SHARE_ACCESS_BOTH:
3659 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3662 for (i = 0; i < 4; i++) {
3663 if ((i & deny) != i)
3669 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3670 struct nfsd4_compound_state *cstate,
3671 struct nfsd4_open_downgrade *od)
3674 struct nfs4_ol_stateid *stp;
3676 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3677 (int)cstate->current_fh.fh_dentry->d_name.len,
3678 cstate->current_fh.fh_dentry->d_name.name);
3680 /* We don't yet support WANT bits: */
3681 if (od->od_deleg_want)
3682 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3686 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3687 &od->od_stateid, &stp);
3690 status = nfserr_inval;
3691 if (!test_access(od->od_share_access, stp)) {
3692 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3693 stp->st_access_bmap, od->od_share_access);
3696 if (!test_deny(od->od_share_deny, stp)) {
3697 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3698 stp->st_deny_bmap, od->od_share_deny);
3701 nfs4_stateid_downgrade(stp, od->od_share_access);
3703 reset_union_bmap_deny(od->od_share_deny, stp);
3705 update_stateid(&stp->st_stid.sc_stateid);
3706 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3709 if (!cstate->replay_owner)
3710 nfs4_unlock_state();
3714 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3716 struct nfs4_openowner *oo;
3717 struct nfs4_ol_stateid *s;
3719 if (!so->so_is_open_owner)
3722 s = oo->oo_last_closed_stid;
3725 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3726 /* Release the last_closed_stid on the next seqid bump: */
3727 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3730 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3731 release_last_closed_stateid(oo);
3734 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3736 unhash_open_stateid(s);
3737 s->st_stid.sc_type = NFS4_CLOSED_STID;
3741 * nfs4_unlock_state() called after encode
3744 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3745 struct nfsd4_close *close)
3748 struct nfs4_openowner *oo;
3749 struct nfs4_ol_stateid *stp;
3751 dprintk("NFSD: nfsd4_close on file %.*s\n",
3752 (int)cstate->current_fh.fh_dentry->d_name.len,
3753 cstate->current_fh.fh_dentry->d_name.name);
3756 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3758 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3762 oo = openowner(stp->st_stateowner);
3764 update_stateid(&stp->st_stid.sc_stateid);
3765 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3767 nfsd4_close_open_stateid(stp);
3768 oo->oo_last_closed_stid = stp;
3770 if (list_empty(&oo->oo_owner.so_stateids)) {
3771 if (cstate->minorversion) {
3772 release_openowner(oo);
3773 cstate->replay_owner = NULL;
3776 * In the 4.0 case we need to keep the owners around a
3777 * little while to handle CLOSE replay.
3779 if (list_empty(&oo->oo_owner.so_stateids))
3780 move_to_close_lru(oo);
3784 if (!cstate->replay_owner)
3785 nfs4_unlock_state();
3790 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3791 struct nfsd4_delegreturn *dr)
3793 struct nfs4_delegation *dp;
3794 stateid_t *stateid = &dr->dr_stateid;
3795 struct nfs4_stid *s;
3796 struct inode *inode;
3799 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3801 inode = cstate->current_fh.fh_dentry->d_inode;
3804 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s);
3807 dp = delegstateid(s);
3808 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3812 unhash_delegation(dp);
3814 nfs4_unlock_state();
3820 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3822 #define LOCKOWNER_INO_HASH_BITS 8
3823 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3824 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3827 end_offset(u64 start, u64 len)
3832 return end >= start ? end: NFS4_MAX_UINT64;
3835 /* last octet in a range */
3837 last_byte_offset(u64 start, u64 len)
3843 return end > start ? end - 1: NFS4_MAX_UINT64;
3846 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3848 return (file_hashval(inode) + cl_id
3849 + opaque_hashval(ownername->data, ownername->len))
3850 & LOCKOWNER_INO_HASH_MASK;
3853 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3856 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3857 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3858 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3859 * locking, this prevents us from being completely protocol-compliant. The
3860 * real solution to this problem is to start using unsigned file offsets in
3861 * the VFS, but this is a very deep change!
