projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
sunrpc: reply AUTH_BADCRED to RPCSEC_GSS with unknown service
[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
fbf4665f 901gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 902{
c237dc03 903 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 904 unsigned short expected_family;
905
906 /* Currently, we only support tcp and tcp6 for the callback channel */
907 if (se->se_callback_netid_len == 3 &&
908 !memcmp(se->se_callback_netid_val, "tcp", 3))
909 expected_family = AF_INET;
910 else if (se->se_callback_netid_len == 4 &&
911 !memcmp(se->se_callback_netid_val, "tcp6", 4))
912 expected_family = AF_INET6;
913 else
1da177e4
LT		 914 goto out_err;
915
aa9a4ec7
JL		 916 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
917 se->se_callback_addr_len,
918 (struct sockaddr *) &cb->cb_addr,
919 sizeof(cb->cb_addr));
920
7077ecba 921 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 922 goto out_err;
aa9a4ec7 923
fbf4665f
JL		 924 if (cb->cb_addr.ss_family == AF_INET6)
925 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
926
ab52ae6d 927 cb->cb_minorversion = 0;
1da177e4
LT		 928 cb->cb_prog = se->se_callback_prog;
929 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 930 return;
931out_err:
aa9a4ec7
JL		 932 cb->cb_addr.ss_family = AF_UNSPEC;
933 cb->cb_addrlen = 0;
849823c5 934 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 935 "will not receive delegations\n",
936 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
937
1da177e4
LT		 938 return;
939}
940
074fe897
AA		 941void
942nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
943{
944 struct nfsd4_compoundres *resp = rqstp->rq_resp;
945
946 resp->cstate.statp = statp;
947}
948
949/*
950 * Dereference the result pages.
951 */
952static void
953nfsd4_release_respages(struct page **respages, short resused)
954{
955 int i;
956
957 dprintk("--> %s\n", __func__);
958 for (i = 0; i < resused; i++) {
959 if (!respages[i])
960 continue;
961 put_page(respages[i]);
962 respages[i] = NULL;
963 }
964}
965
966static void
967nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
968{
969 int i;
970
971 for (i = 0; i < count; i++) {
972 topages[i] = frompages[i];
973 if (!topages[i])
974 continue;
975 get_page(topages[i]);
976 }
977}
978
979/*
980 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
981 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
982 * length of the XDR response is less than se_fmaxresp_cached
983 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
984 * of the reply (e.g. readdir).
985 *
986 * Store the base and length of the rq_req.head[0] page
987 * of the NFSv4.1 data, just past the rpc header.
988 */
989void
990nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
991{
992 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
993 struct svc_rqst *rqstp = resp->rqstp;
994 struct nfsd4_compoundargs *args = rqstp->rq_argp;
995 struct nfsd4_op *op = &args->ops[resp->opcnt];
996 struct kvec *resv = &rqstp->rq_res.head[0];
997
998 dprintk("--> %s entry %p\n", __func__, entry);
999
1000 /* Don't cache a failed OP_SEQUENCE. */
1001 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
1002 return;
bf864a31 1003
074fe897 1004 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1005 entry->ce_opcnt = resp->opcnt;
1006 entry->ce_status = resp->cstate.status;
1007
1008 /*
1009 * Don't need a page to cache just the sequence operation - the slot
1010 * does this for us!
1011 */
1012
1013 if (nfsd4_not_cached(resp)) {
1014 entry->ce_resused = 0;
1015 entry->ce_rpchdrlen = 0;
1016 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1017 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1018 return;
1019 }
074fe897
AA		 1020 entry->ce_resused = rqstp->rq_resused;
1021 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1022 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1023 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1024 entry->ce_resused);
074fe897
AA		 1025 entry->ce_datav.iov_base = resp->cstate.statp;
1026 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1027 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1028 /* Current request rpc header length*/
1029 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1030 (char *)page_address(rqstp->rq_respages[0]);
1031}
1032
1033/*
1034 * We keep the rpc header, but take the nfs reply from the replycache.
1035 */
1036static int
1037nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1038 struct nfsd4_cache_entry *entry)
1039{
1040 struct svc_rqst *rqstp = resp->rqstp;
1041 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1042 int len;
1043
1044 /* Current request rpc header length*/
1045 len = (char *)resp->cstate.statp -
1046 (char *)page_address(rqstp->rq_respages[0]);
1047 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1048 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1049 entry->ce_datav.iov_len);
1050 return 0;
1051 }
1052 /* copy the cached reply nfsd data past the current rpc header */
1053 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1054 entry->ce_datav.iov_len);
1055 resv->iov_len = len + entry->ce_datav.iov_len;
1056 return 1;
1057}
1058
abfabf8c
AA		 1059/*
1060 * Encode the replay sequence operation from the slot values.
1061 * If cachethis is FALSE encode the uncached rep error on the next
1062 * operation which sets resp->p and increments resp->opcnt for
1063 * nfs4svc_encode_compoundres.
1064 *
1065 */
1066static __be32
1067nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1068 struct nfsd4_compoundres *resp)
1069{
1070 struct nfsd4_op *op;
1071 struct nfsd4_slot *slot = resp->cstate.slot;
1072
1073 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1074 resp->opcnt, resp->cstate.slot->sl_cache_entry.ce_cachethis);
1075
1076 /* Encode the replayed sequence operation */
1077 op = &args->ops[resp->opcnt - 1];
1078 nfsd4_encode_operation(resp, op);
1079
1080 /* Return nfserr_retry_uncached_rep in next operation. */
1081 if (args->opcnt > 1 && slot->sl_cache_entry.ce_cachethis == 0) {
1082 op = &args->ops[resp->opcnt++];
1083 op->status = nfserr_retry_uncached_rep;
1084 nfsd4_encode_operation(resp, op);
1085 }
1086 return op->status;
1087}
1088
074fe897
AA		 1089/*
1090 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1091 * cached page. Replace any futher replay pages from the cache.
1092 */
1093__be32
bf864a31
AA		 1094nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1095 struct nfsd4_sequence *seq)
074fe897
AA		 1096{
1097 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1098 __be32 status;
1099
1100 dprintk("--> %s entry %p\n", __func__, entry);
1101
bf864a31
AA		 1102 /*
1103 * If this is just the sequence operation, we did not keep
1104 * a page in the cache entry because we can just use the
1105 * slot info stored in struct nfsd4_sequence that was checked
1106 * against the slot in nfsd4_sequence().
1107 *
1108 * This occurs when seq->cachethis is FALSE, or when the client
1109 * session inactivity timer fires and a solo sequence operation
1110 * is sent (lease renewal).
1111 */
abfabf8c
AA		 1112 seq->maxslots = resp->cstate.session->se_fchannel.maxreqs;
1113
1114 /* Either returns 0 or nfserr_retry_uncached */
1115 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1116 if (status == nfserr_retry_uncached_rep)
1117 return status;
074fe897
AA		 1118
1119 if (!nfsd41_copy_replay_data(resp, entry)) {
1120 /*
1121 * Not enough room to use the replay rpc header, send the
1122 * cached header. Release all the allocated result pages.
1123 */
1124 svc_free_res_pages(resp->rqstp);
1125 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1126 entry->ce_resused);
1127 } else {
1128 /* Release all but the first allocated result page */
1129
1130 resp->rqstp->rq_resused--;
1131 svc_free_res_pages(resp->rqstp);
1132
1133 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1134 &entry->ce_respages[1],
1135 entry->ce_resused - 1);
1136 }
1137
1138 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1139 resp->opcnt = entry->ce_opcnt;
1140 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1141 status = entry->ce_status;
1142
1143 return status;
1144}
1145
0733d213
AA		 1146/*
1147 * Set the exchange_id flags returned by the server.
1148 */
1149static void
1150nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1151{
1152 /* pNFS is not supported */
1153 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1154
1155 /* Referrals are supported, Migration is not. */
1156 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1157
1158 /* set the wire flags to return to client. */
1159 clid->flags = new->cl_exchange_flags;
1160}
1161
069b6ad4
AA		 1162__be32
1163nfsd4_exchange_id(struct svc_rqst *rqstp,
1164 struct nfsd4_compound_state *cstate,
1165 struct nfsd4_exchange_id *exid)
1166{
0733d213
AA		 1167 struct nfs4_client *unconf, *conf, *new;
1168 int status;
1169 unsigned int strhashval;
1170 char dname[HEXDIR_LEN];
363168b4 1171 char addr_str[INET6_ADDRSTRLEN];
0733d213 1172 nfs4_verifier verf = exid->verifier;
363168b4 1173 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1174
363168b4 1175 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1176 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1177 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1178 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1179 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1180
1181 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1182 return nfserr_inval;
1183
1184 /* Currently only support SP4_NONE */
1185 switch (exid->spa_how) {
1186 case SP4_NONE:
1187 break;
1188 case SP4_SSV:
1189 return nfserr_encr_alg_unsupp;
1190 default:
1191 BUG(); /* checked by xdr code */
1192 case SP4_MACH_CRED:
1193 return nfserr_serverfault; /* no excuse :-/ */
1194 }
1195
1196 status = nfs4_make_rec_clidname(dname, &exid->clname);
1197
1198 if (status)
1199 goto error;
1200
1201 strhashval = clientstr_hashval(dname);
1202
1203 nfs4_lock_state();
1204 status = nfs_ok;
1205
a1bcecd2 1206 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1207 if (conf) {
1208 if (!same_verf(&verf, &conf->cl_verifier)) {
1209 /* 18.35.4 case 8 */
1210 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1211 status = nfserr_not_same;
1212 goto out;
1213 }
1214 /* Client reboot: destroy old state */
1215 expire_client(conf);
1216 goto out_new;
1217 }
1218 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1219 /* 18.35.4 case 9 */
1220 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1221 status = nfserr_perm;
1222 goto out;
1223 }
1224 expire_client(conf);
1225 goto out_new;
1226 }
0733d213
AA		 1227 /*
1228 * Set bit when the owner id and verifier map to an already
1229 * confirmed client id (18.35.3).
1230 */
1231 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1232
1233 /*
1234 * Falling into 18.35.4 case 2, possible router replay.
1235 * Leave confirmed record intact and return same result.
1236 */
1237 copy_verf(conf, &verf);
1238 new = conf;
1239 goto out_copy;
6ddbbbfe
MS		 1240 }
1241
1242 /* 18.35.4 case 7 */
1243 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1244 status = nfserr_noent;
1245 goto out;
0733d213
AA		 1246 }
1247
a1bcecd2 1248 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1249 if (unconf) {
1250 /*
1251 * Possible retry or client restart. Per 18.35.4 case 4,
1252 * a new unconfirmed record should be generated regardless
1253 * of whether any properties have changed.
1254 */
1255 expire_client(unconf);
1256 }
1257
1258out_new:
1259 /* Normal case */
1260 new = create_client(exid->clname, dname);
1261 if (new == NULL) {
1262 status = nfserr_resource;
1263 goto out;
1264 }
1265
1266 copy_verf(new, &verf);
1267 copy_cred(&new->cl_cred, &rqstp->rq_cred);
363168b4 1268 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
0733d213
AA		 1269 gen_clid(new);
1270 gen_confirm(new);
1271 add_to_unconfirmed(new, strhashval);
1272out_copy:
1273 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1274 exid->clientid.cl_id = new->cl_clientid.cl_id;
1275
38eb76a5 1276 exid->seqid = 1;
0733d213
AA		 1277 nfsd4_set_ex_flags(new, exid);
1278
1279 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1280 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1281 status = nfs_ok;
1282
1283out:
1284 nfs4_unlock_state();
1285error:
1286 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1287 return status;
069b6ad4
AA		 1288}
1289
b85d4c01 1290static int
88e588d5 1291check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1292{
88e588d5
AA		 1293 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1294 slot_seqid);
b85d4c01
BH		 1295
1296 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1297 if (slot_inuse) {
1298 if (seqid == slot_seqid)
b85d4c01
BH		 1299 return nfserr_jukebox;
1300 else
1301 return nfserr_seq_misordered;
1302 }
1303 /* Normal */
88e588d5 1304 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1305 return nfs_ok;
1306 /* Replay */
88e588d5 1307 if (seqid == slot_seqid)
b85d4c01
BH		 1308 return nfserr_replay_cache;
1309 /* Wraparound */
88e588d5 1310 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1311 return nfs_ok;
1312 /* Misordered replay or misordered new request */
1313 return nfserr_seq_misordered;
1314}
1315
49557cc7
AA		 1316/*
1317 * Cache the create session result into the create session single DRC
1318 * slot cache by saving the xdr structure. sl_seqid has been set.
1319 * Do this for solo or embedded create session operations.
