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 set_access(unsigned int *access, unsigned long bmap) {
456 for (i = 1; i < 4; i++) {
457 if (test_bit(i, &bmap))
463 set_deny(unsigned int *deny, unsigned long bmap) {
467 for (i = 0; i < 4; i++) {
468 if (test_bit(i, &bmap))
474 test_share(struct nfs4_ol_stateid *stp, struct nfsd4_open *open) {
475 unsigned int access, deny;
477 set_access(&access, stp->st_access_bmap);
478 set_deny(&deny, stp->st_deny_bmap);
479 if ((access & open->op_share_deny) || (deny & open->op_share_access))
484 static int nfs4_access_to_omode(u32 access)
486 switch (access & NFS4_SHARE_ACCESS_BOTH) {
487 case NFS4_SHARE_ACCESS_READ:
489 case NFS4_SHARE_ACCESS_WRITE:
491 case NFS4_SHARE_ACCESS_BOTH:
497 static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
499 list_del(&stp->st_perfile);
500 list_del(&stp->st_perstateowner);
503 static void close_generic_stateid(struct nfs4_ol_stateid *stp)
507 if (stp->st_access_bmap) {
508 for (i = 1; i < 4; i++) {
509 if (test_bit(i, &stp->st_access_bmap))
510 nfs4_file_put_access(stp->st_file,
511 nfs4_access_to_omode(i));
512 __clear_bit(i, &stp->st_access_bmap);
515 put_nfs4_file(stp->st_file);
519 static void free_generic_stateid(struct nfs4_ol_stateid *stp)
521 kmem_cache_free(stateid_slab, stp);
524 static void release_lock_stateid(struct nfs4_ol_stateid *stp)
528 unhash_generic_stateid(stp);
529 unhash_stid(&stp->st_stid);
530 file = find_any_file(stp->st_file);
532 locks_remove_posix(file, (fl_owner_t)lockowner(stp->st_stateowner));
533 close_generic_stateid(stp);
534 free_generic_stateid(stp);
537 static void unhash_lockowner(struct nfs4_lockowner *lo)
539 struct nfs4_ol_stateid *stp;
541 list_del(&lo->lo_owner.so_strhash);
542 list_del(&lo->lo_perstateid);
543 list_del(&lo->lo_owner_ino_hash);
544 while (!list_empty(&lo->lo_owner.so_stateids)) {
545 stp = list_first_entry(&lo->lo_owner.so_stateids,
546 struct nfs4_ol_stateid, st_perstateowner);
547 release_lock_stateid(stp);
551 static void release_lockowner(struct nfs4_lockowner *lo)
553 unhash_lockowner(lo);
554 nfs4_free_lockowner(lo);
558 release_stateid_lockowners(struct nfs4_ol_stateid *open_stp)
560 struct nfs4_lockowner *lo;
562 while (!list_empty(&open_stp->st_lockowners)) {
563 lo = list_entry(open_stp->st_lockowners.next,
564 struct nfs4_lockowner, lo_perstateid);
565 release_lockowner(lo);
569 static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
571 unhash_generic_stateid(stp);
572 release_stateid_lockowners(stp);
573 close_generic_stateid(stp);
576 static void release_open_stateid(struct nfs4_ol_stateid *stp)
578 unhash_open_stateid(stp);
579 unhash_stid(&stp->st_stid);
580 free_generic_stateid(stp);
583 static void unhash_openowner(struct nfs4_openowner *oo)
585 struct nfs4_ol_stateid *stp;
587 list_del(&oo->oo_owner.so_strhash);
588 list_del(&oo->oo_perclient);
589 while (!list_empty(&oo->oo_owner.so_stateids)) {
590 stp = list_first_entry(&oo->oo_owner.so_stateids,
591 struct nfs4_ol_stateid, st_perstateowner);
592 release_open_stateid(stp);
596 static void release_last_closed_stateid(struct nfs4_openowner *oo)
598 struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
601 unhash_stid(&s->st_stid);
602 free_generic_stateid(s);
603 oo->oo_last_closed_stid = NULL;
607 static void release_openowner(struct nfs4_openowner *oo)
609 unhash_openowner(oo);
610 list_del(&oo->oo_close_lru);
611 release_last_closed_stateid(oo);
612 nfs4_free_openowner(oo);
615 #define SESSION_HASH_SIZE 512
616 static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
619 hash_sessionid(struct nfs4_sessionid *sessionid)
621 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
623 return sid->sequence % SESSION_HASH_SIZE;
628 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
630 u32 *ptr = (u32 *)(&sessionid->data[0]);
631 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
635 dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
642 gen_sessionid(struct nfsd4_session *ses)
644 struct nfs4_client *clp = ses->se_client;
645 struct nfsd4_sessionid *sid;
647 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
648 sid->clientid = clp->cl_clientid;
649 sid->sequence = current_sessionid++;
654 * The protocol defines ca_maxresponssize_cached to include the size of
655 * the rpc header, but all we need to cache is the data starting after
656 * the end of the initial SEQUENCE operation--the rest we regenerate
657 * each time. Therefore we can advertise a ca_maxresponssize_cached
658 * value that is the number of bytes in our cache plus a few additional
659 * bytes. In order to stay on the safe side, and not promise more than
660 * we can cache, those additional bytes must be the minimum possible: 24
661 * bytes of rpc header (xid through accept state, with AUTH_NULL
662 * verifier), 12 for the compound header (with zero-length tag), and 44
663 * for the SEQUENCE op response:
665 #define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
668 free_session_slots(struct nfsd4_session *ses)
672 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
673 kfree(ses->se_slots[i]);
677 * We don't actually need to cache the rpc and session headers, so we
678 * can allocate a little less for each slot:
680 static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
682 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
685 static int nfsd4_sanitize_slot_size(u32 size)
687 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
688 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
694 * XXX: If we run out of reserved DRC memory we could (up to a point)
695 * re-negotiate active sessions and reduce their slot usage to make
696 * room for new connections. For now we just fail the create session.
698 static int nfsd4_get_drc_mem(int slotsize, u32 num)
702 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
704 spin_lock(&nfsd_drc_lock);
705 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
706 nfsd_drc_max_mem - nfsd_drc_mem_used);
707 num = min_t(int, num, avail / slotsize);
708 nfsd_drc_mem_used += num * slotsize;
709 spin_unlock(&nfsd_drc_lock);
714 static void nfsd4_put_drc_mem(int slotsize, int num)
716 spin_lock(&nfsd_drc_lock);
717 nfsd_drc_mem_used -= slotsize * num;
718 spin_unlock(&nfsd_drc_lock);
721 static struct nfsd4_session *alloc_session(int slotsize, int numslots)
723 struct nfsd4_session *new;
726 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
727 + sizeof(struct nfsd4_session) > PAGE_SIZE);
728 mem = numslots * sizeof(struct nfsd4_slot *);
730 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
733 /* allocate each struct nfsd4_slot and data cache in one piece */
734 for (i = 0; i < numslots; i++) {
735 mem = sizeof(struct nfsd4_slot) + slotsize;
736 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
737 if (!new->se_slots[i])
743 kfree(new->se_slots[i]);
748 static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
750 u32 maxrpc = nfsd_serv->sv_max_mesg;
752 new->maxreqs = numslots;
753 new->maxresp_cached = min_t(u32, req->maxresp_cached,
754 slotsize + NFSD_MIN_HDR_SEQ_SZ);
755 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
756 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
757 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
760 static void free_conn(struct nfsd4_conn *c)
762 svc_xprt_put(c->cn_xprt);
766 static void nfsd4_conn_lost(struct svc_xpt_user *u)
768 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
769 struct nfs4_client *clp = c->cn_session->se_client;
771 spin_lock(&clp->cl_lock);
772 if (!list_empty(&c->cn_persession)) {
773 list_del(&c->cn_persession);
776 spin_unlock(&clp->cl_lock);
777 nfsd4_probe_callback(clp);
780 static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
782 struct nfsd4_conn *conn;
784 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
787 svc_xprt_get(rqstp->rq_xprt);
788 conn->cn_xprt = rqstp->rq_xprt;
789 conn->cn_flags = flags;
790 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
794 static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
796 conn->cn_session = ses;
797 list_add(&conn->cn_persession, &ses->se_conns);
800 static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
802 struct nfs4_client *clp = ses->se_client;
804 spin_lock(&clp->cl_lock);
805 __nfsd4_hash_conn(conn, ses);
806 spin_unlock(&clp->cl_lock);
809 static int nfsd4_register_conn(struct nfsd4_conn *conn)
811 conn->cn_xpt_user.callback = nfsd4_conn_lost;
812 return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
815 static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses, u32 dir)
817 struct nfsd4_conn *conn;
820 conn = alloc_conn(rqstp, dir);
822 return nfserr_jukebox;
823 nfsd4_hash_conn(conn, ses);
824 ret = nfsd4_register_conn(conn);
826 /* oops; xprt is already down: */
827 nfsd4_conn_lost(&conn->cn_xpt_user);
831 static __be32 nfsd4_new_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_session *ses)
833 u32 dir = NFS4_CDFC4_FORE;
835 if (ses->se_flags & SESSION4_BACK_CHAN)
836 dir |= NFS4_CDFC4_BACK;
838 return nfsd4_new_conn(rqstp, ses, dir);
841 /* must be called under client_lock */
842 static void nfsd4_del_conns(struct nfsd4_session *s)
844 struct nfs4_client *clp = s->se_client;
845 struct nfsd4_conn *c;
847 spin_lock(&clp->cl_lock);
848 while (!list_empty(&s->se_conns)) {
849 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
850 list_del_init(&c->cn_persession);
851 spin_unlock(&clp->cl_lock);
853 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
856 spin_lock(&clp->cl_lock);
858 spin_unlock(&clp->cl_lock);
861 static void free_session(struct kref *kref)
863 struct nfsd4_session *ses;
866 BUG_ON(!spin_is_locked(&client_lock));
867 ses = container_of(kref, struct nfsd4_session, se_ref);
868 nfsd4_del_conns(ses);
869 spin_lock(&nfsd_drc_lock);
870 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
871 nfsd_drc_mem_used -= mem;
872 spin_unlock(&nfsd_drc_lock);
873 free_session_slots(ses);
877 void nfsd4_put_session(struct nfsd4_session *ses)
879 spin_lock(&client_lock);
880 nfsd4_put_session_locked(ses);
881 spin_unlock(&client_lock);
884 static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
886 struct nfsd4_session *new;
887 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
888 int numslots, slotsize;
893 * Note decreasing slot size below client's request may
894 * make it difficult for client to function correctly, whereas
895 * decreasing the number of slots will (just?) affect
896 * performance. When short on memory we therefore prefer to
897 * decrease number of slots instead of their size.
899 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
900 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
904 new = alloc_session(slotsize, numslots);
906 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
909 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
911 new->se_client = clp;
914 INIT_LIST_HEAD(&new->se_conns);
916 new->se_cb_seq_nr = 1;
917 new->se_flags = cses->flags;
918 new->se_cb_prog = cses->callback_prog;
919 kref_init(&new->se_ref);
920 idx = hash_sessionid(&new->se_sessionid);
921 spin_lock(&client_lock);
922 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
923 spin_lock(&clp->cl_lock);
924 list_add(&new->se_perclnt, &clp->cl_sessions);
925 spin_unlock(&clp->cl_lock);
926 spin_unlock(&client_lock);
928 status = nfsd4_new_conn_from_crses(rqstp, new);
929 /* whoops: benny points out, status is ignored! (err, or bogus) */
931 spin_lock(&client_lock);
932 free_session(&new->se_ref);
933 spin_unlock(&client_lock);
936 if (cses->flags & SESSION4_BACK_CHAN) {
937 struct sockaddr *sa = svc_addr(rqstp);
939 * This is a little silly; with sessions there's no real
940 * use for the callback address. Use the peer address
941 * as a reasonable default for now, but consider fixing
942 * the rpc client not to require an address in the
945 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
946 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
948 nfsd4_probe_callback(clp);
952 /* caller must hold client_lock */
953 static struct nfsd4_session *
954 find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
956 struct nfsd4_session *elem;
959 dump_sessionid(__func__, sessionid);
960 idx = hash_sessionid(sessionid);
961 /* Search in the appropriate list */
962 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
963 if (!memcmp(elem->se_sessionid.data, sessionid->data,
964 NFS4_MAX_SESSIONID_LEN)) {
969 dprintk("%s: session not found\n", __func__);
973 /* caller must hold client_lock */
975 unhash_session(struct nfsd4_session *ses)
977 list_del(&ses->se_hash);
978 spin_lock(&ses->se_client->cl_lock);
979 list_del(&ses->se_perclnt);
980 spin_unlock(&ses->se_client->cl_lock);
983 /* must be called under the client_lock */
985 renew_client_locked(struct nfs4_client *clp)
987 if (is_client_expired(clp)) {
988 dprintk("%s: client (clientid %08x/%08x) already expired\n",
990 clp->cl_clientid.cl_boot,
991 clp->cl_clientid.cl_id);
995 dprintk("renewing client (clientid %08x/%08x)\n",
996 clp->cl_clientid.cl_boot,
997 clp->cl_clientid.cl_id);
998 list_move_tail(&clp->cl_lru, &client_lru);
999 clp->cl_time = get_seconds();
1003 renew_client(struct nfs4_client *clp)
1005 spin_lock(&client_lock);
1006 renew_client_locked(clp);
1007 spin_unlock(&client_lock);
1010 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1012 STALE_CLIENTID(clientid_t *clid)
1014 if (clid->cl_boot == boot_time)
1016 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1017 clid->cl_boot, clid->cl_id, boot_time);
1022 * XXX Should we use a slab cache ?
