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 nfsd4_init_callback(&dp->dl_recall);
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_name_tree hold confirmed
416 * setclientid_confirmed info.
418 * unconf_id_hashtbl[] and unconf_name_tree 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 * All of the above fields are protected by the client_mutex.
429 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
430 static int reclaim_str_hashtbl_size = 0;
431 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
432 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
433 static struct rb_root conf_name_tree;
434 static struct rb_root unconf_name_tree;
435 static struct list_head client_lru;
436 static struct list_head close_lru;
439 * We store the NONE, READ, WRITE, and BOTH bits separately in the
440 * st_{access,deny}_bmap field of the stateid, in order to track not
441 * only what share bits are currently in force, but also what
442 * combinations of share bits previous opens have used. This allows us
443 * to enforce the recommendation of rfc 3530 14.2.19 that the server
444 * return an error if the client attempt to downgrade to a combination
445 * of share bits not explicable by closing some of its previous opens.
447 * XXX: This enforcement is actually incomplete, since we don't keep
448 * track of access/deny bit combinations; so, e.g., we allow:
450 * OPEN allow read, deny write
451 * OPEN allow both, deny none
452 * DOWNGRADE allow read, deny none
454 * which we should reject.
457 bmap_to_share_mode(unsigned long bmap) {
459 unsigned int access = 0;
461 for (i = 1; i < 4; i++) {
462 if (test_bit(i, &bmap))
469 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
470 unsigned int access, deny;
472 access = bmap_to_share_mode(stp->st_access_bmap);
473 deny = bmap_to_share_mode(stp->st_deny_bmap);
474 if ((access & open->op_share_deny) || (deny & open->op_share_access))
479 /* set share access for a given stateid */
481 set_access(u32 access, struct nfs4_ol_stateid *stp)
483 __set_bit(access, &stp->st_access_bmap);
486 /* clear share access for a given stateid */
488 clear_access(u32 access, struct nfs4_ol_stateid *stp)
490 __clear_bit(access, &stp->st_access_bmap);
493 /* test whether a given stateid has access */
495 test_access(u32 access, struct nfs4_ol_stateid *stp)
497 return test_bit(access, &stp->st_access_bmap);
500 /* set share deny for a given stateid */
502 set_deny(u32 access, struct nfs4_ol_stateid *stp)
504 __set_bit(access, &stp->st_deny_bmap);
507 /* clear share deny for a given stateid */
509 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
511 __clear_bit(access, &stp->st_deny_bmap);
514 /* test whether a given stateid is denying specific access */
516 test_deny(u32 access, struct nfs4_ol_stateid *stp)
518 return test_bit(access, &stp->st_deny_bmap);
521 static int nfs4_access_to_omode(u32 access)
523 switch (access & NFS4_SHARE_ACCESS_BOTH) {
524 case NFS4_SHARE_ACCESS_READ:
526 case NFS4_SHARE_ACCESS_WRITE:
528 case NFS4_SHARE_ACCESS_BOTH:
534 /* release all access and file references for a given stateid */
536 release_all_access(struct nfs4_ol_stateid *stp)
540 for (i = 1; i < 4; i++) {
541 if (test_access(i, stp))
542 nfs4_file_put_access(stp->st_file,
543 nfs4_access_to_omode(i));
544 clear_access(i, stp);
548 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
550 list_del(&stp->st_perfile);
551 list_del(&stp->st_perstateowner);
554 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
556 release_all_access(stp);
557 put_nfs4_file(stp->st_file);
561 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
563 kmem_cache_free(stateid_slab, stp);
566 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
570 unhash_generic_stateid(stp);
571 unhash_stid(&stp->st_stid);
572 file = find_any_file(stp->st_file);
574 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
575 close_generic_stateid(stp);
576 free_generic_stateid(stp);
579 static void unhash_lockowner(struct nfs4_lockowner *lo)
581 struct nfs4_ol_stateid *stp;
583 list_del(&lo->lo_owner.so_strhash);
584 list_del(&lo->lo_perstateid);
585 list_del(&lo->lo_owner_ino_hash);
586 while (!list_empty(&lo->lo_owner.so_stateids)) {
587 stp = list_first_entry(&lo->lo_owner.so_stateids,
588 struct nfs4_ol_stateid, st_perstateowner);
589 release_lock_stateid(stp);
593 static void release_lockowner(struct nfs4_lockowner *lo)
595 unhash_lockowner(lo);
596 nfs4_free_lockowner(lo);
600 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
602 struct nfs4_lockowner *lo;
604 while (!list_empty(&open_stp->st_lockowners)) {
605 lo = list_entry(open_stp->st_lockowners.next,
606 struct nfs4_lockowner, lo_perstateid);
607 release_lockowner(lo);
611 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
613 unhash_generic_stateid(stp);
614 release_stateid_lockowners(stp);
615 close_generic_stateid(stp);
618 static void release_open_stateid(struct nfs4_ol_stateid *stp)
620 unhash_open_stateid(stp);
621 unhash_stid(&stp->st_stid);
622 free_generic_stateid(stp);
625 static void unhash_openowner(struct nfs4_openowner *oo)
627 struct nfs4_ol_stateid *stp;
629 list_del(&oo->oo_owner.so_strhash);
630 list_del(&oo->oo_perclient);
631 while (!list_empty(&oo->oo_owner.so_stateids)) {
632 stp = list_first_entry(&oo->oo_owner.so_stateids,
633 struct nfs4_ol_stateid, st_perstateowner);
634 release_open_stateid(stp);
638 static void release_last_closed_stateid(struct nfs4_openowner *oo)
640 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
643 unhash_stid(&s->st_stid);
644 free_generic_stateid(s);
645 oo->oo_last_closed_stid = NULL;
649 static void release_openowner(struct nfs4_openowner *oo)
651 unhash_openowner(oo);
652 list_del(&oo->oo_close_lru);
653 release_last_closed_stateid(oo);
654 nfs4_free_openowner(oo);
657 #define SESSION_HASH_SIZE 512
658 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
661 hash_sessionid(struct nfs4_sessionid *sessionid)
663 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
665 return sid->sequence % SESSION_HASH_SIZE;
670 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
672 u32 *ptr = (u32 *)(&sessionid->data[0]);
673 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
677 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
684 gen_sessionid(struct nfsd4_session *ses)
686 struct nfs4_client *clp = ses->se_client;
687 struct nfsd4_sessionid *sid;
689 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
690 sid->clientid = clp->cl_clientid;
691 sid->sequence = current_sessionid++;
696 * The protocol defines ca_maxresponssize_cached to include the size of
697 * the rpc header, but all we need to cache is the data starting after
698 * the end of the initial SEQUENCE operation--the rest we regenerate
699 * each time. Therefore we can advertise a ca_maxresponssize_cached
700 * value that is the number of bytes in our cache plus a few additional
701 * bytes. In order to stay on the safe side, and not promise more than
702 * we can cache, those additional bytes must be the minimum possible: 24
703 * bytes of rpc header (xid through accept state, with AUTH_NULL
704 * verifier), 12 for the compound header (with zero-length tag), and 44
705 * for the SEQUENCE op response:
707 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
710 free_session_slots(struct nfsd4_session *ses)
714 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
715 kfree(ses->se_slots[i]);
719 * We don't actually need to cache the rpc and session headers, so we
720 * can allocate a little less for each slot:
722 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
724 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
727 static int nfsd4_sanitize_slot_size(u32 size)
729 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
730 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
736 * XXX: If we run out of reserved DRC memory we could (up to a point)
737 * re-negotiate active sessions and reduce their slot usage to make
738 * room for new connections. For now we just fail the create session.
740 static int nfsd4_get_drc_mem(int slotsize, u32 num)
744 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
746 spin_lock(&nfsd_drc_lock);
747 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
748 nfsd_drc_max_mem - nfsd_drc_mem_used);
749 num = min_t(int, num, avail / slotsize);
750 nfsd_drc_mem_used += num * slotsize;
751 spin_unlock(&nfsd_drc_lock);
756 static void nfsd4_put_drc_mem(int slotsize, int num)
758 spin_lock(&nfsd_drc_lock);
759 nfsd_drc_mem_used -= slotsize * num;
760 spin_unlock(&nfsd_drc_lock);
763 static struct nfsd4_session *__alloc_session(int slotsize, int numslots)
765 struct nfsd4_session *new;
768 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
769 + sizeof(struct nfsd4_session) > PAGE_SIZE);
770 mem = numslots * sizeof(struct nfsd4_slot *);
772 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
775 /* allocate each struct nfsd4_slot and data cache in one piece */
776 for (i = 0; i < numslots; i++) {
777 mem = sizeof(struct nfsd4_slot) + slotsize;
778 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
779 if (!new->se_slots[i])
785 kfree(new->se_slots[i]);
790 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
792 u32 maxrpc = nfsd_serv->sv_max_mesg;
794 new->maxreqs = numslots;
795 new->maxresp_cached = min_t(u32, req->maxresp_cached,
796 slotsize + NFSD_MIN_HDR_SEQ_SZ);
797 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
798 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
799 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
802 static void free_conn(struct nfsd4_conn *c)
804 svc_xprt_put(c->cn_xprt);
808 static void nfsd4_conn_lost(struct svc_xpt_user *u)
810 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
811 struct nfs4_client *clp = c->cn_session->se_client;
813 spin_lock(&clp->cl_lock);
814 if (!list_empty(&c->cn_persession)) {
815 list_del(&c->cn_persession);
818 spin_unlock(&clp->cl_lock);
819 nfsd4_probe_callback(clp);
822 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
824 struct nfsd4_conn *conn;
826 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
829 svc_xprt_get(rqstp->rq_xprt);
830 conn->cn_xprt = rqstp->rq_xprt;
831 conn->cn_flags = flags;
832 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
836 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
838 conn->cn_session = ses;
839 list_add(&conn->cn_persession, &ses->se_conns);
842 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
844 struct nfs4_client *clp = ses->se_client;
846 spin_lock(&clp->cl_lock);
847 __nfsd4_hash_conn(conn, ses);
848 spin_unlock(&clp->cl_lock);
851 static int nfsd4_register_conn(struct nfsd4_conn *conn)
853 conn->cn_xpt_user.callback = nfsd4_conn_lost;
854 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
857 static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
861 nfsd4_hash_conn(conn, ses);
862 ret = nfsd4_register_conn(conn);
864 /* oops; xprt is already down: */
865 nfsd4_conn_lost(&conn->cn_xpt_user);
866 if (conn->cn_flags & NFS4_CDFC4_BACK) {
867 /* callback channel may be back up */
868 nfsd4_probe_callback(ses->se_client);
872 static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
874 u32 dir = NFS4_CDFC4_FORE;
876 if (cses->flags & SESSION4_BACK_CHAN)
877 dir |= NFS4_CDFC4_BACK;
878 return alloc_conn(rqstp, dir);
881 /* must be called under client_lock */
882 static void nfsd4_del_conns(struct nfsd4_session *s)
884 struct nfs4_client *clp = s->se_client;
885 struct nfsd4_conn *c;
887 spin_lock(&clp->cl_lock);
888 while (!list_empty(&s->se_conns)) {
889 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
890 list_del_init(&c->cn_persession);
891 spin_unlock(&clp->cl_lock);
893 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
896 spin_lock(&clp->cl_lock);
898 spin_unlock(&clp->cl_lock);
901 static void __free_session(struct nfsd4_session *ses)
903 nfsd4_put_drc_mem(slot_bytes(&ses->se_fchannel), ses->se_fchannel.maxreqs);
904 free_session_slots(ses);
908 static void free_session(struct kref *kref)
910 struct nfsd4_session *ses;
912 lockdep_assert_held(&client_lock);
913 ses = container_of(kref, struct nfsd4_session, se_ref);
914 nfsd4_del_conns(ses);
918 void nfsd4_put_session(struct nfsd4_session *ses)
920 spin_lock(&client_lock);
921 nfsd4_put_session_locked(ses);
922 spin_unlock(&client_lock);
925 static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fchan)
927 struct nfsd4_session *new;
928 int numslots, slotsize;
930 * Note decreasing slot size below client's request may
931 * make it difficult for client to function correctly, whereas
932 * decreasing the number of slots will (just?) affect
933 * performance. When short on memory we therefore prefer to
934 * decrease number of slots instead of their size.
