projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd41: destroy_session operation
[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{
e10e0cfc
BH		 1425 struct nfsd4_session *ses;
1426 u32 status = nfserr_badsession;
1427
1428 /* Notes:
1429 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1430 * - Should we return nfserr_back_chan_busy if waiting for
1431 * callbacks on to-be-destroyed session?
1432 * - Do we need to clear any callback info from previous session?
1433 */
1434
1435 dump_sessionid(__func__, &sessionid->sessionid);
1436 spin_lock(&sessionid_lock);
1437 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1438 if (!ses) {
1439 spin_unlock(&sessionid_lock);
1440 goto out;
1441 }
1442
1443 unhash_session(ses);
1444 spin_unlock(&sessionid_lock);
1445
1446 /* wait for callbacks */
1447 shutdown_callback_client(ses->se_client);
1448 nfsd4_put_session(ses);
1449 status = nfs_ok;
1450out:
1451 dprintk("%s returns %d\n", __func__, ntohl(status));
1452 return status;
069b6ad4
AA		 1453}
1454
1455__be32
b85d4c01 1456nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1457 struct nfsd4_compound_state *cstate,
1458 struct nfsd4_sequence *seq)
1459{
f9bb94c4 1460 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1461 struct nfsd4_session *session;
1462 struct nfsd4_slot *slot;
1463 int status;
1464
f9bb94c4
AA		 1465 if (resp->opcnt != 1)
1466 return nfserr_sequence_pos;
1467
b85d4c01
BH		 1468 spin_lock(&sessionid_lock);
1469 status = nfserr_badsession;
1470 session = find_in_sessionid_hashtbl(&seq->sessionid);
1471 if (!session)
1472 goto out;
1473
1474 status = nfserr_badslot;
1475 if (seq->slotid >= session->se_fnumslots)
1476 goto out;
1477
1478 slot = &session->se_slots[seq->slotid];
1479 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1480
1481 status = check_slot_seqid(seq->seqid, slot);
1482 if (status == nfserr_replay_cache) {
1483 cstate->slot = slot;
1484 cstate->session = session;
da3846a2 1485 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1486 * for nfsd4_svc_encode_compoundres processing */
1487 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1488 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1489 goto replay_cache;
1490 }
1491 if (status)
1492 goto out;
1493
1494 /* Success! bump slot seqid */
1495 slot->sl_inuse = true;
1496 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1497 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1498 /* Always set the cache entry cachethis for solo sequence */
1499 if (nfsd4_is_solo_sequence(resp))
1500 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1501
1502 cstate->slot = slot;
1503 cstate->session = session;
1504
1505replay_cache:
1506 /* Renew the clientid on success and on replay.
1507 * Hold a session reference until done processing the compound:
1508 * nfsd4_put_session called only if the cstate slot is set.
1509 */
1510 renew_client(session->se_client);
1511 nfsd4_get_session(session);
1512out:
1513 spin_unlock(&sessionid_lock);
1514 dprintk("%s: return %d\n", __func__, ntohl(status));
1515 return status;
069b6ad4
AA		 1516}
1517
b37ad28b 1518__be32
b591480b
BF		 1519nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1520 struct nfsd4_setclientid *setclid)
1da177e4 1521{
27459f09 1522 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT		 1523 struct xdr_netobj clname = {
1524 .len = setclid->se_namelen,
1525 .data = setclid->se_name,
1526 };
1527 nfs4_verifier clverifier = setclid->se_verf;
1528 unsigned int strhashval;
28ce6054 1529 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1530 __be32 status;
68e76ad0 1531 char *princ;
a55370a3 1532 char dname[HEXDIR_LEN];
1da177e4 1533
1da177e4 1534 if (!check_name(clname))
73aea4ec 1535 return nfserr_inval;
1da177e4 1536
a55370a3
N		 1537 status = nfs4_make_rec_clidname(dname, &clname);
1538 if (status)
73aea4ec 1539 return status;
a55370a3 1540
1da177e4
LT		 1541 /*
1542 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1543 * We get here on a DRC miss.
1544 */
1545
a55370a3 1546 strhashval = clientstr_hashval(dname);
1da177e4 1547
1da177e4 1548 nfs4_lock_state();
a1bcecd2 1549 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1550 if (conf) {
a186e767 1551 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1552 status = nfserr_clid_inuse;
026722c2
BF		 1553 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1554 dprintk("NFSD: setclientid: string in use by client"
1555 " at %pI4\n", &conf->cl_addr);
1da177e4
LT		 1556 goto out;
1557 }
1da177e4 1558 }
a186e767
BF		 1559 /*
1560 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1561 * has a description of SETCLIENTID request processing consisting
1562 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1563 */
a1bcecd2 1564 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1565 status = nfserr_resource;
1566 if (!conf) {
a186e767
BF		 1567 /*
1568 * RFC 3530 14.2.33 CASE 4:
1569 * placed first, because it is the normal case
1da177e4
LT		 1570 */
1571 if (unconf)
1572 expire_client(unconf);
a55370a3
N		 1573 new = create_client(clname, dname);
1574 if (new == NULL)
1da177e4 1575 goto out;
1da177e4 1576 gen_clid(new);
599e0a22 1577 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1578 /*
a186e767
BF		 1579 * RFC 3530 14.2.33 CASE 1:
1580 * probable callback update
1da177e4 1581 */
31f4a6c1
N		 1582 if (unconf) {
1583 /* Note this is removing unconfirmed {*x***},
1584 * which is stronger than RFC recommended {vxc**}.
1585 * This has the advantage that there is at most
1586 * one {*x***} in either list at any time.
1587 */
1588 expire_client(unconf);
1da177e4 1589 }
a55370a3
N		 1590 new = create_client(clname, dname);
1591 if (new == NULL)
1da177e4 1592 goto out;
1da177e4 1593 copy_clid(new, conf);
1da177e4
LT		 1594 } else if (!unconf) {
1595 /*
a186e767
BF		 1596 * RFC 3530 14.2.33 CASE 2:
1597 * probable client reboot; state will be removed if
1598 * confirmed.
1da177e4 1599 */
a55370a3
N		 1600 new = create_client(clname, dname);
1601 if (new == NULL)
1da177e4 1602 goto out;
1da177e4 1603 gen_clid(new);
49ba8781 1604 } else {
a186e767
BF		 1605 /*
1606 * RFC 3530 14.2.33 CASE 3:
1607 * probable client reboot; state will be removed if
1608 * confirmed.
1da177e4
LT		 1609 */
1610 expire_client(unconf);
a55370a3
N		 1611 new = create_client(clname, dname);
1612 if (new == NULL)
1da177e4 1613 goto out;
1da177e4 1614 gen_clid(new);
1da177e4 1615 }
c175b83c
BF		 1616 copy_verf(new, &clverifier);
1617 new->cl_addr = sin->sin_addr.s_addr;
61054b14 1618 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1619 princ = svc_gss_principal(rqstp);
1620 if (princ) {
1621 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1622 if (new->cl_principal == NULL) {
1623 free_client(new);
1624 goto out;
1625 }
1626 }
c175b83c
BF		 1627 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1628 gen_confirm(new);
1629 gen_callback(new, setclid);
1630 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1631 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1632 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1633 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1634 status = nfs_ok;
1635out:
1636 nfs4_unlock_state();
1637 return status;
1638}
1639
1640
1641/*
a186e767
BF		 1642 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1643 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1644 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1645 */
b37ad28b 1646__be32
b591480b
BF		 1647nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1648 struct nfsd4_compound_state *cstate,
1649 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1650{
27459f09 1651 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 1652 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1653 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1654 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1655 __be32 status;
1da177e4
LT		 1656
1657 if (STALE_CLIENTID(clid))
1658 return nfserr_stale_clientid;
1659 /*
1660 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1661 * We get here on a DRC miss.
1662 */
1663
1664 nfs4_lock_state();
21ab45a4
N		 1665
1666 conf = find_confirmed_client(clid);
1667 unconf = find_unconfirmed_client(clid);
1668
1669 status = nfserr_clid_inuse;
27459f09 1670 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 1671 goto out;
27459f09 1672 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N		 1673 goto out;
1674
a186e767
BF		 1675 /*
1676 * section 14.2.34 of RFC 3530 has a description of
1677 * SETCLIENTID_CONFIRM request processing consisting
1678 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1679 */
366e0c1d 1680 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1681 /*
1682 * RFC 3530 14.2.34 CASE 1:
1683 * callback update
1684 */
599e0a22 1685 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1686 status = nfserr_clid_inuse;
1687 else {
1a69c179
N		 1688 /* XXX: We just turn off callbacks until we can handle
1689 * change request correctly. */
cb36d634 1690 atomic_set(&conf->cl_callback.cb_set, 0);
21ab45a4 1691 gen_confirm(conf);
46309029 1692 nfsd4_remove_clid_dir(unconf);
1a69c179 1693 expire_client(unconf);
1da177e4 1694 status = nfs_ok;
1a69c179 1695
1da177e4 1696 }
f3aba4e5 1697 } else if (conf && !unconf) {
a186e767
BF		 1698 /*
1699 * RFC 3530 14.2.34 CASE 2:
1700 * probable retransmitted request; play it safe and
1701 * do nothing.
