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)) {
1025 dprintk("%s: client (clientid %08x/%08x) already expired\n",
1027 clp->cl_clientid.cl_boot,
1028 clp->cl_clientid.cl_id);
1032 dprintk("renewing client (clientid %08x/%08x)\n",
1033 clp->cl_clientid.cl_boot,
1034 clp->cl_clientid.cl_id);
1035 list_move_tail(&clp->cl_lru, &client_lru);
1036 clp->cl_time = get_seconds();
1040 renew_client(struct nfs4_client *clp)
1042 spin_lock(&client_lock);
1043 renew_client_locked(clp);
1044 spin_unlock(&client_lock);
1047 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1049 STALE_CLIENTID(clientid_t *clid)
1051 if (clid->cl_boot == boot_time)
1053 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1054 clid->cl_boot, clid->cl_id, boot_time);
1059 * XXX Should we use a slab cache ?
1060 * This type of memory management is somewhat inefficient, but we use it
1061 * anyway since SETCLIENTID is not a common operation.
1063 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1065 struct nfs4_client *clp;
1067 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1070 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1071 if (clp->cl_name.data == NULL) {
1075 clp->cl_name.len = name.len;
1080 free_client(struct nfs4_client *clp)
1082 BUG_ON(!spin_is_locked(&client_lock));
1083 while (!list_empty(&clp->cl_sessions)) {
1084 struct nfsd4_session *ses;
1085 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1087 list_del(&ses->se_perclnt);
1088 nfsd4_put_session_locked(ses);
1090 free_svc_cred(&clp->cl_cred);
1091 kfree(clp->cl_name.data);
1096 release_session_client(struct nfsd4_session *session)
1098 struct nfs4_client *clp = session->se_client;
1100 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1102 if (is_client_expired(clp)) {
1104 session->se_client = NULL;
1106 renew_client_locked(clp);
1107 spin_unlock(&client_lock);
1110 /* must be called under the client_lock */
1112 unhash_client_locked(struct nfs4_client *clp)
1114 struct nfsd4_session *ses;
1116 mark_client_expired(clp);
1117 list_del(&clp->cl_lru);
1118 spin_lock(&clp->cl_lock);
1119 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1120 list_del_init(&ses->se_hash);
1121 spin_unlock(&clp->cl_lock);
1125 expire_client(struct nfs4_client *clp)
1127 struct nfs4_openowner *oo;
1128 struct nfs4_delegation *dp;
1129 struct list_head reaplist;
1131 INIT_LIST_HEAD(&reaplist);
1132 spin_lock(&recall_lock);
1133 while (!list_empty(&clp->cl_delegations)) {
1134 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1135 list_del_init(&dp->dl_perclnt);
1136 list_move(&dp->dl_recall_lru, &reaplist);
1138 spin_unlock(&recall_lock);
1139 while (!list_empty(&reaplist)) {
1140 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1141 unhash_delegation(dp);
1143 while (!list_empty(&clp->cl_openowners)) {
1144 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1145 release_openowner(oo);
1147 nfsd4_shutdown_callback(clp);
1148 if (clp->cl_cb_conn.cb_xprt)
1149 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1150 list_del(&clp->cl_idhash);
1151 list_del(&clp->cl_strhash);
1152 spin_lock(&client_lock);
1153 unhash_client_locked(clp);
1154 if (atomic_read(&clp->cl_refcount) == 0)
1156 spin_unlock(&client_lock);
1159 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1161 memcpy(target->cl_verifier.data, source->data,
1162 sizeof(target->cl_verifier.data));
1165 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1167 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1168 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1171 static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1173 if (source->cr_principal) {
1174 target->cr_principal =
1175 kstrdup(source->cr_principal, GFP_KERNEL);
1176 if (target->cr_principal == NULL)
1179 target->cr_principal = NULL;
1180 target->cr_uid = source->cr_uid;
1181 target->cr_gid = source->cr_gid;
1182 target->cr_group_info = source->cr_group_info;
1183 get_group_info(target->cr_group_info);
1187 static int same_name(const char *n1, const char *n2)
1189 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1193 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1195 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1199 same_clid(clientid_t *cl1, clientid_t *cl2)
1201 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1204 /* XXX what about NGROUP */
1206 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1208 return cr1->cr_uid == cr2->cr_uid;
1211 static void gen_clid(struct nfs4_client *clp)
1213 static u32 current_clientid = 1;
1215 clp->cl_clientid.cl_boot = boot_time;
1216 clp->cl_clientid.cl_id = current_clientid++;
1219 static void gen_confirm(struct nfs4_client *clp)
1224 verf[0] = (__be32)get_seconds();
1225 verf[1] = (__be32)i++;
1226 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1229 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1231 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1234 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1236 struct nfs4_stid *s;
1238 s = find_stateid(cl, t);
1241 if (typemask & s->sc_type)
1246 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1247 struct svc_rqst *rqstp, nfs4_verifier *verf)
1249 struct nfs4_client *clp;
1250 struct sockaddr *sa = svc_addr(rqstp);
1253 clp = alloc_client(name);
1257 INIT_LIST_HEAD(&clp->cl_sessions);
1258 ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1260 spin_lock(&client_lock);
1262 spin_unlock(&client_lock);
1265 idr_init(&clp->cl_stateids);
1266 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1267 atomic_set(&clp->cl_refcount, 0);
1268 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1269 INIT_LIST_HEAD(&clp->cl_idhash);
1270 INIT_LIST_HEAD(&clp->cl_strhash);
1271 INIT_LIST_HEAD(&clp->cl_openowners);
1272 INIT_LIST_HEAD(&clp->cl_delegations);
1273 INIT_LIST_HEAD(&clp->cl_lru);
1274 INIT_LIST_HEAD(&clp->cl_callbacks);
1275 spin_lock_init(&clp->cl_lock);
1276 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1277 clp->cl_time = get_seconds();
1278 clear_bit(0, &clp->cl_cb_slot_busy);
1279 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1280 copy_verf(clp, verf);
1281 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1282 clp->cl_flavor = rqstp->rq_flavor;
1284 clp->cl_cb_session = NULL;
1289 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1291 unsigned int idhashval;
1293 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1294 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1295 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1300 move_to_confirmed(struct nfs4_client *clp)
1302 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1303 unsigned int strhashval;
1305 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1306 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1307 strhashval = clientstr_hashval(clp->cl_recdir);
1308 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1312 static struct nfs4_client *
1313 find_confirmed_client(clientid_t *clid)
1315 struct nfs4_client *clp;
1316 unsigned int idhashval = clientid_hashval(clid->cl_id);
1318 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1319 if (same_clid(&clp->cl_clientid, clid)) {
1327 static struct nfs4_client *
1328 find_unconfirmed_client(clientid_t *clid)
1330 struct nfs4_client *clp;
1331 unsigned int idhashval = clientid_hashval(clid->cl_id);
1333 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1334 if (same_clid(&clp->cl_clientid, clid))
1340 static bool clp_used_exchangeid(struct nfs4_client *clp)
1342 return clp->cl_exchange_flags != 0;
1345 static struct nfs4_client *
1346 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1348 struct nfs4_client *clp;
1350 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1351 if (same_name(clp->cl_recdir, dname))
1357 static struct nfs4_client *
1358 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1360 struct nfs4_client *clp;
1362 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1363 if (same_name(clp->cl_recdir, dname))
1370 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1372 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1373 struct sockaddr *sa = svc_addr(rqstp);
1374 u32 scopeid = rpc_get_scope_id(sa);
1375 unsigned short expected_family;
1377 /* Currently, we only support tcp and tcp6 for the callback channel */
1378 if (se->se_callback_netid_len == 3 &&
1379 !memcmp(se->se_callback_netid_val, "tcp", 3))
1380 expected_family = AF_INET;
1381 else if (se->se_callback_netid_len == 4 &&
1382 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1383 expected_family = AF_INET6;
1387 conn->cb_addrlen = rpc_uaddr2sockaddr(&init_net, se->se_callback_addr_val,
1388 se->se_callback_addr_len,
1389 (struct sockaddr *)&conn->cb_addr,
1390 sizeof(conn->cb_addr));
1392 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1395 if (conn->cb_addr.ss_family == AF_INET6)
1396 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1398 conn->cb_prog = se->se_callback_prog;
1399 conn->cb_ident = se->se_callback_ident;
1400 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1403 conn->cb_addr.ss_family = AF_UNSPEC;
1404 conn->cb_addrlen = 0;
1405 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1406 "will not receive delegations\n",
1407 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1413 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1416 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1418 struct nfsd4_slot *slot = resp->cstate.slot;
1421 dprintk("--> %s slot %p\n", __func__, slot);
1423 slot->sl_opcnt = resp->opcnt;
1424 slot->sl_status = resp->cstate.status;
1426 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1427 if (nfsd4_not_cached(resp)) {
1428 slot->sl_datalen = 0;
1431 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1432 base = (char *)resp->cstate.datap -
1433 (char *)resp->xbuf->head[0].iov_base;
1434 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1436 WARN("%s: sessions DRC could not cache compound\n", __func__);
1441 * Encode the replay sequence operation from the slot values.
1442 * If cachethis is FALSE encode the uncached rep error on the next
1443 * operation which sets resp->p and increments resp->opcnt for
1444 * nfs4svc_encode_compoundres.
1448 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1449 struct nfsd4_compoundres *resp)
1451 struct nfsd4_op *op;
1452 struct nfsd4_slot *slot = resp->cstate.slot;
1454 /* Encode the replayed sequence operation */
1455 op = &args->ops[resp->opcnt - 1];
1456 nfsd4_encode_operation(resp, op);
1458 /* Return nfserr_retry_uncached_rep in next operation. */
1459 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1460 op = &args->ops[resp->opcnt++];
1461 op->status = nfserr_retry_uncached_rep;
1462 nfsd4_encode_operation(resp, op);
1468 * The sequence operation is not cached because we can use the slot and
1472 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1473 struct nfsd4_sequence *seq)
1475 struct nfsd4_slot *slot = resp->cstate.slot;
1478 dprintk("--> %s slot %p\n", __func__, slot);
1480 /* Either returns 0 or nfserr_retry_uncached */
1481 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1482 if (status == nfserr_retry_uncached_rep)
1485 /* The sequence operation has been encoded, cstate->datap set. */
1486 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1488 resp->opcnt = slot->sl_opcnt;
1489 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1490 status = slot->sl_status;
1496 * Set the exchange_id flags returned by the server.