1320 */
1321static void
1322nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1323 struct nfsd4_clid_slot *slot, int nfserr)
1324{
1325 slot->sl_status = nfserr;
1326 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1327}
1328
1329static __be32
1330nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1331 struct nfsd4_clid_slot *slot)
1332{
1333 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1334 return slot->sl_status;
1335}
1336
069b6ad4
AA		 1337__be32
1338nfsd4_create_session(struct svc_rqst *rqstp,
1339 struct nfsd4_compound_state *cstate,
1340 struct nfsd4_create_session *cr_ses)
1341{
363168b4 1342 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1343 struct nfs4_client *conf, *unconf;
49557cc7 1344 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1345 int status = 0;
1346
1347 nfs4_lock_state();
1348 unconf = find_unconfirmed_client(&cr_ses->clientid);
1349 conf = find_confirmed_client(&cr_ses->clientid);
1350
1351 if (conf) {
49557cc7
AA		 1352 cs_slot = &conf->cl_cs_slot;
1353 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1354 if (status == nfserr_replay_cache) {
ec6b5d7b 1355 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1356 cs_slot->sl_seqid);
38eb76a5 1357 /* Return the cached reply status */
49557cc7 1358 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1359 goto out;
49557cc7 1360 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1361 status = nfserr_seq_misordered;
1362 dprintk("Sequence misordered!\n");
1363 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1364 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1365 goto out;
1366 }
49557cc7 1367 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1368 } else if (unconf) {
1369 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1370 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1371 status = nfserr_clid_inuse;
1372 goto out;
1373 }
1374
49557cc7
AA		 1375 cs_slot = &unconf->cl_cs_slot;
1376 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1377 if (status) {
1378 /* an unconfirmed replay returns misordered */
ec6b5d7b 1379 status = nfserr_seq_misordered;
49557cc7 1380 goto out_cache;
ec6b5d7b
AA		 1381 }
1382
49557cc7 1383 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1384 move_to_confirmed(unconf);
1385
1386 /*
1387 * We do not support RDMA or persistent sessions
1388 */
1389 cr_ses->flags &= ~SESSION4_PERSIST;
1390 cr_ses->flags &= ~SESSION4_RDMA;
1391
1392 conf = unconf;
1393 } else {
1394 status = nfserr_stale_clientid;
1395 goto out;
1396 }
1397
1398 status = alloc_init_session(rqstp, conf, cr_ses);
1399 if (status)
1400 goto out;
1401
ec6b5d7b
AA		 1402 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1403 NFS4_MAX_SESSIONID_LEN);
49557cc7 1404 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1405
49557cc7
AA		 1406out_cache:
1407 /* cache solo and embedded create sessions under the state lock */
1408 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1409out:
1410 nfs4_unlock_state();
1411 dprintk("%s returns %d\n", __func__, ntohl(status));
1412 return status;
069b6ad4
AA		 1413}
1414
1415__be32
1416nfsd4_destroy_session(struct svc_rqst *r,
1417 struct nfsd4_compound_state *cstate,
1418 struct nfsd4_destroy_session *sessionid)
1419{
e10e0cfc
BH		 1420 struct nfsd4_session *ses;
1421 u32 status = nfserr_badsession;
1422
1423 /* Notes:
1424 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1425 * - Should we return nfserr_back_chan_busy if waiting for
1426 * callbacks on to-be-destroyed session?
1427 * - Do we need to clear any callback info from previous session?
1428 */
1429
1430 dump_sessionid(__func__, &sessionid->sessionid);
1431 spin_lock(&sessionid_lock);
1432 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1433 if (!ses) {
1434 spin_unlock(&sessionid_lock);
1435 goto out;
1436 }
1437
1438 unhash_session(ses);
1439 spin_unlock(&sessionid_lock);
1440
1441 /* wait for callbacks */
1442 shutdown_callback_client(ses->se_client);
1443 nfsd4_put_session(ses);
1444 status = nfs_ok;
1445out:
1446 dprintk("%s returns %d\n", __func__, ntohl(status));
1447 return status;
069b6ad4
AA		 1448}
1449
1450__be32
b85d4c01 1451nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1452 struct nfsd4_compound_state *cstate,
1453 struct nfsd4_sequence *seq)
1454{
f9bb94c4 1455 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1456 struct nfsd4_session *session;
1457 struct nfsd4_slot *slot;
1458 int status;
1459
f9bb94c4
AA		 1460 if (resp->opcnt != 1)
1461 return nfserr_sequence_pos;
1462
b85d4c01
BH		 1463 spin_lock(&sessionid_lock);
1464 status = nfserr_badsession;
1465 session = find_in_sessionid_hashtbl(&seq->sessionid);
1466 if (!session)
1467 goto out;
1468
1469 status = nfserr_badslot;
6c18ba9f 1470 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1471 goto out;
1472
1473 slot = &session->se_slots[seq->slotid];
1474 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1475
88e588d5 1476 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1477 if (status == nfserr_replay_cache) {
1478 cstate->slot = slot;
1479 cstate->session = session;
da3846a2 1480 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1481 * for nfsd4_svc_encode_compoundres processing */
1482 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1483 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1484 goto replay_cache;
1485 }
1486 if (status)
1487 goto out;
1488
1489 /* Success! bump slot seqid */
1490 slot->sl_inuse = true;
1491 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1492 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1493 /* Always set the cache entry cachethis for solo sequence */
1494 if (nfsd4_is_solo_sequence(resp))
1495 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1496
1497 cstate->slot = slot;
1498 cstate->session = session;
1499
1500replay_cache:
1501 /* Renew the clientid on success and on replay.
1502 * Hold a session reference until done processing the compound:
1503 * nfsd4_put_session called only if the cstate slot is set.
1504 */
1505 renew_client(session->se_client);
1506 nfsd4_get_session(session);
1507out:
1508 spin_unlock(&sessionid_lock);
1509 dprintk("%s: return %d\n", __func__, ntohl(status));
1510 return status;
069b6ad4
AA		 1511}
1512
b37ad28b 1513__be32
b591480b
BF		 1514nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1515 struct nfsd4_setclientid *setclid)
1da177e4 1516{
363168b4 1517 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1518 struct xdr_netobj clname = {
1519 .len = setclid->se_namelen,
1520 .data = setclid->se_name,
1521 };
1522 nfs4_verifier clverifier = setclid->se_verf;
1523 unsigned int strhashval;
28ce6054 1524 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1525 __be32 status;
68e76ad0 1526 char *princ;
a55370a3 1527 char dname[HEXDIR_LEN];
1da177e4 1528
1da177e4 1529 if (!check_name(clname))
73aea4ec 1530 return nfserr_inval;
1da177e4 1531
a55370a3
N		 1532 status = nfs4_make_rec_clidname(dname, &clname);
1533 if (status)
73aea4ec 1534 return status;
a55370a3 1535
1da177e4
LT		 1536 /*
1537 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1538 * We get here on a DRC miss.
1539 */
1540
a55370a3 1541 strhashval = clientstr_hashval(dname);
1da177e4 1542
1da177e4 1543 nfs4_lock_state();
a1bcecd2 1544 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1545 if (conf) {
a186e767 1546 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1547 status = nfserr_clid_inuse;
026722c2 1548 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1549 char addr_str[INET6_ADDRSTRLEN];
1550 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1551 sizeof(addr_str));
1552 dprintk("NFSD: setclientid: string in use by client "
1553 "at %s\n", addr_str);
1da177e4
LT		 1554 goto out;
1555 }
1da177e4 1556 }
a186e767
BF		 1557 /*
1558 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1559 * has a description of SETCLIENTID request processing consisting
1560 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1561 */
a1bcecd2 1562 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1563 status = nfserr_resource;
1564 if (!conf) {
a186e767
BF		 1565 /*
1566 * RFC 3530 14.2.33 CASE 4:
1567 * placed first, because it is the normal case
1da177e4
LT		 1568 */
1569 if (unconf)
1570 expire_client(unconf);
a55370a3
N		 1571 new = create_client(clname, dname);
1572 if (new == NULL)
1da177e4 1573 goto out;
1da177e4 1574 gen_clid(new);
599e0a22 1575 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1576 /*
a186e767
BF		 1577 * RFC 3530 14.2.33 CASE 1:
1578 * probable callback update
1da177e4 1579 */
31f4a6c1
N		 1580 if (unconf) {
1581 /* Note this is removing unconfirmed {*x***},
1582 * which is stronger than RFC recommended {vxc**}.
1583 * This has the advantage that there is at most
1584 * one {*x***} in either list at any time.
1585 */
1586 expire_client(unconf);
1da177e4 1587 }
a55370a3
N		 1588 new = create_client(clname, dname);
1589 if (new == NULL)
1da177e4 1590 goto out;
1da177e4 1591 copy_clid(new, conf);
1da177e4
LT		 1592 } else if (!unconf) {
1593 /*
a186e767
BF		 1594 * RFC 3530 14.2.33 CASE 2:
1595 * probable client reboot; state will be removed if
1596 * confirmed.
1da177e4 1597 */
a55370a3
N		 1598 new = create_client(clname, dname);
1599 if (new == NULL)
1da177e4 1600 goto out;
1da177e4 1601 gen_clid(new);
49ba8781 1602 } else {
a186e767
BF		 1603 /*
1604 * RFC 3530 14.2.33 CASE 3:
1605 * probable client reboot; state will be removed if
1606 * confirmed.
1da177e4
LT		 1607 */
1608 expire_client(unconf);
a55370a3
N		 1609 new = create_client(clname, dname);
1610 if (new == NULL)
1da177e4 1611 goto out;
1da177e4 1612 gen_clid(new);
1da177e4 1613 }
c175b83c 1614 copy_verf(new, &clverifier);
363168b4 1615 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
61054b14 1616 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1617 princ = svc_gss_principal(rqstp);
1618 if (princ) {
1619 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1620 if (new->cl_principal == NULL) {
1621 free_client(new);
1622 goto out;
1623 }
1624 }
c175b83c
BF		 1625 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1626 gen_confirm(new);
fbf4665f 1627 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1628 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1629 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1630 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1631 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1632 status = nfs_ok;
1633out:
1634 nfs4_unlock_state();
1635 return status;
1636}
1637
1638
1639/*
a186e767
BF		 1640 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1641 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1642 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1643 */
b37ad28b 1644__be32
b591480b
BF		 1645nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1646 struct nfsd4_compound_state *cstate,
1647 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1648{
363168b4 1649 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1650 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1651 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1652 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1653 __be32 status;
1da177e4
LT		 1654
1655 if (STALE_CLIENTID(clid))
1656 return nfserr_stale_clientid;
1657 /*
1658 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1659 * We get here on a DRC miss.
1660 */
1661
1662 nfs4_lock_state();
21ab45a4
N		 1663
1664 conf = find_confirmed_client(clid);
1665 unconf = find_unconfirmed_client(clid);
1666
1667 status = nfserr_clid_inuse;
363168b4 1668 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1669 goto out;
363168b4 1670 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1671 goto out;
1672
a186e767
BF		 1673 /*
1674 * section 14.2.34 of RFC 3530 has a description of
1675 * SETCLIENTID_CONFIRM request processing consisting
1676 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1677 */
366e0c1d 1678 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1679 /*
1680 * RFC 3530 14.2.34 CASE 1:
1681 * callback update
1682 */
599e0a22 1683 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1684 status = nfserr_clid_inuse;
1685 else {
1a69c179
N		 1686 /* XXX: We just turn off callbacks until we can handle
1687 * change request correctly. */
c237dc03 1688 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1689 expire_client(unconf);
1da177e4 1690 status = nfs_ok;
1a69c179 1691
1da177e4 1692 }
f3aba4e5 1693 } else if (conf && !unconf) {
a186e767
BF		 1694 /*
1695 * RFC 3530 14.2.34 CASE 2:
1696 * probable retransmitted request; play it safe and
1697 * do nothing.
