[CPUFREQ][1/8] acpi-cpufreq: software coordination and handle all CPUs in the group
[linux-2.6/mini2440.git] / fs / nfsd / nfs4state.c
blobebcf226a9e4a1fb65d04baa8e6ac10dc7f845071
1 /*
2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * All rights reserved.
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <kandros@umich.edu>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/mount.h>
45 #include <linux/workqueue.h>
46 #include <linux/smp_lock.h>
47 #include <linux/kthread.h>
48 #include <linux/nfs4.h>
49 #include <linux/nfsd/state.h>
50 #include <linux/nfsd/xdr4.h>
51 #include <linux/namei.h>
52 #include <linux/mutex.h>
54 #define NFSDDBG_FACILITY NFSDDBG_PROC
56 /* Globals */
57 static time_t lease_time = 90; /* default lease time */
58 static time_t user_lease_time = 90;
59 static time_t boot_time;
60 static int in_grace = 1;
61 static u32 current_clientid = 1;
62 static u32 current_ownerid = 1;
63 static u32 current_fileid = 1;
64 static u32 current_delegid = 1;
65 static u32 nfs4_init;
66 static stateid_t zerostateid; /* bits all 0 */
67 static stateid_t onestateid; /* bits all 1 */
69 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
70 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
72 /* forward declarations */
73 static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
74 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
75 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
76 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
77 static void nfs4_set_recdir(char *recdir);
79 /* Locking:
81 * client_mutex:
82 * protects clientid_hashtbl[], clientstr_hashtbl[],
83 * unconfstr_hashtbl[], uncofid_hashtbl[].
85 static DEFINE_MUTEX(client_mutex);
87 static kmem_cache_t *stateowner_slab = NULL;
88 static kmem_cache_t *file_slab = NULL;
89 static kmem_cache_t *stateid_slab = NULL;
90 static kmem_cache_t *deleg_slab = NULL;
92 void
93 nfs4_lock_state(void)
95 mutex_lock(&client_mutex);
98 void
99 nfs4_unlock_state(void)
101 mutex_unlock(&client_mutex);
104 static inline u32
105 opaque_hashval(const void *ptr, int nbytes)
107 unsigned char *cptr = (unsigned char *) ptr;
109 u32 x = 0;
110 while (nbytes--) {
111 x *= 37;
112 x += *cptr++;
114 return x;
117 /* forward declarations */
118 static void release_stateowner(struct nfs4_stateowner *sop);
119 static void release_stateid(struct nfs4_stateid *stp, int flags);
122 * Delegation state
125 /* recall_lock protects the del_recall_lru */
126 static DEFINE_SPINLOCK(recall_lock);
127 static struct list_head del_recall_lru;
129 static void
130 free_nfs4_file(struct kref *kref)
132 struct nfs4_file *fp = container_of(kref, struct nfs4_file, fi_ref);
133 list_del(&fp->fi_hash);
134 iput(fp->fi_inode);
135 kmem_cache_free(file_slab, fp);
138 static inline void
139 put_nfs4_file(struct nfs4_file *fi)
141 kref_put(&fi->fi_ref, free_nfs4_file);
144 static inline void
145 get_nfs4_file(struct nfs4_file *fi)
147 kref_get(&fi->fi_ref);
150 static int num_delegations;
153 * Open owner state (share locks)
156 /* hash tables for nfs4_stateowner */
157 #define OWNER_HASH_BITS 8
158 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
159 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
161 #define ownerid_hashval(id) \
162 ((id) & OWNER_HASH_MASK)
163 #define ownerstr_hashval(clientid, ownername) \
164 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
166 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
167 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
169 /* hash table for nfs4_file */
170 #define FILE_HASH_BITS 8
171 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
172 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
173 /* hash table for (open)nfs4_stateid */
174 #define STATEID_HASH_BITS 10
175 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
176 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
178 #define file_hashval(x) \
179 hash_ptr(x, FILE_HASH_BITS)
180 #define stateid_hashval(owner_id, file_id) \
181 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
183 static struct list_head file_hashtbl[FILE_HASH_SIZE];
184 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
186 static struct nfs4_delegation *
187 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
189 struct nfs4_delegation *dp;
190 struct nfs4_file *fp = stp->st_file;
191 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
193 dprintk("NFSD alloc_init_deleg\n");
194 if (num_delegations > STATEID_HASH_SIZE * 4)
195 return NULL;
196 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
197 if (dp == NULL)
198 return dp;
199 num_delegations++;
200 INIT_LIST_HEAD(&dp->dl_perfile);
201 INIT_LIST_HEAD(&dp->dl_perclnt);
202 INIT_LIST_HEAD(&dp->dl_recall_lru);
203 dp->dl_client = clp;
204 get_nfs4_file(fp);
205 dp->dl_file = fp;
206 dp->dl_flock = NULL;
207 get_file(stp->st_vfs_file);
208 dp->dl_vfs_file = stp->st_vfs_file;
209 dp->dl_type = type;
210 dp->dl_recall.cbr_dp = NULL;
211 dp->dl_recall.cbr_ident = cb->cb_ident;
212 dp->dl_recall.cbr_trunc = 0;
213 dp->dl_stateid.si_boot = boot_time;
214 dp->dl_stateid.si_stateownerid = current_delegid++;
215 dp->dl_stateid.si_fileid = 0;
216 dp->dl_stateid.si_generation = 0;
217 dp->dl_fhlen = current_fh->fh_handle.fh_size;
218 memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
219 current_fh->fh_handle.fh_size);
220 dp->dl_time = 0;
221 atomic_set(&dp->dl_count, 1);
222 list_add(&dp->dl_perfile, &fp->fi_delegations);
223 list_add(&dp->dl_perclnt, &clp->cl_delegations);
224 return dp;
227 void
228 nfs4_put_delegation(struct nfs4_delegation *dp)
230 if (atomic_dec_and_test(&dp->dl_count)) {
231 dprintk("NFSD: freeing dp %p\n",dp);
232 put_nfs4_file(dp->dl_file);
233 kmem_cache_free(deleg_slab, dp);
234 num_delegations--;
238 /* Remove the associated file_lock first, then remove the delegation.
239 * lease_modify() is called to remove the FS_LEASE file_lock from
240 * the i_flock list, eventually calling nfsd's lock_manager
241 * fl_release_callback.
243 static void
244 nfs4_close_delegation(struct nfs4_delegation *dp)
246 struct file *filp = dp->dl_vfs_file;
248 dprintk("NFSD: close_delegation dp %p\n",dp);
249 dp->dl_vfs_file = NULL;
250 /* The following nfsd_close may not actually close the file,
251 * but we want to remove the lease in any case. */
252 if (dp->dl_flock)
253 setlease(filp, F_UNLCK, &dp->dl_flock);
254 nfsd_close(filp);
257 /* Called under the state lock. */
258 static void
259 unhash_delegation(struct nfs4_delegation *dp)
261 list_del_init(&dp->dl_perfile);
262 list_del_init(&dp->dl_perclnt);
263 spin_lock(&recall_lock);
264 list_del_init(&dp->dl_recall_lru);
265 spin_unlock(&recall_lock);
266 nfs4_close_delegation(dp);
267 nfs4_put_delegation(dp);
271 * SETCLIENTID state
274 /* Hash tables for nfs4_clientid state */
275 #define CLIENT_HASH_BITS 4
276 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
277 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
279 #define clientid_hashval(id) \
280 ((id) & CLIENT_HASH_MASK)
281 #define clientstr_hashval(name) \
282 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
284 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
285 * used in reboot/reset lease grace period processing
287 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
288 * setclientid_confirmed info.
290 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
291 * setclientid info.
293 * client_lru holds client queue ordered by nfs4_client.cl_time
294 * for lease renewal.
296 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
297 * for last close replay.
299 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
300 static int reclaim_str_hashtbl_size = 0;
301 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
302 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
303 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
304 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
305 static struct list_head client_lru;
306 static struct list_head close_lru;
308 static inline void
309 renew_client(struct nfs4_client *clp)
312 * Move client to the end to the LRU list.
314 dprintk("renewing client (clientid %08x/%08x)\n",
315 clp->cl_clientid.cl_boot,
316 clp->cl_clientid.cl_id);
317 list_move_tail(&clp->cl_lru, &client_lru);
318 clp->cl_time = get_seconds();
321 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
322 static int
323 STALE_CLIENTID(clientid_t *clid)
325 if (clid->cl_boot == boot_time)
326 return 0;
327 dprintk("NFSD stale clientid (%08x/%08x)\n",
328 clid->cl_boot, clid->cl_id);
329 return 1;
333 * XXX Should we use a slab cache ?
334 * This type of memory management is somewhat inefficient, but we use it
335 * anyway since SETCLIENTID is not a common operation.