3864 nfs4_transform_lock_offset(struct file_lock *lock)
3866 if (lock->fl_start < 0)
3867 lock->fl_start = OFFSET_MAX;
3868 if (lock->fl_end < 0)
3869 lock->fl_end = OFFSET_MAX;
3872 /* Hack!: For now, we're defining this just so we can use a pointer to it
3873 * as a unique cookie to identify our (NFSv4's) posix locks. */
3874 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3878 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3880 struct nfs4_lockowner *lo;
3882 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3883 lo = (struct nfs4_lockowner *) fl->fl_owner;
3884 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3885 lo->lo_owner.so_owner.len, GFP_KERNEL);
3886 if (!deny->ld_owner.data)
3887 /* We just don't care that much */
3889 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3890 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3893 deny->ld_owner.len = 0;
3894 deny->ld_owner.data = NULL;
3895 deny->ld_clientid.cl_boot = 0;
3896 deny->ld_clientid.cl_id = 0;
3898 deny->ld_start = fl->fl_start;
3899 deny->ld_length = NFS4_MAX_UINT64;
3900 if (fl->fl_end != NFS4_MAX_UINT64)
3901 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3902 deny->ld_type = NFS4_READ_LT;
3903 if (fl->fl_type != F_RDLCK)
3904 deny->ld_type = NFS4_WRITE_LT;
3907 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3909 struct nfs4_ol_stateid *lst;
3911 if (!same_owner_str(&lo->lo_owner, owner, clid))
3913 lst = list_first_entry(&lo->lo_owner.so_stateids,
3914 struct nfs4_ol_stateid, st_perstateowner);
3915 return lst->st_file->fi_inode == inode;
3918 static struct nfs4_lockowner *
3919 find_lockowner_str(struct inode *inode, clientid_t *clid,
3920 struct xdr_netobj *owner)
3922 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3923 struct nfs4_lockowner *lo;
3925 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3926 if (same_lockowner_ino(lo, inode, clid, owner))
3932 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3934 struct inode *inode = open_stp->st_file->fi_inode;
3935 unsigned int inohash = lockowner_ino_hashval(inode,
3936 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3938 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3939 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3940 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3944 * Alloc a lock owner structure.
3945 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3948 * strhashval = ownerstr_hashval
3951 static struct nfs4_lockowner *
3952 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3953 struct nfs4_lockowner *lo;
3955 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3958 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3959 lo->lo_owner.so_is_open_owner = 0;
3960 /* It is the openowner seqid that will be incremented in encode in the
3961 * case of new lockowners; so increment the lock seqid manually: */
3962 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3963 hash_lockowner(lo, strhashval, clp, open_stp);
3967 static struct nfs4_ol_stateid *
3968 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3970 struct nfs4_ol_stateid *stp;
3971 struct nfs4_client *clp = lo->lo_owner.so_client;
3973 stp = nfs4_alloc_stateid(clp);
3976 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3977 list_add(&stp->st_perfile, &fp->fi_stateids);
3978 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3979 stp->st_stateowner = &lo->lo_owner;
3982 stp->st_access_bmap = 0;
3983 stp->st_deny_bmap = open_stp->st_deny_bmap;
3984 stp->st_openstp = open_stp;
3989 check_lock_length(u64 offset, u64 length)
3991 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3992 LOFF_OVERFLOW(offset, length)));
3995 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
3997 struct nfs4_file *fp = lock_stp->st_file;
3998 int oflag = nfs4_access_to_omode(access);
4000 if (test_access(access, lock_stp))