1023 * This type of memory management is somewhat inefficient, but we use it
1024 * anyway since SETCLIENTID is not a common operation.
1026 static struct nfs4_client *alloc_client(struct xdr_netobj name)
1028 struct nfs4_client *clp;
1030 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1033 clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1034 if (clp->cl_name.data == NULL) {
1038 clp->cl_name.len = name.len;
1043 free_client(struct nfs4_client *clp)
1045 BUG_ON(!spin_is_locked(&client_lock));
1046 while (!list_empty(&clp->cl_sessions)) {
1047 struct nfsd4_session *ses;
1048 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1050 list_del(&ses->se_perclnt);
1051 nfsd4_put_session_locked(ses);
1053 if (clp->cl_cred.cr_group_info)
1054 put_group_info(clp->cl_cred.cr_group_info);
1055 kfree(clp->cl_principal);
1056 kfree(clp->cl_name.data);
1061 release_session_client(struct nfsd4_session *session)
1063 struct nfs4_client *clp = session->se_client;
1065 if (!atomic_dec_and_lock(&clp->cl_refcount, &client_lock))
1067 if (is_client_expired(clp)) {
1069 session->se_client = NULL;
1071 renew_client_locked(clp);
1072 spin_unlock(&client_lock);
1075 /* must be called under the client_lock */
1077 unhash_client_locked(struct nfs4_client *clp)
1079 struct nfsd4_session *ses;
1081 mark_client_expired(clp);
1082 list_del(&clp->cl_lru);
1083 spin_lock(&clp->cl_lock);
1084 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1085 list_del_init(&ses->se_hash);
1086 spin_unlock(&clp->cl_lock);
1090 expire_client(struct nfs4_client *clp)
1092 struct nfs4_openowner *oo;
1093 struct nfs4_delegation *dp;
1094 struct list_head reaplist;
1096 INIT_LIST_HEAD(&reaplist);
1097 spin_lock(&recall_lock);
1098 while (!list_empty(&clp->cl_delegations)) {
1099 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1100 list_del_init(&dp->dl_perclnt);
1101 list_move(&dp->dl_recall_lru, &reaplist);
1103 spin_unlock(&recall_lock);
1104 while (!list_empty(&reaplist)) {
1105 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1106 unhash_delegation(dp);
1108 while (!list_empty(&clp->cl_openowners)) {
1109 oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1110 release_openowner(oo);
1112 nfsd4_shutdown_callback(clp);
1113 if (clp->cl_cb_conn.cb_xprt)
1114 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1115 list_del(&clp->cl_idhash);
1116 list_del(&clp->cl_strhash);
1117 spin_lock(&client_lock);
1118 unhash_client_locked(clp);
1119 if (atomic_read(&clp->cl_refcount) == 0)
1121 spin_unlock(&client_lock);
1124 static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1126 memcpy(target->cl_verifier.data, source->data,
1127 sizeof(target->cl_verifier.data));
1130 static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1132 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
1133 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
1136 static void copy_cred(struct svc_cred *target, struct svc_cred *source)
1138 target->cr_uid = source->cr_uid;
1139 target->cr_gid = source->cr_gid;
1140 target->cr_group_info = source->cr_group_info;
1141 get_group_info(target->cr_group_info);
1144 static int same_name(const char *n1, const char *n2)
1146 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1150 same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1152 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1156 same_clid(clientid_t *cl1, clientid_t *cl2)
1158 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1161 /* XXX what about NGROUP */
1163 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1165 return cr1->cr_uid == cr2->cr_uid;
1168 static void gen_clid(struct nfs4_client *clp)
1170 static u32 current_clientid = 1;
1172 clp->cl_clientid.cl_boot = boot_time;
1173 clp->cl_clientid.cl_id = current_clientid++;
1176 static void gen_confirm(struct nfs4_client *clp)
1181 verf[0] = (__be32)get_seconds();
1182 verf[1] = (__be32)i++;
1183 memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1186 static struct nfs4_stid *find_stateid(struct nfs4_client *cl, stateid_t *t)
1188 return idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1191 static struct nfs4_stid *find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1193 struct nfs4_stid *s;
1195 s = find_stateid(cl, t);
1198 if (typemask & s->sc_type)
1203 static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
1204 struct svc_rqst *rqstp, nfs4_verifier *verf)
1206 struct nfs4_client *clp;
1207 struct sockaddr *sa = svc_addr(rqstp);
1210 clp = alloc_client(name);
1214 INIT_LIST_HEAD(&clp->cl_sessions);
1216 princ = svc_gss_principal(rqstp);
1218 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
1219 if (clp->cl_principal == NULL) {
1220 spin_lock(&client_lock);
1222 spin_unlock(&client_lock);
1227 idr_init(&clp->cl_stateids);
1228 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1229 atomic_set(&clp->cl_refcount, 0);
1230 clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1231 INIT_LIST_HEAD(&clp->cl_idhash);
1232 INIT_LIST_HEAD(&clp->cl_strhash);
1233 INIT_LIST_HEAD(&clp->cl_openowners);
1234 INIT_LIST_HEAD(&clp->cl_delegations);
1235 INIT_LIST_HEAD(&clp->cl_lru);
1236 INIT_LIST_HEAD(&clp->cl_callbacks);
1237 spin_lock_init(&clp->cl_lock);
1238 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
1239 clp->cl_time = get_seconds();
1240 clear_bit(0, &clp->cl_cb_slot_busy);
1241 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1242 copy_verf(clp, verf);
1243 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1244 clp->cl_flavor = rqstp->rq_flavor;
1245 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1247 clp->cl_cb_session = NULL;
1252 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
1254 unsigned int idhashval;
1256 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
1257 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1258 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
1263 move_to_confirmed(struct nfs4_client *clp)
1265 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
1266 unsigned int strhashval;
1268 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
1269 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
1270 strhashval = clientstr_hashval(clp->cl_recdir);
1271 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1275 static struct nfs4_client *
1276 find_confirmed_client(clientid_t *clid)
1278 struct nfs4_client *clp;
1279 unsigned int idhashval = clientid_hashval(clid->cl_id);
1281 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
1282 if (same_clid(&clp->cl_clientid, clid)) {
1290 static struct nfs4_client *
1291 find_unconfirmed_client(clientid_t *clid)
1293 struct nfs4_client *clp;
1294 unsigned int idhashval = clientid_hashval(clid->cl_id);
1296 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
1297 if (same_clid(&clp->cl_clientid, clid))
1303 static bool clp_used_exchangeid(struct nfs4_client *clp)
1305 return clp->cl_exchange_flags != 0;
1308 static struct nfs4_client *
1309 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
1311 struct nfs4_client *clp;
1313 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
1314 if (same_name(clp->cl_recdir, dname))
1320 static struct nfs4_client *
1321 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
1323 struct nfs4_client *clp;
1325 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
1326 if (same_name(clp->cl_recdir, dname))
1333 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
1335 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1336 struct sockaddr *sa = svc_addr(rqstp);
1337 u32 scopeid = rpc_get_scope_id(sa);
1338 unsigned short expected_family;
1340 /* Currently, we only support tcp and tcp6 for the callback channel */
1341 if (se->se_callback_netid_len == 3 &&
1342 !memcmp(se->se_callback_netid_val, "tcp", 3))
1343 expected_family = AF_INET;
1344 else if (se->se_callback_netid_len == 4 &&
1345 !memcmp(se->se_callback_netid_val, "tcp6", 4))
1346 expected_family = AF_INET6;
1350 conn->cb_addrlen = rpc_uaddr2sockaddr(&init_net, se->se_callback_addr_val,
1351 se->se_callback_addr_len,
1352 (struct sockaddr *)&conn->cb_addr,
1353 sizeof(conn->cb_addr));
1355 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1358 if (conn->cb_addr.ss_family == AF_INET6)
1359 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1361 conn->cb_prog = se->se_callback_prog;
1362 conn->cb_ident = se->se_callback_ident;
1363 memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
1366 conn->cb_addr.ss_family = AF_UNSPEC;
1367 conn->cb_addrlen = 0;
1368 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1369 "will not receive delegations\n",
1370 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1376 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
1379 nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1381 struct nfsd4_slot *slot = resp->cstate.slot;
1384 dprintk("--> %s slot %p\n", __func__, slot);
1386 slot->sl_opcnt = resp->opcnt;
1387 slot->sl_status = resp->cstate.status;
1389 slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
1390 if (nfsd4_not_cached(resp)) {
1391 slot->sl_datalen = 0;
1394 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1395 base = (char *)resp->cstate.datap -
1396 (char *)resp->xbuf->head[0].iov_base;
1397 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1399 WARN("%s: sessions DRC could not cache compound\n", __func__);
1404 * Encode the replay sequence operation from the slot values.
1405 * If cachethis is FALSE encode the uncached rep error on the next
1406 * operation which sets resp->p and increments resp->opcnt for
1407 * nfs4svc_encode_compoundres.
1411 nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1412 struct nfsd4_compoundres *resp)
1414 struct nfsd4_op *op;
1415 struct nfsd4_slot *slot = resp->cstate.slot;
1417 /* Encode the replayed sequence operation */
1418 op = &args->ops[resp->opcnt - 1];
1419 nfsd4_encode_operation(resp, op);
1421 /* Return nfserr_retry_uncached_rep in next operation. */
1422 if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
1423 op = &args->ops[resp->opcnt++];
1424 op->status = nfserr_retry_uncached_rep;
1425 nfsd4_encode_operation(resp, op);
1431 * The sequence operation is not cached because we can use the slot and
1435 nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1436 struct nfsd4_sequence *seq)
1438 struct nfsd4_slot *slot = resp->cstate.slot;
1441 dprintk("--> %s slot %p\n", __func__, slot);
1443 /* Either returns 0 or nfserr_retry_uncached */
1444 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1445 if (status == nfserr_retry_uncached_rep)
1448 /* The sequence operation has been encoded, cstate->datap set. */
1449 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
1451 resp->opcnt = slot->sl_opcnt;
1452 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1453 status = slot->sl_status;
1459 * Set the exchange_id flags returned by the server.
1462 nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1464 /* pNFS is not supported */
1465 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1467 /* Referrals are supported, Migration is not. */
1468 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1470 /* set the wire flags to return to client. */
1471 clid->flags = new->cl_exchange_flags;
1475 nfsd4_exchange_id(struct svc_rqst *rqstp,
1476 struct nfsd4_compound_state *cstate,
1477 struct nfsd4_exchange_id *exid)
1479 struct nfs4_client *unconf, *conf, *new;
1481 unsigned int strhashval;
1482 char dname[HEXDIR_LEN];
1483 char addr_str[INET6_ADDRSTRLEN];
1484 nfs4_verifier verf = exid->verifier;
1485 struct sockaddr *sa = svc_addr(rqstp);
1487 rpc_ntop(sa, addr_str, sizeof(addr_str));
1488 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1489 "ip_addr=%s flags %x, spa_how %d\n",
1490 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1491 addr_str, exid->flags, exid->spa_how);
1493 if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
1494 return nfserr_inval;
1496 /* Currently only support SP4_NONE */
1497 switch (exid->spa_how) {
1501 return nfserr_serverfault;
1503 BUG(); /* checked by xdr code */
1505 return nfserr_serverfault; /* no excuse :-/ */
1508 status = nfs4_make_rec_clidname(dname, &exid->clname);
1513 strhashval = clientstr_hashval(dname);
1518 conf = find_confirmed_client_by_str(dname, strhashval);
1520 if (!clp_used_exchangeid(conf)) {
1521 status = nfserr_clid_inuse; /* XXX: ? */
1524 if (!same_verf(&verf, &conf->cl_verifier)) {
1525 /* 18.35.4 case 8 */
1526 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1527 status = nfserr_not_same;
1530 /* Client reboot: destroy old state */
1531 expire_client(conf);
1534 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1535 /* 18.35.4 case 9 */
1536 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1537 status = nfserr_perm;
1540 expire_client(conf);
1544 * Set bit when the owner id and verifier map to an already
1545 * confirmed client id (18.35.3).
1547 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1550 * Falling into 18.35.4 case 2, possible router replay.
1551 * Leave confirmed record intact and return same result.
1553 copy_verf(conf, &verf);
1558 /* 18.35.4 case 7 */
1559 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1560 status = nfserr_noent;
1564 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1567 * Possible retry or client restart. Per 18.35.4 case 4,
1568 * a new unconfirmed record should be generated regardless
1569 * of whether any properties have changed.
