projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd: convert nfs4_cb_conn struct to hold address in sockaddr_storage
[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>
363168b4 58#include <linux/sunrpc/clnt.h>
1da177e4
LT		 59
60#define NFSDDBG_FACILITY NFSDDBG_PROC
61
62/* Globals */
63static time_t lease_time = 90; /* default lease time */
d99a05ad 64static time_t user_lease_time = 90;
fd39ca9a 65static time_t boot_time;
1da177e4
LT		 66static u32 current_ownerid = 1;
67static u32 current_fileid = 1;
68static u32 current_delegid = 1;
69static u32 nfs4_init;
fd39ca9a
N		 70static stateid_t zerostateid; /* bits all 0 */
71static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 72static u64 current_sessionid = 1;
fd39ca9a
N		 73
74#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
75#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 76
1da177e4 77/* forward declarations */
fd39ca9a 78static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 79static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 80static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
81static void nfs4_set_recdir(char *recdir);
1da177e4 82
8b671b80
BF		 83/* Locking: */
84
85/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 86static DEFINE_MUTEX(client_mutex);
1da177e4 87
8b671b80
BF		 88/*
89 * Currently used for the del_recall_lru and file hash table. In an
90 * effort to decrease the scope of the client_mutex, this spinlock may
91 * eventually cover more:
92 */
93static DEFINE_SPINLOCK(recall_lock);
94
e18b890b
CL		 95static struct kmem_cache *stateowner_slab = NULL;
96static struct kmem_cache *file_slab = NULL;
97static struct kmem_cache *stateid_slab = NULL;
98static struct kmem_cache *deleg_slab = NULL;
e60d4398 99
1da177e4
LT		 100void
101nfs4_lock_state(void)
102{
353ab6e9 103 mutex_lock(&client_mutex);
1da177e4
LT		 104}
105
106void
107nfs4_unlock_state(void)
108{
353ab6e9 109 mutex_unlock(&client_mutex);
1da177e4
LT		 110}
111
112static inline u32
113opaque_hashval(const void *ptr, int nbytes)
114{
115 unsigned char *cptr = (unsigned char *) ptr;
116
117 u32 x = 0;
118 while (nbytes--) {
119 x *= 37;
120 x += *cptr++;
121 }
122 return x;
123}
124
1da177e4
LT		 125static struct list_head del_recall_lru;
126
13cd2184
N		 127static inline void
128put_nfs4_file(struct nfs4_file *fi)
129{
8b671b80
BF		 130 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
131 list_del(&fi->fi_hash);
132 spin_unlock(&recall_lock);
133 iput(fi->fi_inode);
134 kmem_cache_free(file_slab, fi);
135 }
13cd2184
N		 136}
137
138static inline void
139get_nfs4_file(struct nfs4_file *fi)
140{
8b671b80 141 atomic_inc(&fi->fi_ref);
13cd2184
N		 142}
143
ef0f3390 144static int num_delegations;
c2f1a551 145unsigned int max_delegations;
ef0f3390
N		 146
147/*
148 * Open owner state (share locks)
149 */
150
151/* hash tables for nfs4_stateowner */
152#define OWNER_HASH_BITS 8
153#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
154#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
155
156#define ownerid_hashval(id) \
157 ((id) & OWNER_HASH_MASK)
158#define ownerstr_hashval(clientid, ownername) \
159 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
160
161static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
162static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
163
164/* hash table for nfs4_file */
165#define FILE_HASH_BITS 8
166#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
167#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
168/* hash table for (open)nfs4_stateid */
169#define STATEID_HASH_BITS 10
170#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
171#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
172
173#define file_hashval(x) \
174 hash_ptr(x, FILE_HASH_BITS)
175#define stateid_hashval(owner_id, file_id) \
176 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
177
178static struct list_head file_hashtbl[FILE_HASH_SIZE];
179static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
180
1da177e4
LT		 181static struct nfs4_delegation *
182alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
183{
184 struct nfs4_delegation *dp;
185 struct nfs4_file *fp = stp->st_file;
c237dc03 186 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 187
188 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 189 if (fp->fi_had_conflict)
190 return NULL;
c2f1a551 191 if (num_delegations > max_delegations)
ef0f3390 192 return NULL;
5b2d21c1
N		 193 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
194 if (dp == NULL)
1da177e4 195 return dp;
ef0f3390 196 num_delegations++;
ea1da636
N		 197 INIT_LIST_HEAD(&dp->dl_perfile);
198 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 199 INIT_LIST_HEAD(&dp->dl_recall_lru);
200 dp->dl_client = clp;
13cd2184 201 get_nfs4_file(fp);
1da177e4
LT		 202 dp->dl_file = fp;
203 dp->dl_flock = NULL;
204 get_file(stp->st_vfs_file);
205 dp->dl_vfs_file = stp->st_vfs_file;
206 dp->dl_type = type;
b53d40c5 207 dp->dl_ident = cb->cb_ident;
78155ed7 208 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 209 dp->dl_stateid.si_stateownerid = current_delegid++;
210 dp->dl_stateid.si_fileid = 0;
211 dp->dl_stateid.si_generation = 0;
6c02eaa1 212 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 213 dp->dl_time = 0;
214 atomic_set(&dp->dl_count, 1);
ea1da636
N		 215 list_add(&dp->dl_perfile, &fp->fi_delegations);
216 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 217 return dp;
218}
219
220void
221nfs4_put_delegation(struct nfs4_delegation *dp)
222{
223 if (atomic_dec_and_test(&dp->dl_count)) {
224 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 225 put_nfs4_file(dp->dl_file);
5b2d21c1 226 kmem_cache_free(deleg_slab, dp);
ef0f3390 227 num_delegations--;
1da177e4
LT		 228 }
229}
230
231/* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
235 */
236static void
237nfs4_close_delegation(struct nfs4_delegation *dp)
238{
239 struct file *filp = dp->dl_vfs_file;
240
241 dprintk("NFSD: close_delegation dp %p\n",dp);
242 dp->dl_vfs_file = NULL;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
c907132d 245 if (dp->dl_flock)
a9933cea 246 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 247 nfsd_close(filp);
1da177e4
LT		 248}
249
250/* Called under the state lock. */
251static void
252unhash_delegation(struct nfs4_delegation *dp)
253{
ea1da636
N		 254 list_del_init(&dp->dl_perfile);
255 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 256 spin_lock(&recall_lock);
257 list_del_init(&dp->dl_recall_lru);
258 spin_unlock(&recall_lock);
259 nfs4_close_delegation(dp);
260 nfs4_put_delegation(dp);
261}
262
263/*
264 * SETCLIENTID state
265 */
266
267/* Hash tables for nfs4_clientid state */
268#define CLIENT_HASH_BITS 4
269#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
271
272#define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 274#define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 276/*
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
279 *
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
282 *
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
284 * setclientid info.
285 *
286 * client_lru holds client queue ordered by nfs4_client.cl_time
287 * for lease renewal.
288 *
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
291 */
292static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
293static int reclaim_str_hashtbl_size = 0;
294static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
295static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
298static struct list_head client_lru;
299static struct list_head close_lru;
300
2283963f
BF		 301static void unhash_generic_stateid(struct nfs4_stateid *stp)
302{
303 list_del(&stp->st_hash);
304 list_del(&stp->st_perfile);
305 list_del(&stp->st_perstateowner);
306}
307
308static void free_generic_stateid(struct nfs4_stateid *stp)
309{
310 put_nfs4_file(stp->st_file);
311 kmem_cache_free(stateid_slab, stp);
312}
313
314static void release_lock_stateid(struct nfs4_stateid *stp)
315{
316 unhash_generic_stateid(stp);
317 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
318 free_generic_stateid(stp);
319}
320
f044ff83
BF		 321static void unhash_lockowner(struct nfs4_stateowner *sop)
322{
323 struct nfs4_stateid *stp;
324
325 list_del(&sop->so_idhash);
326 list_del(&sop->so_strhash);
327 list_del(&sop->so_perstateid);
328 while (!list_empty(&sop->so_stateids)) {
329 stp = list_first_entry(&sop->so_stateids,
330 struct nfs4_stateid, st_perstateowner);
331 release_lock_stateid(stp);
332 }
333}
334
335static void release_lockowner(struct nfs4_stateowner *sop)
336{
337 unhash_lockowner(sop);
338 nfs4_put_stateowner(sop);
339}
340
f1d110ca
BF		 341static void
342release_stateid_lockowners(struct nfs4_stateid *open_stp)
343{
344 struct nfs4_stateowner *lock_sop;
345
346 while (!list_empty(&open_stp->st_lockowners)) {
347 lock_sop = list_entry(open_stp->st_lockowners.next,
348 struct nfs4_stateowner, so_perstateid);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 351 release_lockowner(lock_sop);
f1d110ca
BF		 352 }
353}
354
2283963f
BF		 355static void release_open_stateid(struct nfs4_stateid *stp)
356{
357 unhash_generic_stateid(stp);
358 release_stateid_lockowners(stp);
359 nfsd_close(stp->st_vfs_file);
360 free_generic_stateid(stp);
361}
362
f044ff83 363static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 364{
365 struct nfs4_stateid *stp;
366
367 list_del(&sop->so_idhash);
368 list_del(&sop->so_strhash);
f044ff83
BF		 369 list_del(&sop->so_perclient);
370 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 371 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 372 stp = list_first_entry(&sop->so_stateids,
373 struct nfs4_stateid, st_perstateowner);
374 release_open_stateid(stp);
f1d110ca
BF		 375 }
376}
377
f044ff83 378static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 379{
f044ff83 380 unhash_openowner(sop);
f1d110ca
BF		 381 list_del(&sop->so_close_lru);
382 nfs4_put_stateowner(sop);
383}
384
5282fd72
ME		 385static DEFINE_SPINLOCK(sessionid_lock);
386#define SESSION_HASH_SIZE 512
387static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
388
389static inline int
390hash_sessionid(struct nfs4_sessionid *sessionid)
391{
392 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
393
394 return sid->sequence % SESSION_HASH_SIZE;
395}
396
397static inline void
398dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
399{
400 u32 *ptr = (u32 *)(&sessionid->data[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
402}
403
ec6b5d7b
AA		 404static void
405gen_sessionid(struct nfsd4_session *ses)
406{
407 struct nfs4_client *clp = ses->se_client;
408 struct nfsd4_sessionid *sid;
409
410 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
411 sid->clientid = clp->cl_clientid;
412 sid->sequence = current_sessionid++;
413 sid->reserved = 0;
414}
415
416/*
417 * Give the client the number of slots it requests bound by
0c193054 418 * NFSD_MAX_SLOTS_PER_SESSION and by nfsd_drc_max_mem.
ec6b5d7b 419 *
0c193054
AA		 420 * If we run out of reserved DRC memory we should (up to a point) re-negotiate
421 * active sessions and reduce their slot usage to make rooom for new
422 * connections. For now we just fail the create session.
ec6b5d7b
AA		 423 */
424static int set_forechannel_maxreqs(struct nfsd4_channel_attrs *fchan)
425{
0c193054 426 int mem;
ec6b5d7b 427
5d77ddfb
AA		 428 if (fchan->maxreqs < 1)
429 return nfserr_inval;
430 else if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
431 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
432
0c193054 433 mem = fchan->maxreqs * NFSD_SLOT_CACHE_SIZE;
b101ebbc 434
4bd9b0f4 435 spin_lock(&nfsd_drc_lock);
0c193054
AA		 436 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem)
437 mem = ((nfsd_drc_max_mem - nfsd_drc_mem_used) /
438 NFSD_SLOT_CACHE_SIZE) * NFSD_SLOT_CACHE_SIZE;
439 nfsd_drc_mem_used += mem;
4bd9b0f4 440 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 441
0c193054 442 fchan->maxreqs = mem / NFSD_SLOT_CACHE_SIZE;
b101ebbc
AA		 443 if (fchan->maxreqs == 0)
444 return nfserr_resource;
445 return 0;
ec6b5d7b
AA		 446}
447
448/*
449 * fchan holds the client values on input, and the server values on output
450 */
451static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 452 struct nfsd4_channel_attrs *session_fchan,
453 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 454{
455 int status = 0;
456 __u32 maxcount = svc_max_payload(rqstp);
457
458 /* headerpadsz set to zero in encode routine */
459
460 /* Use the client's max request and max response size if possible */
461 if (fchan->maxreq_sz > maxcount)
462 fchan->maxreq_sz = maxcount;
6c18ba9f 463 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 464
465 if (fchan->maxresp_sz > maxcount)
466 fchan->maxresp_sz = maxcount;
6c18ba9f 467 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 468
0c193054 469 session_fchan->maxresp_cached = NFSD_SLOT_CACHE_SIZE;
6c18ba9f 470 fchan->maxresp_cached = session_fchan->maxresp_cached;
ec6b5d7b
AA		 471
472 /* Use the client's maxops if possible */
473 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
474 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 475 session_fchan->maxops = fchan->maxops;
ec6b5d7b 476
ec6b5d7b
AA		 477 /* FIXME: Error means no more DRC pages so the server should
478 * recover pages from existing sessions. For now fail session
479 * creation.
480 */
481 status = set_forechannel_maxreqs(fchan);
482
6c18ba9f 483 session_fchan->maxreqs = fchan->maxreqs;
ec6b5d7b
AA		 484 return status;
485}
486
487static int
488alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
489 struct nfsd4_create_session *cses)
490{
491 struct nfsd4_session *new, tmp;
492 int idx, status = nfserr_resource, slotsize;
493
494 memset(&tmp, 0, sizeof(tmp));
495
496 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 497 tmp.se_bchannel = cses->back_channel;
498 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
499 &cses->fore_channel);
ec6b5d7b
AA		 500 if (status)
501 goto out;
502
503 /* allocate struct nfsd4_session and slot table in one piece */
6c18ba9f 504 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot);
ec6b5d7b
AA		 505 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
506 if (!new)
507 goto out;
508
509 memcpy(new, &tmp, sizeof(*new));
510
511 new->se_client = clp;
512 gen_sessionid(new);
513 idx = hash_sessionid(&new->se_sessionid);
514 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
515 NFS4_MAX_SESSIONID_LEN);
516
517 new->se_flags = cses->flags;
518 kref_init(&new->se_ref);
519 spin_lock(&sessionid_lock);
520 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
521 list_add(&new->se_perclnt, &clp->cl_sessions);
522 spin_unlock(&sessionid_lock);
523
524 status = nfs_ok;
525out:
526 return status;
527}
528
5282fd72
ME		 529/* caller must hold sessionid_lock */
530static struct nfsd4_session *
531find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
532{
533 struct nfsd4_session *elem;
534 int idx;
535
536 dump_sessionid(__func__, sessionid);
537 idx = hash_sessionid(sessionid);
538 dprintk("%s: idx is %d\n", __func__, idx);
539 /* Search in the appropriate list */
540 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
541 dump_sessionid("list traversal", &elem->se_sessionid);
542 if (!memcmp(elem->se_sessionid.data, sessionid->data,
543 NFS4_MAX_SESSIONID_LEN)) {
544 return elem;
545 }
546 }
547
548 dprintk("%s: session not found\n", __func__);
549 return NULL;
550}
551
552/* caller must hold sessionid_lock */
7116ed6b 553static void
5282fd72 554unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 555{
556 list_del(&ses->se_hash);
557 list_del(&ses->se_perclnt);
5282fd72
ME		 558}
559
560static void
561release_session(struct nfsd4_session *ses)
562{
563 spin_lock(&sessionid_lock);
564 unhash_session(ses);
565 spin_unlock(&sessionid_lock);
7116ed6b
AA		 566 nfsd4_put_session(ses);
567}
568
14778a13
AA		 569static void nfsd4_release_respages(struct page **respages, short resused);
570
7116ed6b
AA		 571void
572free_session(struct kref *kref)
573{
574 struct nfsd4_session *ses;
14778a13 575 int i;
7116ed6b
AA		 576
577 ses = container_of(kref, struct nfsd4_session, se_ref);
6c18ba9f 578 for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
14778a13
AA		 579 struct nfsd4_cache_entry *e = &ses->se_slots[i].sl_cache_entry;
580 nfsd4_release_respages(e->ce_respages, e->ce_resused);
581 }
be98d1bb 582 spin_lock(&nfsd_drc_lock);
0c193054 583 nfsd_drc_mem_used -= ses->se_fchannel.maxreqs * NFSD_SLOT_CACHE_SIZE;
be98d1bb 584 spin_unlock(&nfsd_drc_lock);
7116ed6b
AA		 585 kfree(ses);
586}
587
1da177e4
LT		 588static inline void
589renew_client(struct nfs4_client *clp)
590{
591 /*
592 * Move client to the end to the LRU list.