4002 nfs4_file_get_access(fp, oflag);
4003 set_access(access, lock_stp);
4006 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)
4008 struct nfs4_file *fi = ost->st_file;
4009 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4010 struct nfs4_client *cl = oo->oo_owner.so_client;
4011 struct nfs4_lockowner *lo;
4012 unsigned int strhashval;
4014 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4016 if (!cstate->minorversion)
4017 return nfserr_bad_seqid;
4018 /* XXX: a lockowner always has exactly one stateid: */
4019 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4020 struct nfs4_ol_stateid, st_perstateowner);
4023 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4024 &lock->v.new.owner);
4025 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4027 return nfserr_jukebox;
4028 *lst = alloc_init_lock_stateid(lo, fi, ost);
4030 release_lockowner(lo);
4031 return nfserr_jukebox;
4041 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4042 struct nfsd4_lock *lock)
4044 struct nfs4_openowner *open_sop = NULL;
4045 struct nfs4_lockowner *lock_sop = NULL;
4046 struct nfs4_ol_stateid *lock_stp;
4047 struct file *filp = NULL;
4048 struct file_lock file_lock;
4049 struct file_lock conflock;
4051 bool new_state = false;
4054 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4056 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4057 (long long) lock->lk_offset,
4058 (long long) lock->lk_length);
4060 if (check_lock_length(lock->lk_offset, lock->lk_length))
4061 return nfserr_inval;
4063 if ((status = fh_verify(rqstp, &cstate->current_fh,
4064 S_IFREG, NFSD_MAY_LOCK))) {
4065 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4071 if (lock->lk_is_new) {
4072 struct nfs4_ol_stateid *open_stp = NULL;
4074 if (nfsd4_has_session(cstate))
4075 /* See rfc 5661 18.10.3: given clientid is ignored: */
4076 memcpy(&lock->v.new.clientid,
4077 &cstate->session->se_client->cl_clientid,
4078 sizeof(clientid_t));
4080 status = nfserr_stale_clientid;
4081 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4084 /* validate and update open stateid and open seqid */
4085 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4086 lock->lk_new_open_seqid,
4087 &lock->lk_new_open_stateid,
4091 open_sop = openowner(open_stp->st_stateowner);
4092 status = nfserr_bad_stateid;
4093 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4094 &lock->v.new.clientid))
4096 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4097 &lock_stp, &new_state);
4099 status = nfs4_preprocess_seqid_op(cstate,
4100 lock->lk_old_lock_seqid,
4101 &lock->lk_old_lock_stateid,
4102 NFS4_LOCK_STID, &lock_stp);
4105 lock_sop = lockowner(lock_stp->st_stateowner);
4107 lkflg = setlkflg(lock->lk_type);
4108 status = nfs4_check_openmode(lock_stp, lkflg);
4112 status = nfserr_grace;
4113 if (locks_in_grace(SVC_NET(rqstp)) && !lock->lk_reclaim)
4115 status = nfserr_no_grace;
4116 if (!locks_in_grace(SVC_NET(rqstp)) && lock->lk_reclaim)
4119 locks_init_lock(&file_lock);
4120 switch (lock->lk_type) {
4123 filp = find_readable_file(lock_stp->st_file);
4125 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4126 file_lock.fl_type = F_RDLCK;
4129 case NFS4_WRITEW_LT:
4130 filp = find_writeable_file(lock_stp->st_file);
4132 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4133 file_lock.fl_type = F_WRLCK;
4136 status = nfserr_inval;
4140 status = nfserr_openmode;
4143 file_lock.fl_owner = (fl_owner_t)lock_sop;
4144 file_lock.fl_pid = current->tgid;
4145 file_lock.fl_file = filp;
4146 file_lock.fl_flags = FL_POSIX;
4147 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4149 file_lock.fl_start = lock->lk_offset;
4150 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4151 nfs4_transform_lock_offset(&file_lock);
4154 * Try to lock the file in the VFS.
4155 * Note: locks.c uses the BKL to protect the inode's lock list.
