projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd41: non-page DRC for solo sequence responses
[linux-2.6-block.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT		 1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
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.
22*
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.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT		 45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
1da177e4
LT		 58
59#define NFSDDBG_FACILITY NFSDDBG_PROC
60
61/* Globals */
62static time_t lease_time = 90; /* default lease time */
d99a05ad 63static time_t user_lease_time = 90;
fd39ca9a 64static time_t boot_time;
1da177e4
LT		 65static u32 current_ownerid = 1;
66static u32 current_fileid = 1;
67static u32 current_delegid = 1;
68static u32 nfs4_init;
fd39ca9a
N		 69static stateid_t zerostateid; /* bits all 0 */
70static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 71static u64 current_sessionid = 1;
fd39ca9a
N		 72
73#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
74#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 75
1da177e4 76/* forward declarations */
fd39ca9a 77static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 78static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 79static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
80static void nfs4_set_recdir(char *recdir);
1da177e4 81
8b671b80
BF		 82/* Locking: */
83
84/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 85static DEFINE_MUTEX(client_mutex);
1da177e4 86
8b671b80
BF		 87/*
88 * Currently used for the del_recall_lru and file hash table. In an
89 * effort to decrease the scope of the client_mutex, this spinlock may
90 * eventually cover more:
91 */
92static DEFINE_SPINLOCK(recall_lock);
93
e18b890b
CL		 94static struct kmem_cache *stateowner_slab = NULL;
95static struct kmem_cache *file_slab = NULL;
96static struct kmem_cache *stateid_slab = NULL;
97static struct kmem_cache *deleg_slab = NULL;
e60d4398 98
1da177e4
LT		 99void
100nfs4_lock_state(void)
101{
353ab6e9 102 mutex_lock(&client_mutex);
1da177e4
LT		 103}
104
105void
106nfs4_unlock_state(void)
107{
353ab6e9 108 mutex_unlock(&client_mutex);
1da177e4
LT		 109}
110
111static inline u32
112opaque_hashval(const void *ptr, int nbytes)
113{
114 unsigned char *cptr = (unsigned char *) ptr;
115
116 u32 x = 0;
117 while (nbytes--) {
118 x *= 37;
119 x += *cptr++;
120 }
121 return x;
122}
123
1da177e4
LT		 124static struct list_head del_recall_lru;
125
13cd2184
N		 126static inline void
127put_nfs4_file(struct nfs4_file *fi)
128{
8b671b80
BF		 129 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
130 list_del(&fi->fi_hash);
131 spin_unlock(&recall_lock);
132 iput(fi->fi_inode);
133 kmem_cache_free(file_slab, fi);
134 }
13cd2184
N		 135}
136
137static inline void
138get_nfs4_file(struct nfs4_file *fi)
139{
8b671b80 140 atomic_inc(&fi->fi_ref);
13cd2184
N		 141}
142
ef0f3390 143static int num_delegations;
c2f1a551 144unsigned int max_delegations;
ef0f3390
N		 145
146/*
147 * Open owner state (share locks)
148 */
149
150/* hash tables for nfs4_stateowner */
151#define OWNER_HASH_BITS 8
152#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
153#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
154
155#define ownerid_hashval(id) \
156 ((id) & OWNER_HASH_MASK)
157#define ownerstr_hashval(clientid, ownername) \
158 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
159
160static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
161static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
162
163/* hash table for nfs4_file */
164#define FILE_HASH_BITS 8
165#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
166#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
167/* hash table for (open)nfs4_stateid */
168#define STATEID_HASH_BITS 10
169#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
170#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
171
172#define file_hashval(x) \
173 hash_ptr(x, FILE_HASH_BITS)
174#define stateid_hashval(owner_id, file_id) \
175 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
176
177static struct list_head file_hashtbl[FILE_HASH_SIZE];
178static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
179
1da177e4
LT		 180static struct nfs4_delegation *
181alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
182{
183 struct nfs4_delegation *dp;
184 struct nfs4_file *fp = stp->st_file;
185 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
186
187 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 188 if (fp->fi_had_conflict)
189 return NULL;
c2f1a551 190 if (num_delegations > max_delegations)
ef0f3390 191 return NULL;
5b2d21c1
N		 192 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
193 if (dp == NULL)
1da177e4 194 return dp;
ef0f3390 195 num_delegations++;
ea1da636
N		 196 INIT_LIST_HEAD(&dp->dl_perfile);
197 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 198 INIT_LIST_HEAD(&dp->dl_recall_lru);
199 dp->dl_client = clp;
13cd2184 200 get_nfs4_file(fp);
1da177e4
LT		 201 dp->dl_file = fp;
202 dp->dl_flock = NULL;
203 get_file(stp->st_vfs_file);
204 dp->dl_vfs_file = stp->st_vfs_file;
205 dp->dl_type = type;
206 dp->dl_recall.cbr_dp = NULL;
207 dp->dl_recall.cbr_ident = cb->cb_ident;
208 dp->dl_recall.cbr_trunc = 0;
209 dp->dl_stateid.si_boot = boot_time;
210 dp->dl_stateid.si_stateownerid = current_delegid++;
211 dp->dl_stateid.si_fileid = 0;
212 dp->dl_stateid.si_generation = 0;
6c02eaa1 213 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 214 dp->dl_time = 0;
215 atomic_set(&dp->dl_count, 1);
ea1da636
N		 216 list_add(&dp->dl_perfile, &fp->fi_delegations);
217 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 218 return dp;
219}
220
221void
222nfs4_put_delegation(struct nfs4_delegation *dp)
223{
224 if (atomic_dec_and_test(&dp->dl_count)) {
225 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 226 put_nfs4_file(dp->dl_file);
5b2d21c1 227 kmem_cache_free(deleg_slab, dp);
ef0f3390 228 num_delegations--;
1da177e4
LT		 229 }
230}
231
232/* Remove the associated file_lock first, then remove the delegation.
233 * lease_modify() is called to remove the FS_LEASE file_lock from
234 * the i_flock list, eventually calling nfsd's lock_manager
235 * fl_release_callback.
236 */
237static void
238nfs4_close_delegation(struct nfs4_delegation *dp)
239{
240 struct file *filp = dp->dl_vfs_file;
241
242 dprintk("NFSD: close_delegation dp %p\n",dp);
243 dp->dl_vfs_file = NULL;
244 /* The following nfsd_close may not actually close the file,
245 * but we want to remove the lease in any case. */
c907132d 246 if (dp->dl_flock)
a9933cea 247 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 248 nfsd_close(filp);
1da177e4
LT		 249}
250
251/* Called under the state lock. */
252static void
253unhash_delegation(struct nfs4_delegation *dp)
254{
ea1da636
N		 255 list_del_init(&dp->dl_perfile);
256 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 257 spin_lock(&recall_lock);
258 list_del_init(&dp->dl_recall_lru);
259 spin_unlock(&recall_lock);
260 nfs4_close_delegation(dp);
261 nfs4_put_delegation(dp);
262}
263
264/*
265 * SETCLIENTID state
266 */
267
268/* Hash tables for nfs4_clientid state */
269#define CLIENT_HASH_BITS 4
270#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
271#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
272
273#define clientid_hashval(id) \
274 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 275#define clientstr_hashval(name) \
276 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 277/*
278 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
279 * used in reboot/reset lease grace period processing
280 *
281 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
282 * setclientid_confirmed info.
283 *
284 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
285 * setclientid info.
286 *
287 * client_lru holds client queue ordered by nfs4_client.cl_time
288 * for lease renewal.
289 *
290 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
291 * for last close replay.
292 */
293static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
294static int reclaim_str_hashtbl_size = 0;
295static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
298static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
299static struct list_head client_lru;
300static struct list_head close_lru;
301
2283963f
BF		 302static void unhash_generic_stateid(struct nfs4_stateid *stp)
303{
304 list_del(&stp->st_hash);
305 list_del(&stp->st_perfile);
306 list_del(&stp->st_perstateowner);
307}
308
309static void free_generic_stateid(struct nfs4_stateid *stp)
310{
311 put_nfs4_file(stp->st_file);
312 kmem_cache_free(stateid_slab, stp);
313}
314
315static void release_lock_stateid(struct nfs4_stateid *stp)
316{
317 unhash_generic_stateid(stp);
318 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
319 free_generic_stateid(stp);
320}
321
f044ff83
BF		 322static void unhash_lockowner(struct nfs4_stateowner *sop)
323{
324 struct nfs4_stateid *stp;
325
326 list_del(&sop->so_idhash);
327 list_del(&sop->so_strhash);
328 list_del(&sop->so_perstateid);
329 while (!list_empty(&sop->so_stateids)) {
330 stp = list_first_entry(&sop->so_stateids,
331 struct nfs4_stateid, st_perstateowner);
332 release_lock_stateid(stp);
333 }
334}
335
336static void release_lockowner(struct nfs4_stateowner *sop)
337{
338 unhash_lockowner(sop);
339 nfs4_put_stateowner(sop);
340}
341
f1d110ca
BF		 342static void
343release_stateid_lockowners(struct nfs4_stateid *open_stp)
344{
345 struct nfs4_stateowner *lock_sop;
346
347 while (!list_empty(&open_stp->st_lockowners)) {
348 lock_sop = list_entry(open_stp->st_lockowners.next,
349 struct nfs4_stateowner, so_perstateid);
350 /* list_del(&open_stp->st_lockowners); */
351 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 352 release_lockowner(lock_sop);
f1d110ca
BF		 353 }
354}
355
2283963f
BF		 356static void release_open_stateid(struct nfs4_stateid *stp)
357{
358 unhash_generic_stateid(stp);
359 release_stateid_lockowners(stp);
360 nfsd_close(stp->st_vfs_file);
361 free_generic_stateid(stp);
362}
363
f044ff83 364static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 365{
366 struct nfs4_stateid *stp;
367
368 list_del(&sop->so_idhash);
369 list_del(&sop->so_strhash);
f044ff83
BF		 370 list_del(&sop->so_perclient);
371 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 372 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 373 stp = list_first_entry(&sop->so_stateids,
374 struct nfs4_stateid, st_perstateowner);
375 release_open_stateid(stp);
f1d110ca
BF		 376 }
377}
378
f044ff83 379static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 380{
f044ff83 381 unhash_openowner(sop);
f1d110ca
BF		 382 list_del(&sop->so_close_lru);
383 nfs4_put_stateowner(sop);
384}
385
5282fd72
ME		 386static DEFINE_SPINLOCK(sessionid_lock);
387#define SESSION_HASH_SIZE 512
388static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
389
390static inline int
391hash_sessionid(struct nfs4_sessionid *sessionid)
392{
393 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
394
395 return sid->sequence % SESSION_HASH_SIZE;
396}
397
398static inline void
399dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
400{
401 u32 *ptr = (u32 *)(&sessionid->data[0]);
402 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
403}
404
ec6b5d7b
AA		 405static void
406gen_sessionid(struct nfsd4_session *ses)
407{
408 struct nfs4_client *clp = ses->se_client;
409 struct nfsd4_sessionid *sid;
410
411 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
412 sid->clientid = clp->cl_clientid;
413 sid->sequence = current_sessionid++;
414 sid->reserved = 0;
415}
416
417/*
418 * Give the client the number of slots it requests bound by
419 * NFSD_MAX_SLOTS_PER_SESSION and by sv_drc_max_pages.
420 *
421 * If we run out of pages (sv_drc_pages_used == sv_drc_max_pages) we
422 * should (up to a point) re-negotiate active sessions and reduce their
423 * slot usage to make rooom for new connections. For now we just fail the
424 * create session.
425 */
426static int set_forechannel_maxreqs(struct nfsd4_channel_attrs *fchan)
427{
428 int status = 0, np = fchan->maxreqs * NFSD_PAGES_PER_SLOT;
429
430 spin_lock(&nfsd_serv->sv_lock);
431 if (np + nfsd_serv->sv_drc_pages_used > nfsd_serv->sv_drc_max_pages)
432 np = nfsd_serv->sv_drc_max_pages - nfsd_serv->sv_drc_pages_used;
433 nfsd_serv->sv_drc_pages_used += np;
434 spin_unlock(&nfsd_serv->sv_lock);
435
436 if (np <= 0) {
437 status = nfserr_resource;
438 fchan->maxreqs = 0;
439 } else
440 fchan->maxreqs = np / NFSD_PAGES_PER_SLOT;
441
442 return status;
443}
444
445/*
446 * fchan holds the client values on input, and the server values on output
447 */
448static int init_forechannel_attrs(struct svc_rqst *rqstp,
449 struct nfsd4_session *session,
450 struct nfsd4_channel_attrs *fchan)
451{
452 int status = 0;
453 __u32 maxcount = svc_max_payload(rqstp);
454
455 /* headerpadsz set to zero in encode routine */
456
457 /* Use the client's max request and max response size if possible */
458 if (fchan->maxreq_sz > maxcount)
459 fchan->maxreq_sz = maxcount;
460 session->se_fmaxreq_sz = fchan->maxreq_sz;
461
462 if (fchan->maxresp_sz > maxcount)
463 fchan->maxresp_sz = maxcount;
464 session->se_fmaxresp_sz = fchan->maxresp_sz;
465
466 /* Set the max response cached size our default which is
467 * a multiple of PAGE_SIZE and small */
468 session->se_fmaxresp_cached = NFSD_PAGES_PER_SLOT * PAGE_SIZE;
469 fchan->maxresp_cached = session->se_fmaxresp_cached;
470
471 /* Use the client's maxops if possible */
472 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
473 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
474 session->se_fmaxops = fchan->maxops;
475
476 /* try to use the client requested number of slots */
477 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
478 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
479
480 /* FIXME: Error means no more DRC pages so the server should
481 * recover pages from existing sessions. For now fail session
482 * creation.
483 */
484 status = set_forechannel_maxreqs(fchan);
485
486 session->se_fnumslots = fchan->maxreqs;
487 return status;
488}
489
490static int
491alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
492 struct nfsd4_create_session *cses)
493{
494 struct nfsd4_session *new, tmp;
495 int idx, status = nfserr_resource, slotsize;
496
497 memset(&tmp, 0, sizeof(tmp));
498
499 /* FIXME: For now, we just accept the client back channel attributes. */
500 status = init_forechannel_attrs(rqstp, &tmp, &cses->fore_channel);
501 if (status)
502 goto out;
503
504 /* allocate struct nfsd4_session and slot table in one piece */
505 slotsize = tmp.se_fnumslots * sizeof(struct nfsd4_slot);
506 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
507 if (!new)
508 goto out;
509
510 memcpy(new, &tmp, sizeof(*new));
511
512 new->se_client = clp;
513 gen_sessionid(new);
514 idx = hash_sessionid(&new->se_sessionid);
515 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
516 NFS4_MAX_SESSIONID_LEN);
517
518 new->se_flags = cses->flags;
519 kref_init(&new->se_ref);
520 spin_lock(&sessionid_lock);
521 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
522 list_add(&new->se_perclnt, &clp->cl_sessions);
523 spin_unlock(&sessionid_lock);
524
525 status = nfs_ok;
526out:
527 return status;
528}
529
5282fd72
ME		 530/* caller must hold sessionid_lock */
531static struct nfsd4_session *
532find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
533{
534 struct nfsd4_session *elem;
535 int idx;
536
537 dump_sessionid(__func__, sessionid);
538 idx = hash_sessionid(sessionid);
539 dprintk("%s: idx is %d\n", __func__, idx);
540 /* Search in the appropriate list */
541 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
542 dump_sessionid("list traversal", &elem->se_sessionid);
543 if (!memcmp(elem->se_sessionid.data, sessionid->data,
544 NFS4_MAX_SESSIONID_LEN)) {
545 return elem;
546 }
547 }
548
549 dprintk("%s: session not found\n", __func__);
550 return NULL;
551}
552
553/* caller must hold sessionid_lock */
7116ed6b 554static void
5282fd72 555unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 556{
557 list_del(&ses->se_hash);
558 list_del(&ses->se_perclnt);
5282fd72
ME		 559}
560
561static void
562release_session(struct nfsd4_session *ses)
563{
564 spin_lock(&sessionid_lock);
565 unhash_session(ses);
566 spin_unlock(&sessionid_lock);
7116ed6b
AA		 567 nfsd4_put_session(ses);
568}
569
14778a13
AA		 570static void nfsd4_release_respages(struct page **respages, short resused);
571
7116ed6b
AA		 572void
573free_session(struct kref *kref)
574{
575 struct nfsd4_session *ses;
14778a13 576 int i;
7116ed6b
AA		 577
578 ses = container_of(kref, struct nfsd4_session, se_ref);
14778a13
AA		 579 for (i = 0; i < ses->se_fnumslots; i++) {
580 struct nfsd4_cache_entry *e = &ses->se_slots[i].sl_cache_entry;
581 nfsd4_release_respages(e->ce_respages, e->ce_resused);
582 }
7116ed6b
AA		 583 kfree(ses->se_slots);
584 kfree(ses);
585}
586
1da177e4
LT		 587static inline void
588renew_client(struct nfs4_client *clp)
589{
590 /*
591 * Move client to the end to the LRU list.