337 static inline struct nfs4_client *
338 alloc_client(struct xdr_netobj name)
340 struct nfs4_client *clp;
342 if ((clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL))!= NULL) {
343 if ((clp->cl_name.data = kmalloc(name.len, GFP_KERNEL)) != NULL) {
344 memcpy(clp->cl_name.data, name.data, name.len);
345 clp->cl_name.len = name.len;
347 else {
348 kfree(clp);
349 clp = NULL;
352 return clp;
355 static inline void
356 free_client(struct nfs4_client *clp)
358 if (clp->cl_cred.cr_group_info)
359 put_group_info(clp->cl_cred.cr_group_info);
360 kfree(clp->cl_name.data);
361 kfree(clp);
364 void
365 put_nfs4_client(struct nfs4_client *clp)
367 if (atomic_dec_and_test(&clp->cl_count))
368 free_client(clp);
371 static void
372 shutdown_callback_client(struct nfs4_client *clp)
374 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
376 /* shutdown rpc client, ending any outstanding recall rpcs */
377 if (clnt) {
378 clp->cl_callback.cb_client = NULL;
379 rpc_shutdown_client(clnt);
380 rpciod_down();
384 static void
385 expire_client(struct nfs4_client *clp)
387 struct nfs4_stateowner *sop;
388 struct nfs4_delegation *dp;
389 struct list_head reaplist;
391 dprintk("NFSD: expire_client cl_count %d\n",
392 atomic_read(&clp->cl_count));
394 shutdown_callback_client(clp);
396 INIT_LIST_HEAD(&reaplist);
397 spin_lock(&recall_lock);
398 while (!list_empty(&clp->cl_delegations)) {
399 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
400 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
401 dp->dl_flock);
402 list_del_init(&dp->dl_perclnt);
403 list_move(&dp->dl_recall_lru, &reaplist);
405 spin_unlock(&recall_lock);
406 while (!list_empty(&reaplist)) {
407 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
408 list_del_init(&dp->dl_recall_lru);
409 unhash_delegation(dp);
411 list_del(&clp->cl_idhash);
412 list_del(&clp->cl_strhash);
413 list_del(&clp->cl_lru);
414 while (!list_empty(&clp->cl_openowners)) {
415 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
416 release_stateowner(sop);
418 put_nfs4_client(clp);
421 static struct nfs4_client *
422 create_client(struct xdr_netobj name, char *recdir) {
423 struct nfs4_client *clp;
425 if (!(clp = alloc_client(name)))
426 goto out;
427 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
428 atomic_set(&clp->cl_count, 1);
429 atomic_set(&clp->cl_callback.cb_set, 0);
430 INIT_LIST_HEAD(&clp->cl_idhash);
431 INIT_LIST_HEAD(&clp->cl_strhash);
432 INIT_LIST_HEAD(&clp->cl_openowners);
433 INIT_LIST_HEAD(&clp->cl_delegations);
434 INIT_LIST_HEAD(&clp->cl_lru);
435 out:
436 return clp;
439 static void
440 copy_verf(struct nfs4_client *target, nfs4_verifier *source) {
441 memcpy(target->cl_verifier.data, source->data, sizeof(target->cl_verifier.data));
444 static void
445 copy_clid(struct nfs4_client *target, struct nfs4_client *source) {
446 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
447 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
450 static void
451 copy_cred(struct svc_cred *target, struct svc_cred *source) {
453 target->cr_uid = source->cr_uid;
454 target->cr_gid = source->cr_gid;
455 target->cr_group_info = source->cr_group_info;
456 get_group_info(target->cr_group_info);
459 static inline int
460 same_name(const char *n1, const char *n2) {
461 return 0 == memcmp(n1, n2, HEXDIR_LEN);
464 static int
465 cmp_verf(nfs4_verifier *v1, nfs4_verifier *v2) {
466 return(!memcmp(v1->data,v2->data,sizeof(v1->data)));
469 static int
470 cmp_clid(clientid_t * cl1, clientid_t * cl2) {
471 return((cl1->cl_boot == cl2->cl_boot) &&
472 (cl1->cl_id == cl2->cl_id));
475 /* XXX what about NGROUP */
476 static int
477 cmp_creds(struct svc_cred *cr1, struct svc_cred *cr2){
478 return(cr1->cr_uid == cr2->cr_uid);
482 static void
483 gen_clid(struct nfs4_client *clp) {
484 clp->cl_clientid.cl_boot = boot_time;
485 clp->cl_clientid.cl_id = current_clientid++;
488 static void
489 gen_confirm(struct nfs4_client *clp) {
490 struct timespec tv;
491 u32 * p;
493 tv = CURRENT_TIME;
494 p = (u32 *)clp->cl_confirm.data;
495 *p++ = tv.tv_sec;
496 *p++ = tv.tv_nsec;
499 static int
500 check_name(struct xdr_netobj name) {
502 if (name.len == 0)
503 return 0;
504 if (name.len > NFS4_OPAQUE_LIMIT) {
505 printk("NFSD: check_name: name too long(%d)!\n", name.len);
506 return 0;
508 return 1;
511 static void
512 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
514 unsigned int idhashval;
516 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
517 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
518 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
519 list_add_tail(&clp->cl_lru, &client_lru);
520 clp->cl_time = get_seconds();
523 static void
524 move_to_confirmed(struct nfs4_client *clp)
526 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
527 unsigned int strhashval;
529 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
530 list_del_init(&clp->cl_strhash);
531 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
532 strhashval = clientstr_hashval(clp->cl_recdir);
533 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
534 renew_client(clp);
537 static struct nfs4_client *
538 find_confirmed_client(clientid_t *clid)
540 struct nfs4_client *clp;
541 unsigned int idhashval = clientid_hashval(clid->cl_id);
543 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
544 if (cmp_clid(&clp->cl_clientid, clid))
545 return clp;
547 return NULL;
550 static struct nfs4_client *
551 find_unconfirmed_client(clientid_t *clid)
553 struct nfs4_client *clp;
554 unsigned int idhashval = clientid_hashval(clid->cl_id);
556 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
557 if (cmp_clid(&clp->cl_clientid, clid))
558 return clp;
560 return NULL;
563 static struct nfs4_client *
564 find_confirmed_client_by_str(const char *dname, unsigned int hashval)
566 struct nfs4_client *clp;
568 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
569 if (same_name(clp->cl_recdir, dname))
570 return clp;
572 return NULL;
575 static struct nfs4_client *
576 find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
578 struct nfs4_client *clp;
580 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
581 if (same_name(clp->cl_recdir, dname))
582 return clp;
584 return NULL;
587 /* a helper function for parse_callback */
588 static int
589 parse_octet(unsigned int *lenp, char **addrp)
591 unsigned int len = *lenp;
592 char *p = *addrp;
593 int n = -1;
594 char c;
596 for (;;) {
597 if (!len)
598 break;
599 len--;
600 c = *p++;
601 if (c == '.')
602 break;
603 if ((c < '0') || (c > '9')) {
604 n = -1;
605 break;
607 if (n < 0)
608 n = 0;
609 n = (n * 10) + (c - '0');
610 if (n > 255) {
611 n = -1;
612 break;
615 *lenp = len;
616 *addrp = p;
617 return n;
620 /* parse and set the setclientid ipv4 callback address */
621 static int
622 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
624 int temp = 0;
625 u32 cbaddr = 0;
626 u16 cbport = 0;
627 u32 addrlen = addr_len;
628 char *addr = addr_val;
629 int i, shift;
631 /* ipaddress */
632 shift = 24;
633 for(i = 4; i > 0 ; i--) {
634 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
635 return 0;
637 cbaddr |= (temp << shift);
638 if (shift > 0)
639 shift -= 8;
641 *cbaddrp = cbaddr;
643 /* port */
644 shift = 8;
645 for(i = 2; i > 0 ; i--) {
646 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
647 return 0;
649 cbport |= (temp << shift);
650 if (shift > 0)
651 shift -= 8;
653 *cbportp = cbport;
654 return 1;
657 static void
658 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
660 struct nfs4_callback *cb = &clp->cl_callback;
662 /* Currently, we only support tcp for the callback channel */
663 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
664 goto out_err;
666 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
667 &cb->cb_addr, &cb->cb_port)))
668 goto out_err;
669 cb->cb_prog = se->se_callback_prog;
670 cb->cb_ident = se->se_callback_ident;
671 return;
672 out_err:
673 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
674 "will not receive delegations\n",
675 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
677 return;
681 * RFC 3010 has a complex implmentation description of processing a
682 * SETCLIENTID request consisting of 5 bullets, labeled as
683 * CASE0 - CASE4 below.
685 * NOTES:
686 * callback information will be processed in a future patch
688 * an unconfirmed record is added when:
689 * NORMAL (part of CASE 4): there is no confirmed nor unconfirmed record.
690 * CASE 1: confirmed record found with matching name, principal,
691 * verifier, and clientid.
692 * CASE 2: confirmed record found with matching name, principal,
693 * and there is no unconfirmed record with matching
694 * name and principal
696 * an unconfirmed record is replaced when:
697 * CASE 3: confirmed record found with matching name, principal,
698 * and an unconfirmed record is found with matching
699 * name, principal, and with clientid and
700 * confirm that does not match the confirmed record.
701 * CASE 4: there is no confirmed record with matching name and
702 * principal. there is an unconfirmed record with
703 * matching name, principal.
705 * an unconfirmed record is deleted when:
706 * CASE 1: an unconfirmed record that matches input name, verifier,
707 * and confirmed clientid.
708 * CASE 4: any unconfirmed records with matching name and principal
709 * that exist after an unconfirmed record has been replaced
710 * as described above.
714 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
716 u32 ip_addr = rqstp->rq_addr.sin_addr.s_addr;
717 struct xdr_netobj clname = {
718 .len = setclid->se_namelen,
719 .data = setclid->se_name,
721 nfs4_verifier clverifier = setclid->se_verf;
722 unsigned int strhashval;
723 struct nfs4_client *conf, *unconf, *new;
724 int status;
725 char dname[HEXDIR_LEN];
727 if (!check_name(clname))
728 return nfserr_inval;
730 status = nfs4_make_rec_clidname(dname, &clname);
731 if (status)
732 return status;
735 * XXX The Duplicate Request Cache (DRC) has been checked (??)
736 * We get here on a DRC miss.
739 strhashval = clientstr_hashval(dname);
741 nfs4_lock_state();
742 conf = find_confirmed_client_by_str(dname, strhashval);
743 if (conf) {
745 * CASE 0:
746 * clname match, confirmed, different principal
747 * or different ip_address
749 status = nfserr_clid_inuse;
750 if (!cmp_creds(&conf->cl_cred, &rqstp->rq_cred)
751 || conf->cl_addr != ip_addr) {
752 printk("NFSD: setclientid: string in use by client"
753 "(clientid %08x/%08x)\n",
754 conf->cl_clientid.cl_boot, conf->cl_clientid.cl_id);
755 goto out;
758 unconf = find_unconfirmed_client_by_str(dname, strhashval);
759 status = nfserr_resource;
760 if (!conf) {
762 * CASE 4:
763 * placed first, because it is the normal case.
765 if (unconf)
766 expire_client(unconf);
767 new = create_client(clname, dname);
768 if (new == NULL)
769 goto out;
770 copy_verf(new, &clverifier);
771 new->cl_addr = ip_addr;
772 copy_cred(&new->cl_cred,&rqstp->rq_cred);
773 gen_clid(new);
774 gen_confirm(new);
775 gen_callback(new, setclid);
776 add_to_unconfirmed(new, strhashval);
777 } else if (cmp_verf(&conf->cl_verifier, &clverifier)) {
779 * CASE 1:
780 * cl_name match, confirmed, principal match
781 * verifier match: probable callback update
783 * remove any unconfirmed nfs4_client with
784 * matching cl_name, cl_verifier, and cl_clientid
786 * create and insert an unconfirmed nfs4_client with same
787 * cl_name, cl_verifier, and cl_clientid as existing
788 * nfs4_client, but with the new callback info and a
789 * new cl_confirm
791 if (unconf) {
792 /* Note this is removing unconfirmed {*x***},
793 * which is stronger than RFC recommended {vxc**}.
794 * This has the advantage that there is at most
795 * one {*x***} in either list at any time.
797 expire_client(unconf);
799 new = create_client(clname, dname);
800 if (new == NULL)
801 goto out;
802 copy_verf(new,&conf->cl_verifier);
803 new->cl_addr = ip_addr;
804 copy_cred(&new->cl_cred,&rqstp->rq_cred);
805 copy_clid(new, conf);
806 gen_confirm(new);
807 gen_callback(new, setclid);
808 add_to_unconfirmed(new,strhashval);
809 } else if (!unconf) {
811 * CASE 2:
812 * clname match, confirmed, principal match
813 * verfier does not match
814 * no unconfirmed. create a new unconfirmed nfs4_client
815 * using input clverifier, clname, and callback info
816 * and generate a new cl_clientid and cl_confirm.
818 new = create_client(clname, dname);
819 if (new == NULL)
820 goto out;
821 copy_verf(new,&clverifier);
822 new->cl_addr = ip_addr;
823 copy_cred(&new->cl_cred,&rqstp->rq_cred);
824 gen_clid(new);
825 gen_confirm(new);
826 gen_callback(new, setclid);
827 add_to_unconfirmed(new, strhashval);
828 } else if (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm)) {
830 * CASE3:
831 * confirmed found (name, principal match)
832 * confirmed verifier does not match input clverifier
834 * unconfirmed found (name match)
835 * confirmed->cl_confirm != unconfirmed->cl_confirm
837 * remove unconfirmed.
839 * create an unconfirmed nfs4_client
840 * with same cl_name as existing confirmed nfs4_client,
841 * but with new callback info, new cl_clientid,
842 * new cl_verifier and a new cl_confirm
844 expire_client(unconf);
845 new = create_client(clname, dname);
846 if (new == NULL)
847 goto out;
848 copy_verf(new,&clverifier);
849 new->cl_addr = ip_addr;
850 copy_cred(&new->cl_cred,&rqstp->rq_cred);
851 gen_clid(new);
852 gen_confirm(new);
853 gen_callback(new, setclid);
854 add_to_unconfirmed(new, strhashval);
855 } else {
856 /* No cases hit !!! */
857 status = nfserr_inval;
858 goto out;
861 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
862 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
863 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
864 status = nfs_ok;
865 out:
866 nfs4_unlock_state();
867 return status;
872 * RFC 3010 has a complex implmentation description of processing a
873 * SETCLIENTID_CONFIRM request consisting of 4 bullets describing
874 * processing on a DRC miss, labeled as CASE1 - CASE4 below.