1499 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1501 /* pNFS is not supported */
1502 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1504 /* Referrals are supported, Migration is not. */
1505 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1507 /* set the wire flags to return to client. */
1508 clid->flags = new->cl_exchange_flags;
1511 static bool client_has_state(struct nfs4_client *clp)
1514 * Note clp->cl_openowners check isn't quite right: there's no
1515 * need to count owners without stateid's.
1517 * Also note we should probably be using this in 4.0 case too.
1519 return list_empty(&clp->cl_openowners)
1520 && list_empty(&clp->cl_delegations)
1521 && list_empty(&clp->cl_sessions);
1525 nfsd4_exchange_id(struct svc_rqst *rqstp,
1526 struct nfsd4_compound_state *cstate,
1527 struct nfsd4_exchange_id *exid)
1529 struct nfs4_client *unconf, *conf, *new;
1531 unsigned int strhashval;
1532 char dname[HEXDIR_LEN];
1533 char addr_str[INET6_ADDRSTRLEN];
1534 nfs4_verifier verf = exid->verifier;
1535 struct sockaddr *sa = svc_addr(rqstp);
1536 bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
1538 rpc_ntop(sa, addr_str, sizeof(addr_str));
1539 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1540 "ip_addr=%s flags %x, spa_how %d\n",
1541 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1542 addr_str, exid->flags, exid->spa_how);
1544 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1545 return nfserr_inval;
1547 /* Currently only support SP4_NONE */
1548 switch (exid->spa_how) {
1552 return nfserr_serverfault;
1554 BUG(); /* checked by xdr code */
1556 return nfserr_serverfault; /* no excuse :-/ */
1559 status = nfs4_make_rec_clidname(dname, &exid->clname);
1564 strhashval = clientstr_hashval(dname);
1566 /* Cases below refer to rfc 5661 section 18.35.4: */
1568 conf = find_confirmed_client_by_str(dname, strhashval);
1570 bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
1571 bool verfs_match = same_verf(&verf, &conf->cl_verifier);
1574 if (!clp_used_exchangeid(conf)) { /* buggy client */
1575 status = nfserr_inval;
1578 if (!creds_match) { /* case 9 */
1579 status = nfserr_perm;
1582 if (!verfs_match) { /* case 8 */
1583 status = nfserr_not_same;
1587 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1591 if (!creds_match) { /* case 3 */
1592 if (client_has_state(conf)) {
1593 status = nfserr_clid_inuse;
1598 if (verfs_match) { /* case 2 */
1599 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1603 /* case 5, client reboot */
1605 expire_client(conf);
1609 if (update) { /* case 7 */
1610 status = nfserr_noent;
1614 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1615 if (unconf) /* case 4, possible retry or client restart */
1616 expire_client(unconf);
1618 /* case 1 (normal case) */
1620 new = create_client(exid->clname, dname, rqstp, &verf);
1622 status = nfserr_jukebox;
1627 add_to_unconfirmed(new, strhashval);
1629 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1630 exid->clientid.cl_id = new->cl_clientid.cl_id;
1633 nfsd4_set_ex_flags(new, exid);
1635 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1636 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1640 nfs4_unlock_state();
1645 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1647 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1650 /* The slot is in use, and no response has been sent. */
1652 if (seqid == slot_seqid)
1653 return nfserr_jukebox;
1655 return nfserr_seq_misordered;
1657 /* Note unsigned 32-bit arithmetic handles wraparound: */
1658 if (likely(seqid == slot_seqid + 1))
1660 if (seqid == slot_seqid)
1661 return nfserr_replay_cache;
1662 return nfserr_seq_misordered;
1666 * Cache the create session result into the create session single DRC
1667 * slot cache by saving the xdr structure. sl_seqid has been set.
1668 * Do this for solo or embedded create session operations.
1671 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1672 struct nfsd4_clid_slot *slot, __be32 nfserr)
1674 slot->sl_status = nfserr;
1675 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1679 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1680 struct nfsd4_clid_slot *slot)
1682 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1683 return slot->sl_status;
1686 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1687 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1688 1 + /* MIN tag is length with zero, only length */ \
1689 3 + /* version, opcount, opcode */ \
1690 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1691 /* seqid, slotID, slotID, cache */ \
1692 4 ) * sizeof(__be32))
1694 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1695 2 + /* verifier: AUTH_NULL, length 0 */\
1697 1 + /* MIN tag is length with zero, only length */ \
1698 3 + /* opcount, opcode, opstatus*/ \
1699 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1700 /* seqid, slotID, slotID, slotID, status */ \
1701 5 ) * sizeof(__be32))
1703 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1705 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1706 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1710 nfsd4_create_session(struct svc_rqst *rqstp,
1711 struct nfsd4_compound_state *cstate,
1712 struct nfsd4_create_session *cr_ses)
1714 struct sockaddr *sa = svc_addr(rqstp);
1715 struct nfs4_client *conf, *unconf;
1716 struct nfsd4_session *new;
1717 struct nfsd4_clid_slot *cs_slot = NULL;
1718 bool confirm_me = false;
1721 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1722 return nfserr_inval;
1725 unconf = find_unconfirmed_client(&cr_ses->clientid);
1726 conf = find_confirmed_client(&cr_ses->clientid);
1729 cs_slot = &conf->cl_cs_slot;
1730 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1731 if (status == nfserr_replay_cache) {
1732 dprintk("Got a create_session replay! seqid= %d\n",
1734 /* Return the cached reply status */
1735 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1737 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1738 status = nfserr_seq_misordered;
1739 dprintk("Sequence misordered!\n");
1740 dprintk("Expected seqid= %d but got seqid= %d\n",
1741 cs_slot->sl_seqid, cr_ses->seqid);
1744 } else if (unconf) {
1745 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1746 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1747 status = nfserr_clid_inuse;
1751 cs_slot = &unconf->cl_cs_slot;
1752 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1754 /* an unconfirmed replay returns misordered */
1755 status = nfserr_seq_misordered;
1762 status = nfserr_stale_clientid;
1767 * XXX: we should probably set this at creation time, and check
1768 * for consistent minorversion use throughout:
1770 conf->cl_minorversion = 1;
1772 * We do not support RDMA or persistent sessions
1774 cr_ses->flags &= ~SESSION4_PERSIST;
1775 cr_ses->flags &= ~SESSION4_RDMA;
1777 status = nfserr_toosmall;
1778 if (check_forechannel_attrs(cr_ses->fore_channel))
1781 status = nfserr_jukebox;
1782 new = alloc_init_session(rqstp, conf, cr_ses);
1786 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1787 NFS4_MAX_SESSIONID_LEN);
1788 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1789 sizeof(struct nfsd4_channel_attrs));
1790 cs_slot->sl_seqid++;
1791 cr_ses->seqid = cs_slot->sl_seqid;
1793 /* cache solo and embedded create sessions under the state lock */
1794 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1796 move_to_confirmed(conf);
1798 nfs4_unlock_state();
1799 dprintk("%s returns %d\n", __func__, ntohl(status));
1803 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1805 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1806 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1808 return argp->opcnt == resp->opcnt;
1811 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1814 case NFS4_CDFC4_FORE:
1815 case NFS4_CDFC4_BACK:
1817 case NFS4_CDFC4_FORE_OR_BOTH:
1818 case NFS4_CDFC4_BACK_OR_BOTH:
1819 *dir = NFS4_CDFC4_BOTH;
1822 return nfserr_inval;
1825 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1826 struct nfsd4_compound_state *cstate,
1827 struct nfsd4_bind_conn_to_session *bcts)
1831 if (!nfsd4_last_compound_op(rqstp))
1832 return nfserr_not_only_op;
1833 spin_lock(&client_lock);
1834 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1835 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1836 * client_lock iself: */
1837 if (cstate->session) {
1838 nfsd4_get_session(cstate->session);
1839 atomic_inc(&cstate->session->se_client->cl_refcount);
1841 spin_unlock(&client_lock);
1842 if (!cstate->session)
1843 return nfserr_badsession;
1845 status = nfsd4_map_bcts_dir(&bcts->dir);
1847 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1851 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1855 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1859 nfsd4_destroy_session(struct svc_rqst *r,
1860 struct nfsd4_compound_state *cstate,
1861 struct nfsd4_destroy_session *sessionid)
1863 struct nfsd4_session *ses;
1864 __be32 status = nfserr_badsession;
1867 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1868 * - Should we return nfserr_back_chan_busy if waiting for
1869 * callbacks on to-be-destroyed session?
1870 * - Do we need to clear any callback info from previous session?