936 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
937 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
941 new = __alloc_session(slotsize, numslots);
943 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
946 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
950 void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
954 new->se_client = clp;
957 INIT_LIST_HEAD(&new->se_conns);
959 new->se_cb_seq_nr = 1;
960 new->se_flags = cses->flags;
961 new->se_cb_prog = cses->callback_prog;
962 new->se_cb_sec = cses->cb_sec;
963 kref_init(&new->se_ref);
964 idx = hash_sessionid(&new->se_sessionid);
965 spin_lock(&client_lock);
966 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
967 spin_lock(&clp->cl_lock);
968 list_add(&new->se_perclnt, &clp->cl_sessions);
969 spin_unlock(&clp->cl_lock);
970 spin_unlock(&client_lock);
972 if (cses->flags & SESSION4_BACK_CHAN) {
973 struct sockaddr *sa = svc_addr(rqstp);
975 * This is a little silly; with sessions there's no real
976 * use for the callback address. Use the peer address
977 * as a reasonable default for now, but consider fixing
978 * the rpc client not to require an address in the
981 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
982 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
986 /* caller must hold client_lock */
987 static struct nfsd4_session *
988 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
990 struct nfsd4_session *elem;
993 dump_sessionid(__func__, sessionid);
994 idx = hash_sessionid(sessionid);
995 /* Search in the appropriate list */
996 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
997 if (!memcmp(elem->se_sessionid.data, sessionid->data,
998 NFS4_MAX_SESSIONID_LEN)) {
1003 dprintk("%s: session not found\n", __func__);
1007 /* caller must hold client_lock */
1009 unhash_session(struct nfsd4_session *ses)
1011 list_del(&ses->se_hash);
1012 spin_lock(&ses->se_client->cl_lock);
1013 list_del(&ses->se_perclnt);
1014 spin_unlock(&ses->se_client->cl_lock);
1017 /* must be called under the client_lock */
1019 renew_client_locked(struct nfs4_client *clp)
1021 if (is_client_expired(clp)) {
1023 printk("%s: client (clientid %08x/%08x) already expired\n",
1025 clp->cl_clientid.cl_boot,
1026 clp->cl_clientid.cl_id);
1030 dprintk("renewing client (clientid %08x/%08x)\n",
1031 clp->cl_clientid.cl_boot,
1032 clp->cl_clientid.cl_id);
1033 list_move_tail(&clp->cl_lru, &client_lru);
1034 clp->cl_time = get_seconds();
1038 renew_client(struct nfs4_client *clp)
1040 spin_lock(&client_lock);
1041 renew_client_locked(clp);
1042 spin_unlock(&client_lock);
1045 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1047 STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1049 if (clid->cl_boot == nn->boot_time)
1051 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1052 clid->cl_boot, clid->cl_id, nn->boot_time);
1057 * XXX Should we use a slab cache ?
1058 * This type of memory management is somewhat inefficient, but we use it
1059 * anyway since SETCLIENTID is not a common operation.
1061 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1063 struct nfs4_client *clp;
1065 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1068 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1069 if (clp->cl_name.data == NULL) {
1073 clp->cl_name.len = name.len;
1078 free_client(struct nfs4_client *clp)
1080 lockdep_assert_held(&client_lock);
1081 while (!list_empty(&clp->cl_sessions)) {
1082 struct nfsd4_session *ses;
1083 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1085 list_del(&ses->se_perclnt);
1086 nfsd4_put_session_locked(ses);
1088 free_svc_cred(&clp->cl_cred);
1089 kfree(clp->cl_name.data);
1094 release_session_client(struct nfsd4_session *session)
1096 struct nfs4_client *clp = session->se_client;
1098 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1100 if (is_client_expired(clp)) {
1102 session->se_client = NULL;
1104 renew_client_locked(clp);
1105 spin_unlock(&client_lock);
1108 /* must be called under the client_lock */
1110 unhash_client_locked(struct nfs4_client *clp)
1112 struct nfsd4_session *ses;
1114 mark_client_expired(clp);
1115 list_del(&clp->cl_lru);
1116 spin_lock(&clp->cl_lock);
1117 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1118 list_del_init(&ses->se_hash);
1119 spin_unlock(&clp->cl_lock);
1123 destroy_client(struct nfs4_client *clp)
1125 struct nfs4_openowner *oo;
1126 struct nfs4_delegation *dp;
1127 struct list_head reaplist;
1129 INIT_LIST_HEAD(&reaplist);
1130 spin_lock(&recall_lock);
1131 while (!list_empty(&clp->cl_delegations)) {
1132 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1133 list_del_init(&dp->dl_perclnt);
1134 list_move(&dp->dl_recall_lru, &reaplist);
1136 spin_unlock(&recall_lock);
1137 while (!list_empty(&reaplist)) {
1138 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1139 unhash_delegation(dp);
1141 while (!list_empty(&clp->cl_openowners)) {
1142 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1143 release_openowner(oo);
1145 nfsd4_shutdown_callback(clp);
1146 if (clp->cl_cb_conn.cb_xprt)
1147 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1148 list_del(&clp->cl_idhash);
1149 if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1150 rb_erase(&clp->cl_namenode, &conf_name_tree);
1152 rb_erase(&clp->cl_namenode, &unconf_name_tree);
1153 spin_lock(&client_lock);
1154 unhash_client_locked(clp);
1155 if (atomic_read(&clp->cl_refcount) == 0)
1157 spin_unlock(&client_lock);
1160 static void expire_client(struct nfs4_client *clp)
1162 nfsd4_client_record_remove(clp);
1163 destroy_client(clp);
1166 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1168 memcpy(target->cl_verifier.data, source->data,
1169 sizeof(target->cl_verifier.data));
1172 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1174 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1175 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1178 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1180 if (source->cr_principal) {
1181 target->cr_principal =
1182 kstrdup(source->cr_principal, GFP_KERNEL);
1183 if (target->cr_principal == NULL)
1186 target->cr_principal = NULL;
1187 target->cr_flavor = source->cr_flavor;
1188 target->cr_uid = source->cr_uid;
1189 target->cr_gid = source->cr_gid;
1190 target->cr_group_info = source->cr_group_info;
1191 get_group_info(target->cr_group_info);
1196 compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1200 res = o1->len - o2->len;
1203 return (long long)memcmp(o1->data, o2->data, o1->len);
1206 static int same_name(const char *n1, const char *n2)
1208 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1212 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1214 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1218 same_clid(clientid_t *cl1, clientid_t *cl2)
1220 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1223 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1227 if (g1->ngroups != g2->ngroups)
1229 for (i=0; i<g1->ngroups; i++)
1230 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1236 * RFC 3530 language requires clid_inuse be returned when the
1237 * "principal" associated with a requests differs from that previously
1238 * used. We use uid, gid's, and gss principal string as our best
1239 * approximation. We also don't want to allow non-gss use of a client
1240 * established using gss: in theory cr_principal should catch that
1241 * change, but in practice cr_principal can be null even in the gss case
1242 * since gssd doesn't always pass down a principal string.
1244 static bool is_gss_cred(struct svc_cred *cr)
1246 /* Is cr_flavor one of the gss "pseudoflavors"?: */
1247 return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1252 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1254 if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1255 || (cr1->cr_uid != cr2->cr_uid)
1256 || (cr1->cr_gid != cr2->cr_gid)
1257 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1259 if (cr1->cr_principal == cr2->cr_principal)
1261 if (!cr1->cr_principal || !cr2->cr_principal)
1263 return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1266 static void gen_clid(struct nfs4_client *clp)
1268 static u32 current_clientid = 1;
1269 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
1271 clp->cl_clientid.cl_boot = nn->boot_time;
1272 clp->cl_clientid.cl_id = current_clientid++;
1275 static void gen_confirm(struct nfs4_client *clp)
1280 verf[0] = (__be32)get_seconds();
1281 verf[1] = (__be32)i++;
1282 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1285 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1287 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1290 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1292 struct nfs4_stid *s;
1294 s = find_stateid(cl, t);
1297 if (typemask & s->sc_type)
1302 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1303 struct svc_rqst *rqstp, nfs4_verifier *verf)
1305 struct nfs4_client *clp;
1306 struct sockaddr *sa = svc_addr(rqstp);
1309 clp = alloc_client(name);
1313 INIT_LIST_HEAD(&clp->cl_sessions);
1314 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1316 spin_lock(&client_lock);
1318 spin_unlock(&client_lock);
1321 idr_init(&clp->cl_stateids);
1322 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1323 atomic_set(&clp->cl_refcount, 0);
1324 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1325 INIT_LIST_HEAD(&clp->cl_idhash);
1326 INIT_LIST_HEAD(&clp->cl_openowners);
1327 INIT_LIST_HEAD(&clp->cl_delegations);
1328 INIT_LIST_HEAD(&clp->cl_lru);
1329 INIT_LIST_HEAD(&clp->cl_callbacks);
1330 spin_lock_init(&clp->cl_lock);
1331 nfsd4_init_callback(&clp->cl_cb_null);
1332 clp->cl_time = get_seconds();
1333 clear_bit(0, &clp->cl_cb_slot_busy);
1334 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1335 copy_verf(clp, verf);
1336 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1338 clp->cl_cb_session = NULL;
1343 add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1345 struct rb_node **new = &(root->rb_node), *parent = NULL;
1346 struct nfs4_client *clp;
1349 clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1352 if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1353 new = &((*new)->rb_left);
1355 new = &((*new)->rb_right);
1358 rb_link_node(&new_clp->cl_namenode, parent, new);
1359 rb_insert_color(&new_clp->cl_namenode, root);
1362 static struct nfs4_client *
1363 find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1366 struct rb_node *node = root->rb_node;
1367 struct nfs4_client *clp;
1370 clp = rb_entry(node, struct nfs4_client, cl_namenode);
1371 cmp = compare_blob(&clp->cl_name, name);
1373 node = node->rb_left;
1375 node = node->rb_right;
1383 add_to_unconfirmed(struct nfs4_client *clp)
1385 unsigned int idhashval;
1387 clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1388 add_clp_to_name_tree(clp, &unconf_name_tree);
1389 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1390 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1395 move_to_confirmed(struct nfs4_client *clp)
1397 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1399 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1400 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1401 rb_erase(&clp->cl_namenode, &unconf_name_tree);
1402 add_clp_to_name_tree(clp, &conf_name_tree);
1403 set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
1407 static struct nfs4_client *
1408 find_confirmed_client(clientid_t *clid, bool sessions)
1410 struct nfs4_client *clp;
1411 unsigned int idhashval = clientid_hashval(clid->cl_id);
1413 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1414 if (same_clid(&clp->cl_clientid, clid)) {
1415 if ((bool)clp->cl_minorversion != sessions)
1424 static struct nfs4_client *
1425 find_unconfirmed_client(clientid_t *clid, bool sessions)
1427 struct nfs4_client *clp;
1428 unsigned int idhashval = clientid_hashval(clid->cl_id);
1430 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1431 if (same_clid(&clp->cl_clientid, clid)) {
1432 if ((bool)clp->cl_minorversion != sessions)
1440 static bool clp_used_exchangeid(struct nfs4_client *clp)
1442 return clp->cl_exchange_flags != 0;
1445 static struct nfs4_client *
1446 find_confirmed_client_by_name(struct xdr_netobj *name)
1448 return find_clp_in_name_tree(name, &conf_name_tree);
1451 static struct nfs4_client *
1452 find_unconfirmed_client_by_name(struct xdr_netobj *name)
1454 return find_clp_in_name_tree(name, &unconf_name_tree);
1458 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1460 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1461 struct sockaddr *sa = svc_addr(rqstp);
1462 u32 scopeid = rpc_get_scope_id(sa);
1463 unsigned short expected_family;
1465 /* Currently, we only support tcp and tcp6 for the callback channel */
1466 if (se->se_callback_netid_len == 3 &&
1467 !memcmp(se->se_callback_netid_val, "tcp", 3))
1468 expected_family = AF_INET;
1469 else if (se->se_callback_netid_len == 4 &&
1470 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1471 expected_family = AF_INET6;
1475 conn->cb_addrlen = rpc_uaddr2sockaddr(&init_net, se->se_callback_addr_val,
1476 se->se_callback_addr_len,
1477 (struct sockaddr *)&conn->cb_addr,
1478 sizeof(conn->cb_addr));
1480 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1483 if (conn->cb_addr.ss_family == AF_INET6)
1484 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1486 conn->cb_prog = se->se_callback_prog;
1487 conn->cb_ident = se->se_callback_ident;
1488 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1491 conn->cb_addr.ss_family = AF_UNSPEC;
1492 conn->cb_addrlen = 0;
1493 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1494 "will not receive delegations\n",
1495 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1501 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1504 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1506 struct nfsd4_slot *slot = resp->cstate.slot;
1509 dprintk("--> %s slot %p\n", __func__, slot);
1511 slot->sl_opcnt = resp->opcnt;
1512 slot->sl_status = resp->cstate.status;
1514 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1515 if (nfsd4_not_cached(resp)) {
1516 slot->sl_datalen = 0;
1519 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1520 base = (char *)resp->cstate.datap -
1521 (char *)resp->xbuf->head[0].iov_base;
1522 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1524 WARN("%s: sessions DRC could not cache compound\n", __func__);
1529 * Encode the replay sequence operation from the slot values.