7c79f737 1698 */
599e0a22 1699 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1700 status = nfserr_clid_inuse;
21ab45a4 1701 else
1da177e4 1702 status = nfs_ok;
7c79f737 1703 } else if (!conf && unconf
599e0a22 1704 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1705 /*
1706 * RFC 3530 14.2.34 CASE 3:
1707 * Normal case; new or rebooted client:
7c79f737 1708 */
599e0a22 1709 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1710 status = nfserr_clid_inuse;
1711 } else {
1a69c179
N		 1712 unsigned int hash =
1713 clientstr_hashval(unconf->cl_recdir);
1714 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1715 hash, false);
1a69c179 1716 if (conf) {
c7b9a459 1717 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1718 expire_client(conf);
1719 }
1da177e4 1720 move_to_confirmed(unconf);
21ab45a4 1721 conf = unconf;
46f8a64b 1722 nfsd4_probe_callback(conf);
1a69c179 1723 status = nfs_ok;
1da177e4 1724 }
599e0a22
BF		 1725 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1726 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1727 &confirm)))) {
a186e767
BF		 1728 /*
1729 * RFC 3530 14.2.34 CASE 4:
1730 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1731 */
1da177e4 1732 status = nfserr_stale_clientid;
7c79f737 1733 } else {
08e8987c
N		 1734 /* check that we have hit one of the cases...*/
1735 status = nfserr_clid_inuse;
1736 }
1da177e4 1737out:
1da177e4
LT		 1738 nfs4_unlock_state();
1739 return status;
1740}
1741
1da177e4
LT		 1742/* OPEN Share state helper functions */
1743static inline struct nfs4_file *
1744alloc_init_file(struct inode *ino)
1745{
1746 struct nfs4_file *fp;
1747 unsigned int hashval = file_hashval(ino);
1748
e60d4398
N		 1749 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1750 if (fp) {
8b671b80 1751 atomic_set(&fp->fi_ref, 1);
1da177e4 1752 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1753 INIT_LIST_HEAD(&fp->fi_stateids);
1754 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1755 spin_lock(&recall_lock);
1da177e4 1756 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1757 spin_unlock(&recall_lock);
1da177e4
LT		 1758 fp->fi_inode = igrab(ino);
1759 fp->fi_id = current_fileid++;
47f9940c 1760 fp->fi_had_conflict = false;
1da177e4
LT		 1761 return fp;
1762 }
1763 return NULL;
1764}
1765
e60d4398 1766static void
e18b890b 1767nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1768{
e60d4398
N		 1769 if (*slab == NULL)
1770 return;
1a1d92c1 1771 kmem_cache_destroy(*slab);
e60d4398 1772 *slab = NULL;
1da177e4
LT		 1773}
1774
e8ff2a84 1775void
1da177e4
LT		 1776nfsd4_free_slabs(void)
1777{
e60d4398
N		 1778 nfsd4_free_slab(&stateowner_slab);
1779 nfsd4_free_slab(&file_slab);
5ac049ac 1780 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1781 nfsd4_free_slab(&deleg_slab);
e60d4398 1782}
1da177e4 1783
e60d4398
N		 1784static int
1785nfsd4_init_slabs(void)
1786{
1787 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1788 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1789 if (stateowner_slab == NULL)
1790 goto out_nomem;
1791 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1792 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1793 if (file_slab == NULL)
1794 goto out_nomem;
5ac049ac 1795 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1796 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1797 if (stateid_slab == NULL)
1798 goto out_nomem;
5b2d21c1 1799 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1800 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1801 if (deleg_slab == NULL)
1802 goto out_nomem;
e60d4398
N		 1803 return 0;
1804out_nomem:
1805 nfsd4_free_slabs();
1806 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1807 return -ENOMEM;
1da177e4
LT		 1808}
1809
1810void
1811nfs4_free_stateowner(struct kref *kref)
1812{
1813 struct nfs4_stateowner *sop =
1814 container_of(kref, struct nfs4_stateowner, so_ref);
1815 kfree(sop->so_owner.data);
1816 kmem_cache_free(stateowner_slab, sop);
1817}
1818
1819static inline struct nfs4_stateowner *
1820alloc_stateowner(struct xdr_netobj *owner)
1821{
1822 struct nfs4_stateowner *sop;
1823
1824 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1825 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1826 memcpy(sop->so_owner.data, owner->data, owner->len);
1827 sop->so_owner.len = owner->len;
1828 kref_init(&sop->so_ref);
1829 return sop;
1830 }
1831 kmem_cache_free(stateowner_slab, sop);
1832 }
1833 return NULL;
1834}
1835
1836static struct nfs4_stateowner *
1837alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1838 struct nfs4_stateowner *sop;
1839 struct nfs4_replay *rp;
1840 unsigned int idhashval;
1841
1842 if (!(sop = alloc_stateowner(&open->op_owner)))
1843 return NULL;
1844 idhashval = ownerid_hashval(current_ownerid);
1845 INIT_LIST_HEAD(&sop->so_idhash);
1846 INIT_LIST_HEAD(&sop->so_strhash);
1847 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1848 INIT_LIST_HEAD(&sop->so_stateids);
1849 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1850 INIT_LIST_HEAD(&sop->so_close_lru);
1851 sop->so_time = 0;
1852 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1853 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1854 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1855 sop->so_is_open_owner = 1;
1856 sop->so_id = current_ownerid++;
1857 sop->so_client = clp;
1858 sop->so_seqid = open->op_seqid;
1859 sop->so_confirmed = 0;
1860 rp = &sop->so_replay;
de1ae286 1861 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1862 rp->rp_buflen = 0;
1863 rp->rp_buf = rp->rp_ibuf;
1864 return sop;
1865}
1866
1da177e4
LT		 1867static inline void
1868init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1869 struct nfs4_stateowner *sop = open->op_stateowner;
1870 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1871
1872 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1873 INIT_LIST_HEAD(&stp->st_perstateowner);
1874 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1875 INIT_LIST_HEAD(&stp->st_perfile);
1876 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1877 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1878 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1879 stp->st_stateowner = sop;
13cd2184 1880 get_nfs4_file(fp);
1da177e4 1881 stp->st_file = fp;
78155ed7 1882 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1883 stp->st_stateid.si_stateownerid = sop->so_id;
1884 stp->st_stateid.si_fileid = fp->fi_id;
1885 stp->st_stateid.si_generation = 0;
1886 stp->st_access_bmap = 0;
1887 stp->st_deny_bmap = 0;
84459a11
AA		 1888 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1889 &stp->st_access_bmap);
1da177e4 1890 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1891 stp->st_openstp = NULL;
1da177e4
LT		 1892}
1893
fd39ca9a 1894static void
1da177e4
LT		 1895move_to_close_lru(struct nfs4_stateowner *sop)
1896{
1897 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1898
358dd55a 1899 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1900 sop->so_time = get_seconds();
1901}
1902
1da177e4 1903static int
599e0a22
BF		 1904same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1905 clientid_t *clid)
1906{
1907 return (sop->so_owner.len == owner->len) &&
1908 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1909 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1910}
1911
1912static struct nfs4_stateowner *
1913find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1914{
1915 struct nfs4_stateowner *so = NULL;
1916
1917 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1918 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1919 return so;
1920 }
1921 return NULL;
1922}
1923
1924/* search file_hashtbl[] for file */
1925static struct nfs4_file *
1926find_file(struct inode *ino)
1927{
1928 unsigned int hashval = file_hashval(ino);
1929 struct nfs4_file *fp;
1930
8b671b80 1931 spin_lock(&recall_lock);
1da177e4 1932 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1933 if (fp->fi_inode == ino) {
1934 get_nfs4_file(fp);
8b671b80 1935 spin_unlock(&recall_lock);
1da177e4 1936 return fp;
13cd2184 1937 }
1da177e4 1938 }
8b671b80 1939 spin_unlock(&recall_lock);
1da177e4
LT		 1940 return NULL;
1941}
1942
d87a8ade 1943static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1944{
d87a8ade 1945 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1946 return 0;
d87a8ade
AA		 1947 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1948 return 0;
1949 x &= ~NFS4_SHARE_ACCESS_MASK;
1950 if (minorversion && x) {
1951 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1952 return 0;
1953 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1954 return 0;
1955 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1956 }
1957 if (x)
8838dc43
BF		 1958 return 0;
1959 return 1;
ba5a6a19
BF		 1960}
1961
8838dc43 1962static inline int deny_valid(u32 x)
ba5a6a19 1963{
8838dc43
BF		 1964 /* Note: unlike access bits, deny bits may be zero. */
1965 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1966}
1da177e4 1967
4f83aa30
BF		 1968/*
1969 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1970 * st_{access,deny}_bmap field of the stateid, in order to track not
1971 * only what share bits are currently in force, but also what
1972 * combinations of share bits previous opens have used. This allows us
1973 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1974 * return an error if the client attempt to downgrade to a combination
1975 * of share bits not explicable by closing some of its previous opens.
1976 *
1977 * XXX: This enforcement is actually incomplete, since we don't keep
1978 * track of access/deny bit combinations; so, e.g., we allow:
1979 *
1980 * OPEN allow read, deny write
1981 * OPEN allow both, deny none
1982 * DOWNGRADE allow read, deny none
1983 *
1984 * which we should reject.
1985 */
fd39ca9a 1986static void
1da177e4
LT		 1987set_access(unsigned int *access, unsigned long bmap) {
1988 int i;
1989
1990 *access = 0;
1991 for (i = 1; i < 4; i++) {
1992 if (test_bit(i, &bmap))
1993 *access |= i;
1994 }
1995}
1996
fd39ca9a 1997static void
1da177e4
LT		 1998set_deny(unsigned int *deny, unsigned long bmap) {
1999 int i;
2000
2001 *deny = 0;
2002 for (i = 0; i < 4; i++) {
2003 if (test_bit(i, &bmap))
2004 *deny |= i ;
2005 }
2006}
2007
2008static int
2009test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2010 unsigned int access, deny;
2011
2012 set_access(&access, stp->st_access_bmap);
2013 set_deny(&deny, stp->st_deny_bmap);
2014 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2015 return 0;
2016 return 1;
2017}
2018
2019/*
2020 * Called to check deny when READ with all zero stateid or
2021 * WRITE with all zero or all one stateid
2022 */
b37ad28b 2023static __be32
1da177e4
LT		 2024nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2025{
2026 struct inode *ino = current_fh->fh_dentry->d_inode;
2027 struct nfs4_file *fp;
2028 struct nfs4_stateid *stp;
b37ad28b 2029 __be32 ret;
1da177e4
LT		 2030
2031 dprintk("NFSD: nfs4_share_conflict\n");
2032
2033 fp = find_file(ino);
13cd2184
N		 2034 if (!fp)
2035 return nfs_ok;
b700949b 2036 ret = nfserr_locked;
1da177e4 2037 /* Search for conflicting share reservations */
13cd2184
N		 2038 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2039 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2040 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2041 goto out;
1da177e4 2042 }
13cd2184
N		 2043 ret = nfs_ok;
2044out:
2045 put_nfs4_file(fp);
2046 return ret;
1da177e4
LT		 2047}
2048
2049static inline void
2050nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2051{
2052 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2053 drop_file_write_access(filp);
1da177e4
LT		 2054 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2055 }
2056}
2057
1da177e4
LT		 2058/*
2059 * Spawn a thread to perform a recall on the delegation represented
2060 * by the lease (file_lock)
2061 *
2062 * Called from break_lease() with lock_kernel() held.
2063 * Note: we assume break_lease will only call this *once* for any given
2064 * lease.
2065 */
2066static
2067void nfsd_break_deleg_cb(struct file_lock *fl)
2068{
63e4863f 2069 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2070
2071 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2072 if (!dp)
2073 return;
2074
2075 /* We're assuming the state code never drops its reference
2076 * without first removing the lease. Since we're in this lease
2077 * callback (and since the lease code is serialized by the kernel
2078 * lock) we know the server hasn't removed the lease yet, we know
2079 * it's safe to take a reference: */
2080 atomic_inc(&dp->dl_count);
cfdcad4d 2081 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2082
2083 spin_lock(&recall_lock);
2084 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2085 spin_unlock(&recall_lock);
2086
2087 /* only place dl_time is set. protected by lock_kernel*/
2088 dp->dl_time = get_seconds();
2089
0272e1fd
BF		 2090 /*
2091 * We don't want the locks code to timeout the lease for us;
2092 * we'll remove it ourself if the delegation isn't returned
2093 * in time.
2094 */
2095 fl->fl_break_time = 0;
1da177e4 2096
63e4863f
BF		 2097 dp->dl_file->fi_had_conflict = true;
2098 nfsd4_cb_recall(dp);
1da177e4
LT		 2099}
2100
2101/*
2102 * The file_lock is being reapd.
2103 *
2104 * Called by locks_free_lock() with lock_kernel() held.
2105 */
2106static
2107void nfsd_release_deleg_cb(struct file_lock *fl)
2108{
2109 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2110
2111 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2112
2113 if (!(fl->fl_flags & FL_LEASE) || !dp)
2114 return;
2115 dp->dl_flock = NULL;
2116}
2117
2118/*
2119 * Set the delegation file_lock back pointer.