1873 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1874 if (!nfsd4_last_compound_op(r))
1875 return nfserr_not_only_op;
1877 dump_sessionid(__func__, &sessionid->sessionid);
1878 spin_lock(&client_lock);
1879 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1881 spin_unlock(&client_lock);
1885 unhash_session(ses);
1886 spin_unlock(&client_lock);
1889 nfsd4_probe_callback_sync(ses->se_client);
1890 nfs4_unlock_state();
1892 spin_lock(&client_lock);
1893 nfsd4_del_conns(ses);
1894 nfsd4_put_session_locked(ses);
1895 spin_unlock(&client_lock);
1898 dprintk("%s returns %d\n", __func__, ntohl(status));
1902 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1904 struct nfsd4_conn *c;
1906 list_for_each_entry(c, &s->se_conns, cn_persession) {
1907 if (c->cn_xprt == xpt) {
1914 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1916 struct nfs4_client *clp = ses->se_client;
1917 struct nfsd4_conn *c;
1920 spin_lock(&clp->cl_lock);
1921 c = __nfsd4_find_conn(new->cn_xprt, ses);
1923 spin_unlock(&clp->cl_lock);
1927 __nfsd4_hash_conn(new, ses);
1928 spin_unlock(&clp->cl_lock);
1929 ret = nfsd4_register_conn(new);
1931 /* oops; xprt is already down: */
1932 nfsd4_conn_lost(&new->cn_xpt_user);
1936 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1938 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1940 return args->opcnt > session->se_fchannel.maxops;
1943 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1944 struct nfsd4_session *session)
1946 struct xdr_buf *xb = &rqstp->rq_arg;
1948 return xb->len > session->se_fchannel.maxreq_sz;
1952 nfsd4_sequence(struct svc_rqst *rqstp,
1953 struct nfsd4_compound_state *cstate,
1954 struct nfsd4_sequence *seq)
1956 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1957 struct nfsd4_session *session;
1958 struct nfsd4_slot *slot;
1959 struct nfsd4_conn *conn;
1962 if (resp->opcnt != 1)
1963 return nfserr_sequence_pos;
1966 * Will be either used or freed by nfsd4_sequence_check_conn
1969 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1971 return nfserr_jukebox;
1973 spin_lock(&client_lock);
1974 status = nfserr_badsession;
1975 session = find_in_sessionid_hashtbl(&seq->sessionid);
1979 status = nfserr_too_many_ops;
1980 if (nfsd4_session_too_many_ops(rqstp, session))
1983 status = nfserr_req_too_big;
1984 if (nfsd4_request_too_big(rqstp, session))
1987 status = nfserr_badslot;
1988 if (seq->slotid >= session->se_fchannel.maxreqs)
1991 slot = session->se_slots[seq->slotid];
1992 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1994 /* We do not negotiate the number of slots yet, so set the
1995 * maxslots to the session maxreqs which is used to encode
1996 * sr_highest_slotid and the sr_target_slot id to maxslots */
1997 seq->maxslots = session->se_fchannel.maxreqs;
1999 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2000 slot->sl_flags & NFSD4_SLOT_INUSE);
2001 if (status == nfserr_replay_cache) {
2002 status = nfserr_seq_misordered;
2003 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2005 cstate->slot = slot;
2006 cstate->session = session;
2007 /* Return the cached reply status and set cstate->status
2008 * for nfsd4_proc_compound processing */
2009 status = nfsd4_replay_cache_entry(resp, seq);
2010 cstate->status = nfserr_replay_cache;
2016 nfsd4_sequence_check_conn(conn, session);
2019 /* Success! bump slot seqid */
2020 slot->sl_seqid = seq->seqid;
2021 slot->sl_flags |= NFSD4_SLOT_INUSE;
2023 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2025 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2027 cstate->slot = slot;
2028 cstate->session = session;
2031 /* Hold a session reference until done processing the compound. */
2032 if (cstate->session) {
2033 struct nfs4_client *clp = session->se_client;
2035 nfsd4_get_session(cstate->session);
2036 atomic_inc(&clp->cl_refcount);
2037 switch (clp->cl_cb_state) {
2039 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2041 case NFSD4_CB_FAULT:
2042 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2045 seq->status_flags = 0;
2049 spin_unlock(&client_lock);
2050 dprintk("%s: return %d\n", __func__, ntohl(status));
2054 static inline bool has_resources(struct nfs4_client *clp)
2056 return !list_empty(&clp->cl_openowners)
2057 || !list_empty(&clp->cl_delegations)
2058 || !list_empty(&clp->cl_sessions);
2062 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2064 struct nfs4_client *conf, *unconf, *clp;
2068 unconf = find_unconfirmed_client(&dc->clientid);
2069 conf = find_confirmed_client(&dc->clientid);
2074 if (!is_client_expired(conf) && has_resources(conf)) {
2075 status = nfserr_clientid_busy;
2079 /* rfc5661 18.50.3 */
2080 if (cstate->session && conf == cstate->session->se_client) {
2081 status = nfserr_clientid_busy;
2087 status = nfserr_stale_clientid;
2093 nfs4_unlock_state();
2094 dprintk("%s return %d\n", __func__, ntohl(status));
2099 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2103 if (rc->rca_one_fs) {
2104 if (!cstate->current_fh.fh_dentry)
2105 return nfserr_nofilehandle;
2107 * We don't take advantage of the rca_one_fs case.
2108 * That's OK, it's optional, we can safely ignore it.
2114 status = nfserr_complete_already;
2115 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2116 &cstate->session->se_client->cl_flags))
2119 status = nfserr_stale_clientid;
2120 if (is_client_expired(cstate->session->se_client))
2122 * The following error isn't really legal.
2123 * But we only get here if the client just explicitly
2124 * destroyed the client. Surely it no longer cares what
2125 * error it gets back on an operation for the dead
2131 nfsd4_client_record_create(cstate->session->se_client);
2133 nfs4_unlock_state();
2138 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2139 struct nfsd4_setclientid *setclid)
2141 struct xdr_netobj clname = setclid->se_name;
2142 nfs4_verifier clverifier = setclid->se_verf;
2143 unsigned int strhashval;
2144 struct nfs4_client *conf, *unconf, *new;
2146 char dname[HEXDIR_LEN];
2148 status = nfs4_make_rec_clidname(dname, &clname);
2153 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2154 * We get here on a DRC miss.
2157 strhashval = clientstr_hashval(dname);
2160 conf = find_confirmed_client_by_str(dname, strhashval);
2162 /* RFC 3530 14.2.33 CASE 0: */
2163 status = nfserr_clid_inuse;
2164 if (clp_used_exchangeid(conf))
2166 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2167 char addr_str[INET6_ADDRSTRLEN];
2168 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2170 dprintk("NFSD: setclientid: string in use by client "
2171 "at %s\n", addr_str);
2176 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2177 * has a description of SETCLIENTID request processing consisting
2178 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2180 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2181 status = nfserr_jukebox;
2184 * RFC 3530 14.2.33 CASE 4:
2185 * placed first, because it is the normal case
2188 expire_client(unconf);
2189 new = create_client(clname, dname, rqstp, &clverifier);
2193 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
2195 * RFC 3530 14.2.33 CASE 1:
2196 * probable callback update
2199 /* Note this is removing unconfirmed {*x***},
2200 * which is stronger than RFC recommended {vxc**}.
2201 * This has the advantage that there is at most
2202 * one {*x***} in either list at any time.
2204 expire_client(unconf);
2206 new = create_client(clname, dname, rqstp, &clverifier);
2209 copy_clid(new, conf);
2210 } else if (!unconf) {
2212 * RFC 3530 14.2.33 CASE 2:
2213 * probable client reboot; state will be removed if
2216 new = create_client(clname, dname, rqstp, &clverifier);
2222 * RFC 3530 14.2.33 CASE 3:
2223 * probable client reboot; state will be removed if
2226 expire_client(unconf);
2227 new = create_client(clname, dname, rqstp, &clverifier);
2233 * XXX: we should probably set this at creation time, and check
2234 * for consistent minorversion use throughout:
2236 new->cl_minorversion = 0;
2237 gen_callback(new, setclid, rqstp);
2238 add_to_unconfirmed(new, strhashval);
2239 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2240 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2241 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2244 nfs4_unlock_state();
2250 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2251 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2252 * bullets, labeled as CASE1 - CASE4 below.
2255 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2256 struct nfsd4_compound_state *cstate,
2257 struct nfsd4_setclientid_confirm *setclientid_confirm)
2259 struct nfs4_client *conf, *unconf;
2260 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2261 clientid_t * clid = &setclientid_confirm->sc_clientid;
2264 if (STALE_CLIENTID(clid))
2265 return nfserr_stale_clientid;
2267 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2268 * We get here on a DRC miss.
2273 conf = find_confirmed_client(clid);
2274 unconf = find_unconfirmed_client(clid);
2277 * section 14.2.34 of RFC 3530 has a description of
2278 * SETCLIENTID_CONFIRM request processing consisting
2279 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2281 status = nfserr_clid_inuse;
2282 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
2284 * RFC 3530 14.2.34 CASE 1:
2287 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
2288 status = nfserr_clid_inuse;
2290 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2291 nfsd4_probe_callback(conf);
2292 expire_client(unconf);
2295 } else if (conf && !unconf) {
2297 * RFC 3530 14.2.34 CASE 2:
2298 * probable retransmitted request; play it safe and
2301 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
2302 status = nfserr_clid_inuse;
2305 } else if (!conf && unconf
2306 && same_verf(&unconf->cl_confirm, &confirm)) {
2308 * RFC 3530 14.2.34 CASE 3:
2309 * Normal case; new or rebooted client:
2311 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
2312 status = nfserr_clid_inuse;
2315 clientstr_hashval(unconf->cl_recdir);
2316 conf = find_confirmed_client_by_str(unconf->cl_recdir,
2319 nfsd4_client_record_remove(conf);
2320 expire_client(conf);
2322 move_to_confirmed(unconf);
2324 nfsd4_probe_callback(conf);
2327 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
2328 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
2331 * RFC 3530 14.2.34 CASE 4:
2332 * Client probably hasn't noticed that we rebooted yet.
