projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4: create rpc callback client from server thread
[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;
78155ed7 209 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 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;
60adfc50
AA		 606 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
607 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 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);
1a69c179 1691 expire_client(unconf);
1da177e4 1692 status = nfs_ok;
1a69c179 1693
1da177e4 1694 }
f3aba4e5 1695 } else if (conf && !unconf) {
a186e767
BF		 1696 /*
1697 * RFC 3530 14.2.34 CASE 2:
1698 * probable retransmitted request; play it safe and
1699 * do nothing.
7c79f737 1700 */
599e0a22 1701 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1702 status = nfserr_clid_inuse;
21ab45a4 1703 else
1da177e4 1704 status = nfs_ok;
7c79f737 1705 } else if (!conf && unconf
599e0a22 1706 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1707 /*
1708 * RFC 3530 14.2.34 CASE 3:
1709 * Normal case; new or rebooted client:
7c79f737 1710 */
599e0a22 1711 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1712 status = nfserr_clid_inuse;
1713 } else {
1a69c179
N		 1714 unsigned int hash =
1715 clientstr_hashval(unconf->cl_recdir);
1716 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1717 hash, false);
1a69c179 1718 if (conf) {
c7b9a459 1719 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1720 expire_client(conf);
1721 }
1da177e4 1722 move_to_confirmed(unconf);
21ab45a4 1723 conf = unconf;
46f8a64b 1724 nfsd4_probe_callback(conf);
1a69c179 1725 status = nfs_ok;
1da177e4 1726 }
599e0a22
BF		 1727 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1728 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1729 &confirm)))) {
a186e767
BF		 1730 /*
1731 * RFC 3530 14.2.34 CASE 4:
1732 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1733 */
1da177e4 1734 status = nfserr_stale_clientid;
7c79f737 1735 } else {
08e8987c
N		 1736 /* check that we have hit one of the cases...*/
1737 status = nfserr_clid_inuse;
1738 }
1da177e4 1739out:
1da177e4
LT		 1740 nfs4_unlock_state();
1741 return status;
1742}
1743
1da177e4
LT		 1744/* OPEN Share state helper functions */
1745static inline struct nfs4_file *
1746alloc_init_file(struct inode *ino)
1747{
1748 struct nfs4_file *fp;
1749 unsigned int hashval = file_hashval(ino);
1750
e60d4398
N		 1751 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1752 if (fp) {
8b671b80 1753 atomic_set(&fp->fi_ref, 1);
1da177e4 1754 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1755 INIT_LIST_HEAD(&fp->fi_stateids);
1756 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1757 spin_lock(&recall_lock);
1da177e4 1758 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1759 spin_unlock(&recall_lock);
1da177e4
LT		 1760 fp->fi_inode = igrab(ino);
1761 fp->fi_id = current_fileid++;
47f9940c 1762 fp->fi_had_conflict = false;
1da177e4
LT		 1763 return fp;
1764 }
1765 return NULL;
1766}
1767
e60d4398 1768static void
e18b890b 1769nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1770{
e60d4398
N		 1771 if (*slab == NULL)
1772 return;
1a1d92c1 1773 kmem_cache_destroy(*slab);
e60d4398 1774 *slab = NULL;
1da177e4
LT		 1775}
1776
e8ff2a84 1777void
1da177e4
LT		 1778nfsd4_free_slabs(void)
1779{
e60d4398
N		 1780 nfsd4_free_slab(&stateowner_slab);
1781 nfsd4_free_slab(&file_slab);
5ac049ac 1782 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1783 nfsd4_free_slab(&deleg_slab);
e60d4398 1784}
1da177e4 1785
e60d4398
N		 1786static int
1787nfsd4_init_slabs(void)
1788{
1789 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1790 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1791 if (stateowner_slab == NULL)
1792 goto out_nomem;
1793 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1794 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1795 if (file_slab == NULL)
1796 goto out_nomem;
5ac049ac 1797 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1798 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1799 if (stateid_slab == NULL)
1800 goto out_nomem;
5b2d21c1 1801 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1802 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1803 if (deleg_slab == NULL)
1804 goto out_nomem;
e60d4398
N		 1805 return 0;
1806out_nomem:
1807 nfsd4_free_slabs();
1808 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1809 return -ENOMEM;
1da177e4
LT		 1810}
1811
1812void
1813nfs4_free_stateowner(struct kref *kref)
1814{
1815 struct nfs4_stateowner *sop =
1816 container_of(kref, struct nfs4_stateowner, so_ref);
1817 kfree(sop->so_owner.data);
1818 kmem_cache_free(stateowner_slab, sop);
1819}
1820
1821static inline struct nfs4_stateowner *
1822alloc_stateowner(struct xdr_netobj *owner)
1823{
1824 struct nfs4_stateowner *sop;
1825
1826 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1827 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1828 memcpy(sop->so_owner.data, owner->data, owner->len);
1829 sop->so_owner.len = owner->len;
1830 kref_init(&sop->so_ref);
1831 return sop;
1832 }
1833 kmem_cache_free(stateowner_slab, sop);
1834 }
1835 return NULL;
1836}
1837
1838static struct nfs4_stateowner *
1839alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1840 struct nfs4_stateowner *sop;
1841 struct nfs4_replay *rp;
1842 unsigned int idhashval;
1843
1844 if (!(sop = alloc_stateowner(&open->op_owner)))
1845 return NULL;
1846 idhashval = ownerid_hashval(current_ownerid);
1847 INIT_LIST_HEAD(&sop->so_idhash);
1848 INIT_LIST_HEAD(&sop->so_strhash);
1849 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1850 INIT_LIST_HEAD(&sop->so_stateids);
1851 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1852 INIT_LIST_HEAD(&sop->so_close_lru);
1853 sop->so_time = 0;
1854 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1855 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1856 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1857 sop->so_is_open_owner = 1;
1858 sop->so_id = current_ownerid++;
1859 sop->so_client = clp;
1860 sop->so_seqid = open->op_seqid;
1861 sop->so_confirmed = 0;
1862 rp = &sop->so_replay;
de1ae286 1863 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1864 rp->rp_buflen = 0;
1865 rp->rp_buf = rp->rp_ibuf;
1866 return sop;
1867}
1868
1da177e4
LT		 1869static inline void
1870init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1871 struct nfs4_stateowner *sop = open->op_stateowner;
1872 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1873
1874 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1875 INIT_LIST_HEAD(&stp->st_perstateowner);
1876 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1877 INIT_LIST_HEAD(&stp->st_perfile);
1878 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1879 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1880 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1881 stp->st_stateowner = sop;
13cd2184 1882 get_nfs4_file(fp);
1da177e4 1883 stp->st_file = fp;
78155ed7 1884 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1885 stp->st_stateid.si_stateownerid = sop->so_id;
1886 stp->st_stateid.si_fileid = fp->fi_id;
1887 stp->st_stateid.si_generation = 0;
1888 stp->st_access_bmap = 0;
1889 stp->st_deny_bmap = 0;
84459a11
AA		 1890 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1891 &stp->st_access_bmap);
1da177e4 1892 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1893 stp->st_openstp = NULL;
1da177e4
LT		 1894}
1895
fd39ca9a 1896static void
1da177e4
LT		 1897move_to_close_lru(struct nfs4_stateowner *sop)
1898{
1899 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1900
358dd55a 1901 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1902 sop->so_time = get_seconds();
1903}
1904
1da177e4 1905static int
599e0a22
BF		 1906same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1907 clientid_t *clid)
1908{
1909 return (sop->so_owner.len == owner->len) &&
1910 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1911 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1912}
1913
1914static struct nfs4_stateowner *
1915find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1916{
1917 struct nfs4_stateowner *so = NULL;
1918
1919 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1920 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1921 return so;
1922 }
1923 return NULL;
1924}
1925
1926/* search file_hashtbl[] for file */
1927static struct nfs4_file *
1928find_file(struct inode *ino)
1929{
1930 unsigned int hashval = file_hashval(ino);
1931 struct nfs4_file *fp;
1932
8b671b80 1933 spin_lock(&recall_lock);
1da177e4 1934 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1935 if (fp->fi_inode == ino) {
1936 get_nfs4_file(fp);
8b671b80 1937 spin_unlock(&recall_lock);
1da177e4 1938 return fp;
13cd2184 1939 }
1da177e4 1940 }
8b671b80 1941 spin_unlock(&recall_lock);
1da177e4
LT		 1942 return NULL;
1943}
1944
d87a8ade 1945static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1946{
d87a8ade 1947 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1948 return 0;
d87a8ade
AA		 1949 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1950 return 0;
1951 x &= ~NFS4_SHARE_ACCESS_MASK;
1952 if (minorversion && x) {
1953 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1954 return 0;
1955 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1956 return 0;
1957 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1958 }
1959 if (x)
8838dc43
BF		 1960 return 0;
1961 return 1;
ba5a6a19
BF		 1962}
1963
8838dc43 1964static inline int deny_valid(u32 x)
ba5a6a19 1965{
8838dc43
BF		 1966 /* Note: unlike access bits, deny bits may be zero. */
1967 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1968}
1da177e4 1969
4f83aa30
BF		 1970/*
1971 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1972 * st_{access,deny}_bmap field of the stateid, in order to track not
1973 * only what share bits are currently in force, but also what
1974 * combinations of share bits previous opens have used. This allows us
1975 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1976 * return an error if the client attempt to downgrade to a combination
1977 * of share bits not explicable by closing some of its previous opens.
1978 *
1979 * XXX: This enforcement is actually incomplete, since we don't keep
1980 * track of access/deny bit combinations; so, e.g., we allow:
1981 *
1982 * OPEN allow read, deny write
1983 * OPEN allow both, deny none
1984 * DOWNGRADE allow read, deny none
1985 *
1986 * which we should reject.