4158 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4160 case 0: /* success! */
4161 update_stateid(&lock_stp->st_stid.sc_stateid);
4162 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4166 case (EAGAIN): /* conflock holds conflicting lock */
4167 status = nfserr_denied;
4168 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4169 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4172 status = nfserr_deadlock;
4175 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4176 status = nfserrno(err);
4180 if (status && new_state)
4181 release_lockowner(lock_sop);
4182 if (!cstate->replay_owner)
4183 nfs4_unlock_state();
4188 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4189 * so we do a temporary open here just to get an open file to pass to
4190 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4193 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4196 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4198 err = nfserrno(vfs_test_lock(file, lock));
4208 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4209 struct nfsd4_lockt *lockt)
4211 struct inode *inode;
4212 struct file_lock file_lock;
4213 struct nfs4_lockowner *lo;
4215 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4217 if (locks_in_grace(SVC_NET(rqstp)))
4218 return nfserr_grace;
4220 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4221 return nfserr_inval;
4225 status = nfserr_stale_clientid;
4226 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid, nn))
4229 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4232 inode = cstate->current_fh.fh_dentry->d_inode;
4233 locks_init_lock(&file_lock);
4234 switch (lockt->lt_type) {
4237 file_lock.fl_type = F_RDLCK;
4240 case NFS4_WRITEW_LT:
4241 file_lock.fl_type = F_WRLCK;
4244 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4245 status = nfserr_inval;
4249 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4251 file_lock.fl_owner = (fl_owner_t)lo;
4252 file_lock.fl_pid = current->tgid;
4253 file_lock.fl_flags = FL_POSIX;
4255 file_lock.fl_start = lockt->lt_offset;
4256 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4258 nfs4_transform_lock_offset(&file_lock);
4260 status = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4264 if (file_lock.fl_type != F_UNLCK) {
4265 status = nfserr_denied;
4266 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4269 nfs4_unlock_state();
4274 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4275 struct nfsd4_locku *locku)
4277 struct nfs4_ol_stateid *stp;
4278 struct file *filp = NULL;
4279 struct file_lock file_lock;
4283 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4284 (long long) locku->lu_offset,
4285 (long long) locku->lu_length);
4287 if (check_lock_length(locku->lu_offset, locku->lu_length))
4288 return nfserr_inval;
4292 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4293 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4296 filp = find_any_file(stp->st_file);
4298 status = nfserr_lock_range;
4302 locks_init_lock(&file_lock);
4303 file_lock.fl_type = F_UNLCK;
4304 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4305 file_lock.fl_pid = current->tgid;
4306 file_lock.fl_file = filp;
4307 file_lock.fl_flags = FL_POSIX;
4308 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4309 file_lock.fl_start = locku->lu_offset;
4311 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4312 nfs4_transform_lock_offset(&file_lock);
4315 * Try to unlock the file in the VFS.
4317 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4319 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4323 * OK, unlock succeeded; the only thing left to do is update the stateid.
4325 update_stateid(&stp->st_stid.sc_stateid);
4326 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4329 if (!cstate->replay_owner)
4330 nfs4_unlock_state();
4334 status = nfserrno(err);
4340 * 1: locks held by lockowner
4341 * 0: no locks held by lockowner
4344 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4346 struct file_lock **flpp;
4347 struct inode *inode = filp->fi_inode;
4351 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4352 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4363 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4364 struct nfsd4_compound_state *cstate,
4365 struct nfsd4_release_lockowner *rlockowner)
4367 clientid_t *clid = &rlockowner->rl_clientid;
4368 struct nfs4_stateowner *sop;
4369 struct nfs4_lockowner *lo;
4370 struct nfs4_ol_stateid *stp;
4371 struct xdr_netobj *owner = &rlockowner->rl_owner;
4372 struct list_head matches;
4373 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4375 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4377 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4378 clid->cl_boot, clid->cl_id);
4380 /* XXX check for lease expiration */
4382 status = nfserr_stale_clientid;
4383 if (STALE_CLIENTID(clid, nn))
4388 status = nfserr_locks_held;
4389 INIT_LIST_HEAD(&matches);
4391 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4392 if (sop->so_is_open_owner)
4394 if (!same_owner_str(sop, owner, clid))
4396 list_for_each_entry(stp, &sop->so_stateids,
4398 lo = lockowner(sop);
4399 if (check_for_locks(stp->st_file, lo))
4401 list_add(&lo->lo_list, &matches);
4404 /* Clients probably won't expect us to return with some (but not all)
4405 * of the lockowner state released; so don't release any until all
4406 * have been checked. */
4408 while (!list_empty(&matches)) {
4409 lo = list_entry(matches.next, struct nfs4_lockowner,
4411 /* unhash_stateowner deletes so_perclient only
4412 * for openowners. */
4413 list_del(&lo->lo_list);
4414 release_lockowner(lo);
4417 nfs4_unlock_state();
4421 static inline struct nfs4_client_reclaim *
4424 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4428 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4430 unsigned int strhashval = clientstr_hashval(name);
4431 struct nfs4_client *clp;
4433 clp = find_confirmed_client_by_str(name, strhashval);
4436 return test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
4440 * failure => all reset bets are off, nfserr_no_grace...