2334 status = nfserr_stale_clientid;
2337 nfs4_unlock_state();
2341 static struct nfs4_file *nfsd4_alloc_file(void)
2343 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2346 /* OPEN Share state helper functions */
2347 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2349 unsigned int hashval = file_hashval(ino);
2351 atomic_set(&fp->fi_ref, 1);
2352 INIT_LIST_HEAD(&fp->fi_hash);
2353 INIT_LIST_HEAD(&fp->fi_stateids);
2354 INIT_LIST_HEAD(&fp->fi_delegations);
2355 fp->fi_inode = igrab(ino);
2356 fp->fi_had_conflict = false;
2357 fp->fi_lease = NULL;
2358 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2359 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2360 spin_lock(&recall_lock);
2361 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2362 spin_unlock(&recall_lock);
2366 nfsd4_free_slab(struct kmem_cache **slab)
2370 kmem_cache_destroy(*slab);
2375 nfsd4_free_slabs(void)
2377 nfsd4_free_slab(&openowner_slab);
2378 nfsd4_free_slab(&lockowner_slab);
2379 nfsd4_free_slab(&file_slab);
2380 nfsd4_free_slab(&stateid_slab);
2381 nfsd4_free_slab(&deleg_slab);
2385 nfsd4_init_slabs(void)
2387 openowner_slab = kmem_cache_create("nfsd4_openowners",
2388 sizeof(struct nfs4_openowner), 0, 0, NULL);
2389 if (openowner_slab == NULL)
2391 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2392 sizeof(struct nfs4_openowner), 0, 0, NULL);
2393 if (lockowner_slab == NULL)
2395 file_slab = kmem_cache_create("nfsd4_files",
2396 sizeof(struct nfs4_file), 0, 0, NULL);
2397 if (file_slab == NULL)
2399 stateid_slab = kmem_cache_create("nfsd4_stateids",
2400 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2401 if (stateid_slab == NULL)
2403 deleg_slab = kmem_cache_create("nfsd4_delegations",
2404 sizeof(struct nfs4_delegation), 0, 0, NULL);
2405 if (deleg_slab == NULL)
2410 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2414 void nfs4_free_openowner(struct nfs4_openowner *oo)
2416 kfree(oo->oo_owner.so_owner.data);
2417 kmem_cache_free(openowner_slab, oo);
2420 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2422 kfree(lo->lo_owner.so_owner.data);
2423 kmem_cache_free(lockowner_slab, lo);
2426 static void init_nfs4_replay(struct nfs4_replay *rp)
2428 rp->rp_status = nfserr_serverfault;
2430 rp->rp_buf = rp->rp_ibuf;
2433 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2435 struct nfs4_stateowner *sop;
2437 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2441 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2442 if (!sop->so_owner.data) {
2443 kmem_cache_free(slab, sop);
2446 sop->so_owner.len = owner->len;
2448 INIT_LIST_HEAD(&sop->so_stateids);
2449 sop->so_client = clp;
2450 init_nfs4_replay(&sop->so_replay);
2454 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2456 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2457 list_add(&oo->oo_perclient, &clp->cl_openowners);
2460 static struct nfs4_openowner *
2461 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2462 struct nfs4_openowner *oo;
2464 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2467 oo->oo_owner.so_is_open_owner = 1;
2468 oo->oo_owner.so_seqid = open->op_seqid;
2469 oo->oo_flags = NFS4_OO_NEW;
2471 oo->oo_last_closed_stid = NULL;
2472 INIT_LIST_HEAD(&oo->oo_close_lru);
2473 hash_openowner(oo, clp, strhashval);
2477 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2478 struct nfs4_openowner *oo = open->op_openowner;
2479 struct nfs4_client *clp = oo->oo_owner.so_client;
2481 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2482 INIT_LIST_HEAD(&stp->st_lockowners);
2483 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2484 list_add(&stp->st_perfile, &fp->fi_stateids);
2485 stp->st_stateowner = &oo->oo_owner;
2488 stp->st_access_bmap = 0;
2489 stp->st_deny_bmap = 0;
2490 set_access(open->op_share_access, stp);
2491 set_deny(open->op_share_deny, stp);
2492 stp->st_openstp = NULL;
2496 move_to_close_lru(struct nfs4_openowner *oo)
2498 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2500 list_move_tail(&oo->oo_close_lru, &close_lru);
2501 oo->oo_time = get_seconds();
2505 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2508 return (sop->so_owner.len == owner->len) &&
2509 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2510 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2513 static struct nfs4_openowner *
2514 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2516 struct nfs4_stateowner *so;
2517 struct nfs4_openowner *oo;
2519 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2520 if (!so->so_is_open_owner)
2522 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2524 renew_client(oo->oo_owner.so_client);
2531 /* search file_hashtbl[] for file */
2532 static struct nfs4_file *
2533 find_file(struct inode *ino)
2535 unsigned int hashval = file_hashval(ino);
2536 struct nfs4_file *fp;
2538 spin_lock(&recall_lock);
2539 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2540 if (fp->fi_inode == ino) {
2542 spin_unlock(&recall_lock);
2546 spin_unlock(&recall_lock);
2551 * Called to check deny when READ with all zero stateid or
2552 * WRITE with all zero or all one stateid
2555 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2557 struct inode *ino = current_fh->fh_dentry->d_inode;
2558 struct nfs4_file *fp;
2559 struct nfs4_ol_stateid *stp;
2562 dprintk("NFSD: nfs4_share_conflict\n");
2564 fp = find_file(ino);
2567 ret = nfserr_locked;
2568 /* Search for conflicting share reservations */
2569 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2570 if (test_deny(deny_type, stp) ||
2571 test_deny(NFS4_SHARE_DENY_BOTH, stp))
2580 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2582 /* We're assuming the state code never drops its reference
2583 * without first removing the lease. Since we're in this lease
2584 * callback (and since the lease code is serialized by the kernel
2585 * lock) we know the server hasn't removed the lease yet, we know
2586 * it's safe to take a reference: */
2587 atomic_inc(&dp->dl_count);
2589 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2591 /* only place dl_time is set. protected by lock_flocks*/
2592 dp->dl_time = get_seconds();
2594 nfsd4_cb_recall(dp);
2597 /* Called from break_lease() with lock_flocks() held. */
2598 static void nfsd_break_deleg_cb(struct file_lock *fl)
2600 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2601 struct nfs4_delegation *dp;
2604 /* We assume break_lease is only called once per lease: */
2605 BUG_ON(fp->fi_had_conflict);
2607 * We don't want the locks code to timeout the lease for us;
2608 * we'll remove it ourself if a delegation isn't returned
2611 fl->fl_break_time = 0;
2613 spin_lock(&recall_lock);
2614 fp->fi_had_conflict = true;
2615 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2616 nfsd_break_one_deleg(dp);
2617 spin_unlock(&recall_lock);
2621 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2624 return lease_modify(onlist, arg);
2629 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2630 .lm_break = nfsd_break_deleg_cb,
2631 .lm_change = nfsd_change_deleg_cb,
2634 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2636 if (nfsd4_has_session(cstate))
2638 if (seqid == so->so_seqid - 1)
2639 return nfserr_replay_me;
2640 if (seqid == so->so_seqid)
2642 return nfserr_bad_seqid;
2646 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2647 struct nfsd4_open *open)
2649 clientid_t *clientid = &open->op_clientid;
2650 struct nfs4_client *clp = NULL;
2651 unsigned int strhashval;
2652 struct nfs4_openowner *oo = NULL;
2655 if (STALE_CLIENTID(&open->op_clientid))
2656 return nfserr_stale_clientid;
2658 * In case we need it later, after we've already created the
2659 * file and don't want to risk a further failure:
2661 open->op_file = nfsd4_alloc_file();
2662 if (open->op_file == NULL)
2663 return nfserr_jukebox;
2665 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2666 oo = find_openstateowner_str(strhashval, open);
2667 open->op_openowner = oo;
2669 clp = find_confirmed_client(clientid);
2671 return nfserr_expired;
2674 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2675 /* Replace unconfirmed owners without checking for replay. */
2676 clp = oo->oo_owner.so_client;
2677 release_openowner(oo);
2678 open->op_openowner = NULL;
2681 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2684 clp = oo->oo_owner.so_client;
2687 oo = alloc_init_open_stateowner(strhashval, clp, open);
2689 return nfserr_jukebox;
2690 open->op_openowner = oo;
2692 open->op_stp = nfs4_alloc_stateid(clp);
2694 return nfserr_jukebox;
2698 static inline __be32
2699 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2701 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2702 return nfserr_openmode;
2707 static int share_access_to_flags(u32 share_access)
2709 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2712 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2714 struct nfs4_stid *ret;
2716 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2719 return delegstateid(ret);
2722 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2724 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2725 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2729 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2730 struct nfs4_delegation **dp)
2733 __be32 status = nfserr_bad_stateid;
2735 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2738 flags = share_access_to_flags(open->op_share_access);
2739 status = nfs4_check_delegmode(*dp, flags);
2743 if (!nfsd4_is_deleg_cur(open))
2747 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2752 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2754 struct nfs4_ol_stateid *local;
2755 struct nfs4_openowner *oo = open->op_openowner;
2757 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2758 /* ignore lock owners */
2759 if (local->st_stateowner->so_is_open_owner == 0)
2761 /* remember if we have seen this open owner */
2762 if (local->st_stateowner == &oo->oo_owner)
2764 /* check for conflicting share reservations */
2765 if (!test_share(local, open))
2766 return nfserr_share_denied;
2771 static void nfs4_free_stateid(struct nfs4_ol_stateid *s)
2773 kmem_cache_free(stateid_slab, s);
2776 static inline int nfs4_access_to_access(u32 nfs4_access)
2780 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2781 flags |= NFSD_MAY_READ;
2782 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2783 flags |= NFSD_MAY_WRITE;
2787 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2788 struct svc_fh *cur_fh, struct nfsd4_open *open)
2791 int oflag = nfs4_access_to_omode(open->op_share_access);
2792 int access = nfs4_access_to_access(open->op_share_access);
2794 if (!fp->fi_fds[oflag]) {
2795 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2796 &fp->fi_fds[oflag]);
2800 nfs4_file_get_access(fp, oflag);
2805 static inline __be32
2806 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2807 struct nfsd4_open *open)
2809 struct iattr iattr = {
2810 .ia_valid = ATTR_SIZE,
2813 if (!open->op_truncate)
2815 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2816 return nfserr_inval;
2817 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2821 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)
2823 u32 op_share_access = open->op_share_access;
2827 new_access = !test_access(op_share_access, stp);
2829 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2833 status = nfsd4_truncate(rqstp, cur_fh, open);
2836 int oflag = nfs4_access_to_omode(op_share_access);
2837 nfs4_file_put_access(fp, oflag);
2841 /* remember the open */
2842 set_access(op_share_access, stp);
2843 set_deny(open->op_share_deny, stp);
2850 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2852 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2855 /* Should we give out recallable state?: */
2856 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2858 if (clp->cl_cb_state == NFSD4_CB_UP)
2861 * In the sessions case, since we don't have to establish a
2862 * separate connection for callbacks, we assume it's OK
2863 * until we hear otherwise:
2865 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2868 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2870 struct file_lock *fl;
2872 fl = locks_alloc_lock();
2875 locks_init_lock(fl);
2876 fl->fl_lmops = &nfsd_lease_mng_ops;
2877 fl->fl_flags = FL_LEASE;
2878 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2879 fl->fl_end = OFFSET_MAX;
2880 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2881 fl->fl_pid = current->tgid;
2885 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2887 struct nfs4_file *fp = dp->dl_file;
2888 struct file_lock *fl;
2891 fl = nfs4_alloc_init_lease(dp, flag);
2894 fl->fl_file = find_readable_file(fp);
2895 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2896 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2898 list_del_init(&dp->dl_perclnt);
2899 locks_free_lock(fl);
2903 fp->fi_deleg_file = fl->fl_file;
2904 get_file(fp->fi_deleg_file);
2905 atomic_set(&fp->fi_delegees, 1);
2906 list_add(&dp->dl_perfile, &fp->fi_delegations);
2910 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2912 struct nfs4_file *fp = dp->dl_file;
2915 return nfs4_setlease(dp, flag);
2916 spin_lock(&recall_lock);
2917 if (fp->fi_had_conflict) {
2918 spin_unlock(&recall_lock);
2921 atomic_inc(&fp->fi_delegees);
2922 list_add(&dp->dl_perfile, &fp->fi_delegations);
2923 spin_unlock(&recall_lock);
2924 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2928 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2930 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2931 if (status == -EAGAIN)
2932 open->op_why_no_deleg = WND4_CONTENTION;
2934 open->op_why_no_deleg = WND4_RESOURCE;
2935 switch (open->op_deleg_want) {
2936 case NFS4_SHARE_WANT_READ_DELEG:
2937 case NFS4_SHARE_WANT_WRITE_DELEG:
2938 case NFS4_SHARE_WANT_ANY_DELEG:
2940 case NFS4_SHARE_WANT_CANCEL:
2941 open->op_why_no_deleg = WND4_CANCELLED;
2943 case NFS4_SHARE_WANT_NO_DELEG:
2944 BUG(); /* not supposed to get here */
2950 * Attempt to hand out a delegation.
