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/sunrpc/svcauth_gss.h>
42 #include <linux/sunrpc/clnt.h>
45 #include "current_stateid.h"
47 #define NFSDDBG_FACILITY NFSDDBG_PROC
50 time_t nfsd4_lease = 90; /* default lease time */
51 time_t nfsd4_grace = 90;
52 static time_t boot_time;
54 #define all_ones {{~0,~0},~0}
55 static const stateid_t one_stateid = {
57 .si_opaque = all_ones,
59 static const stateid_t zero_stateid = {
62 static const stateid_t currentstateid = {
66 static u64 current_sessionid = 1;
68 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
69 #define ONE_STATEID(stateid) (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
70 #define CURRENT_STATEID(stateid) (!memcmp((stateid), ¤tstateid, sizeof(stateid_t)))
72 /* forward declarations */
73 static int check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner);
77 /* Currently used for almost all code touching nfsv4 state: */
78 static DEFINE_MUTEX(client_mutex);
81 * Currently used for the del_recall_lru and file hash table. In an
82 * effort to decrease the scope of the client_mutex, this spinlock may
83 * eventually cover more:
85 static DEFINE_SPINLOCK(recall_lock);
87 static struct kmem_cache *openowner_slab = NULL;
88 static struct kmem_cache *lockowner_slab = NULL;
89 static struct kmem_cache *file_slab = NULL;
90 static struct kmem_cache *stateid_slab = NULL;
91 static struct kmem_cache *deleg_slab = NULL;
96 mutex_lock(&client_mutex);
99 static void free_session(struct kref *);
101 /* Must be called under the client_lock */
102 static void nfsd4_put_session_locked(struct nfsd4_session *ses)
104 kref_put(&ses->se_ref, free_session);
107 static void nfsd4_get_session(struct nfsd4_session *ses)
109 kref_get(&ses->se_ref);
113 nfs4_unlock_state(void)
115 mutex_unlock(&client_mutex);
119 opaque_hashval(const void *ptr, int nbytes)
121 unsigned char *cptr = (unsigned char *) ptr;
131 static struct list_head del_recall_lru;
133 static void nfsd4_free_file(struct nfs4_file *f)
135 kmem_cache_free(file_slab, f);
139 put_nfs4_file(struct nfs4_file *fi)
141 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
142 list_del(&fi->fi_hash);
143 spin_unlock(&recall_lock);
150 get_nfs4_file(struct nfs4_file *fi)
152 atomic_inc(&fi->fi_ref);
155 static int num_delegations;
156 unsigned int max_delegations;
159 * Open owner state (share locks)
162 /* hash tables for lock and open owners */
163 #define OWNER_HASH_BITS 8
164 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
165 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
167 static unsigned int ownerstr_hashval(u32 clientid, struct xdr_netobj *ownername)
171 ret = opaque_hashval(ownername->data, ownername->len);
173 return ret & OWNER_HASH_MASK;
176 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
178 /* hash table for nfs4_file */
179 #define FILE_HASH_BITS 8
180 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
182 static unsigned int file_hashval(struct inode *ino)
184 /* XXX: why are we hashing on inode pointer, anyway? */
185 return hash_ptr(ino, FILE_HASH_BITS);
188 static struct list_head file_hashtbl[FILE_HASH_SIZE];
190 static void __nfs4_file_get_access(struct nfs4_file *fp, int oflag)
192 BUG_ON(!(fp->fi_fds[oflag] || fp->fi_fds[O_RDWR]));
193 atomic_inc(&fp->fi_access[oflag]);
196 static void nfs4_file_get_access(struct nfs4_file *fp, int oflag)
198 if (oflag == O_RDWR) {
199 __nfs4_file_get_access(fp, O_RDONLY);
200 __nfs4_file_get_access(fp, O_WRONLY);
202 __nfs4_file_get_access(fp, oflag);
205 static void nfs4_file_put_fd(struct nfs4_file *fp, int oflag)
207 if (fp->fi_fds[oflag]) {
208 fput(fp->fi_fds[oflag]);
209 fp->fi_fds[oflag] = NULL;
213 static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
215 if (atomic_dec_and_test(&fp->fi_access[oflag])) {
216 nfs4_file_put_fd(fp, oflag);
218 * It's also safe to get rid of the RDWR open *if*
219 * we no longer have need of the other kind of access
220 * or if we already have the other kind of open:
222 if (fp->fi_fds[1-oflag]
223 || atomic_read(&fp->fi_access[1 - oflag]) == 0)
224 nfs4_file_put_fd(fp, O_RDWR);
228 static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
230 if (oflag == O_RDWR) {
231 __nfs4_file_put_access(fp, O_RDONLY);
232 __nfs4_file_put_access(fp, O_WRONLY);
234 __nfs4_file_put_access(fp, oflag);
237 static inline int get_new_stid(struct nfs4_stid *stid)
239 static int min_stateid = 0;
240 struct idr *stateids = &stid->sc_client->cl_stateids;
244 error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
246 * Note: the necessary preallocation was done in
247 * nfs4_alloc_stateid(). The idr code caps the number of
248 * preallocations that can exist at a time, but the state lock
249 * prevents anyone from using ours before we get here:
253 * It shouldn't be a problem to reuse an opaque stateid value.
254 * I don't think it is for 4.1. But with 4.0 I worry that, for
255 * example, a stray write retransmission could be accepted by
256 * the server when it should have been rejected. Therefore,
257 * adopt a trick from the sctp code to attempt to maximize the
258 * amount of time until an id is reused, by ensuring they always
259 * "increase" (mod INT_MAX):
262 min_stateid = new_stid+1;
263 if (min_stateid == INT_MAX)
268 static void init_stid(struct nfs4_stid *stid, struct nfs4_client *cl, unsigned char type)
270 stateid_t *s = &stid->sc_stateid;
273 stid->sc_type = type;
274 stid->sc_client = cl;
275 s->si_opaque.so_clid = cl->cl_clientid;
276 new_id = get_new_stid(stid);
277 s->si_opaque.so_id = (u32)new_id;
278 /* Will be incremented before return to client: */
279 s->si_generation = 0;
282 static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab)
284 struct idr *stateids = &cl->cl_stateids;
286 if (!idr_pre_get(stateids, GFP_KERNEL))
289 * Note: if we fail here (or any time between now and the time
290 * we actually get the new idr), we won't need to undo the idr
291 * preallocation, since the idr code caps the number of
292 * preallocated entries.
294 return kmem_cache_alloc(slab, GFP_KERNEL);
297 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
299 return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
302 static struct nfs4_delegation *
303 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_ol_stateid *stp, struct svc_fh *current_fh, u32 type)
305 struct nfs4_delegation *dp;
306 struct nfs4_file *fp = stp->st_file;
308 dprintk("NFSD alloc_init_deleg\n");
310 * Major work on the lease subsystem (for example, to support
311 * calbacks on stat) will be required before we can support
312 * write delegations properly.
314 if (type != NFS4_OPEN_DELEGATE_READ)
316 if (fp->fi_had_conflict)
318 if (num_delegations > max_delegations)
320 dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
323 init_stid(&dp->dl_stid, clp, NFS4_DELEG_STID);
325 * delegation seqid's are never incremented. The 4.1 special
326 * meaning of seqid 0 isn't meaningful, really, but let's avoid
327 * 0 anyway just for consistency and use 1:
329 dp->dl_stid.sc_stateid.si_generation = 1;
331 INIT_LIST_HEAD(&dp->dl_perfile);
332 INIT_LIST_HEAD(&dp->dl_perclnt);
333 INIT_LIST_HEAD(&dp->dl_recall_lru);
337 fh_copy_shallow(&dp->dl_fh, ¤t_fh->fh_handle);
339 atomic_set(&dp->dl_count, 1);
340 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
345 nfs4_put_delegation(struct nfs4_delegation *dp)
347 if (atomic_dec_and_test(&dp->dl_count)) {
348 dprintk("NFSD: freeing dp %p\n",dp);
349 put_nfs4_file(dp->dl_file);
350 kmem_cache_free(deleg_slab, dp);
355 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
357 if (atomic_dec_and_test(&fp->fi_delegees)) {
358 vfs_setlease(fp->fi_deleg_file, F_UNLCK, &fp->fi_lease);
360 fput(fp->fi_deleg_file);
361 fp->fi_deleg_file = NULL;
365 static void unhash_stid(struct nfs4_stid *s)
367 struct idr *stateids = &s->sc_client->cl_stateids;
369 idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
372 /* Called under the state lock. */
374 unhash_delegation(struct nfs4_delegation *dp)
376 unhash_stid(&dp->dl_stid);
377 list_del_init(&dp->dl_perclnt);
378 spin_lock(&recall_lock);
379 list_del_init(&dp->dl_perfile);
380 list_del_init(&dp->dl_recall_lru);
381 spin_unlock(&recall_lock);
382 nfs4_put_deleg_lease(dp->dl_file);
383 nfs4_put_delegation(dp);
390 /* client_lock protects the client lru list and session hash table */
391 static DEFINE_SPINLOCK(client_lock);
393 /* Hash tables for nfs4_clientid state */
394 #define CLIENT_HASH_BITS 4
395 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
396 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
398 static unsigned int clientid_hashval(u32 id)
400 return id & CLIENT_HASH_MASK;
403 static unsigned int clientstr_hashval(const char *name)
405 return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
409 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
410 * used in reboot/reset lease grace period processing
412 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
413 * setclientid_confirmed info.
415 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
418 * client_lru holds client queue ordered by nfs4_client.cl_time
421 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
422 * for last close replay.
424 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
425 static int reclaim_str_hashtbl_size = 0;
426 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
427 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
428 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
429 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
430 static struct list_head client_lru;
431 static struct list_head close_lru;
434 * We store the NONE, READ, WRITE, and BOTH bits separately in the
435 * st_{access,deny}_bmap field of the stateid, in order to track not
436 * only what share bits are currently in force, but also what
437 * combinations of share bits previous opens have used. This allows us
438 * to enforce the recommendation of rfc 3530 14.2.19 that the server
439 * return an error if the client attempt to downgrade to a combination
440 * of share bits not explicable by closing some of its previous opens.
442 * XXX: This enforcement is actually incomplete, since we don't keep
443 * track of access/deny bit combinations; so, e.g., we allow:
445 * OPEN allow read, deny write
446 * OPEN allow both, deny none
447 * DOWNGRADE allow read, deny none
449 * which we should reject.