2120 *
a9933cea 2121 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2122 */
2123static
2124void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2125{
2126 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2127
2128 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2129 if (!dp)
2130 return;
2131 dp->dl_flock = new;
2132}
2133
2134/*
a9933cea 2135 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2136 */
2137static
2138int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2139{
2140 struct nfs4_delegation *onlistd =
2141 (struct nfs4_delegation *)onlist->fl_owner;
2142 struct nfs4_delegation *tryd =
2143 (struct nfs4_delegation *)try->fl_owner;
2144
2145 if (onlist->fl_lmops != try->fl_lmops)
2146 return 0;
2147
2148 return onlistd->dl_client == tryd->dl_client;
2149}
2150
2151
2152static
2153int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2154{
2155 if (arg & F_UNLCK)
2156 return lease_modify(onlist, arg);
2157 else
2158 return -EAGAIN;
2159}
2160
fd39ca9a 2161static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2162 .fl_break = nfsd_break_deleg_cb,
2163 .fl_release_private = nfsd_release_deleg_cb,
2164 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2165 .fl_mylease = nfsd_same_client_deleg_cb,
2166 .fl_change = nfsd_change_deleg_cb,
2167};
2168
2169
b37ad28b 2170__be32
6668958f
AA		 2171nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2172 struct nfsd4_open *open)
1da177e4 2173{
1da177e4
LT		 2174 clientid_t *clientid = &open->op_clientid;
2175 struct nfs4_client *clp = NULL;
2176 unsigned int strhashval;
2177 struct nfs4_stateowner *sop = NULL;
2178
1da177e4 2179 if (!check_name(open->op_owner))
0f442aa2 2180 return nfserr_inval;
1da177e4
LT		 2181
2182 if (STALE_CLIENTID(&open->op_clientid))
2183 return nfserr_stale_clientid;
2184
2185 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2186 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2187 open->op_stateowner = sop;
2188 if (!sop) {
2189 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2190 clp = find_confirmed_client(clientid);
2191 if (clp == NULL)
0f442aa2
BF		 2192 return nfserr_expired;
2193 goto renew;
1da177e4 2194 }
6668958f
AA		 2195 /* When sessions are used, skip open sequenceid processing */
2196 if (nfsd4_has_session(cstate))
2197 goto renew;
0f442aa2
BF		 2198 if (!sop->so_confirmed) {
2199 /* Replace unconfirmed owners without checking for replay. */
2200 clp = sop->so_client;
f044ff83 2201 release_openowner(sop);
0f442aa2
BF		 2202 open->op_stateowner = NULL;
2203 goto renew;
2204 }
2205 if (open->op_seqid == sop->so_seqid - 1) {
2206 if (sop->so_replay.rp_buflen)
a90b061c 2207 return nfserr_replay_me;
0f442aa2
BF		 2208 /* The original OPEN failed so spectacularly
2209 * that we don't even have replay data saved!
2210 * Therefore, we have no choice but to continue
2211 * processing this OPEN; presumably, we'll
2212 * fail again for the same reason.
2213 */
2214 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2215 goto renew;
2216 }
2217 if (open->op_seqid != sop->so_seqid)
2218 return nfserr_bad_seqid;
1da177e4 2219renew:
0f442aa2
BF		 2220 if (open->op_stateowner == NULL) {
2221 sop = alloc_init_open_stateowner(strhashval, clp, open);
2222 if (sop == NULL)
2223 return nfserr_resource;
2224 open->op_stateowner = sop;
2225 }
2226 list_del_init(&sop->so_close_lru);
1da177e4 2227 renew_client(sop->so_client);
0f442aa2 2228 return nfs_ok;
1da177e4
LT		 2229}
2230
b37ad28b 2231static inline __be32
4a6e43e6
N		 2232nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2233{
2234 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2235 return nfserr_openmode;
2236 else
2237 return nfs_ok;
2238}
2239
52f4fb43
N		 2240static struct nfs4_delegation *
2241find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2242{
2243 struct nfs4_delegation *dp;
2244
ea1da636 2245 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2246 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2247 return dp;
2248 }
2249 return NULL;
2250}
2251
b37ad28b 2252static __be32
567d9829
N		 2253nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2254 struct nfs4_delegation **dp)
2255{
2256 int flags;
b37ad28b 2257 __be32 status = nfserr_bad_stateid;
567d9829
N		 2258
2259 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2260 if (*dp == NULL)
c44c5eeb 2261 goto out;
567d9829
N		 2262 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2263 RD_STATE : WR_STATE;
2264 status = nfs4_check_delegmode(*dp, flags);
2265 if (status)
2266 *dp = NULL;
c44c5eeb
N		 2267out:
2268 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2269 return nfs_ok;
2270 if (status)
2271 return status;
2272 open->op_stateowner->so_confirmed = 1;
2273 return nfs_ok;
567d9829
N		 2274}
2275
b37ad28b 2276static __be32
1da177e4
LT		 2277nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2278{
2279 struct nfs4_stateid *local;
b37ad28b 2280 __be32 status = nfserr_share_denied;
1da177e4
LT		 2281 struct nfs4_stateowner *sop = open->op_stateowner;
2282
8beefa24 2283 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2284 /* ignore lock owners */
2285 if (local->st_stateowner->so_is_open_owner == 0)
2286 continue;
2287 /* remember if we have seen this open owner */
2288 if (local->st_stateowner == sop)
2289 *stpp = local;
2290 /* check for conflicting share reservations */
2291 if (!test_share(local, open))
2292 goto out;
2293 }
2294 status = 0;
2295out:
2296 return status;
2297}
2298
5ac049ac
N		 2299static inline struct nfs4_stateid *
2300nfs4_alloc_stateid(void)
2301{
2302 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2303}
2304
b37ad28b 2305static __be32
1da177e4 2306nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2307 struct nfs4_delegation *dp,
1da177e4
LT		 2308 struct svc_fh *cur_fh, int flags)
2309{
2310 struct nfs4_stateid *stp;
1da177e4 2311
5ac049ac 2312 stp = nfs4_alloc_stateid();
1da177e4
LT		 2313 if (stp == NULL)
2314 return nfserr_resource;
2315
567d9829
N		 2316 if (dp) {
2317 get_file(dp->dl_vfs_file);
2318 stp->st_vfs_file = dp->dl_vfs_file;
2319 } else {
b37ad28b 2320 __be32 status;
567d9829
N		 2321 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2322 &stp->st_vfs_file);
2323 if (status) {
2324 if (status == nfserr_dropit)
2325 status = nfserr_jukebox;
5ac049ac 2326 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2327 return status;
2328 }
1da177e4 2329 }
1da177e4
LT		 2330 *stpp = stp;
2331 return 0;
2332}
2333
b37ad28b 2334static inline __be32
1da177e4
LT		 2335nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2336 struct nfsd4_open *open)
2337{
2338 struct iattr iattr = {
2339 .ia_valid = ATTR_SIZE,
2340 .ia_size = 0,
2341 };
2342 if (!open->op_truncate)
2343 return 0;
2344 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2345 return nfserr_inval;
1da177e4
LT		 2346 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2347}
2348
b37ad28b 2349static __be32
1da177e4
LT		 2350nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2351{
2352 struct file *filp = stp->st_vfs_file;
7eaa36e2 2353 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2354 unsigned int share_access, new_writer;
b37ad28b 2355 __be32 status;
1da177e4
LT		 2356
2357 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2358 new_writer = (~share_access) & open->op_share_access
2359 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2360
6c26d08f 2361 if (new_writer) {
b37ad28b
AV		 2362 int err = get_write_access(inode);
2363 if (err)
2364 return nfserrno(err);
e518f056
BH		 2365 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2366 if (err)
2367 return nfserrno(err);
2368 file_take_write(filp);
6c26d08f 2369 }
1da177e4
LT		 2370 status = nfsd4_truncate(rqstp, cur_fh, open);
2371 if (status) {
6c26d08f
BF		 2372 if (new_writer)
2373 put_write_access(inode);
1da177e4
LT		 2374 return status;
2375 }
2376 /* remember the open */
6c26d08f 2377 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2378 __set_bit(open->op_share_access, &stp->st_access_bmap);
2379 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2380
2381 return nfs_ok;
2382}
2383
2384
1da177e4 2385static void
37515177 2386nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2387{
37515177
N		 2388 open->op_stateowner->so_confirmed = 1;
2389 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2390}
2391
2392/*
2393 * Attempt to hand out a delegation.
2394 */
2395static void
2396nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2397{
2398 struct nfs4_delegation *dp;
2399 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2400 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2401 struct file_lock fl, *flp = &fl;
2402 int status, flag = 0;
2403
2404 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2405 open->op_recall = 0;
2406 switch (open->op_claim_type) {
2407 case NFS4_OPEN_CLAIM_PREVIOUS:
2408 if (!atomic_read(&cb->cb_set))
2409 open->op_recall = 1;
2410 flag = open->op_delegate_type;
2411 if (flag == NFS4_OPEN_DELEGATE_NONE)
2412 goto out;
2413 break;
2414 case NFS4_OPEN_CLAIM_NULL:
2415 /* Let's not give out any delegations till everyone's
2416 * had the chance to reclaim theirs.... */
af558e33 2417 if (locks_in_grace())
7b190fec
N		 2418 goto out;
2419 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2420 goto out;
2421 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2422 flag = NFS4_OPEN_DELEGATE_WRITE;
2423 else
2424 flag = NFS4_OPEN_DELEGATE_READ;
2425 break;
2426 default:
2427 goto out;
2428 }
1da177e4
LT		 2429
2430 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2431 if (dp == NULL) {
2432 flag = NFS4_OPEN_DELEGATE_NONE;
2433 goto out;
2434 }
2435 locks_init_lock(&fl);
2436 fl.fl_lmops = &nfsd_lease_mng_ops;
2437 fl.fl_flags = FL_LEASE;
9167f501 2438 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2439 fl.fl_end = OFFSET_MAX;
2440 fl.fl_owner = (fl_owner_t)dp;
2441 fl.fl_file = stp->st_vfs_file;
2442 fl.fl_pid = current->tgid;
2443
a9933cea 2444 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2445 * the lock_manager callbacks fl_mylease and fl_change are used
2446 */
9167f501 2447 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2448 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2449 unhash_delegation(dp);
1da177e4
LT		 2450 flag = NFS4_OPEN_DELEGATE_NONE;
2451 goto out;
2452 }
2453
2454 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2455
2456 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2457 dp->dl_stateid.si_boot,
2458 dp->dl_stateid.si_stateownerid,
2459 dp->dl_stateid.si_fileid,
2460 dp->dl_stateid.si_generation);
2461out:
7b190fec
N		 2462 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2463 && flag == NFS4_OPEN_DELEGATE_NONE
2464 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2465 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2466 open->op_delegate_type = flag;
2467}
2468
2469/*
2470 * called with nfs4_lock_state() held.
2471 */
b37ad28b 2472__be32
1da177e4
LT		 2473nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2474{
6668958f 2475 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2476 struct nfs4_file *fp = NULL;
2477 struct inode *ino = current_fh->fh_dentry->d_inode;
2478 struct nfs4_stateid *stp = NULL;
567d9829 2479 struct nfs4_delegation *dp = NULL;
b37ad28b 2480 __be32 status;
1da177e4
LT		 2481
2482 status = nfserr_inval;
d87a8ade 2483 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2484 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2485 goto out;
2486 /*
2487 * Lookup file; if found, lookup stateid and check open request,
2488 * and check for delegations in the process of being recalled.
2489 * If not found, create the nfs4_file struct
2490 */
2491 fp = find_file(ino);
2492 if (fp) {
2493 if ((status = nfs4_check_open(fp, open, &stp)))
2494 goto out;
c44c5eeb
N		 2495 status = nfs4_check_deleg(fp, open, &dp);
2496 if (status)
2497 goto out;
1da177e4 2498 } else {
c44c5eeb
N		 2499 status = nfserr_bad_stateid;
2500 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2501 goto out;
1da177e4
LT		 2502 status = nfserr_resource;
2503 fp = alloc_init_file(ino);
2504 if (fp == NULL)
2505 goto out;
2506 }
2507
2508 /*
2509 * OPEN the file, or upgrade an existing OPEN.
2510 * If truncate fails, the OPEN fails.
2511 */
2512 if (stp) {
2513 /* Stateid was found, this is an OPEN upgrade */
2514 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2515 if (status)
2516 goto out;
444c2c07 2517 update_stateid(&stp->st_stateid);
1da177e4
LT		 2518 } else {
2519 /* Stateid was not found, this is a new OPEN */
2520 int flags = 0;
9ecb6a08 2521 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2522 flags |= NFSD_MAY_READ;
1da177e4 2523 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2524 flags |= NFSD_MAY_WRITE;
567d9829
N		 2525 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2526 if (status)
1da177e4
LT		 2527 goto out;
2528 init_stateid(stp, fp, open);
2529 status = nfsd4_truncate(rqstp, current_fh, open);
2530 if (status) {
2283963f 2531 release_open_stateid(stp);
1da177e4
LT		 2532 goto out;
2533 }
6668958f
AA		 2534 if (nfsd4_has_session(&resp->cstate))
2535 update_stateid(&stp->st_stateid);
1da177e4
LT		 2536 }
2537 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2538
6668958f
AA		 2539 if (nfsd4_has_session(&resp->cstate))
2540 open->op_stateowner->so_confirmed = 1;
2541
1da177e4
LT		 2542 /*
2543 * Attempt to hand out a delegation. No error return, because the
2544 * OPEN succeeds even if we fail.
2545 */
2546 nfs4_open_delegation(current_fh, open, stp);
2547
2548 status = nfs_ok;
2549
2550 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2551 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2552 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2553out:
13cd2184
N		 2554 if (fp)
2555 put_nfs4_file(fp);
37515177
N		 2556 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2557 nfs4_set_claim_prev(open);
1da177e4
LT		 2558 /*
2559 * To finish the open response, we just need to set the rflags.
