]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/nfsd/nfs4state.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd: add support for NFSv4 callbacks over IPv6
[mirror_ubuntu-artful-kernel.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT		 1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
14* 1. Redistributions of source code must retain the above copyright
15* notice, this list of conditions and the following disclaimer.
16* 2. Redistributions in binary form must reproduce the above copyright
17* notice, this list of conditions and the following disclaimer in the
18* documentation and/or other materials provided with the distribution.
19* 3. Neither the name of the University nor the names of its
20* contributors may be used to endorse or promote products derived
21* from this software without specific prior written permission.
22*
23* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT		 45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
68e76ad0 57#include <linux/sunrpc/svcauth_gss.h>
363168b4 58#include <linux/sunrpc/clnt.h>
1da177e4
LT		 59
60#define NFSDDBG_FACILITY NFSDDBG_PROC
61
62/* Globals */
63static time_t lease_time = 90; /* default lease time */
d99a05ad 64static time_t user_lease_time = 90;
fd39ca9a 65static time_t boot_time;
1da177e4
LT		 66static u32 current_ownerid = 1;
67static u32 current_fileid = 1;
68static u32 current_delegid = 1;
69static u32 nfs4_init;
fd39ca9a
N		 70static stateid_t zerostateid; /* bits all 0 */
71static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 72static u64 current_sessionid = 1;
fd39ca9a
N		 73
74#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
75#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 76
1da177e4 77/* forward declarations */
fd39ca9a 78static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 79static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 80static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
81static void nfs4_set_recdir(char *recdir);
1da177e4 82
8b671b80
BF		 83/* Locking: */
84
85/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 86static DEFINE_MUTEX(client_mutex);
1da177e4 87
8b671b80
BF		 88/*
89 * Currently used for the del_recall_lru and file hash table. In an
90 * effort to decrease the scope of the client_mutex, this spinlock may
91 * eventually cover more:
92 */
93static DEFINE_SPINLOCK(recall_lock);
94
e18b890b
CL		 95static struct kmem_cache *stateowner_slab = NULL;
96static struct kmem_cache *file_slab = NULL;
97static struct kmem_cache *stateid_slab = NULL;
98static struct kmem_cache *deleg_slab = NULL;
e60d4398 99
1da177e4
LT		 100void
101nfs4_lock_state(void)
102{
353ab6e9 103 mutex_lock(&client_mutex);
1da177e4
LT		 104}
105
106void
107nfs4_unlock_state(void)
108{
353ab6e9 109 mutex_unlock(&client_mutex);
1da177e4
LT		 110}
111
112static inline u32
113opaque_hashval(const void *ptr, int nbytes)
114{
115 unsigned char *cptr = (unsigned char *) ptr;
116
117 u32 x = 0;
118 while (nbytes--) {
119 x *= 37;
120 x += *cptr++;
121 }
122 return x;
123}
124
1da177e4
LT		 125static struct list_head del_recall_lru;
126
13cd2184
N		 127static inline void
128put_nfs4_file(struct nfs4_file *fi)
129{
8b671b80
BF		 130 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
131 list_del(&fi->fi_hash);
132 spin_unlock(&recall_lock);
133 iput(fi->fi_inode);
134 kmem_cache_free(file_slab, fi);
135 }
13cd2184
N		 136}
137
138static inline void
139get_nfs4_file(struct nfs4_file *fi)
140{
8b671b80 141 atomic_inc(&fi->fi_ref);
13cd2184
N		 142}
143
ef0f3390 144static int num_delegations;
c2f1a551 145unsigned int max_delegations;
ef0f3390
N		 146
147/*
148 * Open owner state (share locks)
149 */
150
151/* hash tables for nfs4_stateowner */
152#define OWNER_HASH_BITS 8
153#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
154#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
155
156#define ownerid_hashval(id) \
157 ((id) & OWNER_HASH_MASK)
158#define ownerstr_hashval(clientid, ownername) \
159 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
160
161static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
162static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
163
164/* hash table for nfs4_file */
165#define FILE_HASH_BITS 8
166#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
167#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
168/* hash table for (open)nfs4_stateid */
169#define STATEID_HASH_BITS 10
170#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
171#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
172
173#define file_hashval(x) \
174 hash_ptr(x, FILE_HASH_BITS)
175#define stateid_hashval(owner_id, file_id) \
176 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
177
178static struct list_head file_hashtbl[FILE_HASH_SIZE];
179static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
180
1da177e4
LT		 181static struct nfs4_delegation *
182alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
183{
184 struct nfs4_delegation *dp;
185 struct nfs4_file *fp = stp->st_file;
c237dc03 186 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 187
188 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 189 if (fp->fi_had_conflict)
190 return NULL;
c2f1a551 191 if (num_delegations > max_delegations)
ef0f3390 192 return NULL;
5b2d21c1
N		 193 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
194 if (dp == NULL)
1da177e4 195 return dp;
ef0f3390 196 num_delegations++;
ea1da636
N		 197 INIT_LIST_HEAD(&dp->dl_perfile);
198 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 199 INIT_LIST_HEAD(&dp->dl_recall_lru);
200 dp->dl_client = clp;
13cd2184 201 get_nfs4_file(fp);
1da177e4
LT		 202 dp->dl_file = fp;
203 dp->dl_flock = NULL;
204 get_file(stp->st_vfs_file);
205 dp->dl_vfs_file = stp->st_vfs_file;
206 dp->dl_type = type;
b53d40c5 207 dp->dl_ident = cb->cb_ident;
78155ed7 208 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 209 dp->dl_stateid.si_stateownerid = current_delegid++;
210 dp->dl_stateid.si_fileid = 0;
211 dp->dl_stateid.si_generation = 0;
6c02eaa1 212 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 213 dp->dl_time = 0;
214 atomic_set(&dp->dl_count, 1);
ea1da636
N		 215 list_add(&dp->dl_perfile, &fp->fi_delegations);
216 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT		 217 return dp;
218}
219
220void
221nfs4_put_delegation(struct nfs4_delegation *dp)
222{
223 if (atomic_dec_and_test(&dp->dl_count)) {
224 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 225 put_nfs4_file(dp->dl_file);
5b2d21c1 226 kmem_cache_free(deleg_slab, dp);
ef0f3390 227 num_delegations--;
1da177e4
LT		 228 }
229}
230
231/* Remove the associated file_lock first, then remove the delegation.