452 bmap_to_share_mode(unsigned long bmap) {
454 unsigned int access = 0;
456 for (i = 1; i < 4; i++) {
457 if (test_bit(i, &bmap))
464 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
465 unsigned int access, deny;
467 access = bmap_to_share_mode(stp->st_access_bmap);
468 deny = bmap_to_share_mode(stp->st_deny_bmap);
469 if ((access & open->op_share_deny) || (deny & open->op_share_access))
474 /* set share access for a given stateid */
476 set_access(u32 access, struct nfs4_ol_stateid *stp)
478 __set_bit(access, &stp->st_access_bmap);
481 /* clear share access for a given stateid */
483 clear_access(u32 access, struct nfs4_ol_stateid *stp)
485 __clear_bit(access, &stp->st_access_bmap);
488 /* test whether a given stateid has access */
490 test_access(u32 access, struct nfs4_ol_stateid *stp)
492 return test_bit(access, &stp->st_access_bmap);
495 /* set share deny for a given stateid */
497 set_deny(u32 access, struct nfs4_ol_stateid *stp)
499 __set_bit(access, &stp->st_deny_bmap);
502 /* clear share deny for a given stateid */
504 clear_deny(u32 access, struct nfs4_ol_stateid *stp)
506 __clear_bit(access, &stp->st_deny_bmap);
509 /* test whether a given stateid is denying specific access */
511 test_deny(u32 access, struct nfs4_ol_stateid *stp)
513 return test_bit(access, &stp->st_deny_bmap);
516 static int nfs4_access_to_omode(u32 access)
518 switch (access & NFS4_SHARE_ACCESS_BOTH) {
519 case NFS4_SHARE_ACCESS_READ:
521 case NFS4_SHARE_ACCESS_WRITE:
523 case NFS4_SHARE_ACCESS_BOTH:
529 /* release all access and file references for a given stateid */
531 release_all_access(struct nfs4_ol_stateid *stp)
535 for (i = 1; i < 4; i++) {
536 if (test_access(i, stp))
537 nfs4_file_put_access(stp->st_file,
538 nfs4_access_to_omode(i));
539 clear_access(i, stp);
543 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
545 list_del(&stp->st_perfile);
546 list_del(&stp->st_perstateowner);
549 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
551 release_all_access(stp);
552 put_nfs4_file(stp->st_file);
556 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
558 kmem_cache_free(stateid_slab, stp);
561 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
565 unhash_generic_stateid(stp);
566 unhash_stid(&stp->st_stid);
567 file = find_any_file(stp->st_file);
569 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
570 close_generic_stateid(stp);
571 free_generic_stateid(stp);
574 static void unhash_lockowner(struct nfs4_lockowner *lo)
576 struct nfs4_ol_stateid *stp;
578 list_del(&lo->lo_owner.so_strhash);
579 list_del(&lo->lo_perstateid);
580 list_del(&lo->lo_owner_ino_hash);
581 while (!list_empty(&lo->lo_owner.so_stateids)) {
582 stp = list_first_entry(&lo->lo_owner.so_stateids,
583 struct nfs4_ol_stateid, st_perstateowner);
584 release_lock_stateid(stp);
588 static void release_lockowner(struct nfs4_lockowner *lo)
590 unhash_lockowner(lo);
591 nfs4_free_lockowner(lo);
595 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
597 struct nfs4_lockowner *lo;
599 while (!list_empty(&open_stp->st_lockowners)) {
600 lo = list_entry(open_stp->st_lockowners.next,
601 struct nfs4_lockowner, lo_perstateid);
602 release_lockowner(lo);
606 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
608 unhash_generic_stateid(stp);
609 release_stateid_lockowners(stp);
610 close_generic_stateid(stp);
613 static void release_open_stateid(struct nfs4_ol_stateid *stp)
615 unhash_open_stateid(stp);
616 unhash_stid(&stp->st_stid);
617 free_generic_stateid(stp);
620 static void unhash_openowner(struct nfs4_openowner *oo)
622 struct nfs4_ol_stateid *stp;
624 list_del(&oo->oo_owner.so_strhash);
625 list_del(&oo->oo_perclient);
626 while (!list_empty(&oo->oo_owner.so_stateids)) {
627 stp = list_first_entry(&oo->oo_owner.so_stateids,
628 struct nfs4_ol_stateid, st_perstateowner);
629 release_open_stateid(stp);
633 static void release_last_closed_stateid(struct nfs4_openowner *oo)
635 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
638 unhash_stid(&s->st_stid);
639 free_generic_stateid(s);
640 oo->oo_last_closed_stid = NULL;
644 static void release_openowner(struct nfs4_openowner *oo)
646 unhash_openowner(oo);
647 list_del(&oo->oo_close_lru);
648 release_last_closed_stateid(oo);
649 nfs4_free_openowner(oo);
652 #define SESSION_HASH_SIZE 512
653 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
656 hash_sessionid(struct nfs4_sessionid *sessionid)
658 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
660 return sid->sequence % SESSION_HASH_SIZE;
665 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
667 u32 *ptr = (u32 *)(&sessionid->data[0]);
668 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
672 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
679 gen_sessionid(struct nfsd4_session *ses)
681 struct nfs4_client *clp = ses->se_client;
682 struct nfsd4_sessionid *sid;
684 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
685 sid->clientid = clp->cl_clientid;
686 sid->sequence = current_sessionid++;
691 * The protocol defines ca_maxresponssize_cached to include the size of
692 * the rpc header, but all we need to cache is the data starting after
693 * the end of the initial SEQUENCE operation--the rest we regenerate
694 * each time. Therefore we can advertise a ca_maxresponssize_cached
695 * value that is the number of bytes in our cache plus a few additional
696 * bytes. In order to stay on the safe side, and not promise more than
697 * we can cache, those additional bytes must be the minimum possible: 24
698 * bytes of rpc header (xid through accept state, with AUTH_NULL
699 * verifier), 12 for the compound header (with zero-length tag), and 44
700 * for the SEQUENCE op response:
702 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
705 free_session_slots(struct nfsd4_session *ses)
709 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
710 kfree(ses->se_slots[i]);
714 * We don't actually need to cache the rpc and session headers, so we
715 * can allocate a little less for each slot:
717 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
719 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
722 static int nfsd4_sanitize_slot_size(u32 size)
724 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
725 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
731 * XXX: If we run out of reserved DRC memory we could (up to a point)
732 * re-negotiate active sessions and reduce their slot usage to make
733 * room for new connections. For now we just fail the create session.
735 static int nfsd4_get_drc_mem(int slotsize, u32 num)
739 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
741 spin_lock(&nfsd_drc_lock);
742 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
743 nfsd_drc_max_mem - nfsd_drc_mem_used);
744 num = min_t(int, num, avail / slotsize);
745 nfsd_drc_mem_used += num * slotsize;
746 spin_unlock(&nfsd_drc_lock);
751 static void nfsd4_put_drc_mem(int slotsize, int num)
753 spin_lock(&nfsd_drc_lock);
754 nfsd_drc_mem_used -= slotsize * num;
755 spin_unlock(&nfsd_drc_lock);
758 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
760 struct nfsd4_session *new;
763 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
764 + sizeof(struct nfsd4_session) > PAGE_SIZE);
765 mem = numslots * sizeof(struct nfsd4_slot *);
767 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
770 /* allocate each struct nfsd4_slot and data cache in one piece */
771 for (i = 0; i < numslots; i++) {
772 mem = sizeof(struct nfsd4_slot) + slotsize;
773 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
774 if (!new->se_slots[i])
780 kfree(new->se_slots[i]);
785 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
787 u32 maxrpc = nfsd_serv->sv_max_mesg;
789 new->maxreqs = numslots;
790 new->maxresp_cached = min_t(u32, req->maxresp_cached,
791 slotsize + NFSD_MIN_HDR_SEQ_SZ);
792 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
793 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
794 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
797 static void free_conn(struct nfsd4_conn *c)
799 svc_xprt_put(c->cn_xprt);
803 static void nfsd4_conn_lost(struct svc_xpt_user *u)
805 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
806 struct nfs4_client *clp = c->cn_session->se_client;
808 spin_lock(&clp->cl_lock);
809 if (!list_empty(&c->cn_persession)) {
810 list_del(&c->cn_persession);
813 spin_unlock(&clp->cl_lock);
814 nfsd4_probe_callback(clp);
817 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
819 struct nfsd4_conn *conn;
821 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
824 svc_xprt_get(rqstp->rq_xprt);
825 conn->cn_xprt = rqstp->rq_xprt;
826 conn->cn_flags = flags;
827 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
831 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
833 conn->cn_session = ses;
834 list_add(&conn->cn_persession, &ses->se_conns);
837 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
839 struct nfs4_client *clp = ses->se_client;
841 spin_lock(&clp->cl_lock);
842 __nfsd4_hash_conn(conn, ses);
843 spin_unlock(&clp->cl_lock);
846 static int nfsd4_register_conn(struct nfsd4_conn *conn)
848 conn->cn_xpt_user.callback = nfsd4_conn_lost;
849 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
852 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
854 struct nfsd4_conn *conn;
857 conn = alloc_conn(rqstp, dir);
859 return nfserr_jukebox;
860 nfsd4_hash_conn(conn, ses);
861 ret = nfsd4_register_conn(conn);
863 /* oops; xprt is already down: */
864 nfsd4_conn_lost(&conn->cn_xpt_user);
868 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
870 u32 dir = NFS4_CDFC4_FORE;
872 if (ses->se_flags & SESSION4_BACK_CHAN)
873 dir |= NFS4_CDFC4_BACK;
875 return nfsd4_new_conn(rqstp, ses, dir);
878 /* must be called under client_lock */
879 static void nfsd4_del_conns(struct nfsd4_session *s)
881 struct nfs4_client *clp = s->se_client;
882 struct nfsd4_conn *c;
884 spin_lock(&clp->cl_lock);
885 while (!list_empty(&s->se_conns)) {
886 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
887 list_del_init(&c->cn_persession);
888 spin_unlock(&clp->cl_lock);
890 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
893 spin_lock(&clp->cl_lock);
895 spin_unlock(&clp->cl_lock);
898 static void free_session(struct kref *kref)
900 struct nfsd4_session *ses;
903 BUG_ON(!spin_is_locked(&client_lock));
904 ses = container_of(kref, struct nfsd4_session, se_ref);
905 nfsd4_del_conns(ses);
906 spin_lock(&nfsd_drc_lock);
907 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
908 nfsd_drc_mem_used -= mem;
909 spin_unlock(&nfsd_drc_lock);
910 free_session_slots(ses);
914 void nfsd4_put_session(struct nfsd4_session *ses)
916 spin_lock(&client_lock);
917 nfsd4_put_session_locked(ses);
918 spin_unlock(&client_lock);
921 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
923 struct nfsd4_session *new;
924 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
925 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);
948 new->se_client = clp;
951 INIT_LIST_HEAD(&new->se_conns);
953 new->se_cb_seq_nr = 1;
954 new->se_flags = cses->flags;
955 new->se_cb_prog = cses->callback_prog;
956 kref_init(&new->se_ref);
957 idx = hash_sessionid(&new->se_sessionid);
958 spin_lock(&client_lock);
959 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
960 spin_lock(&clp->cl_lock);
961 list_add(&new->se_perclnt, &clp->cl_sessions);
962 spin_unlock(&clp->cl_lock);
963 spin_unlock(&client_lock);
965 status = nfsd4_new_conn_from_crses(rqstp, new);
966 /* whoops: benny points out, status is ignored! (err, or bogus) */
968 spin_lock(&client_lock);
969 free_session(&new->se_ref);
970 spin_unlock(&client_lock);
973 if (cses->flags & SESSION4_BACK_CHAN) {
974 struct sockaddr *sa = svc_addr(rqstp);
976 * This is a little silly; with sessions there's no real
977 * use for the callback address. Use the peer address
978 * as a reasonable default for now, but consider fixing
979 * the rpc client not to require an address in the
982 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
983 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
985 nfsd4_probe_callback(clp);
989 /* caller must hold client_lock */
990 static struct nfsd4_session *
991 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
993 struct nfsd4_session *elem;
996 dump_sessionid(__func__, sessionid);
997 idx = hash_sessionid(sessionid);
998 /* Search in the appropriate list */
999 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
1000 if (!memcmp(elem->se_sessionid.data, sessionid->data,
1001 NFS4_MAX_SESSIONID_LEN)) {
1006 dprintk("%s: session not found\n", __func__);
1010 /* caller must hold client_lock */
1012 unhash_session(struct nfsd4_session *ses)
1014 list_del(&ses->se_hash);
1015 spin_lock(&ses->se_client->cl_lock);
1016 list_del(&ses->se_perclnt);
1017 spin_unlock(&ses->se_client->cl_lock);
1020 /* must be called under the client_lock */
1022 renew_client_locked(struct nfs4_client *clp)
1024 if (is_client_expired(clp)) {
1026 printk("%s: client (clientid %08x/%08x) already expired\n",
1028 clp->cl_clientid.cl_boot,
1029 clp->cl_clientid.cl_id);
1033 dprintk("renewing client (clientid %08x/%08x)\n",
1034 clp->cl_clientid.cl_boot,
1035 clp->cl_clientid.cl_id);
1036 list_move_tail(&clp->cl_lru, &client_lru);
1037 clp->cl_time = get_seconds();
1041 renew_client(struct nfs4_client *clp)
1043 spin_lock(&client_lock);
1044 renew_client_locked(clp);
1045 spin_unlock(&client_lock);
1048 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1050 STALE_CLIENTID(clientid_t *clid)
1052 if (clid->cl_boot == boot_time)
1054 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1055 clid->cl_boot, clid->cl_id, boot_time);
1060 * XXX Should we use a slab cache ?
1061 * This type of memory management is somewhat inefficient, but we use it
1062 * anyway since SETCLIENTID is not a common operation.