593 */
594 dprintk("renewing client (clientid %08x/%08x)\n",
595 clp->cl_clientid.cl_boot,
596 clp->cl_clientid.cl_id);
597 list_move_tail(&clp->cl_lru, &client_lru);
598 clp->cl_time = get_seconds();
599}
600
601/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
602static int
603STALE_CLIENTID(clientid_t *clid)
604{
605 if (clid->cl_boot == boot_time)
606 return 0;
60adfc50
AA		 607 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
608 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 609 return 1;
610}
611
612/*
613 * XXX Should we use a slab cache ?
614 * This type of memory management is somewhat inefficient, but we use it
615 * anyway since SETCLIENTID is not a common operation.
616 */
35bba9a3 617static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 618{
619 struct nfs4_client *clp;
620
35bba9a3
BF		 621 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
622 if (clp == NULL)
623 return NULL;
624 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
625 if (clp->cl_name.data == NULL) {
626 kfree(clp);
627 return NULL;
1da177e4 628 }
35bba9a3
BF		 629 memcpy(clp->cl_name.data, name.data, name.len);
630 clp->cl_name.len = name.len;
1da177e4
LT		 631 return clp;
632}
633
1b1a9b31
BF		 634static void
635shutdown_callback_client(struct nfs4_client *clp)
636{
c237dc03 637 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 638
1b1a9b31 639 if (clnt) {
404ec117
BF		 640 /*
641 * Callback threads take a reference on the client, so there
642 * should be no outstanding callbacks at this point.
643 */
c237dc03 644 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 645 rpc_shutdown_client(clnt);
646 }
c237dc03
BF		 647 if (clp->cl_cb_conn.cb_cred) {
648 put_rpccred(clp->cl_cb_conn.cb_cred);
649 clp->cl_cb_conn.cb_cred = NULL;
3cef9ab2 650 }
1b1a9b31
BF		 651}
652
1da177e4
LT		 653static inline void
654free_client(struct nfs4_client *clp)
655{
1b1a9b31 656 shutdown_callback_client(clp);
1da177e4
LT		 657 if (clp->cl_cred.cr_group_info)
658 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 659 kfree(clp->cl_principal);
1da177e4
LT		 660 kfree(clp->cl_name.data);
661 kfree(clp);
662}
663
664void
665put_nfs4_client(struct nfs4_client *clp)
666{
667 if (atomic_dec_and_test(&clp->cl_count))
668 free_client(clp);
669}
670
671static void
672expire_client(struct nfs4_client *clp)
673{
674 struct nfs4_stateowner *sop;
675 struct nfs4_delegation *dp;
1da177e4
LT		 676 struct list_head reaplist;
677
678 dprintk("NFSD: expire_client cl_count %d\n",
679 atomic_read(&clp->cl_count));
680
1da177e4
LT		 681 INIT_LIST_HEAD(&reaplist);
682 spin_lock(&recall_lock);
ea1da636
N		 683 while (!list_empty(&clp->cl_delegations)) {
684 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 685 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
686 dp->dl_flock);
ea1da636 687 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 688 list_move(&dp->dl_recall_lru, &reaplist);
689 }
690 spin_unlock(&recall_lock);
691 while (!list_empty(&reaplist)) {
692 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
693 list_del_init(&dp->dl_recall_lru);
694 unhash_delegation(dp);
695 }
696 list_del(&clp->cl_idhash);
697 list_del(&clp->cl_strhash);
698 list_del(&clp->cl_lru);
ea1da636
N		 699 while (!list_empty(&clp->cl_openowners)) {
700 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 701 release_openowner(sop);
1da177e4 702 }
c4bf7868
ME		 703 while (!list_empty(&clp->cl_sessions)) {
704 struct nfsd4_session *ses;
705 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
706 se_perclnt);
707 release_session(ses);
708 }
1da177e4
LT		 709 put_nfs4_client(clp);
710}
711
35bba9a3
BF		 712static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
713{
1da177e4
LT		 714 struct nfs4_client *clp;
715
35bba9a3
BF		 716 clp = alloc_client(name);
717 if (clp == NULL)
718 return NULL;
a55370a3 719 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 720 atomic_set(&clp->cl_count, 1);
c237dc03 721 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT		 722 INIT_LIST_HEAD(&clp->cl_idhash);
723 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 724 INIT_LIST_HEAD(&clp->cl_openowners);
725 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 726 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 727 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 728 return clp;
729}
730
35bba9a3
BF		 731static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
732{
733 memcpy(target->cl_verifier.data, source->data,
734 sizeof(target->cl_verifier.data));
1da177e4
LT		 735}
736
35bba9a3
BF		 737static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
738{
1da177e4
LT		 739 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
740 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
741}
742
35bba9a3
BF		 743static void copy_cred(struct svc_cred *target, struct svc_cred *source)
744{
1da177e4
LT		 745 target->cr_uid = source->cr_uid;
746 target->cr_gid = source->cr_gid;
747 target->cr_group_info = source->cr_group_info;
748 get_group_info(target->cr_group_info);
749}
750
35bba9a3 751static int same_name(const char *n1, const char *n2)
599e0a22 752{
a55370a3 753 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 754}
755
756static int
599e0a22
BF		 757same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
758{
759 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 760}
761
762static int
599e0a22
BF		 763same_clid(clientid_t *cl1, clientid_t *cl2)
764{
765 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 766}
767
768/* XXX what about NGROUP */
769static int
599e0a22
BF		 770same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
771{
772 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 773}
774
5ec7b46c
BF		 775static void gen_clid(struct nfs4_client *clp)
776{
777 static u32 current_clientid = 1;
778
1da177e4
LT		 779 clp->cl_clientid.cl_boot = boot_time;
780 clp->cl_clientid.cl_id = current_clientid++;
781}
782
deda2faa
BF		 783static void gen_confirm(struct nfs4_client *clp)
784{
785 static u32 i;
786 u32 *p;
1da177e4 787
1da177e4 788 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 789 *p++ = get_seconds();
790 *p++ = i++;
1da177e4
LT		 791}
792
35bba9a3
BF		 793static int check_name(struct xdr_netobj name)
794{
1da177e4
LT		 795 if (name.len == 0)
796 return 0;
797 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 798 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 799 return 0;
800 }
801 return 1;
802}
803
fd39ca9a 804static void
1da177e4
LT		 805add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
806{
807 unsigned int idhashval;
808
809 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
810 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
811 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
812 list_add_tail(&clp->cl_lru, &client_lru);
813 clp->cl_time = get_seconds();
814}
815
fd39ca9a 816static void
1da177e4
LT		 817move_to_confirmed(struct nfs4_client *clp)
818{
819 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
820 unsigned int strhashval;
821
822 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
823 list_del_init(&clp->cl_strhash);
f116629d 824 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 825 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 826 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
827 renew_client(clp);
828}
829
830static struct nfs4_client *
831find_confirmed_client(clientid_t *clid)
832{
833 struct nfs4_client *clp;
834 unsigned int idhashval = clientid_hashval(clid->cl_id);
835
836 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 837 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 838 return clp;
839 }
840 return NULL;
841}
842
843static struct nfs4_client *
844find_unconfirmed_client(clientid_t *clid)
845{
846 struct nfs4_client *clp;
847 unsigned int idhashval = clientid_hashval(clid->cl_id);
848
849 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 850 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 851 return clp;
852 }
853 return NULL;
854}
855
a1bcecd2
AA		 856/*
857 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
858 * parameter. Matching is based on the fact the at least one of the
859 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
860 *
861 * FIXME: we need to unify the clientid namespaces for nfsv4.x
862 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
863 * and SET_CLIENTID{,_CONFIRM}
864 */
865static inline int
866match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
867{
868 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
869 return use_exchange_id == has_exchange_flags;
870}
871
28ce6054 872static struct nfs4_client *
a1bcecd2
AA		 873find_confirmed_client_by_str(const char *dname, unsigned int hashval,
874 bool use_exchange_id)
28ce6054
N		 875{
876 struct nfs4_client *clp;
877
878 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 879 if (same_name(clp->cl_recdir, dname) &&
880 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 881 return clp;
882 }
883 return NULL;
884}
885
886static struct nfs4_client *
a1bcecd2
AA		 887find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
888 bool use_exchange_id)
28ce6054
N		 889{
890 struct nfs4_client *clp;
891
892 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 893 if (same_name(clp->cl_recdir, dname) &&
894 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 895 return clp;
896 }
897 return NULL;
898}
899
fd39ca9a 900static void
1da177e4
LT		 901gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
902{
c237dc03 903 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
1da177e4
LT		 904
905 /* Currently, we only support tcp for the callback channel */
906 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
907 goto out_err;
908
aa9a4ec7
JL		 909 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
910 se->se_callback_addr_len,
911 (struct sockaddr *) &cb->cb_addr,
912 sizeof(cb->cb_addr));
913
914 if (!cb->cb_addrlen || cb->cb_addr.ss_family != AF_INET)
1da177e4 915 goto out_err;
aa9a4ec7 916
ab52ae6d 917 cb->cb_minorversion = 0;
1da177e4
LT		 918 cb->cb_prog = se->se_callback_prog;
919 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 920 return;
921out_err:
aa9a4ec7
JL		 922 cb->cb_addr.ss_family = AF_UNSPEC;
923 cb->cb_addrlen = 0;
849823c5 924 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 925 "will not receive delegations\n",
926 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
927
1da177e4
LT		 928 return;
929}
930
074fe897
AA		 931void
932nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
933{
934 struct nfsd4_compoundres *resp = rqstp->rq_resp;
935
936 resp->cstate.statp = statp;
937}
938
939/*
940 * Dereference the result pages.
941 */
942static void
943nfsd4_release_respages(struct page **respages, short resused)
944{
945 int i;
946
947 dprintk("--> %s\n", __func__);
948 for (i = 0; i < resused; i++) {
949 if (!respages[i])
950 continue;
951 put_page(respages[i]);
952 respages[i] = NULL;
953 }
954}
955
956static void
957nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
958{
959 int i;
960
961 for (i = 0; i < count; i++) {
962 topages[i] = frompages[i];
963 if (!topages[i])
964 continue;
965 get_page(topages[i]);
966 }
967}
968
969/*
970 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
971 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
972 * length of the XDR response is less than se_fmaxresp_cached
973 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
974 * of the reply (e.g. readdir).
975 *
976 * Store the base and length of the rq_req.head[0] page
977 * of the NFSv4.1 data, just past the rpc header.
978 */
979void
980nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
981{
982 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
983 struct svc_rqst *rqstp = resp->rqstp;
984 struct nfsd4_compoundargs *args = rqstp->rq_argp;
985 struct nfsd4_op *op = &args->ops[resp->opcnt];
986 struct kvec *resv = &rqstp->rq_res.head[0];
987
988 dprintk("--> %s entry %p\n", __func__, entry);
989
990 /* Don't cache a failed OP_SEQUENCE. */
991 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
992 return;
bf864a31 993
074fe897 994 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 995 entry->ce_opcnt = resp->opcnt;
996 entry->ce_status = resp->cstate.status;
997
998 /*
999 * Don't need a page to cache just the sequence operation - the slot
1000 * does this for us!
1001 */
1002
1003 if (nfsd4_not_cached(resp)) {
1004 entry->ce_resused = 0;
1005 entry->ce_rpchdrlen = 0;
1006 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1007 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1008 return;
1009 }
074fe897
AA		 1010 entry->ce_resused = rqstp->rq_resused;
1011 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1012 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1013 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1014 entry->ce_resused);
074fe897
AA		 1015 entry->ce_datav.iov_base = resp->cstate.statp;
1016 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1017 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1018 /* Current request rpc header length*/
1019 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1020 (char *)page_address(rqstp->rq_respages[0]);
1021}
1022
1023/*
1024 * We keep the rpc header, but take the nfs reply from the replycache.
1025 */
1026static int
1027nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1028 struct nfsd4_cache_entry *entry)
1029{
1030 struct svc_rqst *rqstp = resp->rqstp;
1031 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1032 int len;
1033
1034 /* Current request rpc header length*/
1035 len = (char *)resp->cstate.statp -
1036 (char *)page_address(rqstp->rq_respages[0]);
1037 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1038 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1039 entry->ce_datav.iov_len);
1040 return 0;
1041 }
1042 /* copy the cached reply nfsd data past the current rpc header */
1043 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1044 entry->ce_datav.iov_len);
1045 resv->iov_len = len + entry->ce_datav.iov_len;
1046 return 1;
1047}
1048
abfabf8c
AA		 1049/*
1050 * Encode the replay sequence operation from the slot values.
1051 * If cachethis is FALSE encode the uncached rep error on the next
1052 * operation which sets resp->p and increments resp->opcnt for
1053 * nfs4svc_encode_compoundres.
1054 *
1055 */
1056static __be32
1057nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1058 struct nfsd4_compoundres *resp)
1059{
1060 struct nfsd4_op *op;
1061 struct nfsd4_slot *slot = resp->cstate.slot;
1062
1063 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1064 resp->opcnt, resp->cstate.slot->sl_cache_entry.ce_cachethis);
1065
1066 /* Encode the replayed sequence operation */
1067 op = &args->ops[resp->opcnt - 1];
1068 nfsd4_encode_operation(resp, op);
1069
1070 /* Return nfserr_retry_uncached_rep in next operation. */
1071 if (args->opcnt > 1 && slot->sl_cache_entry.ce_cachethis == 0) {
1072 op = &args->ops[resp->opcnt++];
1073 op->status = nfserr_retry_uncached_rep;
1074 nfsd4_encode_operation(resp, op);
1075 }
1076 return op->status;
1077}
1078
074fe897
AA		 1079/*
1080 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1081 * cached page. Replace any futher replay pages from the cache.
1082 */
1083__be32
bf864a31
AA		 1084nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1085 struct nfsd4_sequence *seq)
074fe897
AA		 1086{
1087 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1088 __be32 status;
1089
1090 dprintk("--> %s entry %p\n", __func__, entry);
1091
bf864a31
AA		 1092 /*
1093 * If this is just the sequence operation, we did not keep
1094 * a page in the cache entry because we can just use the
1095 * slot info stored in struct nfsd4_sequence that was checked
1096 * against the slot in nfsd4_sequence().
1097 *
1098 * This occurs when seq->cachethis is FALSE, or when the client
1099 * session inactivity timer fires and a solo sequence operation
1100 * is sent (lease renewal).
1101 */
abfabf8c
AA		 1102 seq->maxslots = resp->cstate.session->se_fchannel.maxreqs;
1103
1104 /* Either returns 0 or nfserr_retry_uncached */
1105 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1106 if (status == nfserr_retry_uncached_rep)
1107 return status;
074fe897
AA		 1108
1109 if (!nfsd41_copy_replay_data(resp, entry)) {
1110 /*
1111 * Not enough room to use the replay rpc header, send the
1112 * cached header. Release all the allocated result pages.