1530 * If cachethis is FALSE encode the uncached rep error on the next
1531 * operation which sets resp->p and increments resp->opcnt for
1532 * nfs4svc_encode_compoundres.
1536 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1537 struct nfsd4_compoundres *resp)
1539 struct nfsd4_op *op;
1540 struct nfsd4_slot *slot = resp->cstate.slot;
1542 /* Encode the replayed sequence operation */
1543 op = &args->ops[resp->opcnt - 1];
1544 nfsd4_encode_operation(resp, op);
1546 /* Return nfserr_retry_uncached_rep in next operation. */
1547 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1548 op = &args->ops[resp->opcnt++];
1549 op->status = nfserr_retry_uncached_rep;
1550 nfsd4_encode_operation(resp, op);
1556 * The sequence operation is not cached because we can use the slot and
1560 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1561 struct nfsd4_sequence *seq)
1563 struct nfsd4_slot *slot = resp->cstate.slot;
1566 dprintk("--> %s slot %p\n", __func__, slot);
1568 /* Either returns 0 or nfserr_retry_uncached */
1569 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1570 if (status == nfserr_retry_uncached_rep)
1573 /* The sequence operation has been encoded, cstate->datap set. */
1574 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1576 resp->opcnt = slot->sl_opcnt;
1577 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1578 status = slot->sl_status;
1584 * Set the exchange_id flags returned by the server.
1587 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1589 /* pNFS is not supported */
1590 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1592 /* Referrals are supported, Migration is not. */
1593 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1595 /* set the wire flags to return to client. */
1596 clid->flags = new->cl_exchange_flags;
1599 static bool client_has_state(struct nfs4_client *clp)
1602 * Note clp->cl_openowners check isn't quite right: there's no
1603 * need to count owners without stateid's.
1605 * Also note we should probably be using this in 4.0 case too.
1607 return !list_empty(&clp->cl_openowners)
1608 || !list_empty(&clp->cl_delegations)
1609 || !list_empty(&clp->cl_sessions);
1613 nfsd4_exchange_id(struct svc_rqst *rqstp,
1614 struct nfsd4_compound_state *cstate,
1615 struct nfsd4_exchange_id *exid)
1617 struct nfs4_client *unconf, *conf, *new;
1619 char dname[HEXDIR_LEN];
1620 char addr_str[INET6_ADDRSTRLEN];
1621 nfs4_verifier verf = exid->verifier;
1622 struct sockaddr *sa = svc_addr(rqstp);
1623 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1625 rpc_ntop(sa, addr_str, sizeof(addr_str));
1626 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1627 "ip_addr=%s flags %x, spa_how %d\n",
1628 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1629 addr_str, exid->flags, exid->spa_how);
1631 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1632 return nfserr_inval;
1634 /* Currently only support SP4_NONE */
1635 switch (exid->spa_how) {
1639 return nfserr_serverfault;
1641 BUG(); /* checked by xdr code */
1643 return nfserr_serverfault; /* no excuse :-/ */
1646 status = nfs4_make_rec_clidname(dname, &exid->clname);
1651 /* Cases below refer to rfc 5661 section 18.35.4: */
1653 conf = find_confirmed_client_by_name(&exid->clname);
1655 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1656 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1659 if (!clp_used_exchangeid(conf)) { /* buggy client */
1660 status = nfserr_inval;
1663 if (!creds_match) { /* case 9 */
1664 status = nfserr_perm;
1667 if (!verfs_match) { /* case 8 */
1668 status = nfserr_not_same;
1672 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1676 if (!creds_match) { /* case 3 */
1677 if (client_has_state(conf)) {
1678 status = nfserr_clid_inuse;
1681 expire_client(conf);
1684 if (verfs_match) { /* case 2 */
1685 conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
1689 /* case 5, client reboot */
1693 if (update) { /* case 7 */
1694 status = nfserr_noent;
1698 unconf = find_unconfirmed_client_by_name(&exid->clname);
1699 if (unconf) /* case 4, possible retry or client restart */
1700 expire_client(unconf);
1702 /* case 1 (normal case) */
1704 new = create_client(exid->clname, dname, rqstp, &verf);
1706 status = nfserr_jukebox;
1709 new->cl_minorversion = 1;
1712 add_to_unconfirmed(new);
1714 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1715 exid->clientid.cl_id = new->cl_clientid.cl_id;
1717 exid->seqid = new->cl_cs_slot.sl_seqid + 1;
1718 nfsd4_set_ex_flags(new, exid);
1720 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1721 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1725 nfs4_unlock_state();
1730 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1732 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1735 /* The slot is in use, and no response has been sent. */
1737 if (seqid == slot_seqid)
1738 return nfserr_jukebox;
1740 return nfserr_seq_misordered;
1742 /* Note unsigned 32-bit arithmetic handles wraparound: */
1743 if (likely(seqid == slot_seqid + 1))
1745 if (seqid == slot_seqid)
1746 return nfserr_replay_cache;
1747 return nfserr_seq_misordered;
1751 * Cache the create session result into the create session single DRC
1752 * slot cache by saving the xdr structure. sl_seqid has been set.
1753 * Do this for solo or embedded create session operations.
1756 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1757 struct nfsd4_clid_slot *slot, __be32 nfserr)
1759 slot->sl_status = nfserr;
1760 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1764 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1765 struct nfsd4_clid_slot *slot)
1767 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1768 return slot->sl_status;
1771 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1772 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1773 1 + /* MIN tag is length with zero, only length */ \
1774 3 + /* version, opcount, opcode */ \
1775 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1776 /* seqid, slotID, slotID, cache */ \
1777 4 ) * sizeof(__be32))
1779 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1780 2 + /* verifier: AUTH_NULL, length 0 */\
1782 1 + /* MIN tag is length with zero, only length */ \
1783 3 + /* opcount, opcode, opstatus*/ \
1784 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1785 /* seqid, slotID, slotID, slotID, status */ \
1786 5 ) * sizeof(__be32))
1788 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1790 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1791 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1795 nfsd4_create_session(struct svc_rqst *rqstp,
1796 struct nfsd4_compound_state *cstate,
1797 struct nfsd4_create_session *cr_ses)
1799 struct sockaddr *sa = svc_addr(rqstp);
1800 struct nfs4_client *conf, *unconf;
1801 struct nfsd4_session *new;
1802 struct nfsd4_conn *conn;
1803 struct nfsd4_clid_slot *cs_slot = NULL;
1806 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1807 return nfserr_inval;
1808 if (check_forechannel_attrs(cr_ses->fore_channel))
1809 return nfserr_toosmall;
1810 new = alloc_session(&cr_ses->fore_channel);
1812 return nfserr_jukebox;
1813 status = nfserr_jukebox;
1814 conn = alloc_conn_from_crses(rqstp, cr_ses);
1816 goto out_free_session;
1819 unconf = find_unconfirmed_client(&cr_ses->clientid, true);
1820 conf = find_confirmed_client(&cr_ses->clientid, true);
1823 cs_slot = &conf->cl_cs_slot;
1824 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1825 if (status == nfserr_replay_cache) {
1826 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1828 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1829 status = nfserr_seq_misordered;
1832 } else if (unconf) {
1833 struct nfs4_client *old;
1834 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1835 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1836 status = nfserr_clid_inuse;
1839 cs_slot = &unconf->cl_cs_slot;
1840 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1842 /* an unconfirmed replay returns misordered */
1843 status = nfserr_seq_misordered;
1846 old = find_confirmed_client_by_name(&unconf->cl_name);
1849 move_to_confirmed(unconf);
1852 status = nfserr_stale_clientid;
1857 * We do not support RDMA or persistent sessions
1859 cr_ses->flags &= ~SESSION4_PERSIST;
1860 cr_ses->flags &= ~SESSION4_RDMA;
1862 init_session(rqstp, new, conf, cr_ses);
1863 nfsd4_init_conn(rqstp, conn, new);
1865 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1866 NFS4_MAX_SESSIONID_LEN);
1867 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1868 sizeof(struct nfsd4_channel_attrs));
1869 cs_slot->sl_seqid++;
1870 cr_ses->seqid = cs_slot->sl_seqid;
1872 /* cache solo and embedded create sessions under the state lock */
1873 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1875 nfs4_unlock_state();
1876 dprintk("%s returns %d\n", __func__, ntohl(status));
1881 __free_session(new);
1885 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1887 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1888 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1890 return argp->opcnt == resp->opcnt;
1893 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1896 case NFS4_CDFC4_FORE:
1897 case NFS4_CDFC4_BACK:
1899 case NFS4_CDFC4_FORE_OR_BOTH:
1900 case NFS4_CDFC4_BACK_OR_BOTH:
1901 *dir = NFS4_CDFC4_BOTH;
1904 return nfserr_inval;
1907 __be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
1909 struct nfsd4_session *session = cstate->session;
1911 spin_lock(&client_lock);
1912 session->se_cb_prog = bc->bc_cb_program;
1913 session->se_cb_sec = bc->bc_cb_sec;
1914 spin_unlock(&client_lock);
1916 nfsd4_probe_callback(session->se_client);
1921 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1922 struct nfsd4_compound_state *cstate,
1923 struct nfsd4_bind_conn_to_session *bcts)
1926 struct nfsd4_conn *conn;
1928 if (!nfsd4_last_compound_op(rqstp))
1929 return nfserr_not_only_op;
1930 spin_lock(&client_lock);
1931 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1932 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1933 * client_lock iself: */
1934 if (cstate->session) {
1935 nfsd4_get_session(cstate->session);
1936 atomic_inc(&cstate->session->se_client->cl_refcount);
1938 spin_unlock(&client_lock);
1939 if (!cstate->session)
1940 return nfserr_badsession;
1942 status = nfsd4_map_bcts_dir(&bcts->dir);
1945 conn = alloc_conn(rqstp, bcts->dir);
1947 return nfserr_jukebox;
1948 nfsd4_init_conn(rqstp, conn, cstate->session);
1952 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1956 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1960 nfsd4_destroy_session(struct svc_rqst *r,
1961 struct nfsd4_compound_state *cstate,
1962 struct nfsd4_destroy_session *sessionid)
1964 struct nfsd4_session *ses;
1965 __be32 status = nfserr_badsession;
1968 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1969 * - Should we return nfserr_back_chan_busy if waiting for
1970 * callbacks on to-be-destroyed session?
1971 * - Do we need to clear any callback info from previous session?