1064 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1066 struct nfs4_client *clp;
1068 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1071 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1072 if (clp->cl_name.data == NULL) {
1076 clp->cl_name.len = name.len;
1081 free_client(struct nfs4_client *clp)
1083 BUG_ON(!spin_is_locked(&client_lock));
1084 while (!list_empty(&clp->cl_sessions)) {
1085 struct nfsd4_session *ses;
1086 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1088 list_del(&ses->se_perclnt);
1089 nfsd4_put_session_locked(ses);
1091 free_svc_cred(&clp->cl_cred);
1092 kfree(clp->cl_name.data);
1097 release_session_client(struct nfsd4_session *session)
1099 struct nfs4_client *clp = session->se_client;
1101 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1103 if (is_client_expired(clp)) {
1105 session->se_client = NULL;
1107 renew_client_locked(clp);
1108 spin_unlock(&client_lock);
1111 /* must be called under the client_lock */
1113 unhash_client_locked(struct nfs4_client *clp)
1115 struct nfsd4_session *ses;
1117 mark_client_expired(clp);
1118 list_del(&clp->cl_lru);
1119 spin_lock(&clp->cl_lock);
1120 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1121 list_del_init(&ses->se_hash);
1122 spin_unlock(&clp->cl_lock);
1126 expire_client(struct nfs4_client *clp)
1128 struct nfs4_openowner *oo;
1129 struct nfs4_delegation *dp;
1130 struct list_head reaplist;
1132 INIT_LIST_HEAD(&reaplist);
1133 spin_lock(&recall_lock);
1134 while (!list_empty(&clp->cl_delegations)) {
1135 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1136 list_del_init(&dp->dl_perclnt);
1137 list_move(&dp->dl_recall_lru, &reaplist);
1139 spin_unlock(&recall_lock);
1140 while (!list_empty(&reaplist)) {
1141 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1142 unhash_delegation(dp);
1144 while (!list_empty(&clp->cl_openowners)) {
1145 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1146 release_openowner(oo);
1148 nfsd4_shutdown_callback(clp);
1149 if (clp->cl_cb_conn.cb_xprt)
1150 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1151 list_del(&clp->cl_idhash);
1152 list_del(&clp->cl_strhash);
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 copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1162 memcpy(target->cl_verifier.data, source->data,
1163 sizeof(target->cl_verifier.data));
1166 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1168 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1169 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1172 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1174 if (source->cr_principal) {
1175 target->cr_principal =
1176 kstrdup(source->cr_principal, GFP_KERNEL);
1177 if (target->cr_principal == NULL)
1180 target->cr_principal = NULL;
1181 target->cr_flavor = source->cr_flavor;
1182 target->cr_uid = source->cr_uid;
1183 target->cr_gid = source->cr_gid;
1184 target->cr_group_info = source->cr_group_info;
1185 get_group_info(target->cr_group_info);
1189 static int same_name(const char *n1, const char *n2)
1191 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1195 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1197 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1201 same_clid(clientid_t *cl1, clientid_t *cl2)
1203 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1206 static bool groups_equal(struct group_info *g1, struct group_info *g2)
1210 if (g1->ngroups != g2->ngroups)
1212 for (i=0; i<g1->ngroups; i++)
1213 if (GROUP_AT(g1, i) != GROUP_AT(g2, i))
1219 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1221 if ((cr1->cr_flavor != cr2->cr_flavor)
1222 || (cr1->cr_uid != cr2->cr_uid)
1223 || (cr1->cr_gid != cr2->cr_gid)
1224 || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1226 if (cr1->cr_principal == cr2->cr_principal)
1228 if (!cr1->cr_principal || !cr2->cr_principal)
1230 return 0 == strcmp(cr1->cr_principal, cr1->cr_principal);
1233 static void gen_clid(struct nfs4_client *clp)
1235 static u32 current_clientid = 1;
1237 clp->cl_clientid.cl_boot = boot_time;
1238 clp->cl_clientid.cl_id = current_clientid++;
1241 static void gen_confirm(struct nfs4_client *clp)
1246 verf[0] = (__be32)get_seconds();
1247 verf[1] = (__be32)i++;
1248 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1251 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1253 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1256 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1258 struct nfs4_stid *s;
1260 s = find_stateid(cl, t);
1263 if (typemask & s->sc_type)
1268 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1269 struct svc_rqst *rqstp, nfs4_verifier *verf)
1271 struct nfs4_client *clp;
1272 struct sockaddr *sa = svc_addr(rqstp);
1275 clp = alloc_client(name);
1279 INIT_LIST_HEAD(&clp->cl_sessions);
1280 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1282 spin_lock(&client_lock);
1284 spin_unlock(&client_lock);
1287 idr_init(&clp->cl_stateids);
1288 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1289 atomic_set(&clp->cl_refcount, 0);
1290 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1291 INIT_LIST_HEAD(&clp->cl_idhash);
1292 INIT_LIST_HEAD(&clp->cl_strhash);
1293 INIT_LIST_HEAD(&clp->cl_openowners);
1294 INIT_LIST_HEAD(&clp->cl_delegations);
1295 INIT_LIST_HEAD(&clp->cl_lru);
1296 INIT_LIST_HEAD(&clp->cl_callbacks);
1297 spin_lock_init(&clp->cl_lock);
1298 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1299 clp->cl_time = get_seconds();
1300 clear_bit(0, &clp->cl_cb_slot_busy);
1301 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1302 copy_verf(clp, verf);
1303 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1305 clp->cl_cb_session = NULL;
1310 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1312 unsigned int idhashval;
1314 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1315 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1316 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1321 move_to_confirmed(struct nfs4_client *clp)
1323 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1324 unsigned int strhashval;
1326 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1327 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1328 strhashval = clientstr_hashval(clp->cl_recdir);
1329 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1333 static struct nfs4_client *
1334 find_confirmed_client(clientid_t *clid)
1336 struct nfs4_client *clp;
1337 unsigned int idhashval = clientid_hashval(clid->cl_id);
1339 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1340 if (same_clid(&clp->cl_clientid, clid)) {
1348 static struct nfs4_client *
1349 find_unconfirmed_client(clientid_t *clid)
1351 struct nfs4_client *clp;
1352 unsigned int idhashval = clientid_hashval(clid->cl_id);
1354 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1355 if (same_clid(&clp->cl_clientid, clid))
1361 static bool clp_used_exchangeid(struct nfs4_client *clp)
1363 return clp->cl_exchange_flags != 0;
1366 static struct nfs4_client *
1367 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1369 struct nfs4_client *clp;
1371 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1372 if (same_name(clp->cl_recdir, dname))
1378 static struct nfs4_client *
1379 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1381 struct nfs4_client *clp;
1383 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1384 if (same_name(clp->cl_recdir, dname))
1391 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1393 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1394 struct sockaddr *sa = svc_addr(rqstp);
1395 u32 scopeid = rpc_get_scope_id(sa);
1396 unsigned short expected_family;
1398 /* Currently, we only support tcp and tcp6 for the callback channel */
1399 if (se->se_callback_netid_len == 3 &&
1400 !memcmp(se->se_callback_netid_val, "tcp", 3))
1401 expected_family = AF_INET;
1402 else if (se->se_callback_netid_len == 4 &&
1403 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1404 expected_family = AF_INET6;
1408 conn->cb_addrlen = rpc_uaddr2sockaddr(&init_net, se->se_callback_addr_val,
1409 se->se_callback_addr_len,
1410 (struct sockaddr *)&conn->cb_addr,
1411 sizeof(conn->cb_addr));
1413 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1416 if (conn->cb_addr.ss_family == AF_INET6)
1417 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1419 conn->cb_prog = se->se_callback_prog;
1420 conn->cb_ident = se->se_callback_ident;
1421 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1424 conn->cb_addr.ss_family = AF_UNSPEC;
1425 conn->cb_addrlen = 0;
1426 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1427 "will not receive delegations\n",
1428 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1434 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1437 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1439 struct nfsd4_slot *slot = resp->cstate.slot;
1442 dprintk("--> %s slot %p\n", __func__, slot);
1444 slot->sl_opcnt = resp->opcnt;
1445 slot->sl_status = resp->cstate.status;
1447 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1448 if (nfsd4_not_cached(resp)) {
1449 slot->sl_datalen = 0;
1452 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1453 base = (char *)resp->cstate.datap -
1454 (char *)resp->xbuf->head[0].iov_base;
1455 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1457 WARN("%s: sessions DRC could not cache compound\n", __func__);
1462 * Encode the replay sequence operation from the slot values.
1463 * If cachethis is FALSE encode the uncached rep error on the next
1464 * operation which sets resp->p and increments resp->opcnt for
1465 * nfs4svc_encode_compoundres.
1469 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1470 struct nfsd4_compoundres *resp)
1472 struct nfsd4_op *op;
1473 struct nfsd4_slot *slot = resp->cstate.slot;
1475 /* Encode the replayed sequence operation */
1476 op = &args->ops[resp->opcnt - 1];
1477 nfsd4_encode_operation(resp, op);
1479 /* Return nfserr_retry_uncached_rep in next operation. */
1480 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1481 op = &args->ops[resp->opcnt++];
1482 op->status = nfserr_retry_uncached_rep;
1483 nfsd4_encode_operation(resp, op);
1489 * The sequence operation is not cached because we can use the slot and
1493 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1494 struct nfsd4_sequence *seq)
1496 struct nfsd4_slot *slot = resp->cstate.slot;
1499 dprintk("--> %s slot %p\n", __func__, slot);
1501 /* Either returns 0 or nfserr_retry_uncached */
1502 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1503 if (status == nfserr_retry_uncached_rep)
1506 /* The sequence operation has been encoded, cstate->datap set. */
1507 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1509 resp->opcnt = slot->sl_opcnt;
1510 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1511 status = slot->sl_status;
1517 * Set the exchange_id flags returned by the server.
1520 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1522 /* pNFS is not supported */
1523 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1525 /* Referrals are supported, Migration is not. */
1526 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1528 /* set the wire flags to return to client. */
1529 clid->flags = new->cl_exchange_flags;
1532 static bool client_has_state(struct nfs4_client *clp)
1535 * Note clp->cl_openowners check isn't quite right: there's no
1536 * need to count owners without stateid's.
1538 * Also note we should probably be using this in 4.0 case too.
1540 return list_empty(&clp->cl_openowners)
1541 && list_empty(&clp->cl_delegations)
1542 && list_empty(&clp->cl_sessions);
1546 nfsd4_exchange_id(struct svc_rqst *rqstp,
1547 struct nfsd4_compound_state *cstate,
1548 struct nfsd4_exchange_id *exid)
1550 struct nfs4_client *unconf, *conf, *new;
1552 unsigned int strhashval;
1553 char dname[HEXDIR_LEN];
1554 char addr_str[INET6_ADDRSTRLEN];
1555 nfs4_verifier verf = exid->verifier;
1556 struct sockaddr *sa = svc_addr(rqstp);
1557 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1559 rpc_ntop(sa, addr_str, sizeof(addr_str));
1560 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1561 "ip_addr=%s flags %x, spa_how %d\n",
1562 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1563 addr_str, exid->flags, exid->spa_how);
1565 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1566 return nfserr_inval;
1568 /* Currently only support SP4_NONE */
1569 switch (exid->spa_how) {
1573 return nfserr_serverfault;
1575 BUG(); /* checked by xdr code */
1577 return nfserr_serverfault; /* no excuse :-/ */
1580 status = nfs4_make_rec_clidname(dname, &exid->clname);
1585 strhashval = clientstr_hashval(dname);
1587 /* Cases below refer to rfc 5661 section 18.35.4: */
1589 conf = find_confirmed_client_by_str(dname, strhashval);
1591 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1592 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1595 if (!clp_used_exchangeid(conf)) { /* buggy client */
1596 status = nfserr_inval;
1599 if (!creds_match) { /* case 9 */
1600 status = nfserr_perm;
1603 if (!verfs_match) { /* case 8 */
1604 status = nfserr_not_same;
1608 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1612 if (!creds_match) { /* case 3 */
1613 if (client_has_state(conf)) {
1614 status = nfserr_clid_inuse;
1619 if (verfs_match) { /* case 2 */
1620 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1624 /* case 5, client reboot */
1626 expire_client(conf);
1630 if (update) { /* case 7 */
1631 status = nfserr_noent;
1635 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1636 if (unconf) /* case 4, possible retry or client restart */
1637 expire_client(unconf);
1639 /* case 1 (normal case) */
1641 new = create_client(exid->clname, dname, rqstp, &verf);
1643 status = nfserr_jukebox;
1648 add_to_unconfirmed(new, strhashval);
1650 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1651 exid->clientid.cl_id = new->cl_clientid.cl_id;
1654 nfsd4_set_ex_flags(new, exid);
1656 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1657 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1661 nfs4_unlock_state();
1666 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1668 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1671 /* The slot is in use, and no response has been sent. */
1673 if (seqid == slot_seqid)
1674 return nfserr_jukebox;
1676 return nfserr_seq_misordered;
1678 /* Note unsigned 32-bit arithmetic handles wraparound: */
1679 if (likely(seqid == slot_seqid + 1))
1681 if (seqid == slot_seqid)
1682 return nfserr_replay_cache;
1683 return nfserr_seq_misordered;
1687 * Cache the create session result into the create session single DRC
1688 * slot cache by saving the xdr structure. sl_seqid has been set.
1689 * Do this for solo or embedded create session operations.