1113 */
1114 svc_free_res_pages(resp->rqstp);
1115 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1116 entry->ce_resused);
1117 } else {
1118 /* Release all but the first allocated result page */
1119
1120 resp->rqstp->rq_resused--;
1121 svc_free_res_pages(resp->rqstp);
1122
1123 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1124 &entry->ce_respages[1],
1125 entry->ce_resused - 1);
1126 }
1127
1128 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1129 resp->opcnt = entry->ce_opcnt;
1130 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1131 status = entry->ce_status;
1132
1133 return status;
1134}
1135
0733d213
AA		 1136/*
1137 * Set the exchange_id flags returned by the server.
1138 */
1139static void
1140nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1141{
1142 /* pNFS is not supported */
1143 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1144
1145 /* Referrals are supported, Migration is not. */
1146 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1147
1148 /* set the wire flags to return to client. */
1149 clid->flags = new->cl_exchange_flags;
1150}
1151
069b6ad4
AA		 1152__be32
1153nfsd4_exchange_id(struct svc_rqst *rqstp,
1154 struct nfsd4_compound_state *cstate,
1155 struct nfsd4_exchange_id *exid)
1156{
0733d213
AA		 1157 struct nfs4_client *unconf, *conf, *new;
1158 int status;
1159 unsigned int strhashval;
1160 char dname[HEXDIR_LEN];
363168b4 1161 char addr_str[INET6_ADDRSTRLEN];
0733d213 1162 nfs4_verifier verf = exid->verifier;
363168b4 1163 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1164
363168b4 1165 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1166 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1167 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1168 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1169 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1170
1171 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1172 return nfserr_inval;
1173
1174 /* Currently only support SP4_NONE */
1175 switch (exid->spa_how) {
1176 case SP4_NONE:
1177 break;
1178 case SP4_SSV:
1179 return nfserr_encr_alg_unsupp;
1180 default:
1181 BUG(); /* checked by xdr code */
1182 case SP4_MACH_CRED:
1183 return nfserr_serverfault; /* no excuse :-/ */
1184 }
1185
1186 status = nfs4_make_rec_clidname(dname, &exid->clname);
1187
1188 if (status)
1189 goto error;
1190
1191 strhashval = clientstr_hashval(dname);
1192
1193 nfs4_lock_state();
1194 status = nfs_ok;
1195
a1bcecd2 1196 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1197 if (conf) {
1198 if (!same_verf(&verf, &conf->cl_verifier)) {
1199 /* 18.35.4 case 8 */
1200 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1201 status = nfserr_not_same;
1202 goto out;
1203 }
1204 /* Client reboot: destroy old state */
1205 expire_client(conf);
1206 goto out_new;
1207 }
1208 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1209 /* 18.35.4 case 9 */
1210 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1211 status = nfserr_perm;
1212 goto out;
1213 }
1214 expire_client(conf);
1215 goto out_new;
1216 }
0733d213
AA		 1217 /*
1218 * Set bit when the owner id and verifier map to an already
1219 * confirmed client id (18.35.3).
1220 */
1221 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1222
1223 /*
1224 * Falling into 18.35.4 case 2, possible router replay.
1225 * Leave confirmed record intact and return same result.
1226 */
1227 copy_verf(conf, &verf);
1228 new = conf;
1229 goto out_copy;
6ddbbbfe
MS		 1230 }
1231
1232 /* 18.35.4 case 7 */
1233 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1234 status = nfserr_noent;
1235 goto out;
0733d213
AA		 1236 }
1237
a1bcecd2 1238 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1239 if (unconf) {
1240 /*
1241 * Possible retry or client restart. Per 18.35.4 case 4,
1242 * a new unconfirmed record should be generated regardless
1243 * of whether any properties have changed.
1244 */
1245 expire_client(unconf);
1246 }
1247
1248out_new:
1249 /* Normal case */
1250 new = create_client(exid->clname, dname);
1251 if (new == NULL) {
1252 status = nfserr_resource;
1253 goto out;
1254 }
1255
1256 copy_verf(new, &verf);
1257 copy_cred(&new->cl_cred, &rqstp->rq_cred);
363168b4 1258 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
0733d213
AA		 1259 gen_clid(new);
1260 gen_confirm(new);
1261 add_to_unconfirmed(new, strhashval);
1262out_copy:
1263 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1264 exid->clientid.cl_id = new->cl_clientid.cl_id;
1265
38eb76a5 1266 exid->seqid = 1;
0733d213
AA		 1267 nfsd4_set_ex_flags(new, exid);
1268
1269 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1270 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1271 status = nfs_ok;
1272
1273out:
1274 nfs4_unlock_state();
1275error:
1276 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1277 return status;
069b6ad4
AA		 1278}
1279
b85d4c01 1280static int
88e588d5 1281check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1282{
88e588d5
AA		 1283 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1284 slot_seqid);
b85d4c01
BH		 1285
1286 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1287 if (slot_inuse) {
1288 if (seqid == slot_seqid)
b85d4c01
BH		 1289 return nfserr_jukebox;
1290 else
1291 return nfserr_seq_misordered;
1292 }
1293 /* Normal */
88e588d5 1294 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1295 return nfs_ok;
1296 /* Replay */
88e588d5 1297 if (seqid == slot_seqid)
b85d4c01
BH		 1298 return nfserr_replay_cache;
1299 /* Wraparound */
88e588d5 1300 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1301 return nfs_ok;
1302 /* Misordered replay or misordered new request */
1303 return nfserr_seq_misordered;
1304}
1305
49557cc7
AA		 1306/*
1307 * Cache the create session result into the create session single DRC
1308 * slot cache by saving the xdr structure. sl_seqid has been set.
1309 * Do this for solo or embedded create session operations.
1310 */
1311static void
1312nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1313 struct nfsd4_clid_slot *slot, int nfserr)
1314{
1315 slot->sl_status = nfserr;
1316 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1317}
1318
1319static __be32
1320nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1321 struct nfsd4_clid_slot *slot)
1322{
1323 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1324 return slot->sl_status;
1325}
1326
069b6ad4
AA		 1327__be32
1328nfsd4_create_session(struct svc_rqst *rqstp,
1329 struct nfsd4_compound_state *cstate,
1330 struct nfsd4_create_session *cr_ses)
1331{
363168b4 1332 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1333 struct nfs4_client *conf, *unconf;
49557cc7 1334 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1335 int status = 0;
1336
1337 nfs4_lock_state();
1338 unconf = find_unconfirmed_client(&cr_ses->clientid);
1339 conf = find_confirmed_client(&cr_ses->clientid);
1340
1341 if (conf) {
49557cc7
AA		 1342 cs_slot = &conf->cl_cs_slot;
1343 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1344 if (status == nfserr_replay_cache) {
ec6b5d7b 1345 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1346 cs_slot->sl_seqid);
38eb76a5 1347 /* Return the cached reply status */
49557cc7 1348 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1349 goto out;
49557cc7 1350 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1351 status = nfserr_seq_misordered;
1352 dprintk("Sequence misordered!\n");
1353 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1354 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1355 goto out;
1356 }
49557cc7 1357 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1358 } else if (unconf) {
1359 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1360 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1361 status = nfserr_clid_inuse;
1362 goto out;
1363 }
1364
49557cc7
AA		 1365 cs_slot = &unconf->cl_cs_slot;
1366 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1367 if (status) {
1368 /* an unconfirmed replay returns misordered */
ec6b5d7b 1369 status = nfserr_seq_misordered;
49557cc7 1370 goto out_cache;
ec6b5d7b
AA		 1371 }
1372
49557cc7 1373 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1374 move_to_confirmed(unconf);
1375
1376 /*
1377 * We do not support RDMA or persistent sessions
1378 */
1379 cr_ses->flags &= ~SESSION4_PERSIST;
1380 cr_ses->flags &= ~SESSION4_RDMA;
1381
1382 conf = unconf;
1383 } else {
1384 status = nfserr_stale_clientid;
1385 goto out;
1386 }
1387
1388 status = alloc_init_session(rqstp, conf, cr_ses);
1389 if (status)
1390 goto out;
1391
ec6b5d7b
AA		 1392 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1393 NFS4_MAX_SESSIONID_LEN);
49557cc7 1394 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1395
49557cc7
AA		 1396out_cache:
1397 /* cache solo and embedded create sessions under the state lock */
1398 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1399out:
1400 nfs4_unlock_state();
1401 dprintk("%s returns %d\n", __func__, ntohl(status));
1402 return status;
069b6ad4
AA		 1403}
1404
1405__be32
1406nfsd4_destroy_session(struct svc_rqst *r,
1407 struct nfsd4_compound_state *cstate,
1408 struct nfsd4_destroy_session *sessionid)
1409{
e10e0cfc
BH		 1410 struct nfsd4_session *ses;
1411 u32 status = nfserr_badsession;
1412
1413 /* Notes:
1414 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1415 * - Should we return nfserr_back_chan_busy if waiting for
1416 * callbacks on to-be-destroyed session?
1417 * - Do we need to clear any callback info from previous session?
1418 */
1419
1420 dump_sessionid(__func__, &sessionid->sessionid);
1421 spin_lock(&sessionid_lock);
1422 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1423 if (!ses) {
1424 spin_unlock(&sessionid_lock);
1425 goto out;
1426 }
1427
1428 unhash_session(ses);
1429 spin_unlock(&sessionid_lock);
1430
1431 /* wait for callbacks */
1432 shutdown_callback_client(ses->se_client);
1433 nfsd4_put_session(ses);
1434 status = nfs_ok;
1435out:
1436 dprintk("%s returns %d\n", __func__, ntohl(status));
1437 return status;
069b6ad4
AA		 1438}
1439
1440__be32
b85d4c01 1441nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1442 struct nfsd4_compound_state *cstate,
1443 struct nfsd4_sequence *seq)
1444{
f9bb94c4 1445 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1446 struct nfsd4_session *session;
1447 struct nfsd4_slot *slot;
1448 int status;
1449
f9bb94c4
AA		 1450 if (resp->opcnt != 1)
1451 return nfserr_sequence_pos;
1452
b85d4c01
BH		 1453 spin_lock(&sessionid_lock);
1454 status = nfserr_badsession;
1455 session = find_in_sessionid_hashtbl(&seq->sessionid);
1456 if (!session)
1457 goto out;
1458
1459 status = nfserr_badslot;
6c18ba9f 1460 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1461 goto out;
1462
1463 slot = &session->se_slots[seq->slotid];
1464 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1465
88e588d5 1466 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1467 if (status == nfserr_replay_cache) {
1468 cstate->slot = slot;
1469 cstate->session = session;
da3846a2 1470 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1471 * for nfsd4_svc_encode_compoundres processing */
1472 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1473 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1474 goto replay_cache;
1475 }
1476 if (status)
1477 goto out;
1478
1479 /* Success! bump slot seqid */
1480 slot->sl_inuse = true;
1481 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1482 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1483 /* Always set the cache entry cachethis for solo sequence */
1484 if (nfsd4_is_solo_sequence(resp))
1485 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1486
1487 cstate->slot = slot;
1488 cstate->session = session;
1489
1490replay_cache:
1491 /* Renew the clientid on success and on replay.
1492 * Hold a session reference until done processing the compound:
1493 * nfsd4_put_session called only if the cstate slot is set.
1494 */
1495 renew_client(session->se_client);
1496 nfsd4_get_session(session);
1497out:
1498 spin_unlock(&sessionid_lock);
1499 dprintk("%s: return %d\n", __func__, ntohl(status));
1500 return status;
069b6ad4
AA		 1501}
1502
b37ad28b 1503__be32
b591480b
BF		 1504nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1505 struct nfsd4_setclientid *setclid)
1da177e4 1506{
363168b4 1507 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1508 struct xdr_netobj clname = {
1509 .len = setclid->se_namelen,
1510 .data = setclid->se_name,
1511 };
1512 nfs4_verifier clverifier = setclid->se_verf;
1513 unsigned int strhashval;
28ce6054 1514 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1515 __be32 status;
68e76ad0 1516 char *princ;
a55370a3 1517 char dname[HEXDIR_LEN];
1da177e4 1518
1da177e4 1519 if (!check_name(clname))
73aea4ec 1520 return nfserr_inval;
1da177e4 1521
a55370a3
N		 1522 status = nfs4_make_rec_clidname(dname, &clname);
1523 if (status)
73aea4ec 1524 return status;
a55370a3 1525
1da177e4
LT		 1526 /*
1527 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1528 * We get here on a DRC miss.
1529 */
1530
a55370a3 1531 strhashval = clientstr_hashval(dname);
1da177e4 1532
1da177e4 1533 nfs4_lock_state();
a1bcecd2 1534 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1535 if (conf) {
a186e767 1536 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1537 status = nfserr_clid_inuse;
026722c2 1538 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1539 char addr_str[INET6_ADDRSTRLEN];
1540 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1541 sizeof(addr_str));
1542 dprintk("NFSD: setclientid: string in use by client "
1543 "at %s\n", addr_str);
1da177e4
LT		 1544 goto out;
1545 }
1da177e4 1546 }
a186e767
BF		 1547 /*
1548 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1549 * has a description of SETCLIENTID request processing consisting
1550 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1551 */
a1bcecd2 1552 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1553 status = nfserr_resource;
1554 if (!conf) {
a186e767
BF		 1555 /*
1556 * RFC 3530 14.2.33 CASE 4:
1557 * placed first, because it is the normal case
1da177e4
LT		 1558 */
1559 if (unconf)
1560 expire_client(unconf);
a55370a3
N		 1561 new = create_client(clname, dname);
1562 if (new == NULL)
1da177e4 1563 goto out;
1da177e4 1564 gen_clid(new);
599e0a22 1565 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1566 /*
a186e767
BF		 1567 * RFC 3530 14.2.33 CASE 1:
1568 * probable callback update
1da177e4 1569 */
31f4a6c1
N		 1570 if (unconf) {
1571 /* Note this is removing unconfirmed {*x***},
1572 * which is stronger than RFC recommended {vxc**}.
1573 * This has the advantage that there is at most
1574 * one {*x***} in either list at any time.
1575 */
1576 expire_client(unconf);
1da177e4 1577 }
a55370a3
N		 1578 new = create_client(clname, dname);
1579 if (new == NULL)
1da177e4 1580 goto out;
1da177e4 1581 copy_clid(new, conf);
1da177e4
LT		 1582 } else if (!unconf) {
1583 /*
a186e767
BF		 1584 * RFC 3530 14.2.33 CASE 2:
1585 * probable client reboot; state will be removed if
1586 * confirmed.
1da177e4 1587 */
a55370a3
N		 1588 new = create_client(clname, dname);
1589 if (new == NULL)
1da177e4 1590 goto out;
1da177e4 1591 gen_clid(new);
49ba8781 1592 } else {
a186e767
BF		 1593 /*
1594 * RFC 3530 14.2.33 CASE 3:
1595 * probable client reboot; state will be removed if
1596 * confirmed.
1da177e4
LT		 1597 */
1598 expire_client(unconf);
a55370a3
N		 1599 new = create_client(clname, dname);
1600 if (new == NULL)
1da177e4 1601 goto out;
1da177e4 1602 gen_clid(new);
1da177e4 1603 }
c175b83c 1604 copy_verf(new, &clverifier);
363168b4 1605 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
61054b14 1606 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1607 princ = svc_gss_principal(rqstp);
1608 if (princ) {
1609 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1610 if (new->cl_principal == NULL) {
1611 free_client(new);
1612 goto out;
1613 }
1614 }
c175b83c
BF		 1615 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1616 gen_confirm(new);
1617 gen_callback(new, setclid);
1618 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1619 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1620 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1621 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1622 status = nfs_ok;
1623out:
1624 nfs4_unlock_state();
1625 return status;
1626}
1627
1628
1629/*
a186e767
BF		 1630 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1631 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1632 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1633 */
b37ad28b 1634__be32
b591480b
BF		 1635nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1636 struct nfsd4_compound_state *cstate,
1637 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1638{
363168b4 1639 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1640 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1641 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1642 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1643 __be32 status;
1da177e4
LT		 1644
1645 if (STALE_CLIENTID(clid))
1646 return nfserr_stale_clientid;
1647 /*
1648 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1649 * We get here on a DRC miss.