876 * NOTE: callback information will be processed here in a future patch
879 nfsd4_setclientid_confirm(struct svc_rqst *rqstp, struct nfsd4_setclientid_confirm *setclientid_confirm)
881 u32 ip_addr = rqstp->rq_addr.sin_addr.s_addr;
882 struct nfs4_client *conf, *unconf;
883 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
884 clientid_t * clid = &setclientid_confirm->sc_clientid;
885 int status;
887 if (STALE_CLIENTID(clid))
888 return nfserr_stale_clientid;
890 * XXX The Duplicate Request Cache (DRC) has been checked (??)
891 * We get here on a DRC miss.
894 nfs4_lock_state();
896 conf = find_confirmed_client(clid);
897 unconf = find_unconfirmed_client(clid);
899 status = nfserr_clid_inuse;
900 if (conf && conf->cl_addr != ip_addr)
901 goto out;
902 if (unconf && unconf->cl_addr != ip_addr)
903 goto out;
905 if ((conf && unconf) &&
906 (cmp_verf(&unconf->cl_confirm, &confirm)) &&
907 (cmp_verf(&conf->cl_verifier, &unconf->cl_verifier)) &&
908 (same_name(conf->cl_recdir,unconf->cl_recdir)) &&
909 (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm))) {
910 /* CASE 1:
911 * unconf record that matches input clientid and input confirm.
912 * conf record that matches input clientid.
913 * conf and unconf records match names, verifiers
915 if (!cmp_creds(&conf->cl_cred, &unconf->cl_cred))
916 status = nfserr_clid_inuse;
917 else {
918 /* XXX: We just turn off callbacks until we can handle
919 * change request correctly. */
920 atomic_set(&conf->cl_callback.cb_set, 0);
921 gen_confirm(conf);
922 nfsd4_remove_clid_dir(unconf);
923 expire_client(unconf);
924 status = nfs_ok;
927 } else if ((conf && !unconf) ||
928 ((conf && unconf) &&
929 (!cmp_verf(&conf->cl_verifier, &unconf->cl_verifier) ||
930 !same_name(conf->cl_recdir, unconf->cl_recdir)))) {
931 /* CASE 2:
932 * conf record that matches input clientid.
933 * if unconf record matches input clientid, then
934 * unconf->cl_name or unconf->cl_verifier don't match the
935 * conf record.
937 if (!cmp_creds(&conf->cl_cred,&rqstp->rq_cred))
938 status = nfserr_clid_inuse;
939 else
940 status = nfs_ok;
941 } else if (!conf && unconf
942 && cmp_verf(&unconf->cl_confirm, &confirm)) {
943 /* CASE 3:
944 * conf record not found.
945 * unconf record found.
946 * unconf->cl_confirm matches input confirm
948 if (!cmp_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
949 status = nfserr_clid_inuse;
950 } else {
951 unsigned int hash =
952 clientstr_hashval(unconf->cl_recdir);
953 conf = find_confirmed_client_by_str(unconf->cl_recdir,
954 hash);
955 if (conf) {
956 nfsd4_remove_clid_dir(conf);
957 expire_client(conf);
959 move_to_confirmed(unconf);
960 conf = unconf;
961 status = nfs_ok;
963 } else if ((!conf || (conf && !cmp_verf(&conf->cl_confirm, &confirm)))
964 && (!unconf || (unconf && !cmp_verf(&unconf->cl_confirm,
965 &confirm)))) {
966 /* CASE 4:
967 * conf record not found, or if conf, conf->cl_confirm does not
968 * match input confirm.
969 * unconf record not found, or if unconf, unconf->cl_confirm
970 * does not match input confirm.
972 status = nfserr_stale_clientid;
973 } else {
974 /* check that we have hit one of the cases...*/
975 status = nfserr_clid_inuse;
977 out:
978 if (!status)
979 nfsd4_probe_callback(conf);
980 nfs4_unlock_state();
981 return status;
984 /* OPEN Share state helper functions */
985 static inline struct nfs4_file *
986 alloc_init_file(struct inode *ino)
988 struct nfs4_file *fp;
989 unsigned int hashval = file_hashval(ino);
991 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
992 if (fp) {
993 kref_init(&fp->fi_ref);
994 INIT_LIST_HEAD(&fp->fi_hash);
995 INIT_LIST_HEAD(&fp->fi_stateids);
996 INIT_LIST_HEAD(&fp->fi_delegations);
997 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
998 fp->fi_inode = igrab(ino);
999 fp->fi_id = current_fileid++;
1000 return fp;
1002 return NULL;
1005 static void
1006 nfsd4_free_slab(kmem_cache_t **slab)
1008 if (*slab == NULL)
1009 return;
1010 kmem_cache_destroy(*slab);
1011 *slab = NULL;
1014 static void
1015 nfsd4_free_slabs(void)
1017 nfsd4_free_slab(&stateowner_slab);
1018 nfsd4_free_slab(&file_slab);
1019 nfsd4_free_slab(&stateid_slab);
1020 nfsd4_free_slab(&deleg_slab);
1023 static int
1024 nfsd4_init_slabs(void)
1026 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1027 sizeof(struct nfs4_stateowner), 0, 0, NULL, NULL);
1028 if (stateowner_slab == NULL)
1029 goto out_nomem;
1030 file_slab = kmem_cache_create("nfsd4_files",
1031 sizeof(struct nfs4_file), 0, 0, NULL, NULL);
1032 if (file_slab == NULL)
1033 goto out_nomem;
1034 stateid_slab = kmem_cache_create("nfsd4_stateids",
1035 sizeof(struct nfs4_stateid), 0, 0, NULL, NULL);
1036 if (stateid_slab == NULL)
1037 goto out_nomem;
1038 deleg_slab = kmem_cache_create("nfsd4_delegations",
1039 sizeof(struct nfs4_delegation), 0, 0, NULL, NULL);
1040 if (deleg_slab == NULL)
1041 goto out_nomem;
1042 return 0;
1043 out_nomem:
1044 nfsd4_free_slabs();
1045 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1046 return -ENOMEM;
1049 void
1050 nfs4_free_stateowner(struct kref *kref)
1052 struct nfs4_stateowner *sop =
1053 container_of(kref, struct nfs4_stateowner, so_ref);
1054 kfree(sop->so_owner.data);
1055 kmem_cache_free(stateowner_slab, sop);
1058 static inline struct nfs4_stateowner *
1059 alloc_stateowner(struct xdr_netobj *owner)
1061 struct nfs4_stateowner *sop;
1063 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1064 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1065 memcpy(sop->so_owner.data, owner->data, owner->len);
1066 sop->so_owner.len = owner->len;
1067 kref_init(&sop->so_ref);
1068 return sop;
1070 kmem_cache_free(stateowner_slab, sop);
1072 return NULL;
1075 static struct nfs4_stateowner *
1076 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1077 struct nfs4_stateowner *sop;
1078 struct nfs4_replay *rp;
1079 unsigned int idhashval;
1081 if (!(sop = alloc_stateowner(&open->op_owner)))
1082 return NULL;
1083 idhashval = ownerid_hashval(current_ownerid);
1084 INIT_LIST_HEAD(&sop->so_idhash);
1085 INIT_LIST_HEAD(&sop->so_strhash);
1086 INIT_LIST_HEAD(&sop->so_perclient);
1087 INIT_LIST_HEAD(&sop->so_stateids);
1088 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1089 INIT_LIST_HEAD(&sop->so_close_lru);
1090 sop->so_time = 0;
1091 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1092 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1093 list_add(&sop->so_perclient, &clp->cl_openowners);
1094 sop->so_is_open_owner = 1;
1095 sop->so_id = current_ownerid++;
1096 sop->so_client = clp;
1097 sop->so_seqid = open->op_seqid;
1098 sop->so_confirmed = 0;
1099 rp = &sop->so_replay;
1100 rp->rp_status = nfserr_serverfault;
1101 rp->rp_buflen = 0;
1102 rp->rp_buf = rp->rp_ibuf;
1103 return sop;
1106 static void
1107 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1109 struct nfs4_stateowner *lock_sop;
1111 while (!list_empty(&open_stp->st_lockowners)) {
1112 lock_sop = list_entry(open_stp->st_lockowners.next,
1113 struct nfs4_stateowner, so_perstateid);
1114 /* list_del(&open_stp->st_lockowners); */
1115 BUG_ON(lock_sop->so_is_open_owner);
1116 release_stateowner(lock_sop);
1120 static void
1121 unhash_stateowner(struct nfs4_stateowner *sop)
1123 struct nfs4_stateid *stp;
1125 list_del(&sop->so_idhash);
1126 list_del(&sop->so_strhash);
1127 if (sop->so_is_open_owner)
1128 list_del(&sop->so_perclient);
1129 list_del(&sop->so_perstateid);
1130 while (!list_empty(&sop->so_stateids)) {
1131 stp = list_entry(sop->so_stateids.next,
1132 struct nfs4_stateid, st_perstateowner);
1133 if (sop->so_is_open_owner)
1134 release_stateid(stp, OPEN_STATE);
1135 else
1136 release_stateid(stp, LOCK_STATE);
1140 static void
1141 release_stateowner(struct nfs4_stateowner *sop)
1143 unhash_stateowner(sop);
1144 list_del(&sop->so_close_lru);
1145 nfs4_put_stateowner(sop);
1148 static inline void
1149 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1150 struct nfs4_stateowner *sop = open->op_stateowner;
1151 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1153 INIT_LIST_HEAD(&stp->st_hash);
1154 INIT_LIST_HEAD(&stp->st_perstateowner);
1155 INIT_LIST_HEAD(&stp->st_lockowners);
1156 INIT_LIST_HEAD(&stp->st_perfile);
1157 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1158 list_add(&stp->st_perstateowner, &sop->so_stateids);
1159 list_add(&stp->st_perfile, &fp->fi_stateids);
1160 stp->st_stateowner = sop;
1161 get_nfs4_file(fp);
1162 stp->st_file = fp;
1163 stp->st_stateid.si_boot = boot_time;
1164 stp->st_stateid.si_stateownerid = sop->so_id;
1165 stp->st_stateid.si_fileid = fp->fi_id;
1166 stp->st_stateid.si_generation = 0;
1167 stp->st_access_bmap = 0;
1168 stp->st_deny_bmap = 0;
1169 __set_bit(open->op_share_access, &stp->st_access_bmap);
1170 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1171 stp->st_openstp = NULL;
1174 static void
1175 release_stateid(struct nfs4_stateid *stp, int flags)
1177 struct file *filp = stp->st_vfs_file;
1179 list_del(&stp->st_hash);
1180 list_del(&stp->st_perfile);
1181 list_del(&stp->st_perstateowner);
1182 if (flags & OPEN_STATE) {
1183 release_stateid_lockowners(stp);
1184 stp->st_vfs_file = NULL;
1185 nfsd_close(filp);
1186 } else if (flags & LOCK_STATE)
1187 locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
1188 put_nfs4_file(stp->st_file);
1189 kmem_cache_free(stateid_slab, stp);
1192 static void
1193 move_to_close_lru(struct nfs4_stateowner *sop)
1195 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1197 list_move_tail(&sop->so_close_lru, &close_lru);
1198 sop->so_time = get_seconds();
1201 static int
1202 cmp_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner, clientid_t *clid) {
1203 return ((sop->so_owner.len == owner->len) &&
1204 !memcmp(sop->so_owner.data, owner->data, owner->len) &&
1205 (sop->so_client->cl_clientid.cl_id == clid->cl_id));
1208 static struct nfs4_stateowner *
1209 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1211 struct nfs4_stateowner *so = NULL;
1213 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1214 if (cmp_owner_str(so, &open->op_owner, &open->op_clientid))
1215 return so;
1217 return NULL;
1220 /* search file_hashtbl[] for file */
1221 static struct nfs4_file *
1222 find_file(struct inode *ino)
1224 unsigned int hashval = file_hashval(ino);
1225 struct nfs4_file *fp;
1227 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1228 if (fp->fi_inode == ino) {
1229 get_nfs4_file(fp);
1230 return fp;
1233 return NULL;
1236 static int access_valid(u32 x)
1238 return (x > 0 && x < 4);
1241 static int deny_valid(u32 x)
1243 return (x >= 0 && x < 5);
1246 static void
1247 set_access(unsigned int *access, unsigned long bmap) {
1248 int i;
1250 *access = 0;
1251 for (i = 1; i < 4; i++) {
1252 if (test_bit(i, &bmap))
1253 *access |= i;
1257 static void
1258 set_deny(unsigned int *deny, unsigned long bmap) {
1259 int i;
1261 *deny = 0;
1262 for (i = 0; i < 4; i++) {
1263 if (test_bit(i, &bmap))
1264 *deny |= i ;
1268 static int
1269 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1270 unsigned int access, deny;
1272 set_access(&access, stp->st_access_bmap);
1273 set_deny(&deny, stp->st_deny_bmap);
1274 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1275 return 0;
1276 return 1;
1280 * Called to check deny when READ with all zero stateid or
1281 * WRITE with all zero or all one stateid
1283 static int
1284 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1286 struct inode *ino = current_fh->fh_dentry->d_inode;
1287 struct nfs4_file *fp;
1288 struct nfs4_stateid *stp;
1289 int ret;
1291 dprintk("NFSD: nfs4_share_conflict\n");
1293 fp = find_file(ino);
1294 if (!fp)
1295 return nfs_ok;
1296 ret = nfserr_locked;
1297 /* Search for conflicting share reservations */
1298 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1299 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1300 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1301 goto out;
1303 ret = nfs_ok;
1304 out:
1305 put_nfs4_file(fp);
1306 return ret;
1309 static inline void
1310 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1312 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1313 put_write_access(filp->f_dentry->d_inode);
1314 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1319 * Recall a delegation
1321 static int
1322 do_recall(void *__dp)
1324 struct nfs4_delegation *dp = __dp;
1326 daemonize("nfsv4-recall");
1328 nfsd4_cb_recall(dp);
1329 return 0;
1333 * Spawn a thread to perform a recall on the delegation represented
1334 * by the lease (file_lock)
1336 * Called from break_lease() with lock_kernel() held.