2953 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2955 struct nfs4_delegation *dp;
2956 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2958 int status = 0, flag = 0;
2960 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2961 flag = NFS4_OPEN_DELEGATE_NONE;
2962 open->op_recall = 0;
2963 switch (open->op_claim_type) {
2964 case NFS4_OPEN_CLAIM_PREVIOUS:
2966 open->op_recall = 1;
2967 flag = open->op_delegate_type;
2968 if (flag == NFS4_OPEN_DELEGATE_NONE)
2971 case NFS4_OPEN_CLAIM_NULL:
2972 /* Let's not give out any delegations till everyone's
2973 * had the chance to reclaim theirs.... */
2974 if (locks_in_grace())
2976 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2978 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2979 flag = NFS4_OPEN_DELEGATE_WRITE;
2981 flag = NFS4_OPEN_DELEGATE_READ;
2987 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2990 status = nfs4_set_delegation(dp, flag);
2994 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2996 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2997 STATEID_VAL(&dp->dl_stid.sc_stateid));
2999 open->op_delegate_type = flag;
3000 if (flag == NFS4_OPEN_DELEGATE_NONE) {
3001 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
3002 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
3003 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
3005 /* 4.1 client asking for a delegation? */
3006 if (open->op_deleg_want)
3007 nfsd4_open_deleg_none_ext(open, status);
3011 nfs4_put_delegation(dp);
3013 flag = NFS4_OPEN_DELEGATE_NONE;
3017 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
3018 struct nfs4_delegation *dp)
3020 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
3021 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3022 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3023 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
3024 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
3025 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
3026 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3027 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
3029 /* Otherwise the client must be confused wanting a delegation
3030 * it already has, therefore we don't return
3031 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3036 * called with nfs4_lock_state() held.
3039 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3041 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3042 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3043 struct nfs4_file *fp = NULL;
3044 struct inode *ino = current_fh->fh_dentry->d_inode;
3045 struct nfs4_ol_stateid *stp = NULL;
3046 struct nfs4_delegation *dp = NULL;
3050 * Lookup file; if found, lookup stateid and check open request,
3051 * and check for delegations in the process of being recalled.
3052 * If not found, create the nfs4_file struct
3054 fp = find_file(ino);
3056 if ((status = nfs4_check_open(fp, open, &stp)))
3058 status = nfs4_check_deleg(cl, fp, open, &dp);
3062 status = nfserr_bad_stateid;
3063 if (nfsd4_is_deleg_cur(open))
3065 status = nfserr_jukebox;
3067 open->op_file = NULL;
3068 nfsd4_init_file(fp, ino);
3072 * OPEN the file, or upgrade an existing OPEN.
3073 * If truncate fails, the OPEN fails.
3076 /* Stateid was found, this is an OPEN upgrade */
3077 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3081 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3085 open->op_stp = NULL;
3086 init_open_stateid(stp, fp, open);
3087 status = nfsd4_truncate(rqstp, current_fh, open);
3089 release_open_stateid(stp);
3093 update_stateid(&stp->st_stid.sc_stateid);
3094 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3096 if (nfsd4_has_session(&resp->cstate)) {
3097 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3099 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3100 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3101 open->op_why_no_deleg = WND4_NOT_WANTED;
3107 * Attempt to hand out a delegation. No error return, because the
3108 * OPEN succeeds even if we fail.
3110 nfs4_open_delegation(current_fh, open, stp);
3114 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3115 STATEID_VAL(&stp->st_stid.sc_stateid));
3117 /* 4.1 client trying to upgrade/downgrade delegation? */
3118 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3119 open->op_deleg_want)
3120 nfsd4_deleg_xgrade_none_ext(open, dp);
3124 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3125 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3127 * To finish the open response, we just need to set the rflags.
3129 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3130 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3131 !nfsd4_has_session(&resp->cstate))
3132 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3137 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3139 if (open->op_openowner) {
3140 struct nfs4_openowner *oo = open->op_openowner;
3142 if (!list_empty(&oo->oo_owner.so_stateids))
3143 list_del_init(&oo->oo_close_lru);
3144 if (oo->oo_flags & NFS4_OO_NEW) {
3146 release_openowner(oo);
3147 open->op_openowner = NULL;
3149 oo->oo_flags &= ~NFS4_OO_NEW;
3153 nfsd4_free_file(open->op_file);
3155 nfs4_free_stateid(open->op_stp);
3159 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3162 struct nfs4_client *clp;
3166 dprintk("process_renew(%08x/%08x): starting\n",
3167 clid->cl_boot, clid->cl_id);
3168 status = nfserr_stale_clientid;
3169 if (STALE_CLIENTID(clid))
3171 clp = find_confirmed_client(clid);
3172 status = nfserr_expired;
3174 /* We assume the client took too long to RENEW. */
3175 dprintk("nfsd4_renew: clientid not found!\n");
3178 status = nfserr_cb_path_down;
3179 if (!list_empty(&clp->cl_delegations)
3180 && clp->cl_cb_state != NFSD4_CB_UP)
3184 nfs4_unlock_state();
3188 static struct lock_manager nfsd4_manager = {
3191 static bool grace_ended;
3194 nfsd4_end_grace(void)
3196 /* do nothing if grace period already ended */
3200 dprintk("NFSD: end of grace period\n");
3202 nfsd4_record_grace_done(&init_net, boot_time);
3203 locks_end_grace(&nfsd4_manager);
3205 * Now that every NFSv4 client has had the chance to recover and
3206 * to see the (possibly new, possibly shorter) lease time, we
3207 * can safely set the next grace time to the current lease time:
3209 nfsd4_grace = nfsd4_lease;
3213 nfs4_laundromat(void)
3215 struct nfs4_client *clp;
3216 struct nfs4_openowner *oo;
3217 struct nfs4_delegation *dp;
3218 struct list_head *pos, *next, reaplist;
3219 time_t cutoff = get_seconds() - nfsd4_lease;
3220 time_t t, clientid_val = nfsd4_lease;
3221 time_t u, test_val = nfsd4_lease;
3225 dprintk("NFSD: laundromat service - starting\n");
3227 INIT_LIST_HEAD(&reaplist);
3228 spin_lock(&client_lock);
3229 list_for_each_safe(pos, next, &client_lru) {
3230 clp = list_entry(pos, struct nfs4_client, cl_lru);
3231 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3232 t = clp->cl_time - cutoff;
3233 if (clientid_val > t)
3237 if (atomic_read(&clp->cl_refcount)) {
3238 dprintk("NFSD: client in use (clientid %08x)\n",
3239 clp->cl_clientid.cl_id);
3242 unhash_client_locked(clp);
3243 list_add(&clp->cl_lru, &reaplist);
3245 spin_unlock(&client_lock);
3246 list_for_each_safe(pos, next, &reaplist) {
3247 clp = list_entry(pos, struct nfs4_client, cl_lru);
3248 dprintk("NFSD: purging unused client (clientid %08x)\n",
3249 clp->cl_clientid.cl_id);
3250 nfsd4_client_record_remove(clp);
3253 spin_lock(&recall_lock);
3254 list_for_each_safe(pos, next, &del_recall_lru) {
3255 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3256 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3257 u = dp->dl_time - cutoff;
3262 list_move(&dp->dl_recall_lru, &reaplist);
3264 spin_unlock(&recall_lock);
3265 list_for_each_safe(pos, next, &reaplist) {
3266 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3267 unhash_delegation(dp);
3269 test_val = nfsd4_lease;
3270 list_for_each_safe(pos, next, &close_lru) {
3271 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3272 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3273 u = oo->oo_time - cutoff;
3278 release_openowner(oo);
3280 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3281 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3282 nfs4_unlock_state();
3283 return clientid_val;
3286 static struct workqueue_struct *laundry_wq;
3287 static void laundromat_main(struct work_struct *);
3288 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3291 laundromat_main(struct work_struct *not_used)
3295 t = nfs4_laundromat();
3296 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3297 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3300 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3302 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3303 return nfserr_bad_stateid;
3308 STALE_STATEID(stateid_t *stateid)
3310 if (stateid->si_opaque.so_clid.cl_boot == boot_time)
3312 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3313 STATEID_VAL(stateid));
3318 access_permit_read(struct nfs4_ol_stateid *stp)
3320 return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
3321 test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
3322 test_access(NFS4_SHARE_ACCESS_WRITE, stp);
3326 access_permit_write(struct nfs4_ol_stateid *stp)
3328 return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
3329 test_access(NFS4_SHARE_ACCESS_BOTH, stp);
3333 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3335 __be32 status = nfserr_openmode;
3337 /* For lock stateid's, we test the parent open, not the lock: */
3338 if (stp->st_openstp)
3339 stp = stp->st_openstp;
3340 if ((flags & WR_STATE) && !access_permit_write(stp))
3342 if ((flags & RD_STATE) && !access_permit_read(stp))
3349 static inline __be32
3350 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3352 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3354 else if (locks_in_grace()) {
3355 /* Answer in remaining cases depends on existence of
3356 * conflicting state; so we must wait out the grace period. */
3357 return nfserr_grace;
3358 } else if (flags & WR_STATE)
3359 return nfs4_share_conflict(current_fh,
3360 NFS4_SHARE_DENY_WRITE);
3361 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3362 return nfs4_share_conflict(current_fh,
3363 NFS4_SHARE_DENY_READ);
3367 * Allow READ/WRITE during grace period on recovered state only for files
3368 * that are not able to provide mandatory locking.