1650 */
1651
1652 nfs4_lock_state();
21ab45a4
N		 1653
1654 conf = find_confirmed_client(clid);
1655 unconf = find_unconfirmed_client(clid);
1656
1657 status = nfserr_clid_inuse;
363168b4 1658 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1659 goto out;
363168b4 1660 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1661 goto out;
1662
a186e767
BF		 1663 /*
1664 * section 14.2.34 of RFC 3530 has a description of
1665 * SETCLIENTID_CONFIRM request processing consisting
1666 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1667 */
366e0c1d 1668 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1669 /*
1670 * RFC 3530 14.2.34 CASE 1:
1671 * callback update
1672 */
599e0a22 1673 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1674 status = nfserr_clid_inuse;
1675 else {
1a69c179
N		 1676 /* XXX: We just turn off callbacks until we can handle
1677 * change request correctly. */
c237dc03 1678 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1679 expire_client(unconf);
1da177e4 1680 status = nfs_ok;
1a69c179 1681
1da177e4 1682 }
f3aba4e5 1683 } else if (conf && !unconf) {
a186e767
BF		 1684 /*
1685 * RFC 3530 14.2.34 CASE 2:
1686 * probable retransmitted request; play it safe and
1687 * do nothing.
7c79f737 1688 */
599e0a22 1689 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1690 status = nfserr_clid_inuse;
21ab45a4 1691 else
1da177e4 1692 status = nfs_ok;
7c79f737 1693 } else if (!conf && unconf
599e0a22 1694 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1695 /*
1696 * RFC 3530 14.2.34 CASE 3:
1697 * Normal case; new or rebooted client:
7c79f737 1698 */
599e0a22 1699 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1700 status = nfserr_clid_inuse;
1701 } else {
1a69c179
N		 1702 unsigned int hash =
1703 clientstr_hashval(unconf->cl_recdir);
1704 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1705 hash, false);
1a69c179 1706 if (conf) {
c7b9a459 1707 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1708 expire_client(conf);
1709 }
1da177e4 1710 move_to_confirmed(unconf);
21ab45a4 1711 conf = unconf;
46f8a64b 1712 nfsd4_probe_callback(conf);
1a69c179 1713 status = nfs_ok;
1da177e4 1714 }
599e0a22
BF		 1715 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1716 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1717 &confirm)))) {
a186e767
BF		 1718 /*
1719 * RFC 3530 14.2.34 CASE 4:
1720 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1721 */
1da177e4 1722 status = nfserr_stale_clientid;
7c79f737 1723 } else {
08e8987c
N		 1724 /* check that we have hit one of the cases...*/
1725 status = nfserr_clid_inuse;
1726 }
1da177e4 1727out:
1da177e4
LT		 1728 nfs4_unlock_state();
1729 return status;
1730}
1731
1da177e4
LT		 1732/* OPEN Share state helper functions */
1733static inline struct nfs4_file *
1734alloc_init_file(struct inode *ino)
1735{
1736 struct nfs4_file *fp;
1737 unsigned int hashval = file_hashval(ino);
1738
e60d4398
N		 1739 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1740 if (fp) {
8b671b80 1741 atomic_set(&fp->fi_ref, 1);
1da177e4 1742 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1743 INIT_LIST_HEAD(&fp->fi_stateids);
1744 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1745 spin_lock(&recall_lock);
1da177e4 1746 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1747 spin_unlock(&recall_lock);
1da177e4
LT		 1748 fp->fi_inode = igrab(ino);
1749 fp->fi_id = current_fileid++;
47f9940c 1750 fp->fi_had_conflict = false;
1da177e4
LT		 1751 return fp;
1752 }
1753 return NULL;
1754}
1755
e60d4398 1756static void
e18b890b 1757nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1758{
e60d4398
N		 1759 if (*slab == NULL)
1760 return;
1a1d92c1 1761 kmem_cache_destroy(*slab);
e60d4398 1762 *slab = NULL;
1da177e4
LT		 1763}
1764
e8ff2a84 1765void
1da177e4
LT		 1766nfsd4_free_slabs(void)
1767{
e60d4398
N		 1768 nfsd4_free_slab(&stateowner_slab);
1769 nfsd4_free_slab(&file_slab);
5ac049ac 1770 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1771 nfsd4_free_slab(&deleg_slab);
e60d4398 1772}
1da177e4 1773
e60d4398
N		 1774static int
1775nfsd4_init_slabs(void)
1776{
1777 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1778 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1779 if (stateowner_slab == NULL)
1780 goto out_nomem;
1781 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1782 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1783 if (file_slab == NULL)
1784 goto out_nomem;
5ac049ac 1785 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1786 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1787 if (stateid_slab == NULL)
1788 goto out_nomem;
5b2d21c1 1789 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1790 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1791 if (deleg_slab == NULL)
1792 goto out_nomem;
e60d4398
N		 1793 return 0;
1794out_nomem:
1795 nfsd4_free_slabs();
1796 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1797 return -ENOMEM;
1da177e4
LT		 1798}
1799
1800void
1801nfs4_free_stateowner(struct kref *kref)
1802{
1803 struct nfs4_stateowner *sop =
1804 container_of(kref, struct nfs4_stateowner, so_ref);
1805 kfree(sop->so_owner.data);
1806 kmem_cache_free(stateowner_slab, sop);
1807}
1808
1809static inline struct nfs4_stateowner *
1810alloc_stateowner(struct xdr_netobj *owner)
1811{
1812 struct nfs4_stateowner *sop;
1813
1814 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1815 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1816 memcpy(sop->so_owner.data, owner->data, owner->len);
1817 sop->so_owner.len = owner->len;
1818 kref_init(&sop->so_ref);
1819 return sop;
1820 }
1821 kmem_cache_free(stateowner_slab, sop);
1822 }
1823 return NULL;
1824}
1825
1826static struct nfs4_stateowner *
1827alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1828 struct nfs4_stateowner *sop;
1829 struct nfs4_replay *rp;
1830 unsigned int idhashval;
1831
1832 if (!(sop = alloc_stateowner(&open->op_owner)))
1833 return NULL;
1834 idhashval = ownerid_hashval(current_ownerid);
1835 INIT_LIST_HEAD(&sop->so_idhash);
1836 INIT_LIST_HEAD(&sop->so_strhash);
1837 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1838 INIT_LIST_HEAD(&sop->so_stateids);
1839 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1840 INIT_LIST_HEAD(&sop->so_close_lru);
1841 sop->so_time = 0;
1842 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1843 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1844 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1845 sop->so_is_open_owner = 1;
1846 sop->so_id = current_ownerid++;
1847 sop->so_client = clp;
1848 sop->so_seqid = open->op_seqid;
1849 sop->so_confirmed = 0;
1850 rp = &sop->so_replay;
de1ae286 1851 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1852 rp->rp_buflen = 0;
1853 rp->rp_buf = rp->rp_ibuf;
1854 return sop;
1855}
1856
1da177e4
LT		 1857static inline void
1858init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1859 struct nfs4_stateowner *sop = open->op_stateowner;
1860 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1861
1862 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1863 INIT_LIST_HEAD(&stp->st_perstateowner);
1864 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1865 INIT_LIST_HEAD(&stp->st_perfile);
1866 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1867 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1868 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1869 stp->st_stateowner = sop;
13cd2184 1870 get_nfs4_file(fp);
1da177e4 1871 stp->st_file = fp;
78155ed7 1872 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1873 stp->st_stateid.si_stateownerid = sop->so_id;
1874 stp->st_stateid.si_fileid = fp->fi_id;
1875 stp->st_stateid.si_generation = 0;
1876 stp->st_access_bmap = 0;
1877 stp->st_deny_bmap = 0;
84459a11
AA		 1878 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1879 &stp->st_access_bmap);
1da177e4 1880 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1881 stp->st_openstp = NULL;
1da177e4
LT		 1882}
1883
fd39ca9a 1884static void
1da177e4
LT		 1885move_to_close_lru(struct nfs4_stateowner *sop)
1886{
1887 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1888
358dd55a 1889 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1890 sop->so_time = get_seconds();
1891}
1892
1da177e4 1893static int
599e0a22
BF		 1894same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1895 clientid_t *clid)
1896{
1897 return (sop->so_owner.len == owner->len) &&
1898 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1899 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1900}
1901
1902static struct nfs4_stateowner *
1903find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1904{
1905 struct nfs4_stateowner *so = NULL;
1906
1907 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1908 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1909 return so;
1910 }
1911 return NULL;
1912}
1913
1914/* search file_hashtbl[] for file */
1915static struct nfs4_file *
1916find_file(struct inode *ino)
1917{
1918 unsigned int hashval = file_hashval(ino);
1919 struct nfs4_file *fp;
1920
8b671b80 1921 spin_lock(&recall_lock);
1da177e4 1922 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1923 if (fp->fi_inode == ino) {
1924 get_nfs4_file(fp);
8b671b80 1925 spin_unlock(&recall_lock);
1da177e4 1926 return fp;
13cd2184 1927 }
1da177e4 1928 }
8b671b80 1929 spin_unlock(&recall_lock);
1da177e4
LT		 1930 return NULL;
1931}
1932
d87a8ade 1933static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1934{
d87a8ade 1935 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1936 return 0;
d87a8ade
AA		 1937 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1938 return 0;
1939 x &= ~NFS4_SHARE_ACCESS_MASK;
1940 if (minorversion && x) {
1941 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1942 return 0;
1943 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1944 return 0;
1945 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1946 }
1947 if (x)
8838dc43
BF		 1948 return 0;
1949 return 1;
ba5a6a19
BF		 1950}
1951
8838dc43 1952static inline int deny_valid(u32 x)
ba5a6a19 1953{
8838dc43
BF		 1954 /* Note: unlike access bits, deny bits may be zero. */
1955 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1956}
1da177e4 1957
4f83aa30
BF		 1958/*
1959 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1960 * st_{access,deny}_bmap field of the stateid, in order to track not
1961 * only what share bits are currently in force, but also what
1962 * combinations of share bits previous opens have used. This allows us
1963 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1964 * return an error if the client attempt to downgrade to a combination
1965 * of share bits not explicable by closing some of its previous opens.
1966 *
1967 * XXX: This enforcement is actually incomplete, since we don't keep
1968 * track of access/deny bit combinations; so, e.g., we allow:
1969 *
1970 * OPEN allow read, deny write
1971 * OPEN allow both, deny none
1972 * DOWNGRADE allow read, deny none
1973 *
1974 * which we should reject.
1975 */
fd39ca9a 1976static void
1da177e4
LT		 1977set_access(unsigned int *access, unsigned long bmap) {
1978 int i;
1979
1980 *access = 0;
1981 for (i = 1; i < 4; i++) {
1982 if (test_bit(i, &bmap))
1983 *access |= i;
1984 }
1985}
1986
fd39ca9a 1987static void
1da177e4
LT		 1988set_deny(unsigned int *deny, unsigned long bmap) {
1989 int i;
1990
1991 *deny = 0;
1992 for (i = 0; i < 4; i++) {
1993 if (test_bit(i, &bmap))
1994 *deny |= i ;
1995 }
1996}
1997
1998static int
1999test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2000 unsigned int access, deny;
2001
2002 set_access(&access, stp->st_access_bmap);
2003 set_deny(&deny, stp->st_deny_bmap);
2004 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2005 return 0;
2006 return 1;
2007}
2008
2009/*
2010 * Called to check deny when READ with all zero stateid or
2011 * WRITE with all zero or all one stateid
2012 */
b37ad28b 2013static __be32
1da177e4
LT		 2014nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2015{
2016 struct inode *ino = current_fh->fh_dentry->d_inode;
2017 struct nfs4_file *fp;
2018 struct nfs4_stateid *stp;
b37ad28b 2019 __be32 ret;
1da177e4
LT		 2020
2021 dprintk("NFSD: nfs4_share_conflict\n");
2022
2023 fp = find_file(ino);
13cd2184
N		 2024 if (!fp)
2025 return nfs_ok;
b700949b 2026 ret = nfserr_locked;
1da177e4 2027 /* Search for conflicting share reservations */
13cd2184
N		 2028 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2029 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2030 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2031 goto out;
1da177e4 2032 }
13cd2184
N		 2033 ret = nfs_ok;
2034out:
2035 put_nfs4_file(fp);
2036 return ret;
1da177e4
LT		 2037}
2038
2039static inline void
2040nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2041{
2042 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2043 drop_file_write_access(filp);
1da177e4
LT		 2044 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2045 }
2046}
2047
1da177e4
LT		 2048/*
2049 * Spawn a thread to perform a recall on the delegation represented
2050 * by the lease (file_lock)
2051 *
2052 * Called from break_lease() with lock_kernel() held.
2053 * Note: we assume break_lease will only call this *once* for any given
2054 * lease.
2055 */
2056static
2057void nfsd_break_deleg_cb(struct file_lock *fl)
2058{
63e4863f 2059 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2060
2061 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2062 if (!dp)
2063 return;
2064
2065 /* We're assuming the state code never drops its reference
2066 * without first removing the lease. Since we're in this lease
2067 * callback (and since the lease code is serialized by the kernel
2068 * lock) we know the server hasn't removed the lease yet, we know
2069 * it's safe to take a reference: */
2070 atomic_inc(&dp->dl_count);
cfdcad4d 2071 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2072
2073 spin_lock(&recall_lock);
2074 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2075 spin_unlock(&recall_lock);
2076
2077 /* only place dl_time is set. protected by lock_kernel*/
2078 dp->dl_time = get_seconds();
2079
0272e1fd
BF		 2080 /*
2081 * We don't want the locks code to timeout the lease for us;
2082 * we'll remove it ourself if the delegation isn't returned
2083 * in time.
2084 */
2085 fl->fl_break_time = 0;
1da177e4 2086
63e4863f
BF		 2087 dp->dl_file->fi_had_conflict = true;
2088 nfsd4_cb_recall(dp);
1da177e4
LT		 2089}
2090
2091/*
2092 * The file_lock is being reapd.
2093 *
2094 * Called by locks_free_lock() with lock_kernel() held.
2095 */
2096static
2097void nfsd_release_deleg_cb(struct file_lock *fl)
2098{
2099 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2100
2101 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2102
2103 if (!(fl->fl_flags & FL_LEASE) || !dp)
2104 return;
2105 dp->dl_flock = NULL;
2106}
2107
2108/*
2109 * Set the delegation file_lock back pointer.
2110 *
a9933cea 2111 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2112 */
2113static
2114void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2115{
2116 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2117
2118 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2119 if (!dp)
2120 return;
2121 dp->dl_flock = new;
2122}
2123
2124/*
a9933cea 2125 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2126 */
2127static
2128int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2129{
2130 struct nfs4_delegation *onlistd =
2131 (struct nfs4_delegation *)onlist->fl_owner;
2132 struct nfs4_delegation *tryd =
2133 (struct nfs4_delegation *)try->fl_owner;
2134
2135 if (onlist->fl_lmops != try->fl_lmops)
2136 return 0;
2137
2138 return onlistd->dl_client == tryd->dl_client;
2139}
2140
2141
2142static
2143int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2144{
2145 if (arg & F_UNLCK)
2146 return lease_modify(onlist, arg);
2147 else
2148 return -EAGAIN;
2149}
2150
fd39ca9a 2151static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2152 .fl_break = nfsd_break_deleg_cb,
2153 .fl_release_private = nfsd_release_deleg_cb,
2154 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2155 .fl_mylease = nfsd_same_client_deleg_cb,
2156 .fl_change = nfsd_change_deleg_cb,
2157};
2158
2159
b37ad28b 2160__be32
6668958f
AA		 2161nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2162 struct nfsd4_open *open)
1da177e4 2163{
1da177e4
LT		 2164 clientid_t *clientid = &open->op_clientid;
2165 struct nfs4_client *clp = NULL;
2166 unsigned int strhashval;
2167 struct nfs4_stateowner *sop = NULL;
2168
1da177e4 2169 if (!check_name(open->op_owner))
0f442aa2 2170 return nfserr_inval;
1da177e4
LT		 2171
2172 if (STALE_CLIENTID(&open->op_clientid))
2173 return nfserr_stale_clientid;
2174
2175 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2176 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2177 open->op_stateowner = sop;
2178 if (!sop) {
2179 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2180 clp = find_confirmed_client(clientid);
2181 if (clp == NULL)
0f442aa2
BF		 2182 return nfserr_expired;
2183 goto renew;
1da177e4 2184 }
6668958f
AA		 2185 /* When sessions are used, skip open sequenceid processing */
2186 if (nfsd4_has_session(cstate))
2187 goto renew;
0f442aa2
BF		 2188 if (!sop->so_confirmed) {
2189 /* Replace unconfirmed owners without checking for replay. */
2190 clp = sop->so_client;
f044ff83 2191 release_openowner(sop);
0f442aa2
BF		 2192 open->op_stateowner = NULL;
2193 goto renew;
2194 }
2195 if (open->op_seqid == sop->so_seqid - 1) {
2196 if (sop->so_replay.rp_buflen)
a90b061c 2197 return nfserr_replay_me;
0f442aa2
BF		 2198 /* The original OPEN failed so spectacularly
2199 * that we don't even have replay data saved!