1337 * Note: we assume break_lease will only call this *once* for any given
1338 * lease.
1340 static
1341 void nfsd_break_deleg_cb(struct file_lock *fl)
1343 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1344 struct task_struct *t;
1346 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1347 if (!dp)
1348 return;
1350 /* We're assuming the state code never drops its reference
1351 * without first removing the lease. Since we're in this lease
1352 * callback (and since the lease code is serialized by the kernel
1353 * lock) we know the server hasn't removed the lease yet, we know
1354 * it's safe to take a reference: */
1355 atomic_inc(&dp->dl_count);
1357 spin_lock(&recall_lock);
1358 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1359 spin_unlock(&recall_lock);
1361 /* only place dl_time is set. protected by lock_kernel*/
1362 dp->dl_time = get_seconds();
1364 /* XXX need to merge NFSD_LEASE_TIME with fs/locks.c:lease_break_time */
1365 fl->fl_break_time = jiffies + NFSD_LEASE_TIME * HZ;
1367 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1368 if (IS_ERR(t)) {
1369 struct nfs4_client *clp = dp->dl_client;
1371 printk(KERN_INFO "NFSD: Callback thread failed for "
1372 "for client (clientid %08x/%08x)\n",
1373 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1374 nfs4_put_delegation(dp);
1379 * The file_lock is being reapd.
1381 * Called by locks_free_lock() with lock_kernel() held.
1383 static
1384 void nfsd_release_deleg_cb(struct file_lock *fl)
1386 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1388 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1390 if (!(fl->fl_flags & FL_LEASE) || !dp)
1391 return;
1392 dp->dl_flock = NULL;
1396 * Set the delegation file_lock back pointer.
1398 * Called from __setlease() with lock_kernel() held.
1400 static
1401 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1403 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1405 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1406 if (!dp)
1407 return;
1408 dp->dl_flock = new;
1412 * Called from __setlease() with lock_kernel() held
1414 static
1415 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1417 struct nfs4_delegation *onlistd =
1418 (struct nfs4_delegation *)onlist->fl_owner;
1419 struct nfs4_delegation *tryd =
1420 (struct nfs4_delegation *)try->fl_owner;
1422 if (onlist->fl_lmops != try->fl_lmops)
1423 return 0;
1425 return onlistd->dl_client == tryd->dl_client;
1429 static
1430 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1432 if (arg & F_UNLCK)
1433 return lease_modify(onlist, arg);
1434 else
1435 return -EAGAIN;
1438 static struct lock_manager_operations nfsd_lease_mng_ops = {
1439 .fl_break = nfsd_break_deleg_cb,
1440 .fl_release_private = nfsd_release_deleg_cb,
1441 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1442 .fl_mylease = nfsd_same_client_deleg_cb,
1443 .fl_change = nfsd_change_deleg_cb,
1448 nfsd4_process_open1(struct nfsd4_open *open)
1450 clientid_t *clientid = &open->op_clientid;
1451 struct nfs4_client *clp = NULL;
1452 unsigned int strhashval;
1453 struct nfs4_stateowner *sop = NULL;
1455 if (!check_name(open->op_owner))
1456 return nfserr_inval;
1458 if (STALE_CLIENTID(&open->op_clientid))
1459 return nfserr_stale_clientid;
1461 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1462 sop = find_openstateowner_str(strhashval, open);
1463 open->op_stateowner = sop;
1464 if (!sop) {
1465 /* Make sure the client's lease hasn't expired. */
1466 clp = find_confirmed_client(clientid);
1467 if (clp == NULL)
1468 return nfserr_expired;
1469 goto renew;
1471 if (!sop->so_confirmed) {
1472 /* Replace unconfirmed owners without checking for replay. */
1473 clp = sop->so_client;
1474 release_stateowner(sop);
1475 open->op_stateowner = NULL;
1476 goto renew;
1478 if (open->op_seqid == sop->so_seqid - 1) {
1479 if (sop->so_replay.rp_buflen)
1480 return NFSERR_REPLAY_ME;
1481 /* The original OPEN failed so spectacularly
1482 * that we don't even have replay data saved!
1483 * Therefore, we have no choice but to continue
1484 * processing this OPEN; presumably, we'll
1485 * fail again for the same reason.
1487 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1488 goto renew;
1490 if (open->op_seqid != sop->so_seqid)
1491 return nfserr_bad_seqid;
1492 renew:
1493 if (open->op_stateowner == NULL) {
1494 sop = alloc_init_open_stateowner(strhashval, clp, open);
1495 if (sop == NULL)
1496 return nfserr_resource;
1497 open->op_stateowner = sop;
1499 list_del_init(&sop->so_close_lru);
1500 renew_client(sop->so_client);
1501 return nfs_ok;
1504 static inline int
1505 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1507 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1508 return nfserr_openmode;
1509 else
1510 return nfs_ok;
1513 static struct nfs4_delegation *
1514 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1516 struct nfs4_delegation *dp;
1518 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
1519 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1520 return dp;
1522 return NULL;
1525 static int
1526 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1527 struct nfs4_delegation **dp)
1529 int flags;
1530 int status = nfserr_bad_stateid;
1532 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1533 if (*dp == NULL)
1534 goto out;
1535 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1536 RD_STATE : WR_STATE;
1537 status = nfs4_check_delegmode(*dp, flags);
1538 if (status)
1539 *dp = NULL;
1540 out:
1541 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1542 return nfs_ok;
1543 if (status)
1544 return status;
1545 open->op_stateowner->so_confirmed = 1;
1546 return nfs_ok;
1549 static int
1550 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1552 struct nfs4_stateid *local;
1553 int status = nfserr_share_denied;
1554 struct nfs4_stateowner *sop = open->op_stateowner;
1556 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1557 /* ignore lock owners */
1558 if (local->st_stateowner->so_is_open_owner == 0)
1559 continue;
1560 /* remember if we have seen this open owner */
1561 if (local->st_stateowner == sop)
1562 *stpp = local;
1563 /* check for conflicting share reservations */
1564 if (!test_share(local, open))
1565 goto out;
1567 status = 0;
1568 out:
1569 return status;
1572 static inline struct nfs4_stateid *
1573 nfs4_alloc_stateid(void)
1575 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1578 static int
1579 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1580 struct nfs4_delegation *dp,
1581 struct svc_fh *cur_fh, int flags)
1583 struct nfs4_stateid *stp;
1585 stp = nfs4_alloc_stateid();
1586 if (stp == NULL)
1587 return nfserr_resource;
1589 if (dp) {
1590 get_file(dp->dl_vfs_file);
1591 stp->st_vfs_file = dp->dl_vfs_file;
1592 } else {
1593 int status;
1594 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1595 &stp->st_vfs_file);
1596 if (status) {
1597 if (status == nfserr_dropit)
1598 status = nfserr_jukebox;
1599 kmem_cache_free(stateid_slab, stp);
1600 return status;
1603 *stpp = stp;
1604 return 0;
1607 static inline int
1608 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1609 struct nfsd4_open *open)
1611 struct iattr iattr = {
1612 .ia_valid = ATTR_SIZE,
1613 .ia_size = 0,
1615 if (!open->op_truncate)
1616 return 0;
1617 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1618 return nfserr_inval;
1619 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1622 static int
1623 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1625 struct file *filp = stp->st_vfs_file;
1626 struct inode *inode = filp->f_dentry->d_inode;
1627 unsigned int share_access, new_writer;
1628 int status;
1630 set_access(&share_access, stp->st_access_bmap);
1631 new_writer = (~share_access) & open->op_share_access
1632 & NFS4_SHARE_ACCESS_WRITE;
1634 if (new_writer) {
1635 status = get_write_access(inode);
1636 if (status)
1637 return nfserrno(status);
1639 status = nfsd4_truncate(rqstp, cur_fh, open);
1640 if (status) {
1641 if (new_writer)
1642 put_write_access(inode);
1643 return status;
1645 /* remember the open */
1646 filp->f_mode |= open->op_share_access;
1647 set_bit(open->op_share_access, &stp->st_access_bmap);
1648 set_bit(open->op_share_deny, &stp->st_deny_bmap);
1650 return nfs_ok;
1654 static void
1655 nfs4_set_claim_prev(struct nfsd4_open *open)
1657 open->op_stateowner->so_confirmed = 1;
1658 open->op_stateowner->so_client->cl_firststate = 1;
1662 * Attempt to hand out a delegation.