1987 */
fd39ca9a 1988static void
1da177e4
LT		 1989set_access(unsigned int *access, unsigned long bmap) {
1990 int i;
1991
1992 *access = 0;
1993 for (i = 1; i < 4; i++) {
1994 if (test_bit(i, &bmap))
1995 *access |= i;
1996 }
1997}
1998
fd39ca9a 1999static void
1da177e4
LT		 2000set_deny(unsigned int *deny, unsigned long bmap) {
2001 int i;
2002
2003 *deny = 0;
2004 for (i = 0; i < 4; i++) {
2005 if (test_bit(i, &bmap))
2006 *deny |= i ;
2007 }
2008}
2009
2010static int
2011test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2012 unsigned int access, deny;
2013
2014 set_access(&access, stp->st_access_bmap);
2015 set_deny(&deny, stp->st_deny_bmap);
2016 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2017 return 0;
2018 return 1;
2019}
2020
2021/*
2022 * Called to check deny when READ with all zero stateid or
2023 * WRITE with all zero or all one stateid
2024 */
b37ad28b 2025static __be32
1da177e4
LT		 2026nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2027{
2028 struct inode *ino = current_fh->fh_dentry->d_inode;
2029 struct nfs4_file *fp;
2030 struct nfs4_stateid *stp;
b37ad28b 2031 __be32 ret;
1da177e4
LT		 2032
2033 dprintk("NFSD: nfs4_share_conflict\n");
2034
2035 fp = find_file(ino);
13cd2184
N		 2036 if (!fp)
2037 return nfs_ok;
b700949b 2038 ret = nfserr_locked;
1da177e4 2039 /* Search for conflicting share reservations */
13cd2184
N		 2040 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2041 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2042 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2043 goto out;
1da177e4 2044 }
13cd2184
N		 2045 ret = nfs_ok;
2046out:
2047 put_nfs4_file(fp);
2048 return ret;
1da177e4
LT		 2049}
2050
2051static inline void
2052nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2053{
2054 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2055 drop_file_write_access(filp);
1da177e4
LT		 2056 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2057 }
2058}
2059
2060/*
2061 * Recall a delegation
2062 */
2063static int
2064do_recall(void *__dp)
2065{
2066 struct nfs4_delegation *dp = __dp;
2067
47f9940c 2068 dp->dl_file->fi_had_conflict = true;
1da177e4
LT		 2069 nfsd4_cb_recall(dp);
2070 return 0;
2071}
2072
2073/*
2074 * Spawn a thread to perform a recall on the delegation represented
2075 * by the lease (file_lock)
2076 *
2077 * Called from break_lease() with lock_kernel() held.
2078 * Note: we assume break_lease will only call this *once* for any given
2079 * lease.
2080 */
2081static
2082void nfsd_break_deleg_cb(struct file_lock *fl)
2083{
2084 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
2085 struct task_struct *t;
2086
2087 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2088 if (!dp)
2089 return;
2090
2091 /* We're assuming the state code never drops its reference
2092 * without first removing the lease. Since we're in this lease
2093 * callback (and since the lease code is serialized by the kernel
2094 * lock) we know the server hasn't removed the lease yet, we know
2095 * it's safe to take a reference: */
2096 atomic_inc(&dp->dl_count);
cfdcad4d 2097 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2098
2099 spin_lock(&recall_lock);
2100 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2101 spin_unlock(&recall_lock);
2102
2103 /* only place dl_time is set. protected by lock_kernel*/
2104 dp->dl_time = get_seconds();
2105
0272e1fd
BF		 2106 /*
2107 * We don't want the locks code to timeout the lease for us;
2108 * we'll remove it ourself if the delegation isn't returned
2109 * in time.
2110 */
2111 fl->fl_break_time = 0;
1da177e4
LT		 2112
2113 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
2114 if (IS_ERR(t)) {
2115 struct nfs4_client *clp = dp->dl_client;
2116
2117 printk(KERN_INFO "NFSD: Callback thread failed for "
2118 "for client (clientid %08x/%08x)\n",
2119 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 2120 put_nfs4_client(dp->dl_client);
1da177e4
LT		 2121 nfs4_put_delegation(dp);
2122 }
2123}
2124
2125/*
2126 * The file_lock is being reapd.
2127 *
2128 * Called by locks_free_lock() with lock_kernel() held.
2129 */
2130static
2131void nfsd_release_deleg_cb(struct file_lock *fl)
2132{
2133 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2134
2135 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2136
2137 if (!(fl->fl_flags & FL_LEASE) || !dp)
2138 return;
2139 dp->dl_flock = NULL;
2140}
2141
2142/*
2143 * Set the delegation file_lock back pointer.
2144 *
a9933cea 2145 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2146 */
2147static
2148void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2149{
2150 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2151
2152 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2153 if (!dp)
2154 return;
2155 dp->dl_flock = new;
2156}
2157
2158/*
a9933cea 2159 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2160 */
2161static
2162int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2163{
2164 struct nfs4_delegation *onlistd =
2165 (struct nfs4_delegation *)onlist->fl_owner;
2166 struct nfs4_delegation *tryd =
2167 (struct nfs4_delegation *)try->fl_owner;
2168
2169 if (onlist->fl_lmops != try->fl_lmops)
2170 return 0;
2171
2172 return onlistd->dl_client == tryd->dl_client;
2173}
2174
2175
2176static
2177int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2178{
2179 if (arg & F_UNLCK)
2180 return lease_modify(onlist, arg);
2181 else
2182 return -EAGAIN;
2183}
2184
fd39ca9a 2185static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2186 .fl_break = nfsd_break_deleg_cb,
2187 .fl_release_private = nfsd_release_deleg_cb,
2188 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2189 .fl_mylease = nfsd_same_client_deleg_cb,
2190 .fl_change = nfsd_change_deleg_cb,
2191};
2192
2193
b37ad28b 2194__be32
6668958f
AA		 2195nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2196 struct nfsd4_open *open)
1da177e4 2197{
1da177e4
LT		 2198 clientid_t *clientid = &open->op_clientid;
2199 struct nfs4_client *clp = NULL;
2200 unsigned int strhashval;
2201 struct nfs4_stateowner *sop = NULL;
2202
1da177e4 2203 if (!check_name(open->op_owner))
0f442aa2 2204 return nfserr_inval;
1da177e4
LT		 2205
2206 if (STALE_CLIENTID(&open->op_clientid))
2207 return nfserr_stale_clientid;
2208
2209 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2210 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2211 open->op_stateowner = sop;
2212 if (!sop) {
2213 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2214 clp = find_confirmed_client(clientid);
2215 if (clp == NULL)
0f442aa2
BF		 2216 return nfserr_expired;
2217 goto renew;
1da177e4 2218 }
6668958f
AA		 2219 /* When sessions are used, skip open sequenceid processing */
2220 if (nfsd4_has_session(cstate))
2221 goto renew;
0f442aa2
BF		 2222 if (!sop->so_confirmed) {
2223 /* Replace unconfirmed owners without checking for replay. */
2224 clp = sop->so_client;
f044ff83 2225 release_openowner(sop);
0f442aa2
BF		 2226 open->op_stateowner = NULL;
2227 goto renew;
2228 }
2229 if (open->op_seqid == sop->so_seqid - 1) {
2230 if (sop->so_replay.rp_buflen)
a90b061c 2231 return nfserr_replay_me;
0f442aa2
BF		 2232 /* The original OPEN failed so spectacularly
2233 * that we don't even have replay data saved!
2234 * Therefore, we have no choice but to continue
2235 * processing this OPEN; presumably, we'll
2236 * fail again for the same reason.
2237 */
2238 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2239 goto renew;
2240 }
2241 if (open->op_seqid != sop->so_seqid)
2242 return nfserr_bad_seqid;
1da177e4 2243renew:
0f442aa2
BF		 2244 if (open->op_stateowner == NULL) {
2245 sop = alloc_init_open_stateowner(strhashval, clp, open);
2246 if (sop == NULL)
2247 return nfserr_resource;
2248 open->op_stateowner = sop;
2249 }
2250 list_del_init(&sop->so_close_lru);
1da177e4 2251 renew_client(sop->so_client);
0f442aa2 2252 return nfs_ok;
1da177e4
LT		 2253}
2254
b37ad28b 2255static inline __be32
4a6e43e6
N		 2256nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2257{
2258 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2259 return nfserr_openmode;
2260 else
2261 return nfs_ok;
2262}
2263
52f4fb43
N		 2264static struct nfs4_delegation *
2265find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2266{
2267 struct nfs4_delegation *dp;
2268
ea1da636 2269 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2270 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2271 return dp;
2272 }
2273 return NULL;
2274}
2275
b37ad28b 2276static __be32
567d9829
N		 2277nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2278 struct nfs4_delegation **dp)
2279{
2280 int flags;
b37ad28b 2281 __be32 status = nfserr_bad_stateid;
567d9829
N		 2282
2283 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2284 if (*dp == NULL)
c44c5eeb 2285 goto out;
567d9829
N		 2286 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2287 RD_STATE : WR_STATE;
2288 status = nfs4_check_delegmode(*dp, flags);
2289 if (status)
2290 *dp = NULL;
c44c5eeb
N		 2291out:
2292 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2293 return nfs_ok;
2294 if (status)
2295 return status;
2296 open->op_stateowner->so_confirmed = 1;
2297 return nfs_ok;
567d9829
N		 2298}
2299
b37ad28b 2300static __be32
1da177e4
LT		 2301nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2302{
2303 struct nfs4_stateid *local;
b37ad28b 2304 __be32 status = nfserr_share_denied;
1da177e4
LT		 2305 struct nfs4_stateowner *sop = open->op_stateowner;
2306
8beefa24 2307 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2308 /* ignore lock owners */
2309 if (local->st_stateowner->so_is_open_owner == 0)
2310 continue;
2311 /* remember if we have seen this open owner */
2312 if (local->st_stateowner == sop)
2313 *stpp = local;
2314 /* check for conflicting share reservations */
2315 if (!test_share(local, open))
2316 goto out;
2317 }
2318 status = 0;
2319out:
2320 return status;
2321}
2322
5ac049ac
N		 2323static inline struct nfs4_stateid *
2324nfs4_alloc_stateid(void)
2325{
2326 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2327}
2328
b37ad28b 2329static __be32
1da177e4 2330nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2331 struct nfs4_delegation *dp,
1da177e4
LT		 2332 struct svc_fh *cur_fh, int flags)
2333{
2334 struct nfs4_stateid *stp;
1da177e4 2335
5ac049ac 2336 stp = nfs4_alloc_stateid();
1da177e4
LT		 2337 if (stp == NULL)
2338 return nfserr_resource;
2339
567d9829
N		 2340 if (dp) {
2341 get_file(dp->dl_vfs_file);
2342 stp->st_vfs_file = dp->dl_vfs_file;
2343 } else {
b37ad28b 2344 __be32 status;
567d9829
N		 2345 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2346 &stp->st_vfs_file);
2347 if (status) {
2348 if (status == nfserr_dropit)
2349 status = nfserr_jukebox;
5ac049ac 2350 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2351 return status;
2352 }
1da177e4 2353 }
1da177e4
LT		 2354 *stpp = stp;
2355 return 0;
2356}
2357
b37ad28b 2358static inline __be32
1da177e4
LT		 2359nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2360 struct nfsd4_open *open)
2361{
2362 struct iattr iattr = {
2363 .ia_valid = ATTR_SIZE,
2364 .ia_size = 0,
2365 };
2366 if (!open->op_truncate)
2367 return 0;
2368 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2369 return nfserr_inval;
1da177e4
LT		 2370 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2371}
2372
b37ad28b 2373static __be32
1da177e4
LT		 2374nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2375{
2376 struct file *filp = stp->st_vfs_file;
7eaa36e2 2377 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2378 unsigned int share_access, new_writer;
b37ad28b 2379 __be32 status;
1da177e4
LT		 2380
2381 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2382 new_writer = (~share_access) & open->op_share_access
2383 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2384
6c26d08f 2385 if (new_writer) {
b37ad28b
AV		 2386 int err = get_write_access(inode);
2387 if (err)
2388 return nfserrno(err);
e518f056
BH		 2389 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2390 if (err)
2391 return nfserrno(err);
2392 file_take_write(filp);
6c26d08f 2393 }
1da177e4
LT		 2394 status = nfsd4_truncate(rqstp, cur_fh, open);
2395 if (status) {
6c26d08f
BF		 2396 if (new_writer)
2397 put_write_access(inode);
1da177e4
LT		 2398 return status;
2399 }
2400 /* remember the open */
6c26d08f 2401 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2402 __set_bit(open->op_share_access, &stp->st_access_bmap);
2403 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2404
2405 return nfs_ok;
2406}
2407
2408
1da177e4 2409static void
37515177 2410nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2411{
37515177
N		 2412 open->op_stateowner->so_confirmed = 1;
2413 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2414}
2415
2416/*
2417 * Attempt to hand out a delegation.