2200 * Therefore, we have no choice but to continue
2201 * processing this OPEN; presumably, we'll
2202 * fail again for the same reason.
2203 */
2204 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2205 goto renew;
2206 }
2207 if (open->op_seqid != sop->so_seqid)
2208 return nfserr_bad_seqid;
1da177e4 2209renew:
0f442aa2
BF		 2210 if (open->op_stateowner == NULL) {
2211 sop = alloc_init_open_stateowner(strhashval, clp, open);
2212 if (sop == NULL)
2213 return nfserr_resource;
2214 open->op_stateowner = sop;
2215 }
2216 list_del_init(&sop->so_close_lru);
1da177e4 2217 renew_client(sop->so_client);
0f442aa2 2218 return nfs_ok;
1da177e4
LT		 2219}
2220
b37ad28b 2221static inline __be32
4a6e43e6
N		 2222nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2223{
2224 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2225 return nfserr_openmode;
2226 else
2227 return nfs_ok;
2228}
2229
52f4fb43
N		 2230static struct nfs4_delegation *
2231find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2232{
2233 struct nfs4_delegation *dp;
2234
ea1da636 2235 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2236 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2237 return dp;
2238 }
2239 return NULL;
2240}
2241
b37ad28b 2242static __be32
567d9829
N		 2243nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2244 struct nfs4_delegation **dp)
2245{
2246 int flags;
b37ad28b 2247 __be32 status = nfserr_bad_stateid;
567d9829
N		 2248
2249 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2250 if (*dp == NULL)
c44c5eeb 2251 goto out;
567d9829
N		 2252 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2253 RD_STATE : WR_STATE;
2254 status = nfs4_check_delegmode(*dp, flags);
2255 if (status)
2256 *dp = NULL;
c44c5eeb
N		 2257out:
2258 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2259 return nfs_ok;
2260 if (status)
2261 return status;
2262 open->op_stateowner->so_confirmed = 1;
2263 return nfs_ok;
567d9829
N		 2264}
2265
b37ad28b 2266static __be32
1da177e4
LT		 2267nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2268{
2269 struct nfs4_stateid *local;
b37ad28b 2270 __be32 status = nfserr_share_denied;
1da177e4
LT		 2271 struct nfs4_stateowner *sop = open->op_stateowner;
2272
8beefa24 2273 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2274 /* ignore lock owners */
2275 if (local->st_stateowner->so_is_open_owner == 0)
2276 continue;
2277 /* remember if we have seen this open owner */
2278 if (local->st_stateowner == sop)
2279 *stpp = local;
2280 /* check for conflicting share reservations */
2281 if (!test_share(local, open))
2282 goto out;
2283 }
2284 status = 0;
2285out:
2286 return status;
2287}
2288
5ac049ac
N		 2289static inline struct nfs4_stateid *
2290nfs4_alloc_stateid(void)
2291{
2292 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2293}
2294
b37ad28b 2295static __be32
1da177e4 2296nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2297 struct nfs4_delegation *dp,
1da177e4
LT		 2298 struct svc_fh *cur_fh, int flags)
2299{
2300 struct nfs4_stateid *stp;
1da177e4 2301
5ac049ac 2302 stp = nfs4_alloc_stateid();
1da177e4
LT		 2303 if (stp == NULL)
2304 return nfserr_resource;
2305
567d9829
N		 2306 if (dp) {
2307 get_file(dp->dl_vfs_file);
2308 stp->st_vfs_file = dp->dl_vfs_file;
2309 } else {
b37ad28b 2310 __be32 status;
567d9829
N		 2311 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2312 &stp->st_vfs_file);
2313 if (status) {
2314 if (status == nfserr_dropit)
2315 status = nfserr_jukebox;
5ac049ac 2316 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2317 return status;
2318 }
1da177e4 2319 }
1da177e4
LT		 2320 *stpp = stp;
2321 return 0;
2322}
2323
b37ad28b 2324static inline __be32
1da177e4
LT		 2325nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2326 struct nfsd4_open *open)
2327{
2328 struct iattr iattr = {
2329 .ia_valid = ATTR_SIZE,
2330 .ia_size = 0,
2331 };
2332 if (!open->op_truncate)
2333 return 0;
2334 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2335 return nfserr_inval;
1da177e4
LT		 2336 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2337}
2338
b37ad28b 2339static __be32
1da177e4
LT		 2340nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2341{
2342 struct file *filp = stp->st_vfs_file;
7eaa36e2 2343 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2344 unsigned int share_access, new_writer;
b37ad28b 2345 __be32 status;
1da177e4
LT		 2346
2347 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2348 new_writer = (~share_access) & open->op_share_access
2349 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2350
6c26d08f 2351 if (new_writer) {
b37ad28b
AV		 2352 int err = get_write_access(inode);
2353 if (err)
2354 return nfserrno(err);
e518f056
BH		 2355 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2356 if (err)
2357 return nfserrno(err);
2358 file_take_write(filp);
6c26d08f 2359 }
1da177e4
LT		 2360 status = nfsd4_truncate(rqstp, cur_fh, open);
2361 if (status) {
6c26d08f
BF		 2362 if (new_writer)
2363 put_write_access(inode);
1da177e4
LT		 2364 return status;
2365 }
2366 /* remember the open */
6c26d08f 2367 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2368 __set_bit(open->op_share_access, &stp->st_access_bmap);
2369 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2370
2371 return nfs_ok;
2372}
2373
2374
1da177e4 2375static void
37515177 2376nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2377{
37515177
N		 2378 open->op_stateowner->so_confirmed = 1;
2379 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2380}
2381
2382/*
2383 * Attempt to hand out a delegation.
2384 */
2385static void
2386nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2387{
2388 struct nfs4_delegation *dp;
2389 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2390 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2391 struct file_lock fl, *flp = &fl;
2392 int status, flag = 0;
2393
2394 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2395 open->op_recall = 0;
2396 switch (open->op_claim_type) {
2397 case NFS4_OPEN_CLAIM_PREVIOUS:
2398 if (!atomic_read(&cb->cb_set))
2399 open->op_recall = 1;
2400 flag = open->op_delegate_type;
2401 if (flag == NFS4_OPEN_DELEGATE_NONE)
2402 goto out;
2403 break;
2404 case NFS4_OPEN_CLAIM_NULL:
2405 /* Let's not give out any delegations till everyone's
2406 * had the chance to reclaim theirs.... */
af558e33 2407 if (locks_in_grace())
7b190fec
N		 2408 goto out;
2409 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2410 goto out;
2411 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2412 flag = NFS4_OPEN_DELEGATE_WRITE;
2413 else
2414 flag = NFS4_OPEN_DELEGATE_READ;
2415 break;
2416 default:
2417 goto out;
2418 }
1da177e4
LT		 2419
2420 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2421 if (dp == NULL) {
2422 flag = NFS4_OPEN_DELEGATE_NONE;
2423 goto out;
2424 }
2425 locks_init_lock(&fl);
2426 fl.fl_lmops = &nfsd_lease_mng_ops;
2427 fl.fl_flags = FL_LEASE;
9167f501 2428 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2429 fl.fl_end = OFFSET_MAX;
2430 fl.fl_owner = (fl_owner_t)dp;
2431 fl.fl_file = stp->st_vfs_file;
2432 fl.fl_pid = current->tgid;
2433
a9933cea 2434 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2435 * the lock_manager callbacks fl_mylease and fl_change are used
2436 */
9167f501 2437 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2438 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2439 unhash_delegation(dp);
1da177e4
LT		 2440 flag = NFS4_OPEN_DELEGATE_NONE;
2441 goto out;
2442 }
2443
2444 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2445
2446 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2447 dp->dl_stateid.si_boot,
2448 dp->dl_stateid.si_stateownerid,
2449 dp->dl_stateid.si_fileid,
2450 dp->dl_stateid.si_generation);
2451out:
7b190fec
N		 2452 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2453 && flag == NFS4_OPEN_DELEGATE_NONE
2454 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2455 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2456 open->op_delegate_type = flag;
2457}
2458
2459/*
2460 * called with nfs4_lock_state() held.
2461 */
b37ad28b 2462__be32
1da177e4
LT		 2463nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2464{
6668958f 2465 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2466 struct nfs4_file *fp = NULL;
2467 struct inode *ino = current_fh->fh_dentry->d_inode;
2468 struct nfs4_stateid *stp = NULL;
567d9829 2469 struct nfs4_delegation *dp = NULL;
b37ad28b 2470 __be32 status;
1da177e4
LT		 2471
2472 status = nfserr_inval;
d87a8ade 2473 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2474 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2475 goto out;
2476 /*
2477 * Lookup file; if found, lookup stateid and check open request,
2478 * and check for delegations in the process of being recalled.
2479 * If not found, create the nfs4_file struct
2480 */
2481 fp = find_file(ino);
2482 if (fp) {
2483 if ((status = nfs4_check_open(fp, open, &stp)))
2484 goto out;
c44c5eeb
N		 2485 status = nfs4_check_deleg(fp, open, &dp);
2486 if (status)
2487 goto out;
1da177e4 2488 } else {
c44c5eeb
N		 2489 status = nfserr_bad_stateid;
2490 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2491 goto out;
1da177e4
LT		 2492 status = nfserr_resource;
2493 fp = alloc_init_file(ino);
2494 if (fp == NULL)
2495 goto out;
2496 }
2497
2498 /*
2499 * OPEN the file, or upgrade an existing OPEN.
2500 * If truncate fails, the OPEN fails.
2501 */
2502 if (stp) {
2503 /* Stateid was found, this is an OPEN upgrade */
2504 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2505 if (status)
2506 goto out;
444c2c07 2507 update_stateid(&stp->st_stateid);
1da177e4
LT		 2508 } else {
2509 /* Stateid was not found, this is a new OPEN */
2510 int flags = 0;
9ecb6a08 2511 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2512 flags |= NFSD_MAY_READ;
1da177e4 2513 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2514 flags |= NFSD_MAY_WRITE;
567d9829
N		 2515 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2516 if (status)
1da177e4
LT		 2517 goto out;
2518 init_stateid(stp, fp, open);
2519 status = nfsd4_truncate(rqstp, current_fh, open);
2520 if (status) {
2283963f 2521 release_open_stateid(stp);
1da177e4
LT		 2522 goto out;
2523 }
6668958f
AA		 2524 if (nfsd4_has_session(&resp->cstate))
2525 update_stateid(&stp->st_stateid);
1da177e4
LT		 2526 }
2527 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2528
6668958f
AA		 2529 if (nfsd4_has_session(&resp->cstate))
2530 open->op_stateowner->so_confirmed = 1;
2531
1da177e4
LT		 2532 /*
2533 * Attempt to hand out a delegation. No error return, because the
2534 * OPEN succeeds even if we fail.
2535 */
2536 nfs4_open_delegation(current_fh, open, stp);
2537
2538 status = nfs_ok;
2539
2540 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2541 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2542 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2543out:
13cd2184
N		 2544 if (fp)
2545 put_nfs4_file(fp);
37515177
N		 2546 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2547 nfs4_set_claim_prev(open);
1da177e4
LT		 2548 /*
2549 * To finish the open response, we just need to set the rflags.