1664 static void
1665 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1667 struct nfs4_delegation *dp;
1668 struct nfs4_stateowner *sop = stp->st_stateowner;
1669 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1670 struct file_lock fl, *flp = &fl;
1671 int status, flag = 0;
1673 flag = NFS4_OPEN_DELEGATE_NONE;
1674 open->op_recall = 0;
1675 switch (open->op_claim_type) {
1676 case NFS4_OPEN_CLAIM_PREVIOUS:
1677 if (!atomic_read(&cb->cb_set))
1678 open->op_recall = 1;
1679 flag = open->op_delegate_type;
1680 if (flag == NFS4_OPEN_DELEGATE_NONE)
1681 goto out;
1682 break;
1683 case NFS4_OPEN_CLAIM_NULL:
1684 /* Let's not give out any delegations till everyone's
1685 * had the chance to reclaim theirs.... */
1686 if (nfs4_in_grace())
1687 goto out;
1688 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1689 goto out;
1690 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1691 flag = NFS4_OPEN_DELEGATE_WRITE;
1692 else
1693 flag = NFS4_OPEN_DELEGATE_READ;
1694 break;
1695 default:
1696 goto out;
1699 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1700 if (dp == NULL) {
1701 flag = NFS4_OPEN_DELEGATE_NONE;
1702 goto out;
1704 locks_init_lock(&fl);
1705 fl.fl_lmops = &nfsd_lease_mng_ops;
1706 fl.fl_flags = FL_LEASE;
1707 fl.fl_end = OFFSET_MAX;
1708 fl.fl_owner = (fl_owner_t)dp;
1709 fl.fl_file = stp->st_vfs_file;
1710 fl.fl_pid = current->tgid;
1712 /* setlease checks to see if delegation should be handed out.
1713 * the lock_manager callbacks fl_mylease and fl_change are used
1715 if ((status = setlease(stp->st_vfs_file,
1716 flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK, &flp))) {
1717 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1718 unhash_delegation(dp);
1719 flag = NFS4_OPEN_DELEGATE_NONE;
1720 goto out;
1723 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1725 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1726 dp->dl_stateid.si_boot,
1727 dp->dl_stateid.si_stateownerid,
1728 dp->dl_stateid.si_fileid,
1729 dp->dl_stateid.si_generation);
1730 out:
1731 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1732 && flag == NFS4_OPEN_DELEGATE_NONE
1733 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
1734 printk("NFSD: WARNING: refusing delegation reclaim\n");
1735 open->op_delegate_type = flag;
1739 * called with nfs4_lock_state() held.
1742 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1744 struct nfs4_file *fp = NULL;
1745 struct inode *ino = current_fh->fh_dentry->d_inode;
1746 struct nfs4_stateid *stp = NULL;
1747 struct nfs4_delegation *dp = NULL;
1748 int status;
1750 status = nfserr_inval;
1751 if (!access_valid(open->op_share_access)
1752 || !deny_valid(open->op_share_deny))
1753 goto out;
1755 * Lookup file; if found, lookup stateid and check open request,
1756 * and check for delegations in the process of being recalled.
1757 * If not found, create the nfs4_file struct
1759 fp = find_file(ino);
1760 if (fp) {
1761 if ((status = nfs4_check_open(fp, open, &stp)))
1762 goto out;
1763 status = nfs4_check_deleg(fp, open, &dp);
1764 if (status)
1765 goto out;
1766 } else {
1767 status = nfserr_bad_stateid;
1768 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1769 goto out;
1770 status = nfserr_resource;
1771 fp = alloc_init_file(ino);
1772 if (fp == NULL)
1773 goto out;
1777 * OPEN the file, or upgrade an existing OPEN.
1778 * If truncate fails, the OPEN fails.
1780 if (stp) {
1781 /* Stateid was found, this is an OPEN upgrade */
1782 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1783 if (status)
1784 goto out;
1785 update_stateid(&stp->st_stateid);
1786 } else {
1787 /* Stateid was not found, this is a new OPEN */
1788 int flags = 0;
1789 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
1790 flags |= MAY_READ;
1791 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1792 flags |= MAY_WRITE;
1793 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1794 if (status)
1795 goto out;
1796 init_stateid(stp, fp, open);
1797 status = nfsd4_truncate(rqstp, current_fh, open);
1798 if (status) {
1799 release_stateid(stp, OPEN_STATE);
1800 goto out;
1803 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1806 * Attempt to hand out a delegation. No error return, because the
1807 * OPEN succeeds even if we fail.
1809 nfs4_open_delegation(current_fh, open, stp);
1811 status = nfs_ok;
1813 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1814 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1815 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1816 out:
1817 if (fp)
1818 put_nfs4_file(fp);
1819 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1820 nfs4_set_claim_prev(open);
1822 * To finish the open response, we just need to set the rflags.
1824 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1825 if (!open->op_stateowner->so_confirmed)
1826 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1828 return status;
1831 static struct workqueue_struct *laundry_wq;
1832 static struct work_struct laundromat_work;
1833 static void laundromat_main(void *);
1834 static DECLARE_WORK(laundromat_work, laundromat_main, NULL);
1836 int
1837 nfsd4_renew(clientid_t *clid)
1839 struct nfs4_client *clp;
1840 int status;
1842 nfs4_lock_state();
1843 dprintk("process_renew(%08x/%08x): starting\n",
1844 clid->cl_boot, clid->cl_id);
1845 status = nfserr_stale_clientid;
1846 if (STALE_CLIENTID(clid))
1847 goto out;
1848 clp = find_confirmed_client(clid);
1849 status = nfserr_expired;
1850 if (clp == NULL) {
1851 /* We assume the client took too long to RENEW. */
1852 dprintk("nfsd4_renew: clientid not found!\n");
1853 goto out;
1855 renew_client(clp);
1856 status = nfserr_cb_path_down;
1857 if (!list_empty(&clp->cl_delegations)
1858 && !atomic_read(&clp->cl_callback.cb_set))
1859 goto out;
1860 status = nfs_ok;
1861 out:
1862 nfs4_unlock_state();
1863 return status;
1866 static void
1867 end_grace(void)
1869 dprintk("NFSD: end of grace period\n");
1870 nfsd4_recdir_purge_old();
1871 in_grace = 0;
1874 static time_t
1875 nfs4_laundromat(void)
1877 struct nfs4_client *clp;
1878 struct nfs4_stateowner *sop;
1879 struct nfs4_delegation *dp;
1880 struct list_head *pos, *next, reaplist;
1881 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1882 time_t t, clientid_val = NFSD_LEASE_TIME;
1883 time_t u, test_val = NFSD_LEASE_TIME;
1885 nfs4_lock_state();
1887 dprintk("NFSD: laundromat service - starting\n");
1888 if (in_grace)
1889 end_grace();
1890 list_for_each_safe(pos, next, &client_lru) {
1891 clp = list_entry(pos, struct nfs4_client, cl_lru);
1892 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1893 t = clp->cl_time - cutoff;
1894 if (clientid_val > t)
1895 clientid_val = t;
1896 break;
1898 dprintk("NFSD: purging unused client (clientid %08x)\n",
1899 clp->cl_clientid.cl_id);
1900 nfsd4_remove_clid_dir(clp);
1901 expire_client(clp);
1903 INIT_LIST_HEAD(&reaplist);
1904 spin_lock(&recall_lock);
1905 list_for_each_safe(pos, next, &del_recall_lru) {
1906 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1907 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1908 u = dp->dl_time - cutoff;
1909 if (test_val > u)
1910 test_val = u;
1911 break;
1913 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1914 dp, dp->dl_flock);
1915 list_move(&dp->dl_recall_lru, &reaplist);
1917 spin_unlock(&recall_lock);
1918 list_for_each_safe(pos, next, &reaplist) {
1919 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1920 list_del_init(&dp->dl_recall_lru);
1921 unhash_delegation(dp);
1923 test_val = NFSD_LEASE_TIME;
1924 list_for_each_safe(pos, next, &close_lru) {
1925 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1926 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1927 u = sop->so_time - cutoff;
1928 if (test_val > u)
1929 test_val = u;
1930 break;
1932 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1933 sop->so_id);
1934 release_stateowner(sop);
1936 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1937 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1938 nfs4_unlock_state();
1939 return clientid_val;
1942 void
1943 laundromat_main(void *not_used)
1945 time_t t;
1947 t = nfs4_laundromat();
1948 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1949 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1952 static struct nfs4_stateowner *
1953 search_close_lru(u32 st_id, int flags)
1955 struct nfs4_stateowner *local = NULL;
1957 if (flags & CLOSE_STATE) {
1958 list_for_each_entry(local, &close_lru, so_close_lru) {
1959 if (local->so_id == st_id)
1960 return local;
1963 return NULL;
1966 static inline int
1967 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1969 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_dentry->d_inode;
1972 static int
1973 STALE_STATEID(stateid_t *stateid)
1975 if (stateid->si_boot == boot_time)
1976 return 0;
1977 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1978 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1979 stateid->si_generation);
1980 return 1;
1983 static inline int
1984 access_permit_read(unsigned long access_bmap)
1986 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1987 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1988 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1991 static inline int
1992 access_permit_write(unsigned long access_bmap)
1994 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1995 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
1998 static
1999 int nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2001 int status = nfserr_openmode;
2003 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2004 goto out;
2005 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2006 goto out;
2007 status = nfs_ok;
2008 out:
2009 return status;
2012 static inline int
2013 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2015 /* Trying to call delegreturn with a special stateid? Yuch: */
2016 if (!(flags & (RD_STATE | WR_STATE)))
2017 return nfserr_bad_stateid;
2018 else if (ONE_STATEID(stateid) && (flags & RD_STATE))
2019 return nfs_ok;
2020 else if (nfs4_in_grace()) {
2021 /* Answer in remaining cases depends on existance of
2022 * conflicting state; so we must wait out the grace period. */
2023 return nfserr_grace;
2024 } else if (flags & WR_STATE)
2025 return nfs4_share_conflict(current_fh,
2026 NFS4_SHARE_DENY_WRITE);
2027 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2028 return nfs4_share_conflict(current_fh,
2029 NFS4_SHARE_DENY_READ);
2033 * Allow READ/WRITE during grace period on recovered state only for files
2034 * that are not able to provide mandatory locking.
2036 static inline int
2037 io_during_grace_disallowed(struct inode *inode, int flags)
2039 return nfs4_in_grace() && (flags & (RD_STATE | WR_STATE))
2040 && MANDATORY_LOCK(inode);
2044 * Checks for stateid operations
2047 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2049 struct nfs4_stateid *stp = NULL;
2050 struct nfs4_delegation *dp = NULL;
2051 stateid_t *stidp;
2052 struct inode *ino = current_fh->fh_dentry->d_inode;
2053 int status;
2055 dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2056 stateid->si_boot, stateid->si_stateownerid,
2057 stateid->si_fileid, stateid->si_generation);
2058 if (filpp)
2059 *filpp = NULL;
2061 if (io_during_grace_disallowed(ino, flags))
2062 return nfserr_grace;
2064 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2065 return check_special_stateids(current_fh, stateid, flags);
2067 /* STALE STATEID */
2068 status = nfserr_stale_stateid;
2069 if (STALE_STATEID(stateid))
2070 goto out;
2072 /* BAD STATEID */
2073 status = nfserr_bad_stateid;
2074 if (!stateid->si_fileid) { /* delegation stateid */
2075 if(!(dp = find_delegation_stateid(ino, stateid))) {
2076 dprintk("NFSD: delegation stateid not found\n");
2077 goto out;
2079 stidp = &dp->dl_stateid;
2080 } else { /* open or lock stateid */
2081 if (!(stp = find_stateid(stateid, flags))) {
2082 dprintk("NFSD: open or lock stateid not found\n");
2083 goto out;
2085 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2086 goto out;
2087 if (!stp->st_stateowner->so_confirmed)
2088 goto out;
2089 stidp = &stp->st_stateid;
2091 if (stateid->si_generation > stidp->si_generation)
2092 goto out;
2094 /* OLD STATEID */
2095 status = nfserr_old_stateid;
2096 if (stateid->si_generation < stidp->si_generation)
2097 goto out;
2098 if (stp) {
2099 if ((status = nfs4_check_openmode(stp,flags)))
2100 goto out;
2101 renew_client(stp->st_stateowner->so_client);
2102 if (filpp)
2103 *filpp = stp->st_vfs_file;
2104 } else if (dp) {
2105 if ((status = nfs4_check_delegmode(dp, flags)))
2106 goto out;
2107 renew_client(dp->dl_client);
2108 if (flags & DELEG_RET)
2109 unhash_delegation(dp);
2110 if (filpp)
2111 *filpp = dp->dl_vfs_file;
2113 status = nfs_ok;
2114 out:
2115 return status;
2118 static inline int
2119 setlkflg (int type)
2121 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2122 RD_STATE : WR_STATE;
2126 * Checks for sequence id mutating operations.