592 */
593 dprintk("renewing client (clientid %08x/%08x)\n",
594 clp->cl_clientid.cl_boot,
595 clp->cl_clientid.cl_id);
596 list_move_tail(&clp->cl_lru, &client_lru);
597 clp->cl_time = get_seconds();
598}
599
600/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
601static int
602STALE_CLIENTID(clientid_t *clid)
603{
604 if (clid->cl_boot == boot_time)
605 return 0;
606 dprintk("NFSD stale clientid (%08x/%08x)\n",
607 clid->cl_boot, clid->cl_id);
608 return 1;
609}
610
611/*
612 * XXX Should we use a slab cache ?
613 * This type of memory management is somewhat inefficient, but we use it
614 * anyway since SETCLIENTID is not a common operation.
615 */
35bba9a3 616static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 617{
618 struct nfs4_client *clp;
619
35bba9a3
BF		 620 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
621 if (clp == NULL)
622 return NULL;
623 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
624 if (clp->cl_name.data == NULL) {
625 kfree(clp);
626 return NULL;
1da177e4 627 }
35bba9a3
BF		 628 memcpy(clp->cl_name.data, name.data, name.len);
629 clp->cl_name.len = name.len;
1da177e4
LT		 630 return clp;
631}
632
1b1a9b31
BF		 633static void
634shutdown_callback_client(struct nfs4_client *clp)
635{
636 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
637
1b1a9b31 638 if (clnt) {
404ec117
BF		 639 /*
640 * Callback threads take a reference on the client, so there
641 * should be no outstanding callbacks at this point.
642 */
1b1a9b31
BF		 643 clp->cl_callback.cb_client = NULL;
644 rpc_shutdown_client(clnt);
645 }
646}
647
1da177e4
LT		 648static inline void
649free_client(struct nfs4_client *clp)
650{
1b1a9b31 651 shutdown_callback_client(clp);
38eb76a5
AA		 652 nfsd4_release_respages(clp->cl_slot.sl_cache_entry.ce_respages,
653 clp->cl_slot.sl_cache_entry.ce_resused);
1da177e4
LT		 654 if (clp->cl_cred.cr_group_info)
655 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 656 kfree(clp->cl_principal);
1da177e4
LT		 657 kfree(clp->cl_name.data);
658 kfree(clp);
659}
660
661void
662put_nfs4_client(struct nfs4_client *clp)
663{
664 if (atomic_dec_and_test(&clp->cl_count))
665 free_client(clp);
666}
667
668static void
669expire_client(struct nfs4_client *clp)
670{
671 struct nfs4_stateowner *sop;
672 struct nfs4_delegation *dp;
1da177e4
LT		 673 struct list_head reaplist;
674
675 dprintk("NFSD: expire_client cl_count %d\n",
676 atomic_read(&clp->cl_count));
677
1da177e4
LT		 678 INIT_LIST_HEAD(&reaplist);
679 spin_lock(&recall_lock);
ea1da636
N		 680 while (!list_empty(&clp->cl_delegations)) {
681 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 682 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
683 dp->dl_flock);
ea1da636 684 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 685 list_move(&dp->dl_recall_lru, &reaplist);
686 }
687 spin_unlock(&recall_lock);
688 while (!list_empty(&reaplist)) {
689 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
690 list_del_init(&dp->dl_recall_lru);
691 unhash_delegation(dp);
692 }
693 list_del(&clp->cl_idhash);
694 list_del(&clp->cl_strhash);
695 list_del(&clp->cl_lru);
ea1da636
N		 696 while (!list_empty(&clp->cl_openowners)) {
697 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 698 release_openowner(sop);
1da177e4 699 }
c4bf7868
ME		 700 while (!list_empty(&clp->cl_sessions)) {
701 struct nfsd4_session *ses;
702 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
703 se_perclnt);
704 release_session(ses);
705 }
1da177e4
LT		 706 put_nfs4_client(clp);
707}
708
35bba9a3
BF		 709static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
710{
1da177e4
LT		 711 struct nfs4_client *clp;
712
35bba9a3
BF		 713 clp = alloc_client(name);
714 if (clp == NULL)
715 return NULL;
a55370a3 716 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4
LT		 717 atomic_set(&clp->cl_count, 1);
718 atomic_set(&clp->cl_callback.cb_set, 0);
1da177e4
LT		 719 INIT_LIST_HEAD(&clp->cl_idhash);
720 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 721 INIT_LIST_HEAD(&clp->cl_openowners);
722 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 723 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 724 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 725 return clp;
726}
727
35bba9a3
BF		 728static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
729{
730 memcpy(target->cl_verifier.data, source->data,
731 sizeof(target->cl_verifier.data));
1da177e4
LT		 732}
733
35bba9a3
BF		 734static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
735{
1da177e4
LT		 736 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
737 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
738}
739
35bba9a3
BF		 740static void copy_cred(struct svc_cred *target, struct svc_cred *source)
741{
1da177e4
LT		 742 target->cr_uid = source->cr_uid;
743 target->cr_gid = source->cr_gid;
744 target->cr_group_info = source->cr_group_info;
745 get_group_info(target->cr_group_info);
746}
747
35bba9a3 748static int same_name(const char *n1, const char *n2)
599e0a22 749{
a55370a3 750 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 751}
752
753static int
599e0a22
BF		 754same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
755{
756 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 757}
758
759static int
599e0a22
BF		 760same_clid(clientid_t *cl1, clientid_t *cl2)
761{
762 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 763}
764
765/* XXX what about NGROUP */
766static int
599e0a22
BF		 767same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
768{
769 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 770}
771
5ec7b46c
BF		 772static void gen_clid(struct nfs4_client *clp)
773{
774 static u32 current_clientid = 1;
775
1da177e4
LT		 776 clp->cl_clientid.cl_boot = boot_time;
777 clp->cl_clientid.cl_id = current_clientid++;
778}
779
deda2faa
BF		 780static void gen_confirm(struct nfs4_client *clp)
781{
782 static u32 i;
783 u32 *p;
1da177e4 784
1da177e4 785 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 786 *p++ = get_seconds();
787 *p++ = i++;
1da177e4
LT		 788}
789
35bba9a3
BF		 790static int check_name(struct xdr_netobj name)
791{
1da177e4
LT		 792 if (name.len == 0)
793 return 0;
794 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 795 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 796 return 0;
797 }
798 return 1;
799}
800
fd39ca9a 801static void
1da177e4
LT		 802add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
803{
804 unsigned int idhashval;
805
806 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
807 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
808 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
809 list_add_tail(&clp->cl_lru, &client_lru);
810 clp->cl_time = get_seconds();
811}
812
fd39ca9a 813static void
1da177e4
LT		 814move_to_confirmed(struct nfs4_client *clp)
815{
816 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
817 unsigned int strhashval;
818
819 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
820 list_del_init(&clp->cl_strhash);
f116629d 821 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 822 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 823 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
824 renew_client(clp);
825}
826
827static struct nfs4_client *
828find_confirmed_client(clientid_t *clid)
829{
830 struct nfs4_client *clp;
831 unsigned int idhashval = clientid_hashval(clid->cl_id);
832
833 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 834 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 835 return clp;
836 }
837 return NULL;
838}
839
840static struct nfs4_client *
841find_unconfirmed_client(clientid_t *clid)
842{
843 struct nfs4_client *clp;
844 unsigned int idhashval = clientid_hashval(clid->cl_id);
845
846 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 847 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 848 return clp;
849 }
850 return NULL;
851}
852
a1bcecd2
AA		 853/*
854 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
855 * parameter. Matching is based on the fact the at least one of the
856 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
857 *
858 * FIXME: we need to unify the clientid namespaces for nfsv4.x
859 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
860 * and SET_CLIENTID{,_CONFIRM}
861 */
862static inline int
863match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
864{
865 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
866 return use_exchange_id == has_exchange_flags;
867}
868
28ce6054 869static struct nfs4_client *
a1bcecd2
AA		 870find_confirmed_client_by_str(const char *dname, unsigned int hashval,
871 bool use_exchange_id)
28ce6054
N		 872{
873 struct nfs4_client *clp;
874
875 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 876 if (same_name(clp->cl_recdir, dname) &&
877 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 878 return clp;
879 }
880 return NULL;
881}
882
883static struct nfs4_client *
a1bcecd2
AA		 884find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
885 bool use_exchange_id)
28ce6054
N		 886{
887 struct nfs4_client *clp;
888
889 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 890 if (same_name(clp->cl_recdir, dname) &&
891 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 892 return clp;
893 }
894 return NULL;
895}
896
1da177e4
LT		 897/* a helper function for parse_callback */
898static int
899parse_octet(unsigned int *lenp, char **addrp)
900{
901 unsigned int len = *lenp;
902 char *p = *addrp;
903 int n = -1;
904 char c;
905
906 for (;;) {
907 if (!len)
908 break;
909 len--;
910 c = *p++;
911 if (c == '.')
912 break;
913 if ((c < '0') || (c > '9')) {
914 n = -1;
915 break;
916 }
917 if (n < 0)
918 n = 0;
919 n = (n * 10) + (c - '0');
920 if (n > 255) {
921 n = -1;
922 break;
923 }
924 }
925 *lenp = len;
926 *addrp = p;
927 return n;
928}
929
930/* parse and set the setclientid ipv4 callback address */
fd39ca9a 931static int
1da177e4
LT		 932parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
933{
934 int temp = 0;
935 u32 cbaddr = 0;
936 u16 cbport = 0;
937 u32 addrlen = addr_len;
938 char *addr = addr_val;
939 int i, shift;
940
941 /* ipaddress */
942 shift = 24;
943 for(i = 4; i > 0 ; i--) {
944 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
945 return 0;
946 }
947 cbaddr |= (temp << shift);
948 if (shift > 0)
949 shift -= 8;
950 }
951 *cbaddrp = cbaddr;
952
953 /* port */
954 shift = 8;
955 for(i = 2; i > 0 ; i--) {
956 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
957 return 0;
958 }
959 cbport |= (temp << shift);
960 if (shift > 0)
961 shift -= 8;
962 }
963 *cbportp = cbport;
964 return 1;
965}
966
fd39ca9a 967static void
1da177e4
LT		 968gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
969{
970 struct nfs4_callback *cb = &clp->cl_callback;
971
972 /* Currently, we only support tcp for the callback channel */
973 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
974 goto out_err;
975
976 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
977 &cb->cb_addr, &cb->cb_port)))
978 goto out_err;
979 cb->cb_prog = se->se_callback_prog;
980 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 981 return;
982out_err:
849823c5 983 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 984 "will not receive delegations\n",
985 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
986
1da177e4
LT		 987 return;
988}
989
074fe897
AA		 990void
991nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
992{
993 struct nfsd4_compoundres *resp = rqstp->rq_resp;
994
995 resp->cstate.statp = statp;
996}
997
998/*
999 * Dereference the result pages.
1000 */
1001static void
1002nfsd4_release_respages(struct page **respages, short resused)
1003{
1004 int i;
1005
1006 dprintk("--> %s\n", __func__);
1007 for (i = 0; i < resused; i++) {
1008 if (!respages[i])
1009 continue;
1010 put_page(respages[i]);
1011 respages[i] = NULL;
1012 }
1013}
1014
1015static void
1016nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
1017{
1018 int i;
1019
1020 for (i = 0; i < count; i++) {
1021 topages[i] = frompages[i];
1022 if (!topages[i])
1023 continue;
1024 get_page(topages[i]);
1025 }
1026}
1027
1028/*
1029 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
1030 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
1031 * length of the XDR response is less than se_fmaxresp_cached
1032 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
1033 * of the reply (e.g. readdir).
1034 *
1035 * Store the base and length of the rq_req.head[0] page
1036 * of the NFSv4.1 data, just past the rpc header.
1037 */
1038void
1039nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
1040{
1041 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1042 struct svc_rqst *rqstp = resp->rqstp;
1043 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1044 struct nfsd4_op *op = &args->ops[resp->opcnt];
1045 struct kvec *resv = &rqstp->rq_res.head[0];
1046
1047 dprintk("--> %s entry %p\n", __func__, entry);
1048
1049 /* Don't cache a failed OP_SEQUENCE. */
1050 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1051 return;
bf864a31 1052
074fe897 1053 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1054 entry->ce_opcnt = resp->opcnt;
1055 entry->ce_status = resp->cstate.status;
1056
1057 /*
1058 * Don't need a page to cache just the sequence operation - the slot
1059 * does this for us!
1060 */
1061
1062 if (nfsd4_not_cached(resp)) {
1063 entry->ce_resused = 0;
1064 entry->ce_rpchdrlen = 0;
1065 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1066 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1067 return;
1068 }
074fe897
AA		 1069 entry->ce_resused = rqstp->rq_resused;
1070 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1071 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1072 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1073 entry->ce_resused);
074fe897
AA		 1074 entry->ce_datav.iov_base = resp->cstate.statp;
1075 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1076 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1077 /* Current request rpc header length*/
1078 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1079 (char *)page_address(rqstp->rq_respages[0]);
1080}
1081
1082/*
1083 * We keep the rpc header, but take the nfs reply from the replycache.
1084 */
1085static int
1086nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1087 struct nfsd4_cache_entry *entry)
1088{
1089 struct svc_rqst *rqstp = resp->rqstp;
1090 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1091 int len;
1092
1093 /* Current request rpc header length*/
1094 len = (char *)resp->cstate.statp -
1095 (char *)page_address(rqstp->rq_respages[0]);
1096 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1097 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1098 entry->ce_datav.iov_len);
1099 return 0;
1100 }
1101 /* copy the cached reply nfsd data past the current rpc header */
1102 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1103 entry->ce_datav.iov_len);
1104 resv->iov_len = len + entry->ce_datav.iov_len;
1105 return 1;
1106}
1107
1108/*
1109 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1110 * cached page. Replace any futher replay pages from the cache.
1111 */
1112__be32
bf864a31
AA		 1113nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1114 struct nfsd4_sequence *seq)
074fe897
AA		 1115{
1116 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1117 __be32 status;
1118
1119 dprintk("--> %s entry %p\n", __func__, entry);
1120
bf864a31
AA		 1121 /*
1122 * If this is just the sequence operation, we did not keep
1123 * a page in the cache entry because we can just use the
1124 * slot info stored in struct nfsd4_sequence that was checked
1125 * against the slot in nfsd4_sequence().
1126 *
1127 * This occurs when seq->cachethis is FALSE, or when the client
1128 * session inactivity timer fires and a solo sequence operation
1129 * is sent (lease renewal).