4443 nfs4_client_to_reclaim(const char *name)
4445 unsigned int strhashval;
4446 struct nfs4_client_reclaim *crp = NULL;
4448 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4449 crp = alloc_reclaim();
4452 strhashval = clientstr_hashval(name);
4453 INIT_LIST_HEAD(&crp->cr_strhash);
4454 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4455 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4456 reclaim_str_hashtbl_size++;
4461 nfs4_release_reclaim(void)
4463 struct nfs4_client_reclaim *crp = NULL;
4466 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4467 while (!list_empty(&reclaim_str_hashtbl[i])) {
4468 crp = list_entry(reclaim_str_hashtbl[i].next,
4469 struct nfs4_client_reclaim, cr_strhash);
4470 list_del(&crp->cr_strhash);
4472 reclaim_str_hashtbl_size--;
4475 BUG_ON(reclaim_str_hashtbl_size);
4479 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4480 struct nfs4_client_reclaim *
4481 nfsd4_find_reclaim_client(struct nfs4_client *clp)
4483 unsigned int strhashval;
4484 struct nfs4_client_reclaim *crp = NULL;
4486 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4487 clp->cl_name.len, clp->cl_name.data,
4490 /* find clp->cl_name in reclaim_str_hashtbl */
4491 strhashval = clientstr_hashval(clp->cl_recdir);
4492 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4493 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4501 * Called from OPEN. Look for clientid in reclaim list.
4504 nfs4_check_open_reclaim(clientid_t *clid)
4506 struct nfs4_client *clp;
4508 /* find clientid in conf_id_hashtbl */
4509 clp = find_confirmed_client(clid);
4511 return nfserr_reclaim_bad;
4513 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4516 #ifdef CONFIG_NFSD_FAULT_INJECTION
4518 void nfsd_forget_clients(u64 num)
4520 struct nfs4_client *clp, *next;
4524 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4525 nfsd4_client_record_remove(clp);
4530 nfs4_unlock_state();
4532 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4535 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4537 release_lockowner(lockowner(sop));
4540 static void release_openowner_sop(struct nfs4_stateowner *sop)
4542 release_openowner(openowner(sop));
4545 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4546 void (*release_sop)(struct nfs4_stateowner *))
4549 struct nfs4_stateowner *sop, *next;
4551 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4552 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4553 if (sop->so_is_open_owner != is_open_owner)
4563 void nfsd_forget_locks(u64 num)
4568 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4569 nfs4_unlock_state();
4571 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4574 void nfsd_forget_openowners(u64 num)
4579 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4580 nfs4_unlock_state();
4582 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4585 int nfsd_process_n_delegations(u64 num, struct list_head *list)
4588 struct nfs4_file *fp, *fnext;
4589 struct nfs4_delegation *dp, *dnext;
4591 for (i = 0; i < FILE_HASH_SIZE; i++) {
4592 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4593 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4594 list_move(&dp->dl_recall_lru, list);
4604 void nfsd_forget_delegations(u64 num)
4608 struct nfs4_delegation *dp, *dnext;
4610 spin_lock(&recall_lock);
4611 count = nfsd_process_n_delegations(num, &victims);
4612 spin_unlock(&recall_lock);
4615 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
4616 unhash_delegation(dp);
4617 nfs4_unlock_state();
4619 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4622 void nfsd_recall_delegations(u64 num)
4626 struct nfs4_delegation *dp, *dnext;
4628 spin_lock(&recall_lock);
4629 count = nfsd_process_n_delegations(num, &victims);
4630 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
4631 list_del(&dp->dl_recall_lru);
4632 nfsd_break_one_deleg(dp);
4634 spin_unlock(&recall_lock);
4636 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4639 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4641 /* initialization to perform at module load time: */
4644 nfs4_state_init(void)
4648 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4649 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4650 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4651 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4652 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4653 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4655 for (i = 0; i < SESSION_HASH_SIZE; i++)
4656 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4657 for (i = 0; i < FILE_HASH_SIZE; i++) {
4658 INIT_LIST_HEAD(&file_hashtbl[i]);
4660 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4661 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4663 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4664 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4665 INIT_LIST_HEAD(&close_lru);
4666 INIT_LIST_HEAD(&client_lru);
4667 INIT_LIST_HEAD(&del_recall_lru);
4668 reclaim_str_hashtbl_size = 0;
4672 * Since the lifetime of a delegation isn't limited to that of an open, a
4673 * client may quite reasonably hang on to a delegation as long as it has
4674 * the inode cached. This becomes an obvious problem the first time a
4675 * client's inode cache approaches the size of the server's total memory.