2560 */
2561 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2562 if (!open->op_stateowner->so_confirmed &&
2563 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2564 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2565
2566 return status;
2567}
2568
b37ad28b 2569__be32
b591480b
BF		 2570nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2571 clientid_t *clid)
1da177e4
LT		 2572{
2573 struct nfs4_client *clp;
b37ad28b 2574 __be32 status;
1da177e4
LT		 2575
2576 nfs4_lock_state();
2577 dprintk("process_renew(%08x/%08x): starting\n",
2578 clid->cl_boot, clid->cl_id);
2579 status = nfserr_stale_clientid;
2580 if (STALE_CLIENTID(clid))
2581 goto out;
2582 clp = find_confirmed_client(clid);
2583 status = nfserr_expired;
2584 if (clp == NULL) {
2585 /* We assume the client took too long to RENEW. */
2586 dprintk("nfsd4_renew: clientid not found!\n");
2587 goto out;
2588 }
2589 renew_client(clp);
2590 status = nfserr_cb_path_down;
ea1da636 2591 if (!list_empty(&clp->cl_delegations)
c237dc03 2592 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2593 goto out;
2594 status = nfs_ok;
2595out:
2596 nfs4_unlock_state();
2597 return status;
2598}
2599
af558e33
BF		 2600struct lock_manager nfsd4_manager = {
2601};
2602
a76b4319 2603static void
af558e33 2604nfsd4_end_grace(void)
a76b4319
N		 2605{
2606 dprintk("NFSD: end of grace period\n");
c7b9a459 2607 nfsd4_recdir_purge_old();
af558e33 2608 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2609}
2610
fd39ca9a 2611static time_t
1da177e4
LT		 2612nfs4_laundromat(void)
2613{
2614 struct nfs4_client *clp;
2615 struct nfs4_stateowner *sop;
2616 struct nfs4_delegation *dp;
2617 struct list_head *pos, *next, reaplist;
2618 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2619 time_t t, clientid_val = NFSD_LEASE_TIME;
2620 time_t u, test_val = NFSD_LEASE_TIME;
2621
2622 nfs4_lock_state();
2623
2624 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2625 if (locks_in_grace())
2626 nfsd4_end_grace();
1da177e4
LT		 2627 list_for_each_safe(pos, next, &client_lru) {
2628 clp = list_entry(pos, struct nfs4_client, cl_lru);
2629 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2630 t = clp->cl_time - cutoff;
2631 if (clientid_val > t)
2632 clientid_val = t;
2633 break;
2634 }
2635 dprintk("NFSD: purging unused client (clientid %08x)\n",
2636 clp->cl_clientid.cl_id);
c7b9a459 2637 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2638 expire_client(clp);
2639 }
2640 INIT_LIST_HEAD(&reaplist);
2641 spin_lock(&recall_lock);
2642 list_for_each_safe(pos, next, &del_recall_lru) {
2643 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2644 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2645 u = dp->dl_time - cutoff;
2646 if (test_val > u)
2647 test_val = u;
2648 break;
2649 }
2650 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2651 dp, dp->dl_flock);
2652 list_move(&dp->dl_recall_lru, &reaplist);
2653 }
2654 spin_unlock(&recall_lock);
2655 list_for_each_safe(pos, next, &reaplist) {
2656 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2657 list_del_init(&dp->dl_recall_lru);
2658 unhash_delegation(dp);
2659 }
2660 test_val = NFSD_LEASE_TIME;
2661 list_for_each_safe(pos, next, &close_lru) {
2662 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2663 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2664 u = sop->so_time - cutoff;
2665 if (test_val > u)
2666 test_val = u;
2667 break;
2668 }
2669 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2670 sop->so_id);
f044ff83 2671 release_openowner(sop);
1da177e4
LT		 2672 }
2673 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2674 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2675 nfs4_unlock_state();
2676 return clientid_val;
2677}
2678
a254b246
HH		 2679static struct workqueue_struct *laundry_wq;
2680static void laundromat_main(struct work_struct *);
2681static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2682
2683static void
c4028958 2684laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2685{
2686 time_t t;
2687
2688 t = nfs4_laundromat();
2689 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2690 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2691}
2692
fd39ca9a 2693static struct nfs4_stateowner *
f8816512
N		 2694search_close_lru(u32 st_id, int flags)
2695{
1da177e4
LT		 2696 struct nfs4_stateowner *local = NULL;
2697
1da177e4
LT		 2698 if (flags & CLOSE_STATE) {
2699 list_for_each_entry(local, &close_lru, so_close_lru) {
2700 if (local->so_id == st_id)
2701 return local;
2702 }
2703 }
2704 return NULL;
2705}
2706
2707static inline int
2708nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2709{
7eaa36e2 2710 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2711}
2712
2713static int
2714STALE_STATEID(stateid_t *stateid)
2715{
78155ed7
BN		 2716 if (time_after((unsigned long)boot_time,
2717 (unsigned long)stateid->si_boot)) {
2718 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2719 stateid->si_boot, stateid->si_stateownerid,
2720 stateid->si_fileid, stateid->si_generation);
2721 return 1;
2722 }
2723 return 0;
2724}
2725
2726static int
2727EXPIRED_STATEID(stateid_t *stateid)
2728{
2729 if (time_before((unsigned long)boot_time,
2730 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2731 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2732 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2733 stateid->si_boot, stateid->si_stateownerid,
2734 stateid->si_fileid, stateid->si_generation);
2735 return 1;
2736 }
2737 return 0;
2738}
2739
2740static __be32
2741stateid_error_map(stateid_t *stateid)
2742{
2743 if (STALE_STATEID(stateid))
2744 return nfserr_stale_stateid;
2745 if (EXPIRED_STATEID(stateid))
2746 return nfserr_expired;
2747
2748 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2749 stateid->si_boot, stateid->si_stateownerid,
2750 stateid->si_fileid, stateid->si_generation);
2751 return nfserr_bad_stateid;
1da177e4
LT		 2752}
2753
2754static inline int
2755access_permit_read(unsigned long access_bmap)
2756{
2757 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2758 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2759 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2760}
2761
2762static inline int
2763access_permit_write(unsigned long access_bmap)
2764{
2765 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2766 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2767}
2768
2769static
b37ad28b 2770__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2771{
b37ad28b 2772 __be32 status = nfserr_openmode;
1da177e4
LT		 2773
2774 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2775 goto out;
2776 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2777 goto out;
2778 status = nfs_ok;
2779out:
2780 return status;
2781}
2782
b37ad28b 2783static inline __be32
1da177e4
LT		 2784check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2785{
203a8c8e 2786 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2787 return nfs_ok;
af558e33 2788 else if (locks_in_grace()) {
1da177e4
LT		 2789 /* Answer in remaining cases depends on existance of
2790 * conflicting state; so we must wait out the grace period. */
2791 return nfserr_grace;
2792 } else if (flags & WR_STATE)
2793 return nfs4_share_conflict(current_fh,
2794 NFS4_SHARE_DENY_WRITE);
2795 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2796 return nfs4_share_conflict(current_fh,
2797 NFS4_SHARE_DENY_READ);
2798}
2799
2800/*
2801 * Allow READ/WRITE during grace period on recovered state only for files
2802 * that are not able to provide mandatory locking.
2803 */
2804static inline int
18f82731 2805grace_disallows_io(struct inode *inode)
1da177e4 2806{
203a8c8e 2807 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2808}
2809
6668958f 2810static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2811{
6668958f
AA		 2812 /*
2813 * When sessions are used the stateid generation number is ignored
2814 * when it is zero.
2815 */
2816 if ((flags & HAS_SESSION) && in->si_generation == 0)
2817 goto out;
2818
0836f587
BF		 2819 /* If the client sends us a stateid from the future, it's buggy: */
2820 if (in->si_generation > ref->si_generation)
2821 return nfserr_bad_stateid;
2822 /*
2823 * The following, however, can happen. For example, if the
2824 * client sends an open and some IO at the same time, the open
2825 * may bump si_generation while the IO is still in flight.
2826 * Thanks to hard links and renames, the client never knows what
2827 * file an open will affect. So it could avoid that situation
2828 * only by serializing all opens and IO from the same open
2829 * owner. To recover from the old_stateid error, the client
2830 * will just have to retry the IO:
2831 */
2832 if (in->si_generation < ref->si_generation)
2833 return nfserr_old_stateid;
6668958f 2834out:
0836f587
BF		 2835 return nfs_ok;
2836}
2837
3e633079
BF		 2838static int is_delegation_stateid(stateid_t *stateid)
2839{
2840 return stateid->si_fileid == 0;
2841}
2842
1da177e4
LT		 2843/*
2844* Checks for stateid operations
2845*/
b37ad28b 2846__be32
dd453dfd
BH		 2847nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2848 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2849{
2850 struct nfs4_stateid *stp = NULL;
2851 struct nfs4_delegation *dp = NULL;
dd453dfd 2852 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2853 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2854 __be32 status;
1da177e4 2855
1da177e4
LT		 2856 if (filpp)
2857 *filpp = NULL;
2858
18f82731 2859 if (grace_disallows_io(ino))
1da177e4
LT		 2860 return nfserr_grace;
2861
6668958f
AA		 2862 if (nfsd4_has_session(cstate))
2863 flags |= HAS_SESSION;
2864
1da177e4
LT		 2865 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2866 return check_special_stateids(current_fh, stateid, flags);
2867
1da177e4
LT		 2868 status = nfserr_stale_stateid;
2869 if (STALE_STATEID(stateid))
2870 goto out;
2871
1da177e4 2872 status = nfserr_bad_stateid;
3e633079 2873 if (is_delegation_stateid(stateid)) {
a4455be0 2874 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2875 if (!dp) {
2876 status = stateid_error_map(stateid);
1da177e4 2877 goto out;
78155ed7 2878 }
6668958f
AA		 2879 status = check_stateid_generation(stateid, &dp->dl_stateid,
2880 flags);
0c2a498f
BF		 2881 if (status)
2882 goto out;
dc9bf700
BF		 2883 status = nfs4_check_delegmode(dp, flags);
2884 if (status)
2885 goto out;
2886 renew_client(dp->dl_client);
dc9bf700
BF		 2887 if (filpp)
2888 *filpp = dp->dl_vfs_file;
1da177e4 2889 } else { /* open or lock stateid */
a4455be0 2890 stp = find_stateid(stateid, flags);
78155ed7
BN		 2891 if (!stp) {
2892 status = stateid_error_map(stateid);
1da177e4 2893 goto out;
78155ed7 2894 }
6150ef0d 2895 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2896 goto out;
2897 if (!stp->st_stateowner->so_confirmed)
2898 goto out;
6668958f
AA		 2899 status = check_stateid_generation(stateid, &stp->st_stateid,
2900 flags);
0c2a498f
BF		 2901 if (status)
2902 goto out;
a4455be0
BF		 2903 status = nfs4_check_openmode(stp, flags);
2904 if (status)
1da177e4
LT		 2905 goto out;
2906 renew_client(stp->st_stateowner->so_client);
2907 if (filpp)
2908 *filpp = stp->st_vfs_file;
1da177e4
LT		 2909 }
2910 status = nfs_ok;
2911out:
2912 return status;
2913}
2914
4c4cd222
N		 2915static inline int
2916setlkflg (int type)
2917{
2918 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2919 RD_STATE : WR_STATE;
2920}
1da177e4
LT		 2921
2922/*
2923 * Checks for sequence id mutating operations.
2924 */
b37ad28b 2925static __be32
dd453dfd
BH		 2926nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2927 stateid_t *stateid, int flags,
2928 struct nfs4_stateowner **sopp,
2929 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2930{
1da177e4
LT		 2931 struct nfs4_stateid *stp;
2932 struct nfs4_stateowner *sop;
dd453dfd 2933 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2934 __be32 status;
1da177e4
LT		 2935
2936 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2937 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2938 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2939 stateid->si_generation);
3a4f98bb 2940
1da177e4
LT		 2941 *stpp = NULL;
2942 *sopp = NULL;
2943
1da177e4 2944 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2945 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2946 return nfserr_bad_stateid;
1da177e4
LT		 2947 }
2948
1da177e4 2949 if (STALE_STATEID(stateid))
3a4f98bb 2950 return nfserr_stale_stateid;
6668958f
AA		 2951
2952 if (nfsd4_has_session(cstate))
2953 flags |= HAS_SESSION;
2954
1da177e4
LT		 2955 /*
2956 * We return BAD_STATEID if filehandle doesn't match stateid,
2957 * the confirmed flag is incorrecly set, or the generation
2958 * number is incorrect.