1974 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1975 if (!nfsd4_last_compound_op(r))
1976 return nfserr_not_only_op;
1978 dump_sessionid(__func__, &sessionid->sessionid);
1979 spin_lock(&client_lock);
1980 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1982 spin_unlock(&client_lock);
1986 unhash_session(ses);
1987 spin_unlock(&client_lock);
1990 nfsd4_probe_callback_sync(ses->se_client);
1991 nfs4_unlock_state();
1993 spin_lock(&client_lock);
1994 nfsd4_del_conns(ses);
1995 nfsd4_put_session_locked(ses);
1996 spin_unlock(&client_lock);
1999 dprintk("%s returns %d\n", __func__, ntohl(status));
2003 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2005 struct nfsd4_conn *c;
2007 list_for_each_entry(c, &s->se_conns, cn_persession) {
2008 if (c->cn_xprt == xpt) {
2015 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2017 struct nfs4_client *clp = ses->se_client;
2018 struct nfsd4_conn *c;
2021 spin_lock(&clp->cl_lock);
2022 c = __nfsd4_find_conn(new->cn_xprt, ses);
2024 spin_unlock(&clp->cl_lock);
2028 __nfsd4_hash_conn(new, ses);
2029 spin_unlock(&clp->cl_lock);
2030 ret = nfsd4_register_conn(new);
2032 /* oops; xprt is already down: */
2033 nfsd4_conn_lost(&new->cn_xpt_user);
2037 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2039 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2041 return args->opcnt > session->se_fchannel.maxops;
2044 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2045 struct nfsd4_session *session)
2047 struct xdr_buf *xb = &rqstp->rq_arg;
2049 return xb->len > session->se_fchannel.maxreq_sz;
2053 nfsd4_sequence(struct svc_rqst *rqstp,
2054 struct nfsd4_compound_state *cstate,
2055 struct nfsd4_sequence *seq)
2057 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2058 struct nfsd4_session *session;
2059 struct nfsd4_slot *slot;
2060 struct nfsd4_conn *conn;
2063 if (resp->opcnt != 1)
2064 return nfserr_sequence_pos;
2067 * Will be either used or freed by nfsd4_sequence_check_conn
2070 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2072 return nfserr_jukebox;
2074 spin_lock(&client_lock);
2075 status = nfserr_badsession;
2076 session = find_in_sessionid_hashtbl(&seq->sessionid);
2080 status = nfserr_too_many_ops;
2081 if (nfsd4_session_too_many_ops(rqstp, session))
2084 status = nfserr_req_too_big;
2085 if (nfsd4_request_too_big(rqstp, session))
2088 status = nfserr_badslot;
2089 if (seq->slotid >= session->se_fchannel.maxreqs)
2092 slot = session->se_slots[seq->slotid];
2093 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2095 /* We do not negotiate the number of slots yet, so set the
2096 * maxslots to the session maxreqs which is used to encode
2097 * sr_highest_slotid and the sr_target_slot id to maxslots */
2098 seq->maxslots = session->se_fchannel.maxreqs;
2100 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2101 slot->sl_flags & NFSD4_SLOT_INUSE);
2102 if (status == nfserr_replay_cache) {
2103 status = nfserr_seq_misordered;
2104 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2106 cstate->slot = slot;
2107 cstate->session = session;
2108 /* Return the cached reply status and set cstate->status
2109 * for nfsd4_proc_compound processing */
2110 status = nfsd4_replay_cache_entry(resp, seq);
2111 cstate->status = nfserr_replay_cache;
2117 nfsd4_sequence_check_conn(conn, session);
2120 /* Success! bump slot seqid */
2121 slot->sl_seqid = seq->seqid;
2122 slot->sl_flags |= NFSD4_SLOT_INUSE;
2124 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2126 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2128 cstate->slot = slot;
2129 cstate->session = session;
2132 /* Hold a session reference until done processing the compound. */
2133 if (cstate->session) {
2134 struct nfs4_client *clp = session->se_client;
2136 nfsd4_get_session(cstate->session);
2137 atomic_inc(&clp->cl_refcount);
2138 switch (clp->cl_cb_state) {
2140 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2142 case NFSD4_CB_FAULT:
2143 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2146 seq->status_flags = 0;
2150 spin_unlock(&client_lock);
2151 dprintk("%s: return %d\n", __func__, ntohl(status));
2156 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2158 struct nfs4_client *conf, *unconf, *clp;
2162 unconf = find_unconfirmed_client(&dc->clientid, true);
2163 conf = find_confirmed_client(&dc->clientid, true);
2168 if (!is_client_expired(conf) && client_has_state(conf)) {
2169 status = nfserr_clientid_busy;
2173 /* rfc5661 18.50.3 */
2174 if (cstate->session && conf == cstate->session->se_client) {
2175 status = nfserr_clientid_busy;
2181 status = nfserr_stale_clientid;
2187 nfs4_unlock_state();
2188 dprintk("%s return %d\n", __func__, ntohl(status));
2193 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2197 if (rc->rca_one_fs) {
2198 if (!cstate->current_fh.fh_dentry)
2199 return nfserr_nofilehandle;
2201 * We don't take advantage of the rca_one_fs case.
2202 * That's OK, it's optional, we can safely ignore it.
2208 status = nfserr_complete_already;
2209 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2210 &cstate->session->se_client->cl_flags))
2213 status = nfserr_stale_clientid;
2214 if (is_client_expired(cstate->session->se_client))
2216 * The following error isn't really legal.
2217 * But we only get here if the client just explicitly
2218 * destroyed the client. Surely it no longer cares what
2219 * error it gets back on an operation for the dead
2225 nfsd4_client_record_create(cstate->session->se_client);
2227 nfs4_unlock_state();
2232 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2233 struct nfsd4_setclientid *setclid)
2235 struct xdr_netobj clname = setclid->se_name;
2236 nfs4_verifier clverifier = setclid->se_verf;
2237 struct nfs4_client *conf, *unconf, *new;
2239 char dname[HEXDIR_LEN];
2241 status = nfs4_make_rec_clidname(dname, &clname);
2245 /* Cases below refer to rfc 3530 section 14.2.33: */
2247 conf = find_confirmed_client_by_name(&clname);
2250 status = nfserr_clid_inuse;
2251 if (clp_used_exchangeid(conf))
2253 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2254 char addr_str[INET6_ADDRSTRLEN];
2255 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2257 dprintk("NFSD: setclientid: string in use by client "
2258 "at %s\n", addr_str);
2262 unconf = find_unconfirmed_client_by_name(&clname);
2264 expire_client(unconf);
2265 status = nfserr_jukebox;
2266 new = create_client(clname, dname, rqstp, &clverifier);
2269 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2270 /* case 1: probable callback update */
2271 copy_clid(new, conf);
2272 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2274 new->cl_minorversion = 0;
2275 gen_callback(new, setclid, rqstp);
2276 add_to_unconfirmed(new);
2277 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2278 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2279 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2282 nfs4_unlock_state();
2288 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2289 struct nfsd4_compound_state *cstate,
2290 struct nfsd4_setclientid_confirm *setclientid_confirm)
2292 struct nfs4_client *conf, *unconf;
2293 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2294 clientid_t * clid = &setclientid_confirm->sc_clientid;
2296 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2298 if (STALE_CLIENTID(clid, nn))
2299 return nfserr_stale_clientid;
2302 conf = find_confirmed_client(clid, false);
2303 unconf = find_unconfirmed_client(clid, false);
2305 * We try hard to give out unique clientid's, so if we get an
2306 * attempt to confirm the same clientid with a different cred,
2307 * there's a bug somewhere. Let's charitably assume it's our
2310 status = nfserr_serverfault;
2311 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2313 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2315 /* cases below refer to rfc 3530 section 14.2.34: */
2316 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2317 if (conf && !unconf) /* case 2: probable retransmit */
2319 else /* case 4: client hasn't noticed we rebooted yet? */
2320 status = nfserr_stale_clientid;
2324 if (conf) { /* case 1: callback update */
2325 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2326 nfsd4_probe_callback(conf);
2327 expire_client(unconf);
2328 } else { /* case 3: normal case; new or rebooted client */
2329 conf = find_confirmed_client_by_name(&unconf->cl_name);
2331 expire_client(conf);
2332 move_to_confirmed(unconf);
2333 nfsd4_probe_callback(unconf);
2336 nfs4_unlock_state();
2340 static struct nfs4_file *nfsd4_alloc_file(void)
2342 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2345 /* OPEN Share state helper functions */
2346 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2348 unsigned int hashval = file_hashval(ino);
2350 atomic_set(&fp->fi_ref, 1);
2351 INIT_LIST_HEAD(&fp->fi_hash);
2352 INIT_LIST_HEAD(&fp->fi_stateids);
2353 INIT_LIST_HEAD(&fp->fi_delegations);
2354 fp->fi_inode = igrab(ino);
2355 fp->fi_had_conflict = false;
2356 fp->fi_lease = NULL;
2357 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2358 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2359 spin_lock(&recall_lock);
2360 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2361 spin_unlock(&recall_lock);
2365 nfsd4_free_slab(struct kmem_cache **slab)
2369 kmem_cache_destroy(*slab);
2374 nfsd4_free_slabs(void)
2376 nfsd4_free_slab(&openowner_slab);
2377 nfsd4_free_slab(&lockowner_slab);
2378 nfsd4_free_slab(&file_slab);
2379 nfsd4_free_slab(&stateid_slab);
2380 nfsd4_free_slab(&deleg_slab);
2384 nfsd4_init_slabs(void)
2386 openowner_slab = kmem_cache_create("nfsd4_openowners",
2387 sizeof(struct nfs4_openowner), 0, 0, NULL);
2388 if (openowner_slab == NULL)
2390 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2391 sizeof(struct nfs4_lockowner), 0, 0, NULL);
2392 if (lockowner_slab == NULL)
2394 file_slab = kmem_cache_create("nfsd4_files",
2395 sizeof(struct nfs4_file), 0, 0, NULL);
2396 if (file_slab == NULL)
2398 stateid_slab = kmem_cache_create("nfsd4_stateids",
2399 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2400 if (stateid_slab == NULL)
2402 deleg_slab = kmem_cache_create("nfsd4_delegations",
2403 sizeof(struct nfs4_delegation), 0, 0, NULL);
2404 if (deleg_slab == NULL)
2409 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2413 void nfs4_free_openowner(struct nfs4_openowner *oo)
2415 kfree(oo->oo_owner.so_owner.data);
2416 kmem_cache_free(openowner_slab, oo);
2419 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2421 kfree(lo->lo_owner.so_owner.data);
2422 kmem_cache_free(lockowner_slab, lo);
2425 static void init_nfs4_replay(struct nfs4_replay *rp)
2427 rp->rp_status = nfserr_serverfault;
2429 rp->rp_buf = rp->rp_ibuf;
2432 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2434 struct nfs4_stateowner *sop;
2436 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2440 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2441 if (!sop->so_owner.data) {
2442 kmem_cache_free(slab, sop);
2445 sop->so_owner.len = owner->len;
2447 INIT_LIST_HEAD(&sop->so_stateids);
2448 sop->so_client = clp;
2449 init_nfs4_replay(&sop->so_replay);
2453 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2455 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2456 list_add(&oo->oo_perclient, &clp->cl_openowners);
2459 static struct nfs4_openowner *
2460 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2461 struct nfs4_openowner *oo;
2463 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2466 oo->oo_owner.so_is_open_owner = 1;
2467 oo->oo_owner.so_seqid = open->op_seqid;
2468 oo->oo_flags = NFS4_OO_NEW;
2470 oo->oo_last_closed_stid = NULL;
2471 INIT_LIST_HEAD(&oo->oo_close_lru);
2472 hash_openowner(oo, clp, strhashval);
2476 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2477 struct nfs4_openowner *oo = open->op_openowner;
2478 struct nfs4_client *clp = oo->oo_owner.so_client;
2480 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2481 INIT_LIST_HEAD(&stp->st_lockowners);
2482 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2483 list_add(&stp->st_perfile, &fp->fi_stateids);
2484 stp->st_stateowner = &oo->oo_owner;
2487 stp->st_access_bmap = 0;
2488 stp->st_deny_bmap = 0;
2489 set_access(open->op_share_access, stp);
2490 set_deny(open->op_share_deny, stp);
2491 stp->st_openstp = NULL;
2495 move_to_close_lru(struct nfs4_openowner *oo)
2497 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2499 list_move_tail(&oo->oo_close_lru, &close_lru);
2500 oo->oo_time = get_seconds();
2504 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2507 return (sop->so_owner.len == owner->len) &&
2508 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2509 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2512 static struct nfs4_openowner *
2513 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open, bool sessions)
2515 struct nfs4_stateowner *so;
2516 struct nfs4_openowner *oo;
2517 struct nfs4_client *clp;
2519 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2520 if (!so->so_is_open_owner)
2522 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2524 clp = oo->oo_owner.so_client;
2525 if ((bool)clp->cl_minorversion != sessions)
2527 renew_client(oo->oo_owner.so_client);
2534 /* search file_hashtbl[] for file */
2535 static struct nfs4_file *
2536 find_file(struct inode *ino)
2538 unsigned int hashval = file_hashval(ino);
2539 struct nfs4_file *fp;
2541 spin_lock(&recall_lock);
2542 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2543 if (fp->fi_inode == ino) {
2545 spin_unlock(&recall_lock);
2549 spin_unlock(&recall_lock);