1571 expire_client(unconf);
1576 new = create_client(exid->clname, dname, rqstp, &verf);
1578 status = nfserr_jukebox;
1583 add_to_unconfirmed(new, strhashval);
1585 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1586 exid->clientid.cl_id = new->cl_clientid.cl_id;
1589 nfsd4_set_ex_flags(new, exid);
1591 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
1592 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
1596 nfs4_unlock_state();
1598 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1603 check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
1605 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1608 /* The slot is in use, and no response has been sent. */
1610 if (seqid == slot_seqid)
1611 return nfserr_jukebox;
1613 return nfserr_seq_misordered;
1615 /* Note unsigned 32-bit arithmetic handles wraparound: */
1616 if (likely(seqid == slot_seqid + 1))
1618 if (seqid == slot_seqid)
1619 return nfserr_replay_cache;
1620 return nfserr_seq_misordered;
1624 * Cache the create session result into the create session single DRC
1625 * slot cache by saving the xdr structure. sl_seqid has been set.
1626 * Do this for solo or embedded create session operations.
1629 nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1630 struct nfsd4_clid_slot *slot, __be32 nfserr)
1632 slot->sl_status = nfserr;
1633 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1637 nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1638 struct nfsd4_clid_slot *slot)
1640 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1641 return slot->sl_status;
1644 #define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
1645 2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
1646 1 + /* MIN tag is length with zero, only length */ \
1647 3 + /* version, opcount, opcode */ \
1648 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1649 /* seqid, slotID, slotID, cache */ \
1650 4 ) * sizeof(__be32))
1652 #define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
1653 2 + /* verifier: AUTH_NULL, length 0 */\
1655 1 + /* MIN tag is length with zero, only length */ \
1656 3 + /* opcount, opcode, opstatus*/ \
1657 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
1658 /* seqid, slotID, slotID, slotID, status */ \
1659 5 ) * sizeof(__be32))
1661 static bool check_forechannel_attrs(struct nfsd4_channel_attrs fchannel)
1663 return fchannel.maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ
1664 || fchannel.maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ;
1668 nfsd4_create_session(struct svc_rqst *rqstp,
1669 struct nfsd4_compound_state *cstate,
1670 struct nfsd4_create_session *cr_ses)
1672 struct sockaddr *sa = svc_addr(rqstp);
1673 struct nfs4_client *conf, *unconf;
1674 struct nfsd4_session *new;
1675 struct nfsd4_clid_slot *cs_slot = NULL;
1676 bool confirm_me = false;
1679 if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
1680 return nfserr_inval;
1683 unconf = find_unconfirmed_client(&cr_ses->clientid);
1684 conf = find_confirmed_client(&cr_ses->clientid);
1687 cs_slot = &conf->cl_cs_slot;
1688 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1689 if (status == nfserr_replay_cache) {
1690 dprintk("Got a create_session replay! seqid= %d\n",
1692 /* Return the cached reply status */
1693 status = nfsd4_replay_create_session(cr_ses, cs_slot);
1695 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
1696 status = nfserr_seq_misordered;
1697 dprintk("Sequence misordered!\n");
1698 dprintk("Expected seqid= %d but got seqid= %d\n",
1699 cs_slot->sl_seqid, cr_ses->seqid);
1702 } else if (unconf) {
1703 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1704 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
1705 status = nfserr_clid_inuse;
1709 cs_slot = &unconf->cl_cs_slot;
1710 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
1712 /* an unconfirmed replay returns misordered */
1713 status = nfserr_seq_misordered;
1720 status = nfserr_stale_clientid;
1725 * XXX: we should probably set this at creation time, and check
1726 * for consistent minorversion use throughout:
1728 conf->cl_minorversion = 1;
1730 * We do not support RDMA or persistent sessions
1732 cr_ses->flags &= ~SESSION4_PERSIST;
1733 cr_ses->flags &= ~SESSION4_RDMA;
1735 status = nfserr_toosmall;
1736 if (check_forechannel_attrs(cr_ses->fore_channel))
1739 status = nfserr_jukebox;
1740 new = alloc_init_session(rqstp, conf, cr_ses);
1744 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1745 NFS4_MAX_SESSIONID_LEN);
1746 memcpy(&cr_ses->fore_channel, &new->se_fchannel,
1747 sizeof(struct nfsd4_channel_attrs));
1748 cs_slot->sl_seqid++;
1749 cr_ses->seqid = cs_slot->sl_seqid;
1751 /* cache solo and embedded create sessions under the state lock */
1752 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1754 move_to_confirmed(conf);
1756 nfs4_unlock_state();
1757 dprintk("%s returns %d\n", __func__, ntohl(status));
1761 static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1763 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1764 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1766 return argp->opcnt == resp->opcnt;
1769 static __be32 nfsd4_map_bcts_dir(u32 *dir)
1772 case NFS4_CDFC4_FORE:
1773 case NFS4_CDFC4_BACK:
1775 case NFS4_CDFC4_FORE_OR_BOTH:
1776 case NFS4_CDFC4_BACK_OR_BOTH:
1777 *dir = NFS4_CDFC4_BOTH;
1780 return nfserr_inval;
1783 __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
1784 struct nfsd4_compound_state *cstate,
1785 struct nfsd4_bind_conn_to_session *bcts)
1789 if (!nfsd4_last_compound_op(rqstp))
1790 return nfserr_not_only_op;
1791 spin_lock(&client_lock);
1792 cstate->session = find_in_sessionid_hashtbl(&bcts->sessionid);
1793 /* Sorta weird: we only need the refcnt'ing because new_conn acquires
1794 * client_lock iself: */
1795 if (cstate->session) {
1796 nfsd4_get_session(cstate->session);
1797 atomic_inc(&cstate->session->se_client->cl_refcount);
1799 spin_unlock(&client_lock);
1800 if (!cstate->session)
1801 return nfserr_badsession;
1803 status = nfsd4_map_bcts_dir(&bcts->dir);
1805 nfsd4_new_conn(rqstp, cstate->session, bcts->dir);
1809 static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1813 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1817 nfsd4_destroy_session(struct svc_rqst *r,
1818 struct nfsd4_compound_state *cstate,
1819 struct nfsd4_destroy_session *sessionid)
1821 struct nfsd4_session *ses;
1822 __be32 status = nfserr_badsession;
1825 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1826 * - Should we return nfserr_back_chan_busy if waiting for
1827 * callbacks on to-be-destroyed session?
1828 * - Do we need to clear any callback info from previous session?
1831 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
1832 if (!nfsd4_last_compound_op(r))
1833 return nfserr_not_only_op;
1835 dump_sessionid(__func__, &sessionid->sessionid);
1836 spin_lock(&client_lock);
1837 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1839 spin_unlock(&client_lock);
1843 unhash_session(ses);
1844 spin_unlock(&client_lock);
1847 nfsd4_probe_callback_sync(ses->se_client);
1848 nfs4_unlock_state();
1850 spin_lock(&client_lock);
1851 nfsd4_del_conns(ses);
1852 nfsd4_put_session_locked(ses);
1853 spin_unlock(&client_lock);
1856 dprintk("%s returns %d\n", __func__, ntohl(status));
1860 static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1862 struct nfsd4_conn *c;
1864 list_for_each_entry(c, &s->se_conns, cn_persession) {
1865 if (c->cn_xprt == xpt) {
1872 static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1874 struct nfs4_client *clp = ses->se_client;
1875 struct nfsd4_conn *c;
1878 spin_lock(&clp->cl_lock);
1879 c = __nfsd4_find_conn(new->cn_xprt, ses);
1881 spin_unlock(&clp->cl_lock);
1885 __nfsd4_hash_conn(new, ses);
1886 spin_unlock(&clp->cl_lock);
1887 ret = nfsd4_register_conn(new);
1889 /* oops; xprt is already down: */
1890 nfsd4_conn_lost(&new->cn_xpt_user);
1894 static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
1896 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1898 return args->opcnt > session->se_fchannel.maxops;
1901 static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
1902 struct nfsd4_session *session)
1904 struct xdr_buf *xb = &rqstp->rq_arg;
1906 return xb->len > session->se_fchannel.maxreq_sz;
1910 nfsd4_sequence(struct svc_rqst *rqstp,
1911 struct nfsd4_compound_state *cstate,
1912 struct nfsd4_sequence *seq)
1914 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1915 struct nfsd4_session *session;
1916 struct nfsd4_slot *slot;
1917 struct nfsd4_conn *conn;
1920 if (resp->opcnt != 1)
1921 return nfserr_sequence_pos;
1924 * Will be either used or freed by nfsd4_sequence_check_conn
1927 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1929 return nfserr_jukebox;
1931 spin_lock(&client_lock);
1932 status = nfserr_badsession;
1933 session = find_in_sessionid_hashtbl(&seq->sessionid);
1937 status = nfserr_too_many_ops;
1938 if (nfsd4_session_too_many_ops(rqstp, session))
1941 status = nfserr_req_too_big;
1942 if (nfsd4_request_too_big(rqstp, session))
1945 status = nfserr_badslot;
1946 if (seq->slotid >= session->se_fchannel.maxreqs)
1949 slot = session->se_slots[seq->slotid];
1950 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1952 /* We do not negotiate the number of slots yet, so set the
1953 * maxslots to the session maxreqs which is used to encode
1954 * sr_highest_slotid and the sr_target_slot id to maxslots */
1955 seq->maxslots = session->se_fchannel.maxreqs;
1957 status = check_slot_seqid(seq->seqid, slot->sl_seqid,
1958 slot->sl_flags & NFSD4_SLOT_INUSE);
1959 if (status == nfserr_replay_cache) {
1960 status = nfserr_seq_misordered;
1961 if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
1963 cstate->slot = slot;
1964 cstate->session = session;
1965 /* Return the cached reply status and set cstate->status
1966 * for nfsd4_proc_compound processing */
1967 status = nfsd4_replay_cache_entry(resp, seq);
1968 cstate->status = nfserr_replay_cache;
1974 nfsd4_sequence_check_conn(conn, session);
1977 /* Success! bump slot seqid */
1978 slot->sl_seqid = seq->seqid;
1979 slot->sl_flags |= NFSD4_SLOT_INUSE;
1981 slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
1983 slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
1985 cstate->slot = slot;
1986 cstate->session = session;
1989 /* Hold a session reference until done processing the compound. */
1990 if (cstate->session) {
1991 struct nfs4_client *clp = session->se_client;
1993 nfsd4_get_session(cstate->session);
1994 atomic_inc(&clp->cl_refcount);
1995 switch (clp->cl_cb_state) {
1997 seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
1999 case NFSD4_CB_FAULT:
2000 seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2003 seq->status_flags = 0;
2007 spin_unlock(&client_lock);
2008 dprintk("%s: return %d\n", __func__, ntohl(status));
2012 static inline bool has_resources(struct nfs4_client *clp)
2014 return !list_empty(&clp->cl_openowners)
2015 || !list_empty(&clp->cl_delegations)
2016 || !list_empty(&clp->cl_sessions);
2020 nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2022 struct nfs4_client *conf, *unconf, *clp;
2026 unconf = find_unconfirmed_client(&dc->clientid);
2027 conf = find_confirmed_client(&dc->clientid);
2032 if (!is_client_expired(conf) && has_resources(conf)) {
2033 status = nfserr_clientid_busy;
2037 /* rfc5661 18.50.3 */
2038 if (cstate->session && conf == cstate->session->se_client) {
2039 status = nfserr_clientid_busy;
2045 status = nfserr_stale_clientid;
2051 nfs4_unlock_state();
2052 dprintk("%s return %d\n", __func__, ntohl(status));
2057 nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2061 if (rc->rca_one_fs) {
2062 if (!cstate->current_fh.fh_dentry)
2063 return nfserr_nofilehandle;
2065 * We don't take advantage of the rca_one_fs case.
2066 * That's OK, it's optional, we can safely ignore it.
2072 status = nfserr_complete_already;
2073 if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2074 &cstate->session->se_client->cl_flags))
2077 status = nfserr_stale_clientid;
2078 if (is_client_expired(cstate->session->se_client))
2080 * The following error isn't really legal.
2081 * But we only get here if the client just explicitly
2082 * destroyed the client. Surely it no longer cares what
2083 * error it gets back on an operation for the dead
2089 nfsd4_client_record_create(cstate->session->se_client);
2091 nfs4_unlock_state();
2096 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2097 struct nfsd4_setclientid *setclid)
2099 struct xdr_netobj clname = setclid->se_name;
2100 nfs4_verifier clverifier = setclid->se_verf;
2101 unsigned int strhashval;
2102 struct nfs4_client *conf, *unconf, *new;
2104 char dname[HEXDIR_LEN];
2106 status = nfs4_make_rec_clidname(dname, &clname);
2111 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2112 * We get here on a DRC miss.