1da177e4 2959 */
f8816512
N		 2960 stp = find_stateid(stateid, flags);
2961 if (stp == NULL) {
2962 /*
2963 * Also, we should make sure this isn't just the result of
2964 * a replayed close:
2965 */
2966 sop = search_close_lru(stateid->si_stateownerid, flags);
2967 if (sop == NULL)
78155ed7 2968 return stateid_error_map(stateid);
f8816512
N		 2969 *sopp = sop;
2970 goto check_replay;
2971 }
1da177e4 2972
39325bd0
BF		 2973 *stpp = stp;
2974 *sopp = sop = stp->st_stateowner;
2975
4c4cd222 2976 if (lock) {
4c4cd222 2977 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2978 struct nfs4_client *clp = sop->so_client;
4c4cd222 2979 int lkflg = 0;
b37ad28b 2980 __be32 status;
4c4cd222
N		 2981
2982 lkflg = setlkflg(lock->lk_type);
2983
2984 if (lock->lk_is_new) {
599e0a22
BF		 2985 if (!sop->so_is_open_owner)
2986 return nfserr_bad_stateid;
60adfc50
AA		 2987 if (!(flags & HAS_SESSION) &&
2988 !same_clid(&clp->cl_clientid, lockclid))
2989 return nfserr_bad_stateid;
599e0a22
BF		 2990 /* stp is the open stateid */
2991 status = nfs4_check_openmode(stp, lkflg);
2992 if (status)
2993 return status;
2994 } else {
2995 /* stp is the lock stateid */
2996 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2997 if (status)
2998 return status;
4c4cd222 2999 }
1da177e4
LT		 3000 }
3001
f3362737 3002 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3003 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3004 return nfserr_bad_stateid;
1da177e4
LT		 3005 }
3006
1da177e4
LT		 3007 /*
3008 * We now validate the seqid and stateid generation numbers.
3009 * For the moment, we ignore the possibility of
3010 * generation number wraparound.
3011 */
6668958f 3012 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3013 goto check_replay;
3014
3a4f98bb 3015 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3016 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3017 " unconfirmed stateowner!\n");
3018 return nfserr_bad_stateid;
1da177e4 3019 }
3a4f98bb 3020 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3021 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3022 " confirmed yet!\n");
3023 return nfserr_bad_stateid;
1da177e4 3024 }
6668958f 3025 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3026 if (status)
3027 return status;
52fd004e 3028 renew_client(sop->so_client);
3a4f98bb 3029 return nfs_ok;
1da177e4 3030
1da177e4 3031check_replay:
bd9aac52 3032 if (seqid == sop->so_seqid - 1) {
849823c5 3033 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3034 /* indicate replay to calling function */
a90b061c 3035 return nfserr_replay_me;
1da177e4 3036 }
2fdada03 3037 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3038 sop->so_seqid, seqid);
3039 *sopp = NULL;
3040 return nfserr_bad_seqid;
1da177e4
LT		 3041}
3042
b37ad28b 3043__be32
ca364317 3044nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3045 struct nfsd4_open_confirm *oc)
1da177e4 3046{
b37ad28b 3047 __be32 status;
1da177e4
LT		 3048 struct nfs4_stateowner *sop;
3049 struct nfs4_stateid *stp;
3050
3051 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3052 (int)cstate->current_fh.fh_dentry->d_name.len,
3053 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3054
ca364317 3055 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3056 if (status)
3057 return status;
1da177e4
LT		 3058
3059 nfs4_lock_state();
3060
dd453dfd 3061 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3062 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3063 CONFIRM | OPEN_STATE,
1da177e4
LT		 3064 &oc->oc_stateowner, &stp, NULL)))
3065 goto out;
3066
3067 sop = oc->oc_stateowner;
3068 sop->so_confirmed = 1;
3069 update_stateid(&stp->st_stateid);
3070 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3071 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3072 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3073 stp->st_stateid.si_boot,
3074 stp->st_stateid.si_stateownerid,
3075 stp->st_stateid.si_fileid,
3076 stp->st_stateid.si_generation);
c7b9a459
N		 3077
3078 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3079out:
f2327d9a 3080 if (oc->oc_stateowner) {
1da177e4 3081 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3082 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3083 }
1da177e4
LT		 3084 nfs4_unlock_state();
3085 return status;
3086}
3087
3088
3089/*
3090 * unset all bits in union bitmap (bmap) that
3091 * do not exist in share (from successful OPEN_DOWNGRADE)
3092 */
3093static void
3094reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3095{
3096 int i;
3097 for (i = 1; i < 4; i++) {
3098 if ((i & access) != i)
3099 __clear_bit(i, bmap);
3100 }
3101}
3102
3103static void
3104reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3105{
3106 int i;
3107 for (i = 0; i < 4; i++) {
3108 if ((i & deny) != i)
3109 __clear_bit(i, bmap);
3110 }
3111}
3112
b37ad28b 3113__be32
ca364317
BF		 3114nfsd4_open_downgrade(struct svc_rqst *rqstp,
3115 struct nfsd4_compound_state *cstate,
a4f1706a 3116 struct nfsd4_open_downgrade *od)
1da177e4 3117{
b37ad28b 3118 __be32 status;
1da177e4
LT		 3119 struct nfs4_stateid *stp;
3120 unsigned int share_access;
3121
3122 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3123 (int)cstate->current_fh.fh_dentry->d_name.len,
3124 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3125
d87a8ade 3126 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3127 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3128 return nfserr_inval;
3129
3130 nfs4_lock_state();
dd453dfd 3131 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3132 od->od_seqid,
1da177e4 3133 &od->od_stateid,
f3362737 3134 OPEN_STATE,
1da177e4
LT		 3135 &od->od_stateowner, &stp, NULL)))
3136 goto out;
3137
3138 status = nfserr_inval;
3139 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3140 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3141 stp->st_access_bmap, od->od_share_access);
3142 goto out;
3143 }
3144 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3145 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3146 stp->st_deny_bmap, od->od_share_deny);
3147 goto out;
3148 }
3149 set_access(&share_access, stp->st_access_bmap);
3150 nfs4_file_downgrade(stp->st_vfs_file,
3151 share_access & ~od->od_share_access);
3152
3153 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3154 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3155
3156 update_stateid(&stp->st_stateid);
3157 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3158 status = nfs_ok;
3159out:
f2327d9a 3160 if (od->od_stateowner) {
1da177e4 3161 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3162 cstate->replay_owner = od->od_stateowner;
f2327d9a 3163 }
1da177e4
LT		 3164 nfs4_unlock_state();
3165 return status;
3166}
3167
3168/*
3169 * nfs4_unlock_state() called after encode
3170 */
b37ad28b 3171__be32
ca364317 3172nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3173 struct nfsd4_close *close)
1da177e4 3174{
b37ad28b 3175 __be32 status;
1da177e4
LT		 3176 struct nfs4_stateid *stp;
3177
3178 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3179 (int)cstate->current_fh.fh_dentry->d_name.len,
3180 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3181
3182 nfs4_lock_state();
3183 /* check close_lru for replay */
dd453dfd 3184 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3185 close->cl_seqid,
1da177e4 3186 &close->cl_stateid,
f3362737 3187 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3188 &close->cl_stateowner, &stp, NULL)))
3189 goto out;
1da177e4
LT		 3190 status = nfs_ok;
3191 update_stateid(&stp->st_stateid);
3192 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3193
04ef5954 3194 /* release_stateid() calls nfsd_close() if needed */
2283963f 3195 release_open_stateid(stp);
04ef5954
BF		 3196
3197 /* place unused nfs4_stateowners on so_close_lru list to be
3198 * released by the laundromat service after the lease period
3199 * to enable us to handle CLOSE replay
3200 */
3201 if (list_empty(&close->cl_stateowner->so_stateids))
3202 move_to_close_lru(close->cl_stateowner);
1da177e4 3203out:
f2327d9a 3204 if (close->cl_stateowner) {
1da177e4 3205 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3206 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3207 }
1da177e4
LT		 3208 nfs4_unlock_state();
3209 return status;
3210}
3211
b37ad28b 3212__be32
ca364317
BF		 3213nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3214 struct nfsd4_delegreturn *dr)
1da177e4 3215{
203a8c8e
BF		 3216 struct nfs4_delegation *dp;
3217 stateid_t *stateid = &dr->dr_stateid;
3218 struct inode *inode;
b37ad28b 3219 __be32 status;
6668958f 3220 int flags = 0;
1da177e4 3221
ca364317 3222 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3223 return status;
3224 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3225
6668958f
AA		 3226 if (nfsd4_has_session(cstate))
3227 flags |= HAS_SESSION;
1da177e4 3228 nfs4_lock_state();
203a8c8e
BF		 3229 status = nfserr_bad_stateid;
3230 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3231 goto out;
3232 status = nfserr_stale_stateid;
3233 if (STALE_STATEID(stateid))
3234 goto out;
7e0f7cf5 3235 status = nfserr_bad_stateid;
203a8c8e
BF		 3236 if (!is_delegation_stateid(stateid))
3237 goto out;
203a8c8e 3238 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3239 if (!dp) {
3240 status = stateid_error_map(stateid);
203a8c8e 3241 goto out;
78155ed7 3242 }
6668958f 3243 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3244 if (status)
3245 goto out;
3246 renew_client(dp->dl_client);
3247
3248 unhash_delegation(dp);
1da177e4 3249out:
203a8c8e
BF		 3250 nfs4_unlock_state();
3251
1da177e4
LT		 3252 return status;
3253}
3254
3255
3256/*
3257 * Lock owner state (byte-range locks)
3258 */
3259#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3260#define LOCK_HASH_BITS 8
3261#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3262#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3263
87df4de8
BH		 3264static inline u64
3265end_offset(u64 start, u64 len)
3266{
3267 u64 end;
3268
3269 end = start + len;
3270 return end >= start ? end: NFS4_MAX_UINT64;
3271}
3272
3273/* last octet in a range */
3274static inline u64
3275last_byte_offset(u64 start, u64 len)
3276{
3277 u64 end;
3278
3279 BUG_ON(!len);
3280 end = start + len;
3281 return end > start ? end - 1: NFS4_MAX_UINT64;
3282}
3283
1da177e4
LT		 3284#define lockownerid_hashval(id) \
3285 ((id) & LOCK_HASH_MASK)
3286
3287static inline unsigned int
3288lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3289 struct xdr_netobj *ownername)
3290{
3291 return (file_hashval(inode) + cl_id
3292 + opaque_hashval(ownername->data, ownername->len))
3293 & LOCK_HASH_MASK;
3294}
3295
3296static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3297static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3298static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3299
fd39ca9a 3300static struct nfs4_stateid *
1da177e4
LT		 3301find_stateid(stateid_t *stid, int flags)
3302{
9346eff0 3303 struct nfs4_stateid *local;
1da177e4
LT		 3304 u32 st_id = stid->si_stateownerid;
3305 u32 f_id = stid->si_fileid;
3306 unsigned int hashval;
3307
3308 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3309 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3310 hashval = stateid_hashval(st_id, f_id);
3311 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3312 if ((local->st_stateid.si_stateownerid == st_id) &&
3313 (local->st_stateid.si_fileid == f_id))
3314 return local;
3315 }
3316 }
9346eff0
KK		 3317
3318 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3319 hashval = stateid_hashval(st_id, f_id);
3320 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3321 if ((local->st_stateid.si_stateownerid == st_id) &&
3322 (local->st_stateid.si_fileid == f_id))
3323 return local;
3324 }
849823c5 3325 }
1da177e4
LT		 3326 return NULL;
3327}
3328
3329static struct nfs4_delegation *
3330find_delegation_stateid(struct inode *ino, stateid_t *stid)
3331{
13cd2184
N		 3332 struct nfs4_file *fp;
3333 struct nfs4_delegation *dl;
1da177e4
LT		 3334
3335 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3336 stid->si_boot, stid->si_stateownerid,
3337 stid->si_fileid, stid->si_generation);
3338
1da177e4 3339 fp = find_file(ino);
13cd2184
N		 3340 if (!fp)
3341 return NULL;
3342 dl = find_delegation_file(fp, stid);
3343 put_nfs4_file(fp);
3344 return dl;
1da177e4
LT		 3345}
3346
3347/*
3348 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3349 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3350 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3351 * locking, this prevents us from being completely protocol-compliant. The
3352 * real solution to this problem is to start using unsigned file offsets in
3353 * the VFS, but this is a very deep change!