4677 * For now we avoid this problem by imposing a hard limit on the number
4678 * of delegations, which varies according to the server's memory size.
4681 set_max_delegations(void)
4684 * Allow at most 4 delegations per megabyte of RAM. Quick
4685 * estimates suggest that in the worst case (where every delegation
4686 * is for a different inode), a delegation could take about 1.5K,
4687 * giving a worst case usage of about 6% of memory.
4689 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4692 /* initialization to perform when the nfsd service is started: */
4695 nfs4_state_start(void)
4697 struct net *net = &init_net;
4698 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4702 * FIXME: For now, we hang most of the pernet global stuff off of
4703 * init_net until nfsd is fully containerized. Eventually, we'll
4704 * need to pass a net pointer into this function, take a reference
4705 * to that instead and then do most of the rest of this on a per-net
4709 nfsd4_client_tracking_init(net);
4710 nn->boot_time = get_seconds();
4711 locks_start_grace(net, &nn->nfsd4_manager);
4712 nn->grace_ended = false;
4713 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4715 ret = set_callback_cred();
4720 laundry_wq = create_singlethread_workqueue("nfsd4");
4721 if (laundry_wq == NULL) {
4725 ret = nfsd4_create_callback_queue();
4727 goto out_free_laundry;
4728 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4729 set_max_delegations();
4732 destroy_workqueue(laundry_wq);
4734 nfsd4_client_tracking_exit(net);
4740 __nfs4_state_shutdown(void)
4743 struct nfs4_client *clp = NULL;
4744 struct nfs4_delegation *dp = NULL;
4745 struct list_head *pos, *next, reaplist;
4747 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4748 while (!list_empty(&conf_id_hashtbl[i])) {
4749 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4752 while (!list_empty(&unconf_str_hashtbl[i])) {
4753 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4757 INIT_LIST_HEAD(&reaplist);
4758 spin_lock(&recall_lock);
4759 list_for_each_safe(pos, next, &del_recall_lru) {
4760 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4761 list_move(&dp->dl_recall_lru, &reaplist);
4763 spin_unlock(&recall_lock);
4764 list_for_each_safe(pos, next, &reaplist) {
4765 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4766 unhash_delegation(dp);
4769 nfsd4_client_tracking_exit(&init_net);
4774 nfs4_state_shutdown(void)
4776 struct net *net = &init_net;
4777 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4779 cancel_delayed_work_sync(&laundromat_work);
4780 destroy_workqueue(laundry_wq);
4781 locks_end_grace(&nn->nfsd4_manager);
4783 __nfs4_state_shutdown();
4784 nfs4_unlock_state();
4785 nfsd4_destroy_callback_queue();
4789 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4791 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4792 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4796 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4798 if (cstate->minorversion) {
4799 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4800 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4805 clear_current_stateid(struct nfsd4_compound_state *cstate)
4807 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4811 * functions to set current state id
4814 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4816 put_stateid(cstate, &odp->od_stateid);
4820 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4822 put_stateid(cstate, &open->op_stateid);
4826 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4828 put_stateid(cstate, &close->cl_stateid);
4832 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4834 put_stateid(cstate, &lock->lk_resp_stateid);
4838 * functions to consume current state id
4842 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4844 get_stateid(cstate, &odp->od_stateid);
4848 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4850 get_stateid(cstate, &drp->dr_stateid);
4854 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4856 get_stateid(cstate, &fsp->fr_stateid);
4860 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4862 get_stateid(cstate, &setattr->sa_stateid);
4866 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4868 get_stateid(cstate, &close->cl_stateid);
4872 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4874 get_stateid(cstate, &locku->lu_stateid);
4878 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4880 get_stateid(cstate, &read->rd_stateid);
4884 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4886 get_stateid(cstate, &write->wr_stateid);