2115 strhashval = clientstr_hashval(dname);
2118 conf = find_confirmed_client_by_str(dname, strhashval);
2120 /* RFC 3530 14.2.33 CASE 0: */
2121 status = nfserr_clid_inuse;
2122 if (clp_used_exchangeid(conf))
2124 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
2125 char addr_str[INET6_ADDRSTRLEN];
2126 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
2128 dprintk("NFSD: setclientid: string in use by client "
2129 "at %s\n", addr_str);
2134 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
2135 * has a description of SETCLIENTID request processing consisting
2136 * of 5 bullet points, labeled as CASE0 - CASE4 below.
2138 unconf = find_unconfirmed_client_by_str(dname, strhashval);
2139 status = nfserr_jukebox;
2142 * RFC 3530 14.2.33 CASE 4:
2143 * placed first, because it is the normal case
2146 expire_client(unconf);
2147 new = create_client(clname, dname, rqstp, &clverifier);
2151 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
2153 * RFC 3530 14.2.33 CASE 1:
2154 * probable callback update
2157 /* Note this is removing unconfirmed {*x***},
2158 * which is stronger than RFC recommended {vxc**}.
2159 * This has the advantage that there is at most
2160 * one {*x***} in either list at any time.
2162 expire_client(unconf);
2164 new = create_client(clname, dname, rqstp, &clverifier);
2167 copy_clid(new, conf);
2168 } else if (!unconf) {
2170 * RFC 3530 14.2.33 CASE 2:
2171 * probable client reboot; state will be removed if
2174 new = create_client(clname, dname, rqstp, &clverifier);
2180 * RFC 3530 14.2.33 CASE 3:
2181 * probable client reboot; state will be removed if
2184 expire_client(unconf);
2185 new = create_client(clname, dname, rqstp, &clverifier);
2191 * XXX: we should probably set this at creation time, and check
2192 * for consistent minorversion use throughout:
2194 new->cl_minorversion = 0;
2195 gen_callback(new, setclid, rqstp);
2196 add_to_unconfirmed(new, strhashval);
2197 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
2198 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
2199 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
2202 nfs4_unlock_state();
2208 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
2209 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
2210 * bullets, labeled as CASE1 - CASE4 below.
2213 nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
2214 struct nfsd4_compound_state *cstate,
2215 struct nfsd4_setclientid_confirm *setclientid_confirm)
2217 struct sockaddr *sa = svc_addr(rqstp);
2218 struct nfs4_client *conf, *unconf;
2219 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
2220 clientid_t * clid = &setclientid_confirm->sc_clientid;
2223 if (STALE_CLIENTID(clid))
2224 return nfserr_stale_clientid;
2226 * XXX The Duplicate Request Cache (DRC) has been checked (??)
2227 * We get here on a DRC miss.
2232 conf = find_confirmed_client(clid);
2233 unconf = find_unconfirmed_client(clid);
2235 status = nfserr_clid_inuse;
2236 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
2238 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
2242 * section 14.2.34 of RFC 3530 has a description of
2243 * SETCLIENTID_CONFIRM request processing consisting
2244 * of 4 bullet points, labeled as CASE1 - CASE4 below.
2246 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
2248 * RFC 3530 14.2.34 CASE 1:
2251 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
2252 status = nfserr_clid_inuse;
2254 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
2255 nfsd4_probe_callback(conf);
2256 expire_client(unconf);
2260 } else if (conf && !unconf) {
2262 * RFC 3530 14.2.34 CASE 2:
2263 * probable retransmitted request; play it safe and
2266 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
2267 status = nfserr_clid_inuse;
2270 } else if (!conf && unconf
2271 && same_verf(&unconf->cl_confirm, &confirm)) {
2273 * RFC 3530 14.2.34 CASE 3:
2274 * Normal case; new or rebooted client:
2276 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
2277 status = nfserr_clid_inuse;
2280 clientstr_hashval(unconf->cl_recdir);
2281 conf = find_confirmed_client_by_str(unconf->cl_recdir,
2284 nfsd4_client_record_remove(conf);
2285 expire_client(conf);
2287 move_to_confirmed(unconf);
2289 nfsd4_probe_callback(conf);
2292 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
2293 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
2296 * RFC 3530 14.2.34 CASE 4:
2297 * Client probably hasn't noticed that we rebooted yet.
2299 status = nfserr_stale_clientid;
2301 /* check that we have hit one of the cases...*/
2302 status = nfserr_clid_inuse;
2305 nfs4_unlock_state();
2309 static struct nfs4_file *nfsd4_alloc_file(void)
2311 return kmem_cache_alloc(file_slab, GFP_KERNEL);
2314 /* OPEN Share state helper functions */
2315 static void nfsd4_init_file(struct nfs4_file *fp, struct inode *ino)
2317 unsigned int hashval = file_hashval(ino);
2319 atomic_set(&fp->fi_ref, 1);
2320 INIT_LIST_HEAD(&fp->fi_hash);
2321 INIT_LIST_HEAD(&fp->fi_stateids);
2322 INIT_LIST_HEAD(&fp->fi_delegations);
2323 fp->fi_inode = igrab(ino);
2324 fp->fi_had_conflict = false;
2325 fp->fi_lease = NULL;
2326 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
2327 memset(fp->fi_access, 0, sizeof(fp->fi_access));
2328 spin_lock(&recall_lock);
2329 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
2330 spin_unlock(&recall_lock);
2334 nfsd4_free_slab(struct kmem_cache **slab)
2338 kmem_cache_destroy(*slab);
2343 nfsd4_free_slabs(void)
2345 nfsd4_free_slab(&openowner_slab);
2346 nfsd4_free_slab(&lockowner_slab);
2347 nfsd4_free_slab(&file_slab);
2348 nfsd4_free_slab(&stateid_slab);
2349 nfsd4_free_slab(&deleg_slab);
2353 nfsd4_init_slabs(void)
2355 openowner_slab = kmem_cache_create("nfsd4_openowners",
2356 sizeof(struct nfs4_openowner), 0, 0, NULL);
2357 if (openowner_slab == NULL)
2359 lockowner_slab = kmem_cache_create("nfsd4_lockowners",
2360 sizeof(struct nfs4_openowner), 0, 0, NULL);
2361 if (lockowner_slab == NULL)
2363 file_slab = kmem_cache_create("nfsd4_files",
2364 sizeof(struct nfs4_file), 0, 0, NULL);
2365 if (file_slab == NULL)
2367 stateid_slab = kmem_cache_create("nfsd4_stateids",
2368 sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
2369 if (stateid_slab == NULL)
2371 deleg_slab = kmem_cache_create("nfsd4_delegations",
2372 sizeof(struct nfs4_delegation), 0, 0, NULL);
2373 if (deleg_slab == NULL)
2378 dprintk("nfsd4: out of memory while initializing nfsv4\n");
2382 void nfs4_free_openowner(struct nfs4_openowner *oo)
2384 kfree(oo->oo_owner.so_owner.data);
2385 kmem_cache_free(openowner_slab, oo);
2388 void nfs4_free_lockowner(struct nfs4_lockowner *lo)
2390 kfree(lo->lo_owner.so_owner.data);
2391 kmem_cache_free(lockowner_slab, lo);
2394 static void init_nfs4_replay(struct nfs4_replay *rp)
2396 rp->rp_status = nfserr_serverfault;
2398 rp->rp_buf = rp->rp_ibuf;
2401 static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
2403 struct nfs4_stateowner *sop;
2405 sop = kmem_cache_alloc(slab, GFP_KERNEL);
2409 sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
2410 if (!sop->so_owner.data) {
2411 kmem_cache_free(slab, sop);
2414 sop->so_owner.len = owner->len;
2416 INIT_LIST_HEAD(&sop->so_stateids);
2417 sop->so_client = clp;
2418 init_nfs4_replay(&sop->so_replay);
2422 static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
2424 list_add(&oo->oo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
2425 list_add(&oo->oo_perclient, &clp->cl_openowners);
2428 static struct nfs4_openowner *
2429 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
2430 struct nfs4_openowner *oo;
2432 oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
2435 oo->oo_owner.so_is_open_owner = 1;
2436 oo->oo_owner.so_seqid = open->op_seqid;
2437 oo->oo_flags = NFS4_OO_NEW;
2439 oo->oo_last_closed_stid = NULL;
2440 INIT_LIST_HEAD(&oo->oo_close_lru);
2441 hash_openowner(oo, clp, strhashval);
2445 static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
2446 struct nfs4_openowner *oo = open->op_openowner;
2447 struct nfs4_client *clp = oo->oo_owner.so_client;
2449 init_stid(&stp->st_stid, clp, NFS4_OPEN_STID);
2450 INIT_LIST_HEAD(&stp->st_lockowners);
2451 list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
2452 list_add(&stp->st_perfile, &fp->fi_stateids);
2453 stp->st_stateowner = &oo->oo_owner;
2456 stp->st_access_bmap = 0;
2457 stp->st_deny_bmap = 0;
2458 __set_bit(open->op_share_access, &stp->st_access_bmap);
2459 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2460 stp->st_openstp = NULL;
2464 move_to_close_lru(struct nfs4_openowner *oo)
2466 dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
2468 list_move_tail(&oo->oo_close_lru, &close_lru);
2469 oo->oo_time = get_seconds();
2473 same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
2476 return (sop->so_owner.len == owner->len) &&
2477 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
2478 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
2481 static struct nfs4_openowner *
2482 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
2484 struct nfs4_stateowner *so;
2485 struct nfs4_openowner *oo;
2487 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
2488 if (!so->so_is_open_owner)
2490 if (same_owner_str(so, &open->op_owner, &open->op_clientid)) {
2492 renew_client(oo->oo_owner.so_client);
2499 /* search file_hashtbl[] for file */
2500 static struct nfs4_file *
2501 find_file(struct inode *ino)
2503 unsigned int hashval = file_hashval(ino);
2504 struct nfs4_file *fp;
2506 spin_lock(&recall_lock);
2507 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
2508 if (fp->fi_inode == ino) {
2510 spin_unlock(&recall_lock);
2514 spin_unlock(&recall_lock);
2519 * Called to check deny when READ with all zero stateid or
2520 * WRITE with all zero or all one stateid
2523 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2525 struct inode *ino = current_fh->fh_dentry->d_inode;
2526 struct nfs4_file *fp;
2527 struct nfs4_ol_stateid *stp;
2530 dprintk("NFSD: nfs4_share_conflict\n");
2532 fp = find_file(ino);
2535 ret = nfserr_locked;
2536 /* Search for conflicting share reservations */
2537 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2538 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2539 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2548 static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
2550 /* We're assuming the state code never drops its reference
2551 * without first removing the lease. Since we're in this lease
2552 * callback (and since the lease code is serialized by the kernel
2553 * lock) we know the server hasn't removed the lease yet, we know
2554 * it's safe to take a reference: */
2555 atomic_inc(&dp->dl_count);
2557 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2559 /* only place dl_time is set. protected by lock_flocks*/
2560 dp->dl_time = get_seconds();
2562 nfsd4_cb_recall(dp);
2565 /* Called from break_lease() with lock_flocks() held. */
2566 static void nfsd_break_deleg_cb(struct file_lock *fl)
2568 struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
2569 struct nfs4_delegation *dp;
2572 /* We assume break_lease is only called once per lease: */
2573 BUG_ON(fp->fi_had_conflict);
2575 * We don't want the locks code to timeout the lease for us;
2576 * we'll remove it ourself if a delegation isn't returned
2579 fl->fl_break_time = 0;
2581 spin_lock(&recall_lock);
2582 fp->fi_had_conflict = true;
2583 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
2584 nfsd_break_one_deleg(dp);
2585 spin_unlock(&recall_lock);
2589 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2592 return lease_modify(onlist, arg);
2597 static const struct lock_manager_operations nfsd_lease_mng_ops = {
2598 .lm_break = nfsd_break_deleg_cb,
2599 .lm_change = nfsd_change_deleg_cb,
2602 static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
2604 if (nfsd4_has_session(cstate))
2606 if (seqid == so->so_seqid - 1)
2607 return nfserr_replay_me;
2608 if (seqid == so->so_seqid)
2610 return nfserr_bad_seqid;
2614 nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2615 struct nfsd4_open *open)
2617 clientid_t *clientid = &open->op_clientid;
2618 struct nfs4_client *clp = NULL;
2619 unsigned int strhashval;
2620 struct nfs4_openowner *oo = NULL;
2623 if (STALE_CLIENTID(&open->op_clientid))
2624 return nfserr_stale_clientid;
2626 * In case we need it later, after we've already created the
2627 * file and don't want to risk a further failure:
2629 open->op_file = nfsd4_alloc_file();
2630 if (open->op_file == NULL)
2631 return nfserr_jukebox;
2633 strhashval = ownerstr_hashval(clientid->cl_id, &open->op_owner);
2634 oo = find_openstateowner_str(strhashval, open);
2635 open->op_openowner = oo;
2637 clp = find_confirmed_client(clientid);
2639 return nfserr_expired;
2642 if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
2643 /* Replace unconfirmed owners without checking for replay. */
2644 clp = oo->oo_owner.so_client;
2645 release_openowner(oo);
2646 open->op_openowner = NULL;
2649 status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
2652 clp = oo->oo_owner.so_client;
2655 oo = alloc_init_open_stateowner(strhashval, clp, open);
2657 return nfserr_jukebox;
2658 open->op_openowner = oo;
2660 open->op_stp = nfs4_alloc_stateid(clp);
2662 return nfserr_jukebox;
2666 static inline __be32
2667 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2669 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2670 return nfserr_openmode;
2675 static int share_access_to_flags(u32 share_access)
2677 return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
2680 static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
2682 struct nfs4_stid *ret;
2684 ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
2687 return delegstateid(ret);
2690 static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
2692 return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
2693 open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
2697 nfs4_check_deleg(struct nfs4_client *cl, struct nfs4_file *fp, struct nfsd4_open *open,
2698 struct nfs4_delegation **dp)
2701 __be32 status = nfserr_bad_stateid;
2703 *dp = find_deleg_stateid(cl, &open->op_delegate_stateid);
2706 flags = share_access_to_flags(open->op_share_access);
2707 status = nfs4_check_delegmode(*dp, flags);
2711 if (!nfsd4_is_deleg_cur(open))
2715 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2720 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_ol_stateid **stpp)
2722 struct nfs4_ol_stateid *local;
2723 struct nfs4_openowner *oo = open->op_openowner;
2725 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
2726 /* ignore lock owners */
2727 if (local->st_stateowner->so_is_open_owner == 0)
2729 /* remember if we have seen this open owner */
2730 if (local->st_stateowner == &oo->oo_owner)