2128 static int
2129 nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *stateid, int flags, struct nfs4_stateowner **sopp, struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
2131 struct nfs4_stateid *stp;
2132 struct nfs4_stateowner *sop;
2134 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2135 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2136 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2137 stateid->si_generation);
2139 *stpp = NULL;
2140 *sopp = NULL;
2142 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2143 printk("NFSD: preprocess_seqid_op: magic stateid!\n");
2144 return nfserr_bad_stateid;
2147 if (STALE_STATEID(stateid))
2148 return nfserr_stale_stateid;
2150 * We return BAD_STATEID if filehandle doesn't match stateid,
2151 * the confirmed flag is incorrecly set, or the generation
2152 * number is incorrect.
2154 stp = find_stateid(stateid, flags);
2155 if (stp == NULL) {
2157 * Also, we should make sure this isn't just the result of
2158 * a replayed close:
2160 sop = search_close_lru(stateid->si_stateownerid, flags);
2161 if (sop == NULL)
2162 return nfserr_bad_stateid;
2163 *sopp = sop;
2164 goto check_replay;
2167 if (lock) {
2168 struct nfs4_stateowner *sop = stp->st_stateowner;
2169 clientid_t *lockclid = &lock->v.new.clientid;
2170 struct nfs4_client *clp = sop->so_client;
2171 int lkflg = 0;
2172 int status;
2174 lkflg = setlkflg(lock->lk_type);
2176 if (lock->lk_is_new) {
2177 if (!sop->so_is_open_owner)
2178 return nfserr_bad_stateid;
2179 if (!cmp_clid(&clp->cl_clientid, lockclid))
2180 return nfserr_bad_stateid;
2181 /* stp is the open stateid */
2182 status = nfs4_check_openmode(stp, lkflg);
2183 if (status)
2184 return status;
2185 } else {
2186 /* stp is the lock stateid */
2187 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2188 if (status)
2189 return status;
2194 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp)) {
2195 printk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2196 return nfserr_bad_stateid;
2199 *stpp = stp;
2200 *sopp = sop = stp->st_stateowner;
2203 * We now validate the seqid and stateid generation numbers.
2204 * For the moment, we ignore the possibility of
2205 * generation number wraparound.
2207 if (seqid != sop->so_seqid)
2208 goto check_replay;
2210 if (sop->so_confirmed && flags & CONFIRM) {
2211 printk("NFSD: preprocess_seqid_op: expected"
2212 " unconfirmed stateowner!\n");
2213 return nfserr_bad_stateid;
2215 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2216 printk("NFSD: preprocess_seqid_op: stateowner not"
2217 " confirmed yet!\n");
2218 return nfserr_bad_stateid;
2220 if (stateid->si_generation > stp->st_stateid.si_generation) {
2221 printk("NFSD: preprocess_seqid_op: future stateid?!\n");
2222 return nfserr_bad_stateid;
2225 if (stateid->si_generation < stp->st_stateid.si_generation) {
2226 printk("NFSD: preprocess_seqid_op: old stateid!\n");
2227 return nfserr_old_stateid;
2229 renew_client(sop->so_client);
2230 return nfs_ok;
2232 check_replay:
2233 if (seqid == sop->so_seqid - 1) {
2234 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2235 /* indicate replay to calling function */
2236 return NFSERR_REPLAY_ME;
2238 printk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2239 sop->so_seqid, seqid);
2240 *sopp = NULL;
2241 return nfserr_bad_seqid;
2245 nfsd4_open_confirm(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_confirm *oc, struct nfs4_stateowner **replay_owner)
2247 int status;
2248 struct nfs4_stateowner *sop;
2249 struct nfs4_stateid *stp;
2251 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2252 (int)current_fh->fh_dentry->d_name.len,
2253 current_fh->fh_dentry->d_name.name);
2255 status = fh_verify(rqstp, current_fh, S_IFREG, 0);
2256 if (status)
2257 return status;
2259 nfs4_lock_state();
2261 if ((status = nfs4_preprocess_seqid_op(current_fh, oc->oc_seqid,
2262 &oc->oc_req_stateid,
2263 CHECK_FH | CONFIRM | OPEN_STATE,
2264 &oc->oc_stateowner, &stp, NULL)))
2265 goto out;
2267 sop = oc->oc_stateowner;
2268 sop->so_confirmed = 1;
2269 update_stateid(&stp->st_stateid);
2270 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2271 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2272 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2273 stp->st_stateid.si_boot,
2274 stp->st_stateid.si_stateownerid,
2275 stp->st_stateid.si_fileid,
2276 stp->st_stateid.si_generation);
2278 nfsd4_create_clid_dir(sop->so_client);
2279 out:
2280 if (oc->oc_stateowner) {
2281 nfs4_get_stateowner(oc->oc_stateowner);
2282 *replay_owner = oc->oc_stateowner;
2284 nfs4_unlock_state();
2285 return status;
2290 * unset all bits in union bitmap (bmap) that
2291 * do not exist in share (from successful OPEN_DOWNGRADE)
2293 static void
2294 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2296 int i;
2297 for (i = 1; i < 4; i++) {
2298 if ((i & access) != i)
2299 __clear_bit(i, bmap);
2303 static void
2304 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2306 int i;
2307 for (i = 0; i < 4; i++) {
2308 if ((i & deny) != i)
2309 __clear_bit(i, bmap);
2314 nfsd4_open_downgrade(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_downgrade *od, struct nfs4_stateowner **replay_owner)
2316 int status;
2317 struct nfs4_stateid *stp;
2318 unsigned int share_access;
2320 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2321 (int)current_fh->fh_dentry->d_name.len,
2322 current_fh->fh_dentry->d_name.name);
2324 if (!access_valid(od->od_share_access)
2325 || !deny_valid(od->od_share_deny))
2326 return nfserr_inval;
2328 nfs4_lock_state();
2329 if ((status = nfs4_preprocess_seqid_op(current_fh, od->od_seqid,
2330 &od->od_stateid,
2331 CHECK_FH | OPEN_STATE,
2332 &od->od_stateowner, &stp, NULL)))
2333 goto out;
2335 status = nfserr_inval;
2336 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2337 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2338 stp->st_access_bmap, od->od_share_access);
2339 goto out;
2341 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2342 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2343 stp->st_deny_bmap, od->od_share_deny);
2344 goto out;
2346 set_access(&share_access, stp->st_access_bmap);
2347 nfs4_file_downgrade(stp->st_vfs_file,
2348 share_access & ~od->od_share_access);
2350 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2351 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2353 update_stateid(&stp->st_stateid);
2354 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2355 status = nfs_ok;
2356 out:
2357 if (od->od_stateowner) {
2358 nfs4_get_stateowner(od->od_stateowner);
2359 *replay_owner = od->od_stateowner;
2361 nfs4_unlock_state();
2362 return status;
2366 * nfs4_unlock_state() called after encode
2369 nfsd4_close(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_close *close, struct nfs4_stateowner **replay_owner)
2371 int status;
2372 struct nfs4_stateid *stp;
2374 dprintk("NFSD: nfsd4_close on file %.*s\n",
2375 (int)current_fh->fh_dentry->d_name.len,
2376 current_fh->fh_dentry->d_name.name);
2378 nfs4_lock_state();
2379 /* check close_lru for replay */
2380 if ((status = nfs4_preprocess_seqid_op(current_fh, close->cl_seqid,
2381 &close->cl_stateid,
2382 CHECK_FH | OPEN_STATE | CLOSE_STATE,
2383 &close->cl_stateowner, &stp, NULL)))
2384 goto out;
2385 status = nfs_ok;
2386 update_stateid(&stp->st_stateid);
2387 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2389 /* release_stateid() calls nfsd_close() if needed */
2390 release_stateid(stp, OPEN_STATE);
2392 /* place unused nfs4_stateowners on so_close_lru list to be
2393 * released by the laundromat service after the lease period
2394 * to enable us to handle CLOSE replay
2396 if (list_empty(&close->cl_stateowner->so_stateids))
2397 move_to_close_lru(close->cl_stateowner);
2398 out:
2399 if (close->cl_stateowner) {
2400 nfs4_get_stateowner(close->cl_stateowner);
2401 *replay_owner = close->cl_stateowner;
2403 nfs4_unlock_state();
2404 return status;
2408 nfsd4_delegreturn(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_delegreturn *dr)
2410 int status;
2412 if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0)))
2413 goto out;
2415 nfs4_lock_state();
2416 status = nfs4_preprocess_stateid_op(current_fh, &dr->dr_stateid, DELEG_RET, NULL);
2417 nfs4_unlock_state();
2418 out:
2419 return status;
2424 * Lock owner state (byte-range locks)
2426 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2427 #define LOCK_HASH_BITS 8
2428 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2429 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2431 #define lockownerid_hashval(id) \
2432 ((id) & LOCK_HASH_MASK)
2434 static inline unsigned int
2435 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2436 struct xdr_netobj *ownername)
2438 return (file_hashval(inode) + cl_id
2439 + opaque_hashval(ownername->data, ownername->len))
2440 & LOCK_HASH_MASK;
2443 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2444 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2445 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2447 static struct nfs4_stateid *
2448 find_stateid(stateid_t *stid, int flags)
2450 struct nfs4_stateid *local = NULL;
2451 u32 st_id = stid->si_stateownerid;
2452 u32 f_id = stid->si_fileid;
2453 unsigned int hashval;
2455 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2456 if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2457 hashval = stateid_hashval(st_id, f_id);
2458 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2459 if ((local->st_stateid.si_stateownerid == st_id) &&
2460 (local->st_stateid.si_fileid == f_id))
2461 return local;
2464 if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2465 hashval = stateid_hashval(st_id, f_id);
2466 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2467 if ((local->st_stateid.si_stateownerid == st_id) &&
2468 (local->st_stateid.si_fileid == f_id))
2469 return local;
2472 return NULL;
2475 static struct nfs4_delegation *
2476 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2478 struct nfs4_file *fp;
2479 struct nfs4_delegation *dl;
2481 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2482 stid->si_boot, stid->si_stateownerid,
2483 stid->si_fileid, stid->si_generation);
2485 fp = find_file(ino);
2486 if (!fp)
2487 return NULL;
2488 dl = find_delegation_file(fp, stid);
2489 put_nfs4_file(fp);
2490 return dl;
2494 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2495 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2496 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2497 * locking, this prevents us from being completely protocol-compliant. The
2498 * real solution to this problem is to start using unsigned file offsets in
2499 * the VFS, but this is a very deep change!