2554 * Called to check deny when READ with all zero stateid or
2555 * WRITE with all zero or all one stateid
2558 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2560 struct inode *ino = current_fh->fh_dentry->d_inode;
2561 struct nfs4_file *fp;
2562 struct nfs4_ol_stateid *stp;
2565 dprintk("NFSD: nfs4_share_conflict\n");
2567 fp = find_file(ino);
2570 ret = nfserr_locked;
2571 /* Search for conflicting share reservations */
2572 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2573 if (test_deny(deny_type, stp) ||
2574 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2583 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2585 /* We're assuming the state code never drops its reference
2586 * without first removing the lease. Since we're in this lease
2587 * callback (and since the lease code is serialized by the kernel
2588 * lock) we know the server hasn't removed the lease yet, we know
2589 * it's safe to take a reference: */
2590 atomic_inc(&dp->dl_count);
2592 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2594 /* only place dl_time is set. protected by lock_flocks*/
2595 dp->dl_time = get_seconds();
2597 nfsd4_cb_recall(dp);
2600 /* Called from break_lease() with lock_flocks() held. */
2601 static void nfsd_break_deleg_cb(struct file_lock *fl)
2603 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2604 struct nfs4_delegation *dp;
2607 WARN(1, "(%p)->fl_owner NULL\n", fl);
2610 if (fp->fi_had_conflict) {
2611 WARN(1, "duplicate break on %p\n", fp);
2615 * We don't want the locks code to timeout the lease for us;
2616 * we'll remove it ourself if a delegation isn't returned
2619 fl->fl_break_time = 0;
2621 spin_lock(&recall_lock);
2622 fp->fi_had_conflict = true;
2623 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2624 nfsd_break_one_deleg(dp);
2625 spin_unlock(&recall_lock);
2629 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2632 return lease_modify(onlist, arg);
2637 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2638 .lm_break = nfsd_break_deleg_cb,
2639 .lm_change = nfsd_change_deleg_cb,
2642 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2644 if (nfsd4_has_session(cstate))
2646 if (seqid == so->so_seqid - 1)
2647 return nfserr_replay_me;
2648 if (seqid == so->so_seqid)
2650 return nfserr_bad_seqid;
2654 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2655 struct nfsd4_open *open)
2657 clientid_t *clientid = &open->op_clientid;
2658 struct nfs4_client *clp = NULL;
2659 unsigned int strhashval;
2660 struct nfs4_openowner *oo = NULL;
2662 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
2664 if (STALE_CLIENTID(&open->op_clientid, nn))
2665 return nfserr_stale_clientid;
2667 * In case we need it later, after we've already created the
2668 * file and don't want to risk a further failure:
2670 open->op_file = nfsd4_alloc_file();
2671 if (open->op_file == NULL)
2672 return nfserr_jukebox;
2674 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2675 oo = find_openstateowner_str(strhashval, open, cstate->minorversion);
2676 open->op_openowner = oo;
2678 clp = find_confirmed_client(clientid, cstate->minorversion);
2680 return nfserr_expired;
2683 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2684 /* Replace unconfirmed owners without checking for replay. */
2685 clp = oo->oo_owner.so_client;
2686 release_openowner(oo);
2687 open->op_openowner = NULL;
2690 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2693 clp = oo->oo_owner.so_client;
2696 oo = alloc_init_open_stateowner(strhashval, clp, open);
2698 return nfserr_jukebox;
2699 open->op_openowner = oo;
2701 open->op_stp = nfs4_alloc_stateid(clp);
2703 return nfserr_jukebox;
2707 static inline __be32
2708 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2710 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2711 return nfserr_openmode;
2716 static int share_access_to_flags(u32 share_access)
2718 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2721 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2723 struct nfs4_stid *ret;
2725 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2728 return delegstateid(ret);
2731 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2733 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2734 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2738 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2739 struct nfs4_delegation **dp)
2742 __be32 status = nfserr_bad_stateid;
2744 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2747 flags = share_access_to_flags(open->op_share_access);
2748 status = nfs4_check_delegmode(*dp, flags);
2752 if (!nfsd4_is_deleg_cur(open))
2756 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2761 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2763 struct nfs4_ol_stateid *local;
2764 struct nfs4_openowner *oo = open->op_openowner;
2766 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2767 /* ignore lock owners */
2768 if (local->st_stateowner->so_is_open_owner == 0)
2770 /* remember if we have seen this open owner */
2771 if (local->st_stateowner == &oo->oo_owner)
2773 /* check for conflicting share reservations */
2774 if (!test_share(local, open))
2775 return nfserr_share_denied;
2780 static inline int nfs4_access_to_access(u32 nfs4_access)
2784 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2785 flags |= NFSD_MAY_READ;
2786 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2787 flags |= NFSD_MAY_WRITE;
2791 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2792 struct svc_fh *cur_fh, struct nfsd4_open *open)
2795 int oflag = nfs4_access_to_omode(open->op_share_access);
2796 int access = nfs4_access_to_access(open->op_share_access);
2798 if (!fp->fi_fds[oflag]) {
2799 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2800 &fp->fi_fds[oflag]);
2804 nfs4_file_get_access(fp, oflag);
2809 static inline __be32
2810 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2811 struct nfsd4_open *open)
2813 struct iattr iattr = {
2814 .ia_valid = ATTR_SIZE,
2817 if (!open->op_truncate)
2819 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2820 return nfserr_inval;
2821 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2825 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)
2827 u32 op_share_access = open->op_share_access;
2831 new_access = !test_access(op_share_access, stp);
2833 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2837 status = nfsd4_truncate(rqstp, cur_fh, open);
2840 int oflag = nfs4_access_to_omode(op_share_access);
2841 nfs4_file_put_access(fp, oflag);
2845 /* remember the open */
2846 set_access(op_share_access, stp);
2847 set_deny(open->op_share_deny, stp);
2854 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2856 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2859 /* Should we give out recallable state?: */
2860 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2862 if (clp->cl_cb_state == NFSD4_CB_UP)
2865 * In the sessions case, since we don't have to establish a
2866 * separate connection for callbacks, we assume it's OK
2867 * until we hear otherwise:
2869 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2872 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2874 struct file_lock *fl;
2876 fl = locks_alloc_lock();
2879 locks_init_lock(fl);
2880 fl->fl_lmops = &nfsd_lease_mng_ops;
2881 fl->fl_flags = FL_LEASE;
2882 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2883 fl->fl_end = OFFSET_MAX;
2884 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2885 fl->fl_pid = current->tgid;
2889 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2891 struct nfs4_file *fp = dp->dl_file;
2892 struct file_lock *fl;
2895 fl = nfs4_alloc_init_lease(dp, flag);
2898 fl->fl_file = find_readable_file(fp);
2899 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2900 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2902 list_del_init(&dp->dl_perclnt);
2903 locks_free_lock(fl);
2907 fp->fi_deleg_file = get_file(fl->fl_file);
2908 atomic_set(&fp->fi_delegees, 1);
2909 list_add(&dp->dl_perfile, &fp->fi_delegations);
2913 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2915 struct nfs4_file *fp = dp->dl_file;
2918 return nfs4_setlease(dp, flag);
2919 spin_lock(&recall_lock);
2920 if (fp->fi_had_conflict) {
2921 spin_unlock(&recall_lock);
2924 atomic_inc(&fp->fi_delegees);
2925 list_add(&dp->dl_perfile, &fp->fi_delegations);
2926 spin_unlock(&recall_lock);
2927 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2931 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2933 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2934 if (status == -EAGAIN)
2935 open->op_why_no_deleg = WND4_CONTENTION;
2937 open->op_why_no_deleg = WND4_RESOURCE;
2938 switch (open->op_deleg_want) {
2939 case NFS4_SHARE_WANT_READ_DELEG:
2940 case NFS4_SHARE_WANT_WRITE_DELEG:
2941 case NFS4_SHARE_WANT_ANY_DELEG:
2943 case NFS4_SHARE_WANT_CANCEL:
2944 open->op_why_no_deleg = WND4_CANCELLED;
2946 case NFS4_SHARE_WANT_NO_DELEG:
2947 BUG(); /* not supposed to get here */
2953 * Attempt to hand out a delegation.
2956 nfs4_open_delegation(struct net *net, struct svc_fh *fh,
2957 struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2959 struct nfs4_delegation *dp;
2960 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2962 int status = 0, flag = 0;
2964 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2965 flag = NFS4_OPEN_DELEGATE_NONE;
2966 open->op_recall = 0;
2967 switch (open->op_claim_type) {
2968 case NFS4_OPEN_CLAIM_PREVIOUS:
2970 open->op_recall = 1;
2971 flag = open->op_delegate_type;
2972 if (flag == NFS4_OPEN_DELEGATE_NONE)
2975 case NFS4_OPEN_CLAIM_NULL:
2976 /* Let's not give out any delegations till everyone's
2977 * had the chance to reclaim theirs.... */
2978 if (locks_in_grace(net))
2980 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2982 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2983 flag = NFS4_OPEN_DELEGATE_WRITE;
2985 flag = NFS4_OPEN_DELEGATE_READ;
2991 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2994 status = nfs4_set_delegation(dp, flag);
2998 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
3000 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
3001 STATEID_VAL(&dp->dl_stid.sc_stateid));
3003 open->op_delegate_type = flag;
3004 if (flag == NFS4_OPEN_DELEGATE_NONE) {
3005 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3006 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
3007 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3009 /* 4.1 client asking for a delegation? */
3010 if (open->op_deleg_want)
3011 nfsd4_open_deleg_none_ext(open, status);
3015 nfs4_put_delegation(dp);
3017 flag = NFS4_OPEN_DELEGATE_NONE;
3021 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3022 struct nfs4_delegation *dp)
3024 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3025 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3026 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3027 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3028 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3029 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3030 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3031 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3033 /* Otherwise the client must be confused wanting a delegation
3034 * it already has, therefore we don't return
3035 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3040 * called with nfs4_lock_state() held.
3043 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3045 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3046 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3047 struct nfs4_file *fp = NULL;
3048 struct inode *ino = current_fh->fh_dentry->d_inode;
3049 struct nfs4_ol_stateid *stp = NULL;
3050 struct nfs4_delegation *dp = NULL;
3054 * Lookup file; if found, lookup stateid and check open request,
3055 * and check for delegations in the process of being recalled.
3056 * If not found, create the nfs4_file struct
3058 fp = find_file(ino);
3060 if ((status = nfs4_check_open(fp, open, &stp)))
3062 status = nfs4_check_deleg(cl, fp, open, &dp);
3066 status = nfserr_bad_stateid;
3067 if (nfsd4_is_deleg_cur(open))
3069 status = nfserr_jukebox;
3071 open->op_file = NULL;
3072 nfsd4_init_file(fp, ino);
3076 * OPEN the file, or upgrade an existing OPEN.
3077 * If truncate fails, the OPEN fails.
3080 /* Stateid was found, this is an OPEN upgrade */
3081 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3085 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3088 status = nfsd4_truncate(rqstp, current_fh, open);
3092 open->op_stp = NULL;
3093 init_open_stateid(stp, fp, open);
3095 update_stateid(&stp->st_stid.sc_stateid);
3096 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3098 if (nfsd4_has_session(&resp->cstate)) {
3099 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3101 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3102 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3103 open->op_why_no_deleg = WND4_NOT_WANTED;
3109 * Attempt to hand out a delegation. No error return, because the
3110 * OPEN succeeds even if we fail.
3112 nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
3116 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3117 STATEID_VAL(&stp->st_stid.sc_stateid));
3119 /* 4.1 client trying to upgrade/downgrade delegation? */
3120 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3121 open->op_deleg_want)
3122 nfsd4_deleg_xgrade_none_ext(open, dp);
3126 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3127 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3129 * To finish the open response, we just need to set the rflags.