3354 */
3355static inline void
3356nfs4_transform_lock_offset(struct file_lock *lock)
3357{
3358 if (lock->fl_start < 0)
3359 lock->fl_start = OFFSET_MAX;
3360 if (lock->fl_end < 0)
3361 lock->fl_end = OFFSET_MAX;
3362}
3363
d5b9026a
N		 3364/* Hack!: For now, we're defining this just so we can use a pointer to it
3365 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3366static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3367};
1da177e4
LT		 3368
3369static inline void
3370nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3371{
d5b9026a
N		 3372 struct nfs4_stateowner *sop;
3373 unsigned int hval;
1da177e4 3374
d5b9026a
N		 3375 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3376 sop = (struct nfs4_stateowner *) fl->fl_owner;
3377 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3378 kref_get(&sop->so_ref);
3379 deny->ld_sop = sop;
3380 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3381 } else {
3382 deny->ld_sop = NULL;
3383 deny->ld_clientid.cl_boot = 0;
3384 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3385 }
3386 deny->ld_start = fl->fl_start;
87df4de8
BH		 3387 deny->ld_length = NFS4_MAX_UINT64;
3388 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3389 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3390 deny->ld_type = NFS4_READ_LT;
3391 if (fl->fl_type != F_RDLCK)
3392 deny->ld_type = NFS4_WRITE_LT;
3393}
3394
1da177e4
LT		 3395static struct nfs4_stateowner *
3396find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3397 struct xdr_netobj *owner)
3398{
3399 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3400 struct nfs4_stateowner *op;
3401
3402 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3403 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3404 return op;
3405 }
3406 return NULL;
3407}
3408
3409/*
3410 * Alloc a lock owner structure.
3411 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3412 * occured.
3413 *
3414 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3415 */
3416
3417static struct nfs4_stateowner *
3418alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3419 struct nfs4_stateowner *sop;
3420 struct nfs4_replay *rp;
3421 unsigned int idhashval;
3422
3423 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3424 return NULL;
3425 idhashval = lockownerid_hashval(current_ownerid);
3426 INIT_LIST_HEAD(&sop->so_idhash);
3427 INIT_LIST_HEAD(&sop->so_strhash);
3428 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3429 INIT_LIST_HEAD(&sop->so_stateids);
3430 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3431 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3432 sop->so_time = 0;
3433 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3434 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3435 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3436 sop->so_is_open_owner = 0;
3437 sop->so_id = current_ownerid++;
3438 sop->so_client = clp;
b59e3c0e
NB		 3439 /* It is the openowner seqid that will be incremented in encode in the
3440 * case of new lockowners; so increment the lock seqid manually: */
3441 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3442 sop->so_confirmed = 1;
3443 rp = &sop->so_replay;
de1ae286 3444 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3445 rp->rp_buflen = 0;
3446 rp->rp_buf = rp->rp_ibuf;
3447 return sop;
3448}
3449
fd39ca9a 3450static struct nfs4_stateid *
1da177e4
LT		 3451alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3452{
3453 struct nfs4_stateid *stp;
3454 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3455
5ac049ac
N		 3456 stp = nfs4_alloc_stateid();
3457 if (stp == NULL)
1da177e4
LT		 3458 goto out;
3459 INIT_LIST_HEAD(&stp->st_hash);
3460 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3461 INIT_LIST_HEAD(&stp->st_perstateowner);
3462 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3463 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3464 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3465 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3466 stp->st_stateowner = sop;
13cd2184 3467 get_nfs4_file(fp);
1da177e4 3468 stp->st_file = fp;
78155ed7 3469 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3470 stp->st_stateid.si_stateownerid = sop->so_id;
3471 stp->st_stateid.si_fileid = fp->fi_id;
3472 stp->st_stateid.si_generation = 0;
3473 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3474 stp->st_access_bmap = open_stp->st_access_bmap;
3475 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3476 stp->st_openstp = open_stp;
1da177e4
LT		 3477
3478out:
3479 return stp;
3480}
3481
fd39ca9a 3482static int
1da177e4
LT		 3483check_lock_length(u64 offset, u64 length)
3484{
87df4de8 3485 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3486 LOFF_OVERFLOW(offset, length)));
3487}
3488
3489/*
3490 * LOCK operation
3491 */
b37ad28b 3492__be32
ca364317 3493nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3494 struct nfsd4_lock *lock)
1da177e4 3495{
3e9e3dbe 3496 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3497 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3498 struct nfs4_stateid *lock_stp;
3499 struct file *filp;
3500 struct file_lock file_lock;
8dc7c311 3501 struct file_lock conflock;
b37ad28b 3502 __be32 status = 0;
1da177e4 3503 unsigned int strhashval;
150b3934 3504 unsigned int cmd;
b8dd7b9a 3505 int err;
1da177e4
LT		 3506
3507 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3508 (long long) lock->lk_offset,
3509 (long long) lock->lk_length);
3510
1da177e4
LT		 3511 if (check_lock_length(lock->lk_offset, lock->lk_length))
3512 return nfserr_inval;
3513
ca364317 3514 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3515 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3516 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3517 return status;
3518 }
3519
1da177e4
LT		 3520 nfs4_lock_state();
3521
3522 if (lock->lk_is_new) {
893f8770
N		 3523 /*
3524 * Client indicates that this is a new lockowner.
3525 * Use open owner and open stateid to create lock owner and
3526 * lock stateid.
3527 */
1da177e4
LT		 3528 struct nfs4_stateid *open_stp = NULL;
3529 struct nfs4_file *fp;
3530
3531 status = nfserr_stale_clientid;
60adfc50
AA		 3532 if (!nfsd4_has_session(cstate) &&
3533 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3534 goto out;
1da177e4 3535
1da177e4 3536 /* validate and update open stateid and open seqid */
dd453dfd 3537 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3538 lock->lk_new_open_seqid,
3539 &lock->lk_new_open_stateid,
f3362737 3540 OPEN_STATE,
3a65588a 3541 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3542 lock);
37515177 3543 if (status)
1da177e4 3544 goto out;
3a65588a 3545 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3546 /* create lockowner and lock stateid */
3547 fp = open_stp->st_file;
3548 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3549 open_sop->so_client->cl_clientid.cl_id,
3550 &lock->v.new.owner);
3e9e3dbe
N		 3551 /* XXX: Do we need to check for duplicate stateowners on
3552 * the same file, or should they just be allowed (and
3553 * create new stateids)? */
1da177e4 3554 status = nfserr_resource;
b59e3c0e
NB		 3555 lock_sop = alloc_init_lock_stateowner(strhashval,
3556 open_sop->so_client, open_stp, lock);
3557 if (lock_sop == NULL)
1da177e4 3558 goto out;
b59e3c0e 3559 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3560 if (lock_stp == NULL)
1da177e4 3561 goto out;
1da177e4
LT		 3562 } else {
3563 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3564 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3565 lock->lk_old_lock_seqid,
3566 &lock->lk_old_lock_stateid,
f3362737 3567 LOCK_STATE,
3a65588a 3568 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3569 if (status)
3570 goto out;
3a65588a 3571 lock_sop = lock->lk_replay_owner;
1da177e4 3572 }
3a65588a 3573 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3574 filp = lock_stp->st_vfs_file;
3575
0dd395dc 3576 status = nfserr_grace;
af558e33 3577 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3578 goto out;
3579 status = nfserr_no_grace;
af558e33 3580 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3581 goto out;
3582
1da177e4
LT		 3583 locks_init_lock(&file_lock);
3584 switch (lock->lk_type) {
3585 case NFS4_READ_LT:
3586 case NFS4_READW_LT:
3587 file_lock.fl_type = F_RDLCK;
150b3934 3588 cmd = F_SETLK;
1da177e4
LT		 3589 break;
3590 case NFS4_WRITE_LT:
3591 case NFS4_WRITEW_LT:
3592 file_lock.fl_type = F_WRLCK;
150b3934 3593 cmd = F_SETLK;
1da177e4
LT		 3594 break;
3595 default:
3596 status = nfserr_inval;
3597 goto out;
3598 }
b59e3c0e 3599 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3600 file_lock.fl_pid = current->tgid;
3601 file_lock.fl_file = filp;
3602 file_lock.fl_flags = FL_POSIX;
d5b9026a 3603 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3604
3605 file_lock.fl_start = lock->lk_offset;
87df4de8 3606 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3607 nfs4_transform_lock_offset(&file_lock);
3608
3609 /*
3610 * Try to lock the file in the VFS.
3611 * Note: locks.c uses the BKL to protect the inode's lock list.
3612 */
3613
fd85b817 3614 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3615 switch (-err) {
1da177e4
LT		 3616 case 0: /* success! */
3617 update_stateid(&lock_stp->st_stateid);
3618 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3619 sizeof(stateid_t));
b8dd7b9a 3620 status = 0;
eb76b3fd
AA		 3621 break;
3622 case (EAGAIN): /* conflock holds conflicting lock */
3623 status = nfserr_denied;
3624 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3625 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3626 break;
1da177e4
LT		 3627 case (EDEADLK):
3628 status = nfserr_deadlock;
eb76b3fd 3629 break;
1da177e4 3630 default:
fd85b817 3631 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3632 status = nfserr_resource;
3633 break;
1da177e4 3634 }
1da177e4 3635out:
8a280510 3636 if (status && lock->lk_is_new && lock_sop)
f044ff83 3637 release_lockowner(lock_sop);
3a65588a
BF		 3638 if (lock->lk_replay_owner) {
3639 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3640 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3641 }
1da177e4
LT		 3642 nfs4_unlock_state();
3643 return status;
3644}
3645
55ef1274
BF		 3646/*
3647 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3648 * so we do a temporary open here just to get an open file to pass to
3649 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3650 * inode operation.)
3651 */
3652static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3653{
3654 struct file *file;
3655 int err;
3656
3657 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3658 if (err)
3659 return err;
3660 err = vfs_test_lock(file, lock);
3661 nfsd_close(file);
3662 return err;
3663}
3664
1da177e4
LT		 3665/*
3666 * LOCKT operation
3667 */
b37ad28b 3668__be32
ca364317
BF		 3669nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3670 struct nfsd4_lockt *lockt)
1da177e4
LT		 3671{
3672 struct inode *inode;
1da177e4 3673 struct file_lock file_lock;
fd85b817 3674 int error;
b37ad28b 3675 __be32 status;
1da177e4 3676
af558e33 3677 if (locks_in_grace())
1da177e4
LT		 3678 return nfserr_grace;
3679
3680 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3681 return nfserr_inval;
3682
3683 lockt->lt_stateowner = NULL;
3684 nfs4_lock_state();
3685
3686 status = nfserr_stale_clientid;
60adfc50 3687 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3688 goto out;
1da177e4 3689
ca364317 3690 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3691 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3692 if (status == nfserr_symlink)
3693 status = nfserr_inval;
3694 goto out;
3695 }
3696
ca364317 3697 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3698 locks_init_lock(&file_lock);
3699 switch (lockt->lt_type) {
3700 case NFS4_READ_LT:
3701 case NFS4_READW_LT:
3702 file_lock.fl_type = F_RDLCK;
3703 break;
3704 case NFS4_WRITE_LT:
3705 case NFS4_WRITEW_LT:
3706 file_lock.fl_type = F_WRLCK;
3707 break;
3708 default:
2fdada03 3709 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3710 status = nfserr_inval;
3711 goto out;
3712 }
3713
3714 lockt->lt_stateowner = find_lockstateowner_str(inode,
3715 &lockt->lt_clientid, &lockt->lt_owner);
3716 if (lockt->lt_stateowner)
3717 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3718 file_lock.fl_pid = current->tgid;
3719 file_lock.fl_flags = FL_POSIX;
3720
3721 file_lock.fl_start = lockt->lt_offset;
87df4de8 3722 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3723
3724 nfs4_transform_lock_offset(&file_lock);
3725
1da177e4 3726 status = nfs_ok;
55ef1274 3727 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3728 if (error) {
3729 status = nfserrno(error);
3730 goto out;
3731 }
9d6a8c5c 3732 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3733 status = nfserr_denied;
9d6a8c5c 3734 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3735 }
3736out:
3737 nfs4_unlock_state();
3738 return status;
3739}
3740
b37ad28b 3741__be32
ca364317 3742nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3743 struct nfsd4_locku *locku)
1da177e4
LT		 3744{
3745 struct nfs4_stateid *stp;
3746 struct file *filp = NULL;
3747 struct file_lock file_lock;
b37ad28b 3748 __be32 status;
b8dd7b9a 3749 int err;
1da177e4
LT		 3750
3751 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3752 (long long) locku->lu_offset,
3753 (long long) locku->lu_length);
3754
3755 if (check_lock_length(locku->lu_offset, locku->lu_length))
3756 return nfserr_inval;
3757
3758 nfs4_lock_state();
3759
dd453dfd 3760 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3761 locku->lu_seqid,
3762 &locku->lu_stateid,
f3362737 3763 LOCK_STATE,
1da177e4
LT		 3764 &locku->lu_stateowner, &stp, NULL)))
3765 goto out;
3766
3767 filp = stp->st_vfs_file;
3768 BUG_ON(!filp);
3769 locks_init_lock(&file_lock);
3770 file_lock.fl_type = F_UNLCK;
3771 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3772 file_lock.fl_pid = current->tgid;
3773 file_lock.fl_file = filp;
3774 file_lock.fl_flags = FL_POSIX;
d5b9026a 3775 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3776 file_lock.fl_start = locku->lu_offset;
3777
87df4de8 3778 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3779 nfs4_transform_lock_offset(&file_lock);
3780
3781 /*
3782 * Try to unlock the file in the VFS.