3371 grace_disallows_io(struct inode *inode)
3373 return locks_in_grace() && mandatory_lock(inode);
3376 /* Returns true iff a is later than b: */
3377 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3379 return (s32)a->si_generation - (s32)b->si_generation > 0;
3382 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3385 * When sessions are used the stateid generation number is ignored
3388 if (has_session && in->si_generation == 0)
3391 if (in->si_generation == ref->si_generation)
3394 /* If the client sends us a stateid from the future, it's buggy: */
3395 if (stateid_generation_after(in, ref))
3396 return nfserr_bad_stateid;
3398 * However, we could see a stateid from the past, even from a
3399 * non-buggy client. For example, if the client sends a lock
3400 * while some IO is outstanding, the lock may bump si_generation
3401 * while the IO is still in flight. The client could avoid that
3402 * situation by waiting for responses on all the IO requests,
3403 * but better performance may result in retrying IO that
3404 * receives an old_stateid error if requests are rarely
3405 * reordered in flight:
3407 return nfserr_old_stateid;
3410 __be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3412 struct nfs4_stid *s;
3413 struct nfs4_ol_stateid *ols;
3416 if (STALE_STATEID(stateid))
3417 return nfserr_stale_stateid;
3419 s = find_stateid(cl, stateid);
3421 return nfserr_stale_stateid;
3422 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3425 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3427 ols = openlockstateid(s);
3428 if (ols->st_stateowner->so_is_open_owner
3429 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3430 return nfserr_bad_stateid;
3434 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
3436 struct nfs4_client *cl;
3438 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3439 return nfserr_bad_stateid;
3440 if (STALE_STATEID(stateid))
3441 return nfserr_stale_stateid;
3442 cl = find_confirmed_client(&stateid->si_opaque.so_clid);
3444 return nfserr_expired;
3445 *s = find_stateid_by_type(cl, stateid, typemask);
3447 return nfserr_bad_stateid;
3453 * Checks for stateid operations
3456 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3457 stateid_t *stateid, int flags, struct file **filpp)
3459 struct nfs4_stid *s;
3460 struct nfs4_ol_stateid *stp = NULL;
3461 struct nfs4_delegation *dp = NULL;
3462 struct svc_fh *current_fh = &cstate->current_fh;
3463 struct inode *ino = current_fh->fh_dentry->d_inode;
3469 if (grace_disallows_io(ino))
3470 return nfserr_grace;
3472 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3473 return check_special_stateids(current_fh, stateid, flags);
3475 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
3478 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3481 switch (s->sc_type) {
3482 case NFS4_DELEG_STID:
3483 dp = delegstateid(s);
3484 status = nfs4_check_delegmode(dp, flags);
3488 *filpp = dp->dl_file->fi_deleg_file;
3492 case NFS4_OPEN_STID:
3493 case NFS4_LOCK_STID:
3494 stp = openlockstateid(s);
3495 status = nfs4_check_fh(current_fh, stp);
3498 if (stp->st_stateowner->so_is_open_owner
3499 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3501 status = nfs4_check_openmode(stp, flags);
3505 if (flags & RD_STATE)
3506 *filpp = find_readable_file(stp->st_file);
3508 *filpp = find_writeable_file(stp->st_file);
3512 return nfserr_bad_stateid;
3520 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3522 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3523 return nfserr_locks_held;
3524 release_lock_stateid(stp);
3529 * Test if the stateid is valid
3532 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3533 struct nfsd4_test_stateid *test_stateid)
3535 struct nfsd4_test_stateid_id *stateid;
3536 struct nfs4_client *cl = cstate->session->se_client;
3539 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3540 stateid->ts_id_status = nfs4_validate_stateid(cl, &stateid->ts_id_stateid);
3541 nfs4_unlock_state();
3547 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3548 struct nfsd4_free_stateid *free_stateid)
3550 stateid_t *stateid = &free_stateid->fr_stateid;
3551 struct nfs4_stid *s;
3552 struct nfs4_client *cl = cstate->session->se_client;
3553 __be32 ret = nfserr_bad_stateid;
3556 s = find_stateid(cl, stateid);
3559 switch (s->sc_type) {
3560 case NFS4_DELEG_STID:
3561 ret = nfserr_locks_held;
3563 case NFS4_OPEN_STID:
3564 case NFS4_LOCK_STID:
3565 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3568 if (s->sc_type == NFS4_LOCK_STID)
3569 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3571 ret = nfserr_locks_held;
3574 ret = nfserr_bad_stateid;
3577 nfs4_unlock_state();
3584 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3585 RD_STATE : WR_STATE;
3588 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3590 struct svc_fh *current_fh = &cstate->current_fh;
3591 struct nfs4_stateowner *sop = stp->st_stateowner;
3594 status = nfsd4_check_seqid(cstate, sop, seqid);
3597 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3599 * "Closed" stateid's exist *only* to return
3600 * nfserr_replay_me from the previous step.
3602 return nfserr_bad_stateid;
3603 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3606 return nfs4_check_fh(current_fh, stp);
3610 * Checks for sequence id mutating operations.
3613 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3614 stateid_t *stateid, char typemask,
3615 struct nfs4_ol_stateid **stpp)
3618 struct nfs4_stid *s;
3620 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3621 seqid, STATEID_VAL(stateid));
3624 status = nfsd4_lookup_stateid(stateid, typemask, &s);
3627 *stpp = openlockstateid(s);
3628 cstate->replay_owner = (*stpp)->st_stateowner;
3630 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3633 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3636 struct nfs4_openowner *oo;
3638 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3639 NFS4_OPEN_STID, stpp);
3642 oo = openowner((*stpp)->st_stateowner);
3643 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3644 return nfserr_bad_stateid;
3649 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3650 struct nfsd4_open_confirm *oc)
3653 struct nfs4_openowner *oo;
3654 struct nfs4_ol_stateid *stp;
3656 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3657 (int)cstate->current_fh.fh_dentry->d_name.len,
3658 cstate->current_fh.fh_dentry->d_name.name);
3660 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3666 status = nfs4_preprocess_seqid_op(cstate,
3667 oc->oc_seqid, &oc->oc_req_stateid,
3668 NFS4_OPEN_STID, &stp);
3671 oo = openowner(stp->st_stateowner);
3672 status = nfserr_bad_stateid;
3673 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3675 oo->oo_flags |= NFS4_OO_CONFIRMED;
3676 update_stateid(&stp->st_stid.sc_stateid);
3677 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3678 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3679 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3681 nfsd4_client_record_create(oo->oo_owner.so_client);
3684 if (!cstate->replay_owner)
3685 nfs4_unlock_state();
3689 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3691 if (!test_access(access, stp))
3693 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3694 clear_access(access, stp);
3697 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3699 switch (to_access) {
3700 case NFS4_SHARE_ACCESS_READ:
3701 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3702 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3704 case NFS4_SHARE_ACCESS_WRITE:
3705 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3706 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3708 case NFS4_SHARE_ACCESS_BOTH:
3716 reset_union_bmap_deny(unsigned long deny, struct nfs4_ol_stateid *stp)
3719 for (i = 0; i < 4; i++) {
3720 if ((i & deny) != i)
3726 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3727 struct nfsd4_compound_state *cstate,
3728 struct nfsd4_open_downgrade *od)
3731 struct nfs4_ol_stateid *stp;
3733 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3734 (int)cstate->current_fh.fh_dentry->d_name.len,
3735 cstate->current_fh.fh_dentry->d_name.name);
3737 /* We don't yet support WANT bits: */
3738 if (od->od_deleg_want)
3739 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3743 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3744 &od->od_stateid, &stp);
3747 status = nfserr_inval;
3748 if (!test_access(od->od_share_access, stp)) {
3749 dprintk("NFSD: access not a subset current bitmap: 0x%lx, input access=%08x\n",
3750 stp->st_access_bmap, od->od_share_access);
3753 if (!test_deny(od->od_share_deny, stp)) {
3754 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3755 stp->st_deny_bmap, od->od_share_deny);
3758 nfs4_stateid_downgrade(stp, od->od_share_access);
3760 reset_union_bmap_deny(od->od_share_deny, stp);
3762 update_stateid(&stp->st_stid.sc_stateid);
3763 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3766 if (!cstate->replay_owner)
3767 nfs4_unlock_state();
3771 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3773 struct nfs4_openowner *oo;
3774 struct nfs4_ol_stateid *s;
3776 if (!so->so_is_open_owner)
3779 s = oo->oo_last_closed_stid;
3782 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3783 /* Release the last_closed_stid on the next seqid bump: */
3784 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3787 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3788 release_last_closed_stateid(oo);
3791 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3793 unhash_open_stateid(s);
3794 s->st_stid.sc_type = NFS4_CLOSED_STID;
3798 * nfs4_unlock_state() called after encode
3801 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3802 struct nfsd4_close *close)
3805 struct nfs4_openowner *oo;
3806 struct nfs4_ol_stateid *stp;
3808 dprintk("NFSD: nfsd4_close on file %.*s\n",
3809 (int)cstate->current_fh.fh_dentry->d_name.len,
3810 cstate->current_fh.fh_dentry->d_name.name);
3813 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3815 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3819 oo = openowner(stp->st_stateowner);
3821 update_stateid(&stp->st_stid.sc_stateid);
3822 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3824 nfsd4_close_open_stateid(stp);
3825 oo->oo_last_closed_stid = stp;
3827 /* place unused nfs4_stateowners on so_close_lru list to be
3828 * released by the laundromat service after the lease period
3829 * to enable us to handle CLOSE replay
3831 if (list_empty(&oo->oo_owner.so_stateids))
3832 move_to_close_lru(oo);
3834 if (!cstate->replay_owner)
3835 nfs4_unlock_state();
3840 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3841 struct nfsd4_delegreturn *dr)
3843 struct nfs4_delegation *dp;
3844 stateid_t *stateid = &dr->dr_stateid;
3845 struct nfs4_stid *s;
3846 struct inode *inode;
3849 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3851 inode = cstate->current_fh.fh_dentry->d_inode;
3854 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s);
3857 dp = delegstateid(s);
3858 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3862 unhash_delegation(dp);
3864 nfs4_unlock_state();
3870 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3872 #define LOCKOWNER_INO_HASH_BITS 8
3873 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3874 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3877 end_offset(u64 start, u64 len)
3882 return end >= start ? end: NFS4_MAX_UINT64;
3885 /* last octet in a range */
3887 last_byte_offset(u64 start, u64 len)
3893 return end > start ? end - 1: NFS4_MAX_UINT64;
3896 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3898 return (file_hashval(inode) + cl_id
3899 + opaque_hashval(ownername->data, ownername->len))
3900 & LOCKOWNER_INO_HASH_MASK;
3903 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3906 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3907 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3908 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3909 * locking, this prevents us from being completely protocol-compliant. The
3910 * real solution to this problem is to start using unsigned file offsets in
3911 * the VFS, but this is a very deep change!