3131 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3132 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3133 !nfsd4_has_session(&resp->cstate))
3134 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3139 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3141 if (open->op_openowner) {
3142 struct nfs4_openowner *oo = open->op_openowner;
3144 if (!list_empty(&oo->oo_owner.so_stateids))
3145 list_del_init(&oo->oo_close_lru);
3146 if (oo->oo_flags & NFS4_OO_NEW) {
3148 release_openowner(oo);
3149 open->op_openowner = NULL;
3151 oo->oo_flags &= ~NFS4_OO_NEW;
3155 nfsd4_free_file(open->op_file);
3157 free_generic_stateid(open->op_stp);
3161 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3164 struct nfs4_client *clp;
3166 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3169 dprintk("process_renew(%08x/%08x): starting\n",
3170 clid->cl_boot, clid->cl_id);
3171 status = nfserr_stale_clientid;
3172 if (STALE_CLIENTID(clid, nn))
3174 clp = find_confirmed_client(clid, cstate->minorversion);
3175 status = nfserr_expired;
3177 /* We assume the client took too long to RENEW. */
3178 dprintk("nfsd4_renew: clientid not found!\n");
3181 status = nfserr_cb_path_down;
3182 if (!list_empty(&clp->cl_delegations)
3183 && clp->cl_cb_state != NFSD4_CB_UP)
3187 nfs4_unlock_state();
3192 nfsd4_end_grace(struct net *net)
3194 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3196 /* do nothing if grace period already ended */
3197 if (nn->grace_ended)
3200 dprintk("NFSD: end of grace period\n");
3201 nn->grace_ended = true;
3202 nfsd4_record_grace_done(net, nn->boot_time);
3203 locks_end_grace(&nn->nfsd4_manager);
3205 * Now that every NFSv4 client has had the chance to recover and
3206 * to see the (possibly new, possibly shorter) lease time, we
3207 * can safely set the next grace time to the current lease time:
3209 nfsd4_grace = nfsd4_lease;
3213 nfs4_laundromat(void)
3215 struct nfs4_client *clp;
3216 struct nfs4_openowner *oo;
3217 struct nfs4_delegation *dp;
3218 struct list_head *pos, *next, reaplist;
3219 time_t cutoff = get_seconds() - nfsd4_lease;
3220 time_t t, clientid_val = nfsd4_lease;
3221 time_t u, test_val = nfsd4_lease;
3225 dprintk("NFSD: laundromat service - starting\n");
3226 nfsd4_end_grace(&init_net);
3227 INIT_LIST_HEAD(&reaplist);
3228 spin_lock(&client_lock);
3229 list_for_each_safe(pos, next, &client_lru) {
3230 clp = list_entry(pos, struct nfs4_client, cl_lru);
3231 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3232 t = clp->cl_time - cutoff;
3233 if (clientid_val > t)
3237 if (atomic_read(&clp->cl_refcount)) {
3238 dprintk("NFSD: client in use (clientid %08x)\n",
3239 clp->cl_clientid.cl_id);
3242 unhash_client_locked(clp);
3243 list_add(&clp->cl_lru, &reaplist);
3245 spin_unlock(&client_lock);
3246 list_for_each_safe(pos, next, &reaplist) {
3247 clp = list_entry(pos, struct nfs4_client, cl_lru);
3248 dprintk("NFSD: purging unused client (clientid %08x)\n",
3249 clp->cl_clientid.cl_id);
3252 spin_lock(&recall_lock);
3253 list_for_each_safe(pos, next, &del_recall_lru) {
3254 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3255 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3256 u = dp->dl_time - cutoff;
3261 list_move(&dp->dl_recall_lru, &reaplist);
3263 spin_unlock(&recall_lock);
3264 list_for_each_safe(pos, next, &reaplist) {
3265 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3266 unhash_delegation(dp);
3268 test_val = nfsd4_lease;
3269 list_for_each_safe(pos, next, &close_lru) {
3270 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3271 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3272 u = oo->oo_time - cutoff;
3277 release_openowner(oo);
3279 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3280 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3281 nfs4_unlock_state();
3282 return clientid_val;
3285 static struct workqueue_struct *laundry_wq;
3286 static void laundromat_main(struct work_struct *);
3287 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3290 laundromat_main(struct work_struct *not_used)
3294 t = nfs4_laundromat();
3295 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3296 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3299 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3301 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3302 return nfserr_bad_stateid;
3307 STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
3309 if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
3311 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3312 STATEID_VAL(stateid));
3317 access_permit_read(struct nfs4_ol_stateid *stp)
3319 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3320 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3321 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3325 access_permit_write(struct nfs4_ol_stateid *stp)
3327 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3328 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3332 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3334 __be32 status = nfserr_openmode;
3336 /* For lock stateid's, we test the parent open, not the lock: */
3337 if (stp->st_openstp)
3338 stp = stp->st_openstp;
3339 if ((flags & WR_STATE) && !access_permit_write(stp))
3341 if ((flags & RD_STATE) && !access_permit_read(stp))
3348 static inline __be32
3349 check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
3351 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3353 else if (locks_in_grace(net)) {
3354 /* Answer in remaining cases depends on existence of
3355 * conflicting state; so we must wait out the grace period. */
3356 return nfserr_grace;
3357 } else if (flags & WR_STATE)
3358 return nfs4_share_conflict(current_fh,
3359 NFS4_SHARE_DENY_WRITE);
3360 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3361 return nfs4_share_conflict(current_fh,
3362 NFS4_SHARE_DENY_READ);
3366 * Allow READ/WRITE during grace period on recovered state only for files
3367 * that are not able to provide mandatory locking.
3370 grace_disallows_io(struct net *net, struct inode *inode)
3372 return locks_in_grace(net) && mandatory_lock(inode);
3375 /* Returns true iff a is later than b: */
3376 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3378 return (s32)a->si_generation - (s32)b->si_generation > 0;
3381 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3384 * When sessions are used the stateid generation number is ignored
3387 if (has_session && in->si_generation == 0)
3390 if (in->si_generation == ref->si_generation)
3393 /* If the client sends us a stateid from the future, it's buggy: */
3394 if (stateid_generation_after(in, ref))
3395 return nfserr_bad_stateid;
3397 * However, we could see a stateid from the past, even from a
3398 * non-buggy client. For example, if the client sends a lock
3399 * while some IO is outstanding, the lock may bump si_generation
3400 * while the IO is still in flight. The client could avoid that
3401 * situation by waiting for responses on all the IO requests,
3402 * but better performance may result in retrying IO that
3403 * receives an old_stateid error if requests are rarely
3404 * reordered in flight:
3406 return nfserr_old_stateid;
3409 static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3411 struct nfs4_stid *s;
3412 struct nfs4_ol_stateid *ols;
3415 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3416 return nfserr_bad_stateid;
3417 /* Client debugging aid. */
3418 if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
3419 char addr_str[INET6_ADDRSTRLEN];
3420 rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
3422 pr_warn_ratelimited("NFSD: client %s testing state ID "
3423 "with incorrect client ID\n", addr_str);
3424 return nfserr_bad_stateid;
3426 s = find_stateid(cl, stateid);
3428 return nfserr_bad_stateid;
3429 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3432 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3434 ols = openlockstateid(s);
3435 if (ols->st_stateowner->so_is_open_owner
3436 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3437 return nfserr_bad_stateid;
3441 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s, bool sessions)
3443 struct nfs4_client *cl;
3444 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
3446 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3447 return nfserr_bad_stateid;
3448 if (STALE_STATEID(stateid, nn))
3449 return nfserr_stale_stateid;
3450 cl = find_confirmed_client(&stateid->si_opaque.so_clid, sessions);
3452 return nfserr_expired;
3453 *s = find_stateid_by_type(cl, stateid, typemask);
3455 return nfserr_bad_stateid;
3461 * Checks for stateid operations
3464 nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
3465 stateid_t *stateid, int flags, struct file **filpp)
3467 struct nfs4_stid *s;
3468 struct nfs4_ol_stateid *stp = NULL;
3469 struct nfs4_delegation *dp = NULL;
3470 struct svc_fh *current_fh = &cstate->current_fh;
3471 struct inode *ino = current_fh->fh_dentry->d_inode;
3477 if (grace_disallows_io(net, ino))
3478 return nfserr_grace;
3480 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3481 return check_special_stateids(net, current_fh, stateid, flags);
3483 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s, cstate->minorversion);
3486 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3489 switch (s->sc_type) {
3490 case NFS4_DELEG_STID:
3491 dp = delegstateid(s);
3492 status = nfs4_check_delegmode(dp, flags);
3496 *filpp = dp->dl_file->fi_deleg_file;
3500 case NFS4_OPEN_STID:
3501 case NFS4_LOCK_STID:
3502 stp = openlockstateid(s);
3503 status = nfs4_check_fh(current_fh, stp);
3506 if (stp->st_stateowner->so_is_open_owner
3507 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3509 status = nfs4_check_openmode(stp, flags);
3513 if (flags & RD_STATE)
3514 *filpp = find_readable_file(stp->st_file);
3516 *filpp = find_writeable_file(stp->st_file);
3520 return nfserr_bad_stateid;
3528 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3530 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3531 return nfserr_locks_held;
3532 release_lock_stateid(stp);
3537 * Test if the stateid is valid
3540 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3541 struct nfsd4_test_stateid *test_stateid)
3543 struct nfsd4_test_stateid_id *stateid;
3544 struct nfs4_client *cl = cstate->session->se_client;
3547 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3548 stateid->ts_id_status =
3549 nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
3550 nfs4_unlock_state();
3556 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3557 struct nfsd4_free_stateid *free_stateid)
3559 stateid_t *stateid = &free_stateid->fr_stateid;
3560 struct nfs4_stid *s;
3561 struct nfs4_client *cl = cstate->session->se_client;
3562 __be32 ret = nfserr_bad_stateid;
3565 s = find_stateid(cl, stateid);
3568 switch (s->sc_type) {
3569 case NFS4_DELEG_STID:
3570 ret = nfserr_locks_held;
3572 case NFS4_OPEN_STID:
3573 case NFS4_LOCK_STID:
3574 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3577 if (s->sc_type == NFS4_LOCK_STID)
3578 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3580 ret = nfserr_locks_held;
3583 ret = nfserr_bad_stateid;
3586 nfs4_unlock_state();
3593 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3594 RD_STATE : WR_STATE;
3597 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3599 struct svc_fh *current_fh = &cstate->current_fh;
3600 struct nfs4_stateowner *sop = stp->st_stateowner;
3603 status = nfsd4_check_seqid(cstate, sop, seqid);
3606 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3608 * "Closed" stateid's exist *only* to return
3609 * nfserr_replay_me from the previous step.
3611 return nfserr_bad_stateid;
3612 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3615 return nfs4_check_fh(current_fh, stp);
3619 * Checks for sequence id mutating operations.
3622 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3623 stateid_t *stateid, char typemask,
3624 struct nfs4_ol_stateid **stpp)
3627 struct nfs4_stid *s;
3629 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3630 seqid, STATEID_VAL(stateid));
3633 status = nfsd4_lookup_stateid(stateid, typemask, &s, cstate->minorversion);
3636 *stpp = openlockstateid(s);
3637 cstate->replay_owner = (*stpp)->st_stateowner;
3639 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3642 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3645 struct nfs4_openowner *oo;
3647 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3648 NFS4_OPEN_STID, stpp);
3651 oo = openowner((*stpp)->st_stateowner);
3652 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3653 return nfserr_bad_stateid;
3658 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3659 struct nfsd4_open_confirm *oc)
3662 struct nfs4_openowner *oo;
3663 struct nfs4_ol_stateid *stp;
3665 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3666 (int)cstate->current_fh.fh_dentry->d_name.len,
3667 cstate->current_fh.fh_dentry->d_name.name);
3669 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3675 status = nfs4_preprocess_seqid_op(cstate,
3676 oc->oc_seqid, &oc->oc_req_stateid,
3677 NFS4_OPEN_STID, &stp);
3680 oo = openowner(stp->st_stateowner);
3681 status = nfserr_bad_stateid;
3682 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3684 oo->oo_flags |= NFS4_OO_CONFIRMED;
3685 update_stateid(&stp->st_stid.sc_stateid);
3686 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3687 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3688 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3690 nfsd4_client_record_create(oo->oo_owner.so_client);
3693 if (!cstate->replay_owner)
3694 nfs4_unlock_state();
3698 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3700 if (!test_access(access, stp))
3702 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3703 clear_access(access, stp);
3706 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3708 switch (to_access) {
3709 case NFS4_SHARE_ACCESS_READ:
3710 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3711 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3713 case NFS4_SHARE_ACCESS_WRITE:
3714 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3715 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3717 case NFS4_SHARE_ACCESS_BOTH:
3725 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3728 for (i = 0; i < 4; i++) {
3729 if ((i & deny) != i)
3735 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3736 struct nfsd4_compound_state *cstate,
3737 struct nfsd4_open_downgrade *od)
3740 struct nfs4_ol_stateid *stp;
3742 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3743 (int)cstate->current_fh.fh_dentry->d_name.len,
3744 cstate->current_fh.fh_dentry->d_name.name);
3746 /* We don't yet support WANT bits: */
3747 if (od->od_deleg_want)
3748 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3752 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3753 &od->od_stateid, &stp);
3756 status = nfserr_inval;
3757 if (!test_access(od->od_share_access, stp)) {
3758 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3759 stp->st_access_bmap, od->od_share_access);
3762 if (!test_deny(od->od_share_deny, stp)) {
3763 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3764 stp->st_deny_bmap, od->od_share_deny);
3767 nfs4_stateid_downgrade(stp, od->od_share_access);
3769 reset_union_bmap_deny(od->od_share_deny, stp);
3771 update_stateid(&stp->st_stid.sc_stateid);
3772 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3775 if (!cstate->replay_owner)
3776 nfs4_unlock_state();
3780 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3782 struct nfs4_openowner *oo;
3783 struct nfs4_ol_stateid *s;
3785 if (!so->so_is_open_owner)
3788 s = oo->oo_last_closed_stid;
3791 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3792 /* Release the last_closed_stid on the next seqid bump: */
3793 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3796 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3797 release_last_closed_stateid(oo);
3800 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3802 unhash_open_stateid(s);
3803 s->st_stid.sc_type = NFS4_CLOSED_STID;
3807 * nfs4_unlock_state() called after encode
3810 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3811 struct nfsd4_close *close)
3814 struct nfs4_openowner *oo;
3815 struct nfs4_ol_stateid *stp;
3817 dprintk("NFSD: nfsd4_close on file %.*s\n",
3818 (int)cstate->current_fh.fh_dentry->d_name.len,
3819 cstate->current_fh.fh_dentry->d_name.name);
3822 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3824 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3828 oo = openowner(stp->st_stateowner);
3830 update_stateid(&stp->st_stid.sc_stateid);
3831 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3833 nfsd4_close_open_stateid(stp);
3834 release_last_closed_stateid(oo);
3835 oo->oo_last_closed_stid = stp;
3837 if (list_empty(&oo->oo_owner.so_stateids)) {
3838 if (cstate->minorversion) {
3839 release_openowner(oo);
3840 cstate->replay_owner = NULL;
3843 * In the 4.0 case we need to keep the owners around a
3844 * little while to handle CLOSE replay.