2418 */
2419static void
2420nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2421{
2422 struct nfs4_delegation *dp;
2423 struct nfs4_stateowner *sop = stp->st_stateowner;
2424 struct nfs4_callback *cb = &sop->so_client->cl_callback;
2425 struct file_lock fl, *flp = &fl;
2426 int status, flag = 0;
2427
2428 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2429 open->op_recall = 0;
2430 switch (open->op_claim_type) {
2431 case NFS4_OPEN_CLAIM_PREVIOUS:
2432 if (!atomic_read(&cb->cb_set))
2433 open->op_recall = 1;
2434 flag = open->op_delegate_type;
2435 if (flag == NFS4_OPEN_DELEGATE_NONE)
2436 goto out;
2437 break;
2438 case NFS4_OPEN_CLAIM_NULL:
2439 /* Let's not give out any delegations till everyone's
2440 * had the chance to reclaim theirs.... */
af558e33 2441 if (locks_in_grace())
7b190fec
N		 2442 goto out;
2443 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2444 goto out;
2445 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2446 flag = NFS4_OPEN_DELEGATE_WRITE;
2447 else
2448 flag = NFS4_OPEN_DELEGATE_READ;
2449 break;
2450 default:
2451 goto out;
2452 }
1da177e4
LT		 2453
2454 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2455 if (dp == NULL) {
2456 flag = NFS4_OPEN_DELEGATE_NONE;
2457 goto out;
2458 }
2459 locks_init_lock(&fl);
2460 fl.fl_lmops = &nfsd_lease_mng_ops;
2461 fl.fl_flags = FL_LEASE;
9167f501 2462 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2463 fl.fl_end = OFFSET_MAX;
2464 fl.fl_owner = (fl_owner_t)dp;
2465 fl.fl_file = stp->st_vfs_file;
2466 fl.fl_pid = current->tgid;
2467
a9933cea 2468 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2469 * the lock_manager callbacks fl_mylease and fl_change are used
2470 */
9167f501 2471 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2472 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2473 unhash_delegation(dp);
1da177e4
LT		 2474 flag = NFS4_OPEN_DELEGATE_NONE;
2475 goto out;
2476 }
2477
2478 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2479
2480 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2481 dp->dl_stateid.si_boot,
2482 dp->dl_stateid.si_stateownerid,
2483 dp->dl_stateid.si_fileid,
2484 dp->dl_stateid.si_generation);
2485out:
7b190fec
N		 2486 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2487 && flag == NFS4_OPEN_DELEGATE_NONE
2488 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2489 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2490 open->op_delegate_type = flag;
2491}
2492
2493/*
2494 * called with nfs4_lock_state() held.
2495 */
b37ad28b 2496__be32
1da177e4
LT		 2497nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2498{
6668958f 2499 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2500 struct nfs4_file *fp = NULL;
2501 struct inode *ino = current_fh->fh_dentry->d_inode;
2502 struct nfs4_stateid *stp = NULL;
567d9829 2503 struct nfs4_delegation *dp = NULL;
b37ad28b 2504 __be32 status;
1da177e4
LT		 2505
2506 status = nfserr_inval;
d87a8ade 2507 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2508 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2509 goto out;
2510 /*
2511 * Lookup file; if found, lookup stateid and check open request,
2512 * and check for delegations in the process of being recalled.
2513 * If not found, create the nfs4_file struct
2514 */
2515 fp = find_file(ino);
2516 if (fp) {
2517 if ((status = nfs4_check_open(fp, open, &stp)))
2518 goto out;
c44c5eeb
N		 2519 status = nfs4_check_deleg(fp, open, &dp);
2520 if (status)
2521 goto out;
1da177e4 2522 } else {
c44c5eeb
N		 2523 status = nfserr_bad_stateid;
2524 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2525 goto out;
1da177e4
LT		 2526 status = nfserr_resource;
2527 fp = alloc_init_file(ino);
2528 if (fp == NULL)
2529 goto out;
2530 }
2531
2532 /*
2533 * OPEN the file, or upgrade an existing OPEN.
2534 * If truncate fails, the OPEN fails.
2535 */
2536 if (stp) {
2537 /* Stateid was found, this is an OPEN upgrade */
2538 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2539 if (status)
2540 goto out;
444c2c07 2541 update_stateid(&stp->st_stateid);
1da177e4
LT		 2542 } else {
2543 /* Stateid was not found, this is a new OPEN */
2544 int flags = 0;
9ecb6a08 2545 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2546 flags |= NFSD_MAY_READ;
1da177e4 2547 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2548 flags |= NFSD_MAY_WRITE;
567d9829
N		 2549 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2550 if (status)
1da177e4
LT		 2551 goto out;
2552 init_stateid(stp, fp, open);
2553 status = nfsd4_truncate(rqstp, current_fh, open);
2554 if (status) {
2283963f 2555 release_open_stateid(stp);
1da177e4
LT		 2556 goto out;
2557 }
6668958f
AA		 2558 if (nfsd4_has_session(&resp->cstate))
2559 update_stateid(&stp->st_stateid);
1da177e4
LT		 2560 }
2561 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2562
6668958f
AA		 2563 if (nfsd4_has_session(&resp->cstate))
2564 open->op_stateowner->so_confirmed = 1;
2565
1da177e4
LT		 2566 /*
2567 * Attempt to hand out a delegation. No error return, because the
2568 * OPEN succeeds even if we fail.
2569 */
2570 nfs4_open_delegation(current_fh, open, stp);
2571
2572 status = nfs_ok;
2573
2574 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2575 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2576 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2577out:
13cd2184
N		 2578 if (fp)
2579 put_nfs4_file(fp);
37515177
N		 2580 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2581 nfs4_set_claim_prev(open);
1da177e4
LT		 2582 /*
2583 * To finish the open response, we just need to set the rflags.