7c79f737 1702 */
599e0a22 1703 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1704 status = nfserr_clid_inuse;
21ab45a4 1705 else
1da177e4 1706 status = nfs_ok;
7c79f737 1707 } else if (!conf && unconf
599e0a22 1708 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1709 /*
1710 * RFC 3530 14.2.34 CASE 3:
1711 * Normal case; new or rebooted client:
7c79f737 1712 */
599e0a22 1713 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1714 status = nfserr_clid_inuse;
1715 } else {
1a69c179
N		 1716 unsigned int hash =
1717 clientstr_hashval(unconf->cl_recdir);
1718 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1719 hash, false);
1a69c179 1720 if (conf) {
c7b9a459 1721 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1722 expire_client(conf);
1723 }
1da177e4 1724 move_to_confirmed(unconf);
21ab45a4 1725 conf = unconf;
46f8a64b 1726 nfsd4_probe_callback(conf);
1a69c179 1727 status = nfs_ok;
1da177e4 1728 }
599e0a22
BF		 1729 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1730 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1731 &confirm)))) {
a186e767
BF		 1732 /*
1733 * RFC 3530 14.2.34 CASE 4:
1734 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1735 */
1da177e4 1736 status = nfserr_stale_clientid;
7c79f737 1737 } else {
08e8987c
N		 1738 /* check that we have hit one of the cases...*/
1739 status = nfserr_clid_inuse;
1740 }
1da177e4 1741out:
1da177e4
LT		 1742 nfs4_unlock_state();
1743 return status;
1744}
1745
1da177e4
LT		 1746/* OPEN Share state helper functions */
1747static inline struct nfs4_file *
1748alloc_init_file(struct inode *ino)
1749{
1750 struct nfs4_file *fp;
1751 unsigned int hashval = file_hashval(ino);
1752
e60d4398
N		 1753 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1754 if (fp) {
8b671b80 1755 atomic_set(&fp->fi_ref, 1);
1da177e4 1756 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1757 INIT_LIST_HEAD(&fp->fi_stateids);
1758 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1759 spin_lock(&recall_lock);
1da177e4 1760 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1761 spin_unlock(&recall_lock);
1da177e4
LT		 1762 fp->fi_inode = igrab(ino);
1763 fp->fi_id = current_fileid++;
47f9940c 1764 fp->fi_had_conflict = false;
1da177e4
LT		 1765 return fp;
1766 }
1767 return NULL;
1768}
1769
e60d4398 1770static void
e18b890b 1771nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1772{
e60d4398
N		 1773 if (*slab == NULL)
1774 return;
1a1d92c1 1775 kmem_cache_destroy(*slab);
e60d4398 1776 *slab = NULL;
1da177e4
LT		 1777}
1778
e8ff2a84 1779void
1da177e4
LT		 1780nfsd4_free_slabs(void)
1781{
e60d4398
N		 1782 nfsd4_free_slab(&stateowner_slab);
1783 nfsd4_free_slab(&file_slab);
5ac049ac 1784 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1785 nfsd4_free_slab(&deleg_slab);
e60d4398 1786}
1da177e4 1787
e60d4398
N		 1788static int
1789nfsd4_init_slabs(void)
1790{
1791 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1792 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1793 if (stateowner_slab == NULL)
1794 goto out_nomem;
1795 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1796 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1797 if (file_slab == NULL)
1798 goto out_nomem;
5ac049ac 1799 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1800 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1801 if (stateid_slab == NULL)
1802 goto out_nomem;
5b2d21c1 1803 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1804 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1805 if (deleg_slab == NULL)
1806 goto out_nomem;
e60d4398
N		 1807 return 0;
1808out_nomem:
1809 nfsd4_free_slabs();
1810 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1811 return -ENOMEM;
1da177e4
LT		 1812}
1813
1814void
1815nfs4_free_stateowner(struct kref *kref)
1816{
1817 struct nfs4_stateowner *sop =
1818 container_of(kref, struct nfs4_stateowner, so_ref);
1819 kfree(sop->so_owner.data);
1820 kmem_cache_free(stateowner_slab, sop);
1821}
1822
1823static inline struct nfs4_stateowner *
1824alloc_stateowner(struct xdr_netobj *owner)
1825{
1826 struct nfs4_stateowner *sop;
1827
1828 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1829 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1830 memcpy(sop->so_owner.data, owner->data, owner->len);
1831 sop->so_owner.len = owner->len;
1832 kref_init(&sop->so_ref);
1833 return sop;
1834 }
1835 kmem_cache_free(stateowner_slab, sop);
1836 }
1837 return NULL;
1838}
1839
1840static struct nfs4_stateowner *
1841alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1842 struct nfs4_stateowner *sop;
1843 struct nfs4_replay *rp;
1844 unsigned int idhashval;
1845
1846 if (!(sop = alloc_stateowner(&open->op_owner)))
1847 return NULL;
1848 idhashval = ownerid_hashval(current_ownerid);
1849 INIT_LIST_HEAD(&sop->so_idhash);
1850 INIT_LIST_HEAD(&sop->so_strhash);
1851 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1852 INIT_LIST_HEAD(&sop->so_stateids);
1853 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1854 INIT_LIST_HEAD(&sop->so_close_lru);
1855 sop->so_time = 0;
1856 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1857 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1858 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1859 sop->so_is_open_owner = 1;
1860 sop->so_id = current_ownerid++;
1861 sop->so_client = clp;
1862 sop->so_seqid = open->op_seqid;
1863 sop->so_confirmed = 0;
1864 rp = &sop->so_replay;
de1ae286 1865 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1866 rp->rp_buflen = 0;
1867 rp->rp_buf = rp->rp_ibuf;
1868 return sop;
1869}
1870
1da177e4
LT		 1871static inline void
1872init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1873 struct nfs4_stateowner *sop = open->op_stateowner;
1874 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1875
1876 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1877 INIT_LIST_HEAD(&stp->st_perstateowner);
1878 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1879 INIT_LIST_HEAD(&stp->st_perfile);
1880 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1881 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1882 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1883 stp->st_stateowner = sop;
13cd2184 1884 get_nfs4_file(fp);
1da177e4
LT		 1885 stp->st_file = fp;
1886 stp->st_stateid.si_boot = boot_time;
1887 stp->st_stateid.si_stateownerid = sop->so_id;
1888 stp->st_stateid.si_fileid = fp->fi_id;
1889 stp->st_stateid.si_generation = 0;
1890 stp->st_access_bmap = 0;
1891 stp->st_deny_bmap = 0;
1892 __set_bit(open->op_share_access, &stp->st_access_bmap);
1893 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1894 stp->st_openstp = NULL;
1da177e4
LT		 1895}
1896
fd39ca9a 1897static void
1da177e4
LT		 1898move_to_close_lru(struct nfs4_stateowner *sop)
1899{
1900 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1901
358dd55a 1902 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1903 sop->so_time = get_seconds();
1904}
1905
1da177e4 1906static int
599e0a22
BF		 1907same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1908 clientid_t *clid)
1909{
1910 return (sop->so_owner.len == owner->len) &&
1911 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1912 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1913}
1914
1915static struct nfs4_stateowner *
1916find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1917{
1918 struct nfs4_stateowner *so = NULL;
1919
1920 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1921 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1922 return so;
1923 }
1924 return NULL;
1925}
1926
1927/* search file_hashtbl[] for file */
1928static struct nfs4_file *
1929find_file(struct inode *ino)
1930{
1931 unsigned int hashval = file_hashval(ino);
1932 struct nfs4_file *fp;
1933
8b671b80 1934 spin_lock(&recall_lock);
1da177e4 1935 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1936 if (fp->fi_inode == ino) {
1937 get_nfs4_file(fp);
8b671b80 1938 spin_unlock(&recall_lock);
1da177e4 1939 return fp;
13cd2184 1940 }
1da177e4 1941 }
8b671b80 1942 spin_unlock(&recall_lock);
1da177e4
LT		 1943 return NULL;
1944}
1945
8838dc43 1946static inline int access_valid(u32 x)
ba5a6a19 1947{
8838dc43
BF		 1948 if (x < NFS4_SHARE_ACCESS_READ)
1949 return 0;
1950 if (x > NFS4_SHARE_ACCESS_BOTH)
1951 return 0;
1952 return 1;
ba5a6a19
BF		 1953}
1954
8838dc43 1955static inline int deny_valid(u32 x)
ba5a6a19 1956{
8838dc43
BF		 1957 /* Note: unlike access bits, deny bits may be zero. */
1958 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1959}
1da177e4 1960
4f83aa30
BF		 1961/*
1962 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1963 * st_{access,deny}_bmap field of the stateid, in order to track not
1964 * only what share bits are currently in force, but also what
1965 * combinations of share bits previous opens have used. This allows us
1966 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1967 * return an error if the client attempt to downgrade to a combination
1968 * of share bits not explicable by closing some of its previous opens.
1969 *
1970 * XXX: This enforcement is actually incomplete, since we don't keep
1971 * track of access/deny bit combinations; so, e.g., we allow:
1972 *
1973 * OPEN allow read, deny write
1974 * OPEN allow both, deny none
1975 * DOWNGRADE allow read, deny none
1976 *
1977 * which we should reject.
1978 */
fd39ca9a 1979static void
1da177e4
LT		 1980set_access(unsigned int *access, unsigned long bmap) {
1981 int i;
1982
1983 *access = 0;
1984 for (i = 1; i < 4; i++) {
1985 if (test_bit(i, &bmap))
1986 *access |= i;
1987 }
1988}
1989
fd39ca9a 1990static void
1da177e4
LT		 1991set_deny(unsigned int *deny, unsigned long bmap) {
1992 int i;
1993
1994 *deny = 0;
1995 for (i = 0; i < 4; i++) {
1996 if (test_bit(i, &bmap))
1997 *deny |= i ;
1998 }
1999}
2000
2001static int
2002test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2003 unsigned int access, deny;
2004
2005 set_access(&access, stp->st_access_bmap);
2006 set_deny(&deny, stp->st_deny_bmap);
2007 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2008 return 0;
2009 return 1;
2010}
2011
2012/*
2013 * Called to check deny when READ with all zero stateid or
2014 * WRITE with all zero or all one stateid
2015 */
b37ad28b 2016static __be32
1da177e4
LT		 2017nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2018{
2019 struct inode *ino = current_fh->fh_dentry->d_inode;
2020 struct nfs4_file *fp;
2021 struct nfs4_stateid *stp;
b37ad28b 2022 __be32 ret;
1da177e4
LT		 2023
2024 dprintk("NFSD: nfs4_share_conflict\n");
2025
2026 fp = find_file(ino);
13cd2184
N		 2027 if (!fp)
2028 return nfs_ok;
b700949b 2029 ret = nfserr_locked;
1da177e4 2030 /* Search for conflicting share reservations */
13cd2184
N		 2031 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2032 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2033 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2034 goto out;
1da177e4 2035 }
13cd2184
N		 2036 ret = nfs_ok;
2037out:
2038 put_nfs4_file(fp);
2039 return ret;
1da177e4
LT		 2040}
2041
2042static inline void
2043nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2044{
2045 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2046 drop_file_write_access(filp);
1da177e4
LT		 2047 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2048 }
2049}
2050
2051/*
2052 * Recall a delegation
2053 */
2054static int
2055do_recall(void *__dp)
2056{
2057 struct nfs4_delegation *dp = __dp;
2058
47f9940c 2059 dp->dl_file->fi_had_conflict = true;
1da177e4
LT		 2060 nfsd4_cb_recall(dp);
2061 return 0;
2062}
2063
2064/*
2065 * Spawn a thread to perform a recall on the delegation represented
2066 * by the lease (file_lock)
2067 *
2068 * Called from break_lease() with lock_kernel() held.
2069 * Note: we assume break_lease will only call this *once* for any given
2070 * lease.
2071 */
2072static
2073void nfsd_break_deleg_cb(struct file_lock *fl)
2074{
2075 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
2076 struct task_struct *t;
2077
2078 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2079 if (!dp)
2080 return;
2081
2082 /* We're assuming the state code never drops its reference
2083 * without first removing the lease. Since we're in this lease
2084 * callback (and since the lease code is serialized by the kernel
2085 * lock) we know the server hasn't removed the lease yet, we know
2086 * it's safe to take a reference: */
2087 atomic_inc(&dp->dl_count);
cfdcad4d 2088 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2089
2090 spin_lock(&recall_lock);
2091 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2092 spin_unlock(&recall_lock);
2093
2094 /* only place dl_time is set. protected by lock_kernel*/
2095 dp->dl_time = get_seconds();
2096
0272e1fd
BF		 2097 /*
2098 * We don't want the locks code to timeout the lease for us;
2099 * we'll remove it ourself if the delegation isn't returned
2100 * in time.