2501 static inline void
2502 nfs4_transform_lock_offset(struct file_lock *lock)
2504 if (lock->fl_start < 0)
2505 lock->fl_start = OFFSET_MAX;
2506 if (lock->fl_end < 0)
2507 lock->fl_end = OFFSET_MAX;
2510 /* Hack!: For now, we're defining this just so we can use a pointer to it
2511 * as a unique cookie to identify our (NFSv4's) posix locks. */
2512 static struct lock_manager_operations nfsd_posix_mng_ops = {
2515 static inline void
2516 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2518 struct nfs4_stateowner *sop;
2519 unsigned int hval;
2521 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2522 sop = (struct nfs4_stateowner *) fl->fl_owner;
2523 hval = lockownerid_hashval(sop->so_id);
2524 kref_get(&sop->so_ref);
2525 deny->ld_sop = sop;
2526 deny->ld_clientid = sop->so_client->cl_clientid;
2527 } else {
2528 deny->ld_sop = NULL;
2529 deny->ld_clientid.cl_boot = 0;
2530 deny->ld_clientid.cl_id = 0;
2532 deny->ld_start = fl->fl_start;
2533 deny->ld_length = ~(u64)0;
2534 if (fl->fl_end != ~(u64)0)
2535 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2536 deny->ld_type = NFS4_READ_LT;
2537 if (fl->fl_type != F_RDLCK)
2538 deny->ld_type = NFS4_WRITE_LT;
2541 static struct nfs4_stateowner *
2542 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2543 struct xdr_netobj *owner)
2545 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2546 struct nfs4_stateowner *op;
2548 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2549 if (cmp_owner_str(op, owner, clid))
2550 return op;
2552 return NULL;
2556 * Alloc a lock owner structure.
2557 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2558 * occured.
2560 * strhashval = lock_ownerstr_hashval
2563 static struct nfs4_stateowner *
2564 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2565 struct nfs4_stateowner *sop;
2566 struct nfs4_replay *rp;
2567 unsigned int idhashval;
2569 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2570 return NULL;
2571 idhashval = lockownerid_hashval(current_ownerid);
2572 INIT_LIST_HEAD(&sop->so_idhash);
2573 INIT_LIST_HEAD(&sop->so_strhash);
2574 INIT_LIST_HEAD(&sop->so_perclient);
2575 INIT_LIST_HEAD(&sop->so_stateids);
2576 INIT_LIST_HEAD(&sop->so_perstateid);
2577 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2578 sop->so_time = 0;
2579 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2580 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2581 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
2582 sop->so_is_open_owner = 0;
2583 sop->so_id = current_ownerid++;
2584 sop->so_client = clp;
2585 /* It is the openowner seqid that will be incremented in encode in the
2586 * case of new lockowners; so increment the lock seqid manually: */
2587 sop->so_seqid = lock->lk_new_lock_seqid + 1;
2588 sop->so_confirmed = 1;
2589 rp = &sop->so_replay;
2590 rp->rp_status = nfserr_serverfault;
2591 rp->rp_buflen = 0;
2592 rp->rp_buf = rp->rp_ibuf;
2593 return sop;
2596 static struct nfs4_stateid *
2597 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2599 struct nfs4_stateid *stp;
2600 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2602 stp = nfs4_alloc_stateid();
2603 if (stp == NULL)
2604 goto out;
2605 INIT_LIST_HEAD(&stp->st_hash);
2606 INIT_LIST_HEAD(&stp->st_perfile);
2607 INIT_LIST_HEAD(&stp->st_perstateowner);
2608 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
2609 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2610 list_add(&stp->st_perfile, &fp->fi_stateids);
2611 list_add(&stp->st_perstateowner, &sop->so_stateids);
2612 stp->st_stateowner = sop;
2613 get_nfs4_file(fp);
2614 stp->st_file = fp;
2615 stp->st_stateid.si_boot = boot_time;
2616 stp->st_stateid.si_stateownerid = sop->so_id;
2617 stp->st_stateid.si_fileid = fp->fi_id;
2618 stp->st_stateid.si_generation = 0;
2619 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2620 stp->st_access_bmap = open_stp->st_access_bmap;
2621 stp->st_deny_bmap = open_stp->st_deny_bmap;
2622 stp->st_openstp = open_stp;
2624 out:
2625 return stp;
2628 static int
2629 check_lock_length(u64 offset, u64 length)
2631 return ((length == 0) || ((length != ~(u64)0) &&
2632 LOFF_OVERFLOW(offset, length)));
2636 * LOCK operation
2639 nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock *lock, struct nfs4_stateowner **replay_owner)
2641 struct nfs4_stateowner *open_sop = NULL;
2642 struct nfs4_stateowner *lock_sop = NULL;
2643 struct nfs4_stateid *lock_stp;
2644 struct file *filp;
2645 struct file_lock file_lock;
2646 struct file_lock conflock;
2647 int status = 0;
2648 unsigned int strhashval;
2650 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2651 (long long) lock->lk_offset,
2652 (long long) lock->lk_length);
2654 if (check_lock_length(lock->lk_offset, lock->lk_length))
2655 return nfserr_inval;
2657 if ((status = fh_verify(rqstp, current_fh, S_IFREG, MAY_LOCK))) {
2658 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2659 return status;
2662 nfs4_lock_state();
2664 if (lock->lk_is_new) {
2666 * Client indicates that this is a new lockowner.
2667 * Use open owner and open stateid to create lock owner and
2668 * lock stateid.
2670 struct nfs4_stateid *open_stp = NULL;
2671 struct nfs4_file *fp;
2673 status = nfserr_stale_clientid;
2674 if (STALE_CLIENTID(&lock->lk_new_clientid))
2675 goto out;
2677 /* validate and update open stateid and open seqid */
2678 status = nfs4_preprocess_seqid_op(current_fh,
2679 lock->lk_new_open_seqid,
2680 &lock->lk_new_open_stateid,
2681 CHECK_FH | OPEN_STATE,
2682 &lock->lk_replay_owner, &open_stp,
2683 lock);
2684 if (status)
2685 goto out;
2686 open_sop = lock->lk_replay_owner;
2687 /* create lockowner and lock stateid */
2688 fp = open_stp->st_file;
2689 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2690 open_sop->so_client->cl_clientid.cl_id,
2691 &lock->v.new.owner);
2692 /* XXX: Do we need to check for duplicate stateowners on
2693 * the same file, or should they just be allowed (and
2694 * create new stateids)? */
2695 status = nfserr_resource;
2696 lock_sop = alloc_init_lock_stateowner(strhashval,
2697 open_sop->so_client, open_stp, lock);
2698 if (lock_sop == NULL)
2699 goto out;
2700 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
2701 if (lock_stp == NULL)
2702 goto out;
2703 } else {
2704 /* lock (lock owner + lock stateid) already exists */
2705 status = nfs4_preprocess_seqid_op(current_fh,
2706 lock->lk_old_lock_seqid,
2707 &lock->lk_old_lock_stateid,
2708 CHECK_FH | LOCK_STATE,
2709 &lock->lk_replay_owner, &lock_stp, lock);
2710 if (status)
2711 goto out;
2712 lock_sop = lock->lk_replay_owner;
2714 /* lock->lk_replay_owner and lock_stp have been created or found */
2715 filp = lock_stp->st_vfs_file;
2717 status = nfserr_grace;
2718 if (nfs4_in_grace() && !lock->lk_reclaim)
2719 goto out;
2720 status = nfserr_no_grace;
2721 if (!nfs4_in_grace() && lock->lk_reclaim)
2722 goto out;
2724 locks_init_lock(&file_lock);
2725 switch (lock->lk_type) {
2726 case NFS4_READ_LT:
2727 case NFS4_READW_LT:
2728 file_lock.fl_type = F_RDLCK;
2729 break;
2730 case NFS4_WRITE_LT:
2731 case NFS4_WRITEW_LT:
2732 file_lock.fl_type = F_WRLCK;
2733 break;
2734 default:
2735 status = nfserr_inval;
2736 goto out;
2738 file_lock.fl_owner = (fl_owner_t)lock_sop;
2739 file_lock.fl_pid = current->tgid;
2740 file_lock.fl_file = filp;
2741 file_lock.fl_flags = FL_POSIX;
2742 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2744 file_lock.fl_start = lock->lk_offset;
2745 if ((lock->lk_length == ~(u64)0) ||
2746 LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2747 file_lock.fl_end = ~(u64)0;
2748 else
2749 file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2750 nfs4_transform_lock_offset(&file_lock);
2753 * Try to lock the file in the VFS.
2754 * Note: locks.c uses the BKL to protect the inode's lock list.
2757 /* XXX?: Just to divert the locks_release_private at the start of
2758 * locks_copy_lock: */
2759 conflock.fl_ops = NULL;
2760 conflock.fl_lmops = NULL;
2761 status = posix_lock_file_conf(filp, &file_lock, &conflock);
2762 dprintk("NFSD: nfsd4_lock: posix_lock_file_conf status %d\n",status);
2763 switch (-status) {
2764 case 0: /* success! */
2765 update_stateid(&lock_stp->st_stateid);
2766 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2767 sizeof(stateid_t));
2768 break;
2769 case (EAGAIN): /* conflock holds conflicting lock */
2770 status = nfserr_denied;
2771 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2772 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2773 break;
2774 case (EDEADLK):
2775 status = nfserr_deadlock;
2776 break;
2777 default:
2778 dprintk("NFSD: nfsd4_lock: posix_lock_file_conf() failed! status %d\n",status);
2779 status = nfserr_resource;
2780 break;
2782 out:
2783 if (status && lock->lk_is_new && lock_sop)
2784 release_stateowner(lock_sop);
2785 if (lock->lk_replay_owner) {
2786 nfs4_get_stateowner(lock->lk_replay_owner);
2787 *replay_owner = lock->lk_replay_owner;
2789 nfs4_unlock_state();
2790 return status;
2794 * LOCKT operation
2797 nfsd4_lockt(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lockt *lockt)
2799 struct inode *inode;
2800 struct file file;
2801 struct file_lock file_lock;
2802 struct file_lock conflock;
2803 int status;
2805 if (nfs4_in_grace())
2806 return nfserr_grace;
2808 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2809 return nfserr_inval;
2811 lockt->lt_stateowner = NULL;
2812 nfs4_lock_state();
2814 status = nfserr_stale_clientid;
2815 if (STALE_CLIENTID(&lockt->lt_clientid))
2816 goto out;
2818 if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0))) {
2819 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2820 if (status == nfserr_symlink)
2821 status = nfserr_inval;
2822 goto out;
2825 inode = current_fh->fh_dentry->d_inode;
2826 locks_init_lock(&file_lock);
2827 switch (lockt->lt_type) {
2828 case NFS4_READ_LT:
2829 case NFS4_READW_LT:
2830 file_lock.fl_type = F_RDLCK;
2831 break;
2832 case NFS4_WRITE_LT:
2833 case NFS4_WRITEW_LT:
2834 file_lock.fl_type = F_WRLCK;
2835 break;
2836 default:
2837 printk("NFSD: nfs4_lockt: bad lock type!\n");
2838 status = nfserr_inval;
2839 goto out;
2842 lockt->lt_stateowner = find_lockstateowner_str(inode,
2843 &lockt->lt_clientid, &lockt->lt_owner);
2844 if (lockt->lt_stateowner)
2845 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2846 file_lock.fl_pid = current->tgid;
2847 file_lock.fl_flags = FL_POSIX;
2848 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2850 file_lock.fl_start = lockt->lt_offset;
2851 if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2852 file_lock.fl_end = ~(u64)0;
2853 else
2854 file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2856 nfs4_transform_lock_offset(&file_lock);
2858 /* posix_test_lock uses the struct file _only_ to resolve the inode.
2859 * since LOCKT doesn't require an OPEN, and therefore a struct
2860 * file may not exist, pass posix_test_lock a struct file with
2861 * only the dentry:inode set.