2584 */
2585 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2586 if (!open->op_stateowner->so_confirmed &&
2587 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2588 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2589
2590 return status;
2591}
2592
b37ad28b 2593__be32
b591480b
BF		 2594nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2595 clientid_t *clid)
1da177e4
LT		 2596{
2597 struct nfs4_client *clp;
b37ad28b 2598 __be32 status;
1da177e4
LT		 2599
2600 nfs4_lock_state();
2601 dprintk("process_renew(%08x/%08x): starting\n",
2602 clid->cl_boot, clid->cl_id);
2603 status = nfserr_stale_clientid;
2604 if (STALE_CLIENTID(clid))
2605 goto out;
2606 clp = find_confirmed_client(clid);
2607 status = nfserr_expired;
2608 if (clp == NULL) {
2609 /* We assume the client took too long to RENEW. */
2610 dprintk("nfsd4_renew: clientid not found!\n");
2611 goto out;
2612 }
2613 renew_client(clp);
2614 status = nfserr_cb_path_down;
ea1da636 2615 if (!list_empty(&clp->cl_delegations)
1da177e4
LT		 2616 && !atomic_read(&clp->cl_callback.cb_set))
2617 goto out;
2618 status = nfs_ok;
2619out:
2620 nfs4_unlock_state();
2621 return status;
2622}
2623
af558e33
BF		 2624struct lock_manager nfsd4_manager = {
2625};
2626
a76b4319 2627static void
af558e33 2628nfsd4_end_grace(void)
a76b4319
N		 2629{
2630 dprintk("NFSD: end of grace period\n");
c7b9a459 2631 nfsd4_recdir_purge_old();
af558e33 2632 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2633}
2634
fd39ca9a 2635static time_t
1da177e4
LT		 2636nfs4_laundromat(void)
2637{
2638 struct nfs4_client *clp;
2639 struct nfs4_stateowner *sop;
2640 struct nfs4_delegation *dp;
2641 struct list_head *pos, *next, reaplist;
2642 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2643 time_t t, clientid_val = NFSD_LEASE_TIME;
2644 time_t u, test_val = NFSD_LEASE_TIME;
2645
2646 nfs4_lock_state();
2647
2648 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2649 if (locks_in_grace())
2650 nfsd4_end_grace();
1da177e4
LT		 2651 list_for_each_safe(pos, next, &client_lru) {
2652 clp = list_entry(pos, struct nfs4_client, cl_lru);
2653 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2654 t = clp->cl_time - cutoff;
2655 if (clientid_val > t)
2656 clientid_val = t;
2657 break;
2658 }
2659 dprintk("NFSD: purging unused client (clientid %08x)\n",
2660 clp->cl_clientid.cl_id);
c7b9a459 2661 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2662 expire_client(clp);
2663 }
2664 INIT_LIST_HEAD(&reaplist);
2665 spin_lock(&recall_lock);
2666 list_for_each_safe(pos, next, &del_recall_lru) {
2667 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2668 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2669 u = dp->dl_time - cutoff;
2670 if (test_val > u)
2671 test_val = u;
2672 break;
2673 }
2674 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2675 dp, dp->dl_flock);
2676 list_move(&dp->dl_recall_lru, &reaplist);
2677 }
2678 spin_unlock(&recall_lock);
2679 list_for_each_safe(pos, next, &reaplist) {
2680 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2681 list_del_init(&dp->dl_recall_lru);
2682 unhash_delegation(dp);
2683 }
2684 test_val = NFSD_LEASE_TIME;
2685 list_for_each_safe(pos, next, &close_lru) {
2686 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2687 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2688 u = sop->so_time - cutoff;
2689 if (test_val > u)
2690 test_val = u;
2691 break;
2692 }
2693 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2694 sop->so_id);
f044ff83 2695 release_openowner(sop);
1da177e4
LT		 2696 }
2697 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2698 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2699 nfs4_unlock_state();
2700 return clientid_val;
2701}
2702
a254b246
HH		 2703static struct workqueue_struct *laundry_wq;
2704static void laundromat_main(struct work_struct *);
2705static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2706
2707static void
c4028958 2708laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2709{
2710 time_t t;
2711
2712 t = nfs4_laundromat();
2713 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2714 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2715}
2716
fd39ca9a 2717static struct nfs4_stateowner *
f8816512
N		 2718search_close_lru(u32 st_id, int flags)
2719{
1da177e4
LT		 2720 struct nfs4_stateowner *local = NULL;
2721
1da177e4
LT		 2722 if (flags & CLOSE_STATE) {
2723 list_for_each_entry(local, &close_lru, so_close_lru) {
2724 if (local->so_id == st_id)
2725 return local;
2726 }
2727 }
2728 return NULL;
2729}
2730
2731static inline int
2732nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2733{
7eaa36e2 2734 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2735}
2736
2737static int
2738STALE_STATEID(stateid_t *stateid)
2739{
78155ed7
BN		 2740 if (time_after((unsigned long)boot_time,
2741 (unsigned long)stateid->si_boot)) {
2742 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2743 stateid->si_boot, stateid->si_stateownerid,
2744 stateid->si_fileid, stateid->si_generation);
2745 return 1;
2746 }
2747 return 0;
2748}
2749
2750static int
2751EXPIRED_STATEID(stateid_t *stateid)
2752{
2753 if (time_before((unsigned long)boot_time,
2754 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2755 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2756 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2757 stateid->si_boot, stateid->si_stateownerid,
2758 stateid->si_fileid, stateid->si_generation);
2759 return 1;
2760 }
2761 return 0;
2762}
2763
2764static __be32
2765stateid_error_map(stateid_t *stateid)
2766{
2767 if (STALE_STATEID(stateid))
2768 return nfserr_stale_stateid;
2769 if (EXPIRED_STATEID(stateid))
2770 return nfserr_expired;
2771
2772 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2773 stateid->si_boot, stateid->si_stateownerid,
2774 stateid->si_fileid, stateid->si_generation);
2775 return nfserr_bad_stateid;
1da177e4
LT		 2776}
2777
2778static inline int
2779access_permit_read(unsigned long access_bmap)
2780{
2781 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2782 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2783 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2784}
2785
2786static inline int
2787access_permit_write(unsigned long access_bmap)
2788{
2789 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2790 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2791}
2792
2793static
b37ad28b 2794__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2795{
b37ad28b 2796 __be32 status = nfserr_openmode;
1da177e4
LT		 2797
2798 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2799 goto out;
2800 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2801 goto out;
2802 status = nfs_ok;
2803out:
2804 return status;
2805}
2806
b37ad28b 2807static inline __be32
1da177e4
LT		 2808check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2809{
203a8c8e 2810 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2811 return nfs_ok;
af558e33 2812 else if (locks_in_grace()) {
1da177e4
LT		 2813 /* Answer in remaining cases depends on existance of
2814 * conflicting state; so we must wait out the grace period. */
2815 return nfserr_grace;
2816 } else if (flags & WR_STATE)
2817 return nfs4_share_conflict(current_fh,
2818 NFS4_SHARE_DENY_WRITE);
2819 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2820 return nfs4_share_conflict(current_fh,
2821 NFS4_SHARE_DENY_READ);
2822}
2823
2824/*
2825 * Allow READ/WRITE during grace period on recovered state only for files
2826 * that are not able to provide mandatory locking.
2827 */
2828static inline int
18f82731 2829grace_disallows_io(struct inode *inode)
1da177e4 2830{
203a8c8e 2831 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2832}
2833
6668958f 2834static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2835{
6668958f
AA		 2836 /*
2837 * When sessions are used the stateid generation number is ignored
2838 * when it is zero.
2839 */
2840 if ((flags & HAS_SESSION) && in->si_generation == 0)
2841 goto out;
2842
0836f587
BF		 2843 /* If the client sends us a stateid from the future, it's buggy: */
2844 if (in->si_generation > ref->si_generation)
2845 return nfserr_bad_stateid;
2846 /*
2847 * The following, however, can happen. For example, if the
2848 * client sends an open and some IO at the same time, the open
2849 * may bump si_generation while the IO is still in flight.
2850 * Thanks to hard links and renames, the client never knows what
2851 * file an open will affect. So it could avoid that situation
2852 * only by serializing all opens and IO from the same open
2853 * owner. To recover from the old_stateid error, the client
2854 * will just have to retry the IO:
2855 */
2856 if (in->si_generation < ref->si_generation)
2857 return nfserr_old_stateid;
6668958f 2858out:
0836f587
BF		 2859 return nfs_ok;
2860}
2861
3e633079
BF		 2862static int is_delegation_stateid(stateid_t *stateid)
2863{
2864 return stateid->si_fileid == 0;
2865}
2866
1da177e4
LT		 2867/*
2868* Checks for stateid operations
2869*/
b37ad28b 2870__be32
dd453dfd
BH		 2871nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2872 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2873{
2874 struct nfs4_stateid *stp = NULL;
2875 struct nfs4_delegation *dp = NULL;
dd453dfd 2876 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2877 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2878 __be32 status;
1da177e4 2879
1da177e4
LT		 2880 if (filpp)
2881 *filpp = NULL;
2882
18f82731 2883 if (grace_disallows_io(ino))
1da177e4
LT		 2884 return nfserr_grace;
2885
6668958f
AA		 2886 if (nfsd4_has_session(cstate))
2887 flags |= HAS_SESSION;
2888
1da177e4
LT		 2889 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2890 return check_special_stateids(current_fh, stateid, flags);
2891
1da177e4
LT		 2892 status = nfserr_stale_stateid;
2893 if (STALE_STATEID(stateid))
2894 goto out;
2895
1da177e4 2896 status = nfserr_bad_stateid;
3e633079 2897 if (is_delegation_stateid(stateid)) {
a4455be0 2898 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2899 if (!dp) {
2900 status = stateid_error_map(stateid);
1da177e4 2901 goto out;
78155ed7 2902 }
6668958f
AA		 2903 status = check_stateid_generation(stateid, &dp->dl_stateid,
2904 flags);
0c2a498f
BF		 2905 if (status)
2906 goto out;
dc9bf700
BF		 2907 status = nfs4_check_delegmode(dp, flags);
2908 if (status)
2909 goto out;
2910 renew_client(dp->dl_client);
dc9bf700
BF		 2911 if (filpp)
2912 *filpp = dp->dl_vfs_file;
1da177e4 2913 } else { /* open or lock stateid */
a4455be0 2914 stp = find_stateid(stateid, flags);
78155ed7
BN		 2915 if (!stp) {
2916 status = stateid_error_map(stateid);
1da177e4 2917 goto out;
78155ed7 2918 }
6150ef0d 2919 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2920 goto out;
2921 if (!stp->st_stateowner->so_confirmed)
2922 goto out;
6668958f
AA		 2923 status = check_stateid_generation(stateid, &stp->st_stateid,
2924 flags);
0c2a498f
BF		 2925 if (status)
2926 goto out;
a4455be0
BF		 2927 status = nfs4_check_openmode(stp, flags);
2928 if (status)
1da177e4
LT		 2929 goto out;
2930 renew_client(stp->st_stateowner->so_client);
2931 if (filpp)
2932 *filpp = stp->st_vfs_file;
1da177e4
LT		 2933 }
2934 status = nfs_ok;
2935out:
2936 return status;
2937}
2938
4c4cd222
N		 2939static inline int
2940setlkflg (int type)
2941{
2942 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2943 RD_STATE : WR_STATE;
2944}
1da177e4
LT		 2945
2946/*
2947 * Checks for sequence id mutating operations.