2101 */
2102 fl->fl_break_time = 0;
1da177e4
LT		 2103
2104 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
2105 if (IS_ERR(t)) {
2106 struct nfs4_client *clp = dp->dl_client;
2107
2108 printk(KERN_INFO "NFSD: Callback thread failed for "
2109 "for client (clientid %08x/%08x)\n",
2110 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 2111 put_nfs4_client(dp->dl_client);
1da177e4
LT		 2112 nfs4_put_delegation(dp);
2113 }
2114}
2115
2116/*
2117 * The file_lock is being reapd.
2118 *
2119 * Called by locks_free_lock() with lock_kernel() held.
2120 */
2121static
2122void nfsd_release_deleg_cb(struct file_lock *fl)
2123{
2124 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2125
2126 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2127
2128 if (!(fl->fl_flags & FL_LEASE) || !dp)
2129 return;
2130 dp->dl_flock = NULL;
2131}
2132
2133/*
2134 * Set the delegation file_lock back pointer.
2135 *
a9933cea 2136 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2137 */
2138static
2139void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2140{
2141 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2142
2143 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2144 if (!dp)
2145 return;
2146 dp->dl_flock = new;
2147}
2148
2149/*
a9933cea 2150 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2151 */
2152static
2153int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2154{
2155 struct nfs4_delegation *onlistd =
2156 (struct nfs4_delegation *)onlist->fl_owner;
2157 struct nfs4_delegation *tryd =
2158 (struct nfs4_delegation *)try->fl_owner;
2159
2160 if (onlist->fl_lmops != try->fl_lmops)
2161 return 0;
2162
2163 return onlistd->dl_client == tryd->dl_client;
2164}
2165
2166
2167static
2168int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2169{
2170 if (arg & F_UNLCK)
2171 return lease_modify(onlist, arg);
2172 else
2173 return -EAGAIN;
2174}
2175
fd39ca9a 2176static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2177 .fl_break = nfsd_break_deleg_cb,
2178 .fl_release_private = nfsd_release_deleg_cb,
2179 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2180 .fl_mylease = nfsd_same_client_deleg_cb,
2181 .fl_change = nfsd_change_deleg_cb,
2182};
2183
2184
b37ad28b 2185__be32
1da177e4
LT		 2186nfsd4_process_open1(struct nfsd4_open *open)
2187{
1da177e4
LT		 2188 clientid_t *clientid = &open->op_clientid;
2189 struct nfs4_client *clp = NULL;
2190 unsigned int strhashval;
2191 struct nfs4_stateowner *sop = NULL;
2192
1da177e4 2193 if (!check_name(open->op_owner))
0f442aa2 2194 return nfserr_inval;
1da177e4
LT		 2195
2196 if (STALE_CLIENTID(&open->op_clientid))
2197 return nfserr_stale_clientid;
2198
2199 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2200 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2201 open->op_stateowner = sop;
2202 if (!sop) {
2203 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2204 clp = find_confirmed_client(clientid);
2205 if (clp == NULL)
0f442aa2
BF		 2206 return nfserr_expired;
2207 goto renew;
1da177e4 2208 }
0f442aa2
BF		 2209 if (!sop->so_confirmed) {
2210 /* Replace unconfirmed owners without checking for replay. */
2211 clp = sop->so_client;
f044ff83 2212 release_openowner(sop);
0f442aa2
BF		 2213 open->op_stateowner = NULL;
2214 goto renew;
2215 }
2216 if (open->op_seqid == sop->so_seqid - 1) {
2217 if (sop->so_replay.rp_buflen)
a90b061c 2218 return nfserr_replay_me;
0f442aa2
BF		 2219 /* The original OPEN failed so spectacularly
2220 * that we don't even have replay data saved!
2221 * Therefore, we have no choice but to continue
2222 * processing this OPEN; presumably, we'll
2223 * fail again for the same reason.
2224 */
2225 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2226 goto renew;
2227 }
2228 if (open->op_seqid != sop->so_seqid)
2229 return nfserr_bad_seqid;
1da177e4 2230renew:
0f442aa2
BF		 2231 if (open->op_stateowner == NULL) {
2232 sop = alloc_init_open_stateowner(strhashval, clp, open);
2233 if (sop == NULL)
2234 return nfserr_resource;
2235 open->op_stateowner = sop;
2236 }
2237 list_del_init(&sop->so_close_lru);
1da177e4 2238 renew_client(sop->so_client);
0f442aa2 2239 return nfs_ok;
1da177e4
LT		 2240}
2241
b37ad28b 2242static inline __be32
4a6e43e6
N		 2243nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2244{
2245 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2246 return nfserr_openmode;
2247 else
2248 return nfs_ok;
2249}
2250
52f4fb43
N		 2251static struct nfs4_delegation *
2252find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2253{
2254 struct nfs4_delegation *dp;
2255
ea1da636 2256 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2257 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2258 return dp;
2259 }
2260 return NULL;
2261}
2262
b37ad28b 2263static __be32
567d9829
N		 2264nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2265 struct nfs4_delegation **dp)
2266{
2267 int flags;
b37ad28b 2268 __be32 status = nfserr_bad_stateid;
567d9829
N		 2269
2270 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2271 if (*dp == NULL)
c44c5eeb 2272 goto out;
567d9829
N		 2273 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2274 RD_STATE : WR_STATE;
2275 status = nfs4_check_delegmode(*dp, flags);
2276 if (status)
2277 *dp = NULL;
c44c5eeb
N		 2278out:
2279 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2280 return nfs_ok;
2281 if (status)
2282 return status;
2283 open->op_stateowner->so_confirmed = 1;
2284 return nfs_ok;
567d9829
N		 2285}
2286
b37ad28b 2287static __be32
1da177e4
LT		 2288nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2289{
2290 struct nfs4_stateid *local;
b37ad28b 2291 __be32 status = nfserr_share_denied;
1da177e4
LT		 2292 struct nfs4_stateowner *sop = open->op_stateowner;
2293
8beefa24 2294 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2295 /* ignore lock owners */
2296 if (local->st_stateowner->so_is_open_owner == 0)
2297 continue;
2298 /* remember if we have seen this open owner */
2299 if (local->st_stateowner == sop)
2300 *stpp = local;
2301 /* check for conflicting share reservations */
2302 if (!test_share(local, open))
2303 goto out;
2304 }
2305 status = 0;
2306out:
2307 return status;
2308}
2309
5ac049ac
N		 2310static inline struct nfs4_stateid *
2311nfs4_alloc_stateid(void)
2312{
2313 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2314}
2315
b37ad28b 2316static __be32
1da177e4 2317nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2318 struct nfs4_delegation *dp,
1da177e4
LT		 2319 struct svc_fh *cur_fh, int flags)
2320{
2321 struct nfs4_stateid *stp;
1da177e4 2322
5ac049ac 2323 stp = nfs4_alloc_stateid();
1da177e4
LT		 2324 if (stp == NULL)
2325 return nfserr_resource;
2326
567d9829
N		 2327 if (dp) {
2328 get_file(dp->dl_vfs_file);
2329 stp->st_vfs_file = dp->dl_vfs_file;
2330 } else {
b37ad28b 2331 __be32 status;
567d9829
N		 2332 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2333 &stp->st_vfs_file);
2334 if (status) {
2335 if (status == nfserr_dropit)
2336 status = nfserr_jukebox;
5ac049ac 2337 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2338 return status;
2339 }
1da177e4 2340 }
1da177e4
LT		 2341 *stpp = stp;
2342 return 0;
2343}
2344
b37ad28b 2345static inline __be32
1da177e4
LT		 2346nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2347 struct nfsd4_open *open)
2348{
2349 struct iattr iattr = {
2350 .ia_valid = ATTR_SIZE,
2351 .ia_size = 0,
2352 };
2353 if (!open->op_truncate)
2354 return 0;
2355 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2356 return nfserr_inval;
1da177e4
LT		 2357 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2358}
2359
b37ad28b 2360static __be32
1da177e4
LT		 2361nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2362{
2363 struct file *filp = stp->st_vfs_file;
7eaa36e2 2364 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2365 unsigned int share_access, new_writer;
b37ad28b 2366 __be32 status;
1da177e4
LT		 2367
2368 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2369 new_writer = (~share_access) & open->op_share_access
2370 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2371
6c26d08f 2372 if (new_writer) {
b37ad28b
AV		 2373 int err = get_write_access(inode);
2374 if (err)
2375 return nfserrno(err);
e518f056
BH		 2376 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2377 if (err)
2378 return nfserrno(err);
2379 file_take_write(filp);
6c26d08f 2380 }
1da177e4
LT		 2381 status = nfsd4_truncate(rqstp, cur_fh, open);
2382 if (status) {
6c26d08f
BF		 2383 if (new_writer)
2384 put_write_access(inode);
1da177e4
LT		 2385 return status;
2386 }
2387 /* remember the open */
6c26d08f 2388 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2389 __set_bit(open->op_share_access, &stp->st_access_bmap);
2390 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2391
2392 return nfs_ok;
2393}
2394
2395
1da177e4 2396static void
37515177 2397nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2398{
37515177
N		 2399 open->op_stateowner->so_confirmed = 1;
2400 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2401}
2402
2403/*
2404 * Attempt to hand out a delegation.