3914 nfs4_transform_lock_offset(struct file_lock *lock)
3916 if (lock->fl_start < 0)
3917 lock->fl_start = OFFSET_MAX;
3918 if (lock->fl_end < 0)
3919 lock->fl_end = OFFSET_MAX;
3922 /* Hack!: For now, we're defining this just so we can use a pointer to it
3923 * as a unique cookie to identify our (NFSv4's) posix locks. */
3924 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3928 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3930 struct nfs4_lockowner *lo;
3932 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3933 lo = (struct nfs4_lockowner *) fl->fl_owner;
3934 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3935 lo->lo_owner.so_owner.len, GFP_KERNEL);
3936 if (!deny->ld_owner.data)
3937 /* We just don't care that much */
3939 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3940 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3943 deny->ld_owner.len = 0;
3944 deny->ld_owner.data = NULL;
3945 deny->ld_clientid.cl_boot = 0;
3946 deny->ld_clientid.cl_id = 0;
3948 deny->ld_start = fl->fl_start;
3949 deny->ld_length = NFS4_MAX_UINT64;
3950 if (fl->fl_end != NFS4_MAX_UINT64)
3951 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3952 deny->ld_type = NFS4_READ_LT;
3953 if (fl->fl_type != F_RDLCK)
3954 deny->ld_type = NFS4_WRITE_LT;
3957 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3959 struct nfs4_ol_stateid *lst;
3961 if (!same_owner_str(&lo->lo_owner, owner, clid))
3963 lst = list_first_entry(&lo->lo_owner.so_stateids,
3964 struct nfs4_ol_stateid, st_perstateowner);
3965 return lst->st_file->fi_inode == inode;
3968 static struct nfs4_lockowner *
3969 find_lockowner_str(struct inode *inode, clientid_t *clid,
3970 struct xdr_netobj *owner)
3972 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3973 struct nfs4_lockowner *lo;
3975 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3976 if (same_lockowner_ino(lo, inode, clid, owner))
3982 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3984 struct inode *inode = open_stp->st_file->fi_inode;
3985 unsigned int inohash = lockowner_ino_hashval(inode,
3986 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3988 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3989 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3990 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3994 * Alloc a lock owner structure.
3995 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3998 * strhashval = ownerstr_hashval
4001 static struct nfs4_lockowner *
4002 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
4003 struct nfs4_lockowner *lo;
4005 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
4008 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
4009 lo->lo_owner.so_is_open_owner = 0;
4010 /* It is the openowner seqid that will be incremented in encode in the
4011 * case of new lockowners; so increment the lock seqid manually: */
4012 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
4013 hash_lockowner(lo, strhashval, clp, open_stp);
4017 static struct nfs4_ol_stateid *
4018 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
4020 struct nfs4_ol_stateid *stp;
4021 struct nfs4_client *clp = lo->lo_owner.so_client;
4023 stp = nfs4_alloc_stateid(clp);
4026 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
4027 list_add(&stp->st_perfile, &fp->fi_stateids);
4028 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
4029 stp->st_stateowner = &lo->lo_owner;
4032 stp->st_access_bmap = 0;
4033 stp->st_deny_bmap = open_stp->st_deny_bmap;
4034 stp->st_openstp = open_stp;
4039 check_lock_length(u64 offset, u64 length)
4041 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4042 LOFF_OVERFLOW(offset, length)));
4045 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4047 struct nfs4_file *fp = lock_stp->st_file;
4048 int oflag = nfs4_access_to_omode(access);
4050 if (test_access(access, lock_stp))
4052 nfs4_file_get_access(fp, oflag);
4053 set_access(access, lock_stp);
4056 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)
4058 struct nfs4_file *fi = ost->st_file;
4059 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4060 struct nfs4_client *cl = oo->oo_owner.so_client;
4061 struct nfs4_lockowner *lo;
4062 unsigned int strhashval;
4064 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4066 if (!cstate->minorversion)
4067 return nfserr_bad_seqid;
4068 /* XXX: a lockowner always has exactly one stateid: */
4069 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4070 struct nfs4_ol_stateid, st_perstateowner);
4073 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4074 &lock->v.new.owner);
4075 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4077 return nfserr_jukebox;
4078 *lst = alloc_init_lock_stateid(lo, fi, ost);
4080 release_lockowner(lo);
4081 return nfserr_jukebox;
4091 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4092 struct nfsd4_lock *lock)
4094 struct nfs4_openowner *open_sop = NULL;
4095 struct nfs4_lockowner *lock_sop = NULL;
4096 struct nfs4_ol_stateid *lock_stp;
4097 struct file *filp = NULL;
4098 struct file_lock file_lock;
4099 struct file_lock conflock;
4101 bool new_state = false;
4105 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4106 (long long) lock->lk_offset,
4107 (long long) lock->lk_length);
4109 if (check_lock_length(lock->lk_offset, lock->lk_length))
4110 return nfserr_inval;
4112 if ((status = fh_verify(rqstp, &cstate->current_fh,
4113 S_IFREG, NFSD_MAY_LOCK))) {
4114 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4120 if (lock->lk_is_new) {
4122 * Client indicates that this is a new lockowner.
4123 * Use open owner and open stateid to create lock owner and
4126 struct nfs4_ol_stateid *open_stp = NULL;
4128 if (nfsd4_has_session(cstate))
4129 /* See rfc 5661 18.10.3: given clientid is ignored: */
4130 memcpy(&lock->v.new.clientid,
4131 &cstate->session->se_client->cl_clientid,
4132 sizeof(clientid_t));
4134 status = nfserr_stale_clientid;
4135 if (STALE_CLIENTID(&lock->lk_new_clientid))
4138 /* validate and update open stateid and open seqid */
4139 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4140 lock->lk_new_open_seqid,
4141 &lock->lk_new_open_stateid,
4145 open_sop = openowner(open_stp->st_stateowner);
4146 status = nfserr_bad_stateid;
4147 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4148 &lock->v.new.clientid))
4150 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4151 &lock_stp, &new_state);
4155 /* lock (lock owner + lock stateid) already exists */
4156 status = nfs4_preprocess_seqid_op(cstate,
4157 lock->lk_old_lock_seqid,
4158 &lock->lk_old_lock_stateid,
4159 NFS4_LOCK_STID, &lock_stp);
4163 lock_sop = lockowner(lock_stp->st_stateowner);
4165 lkflg = setlkflg(lock->lk_type);
4166 status = nfs4_check_openmode(lock_stp, lkflg);
4170 status = nfserr_grace;
4171 if (locks_in_grace() && !lock->lk_reclaim)
4173 status = nfserr_no_grace;
4174 if (!locks_in_grace() && lock->lk_reclaim)
4177 locks_init_lock(&file_lock);
4178 switch (lock->lk_type) {
4181 filp = find_readable_file(lock_stp->st_file);
4183 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4184 file_lock.fl_type = F_RDLCK;
4187 case NFS4_WRITEW_LT:
4188 filp = find_writeable_file(lock_stp->st_file);
4190 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4191 file_lock.fl_type = F_WRLCK;
4194 status = nfserr_inval;
4198 status = nfserr_openmode;
4201 file_lock.fl_owner = (fl_owner_t)lock_sop;
4202 file_lock.fl_pid = current->tgid;
4203 file_lock.fl_file = filp;
4204 file_lock.fl_flags = FL_POSIX;
4205 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4207 file_lock.fl_start = lock->lk_offset;
4208 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4209 nfs4_transform_lock_offset(&file_lock);
4212 * Try to lock the file in the VFS.
4213 * Note: locks.c uses the BKL to protect the inode's lock list.
4216 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4218 case 0: /* success! */
4219 update_stateid(&lock_stp->st_stid.sc_stateid);
4220 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4224 case (EAGAIN): /* conflock holds conflicting lock */
4225 status = nfserr_denied;
4226 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4227 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4230 status = nfserr_deadlock;
4233 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4234 status = nfserrno(err);
4238 if (status && new_state)
4239 release_lockowner(lock_sop);
4240 if (!cstate->replay_owner)
4241 nfs4_unlock_state();
4246 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4247 * so we do a temporary open here just to get an open file to pass to
4248 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4251 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4256 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4259 err = vfs_test_lock(file, lock);
4268 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4269 struct nfsd4_lockt *lockt)
4271 struct inode *inode;
4272 struct file_lock file_lock;
4273 struct nfs4_lockowner *lo;
4277 if (locks_in_grace())
4278 return nfserr_grace;
4280 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4281 return nfserr_inval;
4285 status = nfserr_stale_clientid;
4286 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4289 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4292 inode = cstate->current_fh.fh_dentry->d_inode;
4293 locks_init_lock(&file_lock);
4294 switch (lockt->lt_type) {
4297 file_lock.fl_type = F_RDLCK;
4300 case NFS4_WRITEW_LT:
4301 file_lock.fl_type = F_WRLCK;
4304 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4305 status = nfserr_inval;
4309 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4311 file_lock.fl_owner = (fl_owner_t)lo;
4312 file_lock.fl_pid = current->tgid;
4313 file_lock.fl_flags = FL_POSIX;
4315 file_lock.fl_start = lockt->lt_offset;
4316 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4318 nfs4_transform_lock_offset(&file_lock);
4321 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4323 status = nfserrno(error);
4326 if (file_lock.fl_type != F_UNLCK) {
4327 status = nfserr_denied;
4328 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4331 nfs4_unlock_state();
4336 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4337 struct nfsd4_locku *locku)
4339 struct nfs4_ol_stateid *stp;
4340 struct file *filp = NULL;
4341 struct file_lock file_lock;
4345 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4346 (long long) locku->lu_offset,
4347 (long long) locku->lu_length);
4349 if (check_lock_length(locku->lu_offset, locku->lu_length))
4350 return nfserr_inval;
4354 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4355 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4358 filp = find_any_file(stp->st_file);
4360 status = nfserr_lock_range;
4364 locks_init_lock(&file_lock);
4365 file_lock.fl_type = F_UNLCK;
4366 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4367 file_lock.fl_pid = current->tgid;
4368 file_lock.fl_file = filp;
4369 file_lock.fl_flags = FL_POSIX;
4370 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4371 file_lock.fl_start = locku->lu_offset;
4373 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4374 nfs4_transform_lock_offset(&file_lock);
4377 * Try to unlock the file in the VFS.