1692 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1693 struct nfsd4_clid_slot *slot, __be32 nfserr)
1695 slot->sl_status = nfserr;
1696 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1700 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1701 struct nfsd4_clid_slot *slot)
1703 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1704 return slot->sl_status;
1707 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1708 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1709 1 + /* MIN tag is length with zero, only length */ \
1710 3 + /* version, opcount, opcode */ \
1711 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1712 /* seqid, slotID, slotID, cache */ \
1713 4 ) * sizeof(__be32))
1715 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1716 2 + /* verifier: AUTH_NULL, length 0 */\
1718 1 + /* MIN tag is length with zero, only length */ \
1719 3 + /* opcount, opcode, opstatus*/ \
1720 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1721 /* seqid, slotID, slotID, slotID, status */ \
1722 5 ) * sizeof(__be32))
1724 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1726 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1727 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1731 nfsd4_create_session(struct svc_rqst *rqstp,
1732 struct nfsd4_compound_state *cstate,
1733 struct nfsd4_create_session *cr_ses)
1735 struct sockaddr *sa = svc_addr(rqstp);
1736 struct nfs4_client *conf, *unconf;
1737 struct nfsd4_session *new;
1738 struct nfsd4_clid_slot *cs_slot = NULL;
1739 bool confirm_me = false;
1742 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1743 return nfserr_inval;
1746 unconf = find_unconfirmed_client(&cr_ses->clientid);
1747 conf = find_confirmed_client(&cr_ses->clientid);
1750 cs_slot = &conf->cl_cs_slot;
1751 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1752 if (status == nfserr_replay_cache) {
1753 dprintk("Got a create_session replay! seqid= %d\n",
1755 /* Return the cached reply status */
1756 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1758 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1759 status = nfserr_seq_misordered;
1760 dprintk("Sequence misordered!\n");
1761 dprintk("Expected seqid= %d but got seqid= %d\n",
1762 cs_slot->sl_seqid, cr_ses->seqid);
1765 } else if (unconf) {
1766 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1767 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1768 status = nfserr_clid_inuse;
1772 cs_slot = &unconf->cl_cs_slot;
1773 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1775 /* an unconfirmed replay returns misordered */
1776 status = nfserr_seq_misordered;
1783 status = nfserr_stale_clientid;
1788 * XXX: we should probably set this at creation time, and check
1789 * for consistent minorversion use throughout:
1791 conf->cl_minorversion = 1;
1793 * We do not support RDMA or persistent sessions
1795 cr_ses->flags &= ~SESSION4_PERSIST;
1796 cr_ses->flags &= ~SESSION4_RDMA;
1798 status = nfserr_toosmall;
1799 if (check_forechannel_attrs(cr_ses->fore_channel))
1802 status = nfserr_jukebox;
1803 new = alloc_init_session(rqstp, conf, cr_ses);
1807 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1808 NFS4_MAX_SESSIONID_LEN);
1809 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1810 sizeof(struct nfsd4_channel_attrs));
1811 cs_slot->sl_seqid++;
1812 cr_ses->seqid = cs_slot->sl_seqid;
1814 /* cache solo and embedded create sessions under the state lock */
1815 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1817 move_to_confirmed(conf);
1819 nfs4_unlock_state();
1820 dprintk("%s returns %d\n", __func__, ntohl(status));
1824 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1826 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1827 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1829 return argp->opcnt == resp->opcnt;
1832 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1835 case NFS4_CDFC4_FORE:
1836 case NFS4_CDFC4_BACK:
1838 case NFS4_CDFC4_FORE_OR_BOTH:
1839 case NFS4_CDFC4_BACK_OR_BOTH:
1840 *dir = NFS4_CDFC4_BOTH;
1843 return nfserr_inval;
1846 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1847 struct nfsd4_compound_state *cstate,
1848 struct nfsd4_bind_conn_to_session *bcts)
1852 if (!nfsd4_last_compound_op(rqstp))
1853 return nfserr_not_only_op;
1854 spin_lock(&client_lock);
1855 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1856 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1857 * client_lock iself: */
1858 if (cstate->session) {
1859 nfsd4_get_session(cstate->session);
1860 atomic_inc(&cstate->session->se_client->cl_refcount);
1862 spin_unlock(&client_lock);
1863 if (!cstate->session)
1864 return nfserr_badsession;
1866 status = nfsd4_map_bcts_dir(&bcts->dir);
1868 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1872 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1876 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1880 nfsd4_destroy_session(struct svc_rqst *r,
1881 struct nfsd4_compound_state *cstate,
1882 struct nfsd4_destroy_session *sessionid)
1884 struct nfsd4_session *ses;
1885 __be32 status = nfserr_badsession;
1888 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1889 * - Should we return nfserr_back_chan_busy if waiting for
1890 * callbacks on to-be-destroyed session?
1891 * - Do we need to clear any callback info from previous session?
1894 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1895 if (!nfsd4_last_compound_op(r))
1896 return nfserr_not_only_op;
1898 dump_sessionid(__func__, &sessionid->sessionid);
1899 spin_lock(&client_lock);
1900 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1902 spin_unlock(&client_lock);
1906 unhash_session(ses);
1907 spin_unlock(&client_lock);
1910 nfsd4_probe_callback_sync(ses->se_client);
1911 nfs4_unlock_state();
1913 spin_lock(&client_lock);
1914 nfsd4_del_conns(ses);
1915 nfsd4_put_session_locked(ses);
1916 spin_unlock(&client_lock);
1919 dprintk("%s returns %d\n", __func__, ntohl(status));
1923 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1925 struct nfsd4_conn *c;
1927 list_for_each_entry(c, &s->se_conns, cn_persession) {
1928 if (c->cn_xprt == xpt) {
1935 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1937 struct nfs4_client *clp = ses->se_client;
1938 struct nfsd4_conn *c;
1941 spin_lock(&clp->cl_lock);
1942 c = __nfsd4_find_conn(new->cn_xprt, ses);
1944 spin_unlock(&clp->cl_lock);
1948 __nfsd4_hash_conn(new, ses);
1949 spin_unlock(&clp->cl_lock);
1950 ret = nfsd4_register_conn(new);
1952 /* oops; xprt is already down: */
1953 nfsd4_conn_lost(&new->cn_xpt_user);
1957 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1959 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1961 return args->opcnt > session->se_fchannel.maxops;
1964 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1965 struct nfsd4_session *session)
1967 struct xdr_buf *xb = &rqstp->rq_arg;
1969 return xb->len > session->se_fchannel.maxreq_sz;
1973 nfsd4_sequence(struct svc_rqst *rqstp,
1974 struct nfsd4_compound_state *cstate,
1975 struct nfsd4_sequence *seq)
1977 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1978 struct nfsd4_session *session;
1979 struct nfsd4_slot *slot;
1980 struct nfsd4_conn *conn;
1983 if (resp->opcnt != 1)
1984 return nfserr_sequence_pos;
1987 * Will be either used or freed by nfsd4_sequence_check_conn
1990 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1992 return nfserr_jukebox;
1994 spin_lock(&client_lock);
1995 status = nfserr_badsession;
1996 session = find_in_sessionid_hashtbl(&seq->sessionid);
2000 status = nfserr_too_many_ops;
2001 if (nfsd4_session_too_many_ops(rqstp, session))
2004 status = nfserr_req_too_big;
2005 if (nfsd4_request_too_big(rqstp, session))
2008 status = nfserr_badslot;
2009 if (seq->slotid >= session->se_fchannel.maxreqs)
2012 slot = session->se_slots[seq->slotid];
2013 dprintk("%s: slotid %d\n", __func__, seq->slotid);
2015 /* We do not negotiate the number of slots yet, so set the
2016 * maxslots to the session maxreqs which is used to encode
2017 * sr_highest_slotid and the sr_target_slot id to maxslots */
2018 seq->maxslots = session->se_fchannel.maxreqs;
2020 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2021 slot->sl_flags & NFSD4_SLOT_INUSE);
2022 if (status == nfserr_replay_cache) {
2023 status = nfserr_seq_misordered;
2024 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2026 cstate->slot = slot;
2027 cstate->session = session;
2028 /* Return the cached reply status and set cstate->status
2029 * for nfsd4_proc_compound processing */
2030 status = nfsd4_replay_cache_entry(resp, seq);
2031 cstate->status = nfserr_replay_cache;
2037 nfsd4_sequence_check_conn(conn, session);
2040 /* Success! bump slot seqid */
2041 slot->sl_seqid = seq->seqid;
2042 slot->sl_flags |= NFSD4_SLOT_INUSE;
2044 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2046 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2048 cstate->slot = slot;
2049 cstate->session = session;
2052 /* Hold a session reference until done processing the compound. */
2053 if (cstate->session) {
2054 struct nfs4_client *clp = session->se_client;
2056 nfsd4_get_session(cstate->session);
2057 atomic_inc(&clp->cl_refcount);
2058 switch (clp->cl_cb_state) {
2060 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2062 case NFSD4_CB_FAULT:
2063 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2066 seq->status_flags = 0;
2070 spin_unlock(&client_lock);
2071 dprintk("%s: return %d\n", __func__, ntohl(status));
2075 static inline bool has_resources(struct nfs4_client *clp)
2077 return !list_empty(&clp->cl_openowners)
2078 || !list_empty(&clp->cl_delegations)
2079 || !list_empty(&clp->cl_sessions);
2083 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2085 struct nfs4_client *conf, *unconf, *clp;
2089 unconf = find_unconfirmed_client(&dc->clientid);
2090 conf = find_confirmed_client(&dc->clientid);
2095 if (!is_client_expired(conf) && has_resources(conf)) {
2096 status = nfserr_clientid_busy;
2100 /* rfc5661 18.50.3 */
2101 if (cstate->session && conf == cstate->session->se_client) {
2102 status = nfserr_clientid_busy;
2108 status = nfserr_stale_clientid;
2114 nfs4_unlock_state();
2115 dprintk("%s return %d\n", __func__, ntohl(status));
2120 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2124 if (rc->rca_one_fs) {
2125 if (!cstate->current_fh.fh_dentry)
2126 return nfserr_nofilehandle;
2128 * We don't take advantage of the rca_one_fs case.
2129 * That's OK, it's optional, we can safely ignore it.
2135 status = nfserr_complete_already;
2136 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2137 &cstate->session->se_client->cl_flags))
2140 status = nfserr_stale_clientid;
2141 if (is_client_expired(cstate->session->se_client))
2143 * The following error isn't really legal.
2144 * But we only get here if the client just explicitly
2145 * destroyed the client. Surely it no longer cares what
2146 * error it gets back on an operation for the dead
2152 nfsd4_client_record_create(cstate->session->se_client);
2154 nfs4_unlock_state();
2159 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2160 struct nfsd4_setclientid *setclid)
2162 struct xdr_netobj clname = setclid->se_name;
2163 nfs4_verifier clverifier = setclid->se_verf;
2164 unsigned int strhashval;
2165 struct nfs4_client *conf, *unconf, *new;
2167 char dname[HEXDIR_LEN];
2169 status = nfs4_make_rec_clidname(dname, &clname);
2173 strhashval = clientstr_hashval(dname);
2175 /* Cases below refer to rfc 3530 section 14.2.33: */
2177 conf = find_confirmed_client_by_str(dname, strhashval);
2180 status = nfserr_clid_inuse;
2181 if (clp_used_exchangeid(conf))
2183 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2184 char addr_str[INET6_ADDRSTRLEN];
2185 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2187 dprintk("NFSD: setclientid: string in use by client "
2188 "at %s\n", addr_str);
2192 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2194 expire_client(unconf);
2195 status = nfserr_jukebox;
2196 new = create_client(clname, dname, rqstp, &clverifier);
2199 if (conf && same_verf(&conf->cl_verifier, &clverifier))
2200 /* case 1: probable callback update */
2201 copy_clid(new, conf);
2202 else /* case 4 (new client) or cases 2, 3 (client reboot): */
2205 * XXX: we should probably set this at creation time, and check
2206 * for consistent minorversion use throughout:
2208 new->cl_minorversion = 0;
2209 gen_callback(new, setclid, rqstp);
2210 add_to_unconfirmed(new, strhashval);
2211 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2212 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2213 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2216 nfs4_unlock_state();
2222 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2223 struct nfsd4_compound_state *cstate,
2224 struct nfsd4_setclientid_confirm *setclientid_confirm)
2226 struct nfs4_client *conf, *unconf;
2227 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2228 clientid_t * clid = &setclientid_confirm->sc_clientid;
2231 if (STALE_CLIENTID(clid))
2232 return nfserr_stale_clientid;
2235 conf = find_confirmed_client(clid);
2236 unconf = find_unconfirmed_client(clid);
2238 * We try hard to give out unique clientid's, so if we get an
2239 * attempt to confirm the same clientid with a different cred,
2240 * there's a bug somewhere. Let's charitably assume it's our
2243 status = nfserr_serverfault;
2244 if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
2246 if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
2248 /* cases below refer to rfc 3530 section 14.2.34: */
2249 if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
2250 if (conf && !unconf) /* case 2: probable retransmit */
2252 else /* case 4: client hasn't noticed we rebooted yet? */
2253 status = nfserr_stale_clientid;
2257 if (conf) { /* case 1: callback update */
2258 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2259 nfsd4_probe_callback(conf);
2260 expire_client(unconf);
2261 } else { /* case 3: normal case; new or rebooted client */
2262 unsigned int hash = clientstr_hashval(unconf->cl_recdir);
2264 conf = find_confirmed_client_by_str(unconf->cl_recdir, hash);
2266 nfsd4_client_record_remove(conf);
2267 expire_client(conf);
2269 move_to_confirmed(unconf);
2270 nfsd4_probe_callback(unconf);
2273 nfs4_unlock_state();
2277 static struct nfs4_file *nfsd4_alloc_file(void)
2279 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2282 /* OPEN Share state helper functions */
2283 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2285 unsigned int hashval = file_hashval(ino);
2287 atomic_set(&fp->fi_ref, 1);
2288 INIT_LIST_HEAD(&fp->fi_hash);
2289 INIT_LIST_HEAD(&fp->fi_stateids);
2290 INIT_LIST_HEAD(&fp->fi_delegations);
2291 fp->fi_inode = igrab(ino);
2292 fp->fi_had_conflict = false;
2293 fp->fi_lease = NULL;
2294 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2295 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2296 spin_lock(&recall_lock);
2297 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2298 spin_unlock(&recall_lock);
2302 nfsd4_free_slab(struct kmem_cache **slab)
2306 kmem_cache_destroy(*slab);
2311 nfsd4_free_slabs(void)
2313 nfsd4_free_slab(&openowner_slab);
2314 nfsd4_free_slab(&lockowner_slab);
2315 nfsd4_free_slab(&file_slab);
2316 nfsd4_free_slab(&stateid_slab);