2732 /* check for conflicting share reservations */
2733 if (!test_share(local, open))
2734 return nfserr_share_denied;
2739 static void nfs4_free_stateid(struct nfs4_ol_stateid *s)
2741 kmem_cache_free(stateid_slab, s);
2744 static inline int nfs4_access_to_access(u32 nfs4_access)
2748 if (nfs4_access & NFS4_SHARE_ACCESS_READ)
2749 flags |= NFSD_MAY_READ;
2750 if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
2751 flags |= NFSD_MAY_WRITE;
2755 static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
2756 struct svc_fh *cur_fh, struct nfsd4_open *open)
2759 int oflag = nfs4_access_to_omode(open->op_share_access);
2760 int access = nfs4_access_to_access(open->op_share_access);
2762 if (!fp->fi_fds[oflag]) {
2763 status = nfsd_open(rqstp, cur_fh, S_IFREG, access,
2764 &fp->fi_fds[oflag]);
2768 nfs4_file_get_access(fp, oflag);
2773 static inline __be32
2774 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2775 struct nfsd4_open *open)
2777 struct iattr iattr = {
2778 .ia_valid = ATTR_SIZE,
2781 if (!open->op_truncate)
2783 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
2784 return nfserr_inval;
2785 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2789 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)
2791 u32 op_share_access = open->op_share_access;
2795 new_access = !test_bit(op_share_access, &stp->st_access_bmap);
2797 status = nfs4_get_vfs_file(rqstp, fp, cur_fh, open);
2801 status = nfsd4_truncate(rqstp, cur_fh, open);
2804 int oflag = nfs4_access_to_omode(op_share_access);
2805 nfs4_file_put_access(fp, oflag);
2809 /* remember the open */
2810 __set_bit(op_share_access, &stp->st_access_bmap);
2811 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
2818 nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
2820 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
2823 /* Should we give out recallable state?: */
2824 static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
2826 if (clp->cl_cb_state == NFSD4_CB_UP)
2829 * In the sessions case, since we don't have to establish a
2830 * separate connection for callbacks, we assume it's OK
2831 * until we hear otherwise:
2833 return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
2836 static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp, int flag)
2838 struct file_lock *fl;
2840 fl = locks_alloc_lock();
2843 locks_init_lock(fl);
2844 fl->fl_lmops = &nfsd_lease_mng_ops;
2845 fl->fl_flags = FL_LEASE;
2846 fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
2847 fl->fl_end = OFFSET_MAX;
2848 fl->fl_owner = (fl_owner_t)(dp->dl_file);
2849 fl->fl_pid = current->tgid;
2853 static int nfs4_setlease(struct nfs4_delegation *dp, int flag)
2855 struct nfs4_file *fp = dp->dl_file;
2856 struct file_lock *fl;
2859 fl = nfs4_alloc_init_lease(dp, flag);
2862 fl->fl_file = find_readable_file(fp);
2863 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2864 status = vfs_setlease(fl->fl_file, fl->fl_type, &fl);
2866 list_del_init(&dp->dl_perclnt);
2867 locks_free_lock(fl);
2871 fp->fi_deleg_file = fl->fl_file;
2872 get_file(fp->fi_deleg_file);
2873 atomic_set(&fp->fi_delegees, 1);
2874 list_add(&dp->dl_perfile, &fp->fi_delegations);
2878 static int nfs4_set_delegation(struct nfs4_delegation *dp, int flag)
2880 struct nfs4_file *fp = dp->dl_file;
2883 return nfs4_setlease(dp, flag);
2884 spin_lock(&recall_lock);
2885 if (fp->fi_had_conflict) {
2886 spin_unlock(&recall_lock);
2889 atomic_inc(&fp->fi_delegees);
2890 list_add(&dp->dl_perfile, &fp->fi_delegations);
2891 spin_unlock(&recall_lock);
2892 list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
2896 static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
2898 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2899 if (status == -EAGAIN)
2900 open->op_why_no_deleg = WND4_CONTENTION;
2902 open->op_why_no_deleg = WND4_RESOURCE;
2903 switch (open->op_deleg_want) {
2904 case NFS4_SHARE_WANT_READ_DELEG:
2905 case NFS4_SHARE_WANT_WRITE_DELEG:
2906 case NFS4_SHARE_WANT_ANY_DELEG:
2908 case NFS4_SHARE_WANT_CANCEL:
2909 open->op_why_no_deleg = WND4_CANCELLED;
2911 case NFS4_SHARE_WANT_NO_DELEG:
2912 BUG(); /* not supposed to get here */
2918 * Attempt to hand out a delegation.
2921 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
2923 struct nfs4_delegation *dp;
2924 struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
2926 int status = 0, flag = 0;
2928 cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
2929 flag = NFS4_OPEN_DELEGATE_NONE;
2930 open->op_recall = 0;
2931 switch (open->op_claim_type) {
2932 case NFS4_OPEN_CLAIM_PREVIOUS:
2934 open->op_recall = 1;
2935 flag = open->op_delegate_type;
2936 if (flag == NFS4_OPEN_DELEGATE_NONE)
2939 case NFS4_OPEN_CLAIM_NULL:
2940 /* Let's not give out any delegations till everyone's
2941 * had the chance to reclaim theirs.... */
2942 if (locks_in_grace())
2944 if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
2946 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2947 flag = NFS4_OPEN_DELEGATE_WRITE;
2949 flag = NFS4_OPEN_DELEGATE_READ;
2955 dp = alloc_init_deleg(oo->oo_owner.so_client, stp, fh, flag);
2958 status = nfs4_set_delegation(dp, flag);
2962 memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
2964 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2965 STATEID_VAL(&dp->dl_stid.sc_stateid));
2967 open->op_delegate_type = flag;
2968 if (flag == NFS4_OPEN_DELEGATE_NONE) {
2969 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
2970 open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2971 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
2973 /* 4.1 client asking for a delegation? */
2974 if (open->op_deleg_want)
2975 nfsd4_open_deleg_none_ext(open, status);
2979 nfs4_put_delegation(dp);
2981 flag = NFS4_OPEN_DELEGATE_NONE;
2985 static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
2986 struct nfs4_delegation *dp)
2988 if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
2989 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2990 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2991 open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
2992 } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
2993 dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
2994 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
2995 open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
2997 /* Otherwise the client must be confused wanting a delegation
2998 * it already has, therefore we don't return
2999 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
3004 * called with nfs4_lock_state() held.
3007 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
3009 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3010 struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
3011 struct nfs4_file *fp = NULL;
3012 struct inode *ino = current_fh->fh_dentry->d_inode;
3013 struct nfs4_ol_stateid *stp = NULL;
3014 struct nfs4_delegation *dp = NULL;
3018 * Lookup file; if found, lookup stateid and check open request,
3019 * and check for delegations in the process of being recalled.
3020 * If not found, create the nfs4_file struct
3022 fp = find_file(ino);
3024 if ((status = nfs4_check_open(fp, open, &stp)))
3026 status = nfs4_check_deleg(cl, fp, open, &dp);
3030 status = nfserr_bad_stateid;
3031 if (nfsd4_is_deleg_cur(open))
3033 status = nfserr_jukebox;
3035 open->op_file = NULL;
3036 nfsd4_init_file(fp, ino);
3040 * OPEN the file, or upgrade an existing OPEN.
3041 * If truncate fails, the OPEN fails.
3044 /* Stateid was found, this is an OPEN upgrade */
3045 status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
3049 status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
3053 open->op_stp = NULL;
3054 init_open_stateid(stp, fp, open);
3055 status = nfsd4_truncate(rqstp, current_fh, open);
3057 release_open_stateid(stp);
3061 update_stateid(&stp->st_stid.sc_stateid);
3062 memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3064 if (nfsd4_has_session(&resp->cstate)) {
3065 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3067 if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
3068 open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
3069 open->op_why_no_deleg = WND4_NOT_WANTED;
3075 * Attempt to hand out a delegation. No error return, because the
3076 * OPEN succeeds even if we fail.
3078 nfs4_open_delegation(current_fh, open, stp);
3082 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
3083 STATEID_VAL(&stp->st_stid.sc_stateid));
3085 /* 4.1 client trying to upgrade/downgrade delegation? */
3086 if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
3087 open->op_deleg_want)
3088 nfsd4_deleg_xgrade_none_ext(open, dp);
3092 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
3093 nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
3095 * To finish the open response, we just need to set the rflags.
3097 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
3098 if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
3099 !nfsd4_has_session(&resp->cstate))
3100 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
3105 void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
3107 if (open->op_openowner) {
3108 struct nfs4_openowner *oo = open->op_openowner;
3110 if (!list_empty(&oo->oo_owner.so_stateids))
3111 list_del_init(&oo->oo_close_lru);
3112 if (oo->oo_flags & NFS4_OO_NEW) {
3114 release_openowner(oo);
3115 open->op_openowner = NULL;
3117 oo->oo_flags &= ~NFS4_OO_NEW;
3121 nfsd4_free_file(open->op_file);
3123 nfs4_free_stateid(open->op_stp);
3127 nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3130 struct nfs4_client *clp;
3134 dprintk("process_renew(%08x/%08x): starting\n",
3135 clid->cl_boot, clid->cl_id);
3136 status = nfserr_stale_clientid;
3137 if (STALE_CLIENTID(clid))
3139 clp = find_confirmed_client(clid);
3140 status = nfserr_expired;
3142 /* We assume the client took too long to RENEW. */
3143 dprintk("nfsd4_renew: clientid not found!\n");
3146 status = nfserr_cb_path_down;
3147 if (!list_empty(&clp->cl_delegations)
3148 && clp->cl_cb_state != NFSD4_CB_UP)
3152 nfs4_unlock_state();
3156 static struct lock_manager nfsd4_manager = {
3159 static bool grace_ended;
3162 nfsd4_end_grace(void)
3164 /* do nothing if grace period already ended */
3168 dprintk("NFSD: end of grace period\n");
3170 nfsd4_record_grace_done(&init_net, boot_time);
3171 locks_end_grace(&nfsd4_manager);
3173 * Now that every NFSv4 client has had the chance to recover and
3174 * to see the (possibly new, possibly shorter) lease time, we
3175 * can safely set the next grace time to the current lease time:
3177 nfsd4_grace = nfsd4_lease;
3181 nfs4_laundromat(void)
3183 struct nfs4_client *clp;
3184 struct nfs4_openowner *oo;
3185 struct nfs4_delegation *dp;
3186 struct list_head *pos, *next, reaplist;
3187 time_t cutoff = get_seconds() - nfsd4_lease;
3188 time_t t, clientid_val = nfsd4_lease;
3189 time_t u, test_val = nfsd4_lease;
3193 dprintk("NFSD: laundromat service - starting\n");
3195 INIT_LIST_HEAD(&reaplist);
3196 spin_lock(&client_lock);
3197 list_for_each_safe(pos, next, &client_lru) {
3198 clp = list_entry(pos, struct nfs4_client, cl_lru);
3199 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
3200 t = clp->cl_time - cutoff;
3201 if (clientid_val > t)
3205 if (atomic_read(&clp->cl_refcount)) {
3206 dprintk("NFSD: client in use (clientid %08x)\n",
3207 clp->cl_clientid.cl_id);
3210 unhash_client_locked(clp);
3211 list_add(&clp->cl_lru, &reaplist);
3213 spin_unlock(&client_lock);
3214 list_for_each_safe(pos, next, &reaplist) {
3215 clp = list_entry(pos, struct nfs4_client, cl_lru);
3216 dprintk("NFSD: purging unused client (clientid %08x)\n",
3217 clp->cl_clientid.cl_id);
3218 nfsd4_client_record_remove(clp);
3221 spin_lock(&recall_lock);
3222 list_for_each_safe(pos, next, &del_recall_lru) {
3223 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3224 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
3225 u = dp->dl_time - cutoff;
3230 list_move(&dp->dl_recall_lru, &reaplist);
3232 spin_unlock(&recall_lock);
3233 list_for_each_safe(pos, next, &reaplist) {
3234 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3235 unhash_delegation(dp);
3237 test_val = nfsd4_lease;
3238 list_for_each_safe(pos, next, &close_lru) {
3239 oo = container_of(pos, struct nfs4_openowner, oo_close_lru);
3240 if (time_after((unsigned long)oo->oo_time, (unsigned long)cutoff)) {
3241 u = oo->oo_time - cutoff;
3246 release_openowner(oo);
3248 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
3249 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
3250 nfs4_unlock_state();
3251 return clientid_val;
3254 static struct workqueue_struct *laundry_wq;
3255 static void laundromat_main(struct work_struct *);
3256 static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
3259 laundromat_main(struct work_struct *not_used)
3263 t = nfs4_laundromat();
3264 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
3265 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
3268 static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
3270 if (fhp->fh_dentry->d_inode != stp->st_file->fi_inode)
3271 return nfserr_bad_stateid;
3276 STALE_STATEID(stateid_t *stateid)
3278 if (stateid->si_opaque.so_clid.cl_boot == boot_time)
3280 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
3281 STATEID_VAL(stateid));
3286 access_permit_read(unsigned long access_bmap)
3288 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
3289 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
3290 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
3294 access_permit_write(unsigned long access_bmap)
3296 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
3297 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
3301 __be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
3303 __be32 status = nfserr_openmode;
3305 /* For lock stateid's, we test the parent open, not the lock: */
3306 if (stp->st_openstp)
3307 stp = stp->st_openstp;
3308 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
3310 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
3317 static inline __be32
3318 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
3320 if (ONE_STATEID(stateid) && (flags & RD_STATE))
3322 else if (locks_in_grace()) {
3323 /* Answer in remaining cases depends on existence of
3324 * conflicting state; so we must wait out the grace period. */
3325 return nfserr_grace;
3326 } else if (flags & WR_STATE)
3327 return nfs4_share_conflict(current_fh,
3328 NFS4_SHARE_DENY_WRITE);
3329 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
3330 return nfs4_share_conflict(current_fh,
3331 NFS4_SHARE_DENY_READ);
3335 * Allow READ/WRITE during grace period on recovered state only for files
3336 * that are not able to provide mandatory locking.