2948 */
b37ad28b 2949static __be32
dd453dfd
BH		 2950nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2951 stateid_t *stateid, int flags,
2952 struct nfs4_stateowner **sopp,
2953 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2954{
1da177e4
LT		 2955 struct nfs4_stateid *stp;
2956 struct nfs4_stateowner *sop;
dd453dfd 2957 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2958 __be32 status;
1da177e4
LT		 2959
2960 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2961 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2962 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2963 stateid->si_generation);
3a4f98bb 2964
1da177e4
LT		 2965 *stpp = NULL;
2966 *sopp = NULL;
2967
1da177e4 2968 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2969 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2970 return nfserr_bad_stateid;
1da177e4
LT		 2971 }
2972
1da177e4 2973 if (STALE_STATEID(stateid))
3a4f98bb 2974 return nfserr_stale_stateid;
6668958f
AA		 2975
2976 if (nfsd4_has_session(cstate))
2977 flags |= HAS_SESSION;
2978
1da177e4
LT		 2979 /*
2980 * We return BAD_STATEID if filehandle doesn't match stateid,
2981 * the confirmed flag is incorrecly set, or the generation
2982 * number is incorrect.
1da177e4 2983 */
f8816512
N		 2984 stp = find_stateid(stateid, flags);
2985 if (stp == NULL) {
2986 /*
2987 * Also, we should make sure this isn't just the result of
2988 * a replayed close:
2989 */
2990 sop = search_close_lru(stateid->si_stateownerid, flags);
2991 if (sop == NULL)
78155ed7 2992 return stateid_error_map(stateid);
f8816512
N		 2993 *sopp = sop;
2994 goto check_replay;
2995 }
1da177e4 2996
39325bd0
BF		 2997 *stpp = stp;
2998 *sopp = sop = stp->st_stateowner;
2999
4c4cd222 3000 if (lock) {
4c4cd222 3001 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 3002 struct nfs4_client *clp = sop->so_client;
4c4cd222 3003 int lkflg = 0;
b37ad28b 3004 __be32 status;
4c4cd222
N		 3005
3006 lkflg = setlkflg(lock->lk_type);
3007
3008 if (lock->lk_is_new) {
599e0a22
BF		 3009 if (!sop->so_is_open_owner)
3010 return nfserr_bad_stateid;
60adfc50
AA		 3011 if (!(flags & HAS_SESSION) &&
3012 !same_clid(&clp->cl_clientid, lockclid))
3013 return nfserr_bad_stateid;
599e0a22
BF		 3014 /* stp is the open stateid */
3015 status = nfs4_check_openmode(stp, lkflg);
3016 if (status)
3017 return status;
3018 } else {
3019 /* stp is the lock stateid */
3020 status = nfs4_check_openmode(stp->st_openstp, lkflg);
3021 if (status)
3022 return status;
4c4cd222 3023 }
1da177e4
LT		 3024 }
3025
f3362737 3026 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3027 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3028 return nfserr_bad_stateid;
1da177e4
LT		 3029 }
3030
1da177e4
LT		 3031 /*
3032 * We now validate the seqid and stateid generation numbers.
3033 * For the moment, we ignore the possibility of
3034 * generation number wraparound.
3035 */
6668958f 3036 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3037 goto check_replay;
3038
3a4f98bb 3039 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3040 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3041 " unconfirmed stateowner!\n");
3042 return nfserr_bad_stateid;
1da177e4 3043 }
3a4f98bb 3044 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3045 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3046 " confirmed yet!\n");
3047 return nfserr_bad_stateid;
1da177e4 3048 }
6668958f 3049 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3050 if (status)
3051 return status;
52fd004e 3052 renew_client(sop->so_client);
3a4f98bb 3053 return nfs_ok;
1da177e4 3054
1da177e4 3055check_replay:
bd9aac52 3056 if (seqid == sop->so_seqid - 1) {
849823c5 3057 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3058 /* indicate replay to calling function */
a90b061c 3059 return nfserr_replay_me;
1da177e4 3060 }
2fdada03 3061 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3062 sop->so_seqid, seqid);
3063 *sopp = NULL;
3064 return nfserr_bad_seqid;
1da177e4
LT		 3065}
3066
b37ad28b 3067__be32
ca364317 3068nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3069 struct nfsd4_open_confirm *oc)
1da177e4 3070{
b37ad28b 3071 __be32 status;
1da177e4
LT		 3072 struct nfs4_stateowner *sop;
3073 struct nfs4_stateid *stp;
3074
3075 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3076 (int)cstate->current_fh.fh_dentry->d_name.len,
3077 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3078
ca364317 3079 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3080 if (status)
3081 return status;
1da177e4
LT		 3082
3083 nfs4_lock_state();
3084
dd453dfd 3085 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3086 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3087 CONFIRM | OPEN_STATE,
1da177e4
LT		 3088 &oc->oc_stateowner, &stp, NULL)))
3089 goto out;
3090
3091 sop = oc->oc_stateowner;
3092 sop->so_confirmed = 1;
3093 update_stateid(&stp->st_stateid);
3094 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3095 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3096 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3097 stp->st_stateid.si_boot,
3098 stp->st_stateid.si_stateownerid,
3099 stp->st_stateid.si_fileid,
3100 stp->st_stateid.si_generation);
c7b9a459
N		 3101
3102 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3103out:
f2327d9a 3104 if (oc->oc_stateowner) {
1da177e4 3105 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3106 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3107 }
1da177e4
LT		 3108 nfs4_unlock_state();
3109 return status;
3110}
3111
3112
3113/*
3114 * unset all bits in union bitmap (bmap) that
3115 * do not exist in share (from successful OPEN_DOWNGRADE)
3116 */
3117static void
3118reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3119{
3120 int i;
3121 for (i = 1; i < 4; i++) {
3122 if ((i & access) != i)
3123 __clear_bit(i, bmap);
3124 }
3125}
3126
3127static void
3128reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3129{
3130 int i;
3131 for (i = 0; i < 4; i++) {
3132 if ((i & deny) != i)
3133 __clear_bit(i, bmap);
3134 }
3135}
3136
b37ad28b 3137__be32
ca364317
BF		 3138nfsd4_open_downgrade(struct svc_rqst *rqstp,
3139 struct nfsd4_compound_state *cstate,
a4f1706a 3140 struct nfsd4_open_downgrade *od)
1da177e4 3141{
b37ad28b 3142 __be32 status;
1da177e4
LT		 3143 struct nfs4_stateid *stp;
3144 unsigned int share_access;
3145
3146 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3147 (int)cstate->current_fh.fh_dentry->d_name.len,
3148 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3149
d87a8ade 3150 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3151 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3152 return nfserr_inval;
3153
3154 nfs4_lock_state();
dd453dfd 3155 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3156 od->od_seqid,
1da177e4 3157 &od->od_stateid,
f3362737 3158 OPEN_STATE,
1da177e4
LT		 3159 &od->od_stateowner, &stp, NULL)))
3160 goto out;
3161
3162 status = nfserr_inval;
3163 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3164 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3165 stp->st_access_bmap, od->od_share_access);
3166 goto out;
3167 }
3168 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3169 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3170 stp->st_deny_bmap, od->od_share_deny);
3171 goto out;
3172 }
3173 set_access(&share_access, stp->st_access_bmap);
3174 nfs4_file_downgrade(stp->st_vfs_file,
3175 share_access & ~od->od_share_access);
3176
3177 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3178 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3179
3180 update_stateid(&stp->st_stateid);
3181 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3182 status = nfs_ok;
3183out:
f2327d9a 3184 if (od->od_stateowner) {
1da177e4 3185 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3186 cstate->replay_owner = od->od_stateowner;
f2327d9a 3187 }
1da177e4
LT		 3188 nfs4_unlock_state();
3189 return status;
3190}
3191
3192/*
3193 * nfs4_unlock_state() called after encode
3194 */
b37ad28b 3195__be32
ca364317 3196nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3197 struct nfsd4_close *close)
1da177e4 3198{
b37ad28b 3199 __be32 status;
1da177e4
LT		 3200 struct nfs4_stateid *stp;
3201
3202 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3203 (int)cstate->current_fh.fh_dentry->d_name.len,
3204 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3205
3206 nfs4_lock_state();
3207 /* check close_lru for replay */
dd453dfd 3208 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3209 close->cl_seqid,
1da177e4 3210 &close->cl_stateid,
f3362737 3211 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3212 &close->cl_stateowner, &stp, NULL)))
3213 goto out;
1da177e4
LT		 3214 status = nfs_ok;
3215 update_stateid(&stp->st_stateid);
3216 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3217
04ef5954 3218 /* release_stateid() calls nfsd_close() if needed */
2283963f 3219 release_open_stateid(stp);
04ef5954
BF		 3220
3221 /* place unused nfs4_stateowners on so_close_lru list to be
3222 * released by the laundromat service after the lease period
3223 * to enable us to handle CLOSE replay
3224 */
3225 if (list_empty(&close->cl_stateowner->so_stateids))
3226 move_to_close_lru(close->cl_stateowner);
1da177e4 3227out:
f2327d9a 3228 if (close->cl_stateowner) {
1da177e4 3229 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3230 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3231 }
1da177e4
LT		 3232 nfs4_unlock_state();
3233 return status;
3234}
3235
b37ad28b 3236__be32
ca364317
BF		 3237nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3238 struct nfsd4_delegreturn *dr)
1da177e4 3239{
203a8c8e
BF		 3240 struct nfs4_delegation *dp;
3241 stateid_t *stateid = &dr->dr_stateid;
3242 struct inode *inode;
b37ad28b 3243 __be32 status;
6668958f 3244 int flags = 0;
1da177e4 3245
ca364317 3246 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3247 return status;
3248 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3249
6668958f