1130 */
1131 if (seq && nfsd4_not_cached(resp)) {
1132 seq->maxslots = resp->cstate.session->se_fnumslots;
1133 return nfs_ok;
1134 }
074fe897
AA		 1135
1136 if (!nfsd41_copy_replay_data(resp, entry)) {
1137 /*
1138 * Not enough room to use the replay rpc header, send the
1139 * cached header. Release all the allocated result pages.
1140 */
1141 svc_free_res_pages(resp->rqstp);
1142 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1143 entry->ce_resused);
1144 } else {
1145 /* Release all but the first allocated result page */
1146
1147 resp->rqstp->rq_resused--;
1148 svc_free_res_pages(resp->rqstp);
1149
1150 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1151 &entry->ce_respages[1],
1152 entry->ce_resused - 1);
1153 }
1154
1155 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1156 resp->opcnt = entry->ce_opcnt;
1157 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1158 status = entry->ce_status;
1159
1160 return status;
1161}
1162
0733d213
AA		 1163/*
1164 * Set the exchange_id flags returned by the server.
1165 */
1166static void
1167nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1168{
1169 /* pNFS is not supported */
1170 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1171
1172 /* Referrals are supported, Migration is not. */
1173 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1174
1175 /* set the wire flags to return to client. */
1176 clid->flags = new->cl_exchange_flags;
1177}
1178
069b6ad4
AA		 1179__be32
1180nfsd4_exchange_id(struct svc_rqst *rqstp,
1181 struct nfsd4_compound_state *cstate,
1182 struct nfsd4_exchange_id *exid)
1183{
0733d213
AA		 1184 struct nfs4_client *unconf, *conf, *new;
1185 int status;
1186 unsigned int strhashval;
1187 char dname[HEXDIR_LEN];
1188 nfs4_verifier verf = exid->verifier;
1189 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
1190
1191 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
1192 " ip_addr=%u flags %x, spa_how %d\n",
1193 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
1194 ip_addr, exid->flags, exid->spa_how);
1195
1196 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1197 return nfserr_inval;
1198
1199 /* Currently only support SP4_NONE */
1200 switch (exid->spa_how) {
1201 case SP4_NONE:
1202 break;
1203 case SP4_SSV:
1204 return nfserr_encr_alg_unsupp;
1205 default:
1206 BUG(); /* checked by xdr code */
1207 case SP4_MACH_CRED:
1208 return nfserr_serverfault; /* no excuse :-/ */
1209 }
1210
1211 status = nfs4_make_rec_clidname(dname, &exid->clname);
1212
1213 if (status)
1214 goto error;
1215
1216 strhashval = clientstr_hashval(dname);
1217
1218 nfs4_lock_state();
1219 status = nfs_ok;
1220
a1bcecd2 1221 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1222 if (conf) {
1223 if (!same_verf(&verf, &conf->cl_verifier)) {
1224 /* 18.35.4 case 8 */
1225 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1226 status = nfserr_not_same;
1227 goto out;
1228 }
1229 /* Client reboot: destroy old state */
1230 expire_client(conf);
1231 goto out_new;
1232 }
1233 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1234 /* 18.35.4 case 9 */
1235 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1236 status = nfserr_perm;
1237 goto out;
1238 }
1239 expire_client(conf);
1240 goto out_new;
1241 }
1242 if (ip_addr != conf->cl_addr &&
1243 !(exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A)) {
1244 /* Client collision. 18.35.4 case 3 */
1245 status = nfserr_clid_inuse;
1246 goto out;
1247 }
1248 /*
1249 * Set bit when the owner id and verifier map to an already
1250 * confirmed client id (18.35.3).
1251 */
1252 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1253
1254 /*
1255 * Falling into 18.35.4 case 2, possible router replay.
1256 * Leave confirmed record intact and return same result.
1257 */
1258 copy_verf(conf, &verf);
1259 new = conf;
1260 goto out_copy;
1261 } else {
1262 /* 18.35.4 case 7 */
1263 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1264 status = nfserr_noent;
1265 goto out;
1266 }
1267 }
1268
a1bcecd2 1269 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1270 if (unconf) {
1271 /*
1272 * Possible retry or client restart. Per 18.35.4 case 4,
1273 * a new unconfirmed record should be generated regardless
1274 * of whether any properties have changed.
1275 */
1276 expire_client(unconf);
1277 }
1278
1279out_new:
1280 /* Normal case */
1281 new = create_client(exid->clname, dname);
1282 if (new == NULL) {
1283 status = nfserr_resource;
1284 goto out;
1285 }
1286
1287 copy_verf(new, &verf);
1288 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1289 new->cl_addr = ip_addr;
1290 gen_clid(new);
1291 gen_confirm(new);
1292 add_to_unconfirmed(new, strhashval);
1293out_copy:
1294 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1295 exid->clientid.cl_id = new->cl_clientid.cl_id;
1296
38eb76a5
AA		 1297 new->cl_slot.sl_seqid = 0;
1298 exid->seqid = 1;
0733d213
AA		 1299 nfsd4_set_ex_flags(new, exid);
1300
1301 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
38eb76a5 1302 new->cl_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1303 status = nfs_ok;
1304
1305out:
1306 nfs4_unlock_state();
1307error:
1308 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1309 return status;
069b6ad4
AA		 1310}
1311
b85d4c01
BH		 1312static int
1313check_slot_seqid(u32 seqid, struct nfsd4_slot *slot)
1314{
1315 dprintk("%s enter. seqid %d slot->sl_seqid %d\n", __func__, seqid,
1316 slot->sl_seqid);
1317
1318 /* The slot is in use, and no response has been sent. */
1319 if (slot->sl_inuse) {
1320 if (seqid == slot->sl_seqid)
1321 return nfserr_jukebox;
1322 else
1323 return nfserr_seq_misordered;
1324 }
1325 /* Normal */
1326 if (likely(seqid == slot->sl_seqid + 1))
1327 return nfs_ok;
1328 /* Replay */
1329 if (seqid == slot->sl_seqid)
1330 return nfserr_replay_cache;
1331 /* Wraparound */
1332 if (seqid == 1 && (slot->sl_seqid + 1) == 0)
1333 return nfs_ok;
1334 /* Misordered replay or misordered new request */
1335 return nfserr_seq_misordered;
1336}
1337
069b6ad4
AA		 1338__be32
1339nfsd4_create_session(struct svc_rqst *rqstp,
1340 struct nfsd4_compound_state *cstate,
1341 struct nfsd4_create_session *cr_ses)
1342{
ec6b5d7b 1343 u32 ip_addr = svc_addr_in(rqstp)->sin_addr.s_addr;
38eb76a5 1344 struct nfsd4_compoundres *resp = rqstp->rq_resp;
ec6b5d7b 1345 struct nfs4_client *conf, *unconf;
38eb76a5 1346 struct nfsd4_slot *slot = NULL;
ec6b5d7b
AA		 1347 int status = 0;
1348
1349 nfs4_lock_state();
1350 unconf = find_unconfirmed_client(&cr_ses->clientid);
1351 conf = find_confirmed_client(&cr_ses->clientid);
1352
1353 if (conf) {
38eb76a5
AA		 1354 slot = &conf->cl_slot;
1355 status = check_slot_seqid(cr_ses->seqid, slot);
1356 if (status == nfserr_replay_cache) {
ec6b5d7b 1357 dprintk("Got a create_session replay! seqid= %d\n",
38eb76a5
AA		 1358 slot->sl_seqid);
1359 cstate->slot = slot;
1360 cstate->status = status;
1361 /* Return the cached reply status */
bf864a31 1362 status = nfsd4_replay_cache_entry(resp, NULL);
38eb76a5
AA		 1363 goto out;
1364 } else if (cr_ses->seqid != conf->cl_slot.sl_seqid + 1) {
ec6b5d7b
AA		 1365 status = nfserr_seq_misordered;
1366 dprintk("Sequence misordered!\n");
1367 dprintk("Expected seqid= %d but got seqid= %d\n",
38eb76a5 1368 slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1369 goto out;
1370 }
38eb76a5 1371 conf->cl_slot.sl_seqid++;
ec6b5d7b
AA		 1372 } else if (unconf) {
1373 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1374 (ip_addr != unconf->cl_addr)) {
1375 status = nfserr_clid_inuse;
1376 goto out;
1377 }
1378
38eb76a5
AA		 1379 slot = &unconf->cl_slot;
1380 status = check_slot_seqid(cr_ses->seqid, slot);
1381 if (status) {
1382 /* an unconfirmed replay returns misordered */
ec6b5d7b
AA		 1383 status = nfserr_seq_misordered;
1384 goto out;
1385 }
1386
38eb76a5 1387 slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1388 move_to_confirmed(unconf);
1389
1390 /*
1391 * We do not support RDMA or persistent sessions
1392 */
1393 cr_ses->flags &= ~SESSION4_PERSIST;
1394 cr_ses->flags &= ~SESSION4_RDMA;
1395
1396 conf = unconf;
1397 } else {
1398 status = nfserr_stale_clientid;
1399 goto out;
1400 }
1401
1402 status = alloc_init_session(rqstp, conf, cr_ses);
1403 if (status)
1404 goto out;
1405
ec6b5d7b
AA		 1406 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1407 NFS4_MAX_SESSIONID_LEN);
38eb76a5 1408 cr_ses->seqid = slot->sl_seqid;
ec6b5d7b 1409
38eb76a5
AA		 1410 slot->sl_inuse = true;
1411 cstate->slot = slot;
bf864a31
AA		 1412 /* Ensure a page is used for the cache */
1413 slot->sl_cache_entry.ce_cachethis = 1;
ec6b5d7b
AA		 1414out:
1415 nfs4_unlock_state();
1416 dprintk("%s returns %d\n", __func__, ntohl(status));
1417 return status;
069b6ad4
AA		 1418}
1419
1420__be32
1421nfsd4_destroy_session(struct svc_rqst *r,
1422 struct nfsd4_compound_state *cstate,
1423 struct nfsd4_destroy_session *sessionid)
1424{
1425 return -1; /* stub */
1426}
1427
1428__be32
b85d4c01 1429nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1430 struct nfsd4_compound_state *cstate,
1431 struct nfsd4_sequence *seq)
1432{
f9bb94c4 1433 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1434 struct nfsd4_session *session;
1435 struct nfsd4_slot *slot;
1436 int status;
1437
f9bb94c4
AA		 1438 if (resp->opcnt != 1)
1439 return nfserr_sequence_pos;
1440
b85d4c01
BH		 1441 spin_lock(&sessionid_lock);
1442 status = nfserr_badsession;
1443 session = find_in_sessionid_hashtbl(&seq->sessionid);
1444 if (!session)
1445 goto out;
1446
1447 status = nfserr_badslot;
1448 if (seq->slotid >= session->se_fnumslots)
1449 goto out;
1450
1451 slot = &session->se_slots[seq->slotid];
1452 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1453
1454 status = check_slot_seqid(seq->seqid, slot);
1455 if (status == nfserr_replay_cache) {
1456 cstate->slot = slot;
1457 cstate->session = session;
da3846a2 1458 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1459 * for nfsd4_svc_encode_compoundres processing */
1460 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1461 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1462 goto replay_cache;
1463 }
1464 if (status)
1465 goto out;
1466
1467 /* Success! bump slot seqid */
1468 slot->sl_inuse = true;
1469 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1470 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1471 /* Always set the cache entry cachethis for solo sequence */
1472 if (nfsd4_is_solo_sequence(resp))
1473 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1474
1475 cstate->slot = slot;
1476 cstate->session = session;
1477
1478replay_cache:
1479 /* Renew the clientid on success and on replay.
1480 * Hold a session reference until done processing the compound:
1481 * nfsd4_put_session called only if the cstate slot is set.
1482 */
1483 renew_client(session->se_client);
1484 nfsd4_get_session(session);
1485out:
1486 spin_unlock(&sessionid_lock);
1487 dprintk("%s: return %d\n", __func__, ntohl(status));
1488 return status;
069b6ad4
AA		 1489}
1490
b37ad28b 1491__be32
b591480b
BF		 1492nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1493 struct nfsd4_setclientid *setclid)
1da177e4 1494{
27459f09 1495 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1496 struct xdr_netobj clname = {
1497 .len = setclid->se_namelen,
1498 .data = setclid->se_name,
1499 };
1500 nfs4_verifier clverifier = setclid->se_verf;
1501 unsigned int strhashval;
28ce6054 1502 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1503 __be32 status;
68e76ad0 1504 char *princ;
a55370a3 1505 char dname[HEXDIR_LEN];
1da177e4 1506
1da177e4 1507 if (!check_name(clname))
73aea4ec 1508 return nfserr_inval;
1da177e4 1509
a55370a3
N		 1510 status = nfs4_make_rec_clidname(dname, &clname);
1511 if (status)
73aea4ec 1512 return status;
a55370a3 1513
1da177e4
LT		 1514 /*
1515 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1516 * We get here on a DRC miss.
1517 */
1518
a55370a3 1519 strhashval = clientstr_hashval(dname);
1da177e4 1520
1da177e4 1521 nfs4_lock_state();
a1bcecd2 1522 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1523 if (conf) {
a186e767 1524 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1525 status = nfserr_clid_inuse;
026722c2
BF		 1526 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1527 dprintk("NFSD: setclientid: string in use by client"
1528 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1529 goto out;
1530 }
1da177e4 1531 }
a186e767
BF		 1532 /*
1533 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1534 * has a description of SETCLIENTID request processing consisting
1535 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1536 */
a1bcecd2 1537 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1538 status = nfserr_resource;
1539 if (!conf) {
a186e767
BF		 1540 /*
1541 * RFC 3530 14.2.33 CASE 4:
1542 * placed first, because it is the normal case
1da177e4
LT		 1543 */
1544 if (unconf)
1545 expire_client(unconf);
a55370a3
N		 1546 new = create_client(clname, dname);
1547 if (new == NULL)
1da177e4 1548 goto out;
1da177e4 1549 gen_clid(new);
599e0a22 1550 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1551 /*
a186e767
BF		 1552 * RFC 3530 14.2.33 CASE 1:
1553 * probable callback update
1da177e4 1554 */
31f4a6c1
N		 1555 if (unconf) {
1556 /* Note this is removing unconfirmed {*x***},
1557 * which is stronger than RFC recommended {vxc**}.
1558 * This has the advantage that there is at most
1559 * one {*x***} in either list at any time.
1560 */
1561 expire_client(unconf);
1da177e4 1562 }
a55370a3
N		 1563 new = create_client(clname, dname);
1564 if (new == NULL)
1da177e4 1565 goto out;
1da177e4 1566 copy_clid(new, conf);
1da177e4
LT		 1567 } else if (!unconf) {
1568 /*
a186e767
BF		 1569 * RFC 3530 14.2.33 CASE 2:
1570 * probable client reboot; state will be removed if
1571 * confirmed.
1da177e4 1572 */
a55370a3
N		 1573 new = create_client(clname, dname);
1574 if (new == NULL)
1da177e4 1575 goto out;
1da177e4 1576 gen_clid(new);
49ba8781 1577 } else {
a186e767
BF		 1578 /*
1579 * RFC 3530 14.2.33 CASE 3:
1580 * probable client reboot; state will be removed if
1581 * confirmed.