2550 */
2551 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2552 if (!open->op_stateowner->so_confirmed &&
2553 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2554 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2555
2556 return status;
2557}
2558
b37ad28b 2559__be32
b591480b
BF		 2560nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2561 clientid_t *clid)
1da177e4
LT		 2562{
2563 struct nfs4_client *clp;
b37ad28b 2564 __be32 status;
1da177e4
LT		 2565
2566 nfs4_lock_state();
2567 dprintk("process_renew(%08x/%08x): starting\n",
2568 clid->cl_boot, clid->cl_id);
2569 status = nfserr_stale_clientid;
2570 if (STALE_CLIENTID(clid))
2571 goto out;
2572 clp = find_confirmed_client(clid);
2573 status = nfserr_expired;
2574 if (clp == NULL) {
2575 /* We assume the client took too long to RENEW. */
2576 dprintk("nfsd4_renew: clientid not found!\n");
2577 goto out;
2578 }
2579 renew_client(clp);
2580 status = nfserr_cb_path_down;
ea1da636 2581 if (!list_empty(&clp->cl_delegations)
c237dc03 2582 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2583 goto out;
2584 status = nfs_ok;
2585out:
2586 nfs4_unlock_state();
2587 return status;
2588}
2589
af558e33
BF		 2590struct lock_manager nfsd4_manager = {
2591};
2592
a76b4319 2593static void
af558e33 2594nfsd4_end_grace(void)
a76b4319
N		 2595{
2596 dprintk("NFSD: end of grace period\n");
c7b9a459 2597 nfsd4_recdir_purge_old();
af558e33 2598 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2599}
2600
fd39ca9a 2601static time_t
1da177e4
LT		 2602nfs4_laundromat(void)
2603{
2604 struct nfs4_client *clp;
2605 struct nfs4_stateowner *sop;
2606 struct nfs4_delegation *dp;
2607 struct list_head *pos, *next, reaplist;
2608 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2609 time_t t, clientid_val = NFSD_LEASE_TIME;
2610 time_t u, test_val = NFSD_LEASE_TIME;
2611
2612 nfs4_lock_state();
2613
2614 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2615 if (locks_in_grace())
2616 nfsd4_end_grace();
1da177e4
LT		 2617 list_for_each_safe(pos, next, &client_lru) {
2618 clp = list_entry(pos, struct nfs4_client, cl_lru);
2619 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2620 t = clp->cl_time - cutoff;
2621 if (clientid_val > t)
2622 clientid_val = t;
2623 break;
2624 }
2625 dprintk("NFSD: purging unused client (clientid %08x)\n",
2626 clp->cl_clientid.cl_id);
c7b9a459 2627 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2628 expire_client(clp);
2629 }
2630 INIT_LIST_HEAD(&reaplist);
2631 spin_lock(&recall_lock);
2632 list_for_each_safe(pos, next, &del_recall_lru) {
2633 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2634 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2635 u = dp->dl_time - cutoff;
2636 if (test_val > u)
2637 test_val = u;
2638 break;
2639 }
2640 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2641 dp, dp->dl_flock);
2642 list_move(&dp->dl_recall_lru, &reaplist);
2643 }
2644 spin_unlock(&recall_lock);
2645 list_for_each_safe(pos, next, &reaplist) {
2646 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2647 list_del_init(&dp->dl_recall_lru);
2648 unhash_delegation(dp);
2649 }
2650 test_val = NFSD_LEASE_TIME;
2651 list_for_each_safe(pos, next, &close_lru) {
2652 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2653 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2654 u = sop->so_time - cutoff;
2655 if (test_val > u)
2656 test_val = u;
2657 break;
2658 }
2659 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2660 sop->so_id);
f044ff83 2661 release_openowner(sop);
1da177e4
LT		 2662 }
2663 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2664 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2665 nfs4_unlock_state();
2666 return clientid_val;
2667}
2668
a254b246
HH		 2669static struct workqueue_struct *laundry_wq;
2670static void laundromat_main(struct work_struct *);
2671static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2672
2673static void
c4028958 2674laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2675{
2676 time_t t;
2677
2678 t = nfs4_laundromat();
2679 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2680 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2681}
2682
fd39ca9a 2683static struct nfs4_stateowner *
f8816512
N		 2684search_close_lru(u32 st_id, int flags)
2685{
1da177e4
LT		 2686 struct nfs4_stateowner *local = NULL;
2687
1da177e4
LT		 2688 if (flags & CLOSE_STATE) {
2689 list_for_each_entry(local, &close_lru, so_close_lru) {
2690 if (local->so_id == st_id)
2691 return local;
2692 }
2693 }
2694 return NULL;
2695}
2696
2697static inline int
2698nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2699{
7eaa36e2 2700 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2701}
2702
2703static int
2704STALE_STATEID(stateid_t *stateid)
2705{
78155ed7
BN		 2706 if (time_after((unsigned long)boot_time,
2707 (unsigned long)stateid->si_boot)) {
2708 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2709 stateid->si_boot, stateid->si_stateownerid,
2710 stateid->si_fileid, stateid->si_generation);
2711 return 1;
2712 }
2713 return 0;
2714}
2715
2716static int
2717EXPIRED_STATEID(stateid_t *stateid)
2718{
2719 if (time_before((unsigned long)boot_time,
2720 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2721 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2722 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2723 stateid->si_boot, stateid->si_stateownerid,
2724 stateid->si_fileid, stateid->si_generation);
2725 return 1;
2726 }
2727 return 0;
2728}
2729
2730static __be32
2731stateid_error_map(stateid_t *stateid)
2732{
2733 if (STALE_STATEID(stateid))
2734 return nfserr_stale_stateid;
2735 if (EXPIRED_STATEID(stateid))
2736 return nfserr_expired;
2737
2738 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2739 stateid->si_boot, stateid->si_stateownerid,
2740 stateid->si_fileid, stateid->si_generation);
2741 return nfserr_bad_stateid;
1da177e4
LT		 2742}
2743
2744static inline int
2745access_permit_read(unsigned long access_bmap)
2746{
2747 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2748 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2749 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2750}
2751
2752static inline int
2753access_permit_write(unsigned long access_bmap)
2754{
2755 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2756 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2757}
2758
2759static
b37ad28b 2760__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2761{
b37ad28b 2762 __be32 status = nfserr_openmode;
1da177e4
LT		 2763
2764 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2765 goto out;
2766 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2767 goto out;
2768 status = nfs_ok;
2769out:
2770 return status;
2771}
2772
b37ad28b 2773static inline __be32
1da177e4
LT		 2774check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2775{
203a8c8e 2776 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2777 return nfs_ok;
af558e33 2778 else if (locks_in_grace()) {
1da177e4
LT		 2779 /* Answer in remaining cases depends on existance of
2780 * conflicting state; so we must wait out the grace period. */
2781 return nfserr_grace;
2782 } else if (flags & WR_STATE)
2783 return nfs4_share_conflict(current_fh,
2784 NFS4_SHARE_DENY_WRITE);
2785 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2786 return nfs4_share_conflict(current_fh,
2787 NFS4_SHARE_DENY_READ);
2788}
2789
2790/*
2791 * Allow READ/WRITE during grace period on recovered state only for files
2792 * that are not able to provide mandatory locking.
2793 */
2794static inline int
18f82731 2795grace_disallows_io(struct inode *inode)
1da177e4 2796{
203a8c8e 2797 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2798}
2799
6668958f 2800static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2801{
6668958f
AA		 2802 /*
2803 * When sessions are used the stateid generation number is ignored
2804 * when it is zero.
2805 */
2806 if ((flags & HAS_SESSION) && in->si_generation == 0)
2807 goto out;
2808
0836f587
BF		 2809 /* If the client sends us a stateid from the future, it's buggy: */
2810 if (in->si_generation > ref->si_generation)
2811 return nfserr_bad_stateid;
2812 /*
2813 * The following, however, can happen. For example, if the
2814 * client sends an open and some IO at the same time, the open
2815 * may bump si_generation while the IO is still in flight.
2816 * Thanks to hard links and renames, the client never knows what
2817 * file an open will affect. So it could avoid that situation
2818 * only by serializing all opens and IO from the same open
2819 * owner. To recover from the old_stateid error, the client
2820 * will just have to retry the IO:
2821 */
2822 if (in->si_generation < ref->si_generation)
2823 return nfserr_old_stateid;
6668958f 2824out:
0836f587
BF		 2825 return nfs_ok;
2826}
2827
3e633079
BF		 2828static int is_delegation_stateid(stateid_t *stateid)
2829{
2830 return stateid->si_fileid == 0;
2831}
2832
1da177e4
LT		 2833/*
2834* Checks for stateid operations
2835*/
b37ad28b 2836__be32
dd453dfd
BH		 2837nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2838 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2839{
2840 struct nfs4_stateid *stp = NULL;
2841 struct nfs4_delegation *dp = NULL;
dd453dfd 2842 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2843 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2844 __be32 status;
1da177e4 2845
1da177e4
LT		 2846 if (filpp)
2847 *filpp = NULL;
2848
18f82731 2849 if (grace_disallows_io(ino))
1da177e4
LT		 2850 return nfserr_grace;
2851
6668958f
AA		 2852 if (nfsd4_has_session(cstate))
2853 flags |= HAS_SESSION;
2854
1da177e4
LT		 2855 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2856 return check_special_stateids(current_fh, stateid, flags);
2857
1da177e4
LT		 2858 status = nfserr_stale_stateid;
2859 if (STALE_STATEID(stateid))
2860 goto out;
2861
1da177e4 2862 status = nfserr_bad_stateid;
3e633079 2863 if (is_delegation_stateid(stateid)) {
a4455be0 2864 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2865 if (!dp) {
2866 status = stateid_error_map(stateid);
1da177e4 2867 goto out;
78155ed7 2868 }
6668958f
AA		 2869 status = check_stateid_generation(stateid, &dp->dl_stateid,
2870 flags);
0c2a498f
BF		 2871 if (status)
2872 goto out;
dc9bf700
BF		 2873 status = nfs4_check_delegmode(dp, flags);
2874 if (status)
2875 goto out;
2876 renew_client(dp->dl_client);
dc9bf700
BF		 2877 if (filpp)
2878 *filpp = dp->dl_vfs_file;
1da177e4 2879 } else { /* open or lock stateid */
a4455be0 2880 stp = find_stateid(stateid, flags);
78155ed7
BN		 2881 if (!stp) {
2882 status = stateid_error_map(stateid);
1da177e4 2883 goto out;
78155ed7 2884 }
6150ef0d 2885 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2886 goto out;
2887 if (!stp->st_stateowner->so_confirmed)
2888 goto out;
6668958f
AA		 2889 status = check_stateid_generation(stateid, &stp->st_stateid,
2890 flags);
0c2a498f
BF		 2891 if (status)
2892 goto out;
a4455be0
BF		 2893 status = nfs4_check_openmode(stp, flags);
2894 if (status)
1da177e4
LT		 2895 goto out;
2896 renew_client(stp->st_stateowner->so_client);
2897 if (filpp)
2898 *filpp = stp->st_vfs_file;
1da177e4
LT		 2899 }
2900 status = nfs_ok;
2901out:
2902 return status;
2903}
2904
4c4cd222
N		 2905static inline int
2906setlkflg (int type)
2907{
2908 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2909 RD_STATE : WR_STATE;
2910}
1da177e4
LT		 2911
2912/*
2913 * Checks for sequence id mutating operations.
2914 */
b37ad28b 2915static __be32
dd453dfd
BH		 2916nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2917 stateid_t *stateid, int flags,
2918 struct nfs4_stateowner **sopp,
2919 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2920{
1da177e4
LT		 2921 struct nfs4_stateid *stp;
2922 struct nfs4_stateowner *sop;
dd453dfd 2923 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2924 __be32 status;
1da177e4
LT		 2925
2926 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2927 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2928 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2929 stateid->si_generation);
3a4f98bb 2930
1da177e4
LT		 2931 *stpp = NULL;
2932 *sopp = NULL;
2933
1da177e4 2934 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2935 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2936 return nfserr_bad_stateid;
1da177e4
LT		 2937 }
2938
1da177e4 2939 if (STALE_STATEID(stateid))
3a4f98bb 2940 return nfserr_stale_stateid;
6668958f
AA		 2941
2942 if (nfsd4_has_session(cstate))
2943 flags |= HAS_SESSION;
2944
1da177e4
LT		 2945 /*
2946 * We return BAD_STATEID if filehandle doesn't match stateid,
2947 * the confirmed flag is incorrecly set, or the generation
2948 * number is incorrect.
1da177e4 2949 */
f8816512
N		 2950 stp = find_stateid(stateid, flags);
2951 if (stp == NULL) {
2952 /*
2953 * Also, we should make sure this isn't just the result of
2954 * a replayed close:
2955 */
2956 sop = search_close_lru(stateid->si_stateownerid, flags);
2957 if (sop == NULL)
78155ed7 2958 return stateid_error_map(stateid);
f8816512
N		 2959 *sopp = sop;
2960 goto check_replay;
2961 }
1da177e4 2962
39325bd0
BF		 2963 *stpp = stp;
2964 *sopp = sop = stp->st_stateowner;
2965
4c4cd222 2966 if (lock) {
4c4cd222 2967 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2968 struct nfs4_client *clp = sop->so_client;
4c4cd222 2969 int lkflg = 0;
b37ad28b 2970 __be32 status;
4c4cd222
N		 2971
2972 lkflg = setlkflg(lock->lk_type);
2973
2974 if (lock->lk_is_new) {
599e0a22
BF		 2975 if (!sop->so_is_open_owner)
2976 return nfserr_bad_stateid;
60adfc50
AA		 2977 if (!(flags & HAS_SESSION) &&
2978 !same_clid(&clp->cl_clientid, lockclid))
2979 return nfserr_bad_stateid;
599e0a22
BF		 2980 /* stp is the open stateid */
2981 status = nfs4_check_openmode(stp, lkflg);
2982 if (status)
2983 return status;
2984 } else {
2985 /* stp is the lock stateid */
2986 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2987 if (status)
2988 return status;
4c4cd222 2989 }
1da177e4
LT		 2990 }
2991
f3362737 2992 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2993 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2994 return nfserr_bad_stateid;
1da177e4
LT		 2995 }
2996
1da177e4
LT		 2997 /*
2998 * We now validate the seqid and stateid generation numbers.
2999 * For the moment, we ignore the possibility of
3000 * generation number wraparound.
3001 */
6668958f 3002 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3003 goto check_replay;
3004
3a4f98bb 3005 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3006 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3007 " unconfirmed stateowner!\n");
3008 return nfserr_bad_stateid;
1da177e4 3009 }
3a4f98bb 3010 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3011 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3012 " confirmed yet!\n");
3013 return nfserr_bad_stateid;
1da177e4 3014 }
6668958f 3015 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3016 if (status)
3017 return status;
52fd004e 3018 renew_client(sop->so_client);
3a4f98bb 3019 return nfs_ok;
1da177e4 3020
1da177e4 3021check_replay:
bd9aac52 3022 if (seqid == sop->so_seqid - 1) {
849823c5 3023 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3024 /* indicate replay to calling function */
a90b061c 3025 return nfserr_replay_me;
1da177e4 3026 }
2fdada03 3027 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3028 sop->so_seqid, seqid);
3029 *sopp = NULL;
3030 return nfserr_bad_seqid;
1da177e4
LT		 3031}
3032
b37ad28b 3033__be32
ca364317 3034nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3035 struct nfsd4_open_confirm *oc)
1da177e4 3036{
b37ad28b 3037 __be32 status;
1da177e4
LT		 3038 struct nfs4_stateowner *sop;
3039 struct nfs4_stateid *stp;
3040
3041 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3042 (int)cstate->current_fh.fh_dentry->d_name.len,
3043 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3044
ca364317 3045 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3046 if (status)
3047 return status;
1da177e4
LT		 3048
3049 nfs4_lock_state();
3050
dd453dfd 3051 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3052 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3053 CONFIRM | OPEN_STATE,
1da177e4
LT		 3054 &oc->oc_stateowner, &stp, NULL)))
3055 goto out;
3056
3057 sop = oc->oc_stateowner;
3058 sop->so_confirmed = 1;
3059 update_stateid(&stp->st_stateid);
3060 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3061 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3062 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3063 stp->st_stateid.si_boot,
3064 stp->st_stateid.si_stateownerid,
3065 stp->st_stateid.si_fileid,
3066 stp->st_stateid.si_generation);
c7b9a459
N		 3067
3068 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3069out:
f2327d9a 3070 if (oc->oc_stateowner) {
1da177e4 3071 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3072 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3073 }
1da177e4
LT		 3074 nfs4_unlock_state();
3075 return status;
3076}
3077
3078
3079/*
3080 * unset all bits in union bitmap (bmap) that
3081 * do not exist in share (from successful OPEN_DOWNGRADE)
3082 */
3083static void
3084reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3085{
3086 int i;
3087 for (i = 1; i < 4; i++) {
3088 if ((i & access) != i)
3089 __clear_bit(i, bmap);
3090 }
3091}
3092
3093static void
3094reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3095{
3096 int i;
3097 for (i = 0; i < 4; i++) {
3098 if ((i & deny) != i)
3099 __clear_bit(i, bmap);
3100 }
3101}
3102
b37ad28b 3103__be32
ca364317
BF		 3104nfsd4_open_downgrade(struct svc_rqst *rqstp,
3105 struct nfsd4_compound_state *cstate,
a4f1706a 3106 struct nfsd4_open_downgrade *od)
1da177e4 3107{
b37ad28b 3108 __be32 status;
1da177e4
LT		 3109 struct nfs4_stateid *stp;
3110 unsigned int share_access;
3111
3112 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3113 (int)cstate->current_fh.fh_dentry->d_name.len,