2317 nfsd4_free_slab(&deleg_slab);
2321 nfsd4_init_slabs(void)
2323 openowner_slab = kmem_cache_create("nfsd4_openowners",
2324 sizeof(struct nfs4_openowner), 0, 0, NULL);
2325 if (openowner_slab == NULL)
2327 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2328 sizeof(struct nfs4_openowner), 0, 0, NULL);
2329 if (lockowner_slab == NULL)
2331 file_slab = kmem_cache_create("nfsd4_files",
2332 sizeof(struct nfs4_file), 0, 0, NULL);
2333 if (file_slab == NULL)
2335 stateid_slab = kmem_cache_create("nfsd4_stateids",
2336 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2337 if (stateid_slab == NULL)
2339 deleg_slab = kmem_cache_create("nfsd4_delegations",
2340 sizeof(struct nfs4_delegation), 0, 0, NULL);
2341 if (deleg_slab == NULL)
2346 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2350 void nfs4_free_openowner(struct nfs4_openowner *oo)
2352 kfree(oo->oo_owner.so_owner.data);
2353 kmem_cache_free(openowner_slab, oo);
2356 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2358 kfree(lo->lo_owner.so_owner.data);
2359 kmem_cache_free(lockowner_slab, lo);
2362 static void init_nfs4_replay(struct nfs4_replay *rp)
2364 rp->rp_status = nfserr_serverfault;
2366 rp->rp_buf = rp->rp_ibuf;
2369 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2371 struct nfs4_stateowner *sop;
2373 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2377 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2378 if (!sop->so_owner.data) {
2379 kmem_cache_free(slab, sop);
2382 sop->so_owner.len = owner->len;
2384 INIT_LIST_HEAD(&sop->so_stateids);
2385 sop->so_client = clp;
2386 init_nfs4_replay(&sop->so_replay);
2390 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2392 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2393 list_add(&oo->oo_perclient, &clp->cl_openowners);
2396 static struct nfs4_openowner *
2397 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2398 struct nfs4_openowner *oo;
2400 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2403 oo->oo_owner.so_is_open_owner = 1;
2404 oo->oo_owner.so_seqid = open->op_seqid;
2405 oo->oo_flags = NFS4_OO_NEW;
2407 oo->oo_last_closed_stid = NULL;
2408 INIT_LIST_HEAD(&oo->oo_close_lru);
2409 hash_openowner(oo, clp, strhashval);
2413 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2414 struct nfs4_openowner *oo = open->op_openowner;
2415 struct nfs4_client *clp = oo->oo_owner.so_client;
2417 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2418 INIT_LIST_HEAD(&stp->st_lockowners);
2419 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2420 list_add(&stp->st_perfile, &fp->fi_stateids);
2421 stp->st_stateowner = &oo->oo_owner;
2424 stp->st_access_bmap = 0;
2425 stp->st_deny_bmap = 0;
2426 set_access(open->op_share_access, stp);
2427 set_deny(open->op_share_deny, stp);
2428 stp->st_openstp = NULL;
2432 move_to_close_lru(struct nfs4_openowner *oo)
2434 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2436 list_move_tail(&oo->oo_close_lru, &close_lru);
2437 oo->oo_time = get_seconds();
2441 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2444 return (sop->so_owner.len == owner->len) &&
2445 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2446 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2449 static struct nfs4_openowner *
2450 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2452 struct nfs4_stateowner *so;
2453 struct nfs4_openowner *oo;
2455 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2456 if (!so->so_is_open_owner)
2458 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2460 renew_client(oo->oo_owner.so_client);
2467 /* search file_hashtbl[] for file */
2468 static struct nfs4_file *
2469 find_file(struct inode *ino)
2471 unsigned int hashval = file_hashval(ino);
2472 struct nfs4_file *fp;
2474 spin_lock(&recall_lock);
2475 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2476 if (fp->fi_inode == ino) {
2478 spin_unlock(&recall_lock);
2482 spin_unlock(&recall_lock);
2487 * Called to check deny when READ with all zero stateid or
2488 * WRITE with all zero or all one stateid
2491 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2493 struct inode *ino = current_fh->fh_dentry->d_inode;
2494 struct nfs4_file *fp;
2495 struct nfs4_ol_stateid *stp;
2498 dprintk("NFSD: nfs4_share_conflict\n");
2500 fp = find_file(ino);
2503 ret = nfserr_locked;
2504 /* Search for conflicting share reservations */
2505 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2506 if (test_deny(deny_type, stp) ||
2507 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2516 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2518 /* We're assuming the state code never drops its reference
2519 * without first removing the lease. Since we're in this lease
2520 * callback (and since the lease code is serialized by the kernel
2521 * lock) we know the server hasn't removed the lease yet, we know
2522 * it's safe to take a reference: */
2523 atomic_inc(&dp->dl_count);
2525 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2527 /* only place dl_time is set. protected by lock_flocks*/
2528 dp->dl_time = get_seconds();
2530 nfsd4_cb_recall(dp);
2533 /* Called from break_lease() with lock_flocks() held. */
2534 static void nfsd_break_deleg_cb(struct file_lock *fl)
2536 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2537 struct nfs4_delegation *dp;
2540 /* We assume break_lease is only called once per lease: */
2541 BUG_ON(fp->fi_had_conflict);
2543 * We don't want the locks code to timeout the lease for us;
2544 * we'll remove it ourself if a delegation isn't returned
2547 fl->fl_break_time = 0;
2549 spin_lock(&recall_lock);
2550 fp->fi_had_conflict = true;
2551 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2552 nfsd_break_one_deleg(dp);
2553 spin_unlock(&recall_lock);
2557 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2560 return lease_modify(onlist, arg);
2565 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2566 .lm_break = nfsd_break_deleg_cb,
2567 .lm_change = nfsd_change_deleg_cb,
2570 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2572 if (nfsd4_has_session(cstate))
2574 if (seqid == so->so_seqid - 1)
2575 return nfserr_replay_me;
2576 if (seqid == so->so_seqid)
2578 return nfserr_bad_seqid;
2582 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2583 struct nfsd4_open *open)
2585 clientid_t *clientid = &open->op_clientid;
2586 struct nfs4_client *clp = NULL;
2587 unsigned int strhashval;
2588 struct nfs4_openowner *oo = NULL;
2591 if (STALE_CLIENTID(&open->op_clientid))
2592 return nfserr_stale_clientid;
2594 * In case we need it later, after we've already created the
2595 * file and don't want to risk a further failure:
2597 open->op_file = nfsd4_alloc_file();
2598 if (open->op_file == NULL)
2599 return nfserr_jukebox;
2601 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2602 oo = find_openstateowner_str(strhashval, open);
2603 open->op_openowner = oo;
2605 clp = find_confirmed_client(clientid);
2607 return nfserr_expired;
2610 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2611 /* Replace unconfirmed owners without checking for replay. */
2612 clp = oo->oo_owner.so_client;
2613 release_openowner(oo);
2614 open->op_openowner = NULL;
2617 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2620 clp = oo->oo_owner.so_client;
2623 oo = alloc_init_open_stateowner(strhashval, clp, open);
2625 return nfserr_jukebox;
2626 open->op_openowner = oo;
2628 open->op_stp = nfs4_alloc_stateid(clp);
2630 return nfserr_jukebox;
2634 static inline __be32
2635 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2637 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2638 return nfserr_openmode;
2643 static int share_access_to_flags(u32 share_access)
2645 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2648 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2650 struct nfs4_stid *ret;
2652 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2655 return delegstateid(ret);
2658 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2660 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2661 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2665 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2666 struct nfs4_delegation **dp)
2669 __be32 status = nfserr_bad_stateid;
2671 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2674 flags = share_access_to_flags(open->op_share_access);
2675 status = nfs4_check_delegmode(*dp, flags);
2679 if (!nfsd4_is_deleg_cur(open))
2683 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2688 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2690 struct nfs4_ol_stateid *local;
2691 struct nfs4_openowner *oo = open->op_openowner;
2693 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2694 /* ignore lock owners */
2695 if (local->st_stateowner->so_is_open_owner == 0)
2697 /* remember if we have seen this open owner */
2698 if (local->st_stateowner == &oo->oo_owner)
2700 /* check for conflicting share reservations */
2701 if (!test_share(local, open))
2702 return nfserr_share_denied;
2707 static void nfs4_free_stateid(struct nfs4_ol_stateid *s)
2709 kmem_cache_free(stateid_slab, s);
2712 static inline int nfs4_access_to_access(u32 nfs4_access)
2716 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2717 flags |= NFSD_MAY_READ;
2718 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2719 flags |= NFSD_MAY_WRITE;
2723 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2724 struct svc_fh *cur_fh, struct nfsd4_open *open)
2727 int oflag = nfs4_access_to_omode(open->op_share_access);
2728 int access = nfs4_access_to_access(open->op_share_access);
2730 if (!fp->fi_fds[oflag]) {
2731 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2732 &fp->fi_fds[oflag]);
2736 nfs4_file_get_access(fp, oflag);
2741 static inline __be32
2742 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2743 struct nfsd4_open *open)
2745 struct iattr iattr = {
2746 .ia_valid = ATTR_SIZE,
2749 if (!open->op_truncate)
2751 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2752 return nfserr_inval;
2753 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2757 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)
2759 u32 op_share_access = open->op_share_access;
2763 new_access = !test_access(op_share_access, stp);
2765 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2769 status = nfsd4_truncate(rqstp, cur_fh, open);
2772 int oflag = nfs4_access_to_omode(op_share_access);
2773 nfs4_file_put_access(fp, oflag);
2777 /* remember the open */
2778 set_access(op_share_access, stp);
2779 set_deny(open->op_share_deny, stp);
2786 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2788 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2791 /* Should we give out recallable state?: */
2792 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2794 if (clp->cl_cb_state == NFSD4_CB_UP)
2797 * In the sessions case, since we don't have to establish a
2798 * separate connection for callbacks, we assume it's OK
2799 * until we hear otherwise:
2801 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2804 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2806 struct file_lock *fl;
2808 fl = locks_alloc_lock();
2811 locks_init_lock(fl);
2812 fl->fl_lmops = &nfsd_lease_mng_ops;
2813 fl->fl_flags = FL_LEASE;
2814 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2815 fl->fl_end = OFFSET_MAX;
2816 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2817 fl->fl_pid = current->tgid;
2821 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2823 struct nfs4_file *fp = dp->dl_file;
2824 struct file_lock *fl;
2827 fl = nfs4_alloc_init_lease(dp, flag);
2830 fl->fl_file = find_readable_file(fp);
2831 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2832 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2834 list_del_init(&dp->dl_perclnt);
2835 locks_free_lock(fl);
2839 fp->fi_deleg_file = fl->fl_file;
2840 get_file(fp->fi_deleg_file);
2841 atomic_set(&fp->fi_delegees, 1);
2842 list_add(&dp->dl_perfile, &fp->fi_delegations);
2846 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2848 struct nfs4_file *fp = dp->dl_file;
2851 return nfs4_setlease(dp, flag);
2852 spin_lock(&recall_lock);
2853 if (fp->fi_had_conflict) {
2854 spin_unlock(&recall_lock);
2857 atomic_inc(&fp->fi_delegees);
2858 list_add(&dp->dl_perfile, &fp->fi_delegations);
2859 spin_unlock(&recall_lock);
2860 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2864 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2866 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2867 if (status == -EAGAIN)
2868 open->op_why_no_deleg = WND4_CONTENTION;
2870 open->op_why_no_deleg = WND4_RESOURCE;
2871 switch (open->op_deleg_want) {
2872 case NFS4_SHARE_WANT_READ_DELEG:
2873 case NFS4_SHARE_WANT_WRITE_DELEG:
2874 case NFS4_SHARE_WANT_ANY_DELEG:
2876 case NFS4_SHARE_WANT_CANCEL:
2877 open->op_why_no_deleg = WND4_CANCELLED;
2879 case NFS4_SHARE_WANT_NO_DELEG:
2880 BUG(); /* not supposed to get here */
2886 * Attempt to hand out a delegation.
2889 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2891 struct nfs4_delegation *dp;
2892 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2894 int status = 0, flag = 0;
2896 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2897 flag = NFS4_OPEN_DELEGATE_NONE;
2898 open->op_recall = 0;
2899 switch (open->op_claim_type) {
2900 case NFS4_OPEN_CLAIM_PREVIOUS:
2902 open->op_recall = 1;
2903 flag = open->op_delegate_type;
2904 if (flag == NFS4_OPEN_DELEGATE_NONE)
2907 case NFS4_OPEN_CLAIM_NULL:
2908 /* Let's not give out any delegations till everyone's
2909 * had the chance to reclaim theirs.... */
2910 if (locks_in_grace())
2912 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2914 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2915 flag = NFS4_OPEN_DELEGATE_WRITE;
2917 flag = NFS4_OPEN_DELEGATE_READ;
2923 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2926 status = nfs4_set_delegation(dp, flag);
2930 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2932 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2933 STATEID_VAL(&dp->dl_stid.sc_stateid));
2935 open->op_delegate_type = flag;
2936 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2937 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2938 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2939 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2941 /* 4.1 client asking for a delegation? */
2942 if (open->op_deleg_want)
2943 nfsd4_open_deleg_none_ext(open, status);
2947 nfs4_put_delegation(dp);
2949 flag = NFS4_OPEN_DELEGATE_NONE;
2953 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2954 struct nfs4_delegation *dp)
2956 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2957 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2958 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2959 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2960 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2961 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2962 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2963 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
2965 /* Otherwise the client must be confused wanting a delegation
2966 * it already has, therefore we don't return
2967 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
2972 * called with nfs4_lock_state() held.