2405 */
2406static void
2407nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2408{
2409 struct nfs4_delegation *dp;
2410 struct nfs4_stateowner *sop = stp->st_stateowner;
2411 struct nfs4_callback *cb = &sop->so_client->cl_callback;
2412 struct file_lock fl, *flp = &fl;
2413 int status, flag = 0;
2414
2415 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2416 open->op_recall = 0;
2417 switch (open->op_claim_type) {
2418 case NFS4_OPEN_CLAIM_PREVIOUS:
2419 if (!atomic_read(&cb->cb_set))
2420 open->op_recall = 1;
2421 flag = open->op_delegate_type;
2422 if (flag == NFS4_OPEN_DELEGATE_NONE)
2423 goto out;
2424 break;
2425 case NFS4_OPEN_CLAIM_NULL:
2426 /* Let's not give out any delegations till everyone's
2427 * had the chance to reclaim theirs.... */
af558e33 2428 if (locks_in_grace())
7b190fec
N		 2429 goto out;
2430 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2431 goto out;
2432 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2433 flag = NFS4_OPEN_DELEGATE_WRITE;
2434 else
2435 flag = NFS4_OPEN_DELEGATE_READ;
2436 break;
2437 default:
2438 goto out;
2439 }
1da177e4
LT		 2440
2441 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2442 if (dp == NULL) {
2443 flag = NFS4_OPEN_DELEGATE_NONE;
2444 goto out;
2445 }
2446 locks_init_lock(&fl);
2447 fl.fl_lmops = &nfsd_lease_mng_ops;
2448 fl.fl_flags = FL_LEASE;
9167f501 2449 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2450 fl.fl_end = OFFSET_MAX;
2451 fl.fl_owner = (fl_owner_t)dp;
2452 fl.fl_file = stp->st_vfs_file;
2453 fl.fl_pid = current->tgid;
2454
a9933cea 2455 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2456 * the lock_manager callbacks fl_mylease and fl_change are used
2457 */
9167f501 2458 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2459 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2460 unhash_delegation(dp);
1da177e4
LT		 2461 flag = NFS4_OPEN_DELEGATE_NONE;
2462 goto out;
2463 }
2464
2465 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2466
2467 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2468 dp->dl_stateid.si_boot,
2469 dp->dl_stateid.si_stateownerid,
2470 dp->dl_stateid.si_fileid,
2471 dp->dl_stateid.si_generation);
2472out:
7b190fec
N		 2473 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2474 && flag == NFS4_OPEN_DELEGATE_NONE
2475 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2476 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2477 open->op_delegate_type = flag;
2478}
2479
2480/*
2481 * called with nfs4_lock_state() held.
2482 */
b37ad28b 2483__be32
1da177e4
LT		 2484nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2485{
2486 struct nfs4_file *fp = NULL;
2487 struct inode *ino = current_fh->fh_dentry->d_inode;
2488 struct nfs4_stateid *stp = NULL;
567d9829 2489 struct nfs4_delegation *dp = NULL;
b37ad28b 2490 __be32 status;
1da177e4
LT		 2491
2492 status = nfserr_inval;
ba5a6a19
BF		 2493 if (!access_valid(open->op_share_access)
2494 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2495 goto out;
2496 /*
2497 * Lookup file; if found, lookup stateid and check open request,
2498 * and check for delegations in the process of being recalled.
2499 * If not found, create the nfs4_file struct
2500 */
2501 fp = find_file(ino);
2502 if (fp) {
2503 if ((status = nfs4_check_open(fp, open, &stp)))
2504 goto out;
c44c5eeb
N		 2505 status = nfs4_check_deleg(fp, open, &dp);
2506 if (status)
2507 goto out;
1da177e4 2508 } else {
c44c5eeb
N		 2509 status = nfserr_bad_stateid;
2510 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2511 goto out;
1da177e4
LT		 2512 status = nfserr_resource;
2513 fp = alloc_init_file(ino);
2514 if (fp == NULL)
2515 goto out;
2516 }
2517
2518 /*
2519 * OPEN the file, or upgrade an existing OPEN.
2520 * If truncate fails, the OPEN fails.
2521 */
2522 if (stp) {
2523 /* Stateid was found, this is an OPEN upgrade */
2524 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2525 if (status)
2526 goto out;
444c2c07 2527 update_stateid(&stp->st_stateid);
1da177e4
LT		 2528 } else {
2529 /* Stateid was not found, this is a new OPEN */
2530 int flags = 0;
9ecb6a08 2531 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2532 flags |= NFSD_MAY_READ;
1da177e4 2533 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2534 flags |= NFSD_MAY_WRITE;
567d9829
N		 2535 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2536 if (status)
1da177e4
LT		 2537 goto out;
2538 init_stateid(stp, fp, open);
2539 status = nfsd4_truncate(rqstp, current_fh, open);
2540 if (status) {
2283963f 2541 release_open_stateid(stp);
1da177e4
LT		 2542 goto out;
2543 }
2544 }
2545 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2546
2547 /*
2548 * Attempt to hand out a delegation. No error return, because the
2549 * OPEN succeeds even if we fail.
2550 */
2551 nfs4_open_delegation(current_fh, open, stp);
2552
2553 status = nfs_ok;
2554
2555 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2556 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2557 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2558out:
13cd2184
N		 2559 if (fp)
2560 put_nfs4_file(fp);
37515177
N		 2561 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2562 nfs4_set_claim_prev(open);
1da177e4
LT		 2563 /*
2564 * To finish the open response, we just need to set the rflags.
2565 */
2566 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
2567 if (!open->op_stateowner->so_confirmed)
2568 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2569
2570 return status;
2571}
2572
b37ad28b 2573__be32
b591480b
BF		 2574nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2575 clientid_t *clid)
1da177e4
LT		 2576{
2577 struct nfs4_client *clp;
b37ad28b 2578 __be32 status;
1da177e4
LT		 2579
2580 nfs4_lock_state();
2581 dprintk("process_renew(%08x/%08x): starting\n",
2582 clid->cl_boot, clid->cl_id);
2583 status = nfserr_stale_clientid;
2584 if (STALE_CLIENTID(clid))
2585 goto out;
2586 clp = find_confirmed_client(clid);
2587 status = nfserr_expired;
2588 if (clp == NULL) {
2589 /* We assume the client took too long to RENEW. */
2590 dprintk("nfsd4_renew: clientid not found!\n");
2591 goto out;
2592 }
2593 renew_client(clp);
2594 status = nfserr_cb_path_down;
ea1da636 2595 if (!list_empty(&clp->cl_delegations)
1da177e4
LT		 2596 && !atomic_read(&clp->cl_callback.cb_set))
2597 goto out;
2598 status = nfs_ok;
2599out:
2600 nfs4_unlock_state();
2601 return status;
2602}
2603
af558e33
BF		 2604struct lock_manager nfsd4_manager = {
2605};
2606
a76b4319 2607static void
af558e33 2608nfsd4_end_grace(void)
a76b4319
N		 2609{
2610 dprintk("NFSD: end of grace period\n");
c7b9a459 2611 nfsd4_recdir_purge_old();
af558e33 2612 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2613}
2614
fd39ca9a 2615static time_t
1da177e4
LT		 2616nfs4_laundromat(void)
2617{
2618 struct nfs4_client *clp;
2619 struct nfs4_stateowner *sop;
2620 struct nfs4_delegation *dp;
2621 struct list_head *pos, *next, reaplist;
2622 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2623 time_t t, clientid_val = NFSD_LEASE_TIME;
2624 time_t u, test_val = NFSD_LEASE_TIME;
2625
2626 nfs4_lock_state();
2627
2628 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2629 if (locks_in_grace())
2630 nfsd4_end_grace();
1da177e4
LT		 2631 list_for_each_safe(pos, next, &client_lru) {
2632 clp = list_entry(pos, struct nfs4_client, cl_lru);
2633 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2634 t = clp->cl_time - cutoff;
2635 if (clientid_val > t)
2636 clientid_val = t;
2637 break;
2638 }
2639 dprintk("NFSD: purging unused client (clientid %08x)\n",
2640 clp->cl_clientid.cl_id);
c7b9a459 2641 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2642 expire_client(clp);
2643 }
2644 INIT_LIST_HEAD(&reaplist);
2645 spin_lock(&recall_lock);
2646 list_for_each_safe(pos, next, &del_recall_lru) {
2647 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2648 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2649 u = dp->dl_time - cutoff;
2650 if (test_val > u)
2651 test_val = u;
2652 break;
2653 }
2654 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2655 dp, dp->dl_flock);
2656 list_move(&dp->dl_recall_lru, &reaplist);
2657 }
2658 spin_unlock(&recall_lock);
2659 list_for_each_safe(pos, next, &reaplist) {
2660 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2661 list_del_init(&dp->dl_recall_lru);
2662 unhash_delegation(dp);
2663 }
2664 test_val = NFSD_LEASE_TIME;
2665 list_for_each_safe(pos, next, &close_lru) {
2666 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2667 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2668 u = sop->so_time - cutoff;
2669 if (test_val > u)
2670 test_val = u;
2671 break;
2672 }
2673 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2674 sop->so_id);
f044ff83 2675 release_openowner(sop);
1da177e4
LT		 2676 }
2677 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2678 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2679 nfs4_unlock_state();
2680 return clientid_val;
2681}
2682
a254b246
HH		 2683static struct workqueue_struct *laundry_wq;
2684static void laundromat_main(struct work_struct *);
2685static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2686
2687static void
c4028958 2688laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2689{
2690 time_t t;
2691
2692 t = nfs4_laundromat();
2693 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2694 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2695}
2696
fd39ca9a 2697static struct nfs4_stateowner *
f8816512
N		 2698search_close_lru(u32 st_id, int flags)
2699{
1da177e4
LT		 2700 struct nfs4_stateowner *local = NULL;
2701
1da177e4
LT		 2702 if (flags & CLOSE_STATE) {
2703 list_for_each_entry(local, &close_lru, so_close_lru) {
2704 if (local->so_id == st_id)
2705 return local;
2706 }
2707 }
2708 return NULL;
2709}
2710
2711static inline int
2712nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2713{
7eaa36e2 2714 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2715}
2716
2717static int
2718STALE_STATEID(stateid_t *stateid)
2719{
2720 if (stateid->si_boot == boot_time)
2721 return 0;
849823c5 2722 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1da177e4
LT		 2723 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2724 stateid->si_generation);
2725 return 1;
2726}
2727
2728static inline int
2729access_permit_read(unsigned long access_bmap)
2730{
2731 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2732 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2733 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2734}
2735
2736static inline int
2737access_permit_write(unsigned long access_bmap)
2738{
2739 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2740 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2741}
2742
2743static
b37ad28b 2744__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2745{
b37ad28b 2746 __be32 status = nfserr_openmode;
1da177e4
LT		 2747
2748 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2749 goto out;
2750 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2751 goto out;
2752 status = nfs_ok;
2753out:
2754 return status;
2755}
2756
b37ad28b 2757static inline __be32
1da177e4
LT		 2758check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2759{
203a8c8e 2760 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2761 return nfs_ok;
af558e33 2762 else if (locks_in_grace()) {
1da177e4
LT		 2763 /* Answer in remaining cases depends on existance of
2764 * conflicting state; so we must wait out the grace period. */
2765 return nfserr_grace;
2766 } else if (flags & WR_STATE)
2767 return nfs4_share_conflict(current_fh,
2768 NFS4_SHARE_DENY_WRITE);
2769 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2770 return nfs4_share_conflict(current_fh,
2771 NFS4_SHARE_DENY_READ);
2772}
2773
2774/*
2775 * Allow READ/WRITE during grace period on recovered state only for files
2776 * that are not able to provide mandatory locking.
2777 */
2778static inline int
18f82731 2779grace_disallows_io(struct inode *inode)
1da177e4 2780{
203a8c8e 2781 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2782}
2783
0836f587
BF		 2784static int check_stateid_generation(stateid_t *in, stateid_t *ref)
2785{
2786 /* If the client sends us a stateid from the future, it's buggy: */
2787 if (in->si_generation > ref->si_generation)
2788 return nfserr_bad_stateid;
2789 /*
2790 * The following, however, can happen. For example, if the
2791 * client sends an open and some IO at the same time, the open
2792 * may bump si_generation while the IO is still in flight.