1da177e4
LT		 1582 */
1583 expire_client(unconf);
a55370a3
N		 1584 new = create_client(clname, dname);
1585 if (new == NULL)
1da177e4 1586 goto out;
1da177e4 1587 gen_clid(new);
1da177e4 1588 }
c175b83c
BF		 1589 copy_verf(new, &clverifier);
1590 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1591 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1592 princ = svc_gss_principal(rqstp);
1593 if (princ) {
1594 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1595 if (new->cl_principal == NULL) {
1596 free_client(new);
1597 goto out;
1598 }
1599 }
c175b83c
BF		 1600 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1601 gen_confirm(new);
1602 gen_callback(new, setclid);
1603 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1604 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1605 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1606 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1607 status = nfs_ok;
1608out:
1609 nfs4_unlock_state();
1610 return status;
1611}
1612
1613
1614/*
a186e767
BF		 1615 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1616 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1617 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1618 */
b37ad28b 1619__be32
b591480b
BF		 1620nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1621 struct nfsd4_compound_state *cstate,
1622 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1623{
27459f09 1624 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1625 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1626 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1627 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1628 __be32 status;
1da177e4
LT		 1629
1630 if (STALE_CLIENTID(clid))
1631 return nfserr_stale_clientid;
1632 /*
1633 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1634 * We get here on a DRC miss.
1635 */
1636
1637 nfs4_lock_state();
21ab45a4
N		 1638
1639 conf = find_confirmed_client(clid);
1640 unconf = find_unconfirmed_client(clid);
1641
1642 status = nfserr_clid_inuse;
27459f09 1643 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1644 goto out;
27459f09 1645 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1646 goto out;
1647
a186e767
BF		 1648 /*
1649 * section 14.2.34 of RFC 3530 has a description of
1650 * SETCLIENTID_CONFIRM request processing consisting
1651 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1652 */
366e0c1d 1653 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1654 /*
1655 * RFC 3530 14.2.34 CASE 1:
1656 * callback update
1657 */
599e0a22 1658 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1659 status = nfserr_clid_inuse;
1660 else {
1a69c179
N		 1661 /* XXX: We just turn off callbacks until we can handle
1662 * change request correctly. */
cb36d634 1663 atomic_set(&conf->cl_callback.cb_set, 0);
21ab45a4 1664 gen_confirm(conf);
46309029 1665 nfsd4_remove_clid_dir(unconf);
1a69c179 1666 expire_client(unconf);
1da177e4 1667 status = nfs_ok;
1a69c179 1668
1da177e4 1669 }
f3aba4e5 1670 } else if (conf && !unconf) {
a186e767
BF		 1671 /*
1672 * RFC 3530 14.2.34 CASE 2:
1673 * probable retransmitted request; play it safe and
1674 * do nothing.
7c79f737 1675 */
599e0a22 1676 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1677 status = nfserr_clid_inuse;
21ab45a4 1678 else
1da177e4 1679 status = nfs_ok;
7c79f737 1680 } else if (!conf && unconf
599e0a22 1681 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1682 /*
1683 * RFC 3530 14.2.34 CASE 3:
1684 * Normal case; new or rebooted client:
7c79f737 1685 */
599e0a22 1686 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1687 status = nfserr_clid_inuse;
1688 } else {
1a69c179
N		 1689 unsigned int hash =
1690 clientstr_hashval(unconf->cl_recdir);
1691 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1692 hash, false);
1a69c179 1693 if (conf) {
c7b9a459 1694 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1695 expire_client(conf);
1696 }
1da177e4 1697 move_to_confirmed(unconf);
21ab45a4 1698 conf = unconf;
46f8a64b 1699 nfsd4_probe_callback(conf);
1a69c179 1700 status = nfs_ok;
1da177e4 1701 }
599e0a22
BF		 1702 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1703 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1704 &confirm)))) {
a186e767
BF		 1705 /*
1706 * RFC 3530 14.2.34 CASE 4:
1707 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1708 */
1da177e4 1709 status = nfserr_stale_clientid;
7c79f737 1710 } else {
08e8987c
N		 1711 /* check that we have hit one of the cases...*/
1712 status = nfserr_clid_inuse;
1713 }
1da177e4 1714out:
1da177e4
LT		 1715 nfs4_unlock_state();
1716 return status;
1717}
1718
1da177e4
LT		 1719/* OPEN Share state helper functions */
1720static inline struct nfs4_file *
1721alloc_init_file(struct inode *ino)
1722{
1723 struct nfs4_file *fp;
1724 unsigned int hashval = file_hashval(ino);
1725
e60d4398
N		 1726 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1727 if (fp) {
8b671b80 1728 atomic_set(&fp->fi_ref, 1);
1da177e4 1729 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1730 INIT_LIST_HEAD(&fp->fi_stateids);
1731 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1732 spin_lock(&recall_lock);
1da177e4 1733 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1734 spin_unlock(&recall_lock);
1da177e4
LT		 1735 fp->fi_inode = igrab(ino);
1736 fp->fi_id = current_fileid++;
47f9940c 1737 fp->fi_had_conflict = false;
1da177e4
LT		 1738 return fp;
1739 }
1740 return NULL;
1741}
1742
e60d4398 1743static void
e18b890b 1744nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1745{
e60d4398
N		 1746 if (*slab == NULL)
1747 return;
1a1d92c1 1748 kmem_cache_destroy(*slab);
e60d4398 1749 *slab = NULL;
1da177e4
LT		 1750}
1751
e8ff2a84 1752void
1da177e4
LT		 1753nfsd4_free_slabs(void)
1754{
e60d4398
N		 1755 nfsd4_free_slab(&stateowner_slab);
1756 nfsd4_free_slab(&file_slab);
5ac049ac 1757 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1758 nfsd4_free_slab(&deleg_slab);
e60d4398 1759}
1da177e4 1760
e60d4398
N		 1761static int
1762nfsd4_init_slabs(void)
1763{
1764 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1765 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1766 if (stateowner_slab == NULL)
1767 goto out_nomem;
1768 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1769 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1770 if (file_slab == NULL)
1771 goto out_nomem;
5ac049ac 1772 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1773 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1774 if (stateid_slab == NULL)
1775 goto out_nomem;
5b2d21c1 1776 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1777 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1778 if (deleg_slab == NULL)
1779 goto out_nomem;
e60d4398
N		 1780 return 0;
1781out_nomem:
1782 nfsd4_free_slabs();
1783 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1784 return -ENOMEM;
1da177e4
LT		 1785}
1786
1787void
1788nfs4_free_stateowner(struct kref *kref)
1789{
1790 struct nfs4_stateowner *sop =
1791 container_of(kref, struct nfs4_stateowner, so_ref);
1792 kfree(sop->so_owner.data);
1793 kmem_cache_free(stateowner_slab, sop);
1794}
1795
1796static inline struct nfs4_stateowner *
1797alloc_stateowner(struct xdr_netobj *owner)
1798{
1799 struct nfs4_stateowner *sop;
1800
1801 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1802 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1803 memcpy(sop->so_owner.data, owner->data, owner->len);
1804 sop->so_owner.len = owner->len;
1805 kref_init(&sop->so_ref);
1806 return sop;
1807 }
1808 kmem_cache_free(stateowner_slab, sop);
1809 }
1810 return NULL;
1811}
1812
1813static struct nfs4_stateowner *
1814alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1815 struct nfs4_stateowner *sop;
1816 struct nfs4_replay *rp;
1817 unsigned int idhashval;
1818
1819 if (!(sop = alloc_stateowner(&open->op_owner)))
1820 return NULL;
1821 idhashval = ownerid_hashval(current_ownerid);
1822 INIT_LIST_HEAD(&sop->so_idhash);
1823 INIT_LIST_HEAD(&sop->so_strhash);
1824 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1825 INIT_LIST_HEAD(&sop->so_stateids);
1826 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1827 INIT_LIST_HEAD(&sop->so_close_lru);
1828 sop->so_time = 0;
1829 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1830 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1831 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1832 sop->so_is_open_owner = 1;
1833 sop->so_id = current_ownerid++;
1834 sop->so_client = clp;
1835 sop->so_seqid = open->op_seqid;
1836 sop->so_confirmed = 0;
1837 rp = &sop->so_replay;
de1ae286 1838 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1839 rp->rp_buflen = 0;
1840 rp->rp_buf = rp->rp_ibuf;
1841 return sop;
1842}
1843
1da177e4
LT		 1844static inline void
1845init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1846 struct nfs4_stateowner *sop = open->op_stateowner;
1847 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1848
1849 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1850 INIT_LIST_HEAD(&stp->st_perstateowner);
1851 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1852 INIT_LIST_HEAD(&stp->st_perfile);
1853 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1854 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1855 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1856 stp->st_stateowner = sop;
13cd2184 1857 get_nfs4_file(fp);
1da177e4
LT		 1858 stp->st_file = fp;
1859 stp->st_stateid.si_boot = boot_time;
1860 stp->st_stateid.si_stateownerid = sop->so_id;
1861 stp->st_stateid.si_fileid = fp->fi_id;
1862 stp->st_stateid.si_generation = 0;
1863 stp->st_access_bmap = 0;
1864 stp->st_deny_bmap = 0;
1865 __set_bit(open->op_share_access, &stp->st_access_bmap);
1866 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1867 stp->st_openstp = NULL;
1da177e4
LT		 1868}
1869
fd39ca9a 1870static void
1da177e4
LT		 1871move_to_close_lru(struct nfs4_stateowner *sop)
1872{
1873 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1874
358dd55a 1875 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1876 sop->so_time = get_seconds();
1877}
1878
1da177e4 1879static int
599e0a22
BF		 1880same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1881 clientid_t *clid)
1882{
1883 return (sop->so_owner.len == owner->len) &&
1884 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1885 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1886}
1887
1888static struct nfs4_stateowner *
1889find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1890{
1891 struct nfs4_stateowner *so = NULL;
1892
1893 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1894 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1895 return so;
1896 }
1897 return NULL;
1898}
1899
1900/* search file_hashtbl[] for file */
1901static struct nfs4_file *
1902find_file(struct inode *ino)
1903{
1904 unsigned int hashval = file_hashval(ino);
1905 struct nfs4_file *fp;
1906
8b671b80 1907 spin_lock(&recall_lock);
1da177e4 1908 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1909 if (fp->fi_inode == ino) {
1910 get_nfs4_file(fp);
8b671b80 1911 spin_unlock(&recall_lock);
1da177e4 1912 return fp;
13cd2184 1913 }
1da177e4 1914 }
8b671b80 1915 spin_unlock(&recall_lock);
1da177e4
LT		 1916 return NULL;
1917}
1918
8838dc43 1919static inline int access_valid(u32 x)
ba5a6a19 1920{
8838dc43
BF		 1921 if (x < NFS4_SHARE_ACCESS_READ)
1922 return 0;
1923 if (x > NFS4_SHARE_ACCESS_BOTH)
1924 return 0;
1925 return 1;
ba5a6a19
BF		 1926}
1927
8838dc43 1928static inline int deny_valid(u32 x)
ba5a6a19 1929{
8838dc43
BF		 1930 /* Note: unlike access bits, deny bits may be zero. */
1931 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1932}
1da177e4 1933
4f83aa30
BF		 1934/*
1935 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1936 * st_{access,deny}_bmap field of the stateid, in order to track not
1937 * only what share bits are currently in force, but also what
1938 * combinations of share bits previous opens have used. This allows us
1939 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1940 * return an error if the client attempt to downgrade to a combination
1941 * of share bits not explicable by closing some of its previous opens.
1942 *
1943 * XXX: This enforcement is actually incomplete, since we don't keep
1944 * track of access/deny bit combinations; so, e.g., we allow:
1945 *
1946 * OPEN allow read, deny write
1947 * OPEN allow both, deny none
1948 * DOWNGRADE allow read, deny none
1949 *
1950 * which we should reject.
1951 */
fd39ca9a 1952static void
1da177e4
LT		 1953set_access(unsigned int *access, unsigned long bmap) {
1954 int i;
1955
1956 *access = 0;
1957 for (i = 1; i < 4; i++) {
1958 if (test_bit(i, &bmap))
1959 *access |= i;
1960 }
1961}
1962
fd39ca9a 1963static void
1da177e4
LT		 1964set_deny(unsigned int *deny, unsigned long bmap) {
1965 int i;
1966
1967 *deny = 0;
1968 for (i = 0; i < 4; i++) {
1969 if (test_bit(i, &bmap))
1970 *deny |= i ;
1971 }
1972}
1973
1974static int
1975test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1976 unsigned int access, deny;
1977
1978 set_access(&access, stp->st_access_bmap);
1979 set_deny(&deny, stp->st_deny_bmap);
1980 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1981 return 0;
1982 return 1;
1983}
1984
1985/*
1986 * Called to check deny when READ with all zero stateid or
1987 * WRITE with all zero or all one stateid
1988 */
b37ad28b 1989static __be32
1da177e4
LT		 1990nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1991{
1992 struct inode *ino = current_fh->fh_dentry->d_inode;
1993 struct nfs4_file *fp;
1994 struct nfs4_stateid *stp;
b37ad28b 1995 __be32 ret;
1da177e4
LT		 1996
1997 dprintk("NFSD: nfs4_share_conflict\n");
1998
1999 fp = find_file(ino);
13cd2184
N		 2000 if (!fp)
2001 return nfs_ok;
b700949b 2002 ret = nfserr_locked;
1da177e4 2003 /* Search for conflicting share reservations */
13cd2184
N		 2004 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2005 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2006 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2007 goto out;
1da177e4 2008 }
13cd2184
N		 2009 ret = nfs_ok;
2010out:
2011 put_nfs4_file(fp);
2012 return ret;
1da177e4
LT		 2013}
2014
2015static inline void
2016nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2017{
2018 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2019 drop_file_write_access(filp);
1da177e4
LT		 2020 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2021 }
2022}
2023
2024/*
2025 * Recall a delegation
2026 */
2027static int
2028do_recall(void *__dp)
2029{
2030 struct nfs4_delegation *dp = __dp;
2031
47f9940c 2032 dp->dl_file->fi_had_conflict = true;
1da177e4
LT		 2033 nfsd4_cb_recall(dp);
2034 return 0;
2035}
2036
2037/*
2038 * Spawn a thread to perform a recall on the delegation represented
2039 * by the lease (file_lock)
2040 *
2041 * Called from break_lease() with lock_kernel() held.
2042 * Note: we assume break_lease will only call this *once* for any given
2043 * lease.
2044 */
2045static
2046void nfsd_break_deleg_cb(struct file_lock *fl)
2047{
2048 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
2049 struct task_struct *t;
2050
2051 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2052 if (!dp)
2053 return;
2054
2055 /* We're assuming the state code never drops its reference
2056 * without first removing the lease. Since we're in this lease
2057 * callback (and since the lease code is serialized by the kernel
2058 * lock) we know the server hasn't removed the lease yet, we know
2059 * it's safe to take a reference: */
2060 atomic_inc(&dp->dl_count);
cfdcad4d 2061 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2062
2063 spin_lock(&recall_lock);
2064 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2065 spin_unlock(&recall_lock);
2066
2067 /* only place dl_time is set. protected by lock_kernel*/
2068 dp->dl_time = get_seconds();
2069
0272e1fd
BF		 2070 /*
2071 * We don't want the locks code to timeout the lease for us;
2072 * we'll remove it ourself if the delegation isn't returned
2073 * in time.
2074 */
2075 fl->fl_break_time = 0;
1da177e4
LT		 2076
2077 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
2078 if (IS_ERR(t)) {
2079 struct nfs4_client *clp = dp->dl_client;
2080
2081 printk(KERN_INFO "NFSD: Callback thread failed for "
2082 "for client (clientid %08x/%08x)\n",
2083 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 2084 put_nfs4_client(dp->dl_client);
1da177e4
LT		 2085 nfs4_put_delegation(dp);
2086 }
2087}
2088
2089/*
2090 * The file_lock is being reapd.
2091 *
2092 * Called by locks_free_lock() with lock_kernel() held.
2093 */
2094static
2095void nfsd_release_deleg_cb(struct file_lock *fl)
2096{
2097 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2098
2099 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2100
2101 if (!(fl->fl_flags & FL_LEASE) || !dp)
2102 return;
2103 dp->dl_flock = NULL;
2104}
2105
2106/*
2107 * Set the delegation file_lock back pointer.