3783 */
fd85b817 3784 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3785 if (err) {
fd85b817 3786 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3787 goto out_nfserr;
3788 }
3789 /*
3790 * OK, unlock succeeded; the only thing left to do is update the stateid.
3791 */
3792 update_stateid(&stp->st_stateid);
3793 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3794
3795out:
f2327d9a 3796 if (locku->lu_stateowner) {
1da177e4 3797 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3798 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3799 }
1da177e4
LT		 3800 nfs4_unlock_state();
3801 return status;
3802
3803out_nfserr:
b8dd7b9a 3804 status = nfserrno(err);
1da177e4
LT		 3805 goto out;
3806}
3807
3808/*
3809 * returns
3810 * 1: locks held by lockowner
3811 * 0: no locks held by lockowner
3812 */
3813static int
3814check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3815{
3816 struct file_lock **flpp;
7eaa36e2 3817 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3818 int status = 0;
3819
3820 lock_kernel();
3821 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3822 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3823 status = 1;
3824 goto out;
796dadfd 3825 }
1da177e4
LT		 3826 }
3827out:
3828 unlock_kernel();
3829 return status;
3830}
3831
b37ad28b 3832__be32
b591480b
BF		 3833nfsd4_release_lockowner(struct svc_rqst *rqstp,
3834 struct nfsd4_compound_state *cstate,
3835 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3836{
3837 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3838 struct nfs4_stateowner *sop;
3839 struct nfs4_stateid *stp;
1da177e4 3840 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3841 struct list_head matches;
3842 int i;
b37ad28b 3843 __be32 status;
1da177e4
LT		 3844
3845 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3846 clid->cl_boot, clid->cl_id);
3847
3848 /* XXX check for lease expiration */
3849
3850 status = nfserr_stale_clientid;
849823c5 3851 if (STALE_CLIENTID(clid))
1da177e4 3852 return status;
1da177e4
LT		 3853
3854 nfs4_lock_state();
3855
3e9e3dbe
N		 3856 status = nfserr_locks_held;
3857 /* XXX: we're doing a linear search through all the lockowners.
3858 * Yipes! For now we'll just hope clients aren't really using
3859 * release_lockowner much, but eventually we have to fix these
3860 * data structures. */
3861 INIT_LIST_HEAD(&matches);
3862 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3863 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3864 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3865 continue;
3866 list_for_each_entry(stp, &sop->so_stateids,
3867 st_perstateowner) {
3868 if (check_for_locks(stp->st_vfs_file, sop))
3869 goto out;
3870 /* Note: so_perclient unused for lockowners,
3871 * so it's OK to fool with here. */
3872 list_add(&sop->so_perclient, &matches);
3873 }
1da177e4 3874 }
3e9e3dbe
N		 3875 }
3876 /* Clients probably won't expect us to return with some (but not all)
3877 * of the lockowner state released; so don't release any until all
3878 * have been checked. */
3879 status = nfs_ok;
0fa822e4
N		 3880 while (!list_empty(&matches)) {
3881 sop = list_entry(matches.next, struct nfs4_stateowner,
3882 so_perclient);
3883 /* unhash_stateowner deletes so_perclient only
3884 * for openowners. */
3885 list_del(&sop->so_perclient);
f044ff83 3886 release_lockowner(sop);
1da177e4
LT		 3887 }
3888out:
3889 nfs4_unlock_state();
3890 return status;
3891}
3892
3893static inline struct nfs4_client_reclaim *
a55370a3 3894alloc_reclaim(void)
1da177e4 3895{
a55370a3 3896 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3897}
3898
c7b9a459 3899int
a1bcecd2 3900nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3901{
3902 unsigned int strhashval = clientstr_hashval(name);
3903 struct nfs4_client *clp;
3904
a1bcecd2 3905 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3906 return clp ? 1 : 0;
3907}
3908
1da177e4
LT		 3909/*
3910 * failure => all reset bets are off, nfserr_no_grace...
3911 */
190e4fbf
N		 3912int
3913nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3914{
3915 unsigned int strhashval;
3916 struct nfs4_client_reclaim *crp = NULL;
3917
a55370a3
N		 3918 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3919 crp = alloc_reclaim();
1da177e4
LT		 3920 if (!crp)
3921 return 0;
a55370a3 3922 strhashval = clientstr_hashval(name);
1da177e4
LT		 3923 INIT_LIST_HEAD(&crp->cr_strhash);
3924 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3925 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3926 reclaim_str_hashtbl_size++;
3927 return 1;
3928}
3929
3930static void
3931nfs4_release_reclaim(void)
3932{
3933 struct nfs4_client_reclaim *crp = NULL;
3934 int i;
3935
1da177e4
LT		 3936 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3937 while (!list_empty(&reclaim_str_hashtbl[i])) {
3938 crp = list_entry(reclaim_str_hashtbl[i].next,
3939 struct nfs4_client_reclaim, cr_strhash);
3940 list_del(&crp->cr_strhash);
1da177e4
LT		 3941 kfree(crp);
3942 reclaim_str_hashtbl_size--;
3943 }
3944 }
3945 BUG_ON(reclaim_str_hashtbl_size);
3946}
3947
3948/*
3949 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3950static struct nfs4_client_reclaim *
1da177e4
LT		 3951nfs4_find_reclaim_client(clientid_t *clid)
3952{
3953 unsigned int strhashval;
3954 struct nfs4_client *clp;
3955 struct nfs4_client_reclaim *crp = NULL;
3956
3957
3958 /* find clientid in conf_id_hashtbl */
3959 clp = find_confirmed_client(clid);
3960 if (clp == NULL)
3961 return NULL;
3962
a55370a3
N		 3963 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3964 clp->cl_name.len, clp->cl_name.data,
3965 clp->cl_recdir);
1da177e4
LT		 3966
3967 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3968 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3969 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3970 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3971 return crp;
3972 }
3973 }
3974 return NULL;
3975}
3976
3977/*
3978* Called from OPEN. Look for clientid in reclaim list.
3979*/
b37ad28b 3980__be32
1da177e4
LT		 3981nfs4_check_open_reclaim(clientid_t *clid)
3982{
dfc83565 3983 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3984}
3985
ac4d8ff2 3986/* initialization to perform at module load time: */
1da177e4 3987
e8ff2a84 3988int
ac4d8ff2 3989nfs4_state_init(void)
1da177e4 3990{
e8ff2a84 3991 int i, status;
1da177e4 3992
e8ff2a84
BF		 3993 status = nfsd4_init_slabs();
3994 if (status)
3995 return status;
1da177e4
LT		 3996 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3997 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3998 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3999 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
4000 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 4001 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 4002 }
5282fd72
ME		 4003 for (i = 0; i < SESSION_HASH_SIZE; i++)
4004 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4005 for (i = 0; i < FILE_HASH_SIZE; i++) {
4006 INIT_LIST_HEAD(&file_hashtbl[i]);
4007 }
4008 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4009 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4010 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4011 }
4012 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4013 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4014 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4015 }
4016 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4017 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4018 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4019 }
1da177e4 4020 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4021 INIT_LIST_HEAD(&close_lru);
4022 INIT_LIST_HEAD(&client_lru);
4023 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4024 reclaim_str_hashtbl_size = 0;
e8ff2a84 4025 return 0;
ac4d8ff2
N		 4026}
4027
190e4fbf
N		 4028static void
4029nfsd4_load_reboot_recovery_data(void)
4030{
4031 int status;
4032
0964a3d3
N		 4033 nfs4_lock_state();
4034 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4035 status = nfsd4_recdir_load();
0964a3d3 4036 nfs4_unlock_state();
190e4fbf
N		 4037 if (status)
4038 printk("NFSD: Failure reading reboot recovery data\n");
4039}
4040
9a8db97e
ME		 4041unsigned long
4042get_nfs4_grace_period(void)
4043{
4044 return max(user_lease_time, lease_time) * HZ;
4045}
4046
c2f1a551
MS		 4047/*
4048 * Since the lifetime of a delegation isn't limited to that of an open, a
4049 * client may quite reasonably hang on to a delegation as long as it has
4050 * the inode cached. This becomes an obvious problem the first time a
4051 * client's inode cache approaches the size of the server's total memory.
4052 *
4053 * For now we avoid this problem by imposing a hard limit on the number
4054 * of delegations, which varies according to the server's memory size.
4055 */
4056static void
4057set_max_delegations(void)
4058{
4059 /*
4060 * Allow at most 4 delegations per megabyte of RAM. Quick
4061 * estimates suggest that in the worst case (where every delegation
4062 * is for a different inode), a delegation could take about 1.5K,
4063 * giving a worst case usage of about 6% of memory.
4064 */
4065 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4066}
4067
ac4d8ff2
N		 4068/* initialization to perform when the nfsd service is started: */
4069
4070static void
4071__nfs4_state_start(void)
4072{
9a8db97e 4073 unsigned long grace_time;
ac4d8ff2 4074
1da177e4 4075 boot_time = get_seconds();
af558e33 4076 grace_time = get_nfs4_grace_period();
d99a05ad 4077 lease_time = user_lease_time;
af558e33 4078 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4079 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4080 grace_time/HZ);
58da282b 4081 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4082 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4083 set_max_delegations();
1da177e4
LT		 4084}
4085
e8ff2a84 4086void
76a3550e 4087nfs4_state_start(void)
1da177e4 4088{
1da177e4 4089 if (nfs4_init)
e8ff2a84 4090 return;
190e4fbf 4091 nfsd4_load_reboot_recovery_data();
76a3550e 4092 __nfs4_state_start();
1da177e4 4093 nfs4_init = 1;
e8ff2a84 4094 return;
1da177e4
LT		 4095}
4096
1da177e4
LT		 4097time_t
4098nfs4_lease_time(void)
4099{
4100 return lease_time;
4101}
4102
4103static void
4104__nfs4_state_shutdown(void)
4105{
4106 int i;
4107 struct nfs4_client *clp = NULL;
4108 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4109 struct list_head *pos, *next, reaplist;
4110
1da177e4
LT		 4111 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4112 while (!list_empty(&conf_id_hashtbl[i])) {
4113 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4114 expire_client(clp);
4115 }
4116 while (!list_empty(&unconf_str_hashtbl[i])) {
4117 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4118 expire_client(clp);
4119 }
4120 }
4121 INIT_LIST_HEAD(&reaplist);
4122 spin_lock(&recall_lock);
4123 list_for_each_safe(pos, next, &del_recall_lru) {
4124 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4125 list_move(&dp->dl_recall_lru, &reaplist);
4126 }
4127 spin_unlock(&recall_lock);
4128 list_for_each_safe(pos, next, &reaplist) {
4129 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4130 list_del_init(&dp->dl_recall_lru);
4131 unhash_delegation(dp);
4132 }
4133
190e4fbf 4134 nfsd4_shutdown_recdir();
1da177e4 4135 nfs4_init = 0;
1da177e4
LT		 4136}
4137
4138void
4139nfs4_state_shutdown(void)
4140{
5e8d5c29
N		 4141 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4142 destroy_workqueue(laundry_wq);
2c5e7615 4143 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4144 nfs4_lock_state();
4145 nfs4_release_reclaim();
4146 __nfs4_state_shutdown();
1da177e4
LT		 4147 nfs4_unlock_state();
4148}
4149
3dd98a3b
JL		 4150/*
4151 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4152 * accessed when nfsd is starting.
4153 */
0964a3d3
N		 4154static void
4155nfs4_set_recdir(char *recdir)
4156{
0964a3d3 4157 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4158}
4159
4160/*
4161 * Change the NFSv4 recovery directory to recdir.
4162 */
4163int
4164nfs4_reset_recoverydir(char *recdir)
4165{
4166 int status;
a63bb996 4167 struct path path;
0964a3d3 4168
a63bb996 4169 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4170 if (status)
4171 return status;
4172 status = -ENOTDIR;
a63bb996 4173 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4174 nfs4_set_recdir(recdir);
4175 status = 0;
4176 }
a63bb996 4177 path_put(&path);
0964a3d3
N		 4178 return status;
4179}
4180
3dd98a3b
JL		 4181char *
4182nfs4_recoverydir(void)
4183{
4184 return user_recovery_dirname;
4185}
4186
1da177e4
LT		 4187/*
4188 * Called when leasetime is changed.
4189 *
d99a05ad
N		 4190 * The only way the protocol gives us to handle on-the-fly lease changes is to
4191 * simulate a reboot. Instead of doing that, we just wait till the next time
4192 * we start to register any changes in lease time. If the administrator
4193 * really wants to change the lease time *now*, they can go ahead and bring
4194 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4195 *
4196 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4197 * when nfsd is starting
1da177e4
LT		 4198 */
4199void
4200nfs4_reset_lease(time_t leasetime)
4201{
d99a05ad 4202 user_lease_time = leasetime;
1da177e4 4203}
Linux 6.x block layer and io_uring tree(s)