3339 grace_disallows_io(struct inode *inode)
3341 return locks_in_grace() && mandatory_lock(inode);
3344 /* Returns true iff a is later than b: */
3345 static bool stateid_generation_after(stateid_t *a, stateid_t *b)
3347 return (s32)a->si_generation - (s32)b->si_generation > 0;
3350 static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
3353 * When sessions are used the stateid generation number is ignored
3356 if (has_session && in->si_generation == 0)
3359 if (in->si_generation == ref->si_generation)
3362 /* If the client sends us a stateid from the future, it's buggy: */
3363 if (stateid_generation_after(in, ref))
3364 return nfserr_bad_stateid;
3366 * However, we could see a stateid from the past, even from a
3367 * non-buggy client. For example, if the client sends a lock
3368 * while some IO is outstanding, the lock may bump si_generation
3369 * while the IO is still in flight. The client could avoid that
3370 * situation by waiting for responses on all the IO requests,
3371 * but better performance may result in retrying IO that
3372 * receives an old_stateid error if requests are rarely
3373 * reordered in flight:
3375 return nfserr_old_stateid;
3378 __be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
3380 struct nfs4_stid *s;
3381 struct nfs4_ol_stateid *ols;
3384 if (STALE_STATEID(stateid))
3385 return nfserr_stale_stateid;
3387 s = find_stateid(cl, stateid);
3389 return nfserr_stale_stateid;
3390 status = check_stateid_generation(stateid, &s->sc_stateid, 1);
3393 if (!(s->sc_type & (NFS4_OPEN_STID | NFS4_LOCK_STID)))
3395 ols = openlockstateid(s);
3396 if (ols->st_stateowner->so_is_open_owner
3397 && !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3398 return nfserr_bad_stateid;
3402 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
3404 struct nfs4_client *cl;
3406 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3407 return nfserr_bad_stateid;
3408 if (STALE_STATEID(stateid))
3409 return nfserr_stale_stateid;
3410 cl = find_confirmed_client(&stateid->si_opaque.so_clid);
3412 return nfserr_expired;
3413 *s = find_stateid_by_type(cl, stateid, typemask);
3415 return nfserr_bad_stateid;
3421 * Checks for stateid operations
3424 nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
3425 stateid_t *stateid, int flags, struct file **filpp)
3427 struct nfs4_stid *s;
3428 struct nfs4_ol_stateid *stp = NULL;
3429 struct nfs4_delegation *dp = NULL;
3430 struct svc_fh *current_fh = &cstate->current_fh;
3431 struct inode *ino = current_fh->fh_dentry->d_inode;
3437 if (grace_disallows_io(ino))
3438 return nfserr_grace;
3440 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3441 return check_special_stateids(current_fh, stateid, flags);
3443 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
3446 status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
3449 switch (s->sc_type) {
3450 case NFS4_DELEG_STID:
3451 dp = delegstateid(s);
3452 status = nfs4_check_delegmode(dp, flags);
3456 *filpp = dp->dl_file->fi_deleg_file;
3460 case NFS4_OPEN_STID:
3461 case NFS4_LOCK_STID:
3462 stp = openlockstateid(s);
3463 status = nfs4_check_fh(current_fh, stp);
3466 if (stp->st_stateowner->so_is_open_owner
3467 && !(openowner(stp->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
3469 status = nfs4_check_openmode(stp, flags);
3473 if (flags & RD_STATE)
3474 *filpp = find_readable_file(stp->st_file);
3476 *filpp = find_writeable_file(stp->st_file);
3480 return nfserr_bad_stateid;
3488 nfsd4_free_lock_stateid(struct nfs4_ol_stateid *stp)
3490 if (check_for_locks(stp->st_file, lockowner(stp->st_stateowner)))
3491 return nfserr_locks_held;
3492 release_lock_stateid(stp);
3497 * Test if the stateid is valid
3500 nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3501 struct nfsd4_test_stateid *test_stateid)
3503 struct nfsd4_test_stateid_id *stateid;
3504 struct nfs4_client *cl = cstate->session->se_client;
3507 list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
3508 stateid->ts_id_status = nfs4_validate_stateid(cl, &stateid->ts_id_stateid);
3509 nfs4_unlock_state();
3515 nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3516 struct nfsd4_free_stateid *free_stateid)
3518 stateid_t *stateid = &free_stateid->fr_stateid;
3519 struct nfs4_stid *s;
3520 struct nfs4_client *cl = cstate->session->se_client;
3521 __be32 ret = nfserr_bad_stateid;
3524 s = find_stateid(cl, stateid);
3527 switch (s->sc_type) {
3528 case NFS4_DELEG_STID:
3529 ret = nfserr_locks_held;
3531 case NFS4_OPEN_STID:
3532 case NFS4_LOCK_STID:
3533 ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
3536 if (s->sc_type == NFS4_LOCK_STID)
3537 ret = nfsd4_free_lock_stateid(openlockstateid(s));
3539 ret = nfserr_locks_held;
3542 ret = nfserr_bad_stateid;
3545 nfs4_unlock_state();
3552 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
3553 RD_STATE : WR_STATE;
3556 static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
3558 struct svc_fh *current_fh = &cstate->current_fh;
3559 struct nfs4_stateowner *sop = stp->st_stateowner;
3562 status = nfsd4_check_seqid(cstate, sop, seqid);
3565 if (stp->st_stid.sc_type == NFS4_CLOSED_STID)
3567 * "Closed" stateid's exist *only* to return
3568 * nfserr_replay_me from the previous step.
3570 return nfserr_bad_stateid;
3571 status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
3574 return nfs4_check_fh(current_fh, stp);
3578 * Checks for sequence id mutating operations.
3581 nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3582 stateid_t *stateid, char typemask,
3583 struct nfs4_ol_stateid **stpp)
3586 struct nfs4_stid *s;
3588 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
3589 seqid, STATEID_VAL(stateid));
3592 status = nfsd4_lookup_stateid(stateid, typemask, &s);
3595 *stpp = openlockstateid(s);
3596 cstate->replay_owner = (*stpp)->st_stateowner;
3598 return nfs4_seqid_op_checks(cstate, stateid, seqid, *stpp);
3601 static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid, stateid_t *stateid, struct nfs4_ol_stateid **stpp)
3604 struct nfs4_openowner *oo;
3606 status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
3607 NFS4_OPEN_STID, stpp);
3610 oo = openowner((*stpp)->st_stateowner);
3611 if (!(oo->oo_flags & NFS4_OO_CONFIRMED))
3612 return nfserr_bad_stateid;
3617 nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3618 struct nfsd4_open_confirm *oc)
3621 struct nfs4_openowner *oo;
3622 struct nfs4_ol_stateid *stp;
3624 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
3625 (int)cstate->current_fh.fh_dentry->d_name.len,
3626 cstate->current_fh.fh_dentry->d_name.name);
3628 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
3634 status = nfs4_preprocess_seqid_op(cstate,
3635 oc->oc_seqid, &oc->oc_req_stateid,
3636 NFS4_OPEN_STID, &stp);
3639 oo = openowner(stp->st_stateowner);
3640 status = nfserr_bad_stateid;
3641 if (oo->oo_flags & NFS4_OO_CONFIRMED)
3643 oo->oo_flags |= NFS4_OO_CONFIRMED;
3644 update_stateid(&stp->st_stid.sc_stateid);
3645 memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3646 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3647 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
3649 nfsd4_client_record_create(oo->oo_owner.so_client);
3652 if (!cstate->replay_owner)
3653 nfs4_unlock_state();
3657 static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
3659 if (!test_bit(access, &stp->st_access_bmap))
3661 nfs4_file_put_access(stp->st_file, nfs4_access_to_omode(access));
3662 __clear_bit(access, &stp->st_access_bmap);
3665 static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
3667 switch (to_access) {
3668 case NFS4_SHARE_ACCESS_READ:
3669 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
3670 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3672 case NFS4_SHARE_ACCESS_WRITE:
3673 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
3674 nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
3676 case NFS4_SHARE_ACCESS_BOTH:
3684 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3687 for (i = 0; i < 4; i++) {
3688 if ((i & deny) != i)
3689 __clear_bit(i, bmap);
3694 nfsd4_open_downgrade(struct svc_rqst *rqstp,
3695 struct nfsd4_compound_state *cstate,
3696 struct nfsd4_open_downgrade *od)
3699 struct nfs4_ol_stateid *stp;
3701 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
3702 (int)cstate->current_fh.fh_dentry->d_name.len,
3703 cstate->current_fh.fh_dentry->d_name.name);
3705 /* We don't yet support WANT bits: */
3706 if (od->od_deleg_want)
3707 dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
3711 status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
3712 &od->od_stateid, &stp);
3715 status = nfserr_inval;
3716 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3717 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3718 stp->st_access_bmap, od->od_share_access);
3721 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3722 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3723 stp->st_deny_bmap, od->od_share_deny);
3726 nfs4_stateid_downgrade(stp, od->od_share_access);
3728 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3730 update_stateid(&stp->st_stid.sc_stateid);
3731 memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3734 if (!cstate->replay_owner)
3735 nfs4_unlock_state();
3739 void nfsd4_purge_closed_stateid(struct nfs4_stateowner *so)
3741 struct nfs4_openowner *oo;
3742 struct nfs4_ol_stateid *s;
3744 if (!so->so_is_open_owner)
3747 s = oo->oo_last_closed_stid;
3750 if (!(oo->oo_flags & NFS4_OO_PURGE_CLOSE)) {
3751 /* Release the last_closed_stid on the next seqid bump: */
3752 oo->oo_flags |= NFS4_OO_PURGE_CLOSE;
3755 oo->oo_flags &= ~NFS4_OO_PURGE_CLOSE;
3756 release_last_closed_stateid(oo);
3759 static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
3761 unhash_open_stateid(s);
3762 s->st_stid.sc_type = NFS4_CLOSED_STID;
3766 * nfs4_unlock_state() called after encode
3769 nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3770 struct nfsd4_close *close)
3773 struct nfs4_openowner *oo;
3774 struct nfs4_ol_stateid *stp;
3776 dprintk("NFSD: nfsd4_close on file %.*s\n",
3777 (int)cstate->current_fh.fh_dentry->d_name.len,
3778 cstate->current_fh.fh_dentry->d_name.name);
3781 status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
3783 NFS4_OPEN_STID|NFS4_CLOSED_STID,
3787 oo = openowner(stp->st_stateowner);
3789 update_stateid(&stp->st_stid.sc_stateid);
3790 memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
3792 nfsd4_close_open_stateid(stp);
3793 oo->oo_last_closed_stid = stp;
3795 /* place unused nfs4_stateowners on so_close_lru list to be
3796 * released by the laundromat service after the lease period
3797 * to enable us to handle CLOSE replay
3799 if (list_empty(&oo->oo_owner.so_stateids))
3800 move_to_close_lru(oo);
3802 if (!cstate->replay_owner)
3803 nfs4_unlock_state();
3808 nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3809 struct nfsd4_delegreturn *dr)
3811 struct nfs4_delegation *dp;
3812 stateid_t *stateid = &dr->dr_stateid;
3813 struct nfs4_stid *s;
3814 struct inode *inode;
3817 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
3819 inode = cstate->current_fh.fh_dentry->d_inode;
3822 status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID, &s);
3825 dp = delegstateid(s);
3826 status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
3830 unhash_delegation(dp);
3832 nfs4_unlock_state();
3838 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3840 #define LOCKOWNER_INO_HASH_BITS 8
3841 #define LOCKOWNER_INO_HASH_SIZE (1 << LOCKOWNER_INO_HASH_BITS)
3842 #define LOCKOWNER_INO_HASH_MASK (LOCKOWNER_INO_HASH_SIZE - 1)
3845 end_offset(u64 start, u64 len)
3850 return end >= start ? end: NFS4_MAX_UINT64;
3853 /* last octet in a range */
3855 last_byte_offset(u64 start, u64 len)
3861 return end > start ? end - 1: NFS4_MAX_UINT64;
3864 static unsigned int lockowner_ino_hashval(struct inode *inode, u32 cl_id, struct xdr_netobj *ownername)
3866 return (file_hashval(inode) + cl_id
3867 + opaque_hashval(ownername->data, ownername->len))
3868 & LOCKOWNER_INO_HASH_MASK;
3871 static struct list_head lockowner_ino_hashtbl[LOCKOWNER_INO_HASH_SIZE];
3874 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3875 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3876 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3877 * locking, this prevents us from being completely protocol-compliant. The
3878 * real solution to this problem is to start using unsigned file offsets in
3879 * the VFS, but this is a very deep change!