4379 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4381 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4385 * OK, unlock succeeded; the only thing left to do is update the stateid.
4387 update_stateid(&stp->st_stid.sc_stateid);
4388 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4391 if (!cstate->replay_owner)
4392 nfs4_unlock_state();
4396 status = nfserrno(err);
4402 * 1: locks held by lockowner
4403 * 0: no locks held by lockowner
4406 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4408 struct file_lock **flpp;
4409 struct inode *inode = filp->fi_inode;
4413 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4414 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4425 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4426 struct nfsd4_compound_state *cstate,
4427 struct nfsd4_release_lockowner *rlockowner)
4429 clientid_t *clid = &rlockowner->rl_clientid;
4430 struct nfs4_stateowner *sop;
4431 struct nfs4_lockowner *lo;
4432 struct nfs4_ol_stateid *stp;
4433 struct xdr_netobj *owner = &rlockowner->rl_owner;
4434 struct list_head matches;
4435 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4438 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4439 clid->cl_boot, clid->cl_id);
4441 /* XXX check for lease expiration */
4443 status = nfserr_stale_clientid;
4444 if (STALE_CLIENTID(clid))
4449 status = nfserr_locks_held;
4450 INIT_LIST_HEAD(&matches);
4452 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4453 if (sop->so_is_open_owner)
4455 if (!same_owner_str(sop, owner, clid))
4457 list_for_each_entry(stp, &sop->so_stateids,
4459 lo = lockowner(sop);
4460 if (check_for_locks(stp->st_file, lo))
4462 list_add(&lo->lo_list, &matches);
4465 /* Clients probably won't expect us to return with some (but not all)
4466 * of the lockowner state released; so don't release any until all
4467 * have been checked. */
4469 while (!list_empty(&matches)) {
4470 lo = list_entry(matches.next, struct nfs4_lockowner,
4472 /* unhash_stateowner deletes so_perclient only
4473 * for openowners. */
4474 list_del(&lo->lo_list);
4475 release_lockowner(lo);
4478 nfs4_unlock_state();
4482 static inline struct nfs4_client_reclaim *
4485 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4489 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4491 unsigned int strhashval = clientstr_hashval(name);
4492 struct nfs4_client *clp;
4494 clp = find_confirmed_client_by_str(name, strhashval);
4497 return test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
4501 * failure => all reset bets are off, nfserr_no_grace...
4504 nfs4_client_to_reclaim(const char *name)
4506 unsigned int strhashval;
4507 struct nfs4_client_reclaim *crp = NULL;
4509 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4510 crp = alloc_reclaim();
4513 strhashval = clientstr_hashval(name);
4514 INIT_LIST_HEAD(&crp->cr_strhash);
4515 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4516 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4517 reclaim_str_hashtbl_size++;
4522 nfs4_release_reclaim(void)
4524 struct nfs4_client_reclaim *crp = NULL;
4527 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4528 while (!list_empty(&reclaim_str_hashtbl[i])) {
4529 crp = list_entry(reclaim_str_hashtbl[i].next,
4530 struct nfs4_client_reclaim, cr_strhash);
4531 list_del(&crp->cr_strhash);
4533 reclaim_str_hashtbl_size--;
4536 BUG_ON(reclaim_str_hashtbl_size);
4540 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4541 struct nfs4_client_reclaim *
4542 nfsd4_find_reclaim_client(struct nfs4_client *clp)
4544 unsigned int strhashval;
4545 struct nfs4_client_reclaim *crp = NULL;
4547 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4548 clp->cl_name.len, clp->cl_name.data,
4551 /* find clp->cl_name in reclaim_str_hashtbl */
4552 strhashval = clientstr_hashval(clp->cl_recdir);
4553 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4554 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4562 * Called from OPEN. Look for clientid in reclaim list.
4565 nfs4_check_open_reclaim(clientid_t *clid)
4567 struct nfs4_client *clp;
4569 /* find clientid in conf_id_hashtbl */
4570 clp = find_confirmed_client(clid);
4572 return nfserr_reclaim_bad;
4574 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4577 #ifdef CONFIG_NFSD_FAULT_INJECTION
4579 void nfsd_forget_clients(u64 num)
4581 struct nfs4_client *clp, *next;
4585 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4586 nfsd4_client_record_remove(clp);
4591 nfs4_unlock_state();
4593 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4596 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4598 release_lockowner(lockowner(sop));
4601 static void release_openowner_sop(struct nfs4_stateowner *sop)
4603 release_openowner(openowner(sop));
4606 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4607 void (*release_sop)(struct nfs4_stateowner *))
4610 struct nfs4_stateowner *sop, *next;
4612 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4613 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4614 if (sop->so_is_open_owner != is_open_owner)
4624 void nfsd_forget_locks(u64 num)
4629 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4630 nfs4_unlock_state();
4632 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4635 void nfsd_forget_openowners(u64 num)
4640 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4641 nfs4_unlock_state();
4643 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4646 int nfsd_process_n_delegations(u64 num, void (*deleg_func)(struct nfs4_delegation *))
4649 struct nfs4_file *fp, *fnext;
4650 struct nfs4_delegation *dp, *dnext;
4652 for (i = 0; i < FILE_HASH_SIZE; i++) {
4653 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4654 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4665 void nfsd_forget_delegations(u64 num)
4670 count = nfsd_process_n_delegations(num, unhash_delegation);
4671 nfs4_unlock_state();
4673 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4676 void nfsd_recall_delegations(u64 num)
4681 spin_lock(&recall_lock);
4682 count = nfsd_process_n_delegations(num, nfsd_break_one_deleg);
4683 spin_unlock(&recall_lock);
4684 nfs4_unlock_state();
4686 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4689 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4691 /* initialization to perform at module load time: */
4694 nfs4_state_init(void)
4698 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4699 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4700 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4701 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4702 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4703 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4705 for (i = 0; i < SESSION_HASH_SIZE; i++)
4706 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4707 for (i = 0; i < FILE_HASH_SIZE; i++) {
4708 INIT_LIST_HEAD(&file_hashtbl[i]);
4710 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4711 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4713 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4714 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4715 INIT_LIST_HEAD(&close_lru);
4716 INIT_LIST_HEAD(&client_lru);
4717 INIT_LIST_HEAD(&del_recall_lru);
4718 reclaim_str_hashtbl_size = 0;
4722 * Since the lifetime of a delegation isn't limited to that of an open, a
4723 * client may quite reasonably hang on to a delegation as long as it has
4724 * the inode cached. This becomes an obvious problem the first time a
4725 * client's inode cache approaches the size of the server's total memory.
4727 * For now we avoid this problem by imposing a hard limit on the number
4728 * of delegations, which varies according to the server's memory size.
4731 set_max_delegations(void)
4734 * Allow at most 4 delegations per megabyte of RAM. Quick
4735 * estimates suggest that in the worst case (where every delegation
4736 * is for a different inode), a delegation could take about 1.5K,
4737 * giving a worst case usage of about 6% of memory.
4739 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4742 /* initialization to perform when the nfsd service is started: */
4745 nfs4_state_start(void)
4750 * FIXME: For now, we hang most of the pernet global stuff off of
4751 * init_net until nfsd is fully containerized. Eventually, we'll
4752 * need to pass a net pointer into this function, take a reference
4753 * to that instead and then do most of the rest of this on a per-net
4757 nfsd4_client_tracking_init(&init_net);
4758 boot_time = get_seconds();
4759 locks_start_grace(&nfsd4_manager);
4760 grace_ended = false;
4761 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4763 ret = set_callback_cred();
4768 laundry_wq = create_singlethread_workqueue("nfsd4");
4769 if (laundry_wq == NULL) {
4773 ret = nfsd4_create_callback_queue();
4775 goto out_free_laundry;
4776 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4777 set_max_delegations();
4780 destroy_workqueue(laundry_wq);
4782 nfsd4_client_tracking_exit(&init_net);
4788 __nfs4_state_shutdown(void)
4791 struct nfs4_client *clp = NULL;
4792 struct nfs4_delegation *dp = NULL;
4793 struct list_head *pos, *next, reaplist;
4795 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4796 while (!list_empty(&conf_id_hashtbl[i])) {
4797 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4800 while (!list_empty(&unconf_str_hashtbl[i])) {
4801 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4805 INIT_LIST_HEAD(&reaplist);
4806 spin_lock(&recall_lock);
4807 list_for_each_safe(pos, next, &del_recall_lru) {
4808 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4809 list_move(&dp->dl_recall_lru, &reaplist);
4811 spin_unlock(&recall_lock);
4812 list_for_each_safe(pos, next, &reaplist) {
4813 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4814 unhash_delegation(dp);
4817 nfsd4_client_tracking_exit(&init_net);
4822 nfs4_state_shutdown(void)
4824 cancel_delayed_work_sync(&laundromat_work);
4825 destroy_workqueue(laundry_wq);
4826 locks_end_grace(&nfsd4_manager);
4828 __nfs4_state_shutdown();
4829 nfs4_unlock_state();
4830 nfsd4_destroy_callback_queue();
4834 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4836 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4837 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4841 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4843 if (cstate->minorversion) {
4844 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4845 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4850 clear_current_stateid(struct nfsd4_compound_state *cstate)
4852 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4856 * functions to set current state id
4859 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4861 put_stateid(cstate, &odp->od_stateid);
4865 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4867 put_stateid(cstate, &open->op_stateid);
4871 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4873 put_stateid(cstate, &close->cl_stateid);
4877 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4879 put_stateid(cstate, &lock->lk_resp_stateid);
4883 * functions to consume current state id
4887 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4889 get_stateid(cstate, &odp->od_stateid);
4893 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4895 get_stateid(cstate, &drp->dr_stateid);
4899 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4901 get_stateid(cstate, &fsp->fr_stateid);
4905 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4907 get_stateid(cstate, &setattr->sa_stateid);
4911 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4913 get_stateid(cstate, &close->cl_stateid);
4917 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4919 get_stateid(cstate, &locku->lu_stateid);
4923 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4925 get_stateid(cstate, &read->rd_stateid);
4929 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4931 get_stateid(cstate, &write->wr_stateid);