AA		 3250 if (nfsd4_has_session(cstate))
3251 flags |= HAS_SESSION;
1da177e4 3252 nfs4_lock_state();
203a8c8e
BF		 3253 status = nfserr_bad_stateid;
3254 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3255 goto out;
3256 status = nfserr_stale_stateid;
3257 if (STALE_STATEID(stateid))
3258 goto out;
7e0f7cf5 3259 status = nfserr_bad_stateid;
203a8c8e
BF		 3260 if (!is_delegation_stateid(stateid))
3261 goto out;
203a8c8e 3262 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3263 if (!dp) {
3264 status = stateid_error_map(stateid);
203a8c8e 3265 goto out;
78155ed7 3266 }
6668958f 3267 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3268 if (status)
3269 goto out;
3270 renew_client(dp->dl_client);
3271
3272 unhash_delegation(dp);
1da177e4 3273out:
203a8c8e
BF		 3274 nfs4_unlock_state();
3275
1da177e4
LT		 3276 return status;
3277}
3278
3279
3280/*
3281 * Lock owner state (byte-range locks)
3282 */
3283#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3284#define LOCK_HASH_BITS 8
3285#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3286#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3287
87df4de8
BH		 3288static inline u64
3289end_offset(u64 start, u64 len)
3290{
3291 u64 end;
3292
3293 end = start + len;
3294 return end >= start ? end: NFS4_MAX_UINT64;
3295}
3296
3297/* last octet in a range */
3298static inline u64
3299last_byte_offset(u64 start, u64 len)
3300{
3301 u64 end;
3302
3303 BUG_ON(!len);
3304 end = start + len;
3305 return end > start ? end - 1: NFS4_MAX_UINT64;
3306}
3307
1da177e4
LT		 3308#define lockownerid_hashval(id) \
3309 ((id) & LOCK_HASH_MASK)
3310
3311static inline unsigned int
3312lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3313 struct xdr_netobj *ownername)
3314{
3315 return (file_hashval(inode) + cl_id
3316 + opaque_hashval(ownername->data, ownername->len))
3317 & LOCK_HASH_MASK;
3318}
3319
3320static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3321static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3322static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3323
fd39ca9a 3324static struct nfs4_stateid *
1da177e4
LT		 3325find_stateid(stateid_t *stid, int flags)
3326{
9346eff0 3327 struct nfs4_stateid *local;
1da177e4
LT		 3328 u32 st_id = stid->si_stateownerid;
3329 u32 f_id = stid->si_fileid;
3330 unsigned int hashval;
3331
3332 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3333 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3334 hashval = stateid_hashval(st_id, f_id);
3335 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3336 if ((local->st_stateid.si_stateownerid == st_id) &&
3337 (local->st_stateid.si_fileid == f_id))
3338 return local;
3339 }
3340 }
9346eff0
KK		 3341
3342 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3343 hashval = stateid_hashval(st_id, f_id);
3344 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3345 if ((local->st_stateid.si_stateownerid == st_id) &&
3346 (local->st_stateid.si_fileid == f_id))
3347 return local;
3348 }
849823c5 3349 }
1da177e4
LT		 3350 return NULL;
3351}
3352
3353static struct nfs4_delegation *
3354find_delegation_stateid(struct inode *ino, stateid_t *stid)
3355{
13cd2184
N		 3356 struct nfs4_file *fp;
3357 struct nfs4_delegation *dl;
1da177e4
LT		 3358
3359 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3360 stid->si_boot, stid->si_stateownerid,
3361 stid->si_fileid, stid->si_generation);
3362
1da177e4 3363 fp = find_file(ino);
13cd2184
N		 3364 if (!fp)
3365 return NULL;
3366 dl = find_delegation_file(fp, stid);
3367 put_nfs4_file(fp);
3368 return dl;
1da177e4
LT		 3369}
3370
3371/*
3372 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3373 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3374 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3375 * locking, this prevents us from being completely protocol-compliant. The
3376 * real solution to this problem is to start using unsigned file offsets in
3377 * the VFS, but this is a very deep change!
3378 */
3379static inline void
3380nfs4_transform_lock_offset(struct file_lock *lock)
3381{
3382 if (lock->fl_start < 0)
3383 lock->fl_start = OFFSET_MAX;
3384 if (lock->fl_end < 0)
3385 lock->fl_end = OFFSET_MAX;
3386}
3387
d5b9026a
N		 3388/* Hack!: For now, we're defining this just so we can use a pointer to it
3389 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3390static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3391};
1da177e4
LT		 3392
3393static inline void
3394nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3395{
d5b9026a
N		 3396 struct nfs4_stateowner *sop;
3397 unsigned int hval;
1da177e4 3398
d5b9026a
N		 3399 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3400 sop = (struct nfs4_stateowner *) fl->fl_owner;
3401 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3402 kref_get(&sop->so_ref);
3403 deny->ld_sop = sop;
3404 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3405 } else {
3406 deny->ld_sop = NULL;
3407 deny->ld_clientid.cl_boot = 0;
3408 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3409 }
3410 deny->ld_start = fl->fl_start;
87df4de8
BH		 3411 deny->ld_length = NFS4_MAX_UINT64;
3412 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3413 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3414 deny->ld_type = NFS4_READ_LT;
3415 if (fl->fl_type != F_RDLCK)
3416 deny->ld_type = NFS4_WRITE_LT;
3417}
3418
1da177e4
LT		 3419static struct nfs4_stateowner *
3420find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3421 struct xdr_netobj *owner)
3422{
3423 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3424 struct nfs4_stateowner *op;
3425
3426 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3427 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3428 return op;
3429 }
3430 return NULL;
3431}
3432
3433/*
3434 * Alloc a lock owner structure.
3435 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3436 * occured.
3437 *
3438 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3439 */
3440
3441static struct nfs4_stateowner *
3442alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3443 struct nfs4_stateowner *sop;
3444 struct nfs4_replay *rp;
3445 unsigned int idhashval;
3446
3447 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3448 return NULL;
3449 idhashval = lockownerid_hashval(current_ownerid);
3450 INIT_LIST_HEAD(&sop->so_idhash);
3451 INIT_LIST_HEAD(&sop->so_strhash);
3452 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3453 INIT_LIST_HEAD(&sop->so_stateids);
3454 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3455 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3456 sop->so_time = 0;
3457 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3458 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3459 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3460 sop->so_is_open_owner = 0;
3461 sop->so_id = current_ownerid++;
3462 sop->so_client = clp;
b59e3c0e
NB		 3463 /* It is the openowner seqid that will be incremented in encode in the
3464 * case of new lockowners; so increment the lock seqid manually: */
3465 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3466 sop->so_confirmed = 1;
3467 rp = &sop->so_replay;
de1ae286 3468 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3469 rp->rp_buflen = 0;
3470 rp->rp_buf = rp->rp_ibuf;
3471 return sop;
3472}
3473
fd39ca9a 3474static struct nfs4_stateid *
1da177e4
LT		 3475alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3476{
3477 struct nfs4_stateid *stp;
3478 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3479
5ac049ac
N		 3480 stp = nfs4_alloc_stateid();
3481 if (stp == NULL)
1da177e4
LT		 3482 goto out;
3483 INIT_LIST_HEAD(&stp->st_hash);
3484 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3485 INIT_LIST_HEAD(&stp->st_perstateowner);
3486 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3487 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3488 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3489 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3490 stp->st_stateowner = sop;
13cd2184 3491 get_nfs4_file(fp);
1da177e4 3492 stp->st_file = fp;
78155ed7 3493 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3494 stp->st_stateid.si_stateownerid = sop->so_id;
3495 stp->st_stateid.si_fileid = fp->fi_id;
3496 stp->st_stateid.si_generation = 0;
3497 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3498 stp->st_access_bmap = open_stp->st_access_bmap;
3499 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3500 stp->st_openstp = open_stp;
1da177e4
LT		 3501
3502out:
3503 return stp;
3504}
3505
fd39ca9a 3506static int
1da177e4
LT		 3507check_lock_length(u64 offset, u64 length)
3508{
87df4de8 3509 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3510 LOFF_OVERFLOW(offset, length)));
3511}
3512
3513/*
3514 * LOCK operation
3515 */
b37ad28b 3516__be32
ca364317 3517nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3518 struct nfsd4_lock *lock)
1da177e4 3519{
3e9e3dbe 3520 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3521 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3522 struct nfs4_stateid *lock_stp;
3523 struct file *filp;
3524 struct file_lock file_lock;
8dc7c311 3525 struct file_lock conflock;
b37ad28b 3526 __be32 status = 0;
1da177e4 3527 unsigned int strhashval;
150b3934 3528 unsigned int cmd;
b8dd7b9a 3529 int err;
1da177e4
LT		 3530
3531 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3532 (long long) lock->lk_offset,
3533 (long long) lock->lk_length);
3534
1da177e4
LT		 3535 if (check_lock_length(lock->lk_offset, lock->lk_length))
3536 return nfserr_inval;
3537
ca364317 3538 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3539 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3540 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3541 return status;
3542 }
3543
1da177e4
LT		 3544 nfs4_lock_state();
3545
3546 if (lock->lk_is_new) {
893f8770
N		 3547 /*
3548 * Client indicates that this is a new lockowner.
3549 * Use open owner and open stateid to create lock owner and
3550 * lock stateid.