2793 * Thanks to hard links and renames, the client never knows what
2794 * file an open will affect. So it could avoid that situation
2795 * only by serializing all opens and IO from the same open
2796 * owner. To recover from the old_stateid error, the client
2797 * will just have to retry the IO:
2798 */
2799 if (in->si_generation < ref->si_generation)
2800 return nfserr_old_stateid;
2801 return nfs_ok;
2802}
2803
3e633079
BF		 2804static int is_delegation_stateid(stateid_t *stateid)
2805{
2806 return stateid->si_fileid == 0;
2807}
2808
1da177e4
LT		 2809/*
2810* Checks for stateid operations
2811*/
b37ad28b 2812__be32
1da177e4
LT		 2813nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2814{
2815 struct nfs4_stateid *stp = NULL;
2816 struct nfs4_delegation *dp = NULL;
1da177e4 2817 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2818 __be32 status;
1da177e4 2819
1da177e4
LT		 2820 if (filpp)
2821 *filpp = NULL;
2822
18f82731 2823 if (grace_disallows_io(ino))
1da177e4
LT		 2824 return nfserr_grace;
2825
2826 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2827 return check_special_stateids(current_fh, stateid, flags);
2828
1da177e4
LT		 2829 status = nfserr_stale_stateid;
2830 if (STALE_STATEID(stateid))
2831 goto out;
2832
1da177e4 2833 status = nfserr_bad_stateid;
3e633079 2834 if (is_delegation_stateid(stateid)) {
a4455be0 2835 dp = find_delegation_stateid(ino, stateid);
819a8f53 2836 if (!dp)
1da177e4 2837 goto out;
fd03b099 2838 status = check_stateid_generation(stateid, &dp->dl_stateid);
0c2a498f
BF		 2839 if (status)
2840 goto out;
dc9bf700
BF		 2841 status = nfs4_check_delegmode(dp, flags);
2842 if (status)
2843 goto out;
2844 renew_client(dp->dl_client);
dc9bf700
BF		 2845 if (filpp)
2846 *filpp = dp->dl_vfs_file;
1da177e4 2847 } else { /* open or lock stateid */
a4455be0 2848 stp = find_stateid(stateid, flags);
819a8f53 2849 if (!stp)
1da177e4 2850 goto out;
6150ef0d 2851 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2852 goto out;
2853 if (!stp->st_stateowner->so_confirmed)
2854 goto out;
fd03b099 2855 status = check_stateid_generation(stateid, &stp->st_stateid);
0c2a498f
BF		 2856 if (status)
2857 goto out;
a4455be0
BF		 2858 status = nfs4_check_openmode(stp, flags);
2859 if (status)
1da177e4
LT		 2860 goto out;
2861 renew_client(stp->st_stateowner->so_client);
2862 if (filpp)
2863 *filpp = stp->st_vfs_file;
1da177e4
LT		 2864 }
2865 status = nfs_ok;
2866out:
2867 return status;
2868}
2869
4c4cd222
N		 2870static inline int
2871setlkflg (int type)
2872{
2873 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2874 RD_STATE : WR_STATE;
2875}
1da177e4
LT		 2876
2877/*
2878 * Checks for sequence id mutating operations.
2879 */
b37ad28b 2880static __be32
4c4cd222 2881nfs4_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 2882{
1da177e4
LT		 2883 struct nfs4_stateid *stp;
2884 struct nfs4_stateowner *sop;
0836f587 2885 __be32 status;
1da177e4
LT		 2886
2887 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2888 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2889 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2890 stateid->si_generation);
3a4f98bb 2891
1da177e4
LT		 2892 *stpp = NULL;
2893 *sopp = NULL;
2894
1da177e4 2895 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2896 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2897 return nfserr_bad_stateid;
1da177e4
LT		 2898 }
2899
1da177e4 2900 if (STALE_STATEID(stateid))
3a4f98bb 2901 return nfserr_stale_stateid;
1da177e4
LT		 2902 /*
2903 * We return BAD_STATEID if filehandle doesn't match stateid,
2904 * the confirmed flag is incorrecly set, or the generation
2905 * number is incorrect.
1da177e4 2906 */
f8816512
N		 2907 stp = find_stateid(stateid, flags);
2908 if (stp == NULL) {
2909 /*
2910 * Also, we should make sure this isn't just the result of
2911 * a replayed close:
2912 */
2913 sop = search_close_lru(stateid->si_stateownerid, flags);
2914 if (sop == NULL)
2915 return nfserr_bad_stateid;
2916 *sopp = sop;
2917 goto check_replay;
2918 }
1da177e4 2919
39325bd0
BF		 2920 *stpp = stp;
2921 *sopp = sop = stp->st_stateowner;
2922
4c4cd222 2923 if (lock) {
4c4cd222 2924 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2925 struct nfs4_client *clp = sop->so_client;
4c4cd222 2926 int lkflg = 0;
b37ad28b 2927 __be32 status;
4c4cd222
N		 2928
2929 lkflg = setlkflg(lock->lk_type);
2930
2931 if (lock->lk_is_new) {
599e0a22
BF		 2932 if (!sop->so_is_open_owner)
2933 return nfserr_bad_stateid;
2934 if (!same_clid(&clp->cl_clientid, lockclid))
4c4cd222 2935 return nfserr_bad_stateid;
599e0a22
BF		 2936 /* stp is the open stateid */
2937 status = nfs4_check_openmode(stp, lkflg);
2938 if (status)
2939 return status;
2940 } else {
2941 /* stp is the lock stateid */
2942 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2943 if (status)
2944 return status;
4c4cd222 2945 }
1da177e4
LT		 2946 }
2947
f3362737 2948 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2949 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2950 return nfserr_bad_stateid;
1da177e4
LT		 2951 }
2952
1da177e4
LT		 2953 /*
2954 * We now validate the seqid and stateid generation numbers.
2955 * For the moment, we ignore the possibility of
2956 * generation number wraparound.
2957 */
bd9aac52 2958 if (seqid != sop->so_seqid)
1da177e4
LT		 2959 goto check_replay;
2960
3a4f98bb 2961 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2962 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2963 " unconfirmed stateowner!\n");
2964 return nfserr_bad_stateid;
1da177e4 2965 }
3a4f98bb 2966 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2967 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2968 " confirmed yet!\n");
2969 return nfserr_bad_stateid;
1da177e4 2970 }
0836f587
BF		 2971 status = check_stateid_generation(stateid, &stp->st_stateid);
2972 if (status)
2973 return status;
52fd004e 2974 renew_client(sop->so_client);
3a4f98bb 2975 return nfs_ok;
1da177e4 2976
1da177e4 2977check_replay:
bd9aac52 2978 if (seqid == sop->so_seqid - 1) {
849823c5 2979 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2980 /* indicate replay to calling function */
a90b061c 2981 return nfserr_replay_me;
1da177e4 2982 }
2fdada03 2983 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2984 sop->so_seqid, seqid);
2985 *sopp = NULL;
2986 return nfserr_bad_seqid;
1da177e4
LT		 2987}
2988
b37ad28b 2989__be32
ca364317 2990nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2991 struct nfsd4_open_confirm *oc)
1da177e4 2992{
b37ad28b 2993 __be32 status;
1da177e4
LT		 2994 struct nfs4_stateowner *sop;
2995 struct nfs4_stateid *stp;
2996
2997 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 2998 (int)cstate->current_fh.fh_dentry->d_name.len,
2999 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3000
ca364317 3001 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3002 if (status)
3003 return status;
1da177e4
LT		 3004
3005 nfs4_lock_state();
3006
ca364317
BF		 3007 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3008 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3009 CONFIRM | OPEN_STATE,
1da177e4
LT		 3010 &oc->oc_stateowner, &stp, NULL)))
3011 goto out;
3012
3013 sop = oc->oc_stateowner;
3014 sop->so_confirmed = 1;
3015 update_stateid(&stp->st_stateid);
3016 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3017 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3018 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3019 stp->st_stateid.si_boot,
3020 stp->st_stateid.si_stateownerid,
3021 stp->st_stateid.si_fileid,
3022 stp->st_stateid.si_generation);
c7b9a459
N		 3023
3024 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3025out:
f2327d9a 3026 if (oc->oc_stateowner) {
1da177e4 3027 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3028 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3029 }
1da177e4
LT		 3030 nfs4_unlock_state();
3031 return status;
3032}
3033
3034
3035/*
3036 * unset all bits in union bitmap (bmap) that
3037 * do not exist in share (from successful OPEN_DOWNGRADE)
3038 */
3039static void
3040reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3041{
3042 int i;
3043 for (i = 1; i < 4; i++) {
3044 if ((i & access) != i)
3045 __clear_bit(i, bmap);
3046 }
3047}
3048
3049static void
3050reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3051{
3052 int i;
3053 for (i = 0; i < 4; i++) {
3054 if ((i & deny) != i)
3055 __clear_bit(i, bmap);
3056 }
3057}
3058
b37ad28b 3059__be32
ca364317
BF		 3060nfsd4_open_downgrade(struct svc_rqst *rqstp,
3061 struct nfsd4_compound_state *cstate,
a4f1706a 3062 struct nfsd4_open_downgrade *od)
1da177e4 3063{
b37ad28b 3064 __be32 status;
1da177e4
LT		 3065 struct nfs4_stateid *stp;
3066 unsigned int share_access;
3067
3068 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3069 (int)cstate->current_fh.fh_dentry->d_name.len,
3070 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3071
ba5a6a19
BF		 3072 if (!access_valid(od->od_share_access)
3073 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3074 return nfserr_inval;
3075
3076 nfs4_lock_state();
ca364317
BF		 3077 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3078 od->od_seqid,
1da177e4 3079 &od->od_stateid,
f3362737 3080 OPEN_STATE,
1da177e4
LT		 3081 &od->od_stateowner, &stp, NULL)))
3082 goto out;
3083
3084 status = nfserr_inval;
3085 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3086 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3087 stp->st_access_bmap, od->od_share_access);
3088 goto out;
3089 }
3090 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3091 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3092 stp->st_deny_bmap, od->od_share_deny);
3093 goto out;
3094 }
3095 set_access(&share_access, stp->st_access_bmap);
3096 nfs4_file_downgrade(stp->st_vfs_file,
3097 share_access & ~od->od_share_access);
3098
3099 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3100 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3101
3102 update_stateid(&stp->st_stateid);
3103 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3104 status = nfs_ok;
3105out:
f2327d9a 3106 if (od->od_stateowner) {
1da177e4 3107 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3108 cstate->replay_owner = od->od_stateowner;
f2327d9a 3109 }
1da177e4
LT		 3110 nfs4_unlock_state();
3111 return status;
3112}
3113
3114/*
3115 * nfs4_unlock_state() called after encode
3116 */
b37ad28b 3117__be32
ca364317 3118nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3119 struct nfsd4_close *close)
1da177e4 3120{
b37ad28b 3121 __be32 status;
1da177e4
LT		 3122 struct nfs4_stateid *stp;
3123
3124 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3125 (int)cstate->current_fh.fh_dentry->d_name.len,
3126 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3127
3128 nfs4_lock_state();
3129 /* check close_lru for replay */
ca364317
BF		 3130 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
3131 close->cl_seqid,
1da177e4 3132 &close->cl_stateid,
f3362737 3133 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3134 &close->cl_stateowner, &stp, NULL)))