3846 if (list_empty(&oo->oo_owner.so_stateids))
3847 move_to_close_lru(oo);
3851 if (!cstate->replay_owner)
3852 nfs4_unlock_state();
3857 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3858 struct nfsd4_delegreturn *dr)
3860 struct nfs4_delegation *dp;
3861 stateid_t *stateid = &dr->dr_stateid;
3862 struct nfs4_stid *s;
3865 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3869 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s, cstate->minorversion);
3872 dp = delegstateid(s);
3873 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3877 unhash_delegation(dp);
3879 nfs4_unlock_state();
3885 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3887 #define LOCKOWNER_INO_HASH_BITS 8
3888 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3889 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3892 end_offset(u64 start, u64 len)
3897 return end >= start ? end: NFS4_MAX_UINT64;
3900 /* last octet in a range */
3902 last_byte_offset(u64 start, u64 len)
3908 return end > start ? end - 1: NFS4_MAX_UINT64;
3911 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3913 return (file_hashval(inode) + cl_id
3914 + opaque_hashval(ownername->data, ownername->len))
3915 & LOCKOWNER_INO_HASH_MASK;
3918 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3921 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3922 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3923 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3924 * locking, this prevents us from being completely protocol-compliant. The
3925 * real solution to this problem is to start using unsigned file offsets in
3926 * the VFS, but this is a very deep change!
3929 nfs4_transform_lock_offset(struct file_lock *lock)
3931 if (lock->fl_start < 0)
3932 lock->fl_start = OFFSET_MAX;
3933 if (lock->fl_end < 0)
3934 lock->fl_end = OFFSET_MAX;
3937 /* Hack!: For now, we're defining this just so we can use a pointer to it
3938 * as a unique cookie to identify our (NFSv4's) posix locks. */
3939 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3943 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3945 struct nfs4_lockowner *lo;
3947 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3948 lo = (struct nfs4_lockowner *) fl->fl_owner;
3949 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3950 lo->lo_owner.so_owner.len, GFP_KERNEL);
3951 if (!deny->ld_owner.data)
3952 /* We just don't care that much */
3954 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3955 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3958 deny->ld_owner.len = 0;
3959 deny->ld_owner.data = NULL;
3960 deny->ld_clientid.cl_boot = 0;
3961 deny->ld_clientid.cl_id = 0;
3963 deny->ld_start = fl->fl_start;
3964 deny->ld_length = NFS4_MAX_UINT64;
3965 if (fl->fl_end != NFS4_MAX_UINT64)
3966 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3967 deny->ld_type = NFS4_READ_LT;
3968 if (fl->fl_type != F_RDLCK)
3969 deny->ld_type = NFS4_WRITE_LT;
3972 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3974 struct nfs4_ol_stateid *lst;
3976 if (!same_owner_str(&lo->lo_owner, owner, clid))
3978 lst = list_first_entry(&lo->lo_owner.so_stateids,
3979 struct nfs4_ol_stateid, st_perstateowner);
3980 return lst->st_file->fi_inode == inode;
3983 static struct nfs4_lockowner *
3984 find_lockowner_str(struct inode *inode, clientid_t *clid,
3985 struct xdr_netobj *owner)
3987 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3988 struct nfs4_lockowner *lo;
3990 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3991 if (same_lockowner_ino(lo, inode, clid, owner))
3997 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3999 struct inode *inode = open_stp->st_file->fi_inode;
4000 unsigned int inohash = lockowner_ino_hashval(inode,
4001 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
4003 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
4004 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
4005 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
4009 * Alloc a lock owner structure.
4010 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
4013 * strhashval = ownerstr_hashval
4016 static struct nfs4_lockowner *
4017 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4018 struct nfs4_lockowner *lo;
4020 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4023 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4024 lo->lo_owner.so_is_open_owner = 0;
4025 /* It is the openowner seqid that will be incremented in encode in the
4026 * case of new lockowners; so increment the lock seqid manually: */
4027 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4028 hash_lockowner(lo, strhashval, clp, open_stp);
4032 static struct nfs4_ol_stateid *
4033 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4035 struct nfs4_ol_stateid *stp;
4036 struct nfs4_client *clp = lo->lo_owner.so_client;
4038 stp = nfs4_alloc_stateid(clp);
4041 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
4042 list_add(&stp->st_perfile, &fp->fi_stateids);
4043 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4044 stp->st_stateowner = &lo->lo_owner;
4047 stp->st_access_bmap = 0;
4048 stp->st_deny_bmap = open_stp->st_deny_bmap;
4049 stp->st_openstp = open_stp;
4054 check_lock_length(u64 offset, u64 length)
4056 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4057 LOFF_OVERFLOW(offset, length)));
4060 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4062 struct nfs4_file *fp = lock_stp->st_file;
4063 int oflag = nfs4_access_to_omode(access);
4065 if (test_access(access, lock_stp))
4067 nfs4_file_get_access(fp, oflag);
4068 set_access(access, lock_stp);
4071 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)
4073 struct nfs4_file *fi = ost->st_file;
4074 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4075 struct nfs4_client *cl = oo->oo_owner.so_client;
4076 struct nfs4_lockowner *lo;
4077 unsigned int strhashval;
4079 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4081 if (!cstate->minorversion)
4082 return nfserr_bad_seqid;
4083 /* XXX: a lockowner always has exactly one stateid: */
4084 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4085 struct nfs4_ol_stateid, st_perstateowner);
4088 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4089 &lock->v.new.owner);
4090 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4092 return nfserr_jukebox;
4093 *lst = alloc_init_lock_stateid(lo, fi, ost);
4095 release_lockowner(lo);
4096 return nfserr_jukebox;
4106 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4107 struct nfsd4_lock *lock)
4109 struct nfs4_openowner *open_sop = NULL;
4110 struct nfs4_lockowner *lock_sop = NULL;
4111 struct nfs4_ol_stateid *lock_stp;
4112 struct file *filp = NULL;
4113 struct file_lock *file_lock = NULL;
4114 struct file_lock *conflock = NULL;
4116 bool new_state = false;
4119 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4121 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4122 (long long) lock->lk_offset,
4123 (long long) lock->lk_length);
4125 if (check_lock_length(lock->lk_offset, lock->lk_length))
4126 return nfserr_inval;
4128 if ((status = fh_verify(rqstp, &cstate->current_fh,
4129 S_IFREG, NFSD_MAY_LOCK))) {
4130 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4136 if (lock->lk_is_new) {
4137 struct nfs4_ol_stateid *open_stp = NULL;
4139 if (nfsd4_has_session(cstate))
4140 /* See rfc 5661 18.10.3: given clientid is ignored: */
4141 memcpy(&lock->v.new.clientid,
4142 &cstate->session->se_client->cl_clientid,
4143 sizeof(clientid_t));
4145 status = nfserr_stale_clientid;
4146 if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
4149 /* validate and update open stateid and open seqid */
4150 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4151 lock->lk_new_open_seqid,
4152 &lock->lk_new_open_stateid,
4156 open_sop = openowner(open_stp->st_stateowner);
4157 status = nfserr_bad_stateid;
4158 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4159 &lock->v.new.clientid))
4161 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4162 &lock_stp, &new_state);
4164 status = nfs4_preprocess_seqid_op(cstate,
4165 lock->lk_old_lock_seqid,
4166 &lock->lk_old_lock_stateid,
4167 NFS4_LOCK_STID, &lock_stp);
4170 lock_sop = lockowner(lock_stp->st_stateowner);
4172 lkflg = setlkflg(lock->lk_type);
4173 status = nfs4_check_openmode(lock_stp, lkflg);
4177 status = nfserr_grace;
4178 if (locks_in_grace(SVC_NET(rqstp)) && !lock->lk_reclaim)
4180 status = nfserr_no_grace;
4181 if (!locks_in_grace(SVC_NET(rqstp)) && lock->lk_reclaim)
4184 file_lock = locks_alloc_lock();
4186 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4187 status = nfserr_jukebox;
4191 locks_init_lock(file_lock);
4192 switch (lock->lk_type) {
4195 filp = find_readable_file(lock_stp->st_file);
4197 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4198 file_lock->fl_type = F_RDLCK;
4201 case NFS4_WRITEW_LT:
4202 filp = find_writeable_file(lock_stp->st_file);
4204 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4205 file_lock->fl_type = F_WRLCK;
4208 status = nfserr_inval;
4212 status = nfserr_openmode;
4215 file_lock->fl_owner = (fl_owner_t)lock_sop;
4216 file_lock->fl_pid = current->tgid;
4217 file_lock->fl_file = filp;
4218 file_lock->fl_flags = FL_POSIX;
4219 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4220 file_lock->fl_start = lock->lk_offset;
4221 file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4222 nfs4_transform_lock_offset(file_lock);
4224 conflock = locks_alloc_lock();
4226 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4227 status = nfserr_jukebox;
4231 err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
4233 case 0: /* success! */
4234 update_stateid(&lock_stp->st_stid.sc_stateid);
4235 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4239 case (EAGAIN): /* conflock holds conflicting lock */
4240 status = nfserr_denied;
4241 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4242 nfs4_set_lock_denied(conflock, &lock->lk_denied);
4245 status = nfserr_deadlock;
4248 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4249 status = nfserrno(err);
4253 if (status && new_state)
4254 release_lockowner(lock_sop);
4255 if (!cstate->replay_owner)
4256 nfs4_unlock_state();
4258 locks_free_lock(file_lock);
4260 locks_free_lock(conflock);
4265 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4266 * so we do a temporary open here just to get an open file to pass to
4267 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4270 static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4273 __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4275 err = nfserrno(vfs_test_lock(file, lock));
4285 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4286 struct nfsd4_lockt *lockt)
4288 struct inode *inode;
4289 struct file_lock *file_lock = NULL;
4290 struct nfs4_lockowner *lo;
4292 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4294 if (locks_in_grace(SVC_NET(rqstp)))
4295 return nfserr_grace;
4297 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4298 return nfserr_inval;
4302 status = nfserr_stale_clientid;
4303 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid, nn))
4306 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4309 inode = cstate->current_fh.fh_dentry->d_inode;
4310 file_lock = locks_alloc_lock();
4312 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4313 status = nfserr_jukebox;
4316 locks_init_lock(file_lock);
4317 switch (lockt->lt_type) {
4320 file_lock->fl_type = F_RDLCK;
4323 case NFS4_WRITEW_LT:
4324 file_lock->fl_type = F_WRLCK;
4327 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4328 status = nfserr_inval;
4332 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4334 file_lock->fl_owner = (fl_owner_t)lo;
4335 file_lock->fl_pid = current->tgid;
4336 file_lock->fl_flags = FL_POSIX;
4338 file_lock->fl_start = lockt->lt_offset;
4339 file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4341 nfs4_transform_lock_offset(file_lock);
4343 status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
4347 if (file_lock->fl_type != F_UNLCK) {
4348 status = nfserr_denied;
4349 nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
4352 nfs4_unlock_state();
4354 locks_free_lock(file_lock);
4359 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4360 struct nfsd4_locku *locku)
4362 struct nfs4_ol_stateid *stp;
4363 struct file *filp = NULL;
4364 struct file_lock *file_lock = NULL;
4368 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4369 (long long) locku->lu_offset,
4370 (long long) locku->lu_length);
4372 if (check_lock_length(locku->lu_offset, locku->lu_length))
4373 return nfserr_inval;
4377 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4378 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4381 filp = find_any_file(stp->st_file);
4383 status = nfserr_lock_range;
4386 file_lock = locks_alloc_lock();
4388 dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
4389 status = nfserr_jukebox;
4392 locks_init_lock(file_lock);
4393 file_lock->fl_type = F_UNLCK;
4394 file_lock->fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4395 file_lock->fl_pid = current->tgid;
4396 file_lock->fl_file = filp;
4397 file_lock->fl_flags = FL_POSIX;
4398 file_lock->fl_lmops = &nfsd_posix_mng_ops;
4399 file_lock->fl_start = locku->lu_offset;
4401 file_lock->fl_end = last_byte_offset(locku->lu_offset,
4403 nfs4_transform_lock_offset(file_lock);
4406 * Try to unlock the file in the VFS.