2108 *
a9933cea 2109 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2110 */
2111static
2112void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2113{
2114 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2115
2116 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2117 if (!dp)
2118 return;
2119 dp->dl_flock = new;
2120}
2121
2122/*
a9933cea 2123 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2124 */
2125static
2126int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2127{
2128 struct nfs4_delegation *onlistd =
2129 (struct nfs4_delegation *)onlist->fl_owner;
2130 struct nfs4_delegation *tryd =
2131 (struct nfs4_delegation *)try->fl_owner;
2132
2133 if (onlist->fl_lmops != try->fl_lmops)
2134 return 0;
2135
2136 return onlistd->dl_client == tryd->dl_client;
2137}
2138
2139
2140static
2141int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2142{
2143 if (arg & F_UNLCK)
2144 return lease_modify(onlist, arg);
2145 else
2146 return -EAGAIN;
2147}
2148
fd39ca9a 2149static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2150 .fl_break = nfsd_break_deleg_cb,
2151 .fl_release_private = nfsd_release_deleg_cb,
2152 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2153 .fl_mylease = nfsd_same_client_deleg_cb,
2154 .fl_change = nfsd_change_deleg_cb,
2155};
2156
2157
b37ad28b 2158__be32
1da177e4
LT		 2159nfsd4_process_open1(struct nfsd4_open *open)
2160{
1da177e4
LT		 2161 clientid_t *clientid = &open->op_clientid;
2162 struct nfs4_client *clp = NULL;
2163 unsigned int strhashval;
2164 struct nfs4_stateowner *sop = NULL;
2165
1da177e4 2166 if (!check_name(open->op_owner))
0f442aa2 2167 return nfserr_inval;
1da177e4
LT		 2168
2169 if (STALE_CLIENTID(&open->op_clientid))
2170 return nfserr_stale_clientid;
2171
2172 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2173 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2174 open->op_stateowner = sop;
2175 if (!sop) {
2176 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2177 clp = find_confirmed_client(clientid);
2178 if (clp == NULL)
0f442aa2
BF		 2179 return nfserr_expired;
2180 goto renew;
1da177e4 2181 }
0f442aa2
BF		 2182 if (!sop->so_confirmed) {
2183 /* Replace unconfirmed owners without checking for replay. */
2184 clp = sop->so_client;
f044ff83 2185 release_openowner(sop);
0f442aa2
BF		 2186 open->op_stateowner = NULL;
2187 goto renew;
2188 }
2189 if (open->op_seqid == sop->so_seqid - 1) {
2190 if (sop->so_replay.rp_buflen)
a90b061c 2191 return nfserr_replay_me;
0f442aa2
BF		 2192 /* The original OPEN failed so spectacularly
2193 * that we don't even have replay data saved!
2194 * Therefore, we have no choice but to continue
2195 * processing this OPEN; presumably, we'll
2196 * fail again for the same reason.
2197 */
2198 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2199 goto renew;
2200 }
2201 if (open->op_seqid != sop->so_seqid)
2202 return nfserr_bad_seqid;
1da177e4 2203renew:
0f442aa2
BF		 2204 if (open->op_stateowner == NULL) {
2205 sop = alloc_init_open_stateowner(strhashval, clp, open);
2206 if (sop == NULL)
2207 return nfserr_resource;
2208 open->op_stateowner = sop;
2209 }
2210 list_del_init(&sop->so_close_lru);
1da177e4 2211 renew_client(sop->so_client);
0f442aa2 2212 return nfs_ok;
1da177e4
LT		 2213}
2214
b37ad28b 2215static inline __be32
4a6e43e6
N		 2216nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2217{
2218 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2219 return nfserr_openmode;
2220 else
2221 return nfs_ok;
2222}
2223
52f4fb43
N		 2224static struct nfs4_delegation *
2225find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2226{
2227 struct nfs4_delegation *dp;
2228
ea1da636 2229 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2230 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2231 return dp;
2232 }
2233 return NULL;
2234}
2235
b37ad28b 2236static __be32
567d9829
N		 2237nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2238 struct nfs4_delegation **dp)
2239{
2240 int flags;
b37ad28b 2241 __be32 status = nfserr_bad_stateid;
567d9829
N		 2242
2243 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2244 if (*dp == NULL)
c44c5eeb 2245 goto out;
567d9829
N		 2246 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2247 RD_STATE : WR_STATE;
2248 status = nfs4_check_delegmode(*dp, flags);
2249 if (status)
2250 *dp = NULL;
c44c5eeb
N		 2251out:
2252 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2253 return nfs_ok;
2254 if (status)
2255 return status;
2256 open->op_stateowner->so_confirmed = 1;
2257 return nfs_ok;
567d9829
N		 2258}
2259
b37ad28b 2260static __be32
1da177e4
LT		 2261nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2262{
2263 struct nfs4_stateid *local;
b37ad28b 2264 __be32 status = nfserr_share_denied;
1da177e4
LT		 2265 struct nfs4_stateowner *sop = open->op_stateowner;
2266
8beefa24 2267 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2268 /* ignore lock owners */
2269 if (local->st_stateowner->so_is_open_owner == 0)
2270 continue;
2271 /* remember if we have seen this open owner */
2272 if (local->st_stateowner == sop)
2273 *stpp = local;
2274 /* check for conflicting share reservations */
2275 if (!test_share(local, open))
2276 goto out;
2277 }
2278 status = 0;
2279out:
2280 return status;
2281}
2282
5ac049ac
N		 2283static inline struct nfs4_stateid *
2284nfs4_alloc_stateid(void)
2285{
2286 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2287}
2288
b37ad28b 2289static __be32
1da177e4 2290nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2291 struct nfs4_delegation *dp,
1da177e4
LT		 2292 struct svc_fh *cur_fh, int flags)
2293{
2294 struct nfs4_stateid *stp;
1da177e4 2295
5ac049ac 2296 stp = nfs4_alloc_stateid();
1da177e4
LT		 2297 if (stp == NULL)
2298 return nfserr_resource;
2299
567d9829
N		 2300 if (dp) {
2301 get_file(dp->dl_vfs_file);
2302 stp->st_vfs_file = dp->dl_vfs_file;
2303 } else {
b37ad28b 2304 __be32 status;
567d9829
N		 2305 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2306 &stp->st_vfs_file);
2307 if (status) {
2308 if (status == nfserr_dropit)
2309 status = nfserr_jukebox;
5ac049ac 2310 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2311 return status;
2312 }
1da177e4 2313 }
1da177e4
LT		 2314 *stpp = stp;
2315 return 0;
2316}
2317
b37ad28b 2318static inline __be32
1da177e4
LT		 2319nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2320 struct nfsd4_open *open)
2321{
2322 struct iattr iattr = {
2323 .ia_valid = ATTR_SIZE,
2324 .ia_size = 0,
2325 };
2326 if (!open->op_truncate)
2327 return 0;
2328 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2329 return nfserr_inval;
1da177e4
LT		 2330 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2331}
2332
b37ad28b 2333static __be32
1da177e4
LT		 2334nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2335{
2336 struct file *filp = stp->st_vfs_file;
7eaa36e2 2337 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2338 unsigned int share_access, new_writer;
b37ad28b 2339 __be32 status;
1da177e4
LT		 2340
2341 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2342 new_writer = (~share_access) & open->op_share_access
2343 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2344
6c26d08f 2345 if (new_writer) {
b37ad28b
AV		 2346 int err = get_write_access(inode);
2347 if (err)
2348 return nfserrno(err);
e518f056
BH		 2349 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2350 if (err)
2351 return nfserrno(err);
2352 file_take_write(filp);
6c26d08f 2353 }
1da177e4
LT		 2354 status = nfsd4_truncate(rqstp, cur_fh, open);
2355 if (status) {
6c26d08f
BF		 2356 if (new_writer)
2357 put_write_access(inode);
1da177e4
LT		 2358 return status;
2359 }
2360 /* remember the open */
6c26d08f 2361 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2362 __set_bit(open->op_share_access, &stp->st_access_bmap);
2363 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2364
2365 return nfs_ok;
2366}
2367
2368
1da177e4 2369static void
37515177 2370nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2371{
37515177
N		 2372 open->op_stateowner->so_confirmed = 1;
2373 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2374}
2375
2376/*
2377 * Attempt to hand out a delegation.
2378 */
2379static void
2380nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2381{
2382 struct nfs4_delegation *dp;
2383 struct nfs4_stateowner *sop = stp->st_stateowner;
2384 struct nfs4_callback *cb = &sop->so_client->cl_callback;
2385 struct file_lock fl, *flp = &fl;
2386 int status, flag = 0;
2387
2388 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2389 open->op_recall = 0;
2390 switch (open->op_claim_type) {
2391 case NFS4_OPEN_CLAIM_PREVIOUS:
2392 if (!atomic_read(&cb->cb_set))
2393 open->op_recall = 1;
2394 flag = open->op_delegate_type;
2395 if (flag == NFS4_OPEN_DELEGATE_NONE)
2396 goto out;
2397 break;
2398 case NFS4_OPEN_CLAIM_NULL:
2399 /* Let's not give out any delegations till everyone's
2400 * had the chance to reclaim theirs.... */
af558e33 2401 if (locks_in_grace())
7b190fec
N		 2402 goto out;
2403 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2404 goto out;
2405 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2406 flag = NFS4_OPEN_DELEGATE_WRITE;
2407 else
2408 flag = NFS4_OPEN_DELEGATE_READ;
2409 break;
2410 default:
2411 goto out;
2412 }
1da177e4
LT		 2413
2414 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2415 if (dp == NULL) {
2416 flag = NFS4_OPEN_DELEGATE_NONE;
2417 goto out;
2418 }
2419 locks_init_lock(&fl);
2420 fl.fl_lmops = &nfsd_lease_mng_ops;
2421 fl.fl_flags = FL_LEASE;
9167f501 2422 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2423 fl.fl_end = OFFSET_MAX;
2424 fl.fl_owner = (fl_owner_t)dp;
2425 fl.fl_file = stp->st_vfs_file;
2426 fl.fl_pid = current->tgid;
2427
a9933cea 2428 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2429 * the lock_manager callbacks fl_mylease and fl_change are used
2430 */
9167f501 2431 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2432 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2433 unhash_delegation(dp);
1da177e4
LT		 2434 flag = NFS4_OPEN_DELEGATE_NONE;
2435 goto out;
2436 }
2437
2438 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2439
2440 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2441 dp->dl_stateid.si_boot,
2442 dp->dl_stateid.si_stateownerid,
2443 dp->dl_stateid.si_fileid,
2444 dp->dl_stateid.si_generation);
2445out:
7b190fec
N		 2446 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2447 && flag == NFS4_OPEN_DELEGATE_NONE
2448 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2449 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2450 open->op_delegate_type = flag;
2451}
2452
2453/*
2454 * called with nfs4_lock_state() held.
2455 */
b37ad28b 2456__be32
1da177e4
LT		 2457nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2458{
2459 struct nfs4_file *fp = NULL;
2460 struct inode *ino = current_fh->fh_dentry->d_inode;
2461 struct nfs4_stateid *stp = NULL;
567d9829 2462 struct nfs4_delegation *dp = NULL;
b37ad28b 2463 __be32 status;
1da177e4
LT		 2464
2465 status = nfserr_inval;
ba5a6a19
BF		 2466 if (!access_valid(open->op_share_access)
2467 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2468 goto out;
2469 /*
2470 * Lookup file; if found, lookup stateid and check open request,
2471 * and check for delegations in the process of being recalled.
2472 * If not found, create the nfs4_file struct
2473 */
2474 fp = find_file(ino);
2475 if (fp) {
2476 if ((status = nfs4_check_open(fp, open, &stp)))
2477 goto out;
c44c5eeb
N		 2478 status = nfs4_check_deleg(fp, open, &dp);
2479 if (status)
2480 goto out;
1da177e4 2481 } else {
c44c5eeb
N		 2482 status = nfserr_bad_stateid;
2483 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2484 goto out;
1da177e4
LT		 2485 status = nfserr_resource;
2486 fp = alloc_init_file(ino);
2487 if (fp == NULL)
2488 goto out;
2489 }
2490
2491 /*
2492 * OPEN the file, or upgrade an existing OPEN.
2493 * If truncate fails, the OPEN fails.
2494 */
2495 if (stp) {
2496 /* Stateid was found, this is an OPEN upgrade */
2497 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2498 if (status)
2499 goto out;
444c2c07 2500 update_stateid(&stp->st_stateid);
1da177e4
LT		 2501 } else {
2502 /* Stateid was not found, this is a new OPEN */
2503 int flags = 0;
9ecb6a08 2504 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2505 flags |= NFSD_MAY_READ;
1da177e4 2506 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2507 flags |= NFSD_MAY_WRITE;
567d9829
N		 2508 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2509 if (status)
1da177e4
LT		 2510 goto out;
2511 init_stateid(stp, fp, open);
2512 status = nfsd4_truncate(rqstp, current_fh, open);
2513 if (status) {
2283963f 2514 release_open_stateid(stp);
1da177e4
LT		 2515 goto out;
2516 }
2517 }
2518 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2519
2520 /*
2521 * Attempt to hand out a delegation. No error return, because the
2522 * OPEN succeeds even if we fail.
2523 */
2524 nfs4_open_delegation(current_fh, open, stp);
2525
2526 status = nfs_ok;
2527
2528 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2529 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2530 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2531out:
13cd2184
N		 2532 if (fp)
2533 put_nfs4_file(fp);
37515177
N		 2534 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2535 nfs4_set_claim_prev(open);
1da177e4
LT		 2536 /*
2537 * To finish the open response, we just need to set the rflags.