3882 nfs4_transform_lock_offset(struct file_lock *lock)
3884 if (lock->fl_start < 0)
3885 lock->fl_start = OFFSET_MAX;
3886 if (lock->fl_end < 0)
3887 lock->fl_end = OFFSET_MAX;
3890 /* Hack!: For now, we're defining this just so we can use a pointer to it
3891 * as a unique cookie to identify our (NFSv4's) posix locks. */
3892 static const struct lock_manager_operations nfsd_posix_mng_ops = {
3896 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3898 struct nfs4_lockowner *lo;
3900 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3901 lo = (struct nfs4_lockowner *) fl->fl_owner;
3902 deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
3903 lo->lo_owner.so_owner.len, GFP_KERNEL);
3904 if (!deny->ld_owner.data)
3905 /* We just don't care that much */
3907 deny->ld_owner.len = lo->lo_owner.so_owner.len;
3908 deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
3911 deny->ld_owner.len = 0;
3912 deny->ld_owner.data = NULL;
3913 deny->ld_clientid.cl_boot = 0;
3914 deny->ld_clientid.cl_id = 0;
3916 deny->ld_start = fl->fl_start;
3917 deny->ld_length = NFS4_MAX_UINT64;
3918 if (fl->fl_end != NFS4_MAX_UINT64)
3919 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3920 deny->ld_type = NFS4_READ_LT;
3921 if (fl->fl_type != F_RDLCK)
3922 deny->ld_type = NFS4_WRITE_LT;
3925 static bool same_lockowner_ino(struct nfs4_lockowner *lo, struct inode *inode, clientid_t *clid, struct xdr_netobj *owner)
3927 struct nfs4_ol_stateid *lst;
3929 if (!same_owner_str(&lo->lo_owner, owner, clid))
3931 lst = list_first_entry(&lo->lo_owner.so_stateids,
3932 struct nfs4_ol_stateid, st_perstateowner);
3933 return lst->st_file->fi_inode == inode;
3936 static struct nfs4_lockowner *
3937 find_lockowner_str(struct inode *inode, clientid_t *clid,
3938 struct xdr_netobj *owner)
3940 unsigned int hashval = lockowner_ino_hashval(inode, clid->cl_id, owner);
3941 struct nfs4_lockowner *lo;
3943 list_for_each_entry(lo, &lockowner_ino_hashtbl[hashval], lo_owner_ino_hash) {
3944 if (same_lockowner_ino(lo, inode, clid, owner))
3950 static void hash_lockowner(struct nfs4_lockowner *lo, unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp)
3952 struct inode *inode = open_stp->st_file->fi_inode;
3953 unsigned int inohash = lockowner_ino_hashval(inode,
3954 clp->cl_clientid.cl_id, &lo->lo_owner.so_owner);
3956 list_add(&lo->lo_owner.so_strhash, &ownerstr_hashtbl[strhashval]);
3957 list_add(&lo->lo_owner_ino_hash, &lockowner_ino_hashtbl[inohash]);
3958 list_add(&lo->lo_perstateid, &open_stp->st_lockowners);
3962 * Alloc a lock owner structure.
3963 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3966 * strhashval = ownerstr_hashval
3969 static struct nfs4_lockowner *
3970 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_ol_stateid *open_stp, struct nfsd4_lock *lock) {
3971 struct nfs4_lockowner *lo;
3973 lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
3976 INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
3977 lo->lo_owner.so_is_open_owner = 0;
3978 /* It is the openowner seqid that will be incremented in encode in the
3979 * case of new lockowners; so increment the lock seqid manually: */
3980 lo->lo_owner.so_seqid = lock->lk_new_lock_seqid + 1;
3981 hash_lockowner(lo, strhashval, clp, open_stp);
3985 static struct nfs4_ol_stateid *
3986 alloc_init_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp, struct nfs4_ol_stateid *open_stp)
3988 struct nfs4_ol_stateid *stp;
3989 struct nfs4_client *clp = lo->lo_owner.so_client;
3991 stp = nfs4_alloc_stateid(clp);
3994 init_stid(&stp->st_stid, clp, NFS4_LOCK_STID);
3995 list_add(&stp->st_perfile, &fp->fi_stateids);
3996 list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
3997 stp->st_stateowner = &lo->lo_owner;
4000 stp->st_access_bmap = 0;
4001 stp->st_deny_bmap = open_stp->st_deny_bmap;
4002 stp->st_openstp = open_stp;
4007 check_lock_length(u64 offset, u64 length)
4009 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
4010 LOFF_OVERFLOW(offset, length)));
4013 static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
4015 struct nfs4_file *fp = lock_stp->st_file;
4016 int oflag = nfs4_access_to_omode(access);
4018 if (test_bit(access, &lock_stp->st_access_bmap))
4020 nfs4_file_get_access(fp, oflag);
4021 __set_bit(access, &lock_stp->st_access_bmap);
4024 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)
4026 struct nfs4_file *fi = ost->st_file;
4027 struct nfs4_openowner *oo = openowner(ost->st_stateowner);
4028 struct nfs4_client *cl = oo->oo_owner.so_client;
4029 struct nfs4_lockowner *lo;
4030 unsigned int strhashval;
4032 lo = find_lockowner_str(fi->fi_inode, &cl->cl_clientid, &lock->v.new.owner);
4034 if (!cstate->minorversion)
4035 return nfserr_bad_seqid;
4036 /* XXX: a lockowner always has exactly one stateid: */
4037 *lst = list_first_entry(&lo->lo_owner.so_stateids,
4038 struct nfs4_ol_stateid, st_perstateowner);
4041 strhashval = ownerstr_hashval(cl->cl_clientid.cl_id,
4042 &lock->v.new.owner);
4043 lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
4045 return nfserr_jukebox;
4046 *lst = alloc_init_lock_stateid(lo, fi, ost);
4048 release_lockowner(lo);
4049 return nfserr_jukebox;
4059 nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4060 struct nfsd4_lock *lock)
4062 struct nfs4_openowner *open_sop = NULL;
4063 struct nfs4_lockowner *lock_sop = NULL;
4064 struct nfs4_ol_stateid *lock_stp;
4065 struct file *filp = NULL;
4066 struct file_lock file_lock;
4067 struct file_lock conflock;
4069 bool new_state = false;
4073 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
4074 (long long) lock->lk_offset,
4075 (long long) lock->lk_length);
4077 if (check_lock_length(lock->lk_offset, lock->lk_length))
4078 return nfserr_inval;
4080 if ((status = fh_verify(rqstp, &cstate->current_fh,
4081 S_IFREG, NFSD_MAY_LOCK))) {
4082 dprintk("NFSD: nfsd4_lock: permission denied!\n");
4088 if (lock->lk_is_new) {
4090 * Client indicates that this is a new lockowner.
4091 * Use open owner and open stateid to create lock owner and
4094 struct nfs4_ol_stateid *open_stp = NULL;
4096 if (nfsd4_has_session(cstate))
4097 /* See rfc 5661 18.10.3: given clientid is ignored: */
4098 memcpy(&lock->v.new.clientid,
4099 &cstate->session->se_client->cl_clientid,
4100 sizeof(clientid_t));
4102 status = nfserr_stale_clientid;
4103 if (STALE_CLIENTID(&lock->lk_new_clientid))
4106 /* validate and update open stateid and open seqid */
4107 status = nfs4_preprocess_confirmed_seqid_op(cstate,
4108 lock->lk_new_open_seqid,
4109 &lock->lk_new_open_stateid,
4113 open_sop = openowner(open_stp->st_stateowner);
4114 status = nfserr_bad_stateid;
4115 if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
4116 &lock->v.new.clientid))
4118 status = lookup_or_create_lock_state(cstate, open_stp, lock,
4119 &lock_stp, &new_state);
4123 /* lock (lock owner + lock stateid) already exists */
4124 status = nfs4_preprocess_seqid_op(cstate,
4125 lock->lk_old_lock_seqid,
4126 &lock->lk_old_lock_stateid,
4127 NFS4_LOCK_STID, &lock_stp);
4131 lock_sop = lockowner(lock_stp->st_stateowner);
4133 lkflg = setlkflg(lock->lk_type);
4134 status = nfs4_check_openmode(lock_stp, lkflg);
4138 status = nfserr_grace;
4139 if (locks_in_grace() && !lock->lk_reclaim)
4141 status = nfserr_no_grace;
4142 if (!locks_in_grace() && lock->lk_reclaim)
4145 locks_init_lock(&file_lock);
4146 switch (lock->lk_type) {
4149 filp = find_readable_file(lock_stp->st_file);
4151 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
4152 file_lock.fl_type = F_RDLCK;
4155 case NFS4_WRITEW_LT:
4156 filp = find_writeable_file(lock_stp->st_file);
4158 get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
4159 file_lock.fl_type = F_WRLCK;
4162 status = nfserr_inval;
4166 status = nfserr_openmode;
4169 file_lock.fl_owner = (fl_owner_t)lock_sop;
4170 file_lock.fl_pid = current->tgid;
4171 file_lock.fl_file = filp;
4172 file_lock.fl_flags = FL_POSIX;
4173 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4175 file_lock.fl_start = lock->lk_offset;
4176 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
4177 nfs4_transform_lock_offset(&file_lock);
4180 * Try to lock the file in the VFS.
4181 * Note: locks.c uses the BKL to protect the inode's lock list.
4184 err = vfs_lock_file(filp, F_SETLK, &file_lock, &conflock);
4186 case 0: /* success! */
4187 update_stateid(&lock_stp->st_stid.sc_stateid);
4188 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid,
4192 case (EAGAIN): /* conflock holds conflicting lock */
4193 status = nfserr_denied;
4194 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
4195 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
4198 status = nfserr_deadlock;
4201 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
4202 status = nfserrno(err);
4206 if (status && new_state)
4207 release_lockowner(lock_sop);
4208 if (!cstate->replay_owner)
4209 nfs4_unlock_state();
4214 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
4215 * so we do a temporary open here just to get an open file to pass to
4216 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
4219 static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
4224 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
4227 err = vfs_test_lock(file, lock);
4236 nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4237 struct nfsd4_lockt *lockt)
4239 struct inode *inode;
4240 struct file_lock file_lock;
4241 struct nfs4_lockowner *lo;
4245 if (locks_in_grace())
4246 return nfserr_grace;
4248 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
4249 return nfserr_inval;
4253 status = nfserr_stale_clientid;
4254 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
4257 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4260 inode = cstate->current_fh.fh_dentry->d_inode;
4261 locks_init_lock(&file_lock);
4262 switch (lockt->lt_type) {
4265 file_lock.fl_type = F_RDLCK;
4268 case NFS4_WRITEW_LT:
4269 file_lock.fl_type = F_WRLCK;
4272 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
4273 status = nfserr_inval;
4277 lo = find_lockowner_str(inode, &lockt->lt_clientid, &lockt->lt_owner);
4279 file_lock.fl_owner = (fl_owner_t)lo;
4280 file_lock.fl_pid = current->tgid;
4281 file_lock.fl_flags = FL_POSIX;
4283 file_lock.fl_start = lockt->lt_offset;
4284 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
4286 nfs4_transform_lock_offset(&file_lock);
4289 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
4291 status = nfserrno(error);
4294 if (file_lock.fl_type != F_UNLCK) {
4295 status = nfserr_denied;
4296 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
4299 nfs4_unlock_state();
4304 nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4305 struct nfsd4_locku *locku)
4307 struct nfs4_ol_stateid *stp;
4308 struct file *filp = NULL;
4309 struct file_lock file_lock;
4313 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
4314 (long long) locku->lu_offset,
4315 (long long) locku->lu_length);
4317 if (check_lock_length(locku->lu_offset, locku->lu_length))
4318 return nfserr_inval;
4322 status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
4323 &locku->lu_stateid, NFS4_LOCK_STID, &stp);
4326 filp = find_any_file(stp->st_file);
4328 status = nfserr_lock_range;
4332 locks_init_lock(&file_lock);
4333 file_lock.fl_type = F_UNLCK;
4334 file_lock.fl_owner = (fl_owner_t)lockowner(stp->st_stateowner);
4335 file_lock.fl_pid = current->tgid;
4336 file_lock.fl_file = filp;
4337 file_lock.fl_flags = FL_POSIX;
4338 file_lock.fl_lmops = &nfsd_posix_mng_ops;
4339 file_lock.fl_start = locku->lu_offset;
4341 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
4342 nfs4_transform_lock_offset(&file_lock);
4345 * Try to unlock the file in the VFS.