3135 goto out;
1da177e4
LT		 3136 status = nfs_ok;
3137 update_stateid(&stp->st_stateid);
3138 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3139
04ef5954 3140 /* release_stateid() calls nfsd_close() if needed */
2283963f 3141 release_open_stateid(stp);
04ef5954
BF		 3142
3143 /* place unused nfs4_stateowners on so_close_lru list to be
3144 * released by the laundromat service after the lease period
3145 * to enable us to handle CLOSE replay
3146 */
3147 if (list_empty(&close->cl_stateowner->so_stateids))
3148 move_to_close_lru(close->cl_stateowner);
1da177e4 3149out:
f2327d9a 3150 if (close->cl_stateowner) {
1da177e4 3151 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3152 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3153 }
1da177e4
LT		 3154 nfs4_unlock_state();
3155 return status;
3156}
3157
b37ad28b 3158__be32
ca364317
BF		 3159nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3160 struct nfsd4_delegreturn *dr)
1da177e4 3161{
203a8c8e
BF		 3162 struct nfs4_delegation *dp;
3163 stateid_t *stateid = &dr->dr_stateid;
3164 struct inode *inode;
b37ad28b 3165 __be32 status;
1da177e4 3166
ca364317 3167 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3168 return status;
3169 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3170
3171 nfs4_lock_state();
203a8c8e
BF		 3172 status = nfserr_bad_stateid;
3173 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3174 goto out;
3175 status = nfserr_stale_stateid;
3176 if (STALE_STATEID(stateid))
3177 goto out;
7e0f7cf5 3178 status = nfserr_bad_stateid;
203a8c8e
BF		 3179 if (!is_delegation_stateid(stateid))
3180 goto out;
203a8c8e
BF		 3181 dp = find_delegation_stateid(inode, stateid);
3182 if (!dp)
3183 goto out;
3184 status = check_stateid_generation(stateid, &dp->dl_stateid);
3185 if (status)
3186 goto out;
3187 renew_client(dp->dl_client);
3188
3189 unhash_delegation(dp);
1da177e4 3190out:
203a8c8e
BF		 3191 nfs4_unlock_state();
3192
1da177e4
LT		 3193 return status;
3194}
3195
3196
3197/*
3198 * Lock owner state (byte-range locks)
3199 */
3200#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3201#define LOCK_HASH_BITS 8
3202#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3203#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3204
87df4de8
BH		 3205static inline u64
3206end_offset(u64 start, u64 len)
3207{
3208 u64 end;
3209
3210 end = start + len;
3211 return end >= start ? end: NFS4_MAX_UINT64;
3212}
3213
3214/* last octet in a range */
3215static inline u64
3216last_byte_offset(u64 start, u64 len)
3217{
3218 u64 end;
3219
3220 BUG_ON(!len);
3221 end = start + len;
3222 return end > start ? end - 1: NFS4_MAX_UINT64;
3223}
3224
1da177e4
LT		 3225#define lockownerid_hashval(id) \
3226 ((id) & LOCK_HASH_MASK)
3227
3228static inline unsigned int
3229lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3230 struct xdr_netobj *ownername)
3231{
3232 return (file_hashval(inode) + cl_id
3233 + opaque_hashval(ownername->data, ownername->len))
3234 & LOCK_HASH_MASK;
3235}
3236
3237static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3238static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3239static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3240
fd39ca9a 3241static struct nfs4_stateid *
1da177e4
LT		 3242find_stateid(stateid_t *stid, int flags)
3243{
9346eff0 3244 struct nfs4_stateid *local;
1da177e4
LT		 3245 u32 st_id = stid->si_stateownerid;
3246 u32 f_id = stid->si_fileid;
3247 unsigned int hashval;
3248
3249 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3250 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3251 hashval = stateid_hashval(st_id, f_id);
3252 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3253 if ((local->st_stateid.si_stateownerid == st_id) &&
3254 (local->st_stateid.si_fileid == f_id))
3255 return local;
3256 }
3257 }
9346eff0
KK		 3258
3259 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3260 hashval = stateid_hashval(st_id, f_id);
3261 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3262 if ((local->st_stateid.si_stateownerid == st_id) &&
3263 (local->st_stateid.si_fileid == f_id))
3264 return local;
3265 }
849823c5 3266 }
1da177e4
LT		 3267 return NULL;
3268}
3269
3270static struct nfs4_delegation *
3271find_delegation_stateid(struct inode *ino, stateid_t *stid)
3272{
13cd2184
N		 3273 struct nfs4_file *fp;
3274 struct nfs4_delegation *dl;
1da177e4
LT		 3275
3276 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3277 stid->si_boot, stid->si_stateownerid,
3278 stid->si_fileid, stid->si_generation);
3279
1da177e4 3280 fp = find_file(ino);
13cd2184
N		 3281 if (!fp)
3282 return NULL;
3283 dl = find_delegation_file(fp, stid);
3284 put_nfs4_file(fp);
3285 return dl;
1da177e4
LT		 3286}
3287
3288/*
3289 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3290 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3291 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3292 * locking, this prevents us from being completely protocol-compliant. The
3293 * real solution to this problem is to start using unsigned file offsets in
3294 * the VFS, but this is a very deep change!
3295 */
3296static inline void
3297nfs4_transform_lock_offset(struct file_lock *lock)
3298{
3299 if (lock->fl_start < 0)
3300 lock->fl_start = OFFSET_MAX;
3301 if (lock->fl_end < 0)
3302 lock->fl_end = OFFSET_MAX;
3303}
3304
d5b9026a
N		 3305/* Hack!: For now, we're defining this just so we can use a pointer to it
3306 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3307static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3308};
1da177e4
LT		 3309
3310static inline void
3311nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3312{
d5b9026a
N		 3313 struct nfs4_stateowner *sop;
3314 unsigned int hval;
1da177e4 3315
d5b9026a
N		 3316 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3317 sop = (struct nfs4_stateowner *) fl->fl_owner;
3318 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3319 kref_get(&sop->so_ref);
3320 deny->ld_sop = sop;
3321 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3322 } else {
3323 deny->ld_sop = NULL;
3324 deny->ld_clientid.cl_boot = 0;
3325 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3326 }
3327 deny->ld_start = fl->fl_start;
87df4de8
BH		 3328 deny->ld_length = NFS4_MAX_UINT64;
3329 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3330 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3331 deny->ld_type = NFS4_READ_LT;
3332 if (fl->fl_type != F_RDLCK)
3333 deny->ld_type = NFS4_WRITE_LT;
3334}
3335
1da177e4
LT		 3336static struct nfs4_stateowner *
3337find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3338 struct xdr_netobj *owner)
3339{
3340 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3341 struct nfs4_stateowner *op;
3342
3343 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3344 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3345 return op;
3346 }
3347 return NULL;
3348}
3349
3350/*
3351 * Alloc a lock owner structure.
3352 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3353 * occured.
3354 *
3355 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3356 */
3357
3358static struct nfs4_stateowner *
3359alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3360 struct nfs4_stateowner *sop;
3361 struct nfs4_replay *rp;
3362 unsigned int idhashval;
3363
3364 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3365 return NULL;
3366 idhashval = lockownerid_hashval(current_ownerid);
3367 INIT_LIST_HEAD(&sop->so_idhash);
3368 INIT_LIST_HEAD(&sop->so_strhash);
3369 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3370 INIT_LIST_HEAD(&sop->so_stateids);
3371 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3372 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3373 sop->so_time = 0;
3374 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3375 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3376 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3377 sop->so_is_open_owner = 0;
3378 sop->so_id = current_ownerid++;
3379 sop->so_client = clp;
b59e3c0e
NB		 3380 /* It is the openowner seqid that will be incremented in encode in the
3381 * case of new lockowners; so increment the lock seqid manually: */
3382 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3383 sop->so_confirmed = 1;
3384 rp = &sop->so_replay;
de1ae286 3385 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3386 rp->rp_buflen = 0;
3387 rp->rp_buf = rp->rp_ibuf;
3388 return sop;
3389}
3390
fd39ca9a 3391static struct nfs4_stateid *
1da177e4
LT		 3392alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3393{
3394 struct nfs4_stateid *stp;
3395 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3396
5ac049ac
N		 3397 stp = nfs4_alloc_stateid();
3398 if (stp == NULL)
1da177e4
LT		 3399 goto out;
3400 INIT_LIST_HEAD(&stp->st_hash);
3401 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3402 INIT_LIST_HEAD(&stp->st_perstateowner);
3403 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3404 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3405 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3406 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3407 stp->st_stateowner = sop;
13cd2184 3408 get_nfs4_file(fp);
1da177e4
LT		 3409 stp->st_file = fp;
3410 stp->st_stateid.si_boot = boot_time;
3411 stp->st_stateid.si_stateownerid = sop->so_id;
3412 stp->st_stateid.si_fileid = fp->fi_id;
3413 stp->st_stateid.si_generation = 0;
3414 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3415 stp->st_access_bmap = open_stp->st_access_bmap;
3416 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3417 stp->st_openstp = open_stp;
1da177e4
LT		 3418
3419out:
3420 return stp;
3421}
3422
fd39ca9a 3423static int
1da177e4
LT		 3424check_lock_length(u64 offset, u64 length)
3425{
87df4de8 3426 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3427 LOFF_OVERFLOW(offset, length)));
3428}
3429
3430/*
3431 * LOCK operation
3432 */
b37ad28b 3433__be32
ca364317 3434nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3435 struct nfsd4_lock *lock)
1da177e4 3436{
3e9e3dbe 3437 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3438 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3439 struct nfs4_stateid *lock_stp;
3440 struct file *filp;
3441 struct file_lock file_lock;
8dc7c311 3442 struct file_lock conflock;
b37ad28b 3443 __be32 status = 0;
1da177e4 3444 unsigned int strhashval;
150b3934 3445 unsigned int cmd;
b8dd7b9a 3446 int err;
1da177e4
LT		 3447
3448 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3449 (long long) lock->lk_offset,
3450 (long long) lock->lk_length);
3451
1da177e4
LT		 3452 if (check_lock_length(lock->lk_offset, lock->lk_length))
3453 return nfserr_inval;
3454
ca364317 3455 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3456 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3457 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3458 return status;
3459 }
3460
1da177e4
LT		 3461 nfs4_lock_state();
3462
3463 if (lock->lk_is_new) {
893f8770
N		 3464 /*
3465 * Client indicates that this is a new lockowner.