2538 */
2539 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2540 if (!open->op_stateowner->so_confirmed)
2541 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2542
2543 return status;
2544}
2545
b37ad28b 2546__be32
b591480b
BF		 2547nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2548 clientid_t *clid)
1da177e4
LT		 2549{
2550 struct nfs4_client *clp;
b37ad28b 2551 __be32 status;
1da177e4
LT		 2552
2553 nfs4_lock_state();
2554 dprintk("process_renew(%08x/%08x): starting\n",
2555 clid->cl_boot, clid->cl_id);
2556 status = nfserr_stale_clientid;
2557 if (STALE_CLIENTID(clid))
2558 goto out;
2559 clp = find_confirmed_client(clid);
2560 status = nfserr_expired;
2561 if (clp == NULL) {
2562 /* We assume the client took too long to RENEW. */
2563 dprintk("nfsd4_renew: clientid not found!\n");
2564 goto out;
2565 }
2566 renew_client(clp);
2567 status = nfserr_cb_path_down;
ea1da636 2568 if (!list_empty(&clp->cl_delegations)
1da177e4
LT		 2569 && !atomic_read(&clp->cl_callback.cb_set))
2570 goto out;
2571 status = nfs_ok;
2572out:
2573 nfs4_unlock_state();
2574 return status;
2575}
2576
af558e33
BF		 2577struct lock_manager nfsd4_manager = {
2578};
2579
a76b4319 2580static void
af558e33 2581nfsd4_end_grace(void)
a76b4319
N		 2582{
2583 dprintk("NFSD: end of grace period\n");
c7b9a459 2584 nfsd4_recdir_purge_old();
af558e33 2585 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2586}
2587
fd39ca9a 2588static time_t
1da177e4
LT		 2589nfs4_laundromat(void)
2590{
2591 struct nfs4_client *clp;
2592 struct nfs4_stateowner *sop;
2593 struct nfs4_delegation *dp;
2594 struct list_head *pos, *next, reaplist;
2595 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2596 time_t t, clientid_val = NFSD_LEASE_TIME;
2597 time_t u, test_val = NFSD_LEASE_TIME;
2598
2599 nfs4_lock_state();
2600
2601 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2602 if (locks_in_grace())
2603 nfsd4_end_grace();
1da177e4
LT		 2604 list_for_each_safe(pos, next, &client_lru) {
2605 clp = list_entry(pos, struct nfs4_client, cl_lru);
2606 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2607 t = clp->cl_time - cutoff;
2608 if (clientid_val > t)
2609 clientid_val = t;
2610 break;
2611 }
2612 dprintk("NFSD: purging unused client (clientid %08x)\n",
2613 clp->cl_clientid.cl_id);
c7b9a459 2614 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2615 expire_client(clp);
2616 }
2617 INIT_LIST_HEAD(&reaplist);
2618 spin_lock(&recall_lock);
2619 list_for_each_safe(pos, next, &del_recall_lru) {
2620 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2621 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2622 u = dp->dl_time - cutoff;
2623 if (test_val > u)
2624 test_val = u;
2625 break;
2626 }
2627 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2628 dp, dp->dl_flock);
2629 list_move(&dp->dl_recall_lru, &reaplist);
2630 }
2631 spin_unlock(&recall_lock);
2632 list_for_each_safe(pos, next, &reaplist) {
2633 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2634 list_del_init(&dp->dl_recall_lru);
2635 unhash_delegation(dp);
2636 }
2637 test_val = NFSD_LEASE_TIME;
2638 list_for_each_safe(pos, next, &close_lru) {
2639 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2640 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2641 u = sop->so_time - cutoff;
2642 if (test_val > u)
2643 test_val = u;
2644 break;
2645 }
2646 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2647 sop->so_id);
f044ff83 2648 release_openowner(sop);
1da177e4
LT		 2649 }
2650 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2651 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2652 nfs4_unlock_state();
2653 return clientid_val;
2654}
2655
a254b246
HH		 2656static struct workqueue_struct *laundry_wq;
2657static void laundromat_main(struct work_struct *);
2658static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2659
2660static void
c4028958 2661laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2662{
2663 time_t t;
2664
2665 t = nfs4_laundromat();
2666 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2667 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2668}
2669
fd39ca9a 2670static struct nfs4_stateowner *
f8816512
N		 2671search_close_lru(u32 st_id, int flags)
2672{
1da177e4
LT		 2673 struct nfs4_stateowner *local = NULL;
2674
1da177e4
LT		 2675 if (flags & CLOSE_STATE) {
2676 list_for_each_entry(local, &close_lru, so_close_lru) {
2677 if (local->so_id == st_id)
2678 return local;
2679 }
2680 }
2681 return NULL;
2682}
2683
2684static inline int
2685nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2686{
7eaa36e2 2687 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2688}
2689
2690static int
2691STALE_STATEID(stateid_t *stateid)
2692{
2693 if (stateid->si_boot == boot_time)
2694 return 0;
849823c5 2695 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1da177e4
LT		 2696 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2697 stateid->si_generation);
2698 return 1;
2699}
2700
2701static inline int
2702access_permit_read(unsigned long access_bmap)
2703{
2704 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2705 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2706 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2707}
2708
2709static inline int
2710access_permit_write(unsigned long access_bmap)
2711{
2712 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2713 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2714}
2715
2716static
b37ad28b 2717__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2718{
b37ad28b 2719 __be32 status = nfserr_openmode;
1da177e4
LT		 2720
2721 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2722 goto out;
2723 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2724 goto out;
2725 status = nfs_ok;
2726out:
2727 return status;
2728}
2729
b37ad28b 2730static inline __be32
1da177e4
LT		 2731check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2732{
203a8c8e 2733 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2734 return nfs_ok;
af558e33 2735 else if (locks_in_grace()) {
1da177e4
LT		 2736 /* Answer in remaining cases depends on existance of
2737 * conflicting state; so we must wait out the grace period. */
2738 return nfserr_grace;
2739 } else if (flags & WR_STATE)
2740 return nfs4_share_conflict(current_fh,
2741 NFS4_SHARE_DENY_WRITE);
2742 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2743 return nfs4_share_conflict(current_fh,
2744 NFS4_SHARE_DENY_READ);
2745}
2746
2747/*
2748 * Allow READ/WRITE during grace period on recovered state only for files
2749 * that are not able to provide mandatory locking.
2750 */
2751static inline int
18f82731 2752grace_disallows_io(struct inode *inode)
1da177e4 2753{
203a8c8e 2754 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2755}
2756
0836f587
BF		 2757static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2758{
2759 /* If the client sends us a stateid from the future, it's buggy: */
2760 if (in->si_generation > ref->si_generation)
2761 return nfserr_bad_stateid;
2762 /*
2763 * The following, however, can happen. For example, if the
2764 * client sends an open and some IO at the same time, the open
2765 * may bump si_generation while the IO is still in flight.
2766 * Thanks to hard links and renames, the client never knows what
2767 * file an open will affect. So it could avoid that situation
2768 * only by serializing all opens and IO from the same open
2769 * owner. To recover from the old_stateid error, the client
2770 * will just have to retry the IO:
2771 */
2772 if (in->si_generation < ref->si_generation)
2773 return nfserr_old_stateid;
2774 return nfs_ok;
2775}
2776
3e633079
BF		 2777static int is_delegation_stateid(stateid_t *stateid)
2778{
2779 return stateid->si_fileid == 0;
2780}
2781
1da177e4
LT		 2782/*
2783* Checks for stateid operations
2784*/
b37ad28b 2785__be32
1da177e4
LT		 2786nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2787{
2788 struct nfs4_stateid *stp = NULL;
2789 struct nfs4_delegation *dp = NULL;
1da177e4 2790 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2791 __be32 status;
1da177e4 2792
1da177e4
LT		 2793 if (filpp)
2794 *filpp = NULL;
2795
18f82731 2796 if (grace_disallows_io(ino))
1da177e4
LT		 2797 return nfserr_grace;
2798
2799 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2800 return check_special_stateids(current_fh, stateid, flags);
2801
1da177e4
LT		 2802 status = nfserr_stale_stateid;
2803 if (STALE_STATEID(stateid))
2804 goto out;
2805
1da177e4 2806 status = nfserr_bad_stateid;
3e633079 2807 if (is_delegation_stateid(stateid)) {
a4455be0 2808 dp = find_delegation_stateid(ino, stateid);
819a8f53 2809 if (!dp)
1da177e4 2810 goto out;
fd03b099 2811 status = check_stateid_generation(stateid, &dp->dl_stateid);
0c2a498f
BF		 2812 if (status)
2813 goto out;
dc9bf700
BF		 2814 status = nfs4_check_delegmode(dp, flags);
2815 if (status)
2816 goto out;
2817 renew_client(dp->dl_client);
dc9bf700
BF		 2818 if (filpp)
2819 *filpp = dp->dl_vfs_file;
1da177e4 2820 } else { /* open or lock stateid */
a4455be0 2821 stp = find_stateid(stateid, flags);
819a8f53 2822 if (!stp)
1da177e4 2823 goto out;
6150ef0d 2824 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2825 goto out;
2826 if (!stp->st_stateowner->so_confirmed)
2827 goto out;
fd03b099 2828 status = check_stateid_generation(stateid, &stp->st_stateid);
0c2a498f
BF		 2829 if (status)
2830 goto out;
a4455be0
BF		 2831 status = nfs4_check_openmode(stp, flags);
2832 if (status)
1da177e4
LT		 2833 goto out;
2834 renew_client(stp->st_stateowner->so_client);
2835 if (filpp)
2836 *filpp = stp->st_vfs_file;
1da177e4
LT		 2837 }
2838 status = nfs_ok;
2839out:
2840 return status;
2841}
2842
4c4cd222
N		 2843static inline int
2844setlkflg (int type)
2845{
2846 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2847 RD_STATE : WR_STATE;
2848}
1da177e4
LT		 2849
2850/*
2851 * Checks for sequence id mutating operations.
2852 */
b37ad28b 2853static __be32
4c4cd222 2854nfs4_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)
1da177e4 2855{
1da177e4
LT		 2856 struct nfs4_stateid *stp;
2857 struct nfs4_stateowner *sop;
0836f587 2858 __be32 status;
1da177e4
LT		 2859
2860 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2861 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2862 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2863 stateid->si_generation);
3a4f98bb 2864
1da177e4
LT		 2865 *stpp = NULL;
2866 *sopp = NULL;
2867
1da177e4 2868 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2869 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2870 return nfserr_bad_stateid;
1da177e4
LT		 2871 }
2872
1da177e4 2873 if (STALE_STATEID(stateid))
3a4f98bb 2874 return nfserr_stale_stateid;
1da177e4
LT		 2875 /*
2876 * We return BAD_STATEID if filehandle doesn't match stateid,
2877 * the confirmed flag is incorrecly set, or the generation
2878 * number is incorrect.
1da177e4 2879 */
f8816512
N		 2880 stp = find_stateid(stateid, flags);
2881 if (stp == NULL) {
2882 /*
2883 * Also, we should make sure this isn't just the result of
2884 * a replayed close:
2885 */
2886 sop = search_close_lru(stateid->si_stateownerid, flags);
2887 if (sop == NULL)
2888 return nfserr_bad_stateid;
2889 *sopp = sop;
2890 goto check_replay;
2891 }
1da177e4 2892
39325bd0
BF		 2893 *stpp = stp;
2894 *sopp = sop = stp->st_stateowner;
2895
4c4cd222 2896 if (lock) {
4c4cd222 2897 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2898 struct nfs4_client *clp = sop->so_client;
4c4cd222 2899 int lkflg = 0;
b37ad28b 2900 __be32 status;
4c4cd222
N		 2901
2902 lkflg = setlkflg(lock->lk_type);
2903
2904 if (lock->lk_is_new) {
599e0a22
BF		 2905 if (!sop->so_is_open_owner)
2906 return nfserr_bad_stateid;
2907 if (!same_clid(&clp->cl_clientid, lockclid))
4c4cd222 2908 return nfserr_bad_stateid;
599e0a22
BF		 2909 /* stp is the open stateid */
2910 status = nfs4_check_openmode(stp, lkflg);
2911 if (status)
2912 return status;
2913 } else {
2914 /* stp is the lock stateid */
2915 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2916 if (status)
2917 return status;
4c4cd222 2918 }
1da177e4
LT		 2919 }
2920
f3362737 2921 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2922 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2923 return nfserr_bad_stateid;
1da177e4
LT		 2924 }
2925
1da177e4
LT		 2926 /*
2927 * We now validate the seqid and stateid generation numbers.
2928 * For the moment, we ignore the possibility of
2929 * generation number wraparound.
2930 */
bd9aac52 2931 if (seqid != sop->so_seqid)
1da177e4
LT		 2932 goto check_replay;
2933
3a4f98bb 2934 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2935 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2936 " unconfirmed stateowner!\n");
2937 return nfserr_bad_stateid;
1da177e4 2938 }
3a4f98bb 2939 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2940 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2941 " confirmed yet!\n");
2942 return nfserr_bad_stateid;
1da177e4 2943 }
0836f587
BF		 2944 status = check_stateid_generation(stateid, &stp->st_stateid);
2945 if (status)
2946 return status;
52fd004e 2947 renew_client(sop->so_client);
3a4f98bb 2948 return nfs_ok;
1da177e4 2949
1da177e4 2950check_replay:
bd9aac52 2951 if (seqid == sop->so_seqid - 1) {
849823c5 2952 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2953 /* indicate replay to calling function */
a90b061c 2954 return nfserr_replay_me;
1da177e4 2955 }
2fdada03 2956 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2957 sop->so_seqid, seqid);
2958 *sopp = NULL;
2959 return nfserr_bad_seqid;
1da177e4
LT		 2960}
2961
b37ad28b 2962__be32
ca364317 2963nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2964 struct nfsd4_open_confirm *oc)
1da177e4 2965{
b37ad28b 2966 __be32 status;
1da177e4
LT		 2967 struct nfs4_stateowner *sop;
2968 struct nfs4_stateid *stp;
2969
2970 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 2971 (int)cstate->current_fh.fh_dentry->d_name.len,
2972 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2973
ca364317 2974 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 2975 if (status)
2976 return status;
1da177e4
LT		 2977
2978 nfs4_lock_state();
2979
ca364317
BF		 2980 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2981 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 2982 CONFIRM | OPEN_STATE,
1da177e4
LT		 2983 &oc->oc_stateowner, &stp, NULL)))
2984 goto out;
2985
2986 sop = oc->oc_stateowner;
2987 sop->so_confirmed = 1;
2988 update_stateid(&stp->st_stateid);
2989 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2990 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2991 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2992 stp->st_stateid.si_boot,
2993 stp->st_stateid.si_stateownerid,
2994 stp->st_stateid.si_fileid,
2995 stp->st_stateid.si_generation);
c7b9a459
N		 2996
2997 nfsd4_create_clid_dir(sop->so_client);
1da177e4 2998out:
f2327d9a 2999 if (oc->oc_stateowner) {
1da177e4 3000 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3001 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3002 }
1da177e4
LT		 3003 nfs4_unlock_state();
3004 return status;
3005}
3006
3007
3008/*
3009 * unset all bits in union bitmap (bmap) that
3010 * do not exist in share (from successful OPEN_DOWNGRADE)
3011 */
3012static void
3013reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3014{
3015 int i;
3016 for (i = 1; i < 4; i++) {
3017 if ((i & access) != i)
3018 __clear_bit(i, bmap);
3019 }
3020}
3021
3022static void
3023reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3024{
3025 int i;
3026 for (i = 0; i < 4; i++) {
3027 if ((i & deny) != i)
3028 __clear_bit(i, bmap);
3029 }
3030}
3031
b37ad28b 3032__be32
ca364317
BF		 3033nfsd4_open_downgrade(struct svc_rqst *rqstp,
3034 struct nfsd4_compound_state *cstate,
a4f1706a 3035 struct nfsd4_open_downgrade *od)
1da177e4 3036{
b37ad28b 3037 __be32 status;
1da177e4
LT		 3038 struct nfs4_stateid *stp;
3039 unsigned int share_access;
3040
3041 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3042 (int)cstate->current_fh.fh_dentry->d_name.len,
3043 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3044
ba5a6a19
BF		 3045 if (!access_valid(od->od_share_access)
3046 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3047 return nfserr_inval;
3048
3049 nfs4_lock_state();
ca364317
BF		 3050 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3051 od->od_seqid,
1da177e4 3052 &od->od_stateid,
f3362737 3053 OPEN_STATE,
1da177e4
LT		 3054 &od->od_stateowner, &stp, NULL)))
3055 goto out;
3056
3057 status = nfserr_inval;
3058 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3059 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3060 stp->st_access_bmap, od->od_share_access);
3061 goto out;
3062 }
3063 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3064 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3065 stp->st_deny_bmap, od->od_share_deny);
3066 goto out;
3067 }