3551 */
1da177e4
LT		 3552 struct nfs4_stateid *open_stp = NULL;
3553 struct nfs4_file *fp;
3554
3555 status = nfserr_stale_clientid;
60adfc50
AA		 3556 if (!nfsd4_has_session(cstate) &&
3557 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3558 goto out;
1da177e4 3559
1da177e4 3560 /* validate and update open stateid and open seqid */
dd453dfd 3561 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3562 lock->lk_new_open_seqid,
3563 &lock->lk_new_open_stateid,
f3362737 3564 OPEN_STATE,
3a65588a 3565 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3566 lock);
37515177 3567 if (status)
1da177e4 3568 goto out;
3a65588a 3569 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3570 /* create lockowner and lock stateid */
3571 fp = open_stp->st_file;
3572 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3573 open_sop->so_client->cl_clientid.cl_id,
3574 &lock->v.new.owner);
3e9e3dbe
N		 3575 /* XXX: Do we need to check for duplicate stateowners on
3576 * the same file, or should they just be allowed (and
3577 * create new stateids)? */
1da177e4 3578 status = nfserr_resource;
b59e3c0e
NB		 3579 lock_sop = alloc_init_lock_stateowner(strhashval,
3580 open_sop->so_client, open_stp, lock);
3581 if (lock_sop == NULL)
1da177e4 3582 goto out;
b59e3c0e 3583 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3584 if (lock_stp == NULL)
1da177e4 3585 goto out;
1da177e4
LT		 3586 } else {
3587 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3588 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3589 lock->lk_old_lock_seqid,
3590 &lock->lk_old_lock_stateid,
f3362737 3591 LOCK_STATE,
3a65588a 3592 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3593 if (status)
3594 goto out;
3a65588a 3595 lock_sop = lock->lk_replay_owner;
1da177e4 3596 }
3a65588a 3597 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3598 filp = lock_stp->st_vfs_file;
3599
0dd395dc 3600 status = nfserr_grace;
af558e33 3601 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3602 goto out;
3603 status = nfserr_no_grace;
af558e33 3604 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3605 goto out;
3606
1da177e4
LT		 3607 locks_init_lock(&file_lock);
3608 switch (lock->lk_type) {
3609 case NFS4_READ_LT:
3610 case NFS4_READW_LT:
3611 file_lock.fl_type = F_RDLCK;
150b3934 3612 cmd = F_SETLK;
1da177e4
LT		 3613 break;
3614 case NFS4_WRITE_LT:
3615 case NFS4_WRITEW_LT:
3616 file_lock.fl_type = F_WRLCK;
150b3934 3617 cmd = F_SETLK;
1da177e4
LT		 3618 break;
3619 default:
3620 status = nfserr_inval;
3621 goto out;
3622 }
b59e3c0e 3623 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3624 file_lock.fl_pid = current->tgid;
3625 file_lock.fl_file = filp;
3626 file_lock.fl_flags = FL_POSIX;
d5b9026a 3627 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3628
3629 file_lock.fl_start = lock->lk_offset;
87df4de8 3630 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3631 nfs4_transform_lock_offset(&file_lock);
3632
3633 /*
3634 * Try to lock the file in the VFS.
3635 * Note: locks.c uses the BKL to protect the inode's lock list.
3636 */
3637
fd85b817 3638 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3639 switch (-err) {
1da177e4
LT		 3640 case 0: /* success! */
3641 update_stateid(&lock_stp->st_stateid);
3642 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3643 sizeof(stateid_t));
b8dd7b9a 3644 status = 0;
eb76b3fd
AA		 3645 break;
3646 case (EAGAIN): /* conflock holds conflicting lock */
3647 status = nfserr_denied;
3648 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3649 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3650 break;
1da177e4
LT		 3651 case (EDEADLK):
3652 status = nfserr_deadlock;
eb76b3fd 3653 break;
1da177e4 3654 default:
fd85b817 3655 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3656 status = nfserr_resource;
3657 break;
1da177e4 3658 }
1da177e4 3659out:
8a280510 3660 if (status && lock->lk_is_new && lock_sop)
f044ff83 3661 release_lockowner(lock_sop);
3a65588a
BF		 3662 if (lock->lk_replay_owner) {
3663 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3664 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3665 }
1da177e4
LT		 3666 nfs4_unlock_state();
3667 return status;
3668}
3669
55ef1274
BF		 3670/*
3671 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3672 * so we do a temporary open here just to get an open file to pass to
3673 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3674 * inode operation.)
3675 */
3676static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3677{
3678 struct file *file;
3679 int err;
3680
3681 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3682 if (err)
3683 return err;
3684 err = vfs_test_lock(file, lock);
3685 nfsd_close(file);
3686 return err;
3687}
3688
1da177e4
LT		 3689/*
3690 * LOCKT operation
3691 */
b37ad28b 3692__be32
ca364317
BF		 3693nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3694 struct nfsd4_lockt *lockt)
1da177e4
LT		 3695{
3696 struct inode *inode;
1da177e4 3697 struct file_lock file_lock;
fd85b817 3698 int error;
b37ad28b 3699 __be32 status;
1da177e4 3700
af558e33 3701 if (locks_in_grace())
1da177e4
LT		 3702 return nfserr_grace;
3703
3704 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3705 return nfserr_inval;
3706
3707 lockt->lt_stateowner = NULL;
3708 nfs4_lock_state();
3709
3710 status = nfserr_stale_clientid;
60adfc50 3711 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3712 goto out;
1da177e4 3713
ca364317 3714 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3715 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3716 if (status == nfserr_symlink)
3717 status = nfserr_inval;
3718 goto out;
3719 }
3720
ca364317 3721 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3722 locks_init_lock(&file_lock);
3723 switch (lockt->lt_type) {
3724 case NFS4_READ_LT:
3725 case NFS4_READW_LT:
3726 file_lock.fl_type = F_RDLCK;
3727 break;
3728 case NFS4_WRITE_LT:
3729 case NFS4_WRITEW_LT:
3730 file_lock.fl_type = F_WRLCK;
3731 break;
3732 default:
2fdada03 3733 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3734 status = nfserr_inval;
3735 goto out;
3736 }
3737
3738 lockt->lt_stateowner = find_lockstateowner_str(inode,
3739 &lockt->lt_clientid, &lockt->lt_owner);
3740 if (lockt->lt_stateowner)
3741 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3742 file_lock.fl_pid = current->tgid;
3743 file_lock.fl_flags = FL_POSIX;
3744
3745 file_lock.fl_start = lockt->lt_offset;
87df4de8 3746 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3747
3748 nfs4_transform_lock_offset(&file_lock);
3749
1da177e4 3750 status = nfs_ok;
55ef1274 3751 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3752 if (error) {
3753 status = nfserrno(error);
3754 goto out;
3755 }
9d6a8c5c 3756 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3757 status = nfserr_denied;
9d6a8c5c 3758 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3759 }
3760out:
3761 nfs4_unlock_state();
3762 return status;
3763}
3764
b37ad28b 3765__be32
ca364317 3766nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3767 struct nfsd4_locku *locku)
1da177e4
LT		 3768{
3769 struct nfs4_stateid *stp;
3770 struct file *filp = NULL;
3771 struct file_lock file_lock;
b37ad28b 3772 __be32 status;
b8dd7b9a 3773 int err;
1da177e4
LT		 3774
3775 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3776 (long long) locku->lu_offset,
3777 (long long) locku->lu_length);
3778
3779 if (check_lock_length(locku->lu_offset, locku->lu_length))
3780 return nfserr_inval;
3781
3782 nfs4_lock_state();
3783
dd453dfd 3784 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3785 locku->lu_seqid,
3786 &locku->lu_stateid,
f3362737 3787 LOCK_STATE,
1da177e4
LT		 3788 &locku->lu_stateowner, &stp, NULL)))
3789 goto out;
3790
3791 filp = stp->st_vfs_file;
3792 BUG_ON(!filp);
3793 locks_init_lock(&file_lock);
3794 file_lock.fl_type = F_UNLCK;
3795 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3796 file_lock.fl_pid = current->tgid;
3797 file_lock.fl_file = filp;
3798 file_lock.fl_flags = FL_POSIX;
d5b9026a 3799 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3800 file_lock.fl_start = locku->lu_offset;
3801
87df4de8 3802 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3803 nfs4_transform_lock_offset(&file_lock);
3804
3805 /*
3806 * Try to unlock the file in the VFS.
3807 */
fd85b817 3808 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3809 if (err) {
fd85b817 3810 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3811 goto out_nfserr;
3812 }
3813 /*
3814 * OK, unlock succeeded; the only thing left to do is update the stateid.
3815 */
3816 update_stateid(&stp->st_stateid);
3817 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3818
3819out:
f2327d9a 3820 if (locku->lu_stateowner) {
1da177e4 3821 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3822 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3823 }
1da177e4
LT		 3824 nfs4_unlock_state();
3825 return status;
3826
3827out_nfserr:
b8dd7b9a 3828 status = nfserrno(err);
1da177e4
LT		 3829 goto out;
3830}
3831
3832/*
3833 * returns
3834 * 1: locks held by lockowner
3835 * 0: no locks held by lockowner
3836 */
3837static int
3838check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3839{
3840 struct file_lock **flpp;
7eaa36e2 3841 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3842 int status = 0;
3843
3844 lock_kernel();
3845 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3846 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3847 status = 1;
3848 goto out;
796dadfd 3849 }
1da177e4
LT		 3850 }
3851out:
3852 unlock_kernel();
3853 return status;
3854}
3855
b37ad28b 3856__be32
b591480b
BF		 3857nfsd4_release_lockowner(struct svc_rqst *rqstp,
3858 struct nfsd4_compound_state *cstate,
3859 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3860{
3861 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3862 struct nfs4_stateowner *sop;
3863 struct nfs4_stateid *stp;
1da177e4 3864 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3865 struct list_head matches;
3866 int i;
b37ad28b 3867 __be32 status;
1da177e4
LT		 3868
3869 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3870 clid->cl_boot, clid->cl_id);
3871
3872 /* XXX check for lease expiration */
3873
3874 status = nfserr_stale_clientid;
849823c5 3875 if (STALE_CLIENTID(clid))
1da177e4 3876 return status;
1da177e4
LT		 3877
3878 nfs4_lock_state();
3879
3e9e3dbe
N		 3880 status = nfserr_locks_held;
3881 /* XXX: we're doing a linear search through all the lockowners.