2975 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2977 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2978 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
2979 struct nfs4_file *fp = NULL;
2980 struct inode *ino = current_fh->fh_dentry->d_inode;
2981 struct nfs4_ol_stateid *stp = NULL;
2982 struct nfs4_delegation *dp = NULL;
2986 * Lookup file; if found, lookup stateid and check open request,
2987 * and check for delegations in the process of being recalled.
2988 * If not found, create the nfs4_file struct
2990 fp = find_file(ino);
2992 if ((status = nfs4_check_open(fp, open, &stp)))
2994 status = nfs4_check_deleg(cl, fp, open, &dp);
2998 status = nfserr_bad_stateid;
2999 if (nfsd4_is_deleg_cur(open))
3001 status = nfserr_jukebox;
3003 open->op_file = NULL;
3004 nfsd4_init_file(fp, ino);
3008 * OPEN the file, or upgrade an existing OPEN.
3009 * If truncate fails, the OPEN fails.
3012 /* Stateid was found, this is an OPEN upgrade */
3013 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3017 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3021 open->op_stp = NULL;
3022 init_open_stateid(stp, fp, open);
3023 status = nfsd4_truncate(rqstp, current_fh, open);
3025 release_open_stateid(stp);
3029 update_stateid(&stp->st_stid.sc_stateid);
3030 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3032 if (nfsd4_has_session(&resp->cstate)) {
3033 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3035 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3036 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3037 open->op_why_no_deleg = WND4_NOT_WANTED;
3043 * Attempt to hand out a delegation. No error return, because the
3044 * OPEN succeeds even if we fail.
3046 nfs4_open_delegation(current_fh, open, stp);
3050 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3051 STATEID_VAL(&stp->st_stid.sc_stateid));
3053 /* 4.1 client trying to upgrade/downgrade delegation? */
3054 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3055 open->op_deleg_want)
3056 nfsd4_deleg_xgrade_none_ext(open, dp);
3060 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3061 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3063 * To finish the open response, we just need to set the rflags.
3065 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3066 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3067 !nfsd4_has_session(&resp->cstate))
3068 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3073 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3075 if (open->op_openowner) {
3076 struct nfs4_openowner *oo = open->op_openowner;
3078 if (!list_empty(&oo->oo_owner.so_stateids))
3079 list_del_init(&oo->oo_close_lru);
3080 if (oo->oo_flags & NFS4_OO_NEW) {
3082 release_openowner(oo);
3083 open->op_openowner = NULL;
3085 oo->oo_flags &= ~NFS4_OO_NEW;
3089 nfsd4_free_file(open->op_file);
3091 nfs4_free_stateid(open->op_stp);
3095 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3098 struct nfs4_client *clp;
3102 dprintk("process_renew(%08x/%08x): starting\n",
3103 clid->cl_boot, clid->cl_id);
3104 status = nfserr_stale_clientid;
3105 if (STALE_CLIENTID(clid))
3107 clp = find_confirmed_client(clid);
3108 status = nfserr_expired;
3110 /* We assume the client took too long to RENEW. */
3111 dprintk("nfsd4_renew: clientid not found!\n");
3114 status = nfserr_cb_path_down;
3115 if (!list_empty(&clp->cl_delegations)
3116 && clp->cl_cb_state != NFSD4_CB_UP)
3120 nfs4_unlock_state();
3124 static struct lock_manager nfsd4_manager = {
3127 static bool grace_ended;
3130 nfsd4_end_grace(void)
3132 /* do nothing if grace period already ended */
3136 dprintk("NFSD: end of grace period\n");
3138 nfsd4_record_grace_done(&init_net, boot_time);
3139 locks_end_grace(&nfsd4_manager);
3141 * Now that every NFSv4 client has had the chance to recover and
3142 * to see the (possibly new, possibly shorter) lease time, we
3143 * can safely set the next grace time to the current lease time:
3145 nfsd4_grace = nfsd4_lease;
3149 nfs4_laundromat(void)
3151 struct nfs4_client *clp;
3152 struct nfs4_openowner *oo;
3153 struct nfs4_delegation *dp;
3154 struct list_head *pos, *next, reaplist;
3155 time_t cutoff = get_seconds() - nfsd4_lease;
3156 time_t t, clientid_val = nfsd4_lease;
3157 time_t u, test_val = nfsd4_lease;
3161 dprintk("NFSD: laundromat service - starting\n");
3163 INIT_LIST_HEAD(&reaplist);
3164 spin_lock(&client_lock);
3165 list_for_each_safe(pos, next, &client_lru) {
3166 clp = list_entry(pos, struct nfs4_client, cl_lru);
3167 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3168 t = clp->cl_time - cutoff;
3169 if (clientid_val > t)
3173 if (atomic_read(&clp->cl_refcount)) {
3174 dprintk("NFSD: client in use (clientid %08x)\n",
3175 clp->cl_clientid.cl_id);
3178 unhash_client_locked(clp);
3179 list_add(&clp->cl_lru, &reaplist);
3181 spin_unlock(&client_lock);
3182 list_for_each_safe(pos, next, &reaplist) {
3183 clp = list_entry(pos, struct nfs4_client, cl_lru);
3184 dprintk("NFSD: purging unused client (clientid %08x)\n",
3185 clp->cl_clientid.cl_id);
3186 nfsd4_client_record_remove(clp);
3189 spin_lock(&recall_lock);
3190 list_for_each_safe(pos, next, &del_recall_lru) {
3191 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3192 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3193 u = dp->dl_time - cutoff;
3198 list_move(&dp->dl_recall_lru, &reaplist);
3200 spin_unlock(&recall_lock);
3201 list_for_each_safe(pos, next, &reaplist) {
3202 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3203 unhash_delegation(dp);
3205 test_val = nfsd4_lease;
3206 list_for_each_safe(pos, next, &close_lru) {
3207 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3208 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3209 u = oo->oo_time - cutoff;
3214 release_openowner(oo);
3216 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3217 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3218 nfs4_unlock_state();
3219 return clientid_val;
3222 static struct workqueue_struct *laundry_wq;
3223 static void laundromat_main(struct work_struct *);
3224 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3227 laundromat_main(struct work_struct *not_used)
3231 t = nfs4_laundromat();
3232 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3233 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3236 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3238 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3239 return nfserr_bad_stateid;
3244 STALE_STATEID(stateid_t *stateid)
3246 if (stateid->si_opaque.so_clid.cl_boot == boot_time)
3248 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3249 STATEID_VAL(stateid));
3254 access_permit_read(struct nfs4_ol_stateid *stp)
3256 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3257 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3258 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3262 access_permit_write(struct nfs4_ol_stateid *stp)
3264 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3265 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3269 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3271 __be32 status = nfserr_openmode;
3273 /* For lock stateid's, we test the parent open, not the lock: */
3274 if (stp->st_openstp)
3275 stp = stp->st_openstp;
3276 if ((flags & WR_STATE) && !access_permit_write(stp))
3278 if ((flags & RD_STATE) && !access_permit_read(stp))
3285 static inline __be32
3286 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3288 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3290 else if (locks_in_grace()) {
3291 /* Answer in remaining cases depends on existence of
3292 * conflicting state; so we must wait out the grace period. */
3293 return nfserr_grace;
3294 } else if (flags & WR_STATE)
3295 return nfs4_share_conflict(current_fh,
3296 NFS4_SHARE_DENY_WRITE);
3297 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3298 return nfs4_share_conflict(current_fh,
3299 NFS4_SHARE_DENY_READ);
3303 * Allow READ/WRITE during grace period on recovered state only for files
3304 * that are not able to provide mandatory locking.
3307 grace_disallows_io(struct inode *inode)
3309 return locks_in_grace() && mandatory_lock(inode);
3312 /* Returns true iff a is later than b: */
3313 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3315 return (s32)a->si_generation - (s32)b->si_generation > 0;
3318 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3321 * When sessions are used the stateid generation number is ignored
3324 if (has_session && in->si_generation == 0)
3327 if (in->si_generation == ref->si_generation)
3330 /* If the client sends us a stateid from the future, it's buggy: */
3331 if (stateid_generation_after(in, ref))
3332 return nfserr_bad_stateid;
3334 * However, we could see a stateid from the past, even from a
3335 * non-buggy client. For example, if the client sends a lock
3336 * while some IO is outstanding, the lock may bump si_generation
3337 * while the IO is still in flight. The client could avoid that
3338 * situation by waiting for responses on all the IO requests,
3339 * but better performance may result in retrying IO that
3340 * receives an old_stateid error if requests are rarely
3341 * reordered in flight:
3343 return nfserr_old_stateid;
3346 __be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3348 struct nfs4_stid *s;
3349 struct nfs4_ol_stateid *ols;
3352 if (STALE_STATEID(stateid))
3353 return nfserr_stale_stateid;
3355 s = find_stateid(cl, stateid);
3357 return nfserr_stale_stateid;
3358 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3361 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3363 ols = openlockstateid(s);
3364 if (ols->st_stateowner->so_is_open_owner
3365 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3366 return nfserr_bad_stateid;
3370 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
3372 struct nfs4_client *cl;
3374 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3375 return nfserr_bad_stateid;
3376 if (STALE_STATEID(stateid))
3377 return nfserr_stale_stateid;
3378 cl = find_confirmed_client(&stateid->si_opaque.so_clid);
3380 return nfserr_expired;
3381 *s = find_stateid_by_type(cl, stateid, typemask);
3383 return nfserr_bad_stateid;
3389 * Checks for stateid operations
3392 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3393 stateid_t *stateid, int flags, struct file **filpp)
3395 struct nfs4_stid *s;
3396 struct nfs4_ol_stateid *stp = NULL;
3397 struct nfs4_delegation *dp = NULL;
3398 struct svc_fh *current_fh = &cstate->current_fh;
3399 struct inode *ino = current_fh->fh_dentry->d_inode;
3405 if (grace_disallows_io(ino))
3406 return nfserr_grace;
3408 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3409 return check_special_stateids(current_fh, stateid, flags);
3411 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
3414 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3417 switch (s->sc_type) {
3418 case NFS4_DELEG_STID:
3419 dp = delegstateid(s);
3420 status = nfs4_check_delegmode(dp, flags);
3424 *filpp = dp->dl_file->fi_deleg_file;
3428 case NFS4_OPEN_STID:
3429 case NFS4_LOCK_STID:
3430 stp = openlockstateid(s);
3431 status = nfs4_check_fh(current_fh, stp);
3434 if (stp->st_stateowner->so_is_open_owner
3435 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3437 status = nfs4_check_openmode(stp, flags);
3441 if (flags & RD_STATE)
3442 *filpp = find_readable_file(stp->st_file);
3444 *filpp = find_writeable_file(stp->st_file);
3448 return nfserr_bad_stateid;
3456 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3458 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3459 return nfserr_locks_held;
3460 release_lock_stateid(stp);
3465 * Test if the stateid is valid
3468 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3469 struct nfsd4_test_stateid *test_stateid)
3471 struct nfsd4_test_stateid_id *stateid;
3472 struct nfs4_client *cl = cstate->session->se_client;
3475 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3476 stateid->ts_id_status = nfs4_validate_stateid(cl, &stateid->ts_id_stateid);
3477 nfs4_unlock_state();
3483 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3484 struct nfsd4_free_stateid *free_stateid)
3486 stateid_t *stateid = &free_stateid->fr_stateid;
3487 struct nfs4_stid *s;
3488 struct nfs4_client *cl = cstate->session->se_client;
3489 __be32 ret = nfserr_bad_stateid;
3492 s = find_stateid(cl, stateid);
3495 switch (s->sc_type) {
3496 case NFS4_DELEG_STID:
3497 ret = nfserr_locks_held;
3499 case NFS4_OPEN_STID:
3500 case NFS4_LOCK_STID:
3501 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3504 if (s->sc_type == NFS4_LOCK_STID)
3505 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3507 ret = nfserr_locks_held;
3510 ret = nfserr_bad_stateid;
3513 nfs4_unlock_state();
3520 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3521 RD_STATE : WR_STATE;
3524 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3526 struct svc_fh *current_fh = &cstate->current_fh;
3527 struct nfs4_stateowner *sop = stp->st_stateowner;
3530 status = nfsd4_check_seqid(cstate, sop, seqid);
3533 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3535 * "Closed" stateid's exist *only* to return
3536 * nfserr_replay_me from the previous step.