3068 set_access(&share_access, stp->st_access_bmap);
3069 nfs4_file_downgrade(stp->st_vfs_file,
3070 share_access & ~od->od_share_access);
3071
3072 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3073 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3074
3075 update_stateid(&stp->st_stateid);
3076 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3077 status = nfs_ok;
3078out:
f2327d9a 3079 if (od->od_stateowner) {
1da177e4 3080 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3081 cstate->replay_owner = od->od_stateowner;
f2327d9a 3082 }
1da177e4
LT		 3083 nfs4_unlock_state();
3084 return status;
3085}
3086
3087/*
3088 * nfs4_unlock_state() called after encode
3089 */
b37ad28b 3090__be32
ca364317 3091nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3092 struct nfsd4_close *close)
1da177e4 3093{
b37ad28b 3094 __be32 status;
1da177e4
LT		 3095 struct nfs4_stateid *stp;
3096
3097 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3098 (int)cstate->current_fh.fh_dentry->d_name.len,
3099 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3100
3101 nfs4_lock_state();
3102 /* check close_lru for replay */
ca364317
BF		 3103 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3104 close->cl_seqid,
1da177e4 3105 &close->cl_stateid,
f3362737 3106 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3107 &close->cl_stateowner, &stp, NULL)))
3108 goto out;
1da177e4
LT		 3109 status = nfs_ok;
3110 update_stateid(&stp->st_stateid);
3111 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3112
04ef5954 3113 /* release_stateid() calls nfsd_close() if needed */
2283963f 3114 release_open_stateid(stp);
04ef5954
BF		 3115
3116 /* place unused nfs4_stateowners on so_close_lru list to be
3117 * released by the laundromat service after the lease period
3118 * to enable us to handle CLOSE replay
3119 */
3120 if (list_empty(&close->cl_stateowner->so_stateids))
3121 move_to_close_lru(close->cl_stateowner);
1da177e4 3122out:
f2327d9a 3123 if (close->cl_stateowner) {
1da177e4 3124 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3125 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3126 }
1da177e4
LT		 3127 nfs4_unlock_state();
3128 return status;
3129}
3130
b37ad28b 3131__be32
ca364317
BF		 3132nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3133 struct nfsd4_delegreturn *dr)
1da177e4 3134{
203a8c8e
BF		 3135 struct nfs4_delegation *dp;
3136 stateid_t *stateid = &dr->dr_stateid;
3137 struct inode *inode;
b37ad28b 3138 __be32 status;
1da177e4 3139
ca364317 3140 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3141 return status;
3142 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3143
3144 nfs4_lock_state();
203a8c8e
BF		 3145 status = nfserr_bad_stateid;
3146 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3147 goto out;
3148 status = nfserr_stale_stateid;
3149 if (STALE_STATEID(stateid))
3150 goto out;
7e0f7cf5 3151 status = nfserr_bad_stateid;
203a8c8e
BF		 3152 if (!is_delegation_stateid(stateid))
3153 goto out;
203a8c8e
BF		 3154 dp = find_delegation_stateid(inode, stateid);
3155 if (!dp)
3156 goto out;
3157 status = check_stateid_generation(stateid, &dp->dl_stateid);
3158 if (status)
3159 goto out;
3160 renew_client(dp->dl_client);
3161
3162 unhash_delegation(dp);
1da177e4 3163out:
203a8c8e
BF		 3164 nfs4_unlock_state();
3165
1da177e4
LT		 3166 return status;
3167}
3168
3169
3170/*
3171 * Lock owner state (byte-range locks)
3172 */
3173#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3174#define LOCK_HASH_BITS 8
3175#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3176#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3177
87df4de8
BH		 3178static inline u64
3179end_offset(u64 start, u64 len)
3180{
3181 u64 end;
3182
3183 end = start + len;
3184 return end >= start ? end: NFS4_MAX_UINT64;
3185}
3186
3187/* last octet in a range */
3188static inline u64
3189last_byte_offset(u64 start, u64 len)
3190{
3191 u64 end;
3192
3193 BUG_ON(!len);
3194 end = start + len;
3195 return end > start ? end - 1: NFS4_MAX_UINT64;
3196}
3197
1da177e4
LT		 3198#define lockownerid_hashval(id) \
3199 ((id) & LOCK_HASH_MASK)
3200
3201static inline unsigned int
3202lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3203 struct xdr_netobj *ownername)
3204{
3205 return (file_hashval(inode) + cl_id
3206 + opaque_hashval(ownername->data, ownername->len))
3207 & LOCK_HASH_MASK;
3208}
3209
3210static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3211static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3212static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3213
fd39ca9a 3214static struct nfs4_stateid *
1da177e4
LT		 3215find_stateid(stateid_t *stid, int flags)
3216{
9346eff0 3217 struct nfs4_stateid *local;
1da177e4
LT		 3218 u32 st_id = stid->si_stateownerid;
3219 u32 f_id = stid->si_fileid;
3220 unsigned int hashval;
3221
3222 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3223 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3224 hashval = stateid_hashval(st_id, f_id);
3225 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3226 if ((local->st_stateid.si_stateownerid == st_id) &&
3227 (local->st_stateid.si_fileid == f_id))
3228 return local;
3229 }
3230 }
9346eff0
KK		 3231
3232 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3233 hashval = stateid_hashval(st_id, f_id);
3234 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3235 if ((local->st_stateid.si_stateownerid == st_id) &&
3236 (local->st_stateid.si_fileid == f_id))
3237 return local;
3238 }
849823c5 3239 }
1da177e4
LT		 3240 return NULL;
3241}
3242
3243static struct nfs4_delegation *
3244find_delegation_stateid(struct inode *ino, stateid_t *stid)
3245{
13cd2184
N		 3246 struct nfs4_file *fp;
3247 struct nfs4_delegation *dl;
1da177e4
LT		 3248
3249 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3250 stid->si_boot, stid->si_stateownerid,
3251 stid->si_fileid, stid->si_generation);
3252
1da177e4 3253 fp = find_file(ino);
13cd2184
N		 3254 if (!fp)
3255 return NULL;
3256 dl = find_delegation_file(fp, stid);
3257 put_nfs4_file(fp);
3258 return dl;
1da177e4
LT		 3259}
3260
3261/*
3262 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3263 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3264 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3265 * locking, this prevents us from being completely protocol-compliant. The
3266 * real solution to this problem is to start using unsigned file offsets in
3267 * the VFS, but this is a very deep change!
3268 */
3269static inline void
3270nfs4_transform_lock_offset(struct file_lock *lock)
3271{
3272 if (lock->fl_start < 0)
3273 lock->fl_start = OFFSET_MAX;
3274 if (lock->fl_end < 0)
3275 lock->fl_end = OFFSET_MAX;
3276}
3277
d5b9026a
N		 3278/* Hack!: For now, we're defining this just so we can use a pointer to it
3279 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3280static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3281};
1da177e4
LT		 3282
3283static inline void
3284nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3285{
d5b9026a
N		 3286 struct nfs4_stateowner *sop;
3287 unsigned int hval;
1da177e4 3288
d5b9026a
N		 3289 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3290 sop = (struct nfs4_stateowner *) fl->fl_owner;
3291 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3292 kref_get(&sop->so_ref);
3293 deny->ld_sop = sop;
3294 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3295 } else {
3296 deny->ld_sop = NULL;
3297 deny->ld_clientid.cl_boot = 0;
3298 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3299 }
3300 deny->ld_start = fl->fl_start;
87df4de8
BH		 3301 deny->ld_length = NFS4_MAX_UINT64;
3302 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3303 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3304 deny->ld_type = NFS4_READ_LT;
3305 if (fl->fl_type != F_RDLCK)
3306 deny->ld_type = NFS4_WRITE_LT;
3307}
3308
1da177e4
LT		 3309static struct nfs4_stateowner *
3310find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3311 struct xdr_netobj *owner)
3312{
3313 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3314 struct nfs4_stateowner *op;
3315
3316 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3317 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3318 return op;
3319 }
3320 return NULL;
3321}
3322
3323/*
3324 * Alloc a lock owner structure.
3325 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3326 * occured.
3327 *
3328 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3329 */
3330
3331static struct nfs4_stateowner *
3332alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3333 struct nfs4_stateowner *sop;
3334 struct nfs4_replay *rp;
3335 unsigned int idhashval;
3336
3337 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3338 return NULL;
3339 idhashval = lockownerid_hashval(current_ownerid);
3340 INIT_LIST_HEAD(&sop->so_idhash);
3341 INIT_LIST_HEAD(&sop->so_strhash);
3342 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3343 INIT_LIST_HEAD(&sop->so_stateids);
3344 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3345 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3346 sop->so_time = 0;
3347 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3348 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3349 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3350 sop->so_is_open_owner = 0;
3351 sop->so_id = current_ownerid++;
3352 sop->so_client = clp;
b59e3c0e
NB		 3353 /* It is the openowner seqid that will be incremented in encode in the
3354 * case of new lockowners; so increment the lock seqid manually: */
3355 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3356 sop->so_confirmed = 1;
3357 rp = &sop->so_replay;
de1ae286 3358 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3359 rp->rp_buflen = 0;
3360 rp->rp_buf = rp->rp_ibuf;
3361 return sop;
3362}
3363
fd39ca9a 3364static struct nfs4_stateid *
1da177e4
LT		 3365alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3366{
3367 struct nfs4_stateid *stp;
3368 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3369
5ac049ac
N		 3370 stp = nfs4_alloc_stateid();
3371 if (stp == NULL)
1da177e4
LT		 3372 goto out;
3373 INIT_LIST_HEAD(&stp->st_hash);
3374 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3375 INIT_LIST_HEAD(&stp->st_perstateowner);
3376 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3377 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3378 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3379 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3380 stp->st_stateowner = sop;
13cd2184 3381 get_nfs4_file(fp);
1da177e4
LT		 3382 stp->st_file = fp;
3383 stp->st_stateid.si_boot = boot_time;
3384 stp->st_stateid.si_stateownerid = sop->so_id;
3385 stp->st_stateid.si_fileid = fp->fi_id;
3386 stp->st_stateid.si_generation = 0;
3387 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3388 stp->st_access_bmap = open_stp->st_access_bmap;
3389 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3390 stp->st_openstp = open_stp;
1da177e4
LT		 3391
3392out:
3393 return stp;
3394}
3395
fd39ca9a 3396static int
1da177e4
LT		 3397check_lock_length(u64 offset, u64 length)
3398{
87df4de8 3399 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3400 LOFF_OVERFLOW(offset, length)));
3401}
3402
3403/*
3404 * LOCK operation
3405 */
b37ad28b 3406__be32
ca364317 3407nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3408 struct nfsd4_lock *lock)
1da177e4 3409{
3e9e3dbe 3410 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3411 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3412 struct nfs4_stateid *lock_stp;
3413 struct file *filp;
3414 struct file_lock file_lock;
8dc7c311 3415 struct file_lock conflock;
b37ad28b 3416 __be32 status = 0;
1da177e4 3417 unsigned int strhashval;
150b3934 3418 unsigned int cmd;
b8dd7b9a 3419 int err;
1da177e4
LT		 3420
3421 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3422 (long long) lock->lk_offset,
3423 (long long) lock->lk_length);
3424
1da177e4
LT		 3425 if (check_lock_length(lock->lk_offset, lock->lk_length))
3426 return nfserr_inval;
3427
ca364317 3428 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3429 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3430 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3431 return status;
3432 }
3433
1da177e4
LT		 3434 nfs4_lock_state();
3435
3436 if (lock->lk_is_new) {
893f8770
N		 3437 /*
3438 * Client indicates that this is a new lockowner.
3439 * Use open owner and open stateid to create lock owner and
3440 * lock stateid.
3441 */
1da177e4
LT		 3442 struct nfs4_stateid *open_stp = NULL;
3443 struct nfs4_file *fp;
3444
3445 status = nfserr_stale_clientid;
849823c5 3446 if (STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3447 goto out;
1da177e4 3448
1da177e4 3449 /* validate and update open stateid and open seqid */
ca364317 3450 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3451 lock->lk_new_open_seqid,
3452 &lock->lk_new_open_stateid,
f3362737 3453 OPEN_STATE,
3a65588a 3454 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3455 lock);
37515177 3456 if (status)
1da177e4 3457 goto out;
3a65588a 3458 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3459 /* create lockowner and lock stateid */
3460 fp = open_stp->st_file;
3461 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3462 open_sop->so_client->cl_clientid.cl_id,
3463 &lock->v.new.owner);
3e9e3dbe
N		 3464 /* XXX: Do we need to check for duplicate stateowners on
3465 * the same file, or should they just be allowed (and
3466 * create new stateids)? */
1da177e4 3467 status = nfserr_resource;
b59e3c0e
NB		 3468 lock_sop = alloc_init_lock_stateowner(strhashval,
3469 open_sop->so_client, open_stp, lock);
3470 if (lock_sop == NULL)
1da177e4 3471 goto out;
b59e3c0e 3472 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3473 if (lock_stp == NULL)
1da177e4 3474 goto out;
1da177e4
LT		 3475 } else {
3476 /* lock (lock owner + lock stateid) already exists */
ca364317 3477 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3478 lock->lk_old_lock_seqid,
3479 &lock->lk_old_lock_stateid,
f3362737 3480 LOCK_STATE,
3a65588a 3481 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3482 if (status)
3483 goto out;
3a65588a 3484 lock_sop = lock->lk_replay_owner;
1da177e4 3485 }
3a65588a 3486 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3487 filp = lock_stp->st_vfs_file;
3488
0dd395dc 3489 status = nfserr_grace;
af558e33 3490 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3491 goto out;
3492 status = nfserr_no_grace;
af558e33 3493 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3494 goto out;
3495
1da177e4
LT		 3496 locks_init_lock(&file_lock);
3497 switch (lock->lk_type) {
3498 case NFS4_READ_LT:
3499 case NFS4_READW_LT:
3500 file_lock.fl_type = F_RDLCK;
150b3934 3501 cmd = F_SETLK;
1da177e4
LT		 3502 break;
3503 case NFS4_WRITE_LT:
3504 case NFS4_WRITEW_LT:
3505 file_lock.fl_type = F_WRLCK;
150b3934 3506 cmd = F_SETLK;
1da177e4
LT		 3507 break;
3508 default:
3509 status = nfserr_inval;
3510 goto out;
3511 }
b59e3c0e 3512 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3513 file_lock.fl_pid = current->tgid;
3514 file_lock.fl_file = filp;
3515 file_lock.fl_flags = FL_POSIX;
d5b9026a 3516 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3517
3518 file_lock.fl_start = lock->lk_offset;
87df4de8 3519 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3520 nfs4_transform_lock_offset(&file_lock);
3521
3522 /*
3523 * Try to lock the file in the VFS.
3524 * Note: locks.c uses the BKL to protect the inode's lock list.
3525 */
3526
fd85b817 3527 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3528 switch (-err) {
1da177e4
LT		 3529 case 0: /* success! */
3530 update_stateid(&lock_stp->st_stateid);
3531 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3532 sizeof(stateid_t));
b8dd7b9a 3533 status = 0;
eb76b3fd
AA		 3534 break;
3535 case (EAGAIN): /* conflock holds conflicting lock */
3536 status = nfserr_denied;
3537 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3538 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3539 break;
1da177e4
LT		 3540 case (EDEADLK):
3541 status = nfserr_deadlock;
eb76b3fd 3542 break;
1da177e4 3543 default:
fd85b817 3544 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3545 status = nfserr_resource;
3546 break;
1da177e4 3547 }
1da177e4 3548out:
8a280510 3549 if (status && lock->lk_is_new && lock_sop)
f044ff83 3550 release_lockowner(lock_sop);
3a65588a
BF		 3551 if (lock->lk_replay_owner) {
3552 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3553 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3554 }
1da177e4
LT		 3555 nfs4_unlock_state();
3556 return status;
3557}
3558
55ef1274
BF		 3559/*
3560 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3561 * so we do a temporary open here just to get an open file to pass to
3562 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3563 * inode operation.)