3114 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3115
d87a8ade 3116 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3117 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3118 return nfserr_inval;
3119
3120 nfs4_lock_state();
dd453dfd 3121 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3122 od->od_seqid,
1da177e4 3123 &od->od_stateid,
f3362737 3124 OPEN_STATE,
1da177e4
LT		 3125 &od->od_stateowner, &stp, NULL)))
3126 goto out;
3127
3128 status = nfserr_inval;
3129 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3130 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3131 stp->st_access_bmap, od->od_share_access);
3132 goto out;
3133 }
3134 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3135 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3136 stp->st_deny_bmap, od->od_share_deny);
3137 goto out;
3138 }
3139 set_access(&share_access, stp->st_access_bmap);
3140 nfs4_file_downgrade(stp->st_vfs_file,
3141 share_access & ~od->od_share_access);
3142
3143 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3144 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3145
3146 update_stateid(&stp->st_stateid);
3147 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3148 status = nfs_ok;
3149out:
f2327d9a 3150 if (od->od_stateowner) {
1da177e4 3151 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3152 cstate->replay_owner = od->od_stateowner;
f2327d9a 3153 }
1da177e4
LT		 3154 nfs4_unlock_state();
3155 return status;
3156}
3157
3158/*
3159 * nfs4_unlock_state() called after encode
3160 */
b37ad28b 3161__be32
ca364317 3162nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3163 struct nfsd4_close *close)
1da177e4 3164{
b37ad28b 3165 __be32 status;
1da177e4
LT		 3166 struct nfs4_stateid *stp;
3167
3168 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3169 (int)cstate->current_fh.fh_dentry->d_name.len,
3170 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3171
3172 nfs4_lock_state();
3173 /* check close_lru for replay */
dd453dfd 3174 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3175 close->cl_seqid,
1da177e4 3176 &close->cl_stateid,
f3362737 3177 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3178 &close->cl_stateowner, &stp, NULL)))
3179 goto out;
1da177e4
LT		 3180 status = nfs_ok;
3181 update_stateid(&stp->st_stateid);
3182 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3183
04ef5954 3184 /* release_stateid() calls nfsd_close() if needed */
2283963f 3185 release_open_stateid(stp);
04ef5954
BF		 3186
3187 /* place unused nfs4_stateowners on so_close_lru list to be
3188 * released by the laundromat service after the lease period
3189 * to enable us to handle CLOSE replay
3190 */
3191 if (list_empty(&close->cl_stateowner->so_stateids))
3192 move_to_close_lru(close->cl_stateowner);
1da177e4 3193out:
f2327d9a 3194 if (close->cl_stateowner) {
1da177e4 3195 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3196 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3197 }
1da177e4
LT		 3198 nfs4_unlock_state();
3199 return status;
3200}
3201
b37ad28b 3202__be32
ca364317
BF		 3203nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3204 struct nfsd4_delegreturn *dr)
1da177e4 3205{
203a8c8e
BF		 3206 struct nfs4_delegation *dp;
3207 stateid_t *stateid = &dr->dr_stateid;
3208 struct inode *inode;
b37ad28b 3209 __be32 status;
6668958f 3210 int flags = 0;
1da177e4 3211
ca364317 3212 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3213 return status;
3214 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3215
6668958f
AA		 3216 if (nfsd4_has_session(cstate))
3217 flags |= HAS_SESSION;
1da177e4 3218 nfs4_lock_state();
203a8c8e
BF		 3219 status = nfserr_bad_stateid;
3220 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3221 goto out;
3222 status = nfserr_stale_stateid;
3223 if (STALE_STATEID(stateid))
3224 goto out;
7e0f7cf5 3225 status = nfserr_bad_stateid;
203a8c8e
BF		 3226 if (!is_delegation_stateid(stateid))
3227 goto out;
203a8c8e 3228 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3229 if (!dp) {
3230 status = stateid_error_map(stateid);
203a8c8e 3231 goto out;
78155ed7 3232 }
6668958f 3233 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3234 if (status)
3235 goto out;
3236 renew_client(dp->dl_client);
3237
3238 unhash_delegation(dp);
1da177e4 3239out:
203a8c8e
BF		 3240 nfs4_unlock_state();
3241
1da177e4
LT		 3242 return status;
3243}
3244
3245
3246/*
3247 * Lock owner state (byte-range locks)
3248 */
3249#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3250#define LOCK_HASH_BITS 8
3251#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3252#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3253
87df4de8
BH		 3254static inline u64
3255end_offset(u64 start, u64 len)
3256{
3257 u64 end;
3258
3259 end = start + len;
3260 return end >= start ? end: NFS4_MAX_UINT64;
3261}
3262
3263/* last octet in a range */
3264static inline u64
3265last_byte_offset(u64 start, u64 len)
3266{
3267 u64 end;
3268
3269 BUG_ON(!len);
3270 end = start + len;
3271 return end > start ? end - 1: NFS4_MAX_UINT64;
3272}
3273
1da177e4
LT		 3274#define lockownerid_hashval(id) \
3275 ((id) & LOCK_HASH_MASK)
3276
3277static inline unsigned int
3278lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3279 struct xdr_netobj *ownername)
3280{
3281 return (file_hashval(inode) + cl_id
3282 + opaque_hashval(ownername->data, ownername->len))
3283 & LOCK_HASH_MASK;
3284}
3285
3286static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3287static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3288static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3289
fd39ca9a 3290static struct nfs4_stateid *
1da177e4
LT		 3291find_stateid(stateid_t *stid, int flags)
3292{
9346eff0 3293 struct nfs4_stateid *local;
1da177e4
LT		 3294 u32 st_id = stid->si_stateownerid;
3295 u32 f_id = stid->si_fileid;
3296 unsigned int hashval;
3297
3298 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3299 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3300 hashval = stateid_hashval(st_id, f_id);
3301 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3302 if ((local->st_stateid.si_stateownerid == st_id) &&
3303 (local->st_stateid.si_fileid == f_id))
3304 return local;
3305 }
3306 }
9346eff0
KK		 3307
3308 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3309 hashval = stateid_hashval(st_id, f_id);
3310 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3311 if ((local->st_stateid.si_stateownerid == st_id) &&
3312 (local->st_stateid.si_fileid == f_id))
3313 return local;
3314 }
849823c5 3315 }
1da177e4
LT		 3316 return NULL;
3317}
3318
3319static struct nfs4_delegation *
3320find_delegation_stateid(struct inode *ino, stateid_t *stid)
3321{
13cd2184
N		 3322 struct nfs4_file *fp;
3323 struct nfs4_delegation *dl;
1da177e4
LT		 3324
3325 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3326 stid->si_boot, stid->si_stateownerid,
3327 stid->si_fileid, stid->si_generation);
3328
1da177e4 3329 fp = find_file(ino);
13cd2184
N		 3330 if (!fp)
3331 return NULL;
3332 dl = find_delegation_file(fp, stid);
3333 put_nfs4_file(fp);
3334 return dl;
1da177e4
LT		 3335}
3336
3337/*
3338 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3339 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3340 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3341 * locking, this prevents us from being completely protocol-compliant. The
3342 * real solution to this problem is to start using unsigned file offsets in
3343 * the VFS, but this is a very deep change!
3344 */
3345static inline void
3346nfs4_transform_lock_offset(struct file_lock *lock)
3347{
3348 if (lock->fl_start < 0)
3349 lock->fl_start = OFFSET_MAX;
3350 if (lock->fl_end < 0)
3351 lock->fl_end = OFFSET_MAX;
3352}
3353
d5b9026a
N		 3354/* Hack!: For now, we're defining this just so we can use a pointer to it
3355 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3356static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3357};
1da177e4
LT		 3358
3359static inline void
3360nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3361{
d5b9026a
N		 3362 struct nfs4_stateowner *sop;
3363 unsigned int hval;
1da177e4 3364
d5b9026a
N		 3365 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3366 sop = (struct nfs4_stateowner *) fl->fl_owner;
3367 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3368 kref_get(&sop->so_ref);
3369 deny->ld_sop = sop;
3370 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3371 } else {
3372 deny->ld_sop = NULL;
3373 deny->ld_clientid.cl_boot = 0;
3374 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3375 }
3376 deny->ld_start = fl->fl_start;
87df4de8
BH		 3377 deny->ld_length = NFS4_MAX_UINT64;
3378 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3379 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3380 deny->ld_type = NFS4_READ_LT;
3381 if (fl->fl_type != F_RDLCK)
3382 deny->ld_type = NFS4_WRITE_LT;
3383}
3384
1da177e4
LT		 3385static struct nfs4_stateowner *
3386find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3387 struct xdr_netobj *owner)
3388{
3389 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3390 struct nfs4_stateowner *op;
3391
3392 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3393 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3394 return op;
3395 }
3396 return NULL;
3397}
3398
3399/*
3400 * Alloc a lock owner structure.
3401 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3402 * occured.
3403 *
3404 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3405 */
3406
3407static struct nfs4_stateowner *
3408alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3409 struct nfs4_stateowner *sop;
3410 struct nfs4_replay *rp;
3411 unsigned int idhashval;
3412
3413 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3414 return NULL;
3415 idhashval = lockownerid_hashval(current_ownerid);
3416 INIT_LIST_HEAD(&sop->so_idhash);
3417 INIT_LIST_HEAD(&sop->so_strhash);
3418 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3419 INIT_LIST_HEAD(&sop->so_stateids);
3420 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3421 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3422 sop->so_time = 0;
3423 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3424 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3425 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3426 sop->so_is_open_owner = 0;
3427 sop->so_id = current_ownerid++;
3428 sop->so_client = clp;
b59e3c0e
NB		 3429 /* It is the openowner seqid that will be incremented in encode in the
3430 * case of new lockowners; so increment the lock seqid manually: */
3431 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3432 sop->so_confirmed = 1;
3433 rp = &sop->so_replay;
de1ae286 3434 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3435 rp->rp_buflen = 0;
3436 rp->rp_buf = rp->rp_ibuf;
3437 return sop;
3438}
3439
fd39ca9a 3440static struct nfs4_stateid *
1da177e4
LT		 3441alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3442{
3443 struct nfs4_stateid *stp;
3444 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3445
5ac049ac
N		 3446 stp = nfs4_alloc_stateid();
3447 if (stp == NULL)
1da177e4
LT		 3448 goto out;
3449 INIT_LIST_HEAD(&stp->st_hash);
3450 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3451 INIT_LIST_HEAD(&stp->st_perstateowner);
3452 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3453 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3454 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3455 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3456 stp->st_stateowner = sop;
13cd2184 3457 get_nfs4_file(fp);
1da177e4 3458 stp->st_file = fp;
78155ed7 3459 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3460 stp->st_stateid.si_stateownerid = sop->so_id;
3461 stp->st_stateid.si_fileid = fp->fi_id;
3462 stp->st_stateid.si_generation = 0;
3463 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3464 stp->st_access_bmap = open_stp->st_access_bmap;
3465 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3466 stp->st_openstp = open_stp;
1da177e4
LT		 3467
3468out:
3469 return stp;
3470}
3471
fd39ca9a 3472static int
1da177e4
LT		 3473check_lock_length(u64 offset, u64 length)
3474{
87df4de8 3475 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3476 LOFF_OVERFLOW(offset, length)));
3477}
3478
3479/*
3480 * LOCK operation
3481 */
b37ad28b 3482__be32
ca364317 3483nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3484 struct nfsd4_lock *lock)
1da177e4 3485{
3e9e3dbe 3486 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3487 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3488 struct nfs4_stateid *lock_stp;
3489 struct file *filp;
3490 struct file_lock file_lock;
8dc7c311 3491 struct file_lock conflock;
b37ad28b 3492 __be32 status = 0;
1da177e4 3493 unsigned int strhashval;
150b3934 3494 unsigned int cmd;
b8dd7b9a 3495 int err;
1da177e4
LT		 3496
3497 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3498 (long long) lock->lk_offset,
3499 (long long) lock->lk_length);
3500
1da177e4
LT		 3501 if (check_lock_length(lock->lk_offset, lock->lk_length))
3502 return nfserr_inval;
3503
ca364317 3504 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3505 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3506 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3507 return status;
3508 }
3509
1da177e4
LT		 3510 nfs4_lock_state();
3511
3512 if (lock->lk_is_new) {
893f8770
N		 3513 /*
3514 * Client indicates that this is a new lockowner.
3515 * Use open owner and open stateid to create lock owner and
3516 * lock stateid.
3517 */
1da177e4
LT		 3518 struct nfs4_stateid *open_stp = NULL;
3519 struct nfs4_file *fp;
3520
3521 status = nfserr_stale_clientid;
60adfc50
AA		 3522 if (!nfsd4_has_session(cstate) &&
3523 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3524 goto out;
1da177e4 3525
1da177e4 3526 /* validate and update open stateid and open seqid */
dd453dfd 3527 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3528 lock->lk_new_open_seqid,
3529 &lock->lk_new_open_stateid,
f3362737 3530 OPEN_STATE,
3a65588a 3531 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3532 lock);
37515177 3533 if (status)
1da177e4 3534 goto out;
3a65588a 3535 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3536 /* create lockowner and lock stateid */
3537 fp = open_stp->st_file;
3538 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3539 open_sop->so_client->cl_clientid.cl_id,
3540 &lock->v.new.owner);
3e9e3dbe
N		 3541 /* XXX: Do we need to check for duplicate stateowners on
3542 * the same file, or should they just be allowed (and
3543 * create new stateids)? */
1da177e4 3544 status = nfserr_resource;
b59e3c0e
NB		 3545 lock_sop = alloc_init_lock_stateowner(strhashval,
3546 open_sop->so_client, open_stp, lock);
3547 if (lock_sop == NULL)
1da177e4 3548 goto out;
b59e3c0e 3549 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3550 if (lock_stp == NULL)
1da177e4 3551 goto out;
1da177e4
LT		 3552 } else {
3553 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3554 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3555 lock->lk_old_lock_seqid,
3556 &lock->lk_old_lock_stateid,
f3362737 3557 LOCK_STATE,
3a65588a 3558 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3559 if (status)
3560 goto out;
3a65588a 3561 lock_sop = lock->lk_replay_owner;
1da177e4 3562 }
3a65588a 3563 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3564 filp = lock_stp->st_vfs_file;
3565
0dd395dc 3566 status = nfserr_grace;
af558e33 3567 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3568 goto out;
3569 status = nfserr_no_grace;
af558e33 3570 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3571 goto out;
3572
1da177e4
LT		 3573 locks_init_lock(&file_lock);
3574 switch (lock->lk_type) {
3575 case NFS4_READ_LT:
3576 case NFS4_READW_LT:
3577 file_lock.fl_type = F_RDLCK;
150b3934 3578 cmd = F_SETLK;
1da177e4
LT		 3579 break;
3580 case NFS4_WRITE_LT:
3581 case NFS4_WRITEW_LT:
3582 file_lock.fl_type = F_WRLCK;
150b3934 3583 cmd = F_SETLK;
1da177e4
LT		 3584 break;
3585 default:
3586 status = nfserr_inval;
3587 goto out;
3588 }
b59e3c0e 3589 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3590 file_lock.fl_pid = current->tgid;
3591 file_lock.fl_file = filp;
3592 file_lock.fl_flags = FL_POSIX;
d5b9026a 3593 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3594
3595 file_lock.fl_start = lock->lk_offset;
87df4de8 3596 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3597 nfs4_transform_lock_offset(&file_lock);
3598
3599 /*
3600 * Try to lock the file in the VFS.
3601 * Note: locks.c uses the BKL to protect the inode's lock list.
3602 */
3603
fd85b817 3604 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3605 switch (-err) {
1da177e4
LT		 3606 case 0: /* success! */
3607 update_stateid(&lock_stp->st_stateid);
3608 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3609 sizeof(stateid_t));
b8dd7b9a 3610 status = 0;
eb76b3fd
AA		 3611 break;
3612 case (EAGAIN): /* conflock holds conflicting lock */
3613 status = nfserr_denied;
3614 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3615 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3616 break;
1da177e4
LT		 3617 case (EDEADLK):
3618 status = nfserr_deadlock;
eb76b3fd 3619 break;
1da177e4 3620 default:
fd85b817 3621 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3622 status = nfserr_resource;
3623 break;
1da177e4 3624 }
1da177e4 3625out:
8a280510 3626 if (status && lock->lk_is_new && lock_sop)
f044ff83 3627 release_lockowner(lock_sop);
3a65588a
BF		 3628 if (lock->lk_replay_owner) {
3629 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3630 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3631 }
1da177e4
LT		 3632 nfs4_unlock_state();
3633 return status;
3634}
3635
55ef1274
BF		 3636/*
3637 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3638 * so we do a temporary open here just to get an open file to pass to
3639 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3640 * inode operation.)