4408 err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
4410 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4414 * OK, unlock succeeded; the only thing left to do is update the stateid.
4416 update_stateid(&stp->st_stid.sc_stateid);
4417 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4420 if (!cstate->replay_owner)
4421 nfs4_unlock_state();
4423 locks_free_lock(file_lock);
4427 status = nfserrno(err);
4433 * 1: locks held by lockowner
4434 * 0: no locks held by lockowner
4437 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4439 struct file_lock **flpp;
4440 struct inode *inode = filp->fi_inode;
4444 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4445 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4456 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4457 struct nfsd4_compound_state *cstate,
4458 struct nfsd4_release_lockowner *rlockowner)
4460 clientid_t *clid = &rlockowner->rl_clientid;
4461 struct nfs4_stateowner *sop;
4462 struct nfs4_lockowner *lo;
4463 struct nfs4_ol_stateid *stp;
4464 struct xdr_netobj *owner = &rlockowner->rl_owner;
4465 struct list_head matches;
4466 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4468 struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
4470 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4471 clid->cl_boot, clid->cl_id);
4473 /* XXX check for lease expiration */
4475 status = nfserr_stale_clientid;
4476 if (STALE_CLIENTID(clid, nn))
4481 status = nfserr_locks_held;
4482 INIT_LIST_HEAD(&matches);
4484 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4485 if (sop->so_is_open_owner)
4487 if (!same_owner_str(sop, owner, clid))
4489 list_for_each_entry(stp, &sop->so_stateids,
4491 lo = lockowner(sop);
4492 if (check_for_locks(stp->st_file, lo))
4494 list_add(&lo->lo_list, &matches);
4497 /* Clients probably won't expect us to return with some (but not all)
4498 * of the lockowner state released; so don't release any until all
4499 * have been checked. */
4501 while (!list_empty(&matches)) {
4502 lo = list_entry(matches.next, struct nfs4_lockowner,
4504 /* unhash_stateowner deletes so_perclient only
4505 * for openowners. */
4506 list_del(&lo->lo_list);
4507 release_lockowner(lo);
4510 nfs4_unlock_state();
4514 static inline struct nfs4_client_reclaim *
4517 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4521 nfs4_has_reclaimed_state(const char *name)
4523 struct nfs4_client_reclaim *crp;
4525 crp = nfsd4_find_reclaim_client(name);
4526 return (crp && crp->cr_clp);
4530 * failure => all reset bets are off, nfserr_no_grace...
4532 struct nfs4_client_reclaim *
4533 nfs4_client_to_reclaim(const char *name)
4535 unsigned int strhashval;
4536 struct nfs4_client_reclaim *crp;
4538 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4539 crp = alloc_reclaim();
4541 strhashval = clientstr_hashval(name);
4542 INIT_LIST_HEAD(&crp->cr_strhash);
4543 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4544 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4546 reclaim_str_hashtbl_size++;
4552 nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp)
4554 list_del(&crp->cr_strhash);
4556 reclaim_str_hashtbl_size--;
4560 nfs4_release_reclaim(void)
4562 struct nfs4_client_reclaim *crp = NULL;
4565 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4566 while (!list_empty(&reclaim_str_hashtbl[i])) {
4567 crp = list_entry(reclaim_str_hashtbl[i].next,
4568 struct nfs4_client_reclaim, cr_strhash);
4569 nfs4_remove_reclaim_record(crp);
4572 BUG_ON(reclaim_str_hashtbl_size);
4576 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4577 struct nfs4_client_reclaim *
4578 nfsd4_find_reclaim_client(const char *recdir)
4580 unsigned int strhashval;
4581 struct nfs4_client_reclaim *crp = NULL;
4583 dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
4585 strhashval = clientstr_hashval(recdir);
4586 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4587 if (same_name(crp->cr_recdir, recdir)) {
4595 * Called from OPEN. Look for clientid in reclaim list.
4598 nfs4_check_open_reclaim(clientid_t *clid, bool sessions)
4600 struct nfs4_client *clp;
4602 /* find clientid in conf_id_hashtbl */
4603 clp = find_confirmed_client(clid, sessions);
4605 return nfserr_reclaim_bad;
4607 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4610 #ifdef CONFIG_NFSD_FAULT_INJECTION
4612 void nfsd_forget_clients(u64 num)
4614 struct nfs4_client *clp, *next;
4618 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4623 nfs4_unlock_state();
4625 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4628 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4630 release_lockowner(lockowner(sop));
4633 static void release_openowner_sop(struct nfs4_stateowner *sop)
4635 release_openowner(openowner(sop));
4638 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4639 void (*release_sop)(struct nfs4_stateowner *))
4642 struct nfs4_stateowner *sop, *next;
4644 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4645 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4646 if (sop->so_is_open_owner != is_open_owner)
4656 void nfsd_forget_locks(u64 num)
4661 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4662 nfs4_unlock_state();
4664 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4667 void nfsd_forget_openowners(u64 num)
4672 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4673 nfs4_unlock_state();
4675 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4678 static int nfsd_process_n_delegations(u64 num, struct list_head *list)
4681 struct nfs4_file *fp, *fnext;
4682 struct nfs4_delegation *dp, *dnext;
4684 for (i = 0; i < FILE_HASH_SIZE; i++) {
4685 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4686 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4687 list_move(&dp->dl_recall_lru, list);
4697 void nfsd_forget_delegations(u64 num)
4701 struct nfs4_delegation *dp, *dnext;
4703 spin_lock(&recall_lock);
4704 count = nfsd_process_n_delegations(num, &victims);
4705 spin_unlock(&recall_lock);
4708 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
4709 unhash_delegation(dp);
4710 nfs4_unlock_state();
4712 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4715 void nfsd_recall_delegations(u64 num)
4719 struct nfs4_delegation *dp, *dnext;
4721 spin_lock(&recall_lock);
4722 count = nfsd_process_n_delegations(num, &victims);
4723 list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
4724 list_del(&dp->dl_recall_lru);
4725 nfsd_break_one_deleg(dp);
4727 spin_unlock(&recall_lock);
4729 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4732 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4734 /* initialization to perform at module load time: */
4737 nfs4_state_init(void)
4741 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4742 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4743 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4744 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4746 conf_name_tree = RB_ROOT;
4747 unconf_name_tree = RB_ROOT;
4748 for (i = 0; i < SESSION_HASH_SIZE; i++)
4749 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4750 for (i = 0; i < FILE_HASH_SIZE; i++) {
4751 INIT_LIST_HEAD(&file_hashtbl[i]);
4753 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4754 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4756 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4757 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4758 INIT_LIST_HEAD(&close_lru);
4759 INIT_LIST_HEAD(&client_lru);
4760 INIT_LIST_HEAD(&del_recall_lru);
4761 reclaim_str_hashtbl_size = 0;
4765 * Since the lifetime of a delegation isn't limited to that of an open, a
4766 * client may quite reasonably hang on to a delegation as long as it has
4767 * the inode cached. This becomes an obvious problem the first time a
4768 * client's inode cache approaches the size of the server's total memory.
4770 * For now we avoid this problem by imposing a hard limit on the number
4771 * of delegations, which varies according to the server's memory size.
4774 set_max_delegations(void)
4777 * Allow at most 4 delegations per megabyte of RAM. Quick
4778 * estimates suggest that in the worst case (where every delegation
4779 * is for a different inode), a delegation could take about 1.5K,
4780 * giving a worst case usage of about 6% of memory.
4782 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4785 /* initialization to perform when the nfsd service is started: */
4788 nfs4_state_start(void)
4790 struct net *net = &init_net;
4791 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4795 * FIXME: For now, we hang most of the pernet global stuff off of
4796 * init_net until nfsd is fully containerized. Eventually, we'll
4797 * need to pass a net pointer into this function, take a reference
4798 * to that instead and then do most of the rest of this on a per-net
4802 nfsd4_client_tracking_init(net);
4803 nn->boot_time = get_seconds();
4804 locks_start_grace(net, &nn->nfsd4_manager);
4805 nn->grace_ended = false;
4806 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4808 ret = set_callback_cred();
4813 laundry_wq = create_singlethread_workqueue("nfsd4");
4814 if (laundry_wq == NULL) {
4818 ret = nfsd4_create_callback_queue();
4820 goto out_free_laundry;
4821 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4822 set_max_delegations();
4825 destroy_workqueue(laundry_wq);
4827 nfsd4_client_tracking_exit(net);
4832 /* should be called with the state lock held */
4834 __nfs4_state_shutdown(void)
4837 struct nfs4_client *clp = NULL;
4838 struct nfs4_delegation *dp = NULL;
4839 struct list_head *pos, *next, reaplist;
4840 struct rb_node *node, *tmp;
4842 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4843 while (!list_empty(&conf_id_hashtbl[i])) {
4844 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4845 destroy_client(clp);
4849 node = rb_first(&unconf_name_tree);
4850 while (node != NULL) {
4852 node = rb_next(tmp);
4853 clp = rb_entry(tmp, struct nfs4_client, cl_namenode);
4854 rb_erase(tmp, &unconf_name_tree);
4855 destroy_client(clp);
4858 INIT_LIST_HEAD(&reaplist);
4859 spin_lock(&recall_lock);
4860 list_for_each_safe(pos, next, &del_recall_lru) {
4861 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4862 list_move(&dp->dl_recall_lru, &reaplist);
4864 spin_unlock(&recall_lock);
4865 list_for_each_safe(pos, next, &reaplist) {
4866 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4867 unhash_delegation(dp);
4870 nfsd4_client_tracking_exit(&init_net);
4875 nfs4_state_shutdown(void)
4877 struct net *net = &init_net;
4878 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4880 cancel_delayed_work_sync(&laundromat_work);
4881 destroy_workqueue(laundry_wq);
4882 locks_end_grace(&nn->nfsd4_manager);
4884 __nfs4_state_shutdown();
4885 nfs4_unlock_state();
4886 nfsd4_destroy_callback_queue();
4890 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4892 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4893 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4897 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4899 if (cstate->minorversion) {
4900 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4901 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4906 clear_current_stateid(struct nfsd4_compound_state *cstate)
4908 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4912 * functions to set current state id
4915 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4917 put_stateid(cstate, &odp->od_stateid);
4921 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4923 put_stateid(cstate, &open->op_stateid);
4927 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4929 put_stateid(cstate, &close->cl_stateid);
4933 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4935 put_stateid(cstate, &lock->lk_resp_stateid);
4939 * functions to consume current state id
4943 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4945 get_stateid(cstate, &odp->od_stateid);
4949 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4951 get_stateid(cstate, &drp->dr_stateid);
4955 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4957 get_stateid(cstate, &fsp->fr_stateid);
4961 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4963 get_stateid(cstate, &setattr->sa_stateid);
4967 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4969 get_stateid(cstate, &close->cl_stateid);
4973 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4975 get_stateid(cstate, &locku->lu_stateid);
4979 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4981 get_stateid(cstate, &read->rd_stateid);
4985 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4987 get_stateid(cstate, &write->wr_stateid);