3466 * Use open owner and open stateid to create lock owner and
3467 * lock stateid.
3468 */
1da177e4
LT		 3469 struct nfs4_stateid *open_stp = NULL;
3470 struct nfs4_file *fp;
3471
3472 status = nfserr_stale_clientid;
849823c5 3473 if (STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3474 goto out;
1da177e4 3475
1da177e4 3476 /* validate and update open stateid and open seqid */
ca364317 3477 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3478 lock->lk_new_open_seqid,
3479 &lock->lk_new_open_stateid,
f3362737 3480 OPEN_STATE,
3a65588a 3481 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3482 lock);
37515177 3483 if (status)
1da177e4 3484 goto out;
3a65588a 3485 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3486 /* create lockowner and lock stateid */
3487 fp = open_stp->st_file;
3488 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3489 open_sop->so_client->cl_clientid.cl_id,
3490 &lock->v.new.owner);
3e9e3dbe
N		 3491 /* XXX: Do we need to check for duplicate stateowners on
3492 * the same file, or should they just be allowed (and
3493 * create new stateids)? */
1da177e4 3494 status = nfserr_resource;
b59e3c0e
NB		 3495 lock_sop = alloc_init_lock_stateowner(strhashval,
3496 open_sop->so_client, open_stp, lock);
3497 if (lock_sop == NULL)
1da177e4 3498 goto out;
b59e3c0e 3499 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3500 if (lock_stp == NULL)
1da177e4 3501 goto out;
1da177e4
LT		 3502 } else {
3503 /* lock (lock owner + lock stateid) already exists */
ca364317 3504 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3505 lock->lk_old_lock_seqid,
3506 &lock->lk_old_lock_stateid,
f3362737 3507 LOCK_STATE,
3a65588a 3508 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3509 if (status)
3510 goto out;
3a65588a 3511 lock_sop = lock->lk_replay_owner;
1da177e4 3512 }
3a65588a 3513 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3514 filp = lock_stp->st_vfs_file;
3515
0dd395dc 3516 status = nfserr_grace;
af558e33 3517 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3518 goto out;
3519 status = nfserr_no_grace;
af558e33 3520 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3521 goto out;
3522
1da177e4
LT		 3523 locks_init_lock(&file_lock);
3524 switch (lock->lk_type) {
3525 case NFS4_READ_LT:
3526 case NFS4_READW_LT:
3527 file_lock.fl_type = F_RDLCK;
150b3934 3528 cmd = F_SETLK;
1da177e4
LT		 3529 break;
3530 case NFS4_WRITE_LT:
3531 case NFS4_WRITEW_LT:
3532 file_lock.fl_type = F_WRLCK;
150b3934 3533 cmd = F_SETLK;
1da177e4
LT		 3534 break;
3535 default:
3536 status = nfserr_inval;
3537 goto out;
3538 }
b59e3c0e 3539 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3540 file_lock.fl_pid = current->tgid;
3541 file_lock.fl_file = filp;
3542 file_lock.fl_flags = FL_POSIX;
d5b9026a 3543 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3544
3545 file_lock.fl_start = lock->lk_offset;
87df4de8 3546 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3547 nfs4_transform_lock_offset(&file_lock);
3548
3549 /*
3550 * Try to lock the file in the VFS.
3551 * Note: locks.c uses the BKL to protect the inode's lock list.
3552 */
3553
fd85b817 3554 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3555 switch (-err) {
1da177e4
LT		 3556 case 0: /* success! */
3557 update_stateid(&lock_stp->st_stateid);
3558 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3559 sizeof(stateid_t));
b8dd7b9a 3560 status = 0;
eb76b3fd
AA		 3561 break;
3562 case (EAGAIN): /* conflock holds conflicting lock */
3563 status = nfserr_denied;
3564 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3565 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3566 break;
1da177e4
LT		 3567 case (EDEADLK):
3568 status = nfserr_deadlock;
eb76b3fd 3569 break;
1da177e4 3570 default:
fd85b817 3571 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3572 status = nfserr_resource;
3573 break;
1da177e4 3574 }
1da177e4 3575out:
8a280510 3576 if (status && lock->lk_is_new && lock_sop)
f044ff83 3577 release_lockowner(lock_sop);
3a65588a
BF		 3578 if (lock->lk_replay_owner) {
3579 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3580 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3581 }
1da177e4
LT		 3582 nfs4_unlock_state();
3583 return status;
3584}
3585
55ef1274
BF		 3586/*
3587 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3588 * so we do a temporary open here just to get an open file to pass to
3589 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3590 * inode operation.)
3591 */
3592static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3593{
3594 struct file *file;
3595 int err;
3596
3597 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3598 if (err)
3599 return err;
3600 err = vfs_test_lock(file, lock);
3601 nfsd_close(file);
3602 return err;
3603}
3604
1da177e4
LT		 3605/*
3606 * LOCKT operation
3607 */
b37ad28b 3608__be32
ca364317
BF		 3609nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3610 struct nfsd4_lockt *lockt)
1da177e4
LT		 3611{
3612 struct inode *inode;
1da177e4 3613 struct file_lock file_lock;
fd85b817 3614 int error;
b37ad28b 3615 __be32 status;
1da177e4 3616
af558e33 3617 if (locks_in_grace())
1da177e4
LT		 3618 return nfserr_grace;
3619
3620 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3621 return nfserr_inval;
3622
3623 lockt->lt_stateowner = NULL;
3624 nfs4_lock_state();
3625
3626 status = nfserr_stale_clientid;
849823c5 3627 if (STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3628 goto out;
1da177e4 3629
ca364317 3630 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3631 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3632 if (status == nfserr_symlink)
3633 status = nfserr_inval;
3634 goto out;
3635 }
3636
ca364317 3637 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3638 locks_init_lock(&file_lock);
3639 switch (lockt->lt_type) {
3640 case NFS4_READ_LT:
3641 case NFS4_READW_LT:
3642 file_lock.fl_type = F_RDLCK;
3643 break;
3644 case NFS4_WRITE_LT:
3645 case NFS4_WRITEW_LT:
3646 file_lock.fl_type = F_WRLCK;
3647 break;
3648 default:
2fdada03 3649 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3650 status = nfserr_inval;
3651 goto out;
3652 }
3653
3654 lockt->lt_stateowner = find_lockstateowner_str(inode,
3655 &lockt->lt_clientid, &lockt->lt_owner);
3656 if (lockt->lt_stateowner)
3657 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3658 file_lock.fl_pid = current->tgid;
3659 file_lock.fl_flags = FL_POSIX;
3660
3661 file_lock.fl_start = lockt->lt_offset;
87df4de8 3662 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3663
3664 nfs4_transform_lock_offset(&file_lock);
3665
1da177e4 3666 status = nfs_ok;
55ef1274 3667 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3668 if (error) {
3669 status = nfserrno(error);
3670 goto out;
3671 }
9d6a8c5c 3672 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3673 status = nfserr_denied;
9d6a8c5c 3674 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3675 }
3676out:
3677 nfs4_unlock_state();
3678 return status;
3679}
3680
b37ad28b 3681__be32
ca364317 3682nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3683 struct nfsd4_locku *locku)
1da177e4
LT		 3684{
3685 struct nfs4_stateid *stp;
3686 struct file *filp = NULL;
3687 struct file_lock file_lock;
b37ad28b 3688 __be32 status;
b8dd7b9a 3689 int err;
1da177e4
LT		 3690
3691 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3692 (long long) locku->lu_offset,
3693 (long long) locku->lu_length);
3694
3695 if (check_lock_length(locku->lu_offset, locku->lu_length))
3696 return nfserr_inval;
3697
3698 nfs4_lock_state();
3699
ca364317 3700 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT		 3701 locku->lu_seqid,
3702 &locku->lu_stateid,
f3362737 3703 LOCK_STATE,
1da177e4
LT		 3704 &locku->lu_stateowner, &stp, NULL)))
3705 goto out;
3706
3707 filp = stp->st_vfs_file;
3708 BUG_ON(!filp);
3709 locks_init_lock(&file_lock);
3710 file_lock.fl_type = F_UNLCK;
3711 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3712 file_lock.fl_pid = current->tgid;
3713 file_lock.fl_file = filp;
3714 file_lock.fl_flags = FL_POSIX;
d5b9026a 3715 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3716 file_lock.fl_start = locku->lu_offset;
3717
87df4de8 3718 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3719 nfs4_transform_lock_offset(&file_lock);
3720
3721 /*
3722 * Try to unlock the file in the VFS.
3723 */
fd85b817 3724 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3725 if (err) {
fd85b817 3726 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3727 goto out_nfserr;
3728 }
3729 /*
3730 * OK, unlock succeeded; the only thing left to do is update the stateid.
3731 */
3732 update_stateid(&stp->st_stateid);
3733 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3734
3735out:
f2327d9a 3736 if (locku->lu_stateowner) {
1da177e4 3737 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3738 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3739 }
1da177e4
LT		 3740 nfs4_unlock_state();
3741 return status;
3742
3743out_nfserr:
b8dd7b9a 3744 status = nfserrno(err);
1da177e4
LT		 3745 goto out;
3746}
3747
3748/*
3749 * returns
3750 * 1: locks held by lockowner
3751 * 0: no locks held by lockowner
3752 */
3753static int
3754check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3755{
3756 struct file_lock **flpp;
7eaa36e2 3757 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3758 int status = 0;
3759
3760 lock_kernel();
3761 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3762 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3763 status = 1;
3764 goto out;
796dadfd 3765 }
1da177e4
LT		 3766 }
3767out:
3768 unlock_kernel();
3769 return status;
3770}
3771
b37ad28b 3772__be32
b591480b
BF		 3773nfsd4_release_lockowner(struct svc_rqst *rqstp,
3774 struct nfsd4_compound_state *cstate,
3775 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3776{
3777 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3778 struct nfs4_stateowner *sop;
3779 struct nfs4_stateid *stp;
1da177e4 3780 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3781 struct list_head matches;
3782 int i;
b37ad28b 3783 __be32 status;
1da177e4
LT		 3784
3785 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3786 clid->cl_boot, clid->cl_id);
3787
3788 /* XXX check for lease expiration */
3789
3790 status = nfserr_stale_clientid;
849823c5 3791 if (STALE_CLIENTID(clid))
1da177e4 3792 return status;
1da177e4
LT		 3793
3794 nfs4_lock_state();
3795
3e9e3dbe
N		 3796 status = nfserr_locks_held;
3797 /* XXX: we're doing a linear search through all the lockowners.