3564 */
3565static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3566{
3567 struct file *file;
3568 int err;
3569
3570 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3571 if (err)
3572 return err;
3573 err = vfs_test_lock(file, lock);
3574 nfsd_close(file);
3575 return err;
3576}
3577
1da177e4
LT		 3578/*
3579 * LOCKT operation
3580 */
b37ad28b 3581__be32
ca364317
BF		 3582nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3583 struct nfsd4_lockt *lockt)
1da177e4
LT		 3584{
3585 struct inode *inode;
1da177e4 3586 struct file_lock file_lock;
fd85b817 3587 int error;
b37ad28b 3588 __be32 status;
1da177e4 3589
af558e33 3590 if (locks_in_grace())
1da177e4
LT		 3591 return nfserr_grace;
3592
3593 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3594 return nfserr_inval;
3595
3596 lockt->lt_stateowner = NULL;
3597 nfs4_lock_state();
3598
3599 status = nfserr_stale_clientid;
849823c5 3600 if (STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3601 goto out;
1da177e4 3602
ca364317 3603 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3604 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3605 if (status == nfserr_symlink)
3606 status = nfserr_inval;
3607 goto out;
3608 }
3609
ca364317 3610 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3611 locks_init_lock(&file_lock);
3612 switch (lockt->lt_type) {
3613 case NFS4_READ_LT:
3614 case NFS4_READW_LT:
3615 file_lock.fl_type = F_RDLCK;
3616 break;
3617 case NFS4_WRITE_LT:
3618 case NFS4_WRITEW_LT:
3619 file_lock.fl_type = F_WRLCK;
3620 break;
3621 default:
2fdada03 3622 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3623 status = nfserr_inval;
3624 goto out;
3625 }
3626
3627 lockt->lt_stateowner = find_lockstateowner_str(inode,
3628 &lockt->lt_clientid, &lockt->lt_owner);
3629 if (lockt->lt_stateowner)
3630 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3631 file_lock.fl_pid = current->tgid;
3632 file_lock.fl_flags = FL_POSIX;
3633
3634 file_lock.fl_start = lockt->lt_offset;
87df4de8 3635 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3636
3637 nfs4_transform_lock_offset(&file_lock);
3638
1da177e4 3639 status = nfs_ok;
55ef1274 3640 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3641 if (error) {
3642 status = nfserrno(error);
3643 goto out;
3644 }
9d6a8c5c 3645 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3646 status = nfserr_denied;
9d6a8c5c 3647 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3648 }
3649out:
3650 nfs4_unlock_state();
3651 return status;
3652}
3653
b37ad28b 3654__be32
ca364317 3655nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3656 struct nfsd4_locku *locku)
1da177e4
LT		 3657{
3658 struct nfs4_stateid *stp;
3659 struct file *filp = NULL;
3660 struct file_lock file_lock;
b37ad28b 3661 __be32 status;
b8dd7b9a 3662 int err;
1da177e4
LT		 3663
3664 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3665 (long long) locku->lu_offset,
3666 (long long) locku->lu_length);
3667
3668 if (check_lock_length(locku->lu_offset, locku->lu_length))
3669 return nfserr_inval;
3670
3671 nfs4_lock_state();
3672
ca364317 3673 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3674 locku->lu_seqid,
3675 &locku->lu_stateid,
f3362737 3676 LOCK_STATE,
1da177e4
LT		 3677 &locku->lu_stateowner, &stp, NULL)))
3678 goto out;
3679
3680 filp = stp->st_vfs_file;
3681 BUG_ON(!filp);
3682 locks_init_lock(&file_lock);
3683 file_lock.fl_type = F_UNLCK;
3684 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3685 file_lock.fl_pid = current->tgid;
3686 file_lock.fl_file = filp;
3687 file_lock.fl_flags = FL_POSIX;
d5b9026a 3688 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3689 file_lock.fl_start = locku->lu_offset;
3690
87df4de8 3691 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3692 nfs4_transform_lock_offset(&file_lock);
3693
3694 /*
3695 * Try to unlock the file in the VFS.
3696 */
fd85b817 3697 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3698 if (err) {
fd85b817 3699 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3700 goto out_nfserr;
3701 }
3702 /*
3703 * OK, unlock succeeded; the only thing left to do is update the stateid.
3704 */
3705 update_stateid(&stp->st_stateid);
3706 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3707
3708out:
f2327d9a 3709 if (locku->lu_stateowner) {
1da177e4 3710 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3711 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3712 }
1da177e4
LT		 3713 nfs4_unlock_state();
3714 return status;
3715
3716out_nfserr:
b8dd7b9a 3717 status = nfserrno(err);
1da177e4
LT		 3718 goto out;
3719}
3720
3721/*
3722 * returns
3723 * 1: locks held by lockowner
3724 * 0: no locks held by lockowner
3725 */
3726static int
3727check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3728{
3729 struct file_lock **flpp;
7eaa36e2 3730 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3731 int status = 0;
3732
3733 lock_kernel();
3734 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3735 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3736 status = 1;
3737 goto out;
796dadfd 3738 }
1da177e4
LT		 3739 }
3740out:
3741 unlock_kernel();
3742 return status;
3743}
3744
b37ad28b 3745__be32
b591480b
BF		 3746nfsd4_release_lockowner(struct svc_rqst *rqstp,
3747 struct nfsd4_compound_state *cstate,
3748 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3749{
3750 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3751 struct nfs4_stateowner *sop;
3752 struct nfs4_stateid *stp;
1da177e4 3753 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3754 struct list_head matches;
3755 int i;
b37ad28b 3756 __be32 status;
1da177e4
LT		 3757
3758 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3759 clid->cl_boot, clid->cl_id);
3760
3761 /* XXX check for lease expiration */
3762
3763 status = nfserr_stale_clientid;
849823c5 3764 if (STALE_CLIENTID(clid))
1da177e4 3765 return status;
1da177e4
LT		 3766
3767 nfs4_lock_state();
3768
3e9e3dbe
N		 3769 status = nfserr_locks_held;
3770 /* XXX: we're doing a linear search through all the lockowners.
3771 * Yipes! For now we'll just hope clients aren't really using
3772 * release_lockowner much, but eventually we have to fix these
3773 * data structures. */
3774 INIT_LIST_HEAD(&matches);
3775 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3776 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3777 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3778 continue;
3779 list_for_each_entry(stp, &sop->so_stateids,
3780 st_perstateowner) {
3781 if (check_for_locks(stp->st_vfs_file, sop))
3782 goto out;
3783 /* Note: so_perclient unused for lockowners,
3784 * so it's OK to fool with here. */
3785 list_add(&sop->so_perclient, &matches);
3786 }
1da177e4 3787 }
3e9e3dbe
N		 3788 }
3789 /* Clients probably won't expect us to return with some (but not all)
3790 * of the lockowner state released; so don't release any until all
3791 * have been checked. */
3792 status = nfs_ok;
0fa822e4
N		 3793 while (!list_empty(&matches)) {
3794 sop = list_entry(matches.next, struct nfs4_stateowner,
3795 so_perclient);
3796 /* unhash_stateowner deletes so_perclient only
3797 * for openowners. */
3798 list_del(&sop->so_perclient);
f044ff83 3799 release_lockowner(sop);
1da177e4
LT		 3800 }
3801out:
3802 nfs4_unlock_state();
3803 return status;
3804}
3805
3806static inline struct nfs4_client_reclaim *
a55370a3 3807alloc_reclaim(void)
1da177e4 3808{
a55370a3 3809 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3810}
3811
c7b9a459 3812int
a1bcecd2 3813nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3814{
3815 unsigned int strhashval = clientstr_hashval(name);
3816 struct nfs4_client *clp;
3817
a1bcecd2 3818 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3819 return clp ? 1 : 0;
3820}
3821
1da177e4
LT		 3822/*
3823 * failure => all reset bets are off, nfserr_no_grace...
3824 */
190e4fbf
N		 3825int
3826nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3827{
3828 unsigned int strhashval;
3829 struct nfs4_client_reclaim *crp = NULL;
3830
a55370a3
N		 3831 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3832 crp = alloc_reclaim();
1da177e4
LT		 3833 if (!crp)
3834 return 0;
a55370a3 3835 strhashval = clientstr_hashval(name);
1da177e4
LT		 3836 INIT_LIST_HEAD(&crp->cr_strhash);
3837 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3838 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3839 reclaim_str_hashtbl_size++;
3840 return 1;
3841}
3842
3843static void
3844nfs4_release_reclaim(void)
3845{
3846 struct nfs4_client_reclaim *crp = NULL;
3847 int i;
3848
1da177e4
LT		 3849 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3850 while (!list_empty(&reclaim_str_hashtbl[i])) {
3851 crp = list_entry(reclaim_str_hashtbl[i].next,
3852 struct nfs4_client_reclaim, cr_strhash);
3853 list_del(&crp->cr_strhash);
1da177e4
LT		 3854 kfree(crp);
3855 reclaim_str_hashtbl_size--;
3856 }
3857 }
3858 BUG_ON(reclaim_str_hashtbl_size);
3859}
3860
3861/*
3862 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3863static struct nfs4_client_reclaim *
1da177e4
LT		 3864nfs4_find_reclaim_client(clientid_t *clid)
3865{
3866 unsigned int strhashval;
3867 struct nfs4_client *clp;
3868 struct nfs4_client_reclaim *crp = NULL;
3869
3870
3871 /* find clientid in conf_id_hashtbl */
3872 clp = find_confirmed_client(clid);
3873 if (clp == NULL)
3874 return NULL;
3875
a55370a3
N		 3876 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3877 clp->cl_name.len, clp->cl_name.data,
3878 clp->cl_recdir);
1da177e4
LT		 3879
3880 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3881 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3882 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3883 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3884 return crp;
3885 }
3886 }
3887 return NULL;
3888}
3889
3890/*
3891* Called from OPEN. Look for clientid in reclaim list.
3892*/
b37ad28b 3893__be32
1da177e4
LT		 3894nfs4_check_open_reclaim(clientid_t *clid)
3895{
dfc83565 3896 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3897}
3898
ac4d8ff2 3899/* initialization to perform at module load time: */
1da177e4 3900
e8ff2a84 3901int
ac4d8ff2 3902nfs4_state_init(void)
1da177e4 3903{
e8ff2a84 3904 int i, status;
1da177e4 3905
e8ff2a84
BF		 3906 status = nfsd4_init_slabs();
3907 if (status)
3908 return status;
1da177e4
LT		 3909 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3910 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3911 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3912 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3913 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3914 }
5282fd72
ME		 3915 for (i = 0; i < SESSION_HASH_SIZE; i++)
3916 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3917 for (i = 0; i < FILE_HASH_SIZE; i++) {
3918 INIT_LIST_HEAD(&file_hashtbl[i]);
3919 }
3920 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3921 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3922 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3923 }
3924 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3925 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3926 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3927 }
3928 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3929 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3930 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3931 }
1da177e4 3932 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3933 INIT_LIST_HEAD(&close_lru);
3934 INIT_LIST_HEAD(&client_lru);
3935 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N		 3936 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3937 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3938 reclaim_str_hashtbl_size = 0;
e8ff2a84 3939 return 0;
ac4d8ff2
N		 3940}
3941
190e4fbf
N		 3942static void
3943nfsd4_load_reboot_recovery_data(void)
3944{
3945 int status;
3946
0964a3d3
N		 3947 nfs4_lock_state();
3948 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3949 status = nfsd4_recdir_load();
0964a3d3 3950 nfs4_unlock_state();
190e4fbf
N		 3951 if (status)
3952 printk("NFSD: Failure reading reboot recovery data\n");
3953}
3954
9a8db97e
ME		 3955unsigned long
3956get_nfs4_grace_period(void)
3957{
3958 return max(user_lease_time, lease_time) * HZ;
3959}
3960
c2f1a551
MS		 3961/*
3962 * Since the lifetime of a delegation isn't limited to that of an open, a
3963 * client may quite reasonably hang on to a delegation as long as it has
3964 * the inode cached. This becomes an obvious problem the first time a
3965 * client's inode cache approaches the size of the server's total memory.
3966 *
3967 * For now we avoid this problem by imposing a hard limit on the number
3968 * of delegations, which varies according to the server's memory size.
3969 */
3970static void
3971set_max_delegations(void)
3972{
3973 /*
3974 * Allow at most 4 delegations per megabyte of RAM. Quick
3975 * estimates suggest that in the worst case (where every delegation
3976 * is for a different inode), a delegation could take about 1.5K,
3977 * giving a worst case usage of about 6% of memory.
3978 */
3979 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3980}
3981
ac4d8ff2
N		 3982/* initialization to perform when the nfsd service is started: */
3983
3984static void
3985__nfs4_state_start(void)
3986{
9a8db97e 3987 unsigned long grace_time;
ac4d8ff2 3988
1da177e4 3989 boot_time = get_seconds();
af558e33 3990 grace_time = get_nfs4_grace_period();
d99a05ad 3991 lease_time = user_lease_time;
af558e33 3992 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 3993 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3994 grace_time/HZ);
58da282b 3995 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 3996 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 3997 set_max_delegations();
1da177e4
LT		 3998}
3999
e8ff2a84 4000void
76a3550e 4001nfs4_state_start(void)
1da177e4 4002{
1da177e4 4003 if (nfs4_init)
e8ff2a84 4004 return;
190e4fbf 4005 nfsd4_load_reboot_recovery_data();
76a3550e 4006 __nfs4_state_start();
1da177e4 4007 nfs4_init = 1;
e8ff2a84 4008 return;
1da177e4
LT		 4009}
4010
1da177e4
LT		 4011time_t
4012nfs4_lease_time(void)
4013{
4014 return lease_time;
4015}
4016
4017static void
4018__nfs4_state_shutdown(void)
4019{
4020 int i;
4021 struct nfs4_client *clp = NULL;
4022 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4023 struct list_head *pos, *next, reaplist;
4024
1da177e4
LT		 4025 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4026 while (!list_empty(&conf_id_hashtbl[i])) {
4027 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4028 expire_client(clp);
4029 }
4030 while (!list_empty(&unconf_str_hashtbl[i])) {
4031 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4032 expire_client(clp);
4033 }
4034 }
4035 INIT_LIST_HEAD(&reaplist);
4036 spin_lock(&recall_lock);
4037 list_for_each_safe(pos, next, &del_recall_lru) {
4038 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4039 list_move(&dp->dl_recall_lru, &reaplist);
4040 }
4041 spin_unlock(&recall_lock);
4042 list_for_each_safe(pos, next, &reaplist) {
4043 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4044 list_del_init(&dp->dl_recall_lru);
4045 unhash_delegation(dp);
4046 }
4047
190e4fbf 4048 nfsd4_shutdown_recdir();
1da177e4 4049 nfs4_init = 0;
1da177e4
LT		 4050}
4051
4052void
4053nfs4_state_shutdown(void)
4054{
5e8d5c29
N		 4055 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4056 destroy_workqueue(laundry_wq);
2c5e7615 4057 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4058 nfs4_lock_state();
4059 nfs4_release_reclaim();
4060 __nfs4_state_shutdown();
1da177e4
LT		 4061 nfs4_unlock_state();
4062}
4063
3dd98a3b
JL		 4064/*
4065 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4066 * accessed when nfsd is starting.
4067 */
0964a3d3
N		 4068static void
4069nfs4_set_recdir(char *recdir)
4070{
0964a3d3 4071 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4072}
4073
4074/*
4075 * Change the NFSv4 recovery directory to recdir.
4076 */
4077int
4078nfs4_reset_recoverydir(char *recdir)
4079{
4080 int status;
a63bb996 4081 struct path path;
0964a3d3 4082
a63bb996 4083 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4084 if (status)
4085 return status;
4086 status = -ENOTDIR;
a63bb996 4087 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4088 nfs4_set_recdir(recdir);
4089 status = 0;
4090 }
a63bb996 4091 path_put(&path);
0964a3d3
N		 4092 return status;
4093}
4094
3dd98a3b
JL		 4095char *
4096nfs4_recoverydir(void)
4097{
4098 return user_recovery_dirname;
4099}
4100
1da177e4
LT		 4101/*
4102 * Called when leasetime is changed.
4103 *
d99a05ad
N		 4104 * The only way the protocol gives us to handle on-the-fly lease changes is to
4105 * simulate a reboot. Instead of doing that, we just wait till the next time
4106 * we start to register any changes in lease time. If the administrator
4107 * really wants to change the lease time *now*, they can go ahead and bring
4108 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4109 *
4110 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4111 * when nfsd is starting
1da177e4
LT		 4112 */
4113void
4114nfs4_reset_lease(time_t leasetime)
4115{
d99a05ad 4116 user_lease_time = leasetime;
1da177e4 4117}
Linux 6.x block layer and io_uring tree(s)