3538 return nfserr_bad_stateid;
3539 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3542 return nfs4_check_fh(current_fh, stp);
3546 * Checks for sequence id mutating operations.
3549 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3550 stateid_t *stateid, char typemask,
3551 struct nfs4_ol_stateid **stpp)
3554 struct nfs4_stid *s;
3556 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3557 seqid, STATEID_VAL(stateid));
3560 status = nfsd4_lookup_stateid(stateid, typemask, &s);
3563 *stpp = openlockstateid(s);
3564 cstate->replay_owner = (*stpp)->st_stateowner;
3566 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3569 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3572 struct nfs4_openowner *oo;
3574 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3575 NFS4_OPEN_STID, stpp);
3578 oo = openowner((*stpp)->st_stateowner);
3579 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3580 return nfserr_bad_stateid;
3585 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3586 struct nfsd4_open_confirm *oc)
3589 struct nfs4_openowner *oo;
3590 struct nfs4_ol_stateid *stp;
3592 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3593 (int)cstate->current_fh.fh_dentry->d_name.len,
3594 cstate->current_fh.fh_dentry->d_name.name);
3596 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3602 status = nfs4_preprocess_seqid_op(cstate,
3603 oc->oc_seqid, &oc->oc_req_stateid,
3604 NFS4_OPEN_STID, &stp);
3607 oo = openowner(stp->st_stateowner);
3608 status = nfserr_bad_stateid;
3609 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3611 oo->oo_flags |= NFS4_OO_CONFIRMED;
3612 update_stateid(&stp->st_stid.sc_stateid);
3613 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3614 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3615 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3617 nfsd4_client_record_create(oo->oo_owner.so_client);
3620 if (!cstate->replay_owner)
3621 nfs4_unlock_state();
3625 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3627 if (!test_access(access, stp))
3629 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3630 clear_access(access, stp);
3633 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3635 switch (to_access) {
3636 case NFS4_SHARE_ACCESS_READ:
3637 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3638 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3640 case NFS4_SHARE_ACCESS_WRITE:
3641 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3642 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3644 case NFS4_SHARE_ACCESS_BOTH:
3652 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3655 for (i = 0; i < 4; i++) {
3656 if ((i & deny) != i)
3662 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3663 struct nfsd4_compound_state *cstate,
3664 struct nfsd4_open_downgrade *od)
3667 struct nfs4_ol_stateid *stp;
3669 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3670 (int)cstate->current_fh.fh_dentry->d_name.len,
3671 cstate->current_fh.fh_dentry->d_name.name);
3673 /* We don't yet support WANT bits: */
3674 if (od->od_deleg_want)
3675 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3679 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3680 &od->od_stateid, &stp);
3683 status = nfserr_inval;
3684 if (!test_access(od->od_share_access, stp)) {
3685 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3686 stp->st_access_bmap, od->od_share_access);
3689 if (!test_deny(od->od_share_deny, stp)) {
3690 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3691 stp->st_deny_bmap, od->od_share_deny);
3694 nfs4_stateid_downgrade(stp, od->od_share_access);
3696 reset_union_bmap_deny(od->od_share_deny, stp);
3698 update_stateid(&stp->st_stid.sc_stateid);
3699 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3702 if (!cstate->replay_owner)
3703 nfs4_unlock_state();
3707 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3709 struct nfs4_openowner *oo;
3710 struct nfs4_ol_stateid *s;
3712 if (!so->so_is_open_owner)
3715 s = oo->oo_last_closed_stid;
3718 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3719 /* Release the last_closed_stid on the next seqid bump: */
3720 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3723 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3724 release_last_closed_stateid(oo);
3727 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3729 unhash_open_stateid(s);
3730 s->st_stid.sc_type = NFS4_CLOSED_STID;
3734 * nfs4_unlock_state() called after encode
3737 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3738 struct nfsd4_close *close)
3741 struct nfs4_openowner *oo;
3742 struct nfs4_ol_stateid *stp;
3744 dprintk("NFSD: nfsd4_close on file %.*s\n",
3745 (int)cstate->current_fh.fh_dentry->d_name.len,
3746 cstate->current_fh.fh_dentry->d_name.name);
3749 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3751 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3755 oo = openowner(stp->st_stateowner);
3757 update_stateid(&stp->st_stid.sc_stateid);
3758 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3760 nfsd4_close_open_stateid(stp);
3761 oo->oo_last_closed_stid = stp;
3763 /* place unused nfs4_stateowners on so_close_lru list to be
3764 * released by the laundromat service after the lease period
3765 * to enable us to handle CLOSE replay
3767 if (list_empty(&oo->oo_owner.so_stateids))
3768 move_to_close_lru(oo);
3770 if (!cstate->replay_owner)
3771 nfs4_unlock_state();
3776 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3777 struct nfsd4_delegreturn *dr)
3779 struct nfs4_delegation *dp;
3780 stateid_t *stateid = &dr->dr_stateid;
3781 struct nfs4_stid *s;
3782 struct inode *inode;
3785 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3787 inode = cstate->current_fh.fh_dentry->d_inode;
3790 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s);
3793 dp = delegstateid(s);
3794 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3798 unhash_delegation(dp);
3800 nfs4_unlock_state();
3806 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3808 #define LOCKOWNER_INO_HASH_BITS 8
3809 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3810 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3813 end_offset(u64 start, u64 len)
3818 return end >= start ? end: NFS4_MAX_UINT64;
3821 /* last octet in a range */
3823 last_byte_offset(u64 start, u64 len)
3829 return end > start ? end - 1: NFS4_MAX_UINT64;
3832 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3834 return (file_hashval(inode) + cl_id
3835 + opaque_hashval(ownername->data, ownername->len))
3836 & LOCKOWNER_INO_HASH_MASK;
3839 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3842 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3843 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3844 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3845 * locking, this prevents us from being completely protocol-compliant. The
3846 * real solution to this problem is to start using unsigned file offsets in
3847 * the VFS, but this is a very deep change!
3850 nfs4_transform_lock_offset(struct file_lock *lock)
3852 if (lock->fl_start < 0)
3853 lock->fl_start = OFFSET_MAX;
3854 if (lock->fl_end < 0)
3855 lock->fl_end = OFFSET_MAX;
3858 /* Hack!: For now, we're defining this just so we can use a pointer to it
3859 * as a unique cookie to identify our (NFSv4's) posix locks. */
3860 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3864 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3866 struct nfs4_lockowner *lo;
3868 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3869 lo = (struct nfs4_lockowner *) fl->fl_owner;
3870 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3871 lo->lo_owner.so_owner.len, GFP_KERNEL);
3872 if (!deny->ld_owner.data)
3873 /* We just don't care that much */
3875 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3876 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3879 deny->ld_owner.len = 0;
3880 deny->ld_owner.data = NULL;
3881 deny->ld_clientid.cl_boot = 0;
3882 deny->ld_clientid.cl_id = 0;
3884 deny->ld_start = fl->fl_start;
3885 deny->ld_length = NFS4_MAX_UINT64;
3886 if (fl->fl_end != NFS4_MAX_UINT64)
3887 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3888 deny->ld_type = NFS4_READ_LT;
3889 if (fl->fl_type != F_RDLCK)
3890 deny->ld_type = NFS4_WRITE_LT;
3893 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3895 struct nfs4_ol_stateid *lst;
3897 if (!same_owner_str(&lo->lo_owner, owner, clid))
3899 lst = list_first_entry(&lo->lo_owner.so_stateids,
3900 struct nfs4_ol_stateid, st_perstateowner);
3901 return lst->st_file->fi_inode == inode;
3904 static struct nfs4_lockowner *
3905 find_lockowner_str(struct inode *inode, clientid_t *clid,
3906 struct xdr_netobj *owner)
3908 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3909 struct nfs4_lockowner *lo;
3911 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3912 if (same_lockowner_ino(lo, inode, clid, owner))
3918 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3920 struct inode *inode = open_stp->st_file->fi_inode;
3921 unsigned int inohash = lockowner_ino_hashval(inode,
3922 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3924 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3925 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3926 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3930 * Alloc a lock owner structure.
3931 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3934 * strhashval = ownerstr_hashval
3937 static struct nfs4_lockowner *
3938 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3939 struct nfs4_lockowner *lo;
3941 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3944 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3945 lo->lo_owner.so_is_open_owner = 0;
3946 /* It is the openowner seqid that will be incremented in encode in the
3947 * case of new lockowners; so increment the lock seqid manually: */
3948 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3949 hash_lockowner(lo, strhashval, clp, open_stp);
3953 static struct nfs4_ol_stateid *
3954 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3956 struct nfs4_ol_stateid *stp;
3957 struct nfs4_client *clp = lo->lo_owner.so_client;
3959 stp = nfs4_alloc_stateid(clp);
3962 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3963 list_add(&stp->st_perfile, &fp->fi_stateids);
3964 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3965 stp->st_stateowner = &lo->lo_owner;
3968 stp->st_access_bmap = 0;
3969 stp->st_deny_bmap = open_stp->st_deny_bmap;
3970 stp->st_openstp = open_stp;
3975 check_lock_length(u64 offset, u64 length)
3977 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
3978 LOFF_OVERFLOW(offset, length)));
3981 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
3983 struct nfs4_file *fp = lock_stp->st_file;
3984 int oflag = nfs4_access_to_omode(access);
3986 if (test_access(access, lock_stp))
3988 nfs4_file_get_access(fp, oflag);
3989 set_access(access, lock_stp);
3992 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)
3994 struct nfs4_file *fi = ost->st_file;
3995 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
3996 struct nfs4_client *cl = oo->oo_owner.so_client;
3997 struct nfs4_lockowner *lo;
3998 unsigned int strhashval;
4000 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4002 if (!cstate->minorversion)
4003 return nfserr_bad_seqid;
4004 /* XXX: a lockowner always has exactly one stateid: */
4005 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4006 struct nfs4_ol_stateid, st_perstateowner);
4009 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4010 &lock->v.new.owner);
4011 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4013 return nfserr_jukebox;
4014 *lst = alloc_init_lock_stateid(lo, fi, ost);
4016 release_lockowner(lo);
4017 return nfserr_jukebox;
4027 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4028 struct nfsd4_lock *lock)
4030 struct nfs4_openowner *open_sop = NULL;
4031 struct nfs4_lockowner *lock_sop = NULL;
4032 struct nfs4_ol_stateid *lock_stp;
4033 struct file *filp = NULL;
4034 struct file_lock file_lock;
4035 struct file_lock conflock;
4037 bool new_state = false;
4041 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4042 (long long) lock->lk_offset,
4043 (long long) lock->lk_length);
4045 if (check_lock_length(lock->lk_offset, lock->lk_length))
4046 return nfserr_inval;
4048 if ((status = fh_verify(rqstp, &cstate->current_fh,
4049 S_IFREG, NFSD_MAY_LOCK))) {
4050 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4056 if (lock->lk_is_new) {
4058 * Client indicates that this is a new lockowner.
4059 * Use open owner and open stateid to create lock owner and
4062 struct nfs4_ol_stateid *open_stp = NULL;
4064 if (nfsd4_has_session(cstate))
4065 /* See rfc 5661 18.10.3: given clientid is ignored: */
4066 memcpy(&lock->v.new.clientid,
4067 &cstate->session->se_client->cl_clientid,
4068 sizeof(clientid_t));
4070 status = nfserr_stale_clientid;
4071 if (STALE_CLIENTID(&lock->lk_new_clientid))
4074 /* validate and update open stateid and open seqid */
4075 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4076 lock->lk_new_open_seqid,
4077 &lock->lk_new_open_stateid,
4081 open_sop = openowner(open_stp->st_stateowner);
4082 status = nfserr_bad_stateid;
4083 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4084 &lock->v.new.clientid))
4086 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4087 &lock_stp, &new_state);
4091 /* lock (lock owner + lock stateid) already exists */
4092 status = nfs4_preprocess_seqid_op(cstate,
4093 lock->lk_old_lock_seqid,
4094 &lock->lk_old_lock_stateid,
4095 NFS4_LOCK_STID, &lock_stp);
4099 lock_sop = lockowner(lock_stp->st_stateowner);
4101 lkflg = setlkflg(lock->lk_type);
4102 status = nfs4_check_openmode(lock_stp, lkflg);
4106 status = nfserr_grace;
4107 if (locks_in_grace() && !lock->lk_reclaim)
4109 status = nfserr_no_grace;
4110 if (!locks_in_grace() && lock->lk_reclaim)
4113 locks_init_lock(&file_lock);
4114 switch (lock->lk_type) {
4117 filp = find_readable_file(lock_stp->st_file);
4119 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4120 file_lock.fl_type = F_RDLCK;
4123 case NFS4_WRITEW_LT:
4124 filp = find_writeable_file(lock_stp->st_file);
4126 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4127 file_lock.fl_type = F_WRLCK;
4130 status = nfserr_inval;
4134 status = nfserr_openmode;
4137 file_lock.fl_owner = (fl_owner_t)lock_sop;
4138 file_lock.fl_pid = current->tgid;
4139 file_lock.fl_file = filp;
4140 file_lock.fl_flags = FL_POSIX;
4141 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4143 file_lock.fl_start = lock->lk_offset;
4144 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4145 nfs4_transform_lock_offset(&file_lock);
4148 * Try to lock the file in the VFS.