3882 * Yipes! For now we'll just hope clients aren't really using
3883 * release_lockowner much, but eventually we have to fix these
3884 * data structures. */
3885 INIT_LIST_HEAD(&matches);
3886 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3887 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3888 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3889 continue;
3890 list_for_each_entry(stp, &sop->so_stateids,
3891 st_perstateowner) {
3892 if (check_for_locks(stp->st_vfs_file, sop))
3893 goto out;
3894 /* Note: so_perclient unused for lockowners,
3895 * so it's OK to fool with here. */
3896 list_add(&sop->so_perclient, &matches);
3897 }
1da177e4 3898 }
3e9e3dbe
N		 3899 }
3900 /* Clients probably won't expect us to return with some (but not all)
3901 * of the lockowner state released; so don't release any until all
3902 * have been checked. */
3903 status = nfs_ok;
0fa822e4
N		 3904 while (!list_empty(&matches)) {
3905 sop = list_entry(matches.next, struct nfs4_stateowner,
3906 so_perclient);
3907 /* unhash_stateowner deletes so_perclient only
3908 * for openowners. */
3909 list_del(&sop->so_perclient);
f044ff83 3910 release_lockowner(sop);
1da177e4
LT		 3911 }
3912out:
3913 nfs4_unlock_state();
3914 return status;
3915}
3916
3917static inline struct nfs4_client_reclaim *
a55370a3 3918alloc_reclaim(void)
1da177e4 3919{
a55370a3 3920 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3921}
3922
c7b9a459 3923int
a1bcecd2 3924nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3925{
3926 unsigned int strhashval = clientstr_hashval(name);
3927 struct nfs4_client *clp;
3928
a1bcecd2 3929 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3930 return clp ? 1 : 0;
3931}
3932
1da177e4
LT		 3933/*
3934 * failure => all reset bets are off, nfserr_no_grace...
3935 */
190e4fbf
N		 3936int
3937nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3938{
3939 unsigned int strhashval;
3940 struct nfs4_client_reclaim *crp = NULL;
3941
a55370a3
N		 3942 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3943 crp = alloc_reclaim();
1da177e4
LT		 3944 if (!crp)
3945 return 0;
a55370a3 3946 strhashval = clientstr_hashval(name);
1da177e4
LT		 3947 INIT_LIST_HEAD(&crp->cr_strhash);
3948 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3949 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3950 reclaim_str_hashtbl_size++;
3951 return 1;
3952}
3953
3954static void
3955nfs4_release_reclaim(void)
3956{
3957 struct nfs4_client_reclaim *crp = NULL;
3958 int i;
3959
1da177e4
LT		 3960 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3961 while (!list_empty(&reclaim_str_hashtbl[i])) {
3962 crp = list_entry(reclaim_str_hashtbl[i].next,
3963 struct nfs4_client_reclaim, cr_strhash);
3964 list_del(&crp->cr_strhash);
1da177e4
LT		 3965 kfree(crp);
3966 reclaim_str_hashtbl_size--;
3967 }
3968 }
3969 BUG_ON(reclaim_str_hashtbl_size);
3970}
3971
3972/*
3973 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3974static struct nfs4_client_reclaim *
1da177e4
LT		 3975nfs4_find_reclaim_client(clientid_t *clid)
3976{
3977 unsigned int strhashval;
3978 struct nfs4_client *clp;
3979 struct nfs4_client_reclaim *crp = NULL;
3980
3981
3982 /* find clientid in conf_id_hashtbl */
3983 clp = find_confirmed_client(clid);
3984 if (clp == NULL)
3985 return NULL;
3986
a55370a3
N		 3987 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3988 clp->cl_name.len, clp->cl_name.data,
3989 clp->cl_recdir);
1da177e4
LT		 3990
3991 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3992 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3993 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3994 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3995 return crp;
3996 }
3997 }
3998 return NULL;
3999}
4000
4001/*
4002* Called from OPEN. Look for clientid in reclaim list.
4003*/
b37ad28b 4004__be32
1da177e4
LT		 4005nfs4_check_open_reclaim(clientid_t *clid)
4006{
dfc83565 4007 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 4008}
4009
ac4d8ff2 4010/* initialization to perform at module load time: */
1da177e4 4011
e8ff2a84 4012int
ac4d8ff2 4013nfs4_state_init(void)
1da177e4 4014{
e8ff2a84 4015 int i, status;
1da177e4 4016
e8ff2a84
BF		 4017 status = nfsd4_init_slabs();
4018 if (status)
4019 return status;
1da177e4
LT		 4020 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4021 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
4022 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
4023 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4024 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
4025 }
5282fd72
ME		 4026 for (i = 0; i < SESSION_HASH_SIZE; i++)
4027 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4028 for (i = 0; i < FILE_HASH_SIZE; i++) {
4029 INIT_LIST_HEAD(&file_hashtbl[i]);
4030 }
4031 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4032 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4033 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4034 }
4035 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4036 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4037 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4038 }
4039 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4040 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4041 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4042 }
1da177e4 4043 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4044 INIT_LIST_HEAD(&close_lru);
4045 INIT_LIST_HEAD(&client_lru);
4046 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N		 4047 for (i = 0; i < CLIENT_HASH_SIZE; i++)
4048 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
4049 reclaim_str_hashtbl_size = 0;
e8ff2a84 4050 return 0;
ac4d8ff2
N		 4051}
4052
190e4fbf
N		 4053static void
4054nfsd4_load_reboot_recovery_data(void)
4055{
4056 int status;
4057
0964a3d3
N		 4058 nfs4_lock_state();
4059 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4060 status = nfsd4_recdir_load();
0964a3d3 4061 nfs4_unlock_state();
190e4fbf
N		 4062 if (status)
4063 printk("NFSD: Failure reading reboot recovery data\n");
4064}
4065
9a8db97e
ME		 4066unsigned long
4067get_nfs4_grace_period(void)
4068{
4069 return max(user_lease_time, lease_time) * HZ;
4070}
4071
c2f1a551
MS		 4072/*
4073 * Since the lifetime of a delegation isn't limited to that of an open, a
4074 * client may quite reasonably hang on to a delegation as long as it has
4075 * the inode cached. This becomes an obvious problem the first time a
4076 * client's inode cache approaches the size of the server's total memory.
4077 *
4078 * For now we avoid this problem by imposing a hard limit on the number
4079 * of delegations, which varies according to the server's memory size.
4080 */
4081static void
4082set_max_delegations(void)
4083{
4084 /*
4085 * Allow at most 4 delegations per megabyte of RAM. Quick
4086 * estimates suggest that in the worst case (where every delegation
4087 * is for a different inode), a delegation could take about 1.5K,
4088 * giving a worst case usage of about 6% of memory.
4089 */
4090 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4091}
4092
ac4d8ff2
N		 4093/* initialization to perform when the nfsd service is started: */
4094
4095static void
4096__nfs4_state_start(void)
4097{
9a8db97e 4098 unsigned long grace_time;
ac4d8ff2 4099
1da177e4 4100 boot_time = get_seconds();
af558e33 4101 grace_time = get_nfs4_grace_period();
d99a05ad 4102 lease_time = user_lease_time;
af558e33 4103 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4104 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4105 grace_time/HZ);
58da282b 4106 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4107 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4108 set_max_delegations();
1da177e4
LT		 4109}
4110
e8ff2a84 4111void
76a3550e 4112nfs4_state_start(void)
1da177e4 4113{
1da177e4 4114 if (nfs4_init)
e8ff2a84 4115 return;
190e4fbf 4116 nfsd4_load_reboot_recovery_data();
76a3550e 4117 __nfs4_state_start();
1da177e4 4118 nfs4_init = 1;
e8ff2a84 4119 return;
1da177e4
LT		 4120}
4121
1da177e4
LT		 4122time_t
4123nfs4_lease_time(void)
4124{
4125 return lease_time;
4126}
4127
4128static void
4129__nfs4_state_shutdown(void)
4130{
4131 int i;
4132 struct nfs4_client *clp = NULL;
4133 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4134 struct list_head *pos, *next, reaplist;
4135
1da177e4
LT		 4136 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4137 while (!list_empty(&conf_id_hashtbl[i])) {
4138 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4139 expire_client(clp);
4140 }
4141 while (!list_empty(&unconf_str_hashtbl[i])) {
4142 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4143 expire_client(clp);
4144 }
4145 }
4146 INIT_LIST_HEAD(&reaplist);
4147 spin_lock(&recall_lock);
4148 list_for_each_safe(pos, next, &del_recall_lru) {
4149 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4150 list_move(&dp->dl_recall_lru, &reaplist);
4151 }
4152 spin_unlock(&recall_lock);
4153 list_for_each_safe(pos, next, &reaplist) {
4154 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4155 list_del_init(&dp->dl_recall_lru);
4156 unhash_delegation(dp);
4157 }
4158
190e4fbf 4159 nfsd4_shutdown_recdir();
1da177e4 4160 nfs4_init = 0;
1da177e4
LT		 4161}
4162
4163void
4164nfs4_state_shutdown(void)
4165{
5e8d5c29
N		 4166 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4167 destroy_workqueue(laundry_wq);
2c5e7615 4168 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4169 nfs4_lock_state();
4170 nfs4_release_reclaim();
4171 __nfs4_state_shutdown();
1da177e4
LT		 4172 nfs4_unlock_state();
4173}
4174
3dd98a3b
JL		 4175/*
4176 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4177 * accessed when nfsd is starting.
4178 */
0964a3d3
N		 4179static void
4180nfs4_set_recdir(char *recdir)
4181{
0964a3d3 4182 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4183}
4184
4185/*
4186 * Change the NFSv4 recovery directory to recdir.
4187 */
4188int
4189nfs4_reset_recoverydir(char *recdir)
4190{
4191 int status;
a63bb996 4192 struct path path;
0964a3d3 4193
a63bb996 4194 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4195 if (status)
4196 return status;
4197 status = -ENOTDIR;
a63bb996 4198 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4199 nfs4_set_recdir(recdir);
4200 status = 0;
4201 }
a63bb996 4202 path_put(&path);
0964a3d3
N		 4203 return status;
4204}
4205
3dd98a3b
JL		 4206char *
4207nfs4_recoverydir(void)
4208{
4209 return user_recovery_dirname;
4210}
4211
1da177e4
LT		 4212/*
4213 * Called when leasetime is changed.
4214 *
d99a05ad
N		 4215 * The only way the protocol gives us to handle on-the-fly lease changes is to
4216 * simulate a reboot. Instead of doing that, we just wait till the next time
4217 * we start to register any changes in lease time. If the administrator
4218 * really wants to change the lease time *now*, they can go ahead and bring
4219 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4220 *
4221 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4222 * when nfsd is starting
1da177e4
LT		 4223 */
4224void
4225nfs4_reset_lease(time_t leasetime)
4226{
d99a05ad 4227 user_lease_time = leasetime;
1da177e4 4228}
Linux 6.x block layer and io_uring tree(s)