3798 * Yipes! For now we'll just hope clients aren't really using
3799 * release_lockowner much, but eventually we have to fix these
3800 * data structures. */
3801 INIT_LIST_HEAD(&matches);
3802 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3803 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3804 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3805 continue;
3806 list_for_each_entry(stp, &sop->so_stateids,
3807 st_perstateowner) {
3808 if (check_for_locks(stp->st_vfs_file, sop))
3809 goto out;
3810 /* Note: so_perclient unused for lockowners,
3811 * so it's OK to fool with here. */
3812 list_add(&sop->so_perclient, &matches);
3813 }
1da177e4 3814 }
3e9e3dbe
N		 3815 }
3816 /* Clients probably won't expect us to return with some (but not all)
3817 * of the lockowner state released; so don't release any until all
3818 * have been checked. */
3819 status = nfs_ok;
0fa822e4
N		 3820 while (!list_empty(&matches)) {
3821 sop = list_entry(matches.next, struct nfs4_stateowner,
3822 so_perclient);
3823 /* unhash_stateowner deletes so_perclient only
3824 * for openowners. */
3825 list_del(&sop->so_perclient);
f044ff83 3826 release_lockowner(sop);
1da177e4
LT		 3827 }
3828out:
3829 nfs4_unlock_state();
3830 return status;
3831}
3832
3833static inline struct nfs4_client_reclaim *
a55370a3 3834alloc_reclaim(void)
1da177e4 3835{
a55370a3 3836 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3837}
3838
c7b9a459 3839int
a1bcecd2 3840nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3841{
3842 unsigned int strhashval = clientstr_hashval(name);
3843 struct nfs4_client *clp;
3844
a1bcecd2 3845 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3846 return clp ? 1 : 0;
3847}
3848
1da177e4
LT		 3849/*
3850 * failure => all reset bets are off, nfserr_no_grace...
3851 */
190e4fbf
N		 3852int
3853nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3854{
3855 unsigned int strhashval;
3856 struct nfs4_client_reclaim *crp = NULL;
3857
a55370a3
N		 3858 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3859 crp = alloc_reclaim();
1da177e4
LT		 3860 if (!crp)
3861 return 0;
a55370a3 3862 strhashval = clientstr_hashval(name);
1da177e4
LT		 3863 INIT_LIST_HEAD(&crp->cr_strhash);
3864 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3865 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3866 reclaim_str_hashtbl_size++;
3867 return 1;
3868}
3869
3870static void
3871nfs4_release_reclaim(void)
3872{
3873 struct nfs4_client_reclaim *crp = NULL;
3874 int i;
3875
1da177e4
LT		 3876 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3877 while (!list_empty(&reclaim_str_hashtbl[i])) {
3878 crp = list_entry(reclaim_str_hashtbl[i].next,
3879 struct nfs4_client_reclaim, cr_strhash);
3880 list_del(&crp->cr_strhash);
1da177e4
LT		 3881 kfree(crp);
3882 reclaim_str_hashtbl_size--;
3883 }
3884 }
3885 BUG_ON(reclaim_str_hashtbl_size);
3886}
3887
3888/*
3889 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3890static struct nfs4_client_reclaim *
1da177e4
LT		 3891nfs4_find_reclaim_client(clientid_t *clid)
3892{
3893 unsigned int strhashval;
3894 struct nfs4_client *clp;
3895 struct nfs4_client_reclaim *crp = NULL;
3896
3897
3898 /* find clientid in conf_id_hashtbl */
3899 clp = find_confirmed_client(clid);
3900 if (clp == NULL)
3901 return NULL;
3902
a55370a3
N		 3903 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3904 clp->cl_name.len, clp->cl_name.data,
3905 clp->cl_recdir);
1da177e4
LT		 3906
3907 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3908 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3909 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3910 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3911 return crp;
3912 }
3913 }
3914 return NULL;
3915}
3916
3917/*
3918* Called from OPEN. Look for clientid in reclaim list.
3919*/
b37ad28b 3920__be32
1da177e4
LT		 3921nfs4_check_open_reclaim(clientid_t *clid)
3922{
dfc83565 3923 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3924}
3925
ac4d8ff2 3926/* initialization to perform at module load time: */
1da177e4 3927
e8ff2a84 3928int
ac4d8ff2 3929nfs4_state_init(void)
1da177e4 3930{
e8ff2a84 3931 int i, status;
1da177e4 3932
e8ff2a84
BF		 3933 status = nfsd4_init_slabs();
3934 if (status)
3935 return status;
1da177e4
LT		 3936 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3937 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3938 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3939 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3940 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3941 }
5282fd72
ME		 3942 for (i = 0; i < SESSION_HASH_SIZE; i++)
3943 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3944 for (i = 0; i < FILE_HASH_SIZE; i++) {
3945 INIT_LIST_HEAD(&file_hashtbl[i]);
3946 }
3947 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3948 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3949 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3950 }
3951 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3952 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3953 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3954 }
3955 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3956 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3957 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3958 }
1da177e4 3959 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3960 INIT_LIST_HEAD(&close_lru);
3961 INIT_LIST_HEAD(&client_lru);
3962 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N		 3963 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3964 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3965 reclaim_str_hashtbl_size = 0;
e8ff2a84 3966 return 0;
ac4d8ff2
N		 3967}
3968
190e4fbf
N		 3969static void
3970nfsd4_load_reboot_recovery_data(void)
3971{
3972 int status;
3973
0964a3d3
N		 3974 nfs4_lock_state();
3975 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3976 status = nfsd4_recdir_load();
0964a3d3 3977 nfs4_unlock_state();
190e4fbf
N		 3978 if (status)
3979 printk("NFSD: Failure reading reboot recovery data\n");
3980}
3981
9a8db97e
ME		 3982unsigned long
3983get_nfs4_grace_period(void)
3984{
3985 return max(user_lease_time, lease_time) * HZ;
3986}
3987
c2f1a551
MS		 3988/*
3989 * Since the lifetime of a delegation isn't limited to that of an open, a
3990 * client may quite reasonably hang on to a delegation as long as it has
3991 * the inode cached. This becomes an obvious problem the first time a
3992 * client's inode cache approaches the size of the server's total memory.
3993 *
3994 * For now we avoid this problem by imposing a hard limit on the number
3995 * of delegations, which varies according to the server's memory size.
3996 */
3997static void
3998set_max_delegations(void)
3999{
4000 /*
4001 * Allow at most 4 delegations per megabyte of RAM. Quick
4002 * estimates suggest that in the worst case (where every delegation
4003 * is for a different inode), a delegation could take about 1.5K,
4004 * giving a worst case usage of about 6% of memory.
4005 */
4006 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4007}
4008
ac4d8ff2
N		 4009/* initialization to perform when the nfsd service is started: */
4010
4011static void
4012__nfs4_state_start(void)
4013{
9a8db97e 4014 unsigned long grace_time;
ac4d8ff2 4015
1da177e4 4016 boot_time = get_seconds();
af558e33 4017 grace_time = get_nfs4_grace_period();
d99a05ad 4018 lease_time = user_lease_time;
af558e33 4019 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4020 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4021 grace_time/HZ);
58da282b 4022 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4023 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4024 set_max_delegations();
1da177e4
LT		 4025}
4026
e8ff2a84 4027void
76a3550e 4028nfs4_state_start(void)
1da177e4 4029{
1da177e4 4030 if (nfs4_init)
e8ff2a84 4031 return;
190e4fbf 4032 nfsd4_load_reboot_recovery_data();
76a3550e 4033 __nfs4_state_start();
1da177e4 4034 nfs4_init = 1;
e8ff2a84 4035 return;
1da177e4
LT		 4036}
4037
1da177e4
LT		 4038time_t
4039nfs4_lease_time(void)
4040{
4041 return lease_time;
4042}
4043
4044static void
4045__nfs4_state_shutdown(void)
4046{
4047 int i;
4048 struct nfs4_client *clp = NULL;
4049 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4050 struct list_head *pos, *next, reaplist;
4051
1da177e4
LT		 4052 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4053 while (!list_empty(&conf_id_hashtbl[i])) {
4054 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4055 expire_client(clp);
4056 }
4057 while (!list_empty(&unconf_str_hashtbl[i])) {
4058 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4059 expire_client(clp);
4060 }
4061 }
4062 INIT_LIST_HEAD(&reaplist);
4063 spin_lock(&recall_lock);
4064 list_for_each_safe(pos, next, &del_recall_lru) {
4065 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4066 list_move(&dp->dl_recall_lru, &reaplist);
4067 }
4068 spin_unlock(&recall_lock);
4069 list_for_each_safe(pos, next, &reaplist) {
4070 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4071 list_del_init(&dp->dl_recall_lru);
4072 unhash_delegation(dp);
4073 }
4074
190e4fbf 4075 nfsd4_shutdown_recdir();
1da177e4 4076 nfs4_init = 0;
1da177e4
LT		 4077}
4078
4079void
4080nfs4_state_shutdown(void)
4081{
5e8d5c29
N		 4082 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4083 destroy_workqueue(laundry_wq);
2c5e7615 4084 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4085 nfs4_lock_state();
4086 nfs4_release_reclaim();
4087 __nfs4_state_shutdown();
1da177e4
LT		 4088 nfs4_unlock_state();
4089}
4090
3dd98a3b
JL		 4091/*
4092 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4093 * accessed when nfsd is starting.
4094 */
0964a3d3
N		 4095static void
4096nfs4_set_recdir(char *recdir)
4097{
0964a3d3 4098 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4099}
4100
4101/*
4102 * Change the NFSv4 recovery directory to recdir.
4103 */
4104int
4105nfs4_reset_recoverydir(char *recdir)
4106{
4107 int status;
a63bb996 4108 struct path path;
0964a3d3 4109
a63bb996 4110 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4111 if (status)
4112 return status;
4113 status = -ENOTDIR;
a63bb996 4114 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4115 nfs4_set_recdir(recdir);
4116 status = 0;
4117 }
a63bb996 4118 path_put(&path);
0964a3d3
N		 4119 return status;
4120}
4121
3dd98a3b
JL		 4122char *
4123nfs4_recoverydir(void)
4124{
4125 return user_recovery_dirname;
4126}
4127
1da177e4
LT		 4128/*
4129 * Called when leasetime is changed.
4130 *
d99a05ad
N		 4131 * The only way the protocol gives us to handle on-the-fly lease changes is to
4132 * simulate a reboot. Instead of doing that, we just wait till the next time
4133 * we start to register any changes in lease time. If the administrator
4134 * really wants to change the lease time *now*, they can go ahead and bring
4135 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4136 *
4137 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4138 * when nfsd is starting
1da177e4
LT		 4139 */
4140void
4141nfs4_reset_lease(time_t leasetime)
4142{
d99a05ad 4143 user_lease_time = leasetime;
1da177e4 4144}
Linux 6.x block layer and io_uring tree(s)