4149 * Note: locks.c uses the BKL to protect the inode's lock list.
4152 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4154 case 0: /* success! */
4155 update_stateid(&lock_stp->st_stid.sc_stateid);
4156 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4160 case (EAGAIN): /* conflock holds conflicting lock */
4161 status = nfserr_denied;
4162 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4163 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4166 status = nfserr_deadlock;
4169 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4170 status = nfserrno(err);
4174 if (status && new_state)
4175 release_lockowner(lock_sop);
4176 if (!cstate->replay_owner)
4177 nfs4_unlock_state();
4182 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4183 * so we do a temporary open here just to get an open file to pass to
4184 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4187 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4192 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4195 err = vfs_test_lock(file, lock);
4204 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4205 struct nfsd4_lockt *lockt)
4207 struct inode *inode;
4208 struct file_lock file_lock;
4209 struct nfs4_lockowner *lo;
4213 if (locks_in_grace())
4214 return nfserr_grace;
4216 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4217 return nfserr_inval;
4221 status = nfserr_stale_clientid;
4222 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4225 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4228 inode = cstate->current_fh.fh_dentry->d_inode;
4229 locks_init_lock(&file_lock);
4230 switch (lockt->lt_type) {
4233 file_lock.fl_type = F_RDLCK;
4236 case NFS4_WRITEW_LT:
4237 file_lock.fl_type = F_WRLCK;
4240 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4241 status = nfserr_inval;
4245 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4247 file_lock.fl_owner = (fl_owner_t)lo;
4248 file_lock.fl_pid = current->tgid;
4249 file_lock.fl_flags = FL_POSIX;
4251 file_lock.fl_start = lockt->lt_offset;
4252 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4254 nfs4_transform_lock_offset(&file_lock);
4257 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4259 status = nfserrno(error);
4262 if (file_lock.fl_type != F_UNLCK) {
4263 status = nfserr_denied;
4264 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4267 nfs4_unlock_state();
4272 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4273 struct nfsd4_locku *locku)
4275 struct nfs4_ol_stateid *stp;
4276 struct file *filp = NULL;
4277 struct file_lock file_lock;
4281 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4282 (long long) locku->lu_offset,
4283 (long long) locku->lu_length);
4285 if (check_lock_length(locku->lu_offset, locku->lu_length))
4286 return nfserr_inval;
4290 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4291 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4294 filp = find_any_file(stp->st_file);
4296 status = nfserr_lock_range;
4300 locks_init_lock(&file_lock);
4301 file_lock.fl_type = F_UNLCK;
4302 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4303 file_lock.fl_pid = current->tgid;
4304 file_lock.fl_file = filp;
4305 file_lock.fl_flags = FL_POSIX;
4306 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4307 file_lock.fl_start = locku->lu_offset;
4309 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4310 nfs4_transform_lock_offset(&file_lock);
4313 * Try to unlock the file in the VFS.
4315 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4317 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4321 * OK, unlock succeeded; the only thing left to do is update the stateid.
4323 update_stateid(&stp->st_stid.sc_stateid);
4324 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4327 if (!cstate->replay_owner)
4328 nfs4_unlock_state();
4332 status = nfserrno(err);
4338 * 1: locks held by lockowner
4339 * 0: no locks held by lockowner
4342 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4344 struct file_lock **flpp;
4345 struct inode *inode = filp->fi_inode;
4349 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4350 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4361 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4362 struct nfsd4_compound_state *cstate,
4363 struct nfsd4_release_lockowner *rlockowner)
4365 clientid_t *clid = &rlockowner->rl_clientid;
4366 struct nfs4_stateowner *sop;
4367 struct nfs4_lockowner *lo;
4368 struct nfs4_ol_stateid *stp;
4369 struct xdr_netobj *owner = &rlockowner->rl_owner;
4370 struct list_head matches;
4371 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4374 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4375 clid->cl_boot, clid->cl_id);
4377 /* XXX check for lease expiration */
4379 status = nfserr_stale_clientid;
4380 if (STALE_CLIENTID(clid))
4385 status = nfserr_locks_held;
4386 INIT_LIST_HEAD(&matches);
4388 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4389 if (sop->so_is_open_owner)
4391 if (!same_owner_str(sop, owner, clid))
4393 list_for_each_entry(stp, &sop->so_stateids,
4395 lo = lockowner(sop);
4396 if (check_for_locks(stp->st_file, lo))
4398 list_add(&lo->lo_list, &matches);
4401 /* Clients probably won't expect us to return with some (but not all)
4402 * of the lockowner state released; so don't release any until all
4403 * have been checked. */
4405 while (!list_empty(&matches)) {
4406 lo = list_entry(matches.next, struct nfs4_lockowner,
4408 /* unhash_stateowner deletes so_perclient only
4409 * for openowners. */
4410 list_del(&lo->lo_list);
4411 release_lockowner(lo);
4414 nfs4_unlock_state();
4418 static inline struct nfs4_client_reclaim *
4421 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4425 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4427 unsigned int strhashval = clientstr_hashval(name);
4428 struct nfs4_client *clp;
4430 clp = find_confirmed_client_by_str(name, strhashval);
4433 return test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
4437 * failure => all reset bets are off, nfserr_no_grace...
4440 nfs4_client_to_reclaim(const char *name)
4442 unsigned int strhashval;
4443 struct nfs4_client_reclaim *crp = NULL;
4445 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4446 crp = alloc_reclaim();
4449 strhashval = clientstr_hashval(name);
4450 INIT_LIST_HEAD(&crp->cr_strhash);
4451 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4452 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4453 reclaim_str_hashtbl_size++;
4458 nfs4_release_reclaim(void)
4460 struct nfs4_client_reclaim *crp = NULL;
4463 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4464 while (!list_empty(&reclaim_str_hashtbl[i])) {
4465 crp = list_entry(reclaim_str_hashtbl[i].next,
4466 struct nfs4_client_reclaim, cr_strhash);
4467 list_del(&crp->cr_strhash);
4469 reclaim_str_hashtbl_size--;
4472 BUG_ON(reclaim_str_hashtbl_size);
4476 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4477 struct nfs4_client_reclaim *
4478 nfsd4_find_reclaim_client(struct nfs4_client *clp)
4480 unsigned int strhashval;
4481 struct nfs4_client_reclaim *crp = NULL;
4483 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4484 clp->cl_name.len, clp->cl_name.data,
4487 /* find clp->cl_name in reclaim_str_hashtbl */
4488 strhashval = clientstr_hashval(clp->cl_recdir);
4489 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4490 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4498 * Called from OPEN. Look for clientid in reclaim list.
4501 nfs4_check_open_reclaim(clientid_t *clid)
4503 struct nfs4_client *clp;
4505 /* find clientid in conf_id_hashtbl */
4506 clp = find_confirmed_client(clid);
4508 return nfserr_reclaim_bad;
4510 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4513 #ifdef CONFIG_NFSD_FAULT_INJECTION
4515 void nfsd_forget_clients(u64 num)
4517 struct nfs4_client *clp, *next;
4521 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4522 nfsd4_client_record_remove(clp);
4527 nfs4_unlock_state();
4529 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4532 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4534 release_lockowner(lockowner(sop));
4537 static void release_openowner_sop(struct nfs4_stateowner *sop)
4539 release_openowner(openowner(sop));
4542 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4543 void (*release_sop)(struct nfs4_stateowner *))
4546 struct nfs4_stateowner *sop, *next;
4548 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4549 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4550 if (sop->so_is_open_owner != is_open_owner)
4560 void nfsd_forget_locks(u64 num)
4565 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4566 nfs4_unlock_state();
4568 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4571 void nfsd_forget_openowners(u64 num)
4576 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4577 nfs4_unlock_state();
4579 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4582 int nfsd_process_n_delegations(u64 num, void (*deleg_func)(struct nfs4_delegation *))
4585 struct nfs4_file *fp, *fnext;
4586 struct nfs4_delegation *dp, *dnext;
4588 for (i = 0; i < FILE_HASH_SIZE; i++) {
4589 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4590 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4601 void nfsd_forget_delegations(u64 num)
4606 count = nfsd_process_n_delegations(num, unhash_delegation);
4607 nfs4_unlock_state();
4609 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4612 void nfsd_recall_delegations(u64 num)
4617 spin_lock(&recall_lock);
4618 count = nfsd_process_n_delegations(num, nfsd_break_one_deleg);
4619 spin_unlock(&recall_lock);
4620 nfs4_unlock_state();
4622 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4625 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4627 /* initialization to perform at module load time: */
4630 nfs4_state_init(void)
4634 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4635 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4636 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4637 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4638 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4639 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4641 for (i = 0; i < SESSION_HASH_SIZE; i++)
4642 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4643 for (i = 0; i < FILE_HASH_SIZE; i++) {
4644 INIT_LIST_HEAD(&file_hashtbl[i]);
4646 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4647 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4649 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4650 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4651 INIT_LIST_HEAD(&close_lru);
4652 INIT_LIST_HEAD(&client_lru);
4653 INIT_LIST_HEAD(&del_recall_lru);
4654 reclaim_str_hashtbl_size = 0;
4658 * Since the lifetime of a delegation isn't limited to that of an open, a
4659 * client may quite reasonably hang on to a delegation as long as it has
4660 * the inode cached. This becomes an obvious problem the first time a
4661 * client's inode cache approaches the size of the server's total memory.
4663 * For now we avoid this problem by imposing a hard limit on the number
4664 * of delegations, which varies according to the server's memory size.
4667 set_max_delegations(void)
4670 * Allow at most 4 delegations per megabyte of RAM. Quick
4671 * estimates suggest that in the worst case (where every delegation
4672 * is for a different inode), a delegation could take about 1.5K,
4673 * giving a worst case usage of about 6% of memory.
4675 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4678 /* initialization to perform when the nfsd service is started: */
4681 nfs4_state_start(void)
4686 * FIXME: For now, we hang most of the pernet global stuff off of
4687 * init_net until nfsd is fully containerized. Eventually, we'll
4688 * need to pass a net pointer into this function, take a reference
4689 * to that instead and then do most of the rest of this on a per-net
4693 nfsd4_client_tracking_init(&init_net);
4694 boot_time = get_seconds();
4695 locks_start_grace(&nfsd4_manager);
4696 grace_ended = false;
4697 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4699 ret = set_callback_cred();
4704 laundry_wq = create_singlethread_workqueue("nfsd4");
4705 if (laundry_wq == NULL) {
4709 ret = nfsd4_create_callback_queue();
4711 goto out_free_laundry;
4712 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4713 set_max_delegations();
4716 destroy_workqueue(laundry_wq);
4718 nfsd4_client_tracking_exit(&init_net);
4724 __nfs4_state_shutdown(void)
4727 struct nfs4_client *clp = NULL;
4728 struct nfs4_delegation *dp = NULL;
4729 struct list_head *pos, *next, reaplist;
4731 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4732 while (!list_empty(&conf_id_hashtbl[i])) {
4733 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4736 while (!list_empty(&unconf_str_hashtbl[i])) {
4737 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4741 INIT_LIST_HEAD(&reaplist);
4742 spin_lock(&recall_lock);
4743 list_for_each_safe(pos, next, &del_recall_lru) {
4744 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4745 list_move(&dp->dl_recall_lru, &reaplist);
4747 spin_unlock(&recall_lock);
4748 list_for_each_safe(pos, next, &reaplist) {
4749 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4750 unhash_delegation(dp);
4753 nfsd4_client_tracking_exit(&init_net);
4758 nfs4_state_shutdown(void)
4760 cancel_delayed_work_sync(&laundromat_work);
4761 destroy_workqueue(laundry_wq);
4762 locks_end_grace(&nfsd4_manager);
4764 __nfs4_state_shutdown();
4765 nfs4_unlock_state();
4766 nfsd4_destroy_callback_queue();
4770 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4772 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4773 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4777 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4779 if (cstate->minorversion) {
4780 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4781 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4786 clear_current_stateid(struct nfsd4_compound_state *cstate)
4788 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4792 * functions to set current state id
4795 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4797 put_stateid(cstate, &odp->od_stateid);
4801 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4803 put_stateid(cstate, &open->op_stateid);
4807 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4809 put_stateid(cstate, &close->cl_stateid);
4813 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4815 put_stateid(cstate, &lock->lk_resp_stateid);
4819 * functions to consume current state id
4823 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4825 get_stateid(cstate, &odp->od_stateid);
4829 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4831 get_stateid(cstate, &drp->dr_stateid);
4835 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4837 get_stateid(cstate, &fsp->fr_stateid);
4841 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4843 get_stateid(cstate, &setattr->sa_stateid);
4847 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4849 get_stateid(cstate, &close->cl_stateid);
4853 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4855 get_stateid(cstate, &locku->lu_stateid);
4859 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4861 get_stateid(cstate, &read->rd_stateid);
4865 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4867 get_stateid(cstate, &write->wr_stateid);