4347 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
4349 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
4353 * OK, unlock succeeded; the only thing left to do is update the stateid.
4355 update_stateid(&stp->st_stid.sc_stateid);
4356 memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4359 if (!cstate->replay_owner)
4360 nfs4_unlock_state();
4364 status = nfserrno(err);
4370 * 1: locks held by lockowner
4371 * 0: no locks held by lockowner
4374 check_for_locks(struct nfs4_file *filp, struct nfs4_lockowner *lowner)
4376 struct file_lock **flpp;
4377 struct inode *inode = filp->fi_inode;
4381 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
4382 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
4393 nfsd4_release_lockowner(struct svc_rqst *rqstp,
4394 struct nfsd4_compound_state *cstate,
4395 struct nfsd4_release_lockowner *rlockowner)
4397 clientid_t *clid = &rlockowner->rl_clientid;
4398 struct nfs4_stateowner *sop;
4399 struct nfs4_lockowner *lo;
4400 struct nfs4_ol_stateid *stp;
4401 struct xdr_netobj *owner = &rlockowner->rl_owner;
4402 struct list_head matches;
4403 unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
4406 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
4407 clid->cl_boot, clid->cl_id);
4409 /* XXX check for lease expiration */
4411 status = nfserr_stale_clientid;
4412 if (STALE_CLIENTID(clid))
4417 status = nfserr_locks_held;
4418 INIT_LIST_HEAD(&matches);
4420 list_for_each_entry(sop, &ownerstr_hashtbl[hashval], so_strhash) {
4421 if (sop->so_is_open_owner)
4423 if (!same_owner_str(sop, owner, clid))
4425 list_for_each_entry(stp, &sop->so_stateids,
4427 lo = lockowner(sop);
4428 if (check_for_locks(stp->st_file, lo))
4430 list_add(&lo->lo_list, &matches);
4433 /* Clients probably won't expect us to return with some (but not all)
4434 * of the lockowner state released; so don't release any until all
4435 * have been checked. */
4437 while (!list_empty(&matches)) {
4438 lo = list_entry(matches.next, struct nfs4_lockowner,
4440 /* unhash_stateowner deletes so_perclient only
4441 * for openowners. */
4442 list_del(&lo->lo_list);
4443 release_lockowner(lo);
4446 nfs4_unlock_state();
4450 static inline struct nfs4_client_reclaim *
4453 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
4457 nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
4459 unsigned int strhashval = clientstr_hashval(name);
4460 struct nfs4_client *clp;
4462 clp = find_confirmed_client_by_str(name, strhashval);
4465 return test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
4469 * failure => all reset bets are off, nfserr_no_grace...
4472 nfs4_client_to_reclaim(const char *name)
4474 unsigned int strhashval;
4475 struct nfs4_client_reclaim *crp = NULL;
4477 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
4478 crp = alloc_reclaim();
4481 strhashval = clientstr_hashval(name);
4482 INIT_LIST_HEAD(&crp->cr_strhash);
4483 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
4484 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
4485 reclaim_str_hashtbl_size++;
4490 nfs4_release_reclaim(void)
4492 struct nfs4_client_reclaim *crp = NULL;
4495 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4496 while (!list_empty(&reclaim_str_hashtbl[i])) {
4497 crp = list_entry(reclaim_str_hashtbl[i].next,
4498 struct nfs4_client_reclaim, cr_strhash);
4499 list_del(&crp->cr_strhash);
4501 reclaim_str_hashtbl_size--;
4504 BUG_ON(reclaim_str_hashtbl_size);
4508 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
4509 struct nfs4_client_reclaim *
4510 nfsd4_find_reclaim_client(struct nfs4_client *clp)
4512 unsigned int strhashval;
4513 struct nfs4_client_reclaim *crp = NULL;
4515 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
4516 clp->cl_name.len, clp->cl_name.data,
4519 /* find clp->cl_name in reclaim_str_hashtbl */
4520 strhashval = clientstr_hashval(clp->cl_recdir);
4521 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
4522 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
4530 * Called from OPEN. Look for clientid in reclaim list.
4533 nfs4_check_open_reclaim(clientid_t *clid)
4535 struct nfs4_client *clp;
4537 /* find clientid in conf_id_hashtbl */
4538 clp = find_confirmed_client(clid);
4540 return nfserr_reclaim_bad;
4542 return nfsd4_client_record_check(clp) ? nfserr_reclaim_bad : nfs_ok;
4545 #ifdef CONFIG_NFSD_FAULT_INJECTION
4547 void nfsd_forget_clients(u64 num)
4549 struct nfs4_client *clp, *next;
4553 list_for_each_entry_safe(clp, next, &client_lru, cl_lru) {
4554 nfsd4_client_record_remove(clp);
4559 nfs4_unlock_state();
4561 printk(KERN_INFO "NFSD: Forgot %d clients", count);
4564 static void release_lockowner_sop(struct nfs4_stateowner *sop)
4566 release_lockowner(lockowner(sop));
4569 static void release_openowner_sop(struct nfs4_stateowner *sop)
4571 release_openowner(openowner(sop));
4574 static int nfsd_release_n_owners(u64 num, bool is_open_owner,
4575 void (*release_sop)(struct nfs4_stateowner *))
4578 struct nfs4_stateowner *sop, *next;
4580 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4581 list_for_each_entry_safe(sop, next, &ownerstr_hashtbl[i], so_strhash) {
4582 if (sop->so_is_open_owner != is_open_owner)
4592 void nfsd_forget_locks(u64 num)
4597 count = nfsd_release_n_owners(num, false, release_lockowner_sop);
4598 nfs4_unlock_state();
4600 printk(KERN_INFO "NFSD: Forgot %d locks", count);
4603 void nfsd_forget_openowners(u64 num)
4608 count = nfsd_release_n_owners(num, true, release_openowner_sop);
4609 nfs4_unlock_state();
4611 printk(KERN_INFO "NFSD: Forgot %d open owners", count);
4614 int nfsd_process_n_delegations(u64 num, void (*deleg_func)(struct nfs4_delegation *))
4617 struct nfs4_file *fp, *fnext;
4618 struct nfs4_delegation *dp, *dnext;
4620 for (i = 0; i < FILE_HASH_SIZE; i++) {
4621 list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
4622 list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
4633 void nfsd_forget_delegations(u64 num)
4638 count = nfsd_process_n_delegations(num, unhash_delegation);
4639 nfs4_unlock_state();
4641 printk(KERN_INFO "NFSD: Forgot %d delegations", count);
4644 void nfsd_recall_delegations(u64 num)
4649 spin_lock(&recall_lock);
4650 count = nfsd_process_n_delegations(num, nfsd_break_one_deleg);
4651 spin_unlock(&recall_lock);
4652 nfs4_unlock_state();
4654 printk(KERN_INFO "NFSD: Recalled %d delegations", count);
4657 #endif /* CONFIG_NFSD_FAULT_INJECTION */
4659 /* initialization to perform at module load time: */
4662 nfs4_state_init(void)
4666 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4667 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4668 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4669 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4670 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4671 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4673 for (i = 0; i < SESSION_HASH_SIZE; i++)
4674 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
4675 for (i = 0; i < FILE_HASH_SIZE; i++) {
4676 INIT_LIST_HEAD(&file_hashtbl[i]);
4678 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4679 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4681 for (i = 0; i < LOCKOWNER_INO_HASH_SIZE; i++)
4682 INIT_LIST_HEAD(&lockowner_ino_hashtbl[i]);
4683 INIT_LIST_HEAD(&close_lru);
4684 INIT_LIST_HEAD(&client_lru);
4685 INIT_LIST_HEAD(&del_recall_lru);
4686 reclaim_str_hashtbl_size = 0;
4690 * Since the lifetime of a delegation isn't limited to that of an open, a
4691 * client may quite reasonably hang on to a delegation as long as it has
4692 * the inode cached. This becomes an obvious problem the first time a
4693 * client's inode cache approaches the size of the server's total memory.
4695 * For now we avoid this problem by imposing a hard limit on the number
4696 * of delegations, which varies according to the server's memory size.
4699 set_max_delegations(void)
4702 * Allow at most 4 delegations per megabyte of RAM. Quick
4703 * estimates suggest that in the worst case (where every delegation
4704 * is for a different inode), a delegation could take about 1.5K,
4705 * giving a worst case usage of about 6% of memory.
4707 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4710 /* initialization to perform when the nfsd service is started: */
4713 nfs4_state_start(void)
4718 * FIXME: For now, we hang most of the pernet global stuff off of
4719 * init_net until nfsd is fully containerized. Eventually, we'll
4720 * need to pass a net pointer into this function, take a reference
4721 * to that instead and then do most of the rest of this on a per-net
4725 nfsd4_client_tracking_init(&init_net);
4726 boot_time = get_seconds();
4727 locks_start_grace(&nfsd4_manager);
4728 grace_ended = false;
4729 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4731 ret = set_callback_cred();
4736 laundry_wq = create_singlethread_workqueue("nfsd4");
4737 if (laundry_wq == NULL) {
4741 ret = nfsd4_create_callback_queue();
4743 goto out_free_laundry;
4744 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
4745 set_max_delegations();
4748 destroy_workqueue(laundry_wq);
4750 nfsd4_client_tracking_exit(&init_net);
4756 __nfs4_state_shutdown(void)
4759 struct nfs4_client *clp = NULL;
4760 struct nfs4_delegation *dp = NULL;
4761 struct list_head *pos, *next, reaplist;
4763 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4764 while (!list_empty(&conf_id_hashtbl[i])) {
4765 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4768 while (!list_empty(&unconf_str_hashtbl[i])) {
4769 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4773 INIT_LIST_HEAD(&reaplist);
4774 spin_lock(&recall_lock);
4775 list_for_each_safe(pos, next, &del_recall_lru) {
4776 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4777 list_move(&dp->dl_recall_lru, &reaplist);
4779 spin_unlock(&recall_lock);
4780 list_for_each_safe(pos, next, &reaplist) {
4781 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4782 unhash_delegation(dp);
4785 nfsd4_client_tracking_exit(&init_net);
4790 nfs4_state_shutdown(void)
4792 cancel_delayed_work_sync(&laundromat_work);
4793 destroy_workqueue(laundry_wq);
4794 locks_end_grace(&nfsd4_manager);
4796 __nfs4_state_shutdown();
4797 nfs4_unlock_state();
4798 nfsd4_destroy_callback_queue();
4802 get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4804 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
4805 memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
4809 put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
4811 if (cstate->minorversion) {
4812 memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
4813 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4818 clear_current_stateid(struct nfsd4_compound_state *cstate)
4820 CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
4824 * functions to set current state id
4827 nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4829 put_stateid(cstate, &odp->od_stateid);
4833 nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
4835 put_stateid(cstate, &open->op_stateid);
4839 nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4841 put_stateid(cstate, &close->cl_stateid);
4845 nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
4847 put_stateid(cstate, &lock->lk_resp_stateid);
4851 * functions to consume current state id
4855 nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
4857 get_stateid(cstate, &odp->od_stateid);
4861 nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
4863 get_stateid(cstate, &drp->dr_stateid);
4867 nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
4869 get_stateid(cstate, &fsp->fr_stateid);
4873 nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
4875 get_stateid(cstate, &setattr->sa_stateid);
4879 nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
4881 get_stateid(cstate, &close->cl_stateid);
4885 nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
4887 get_stateid(cstate, &locku->lu_stateid);
4891 nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
4893 get_stateid(cstate, &read->rd_stateid);
4897 nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
4899 get_stateid(cstate, &write->wr_stateid);