3641 */
3642static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3643{
3644 struct file *file;
3645 int err;
3646
3647 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3648 if (err)
3649 return err;
3650 err = vfs_test_lock(file, lock);
3651 nfsd_close(file);
3652 return err;
3653}
3654
1da177e4
LT		 3655/*
3656 * LOCKT operation
3657 */
b37ad28b 3658__be32
ca364317
BF		 3659nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3660 struct nfsd4_lockt *lockt)
1da177e4
LT		 3661{
3662 struct inode *inode;
1da177e4 3663 struct file_lock file_lock;
fd85b817 3664 int error;
b37ad28b 3665 __be32 status;
1da177e4 3666
af558e33 3667 if (locks_in_grace())
1da177e4
LT		 3668 return nfserr_grace;
3669
3670 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3671 return nfserr_inval;
3672
3673 lockt->lt_stateowner = NULL;
3674 nfs4_lock_state();
3675
3676 status = nfserr_stale_clientid;
60adfc50 3677 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3678 goto out;
1da177e4 3679
ca364317 3680 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3681 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3682 if (status == nfserr_symlink)
3683 status = nfserr_inval;
3684 goto out;
3685 }
3686
ca364317 3687 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3688 locks_init_lock(&file_lock);
3689 switch (lockt->lt_type) {
3690 case NFS4_READ_LT:
3691 case NFS4_READW_LT:
3692 file_lock.fl_type = F_RDLCK;
3693 break;
3694 case NFS4_WRITE_LT:
3695 case NFS4_WRITEW_LT:
3696 file_lock.fl_type = F_WRLCK;
3697 break;
3698 default:
2fdada03 3699 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3700 status = nfserr_inval;
3701 goto out;
3702 }
3703
3704 lockt->lt_stateowner = find_lockstateowner_str(inode,
3705 &lockt->lt_clientid, &lockt->lt_owner);
3706 if (lockt->lt_stateowner)
3707 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3708 file_lock.fl_pid = current->tgid;
3709 file_lock.fl_flags = FL_POSIX;
3710
3711 file_lock.fl_start = lockt->lt_offset;
87df4de8 3712 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3713
3714 nfs4_transform_lock_offset(&file_lock);
3715
1da177e4 3716 status = nfs_ok;
55ef1274 3717 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3718 if (error) {
3719 status = nfserrno(error);
3720 goto out;
3721 }
9d6a8c5c 3722 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3723 status = nfserr_denied;
9d6a8c5c 3724 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3725 }
3726out:
3727 nfs4_unlock_state();
3728 return status;
3729}
3730
b37ad28b 3731__be32
ca364317 3732nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3733 struct nfsd4_locku *locku)
1da177e4
LT		 3734{
3735 struct nfs4_stateid *stp;
3736 struct file *filp = NULL;
3737 struct file_lock file_lock;
b37ad28b 3738 __be32 status;
b8dd7b9a 3739 int err;
1da177e4
LT		 3740
3741 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3742 (long long) locku->lu_offset,
3743 (long long) locku->lu_length);
3744
3745 if (check_lock_length(locku->lu_offset, locku->lu_length))
3746 return nfserr_inval;
3747
3748 nfs4_lock_state();
3749
dd453dfd 3750 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3751 locku->lu_seqid,
3752 &locku->lu_stateid,
f3362737 3753 LOCK_STATE,
1da177e4
LT		 3754 &locku->lu_stateowner, &stp, NULL)))
3755 goto out;
3756
3757 filp = stp->st_vfs_file;
3758 BUG_ON(!filp);
3759 locks_init_lock(&file_lock);
3760 file_lock.fl_type = F_UNLCK;
3761 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3762 file_lock.fl_pid = current->tgid;
3763 file_lock.fl_file = filp;
3764 file_lock.fl_flags = FL_POSIX;
d5b9026a 3765 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3766 file_lock.fl_start = locku->lu_offset;
3767
87df4de8 3768 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3769 nfs4_transform_lock_offset(&file_lock);
3770
3771 /*
3772 * Try to unlock the file in the VFS.
3773 */
fd85b817 3774 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3775 if (err) {
fd85b817 3776 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3777 goto out_nfserr;
3778 }
3779 /*
3780 * OK, unlock succeeded; the only thing left to do is update the stateid.
3781 */
3782 update_stateid(&stp->st_stateid);
3783 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3784
3785out:
f2327d9a 3786 if (locku->lu_stateowner) {
1da177e4 3787 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3788 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3789 }
1da177e4
LT		 3790 nfs4_unlock_state();
3791 return status;
3792
3793out_nfserr:
b8dd7b9a 3794 status = nfserrno(err);
1da177e4
LT		 3795 goto out;
3796}
3797
3798/*
3799 * returns
3800 * 1: locks held by lockowner
3801 * 0: no locks held by lockowner
3802 */
3803static int
3804check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3805{
3806 struct file_lock **flpp;
7eaa36e2 3807 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3808 int status = 0;
3809
3810 lock_kernel();
3811 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3812 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3813 status = 1;
3814 goto out;
796dadfd 3815 }
1da177e4
LT		 3816 }
3817out:
3818 unlock_kernel();
3819 return status;
3820}
3821
b37ad28b 3822__be32
b591480b
BF		 3823nfsd4_release_lockowner(struct svc_rqst *rqstp,
3824 struct nfsd4_compound_state *cstate,
3825 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3826{
3827 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3828 struct nfs4_stateowner *sop;
3829 struct nfs4_stateid *stp;
1da177e4 3830 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3831 struct list_head matches;
3832 int i;
b37ad28b 3833 __be32 status;
1da177e4
LT		 3834
3835 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3836 clid->cl_boot, clid->cl_id);
3837
3838 /* XXX check for lease expiration */
3839
3840 status = nfserr_stale_clientid;
849823c5 3841 if (STALE_CLIENTID(clid))
1da177e4 3842 return status;
1da177e4
LT		 3843
3844 nfs4_lock_state();
3845
3e9e3dbe
N		 3846 status = nfserr_locks_held;
3847 /* XXX: we're doing a linear search through all the lockowners.
3848 * Yipes! For now we'll just hope clients aren't really using
3849 * release_lockowner much, but eventually we have to fix these
3850 * data structures. */
3851 INIT_LIST_HEAD(&matches);
3852 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3853 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3854 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3855 continue;
3856 list_for_each_entry(stp, &sop->so_stateids,
3857 st_perstateowner) {
3858 if (check_for_locks(stp->st_vfs_file, sop))
3859 goto out;
3860 /* Note: so_perclient unused for lockowners,
3861 * so it's OK to fool with here. */
3862 list_add(&sop->so_perclient, &matches);
3863 }
1da177e4 3864 }
3e9e3dbe
N		 3865 }
3866 /* Clients probably won't expect us to return with some (but not all)
3867 * of the lockowner state released; so don't release any until all
3868 * have been checked. */
3869 status = nfs_ok;
0fa822e4
N		 3870 while (!list_empty(&matches)) {
3871 sop = list_entry(matches.next, struct nfs4_stateowner,
3872 so_perclient);
3873 /* unhash_stateowner deletes so_perclient only
3874 * for openowners. */
3875 list_del(&sop->so_perclient);
f044ff83 3876 release_lockowner(sop);
1da177e4
LT		 3877 }
3878out:
3879 nfs4_unlock_state();
3880 return status;
3881}
3882
3883static inline struct nfs4_client_reclaim *
a55370a3 3884alloc_reclaim(void)
1da177e4 3885{
a55370a3 3886 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3887}
3888
c7b9a459 3889int
a1bcecd2 3890nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3891{
3892 unsigned int strhashval = clientstr_hashval(name);
3893 struct nfs4_client *clp;
3894
a1bcecd2 3895 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3896 return clp ? 1 : 0;
3897}
3898
1da177e4
LT		 3899/*
3900 * failure => all reset bets are off, nfserr_no_grace...
3901 */
190e4fbf
N		 3902int
3903nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3904{
3905 unsigned int strhashval;
3906 struct nfs4_client_reclaim *crp = NULL;
3907
a55370a3
N		 3908 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3909 crp = alloc_reclaim();
1da177e4
LT		 3910 if (!crp)
3911 return 0;
a55370a3 3912 strhashval = clientstr_hashval(name);
1da177e4
LT		 3913 INIT_LIST_HEAD(&crp->cr_strhash);
3914 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3915 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3916 reclaim_str_hashtbl_size++;
3917 return 1;
3918}
3919
3920static void
3921nfs4_release_reclaim(void)
3922{
3923 struct nfs4_client_reclaim *crp = NULL;
3924 int i;
3925
1da177e4
LT		 3926 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3927 while (!list_empty(&reclaim_str_hashtbl[i])) {
3928 crp = list_entry(reclaim_str_hashtbl[i].next,
3929 struct nfs4_client_reclaim, cr_strhash);
3930 list_del(&crp->cr_strhash);
1da177e4
LT		 3931 kfree(crp);
3932 reclaim_str_hashtbl_size--;
3933 }
3934 }
3935 BUG_ON(reclaim_str_hashtbl_size);
3936}
3937
3938/*
3939 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3940static struct nfs4_client_reclaim *
1da177e4
LT		 3941nfs4_find_reclaim_client(clientid_t *clid)
3942{
3943 unsigned int strhashval;
3944 struct nfs4_client *clp;
3945 struct nfs4_client_reclaim *crp = NULL;
3946
3947
3948 /* find clientid in conf_id_hashtbl */
3949 clp = find_confirmed_client(clid);
3950 if (clp == NULL)
3951 return NULL;
3952
a55370a3
N		 3953 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3954 clp->cl_name.len, clp->cl_name.data,
3955 clp->cl_recdir);
1da177e4
LT		 3956
3957 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3958 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3959 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3960 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3961 return crp;
3962 }
3963 }
3964 return NULL;
3965}
3966
3967/*
3968* Called from OPEN. Look for clientid in reclaim list.
3969*/
b37ad28b 3970__be32
1da177e4
LT		 3971nfs4_check_open_reclaim(clientid_t *clid)
3972{
dfc83565 3973 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3974}
3975
ac4d8ff2 3976/* initialization to perform at module load time: */
1da177e4 3977
e8ff2a84 3978int
ac4d8ff2 3979nfs4_state_init(void)
1da177e4 3980{
e8ff2a84 3981 int i, status;
1da177e4 3982
e8ff2a84
BF		 3983 status = nfsd4_init_slabs();
3984 if (status)
3985 return status;
1da177e4
LT		 3986 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3987 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3988 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3989 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3990 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3991 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3992 }
5282fd72
ME		 3993 for (i = 0; i < SESSION_HASH_SIZE; i++)
3994 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3995 for (i = 0; i < FILE_HASH_SIZE; i++) {
3996 INIT_LIST_HEAD(&file_hashtbl[i]);
3997 }
3998 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3999 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4000 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4001 }
4002 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4003 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4004 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4005 }
4006 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4007 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4008 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4009 }
1da177e4 4010 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4011 INIT_LIST_HEAD(&close_lru);
4012 INIT_LIST_HEAD(&client_lru);
4013 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4014 reclaim_str_hashtbl_size = 0;
e8ff2a84 4015 return 0;
ac4d8ff2
N		 4016}
4017
190e4fbf
N		 4018static void
4019nfsd4_load_reboot_recovery_data(void)
4020{
4021 int status;
4022
0964a3d3
N		 4023 nfs4_lock_state();
4024 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4025 status = nfsd4_recdir_load();
0964a3d3 4026 nfs4_unlock_state();
190e4fbf
N		 4027 if (status)
4028 printk("NFSD: Failure reading reboot recovery data\n");
4029}
4030
9a8db97e
ME		 4031unsigned long
4032get_nfs4_grace_period(void)
4033{
4034 return max(user_lease_time, lease_time) * HZ;
4035}
4036
c2f1a551
MS		 4037/*
4038 * Since the lifetime of a delegation isn't limited to that of an open, a
4039 * client may quite reasonably hang on to a delegation as long as it has
4040 * the inode cached. This becomes an obvious problem the first time a
4041 * client's inode cache approaches the size of the server's total memory.
4042 *
4043 * For now we avoid this problem by imposing a hard limit on the number
4044 * of delegations, which varies according to the server's memory size.
4045 */
4046static void
4047set_max_delegations(void)
4048{
4049 /*
4050 * Allow at most 4 delegations per megabyte of RAM. Quick
4051 * estimates suggest that in the worst case (where every delegation
4052 * is for a different inode), a delegation could take about 1.5K,
4053 * giving a worst case usage of about 6% of memory.
4054 */
4055 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4056}
4057
ac4d8ff2
N		 4058/* initialization to perform when the nfsd service is started: */
4059
4060static void
4061__nfs4_state_start(void)
4062{
9a8db97e 4063 unsigned long grace_time;
ac4d8ff2 4064
1da177e4 4065 boot_time = get_seconds();
af558e33 4066 grace_time = get_nfs4_grace_period();
d99a05ad 4067 lease_time = user_lease_time;
af558e33 4068 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4069 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4070 grace_time/HZ);
58da282b 4071 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4072 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4073 set_max_delegations();
1da177e4
LT		 4074}
4075
e8ff2a84 4076void
76a3550e 4077nfs4_state_start(void)
1da177e4 4078{
1da177e4 4079 if (nfs4_init)
e8ff2a84 4080 return;
190e4fbf 4081 nfsd4_load_reboot_recovery_data();
76a3550e 4082 __nfs4_state_start();
1da177e4 4083 nfs4_init = 1;
e8ff2a84 4084 return;
1da177e4
LT		 4085}
4086
1da177e4
LT		 4087time_t
4088nfs4_lease_time(void)
4089{
4090 return lease_time;
4091}
4092
4093static void
4094__nfs4_state_shutdown(void)
4095{
4096 int i;
4097 struct nfs4_client *clp = NULL;
4098 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4099 struct list_head *pos, *next, reaplist;
4100
1da177e4
LT		 4101 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4102 while (!list_empty(&conf_id_hashtbl[i])) {
4103 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4104 expire_client(clp);
4105 }
4106 while (!list_empty(&unconf_str_hashtbl[i])) {
4107 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4108 expire_client(clp);
4109 }
4110 }
4111 INIT_LIST_HEAD(&reaplist);
4112 spin_lock(&recall_lock);
4113 list_for_each_safe(pos, next, &del_recall_lru) {
4114 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4115 list_move(&dp->dl_recall_lru, &reaplist);
4116 }
4117 spin_unlock(&recall_lock);
4118 list_for_each_safe(pos, next, &reaplist) {
4119 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4120 list_del_init(&dp->dl_recall_lru);
4121 unhash_delegation(dp);
4122 }
4123
190e4fbf 4124 nfsd4_shutdown_recdir();
1da177e4 4125 nfs4_init = 0;
1da177e4
LT		 4126}
4127
4128void
4129nfs4_state_shutdown(void)
4130{
5e8d5c29
N		 4131 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4132 destroy_workqueue(laundry_wq);
2c5e7615 4133 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4134 nfs4_lock_state();
4135 nfs4_release_reclaim();
4136 __nfs4_state_shutdown();
1da177e4
LT		 4137 nfs4_unlock_state();
4138}
4139
3dd98a3b
JL		 4140/*
4141 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4142 * accessed when nfsd is starting.
4143 */
0964a3d3
N		 4144static void
4145nfs4_set_recdir(char *recdir)
4146{
0964a3d3 4147 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4148}
4149
4150/*
4151 * Change the NFSv4 recovery directory to recdir.
4152 */
4153int
4154nfs4_reset_recoverydir(char *recdir)
4155{
4156 int status;
a63bb996 4157 struct path path;
0964a3d3 4158
a63bb996 4159 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4160 if (status)
4161 return status;
4162 status = -ENOTDIR;
a63bb996 4163 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4164 nfs4_set_recdir(recdir);
4165 status = 0;
4166 }
a63bb996 4167 path_put(&path);
0964a3d3
N		 4168 return status;
4169}
4170
3dd98a3b
JL		 4171char *
4172nfs4_recoverydir(void)
4173{
4174 return user_recovery_dirname;
4175}
4176
1da177e4
LT		 4177/*
4178 * Called when leasetime is changed.
4179 *
d99a05ad
N		 4180 * The only way the protocol gives us to handle on-the-fly lease changes is to
4181 * simulate a reboot. Instead of doing that, we just wait till the next time
4182 * we start to register any changes in lease time. If the administrator
4183 * really wants to change the lease time *now*, they can go ahead and bring
4184 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4185 *
4186 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4187 * when nfsd is starting
1da177e4
LT		 4188 */
4189void
4190nfs4_reset_lease(time_t leasetime)
4191{
d99a05ad 4192 user_lease_time = leasetime;
1da177e4 4193}
Linux 6.x block layer and io_uring tree(s)