2863 memset(&file, 0, sizeof (struct file));
2864 file.f_dentry = current_fh->fh_dentry;
2866 status = nfs_ok;
2867 if (posix_test_lock(&file, &file_lock, &conflock)) {
2868 status = nfserr_denied;
2869 nfs4_set_lock_denied(&conflock, &lockt->lt_denied);
2871 out:
2872 nfs4_unlock_state();
2873 return status;
2877 nfsd4_locku(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_locku *locku, struct nfs4_stateowner **replay_owner)
2879 struct nfs4_stateid *stp;
2880 struct file *filp = NULL;
2881 struct file_lock file_lock;
2882 int status;
2884 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2885 (long long) locku->lu_offset,
2886 (long long) locku->lu_length);
2888 if (check_lock_length(locku->lu_offset, locku->lu_length))
2889 return nfserr_inval;
2891 nfs4_lock_state();
2893 if ((status = nfs4_preprocess_seqid_op(current_fh,
2894 locku->lu_seqid,
2895 &locku->lu_stateid,
2896 CHECK_FH | LOCK_STATE,
2897 &locku->lu_stateowner, &stp, NULL)))
2898 goto out;
2900 filp = stp->st_vfs_file;
2901 BUG_ON(!filp);
2902 locks_init_lock(&file_lock);
2903 file_lock.fl_type = F_UNLCK;
2904 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2905 file_lock.fl_pid = current->tgid;
2906 file_lock.fl_file = filp;
2907 file_lock.fl_flags = FL_POSIX;
2908 file_lock.fl_lmops = &nfsd_posix_mng_ops;
2909 file_lock.fl_start = locku->lu_offset;
2911 if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2912 file_lock.fl_end = ~(u64)0;
2913 else
2914 file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2915 nfs4_transform_lock_offset(&file_lock);
2918 * Try to unlock the file in the VFS.
2920 status = posix_lock_file(filp, &file_lock);
2921 if (status) {
2922 dprintk("NFSD: nfs4_locku: posix_lock_file failed!\n");
2923 goto out_nfserr;
2926 * OK, unlock succeeded; the only thing left to do is update the stateid.
2928 update_stateid(&stp->st_stateid);
2929 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2931 out:
2932 if (locku->lu_stateowner) {
2933 nfs4_get_stateowner(locku->lu_stateowner);
2934 *replay_owner = locku->lu_stateowner;
2936 nfs4_unlock_state();
2937 return status;
2939 out_nfserr:
2940 status = nfserrno(status);
2941 goto out;
2945 * returns
2946 * 1: locks held by lockowner
2947 * 0: no locks held by lockowner
2949 static int
2950 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2952 struct file_lock **flpp;
2953 struct inode *inode = filp->f_dentry->d_inode;
2954 int status = 0;
2956 lock_kernel();
2957 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
2958 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
2959 status = 1;
2960 goto out;
2963 out:
2964 unlock_kernel();
2965 return status;
2969 nfsd4_release_lockowner(struct svc_rqst *rqstp, struct nfsd4_release_lockowner *rlockowner)
2971 clientid_t *clid = &rlockowner->rl_clientid;
2972 struct nfs4_stateowner *sop;
2973 struct nfs4_stateid *stp;
2974 struct xdr_netobj *owner = &rlockowner->rl_owner;
2975 struct list_head matches;
2976 int i;
2977 int status;
2979 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
2980 clid->cl_boot, clid->cl_id);
2982 /* XXX check for lease expiration */
2984 status = nfserr_stale_clientid;
2985 if (STALE_CLIENTID(clid))
2986 return status;
2988 nfs4_lock_state();
2990 status = nfserr_locks_held;
2991 /* XXX: we're doing a linear search through all the lockowners.
2992 * Yipes! For now we'll just hope clients aren't really using
2993 * release_lockowner much, but eventually we have to fix these
2994 * data structures. */
2995 INIT_LIST_HEAD(&matches);
2996 for (i = 0; i < LOCK_HASH_SIZE; i++) {
2997 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
2998 if (!cmp_owner_str(sop, owner, clid))
2999 continue;
3000 list_for_each_entry(stp, &sop->so_stateids,
3001 st_perstateowner) {
3002 if (check_for_locks(stp->st_vfs_file, sop))
3003 goto out;
3004 /* Note: so_perclient unused for lockowners,
3005 * so it's OK to fool with here. */
3006 list_add(&sop->so_perclient, &matches);
3010 /* Clients probably won't expect us to return with some (but not all)
3011 * of the lockowner state released; so don't release any until all
3012 * have been checked. */
3013 status = nfs_ok;
3014 while (!list_empty(&matches)) {
3015 sop = list_entry(matches.next, struct nfs4_stateowner,
3016 so_perclient);
3017 /* unhash_stateowner deletes so_perclient only
3018 * for openowners. */
3019 list_del(&sop->so_perclient);
3020 release_stateowner(sop);
3022 out:
3023 nfs4_unlock_state();
3024 return status;
3027 static inline struct nfs4_client_reclaim *
3028 alloc_reclaim(void)
3030 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3034 nfs4_has_reclaimed_state(const char *name)
3036 unsigned int strhashval = clientstr_hashval(name);
3037 struct nfs4_client *clp;
3039 clp = find_confirmed_client_by_str(name, strhashval);
3040 return clp ? 1 : 0;
3044 * failure => all reset bets are off, nfserr_no_grace...
3047 nfs4_client_to_reclaim(const char *name)
3049 unsigned int strhashval;
3050 struct nfs4_client_reclaim *crp = NULL;
3052 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3053 crp = alloc_reclaim();
3054 if (!crp)
3055 return 0;
3056 strhashval = clientstr_hashval(name);
3057 INIT_LIST_HEAD(&crp->cr_strhash);
3058 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3059 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
3060 reclaim_str_hashtbl_size++;
3061 return 1;
3064 static void
3065 nfs4_release_reclaim(void)
3067 struct nfs4_client_reclaim *crp = NULL;
3068 int i;
3070 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3071 while (!list_empty(&reclaim_str_hashtbl[i])) {
3072 crp = list_entry(reclaim_str_hashtbl[i].next,
3073 struct nfs4_client_reclaim, cr_strhash);
3074 list_del(&crp->cr_strhash);
3075 kfree(crp);
3076 reclaim_str_hashtbl_size--;
3079 BUG_ON(reclaim_str_hashtbl_size);
3083 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3084 static struct nfs4_client_reclaim *
3085 nfs4_find_reclaim_client(clientid_t *clid)
3087 unsigned int strhashval;
3088 struct nfs4_client *clp;
3089 struct nfs4_client_reclaim *crp = NULL;
3092 /* find clientid in conf_id_hashtbl */
3093 clp = find_confirmed_client(clid);
3094 if (clp == NULL)
3095 return NULL;
3097 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3098 clp->cl_name.len, clp->cl_name.data,
3099 clp->cl_recdir);
3101 /* find clp->cl_name in reclaim_str_hashtbl */
3102 strhashval = clientstr_hashval(clp->cl_recdir);
3103 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3104 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
3105 return crp;
3108 return NULL;
3112 * Called from OPEN. Look for clientid in reclaim list.
3115 nfs4_check_open_reclaim(clientid_t *clid)
3117 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
3120 /* initialization to perform at module load time: */
3122 void
3123 nfs4_state_init(void)
3125 int i;
3127 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3128 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3129 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3130 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3131 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3133 for (i = 0; i < FILE_HASH_SIZE; i++) {
3134 INIT_LIST_HEAD(&file_hashtbl[i]);
3136 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3137 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3138 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3140 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3141 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3142 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3144 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3145 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3146 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3148 memset(&onestateid, ~0, sizeof(stateid_t));
3149 INIT_LIST_HEAD(&close_lru);
3150 INIT_LIST_HEAD(&client_lru);
3151 INIT_LIST_HEAD(&del_recall_lru);
3152 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3153 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3154 reclaim_str_hashtbl_size = 0;
3157 static void
3158 nfsd4_load_reboot_recovery_data(void)
3160 int status;
3162 nfs4_lock_state();
3163 nfsd4_init_recdir(user_recovery_dirname);
3164 status = nfsd4_recdir_load();
3165 nfs4_unlock_state();
3166 if (status)
3167 printk("NFSD: Failure reading reboot recovery data\n");
3170 /* initialization to perform when the nfsd service is started: */
3172 static void
3173 __nfs4_state_start(void)
3175 time_t grace_time;
3177 boot_time = get_seconds();
3178 grace_time = max(user_lease_time, lease_time);
3179 lease_time = user_lease_time;
3180 in_grace = 1;
3181 printk("NFSD: starting %ld-second grace period\n", grace_time);
3182 laundry_wq = create_singlethread_workqueue("nfsd4");
3183 queue_delayed_work(laundry_wq, &laundromat_work, grace_time*HZ);
3187 nfs4_state_start(void)
3189 int status;
3191 if (nfs4_init)
3192 return 0;
3193 status = nfsd4_init_slabs();
3194 if (status)
3195 return status;
3196 nfsd4_load_reboot_recovery_data();
3197 __nfs4_state_start();
3198 nfs4_init = 1;
3199 return 0;
3203 nfs4_in_grace(void)
3205 return in_grace;
3208 time_t
3209 nfs4_lease_time(void)
3211 return lease_time;
3214 static void
3215 __nfs4_state_shutdown(void)
3217 int i;
3218 struct nfs4_client *clp = NULL;
3219 struct nfs4_delegation *dp = NULL;
3220 struct list_head *pos, *next, reaplist;
3222 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3223 while (!list_empty(&conf_id_hashtbl[i])) {
3224 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3225 expire_client(clp);
3227 while (!list_empty(&unconf_str_hashtbl[i])) {
3228 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3229 expire_client(clp);
3232 INIT_LIST_HEAD(&reaplist);
3233 spin_lock(&recall_lock);
3234 list_for_each_safe(pos, next, &del_recall_lru) {
3235 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3236 list_move(&dp->dl_recall_lru, &reaplist);
3238 spin_unlock(&recall_lock);
3239 list_for_each_safe(pos, next, &reaplist) {
3240 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3241 list_del_init(&dp->dl_recall_lru);
3242 unhash_delegation(dp);
3245 cancel_delayed_work(&laundromat_work);
3246 nfsd4_shutdown_recdir();
3247 nfs4_init = 0;
3250 void
3251 nfs4_state_shutdown(void)
3253 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3254 destroy_workqueue(laundry_wq);
3255 nfs4_lock_state();
3256 nfs4_release_reclaim();
3257 __nfs4_state_shutdown();
3258 nfsd4_free_slabs();
3259 nfs4_unlock_state();
3262 static void
3263 nfs4_set_recdir(char *recdir)
3265 nfs4_lock_state();
3266 strcpy(user_recovery_dirname, recdir);
3267 nfs4_unlock_state();
3271 * Change the NFSv4 recovery directory to recdir.
3274 nfs4_reset_recoverydir(char *recdir)
3276 int status;
3277 struct nameidata nd;
3279 status = path_lookup(recdir, LOOKUP_FOLLOW, &nd);
3280 if (status)
3281 return status;
3282 status = -ENOTDIR;
3283 if (S_ISDIR(nd.dentry->d_inode->i_mode)) {
3284 nfs4_set_recdir(recdir);
3285 status = 0;
3287 path_release(&nd);
3288 return status;
3292 * Called when leasetime is changed.
3294 * The only way the protocol gives us to handle on-the-fly lease changes is to
3295 * simulate a reboot. Instead of doing that, we just wait till the next time
3296 * we start to register any changes in lease time. If the administrator
3297 * really wants to change the lease time *now*, they can go ahead and bring
3298 * nfsd down and then back up again after changing the lease time.
3300 void
3301 nfs4_reset_lease(time_t leasetime)
3303 lock_kernel();
3304 user_lease_time = leasetime;
3305 unlock_kernel();