232 * lease_modify() is called to remove the FS_LEASE file_lock from
233 * the i_flock list, eventually calling nfsd's lock_manager
234 * fl_release_callback.
235 */
236static void
237nfs4_close_delegation(struct nfs4_delegation *dp)
238{
239 struct file *filp = dp->dl_vfs_file;
240
241 dprintk("NFSD: close_delegation dp %p\n",dp);
242 dp->dl_vfs_file = NULL;
243 /* The following nfsd_close may not actually close the file,
244 * but we want to remove the lease in any case. */
c907132d 245 if (dp->dl_flock)
a9933cea 246 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 247 nfsd_close(filp);
1da177e4
LT		 248}
249
250/* Called under the state lock. */
251static void
252unhash_delegation(struct nfs4_delegation *dp)
253{
ea1da636
N		 254 list_del_init(&dp->dl_perfile);
255 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 256 spin_lock(&recall_lock);
257 list_del_init(&dp->dl_recall_lru);
258 spin_unlock(&recall_lock);
259 nfs4_close_delegation(dp);
260 nfs4_put_delegation(dp);
261}
262
263/*
264 * SETCLIENTID state
265 */
266
267/* Hash tables for nfs4_clientid state */
268#define CLIENT_HASH_BITS 4
269#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
270#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
271
272#define clientid_hashval(id) \
273 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 274#define clientstr_hashval(name) \
275 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 276/*
277 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
278 * used in reboot/reset lease grace period processing
279 *
280 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
281 * setclientid_confirmed info.
282 *
283 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
284 * setclientid info.
285 *
286 * client_lru holds client queue ordered by nfs4_client.cl_time
287 * for lease renewal.
288 *
289 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
290 * for last close replay.
291 */
292static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
293static int reclaim_str_hashtbl_size = 0;
294static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
295static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
296static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
297static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
298static struct list_head client_lru;
299static struct list_head close_lru;
300
2283963f
BF		 301static void unhash_generic_stateid(struct nfs4_stateid *stp)
302{
303 list_del(&stp->st_hash);
304 list_del(&stp->st_perfile);
305 list_del(&stp->st_perstateowner);
306}
307
308static void free_generic_stateid(struct nfs4_stateid *stp)
309{
310 put_nfs4_file(stp->st_file);
311 kmem_cache_free(stateid_slab, stp);
312}
313
314static void release_lock_stateid(struct nfs4_stateid *stp)
315{
316 unhash_generic_stateid(stp);
317 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
318 free_generic_stateid(stp);
319}
320
f044ff83
BF		 321static void unhash_lockowner(struct nfs4_stateowner *sop)
322{
323 struct nfs4_stateid *stp;
324
325 list_del(&sop->so_idhash);
326 list_del(&sop->so_strhash);
327 list_del(&sop->so_perstateid);
328 while (!list_empty(&sop->so_stateids)) {
329 stp = list_first_entry(&sop->so_stateids,
330 struct nfs4_stateid, st_perstateowner);
331 release_lock_stateid(stp);
332 }
333}
334
335static void release_lockowner(struct nfs4_stateowner *sop)
336{
337 unhash_lockowner(sop);
338 nfs4_put_stateowner(sop);
339}
340
f1d110ca
BF		 341static void
342release_stateid_lockowners(struct nfs4_stateid *open_stp)
343{
344 struct nfs4_stateowner *lock_sop;
345
346 while (!list_empty(&open_stp->st_lockowners)) {
347 lock_sop = list_entry(open_stp->st_lockowners.next,
348 struct nfs4_stateowner, so_perstateid);
349 /* list_del(&open_stp->st_lockowners); */
350 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 351 release_lockowner(lock_sop);
f1d110ca
BF		 352 }
353}
354
2283963f
BF		 355static void release_open_stateid(struct nfs4_stateid *stp)
356{
357 unhash_generic_stateid(stp);
358 release_stateid_lockowners(stp);
359 nfsd_close(stp->st_vfs_file);
360 free_generic_stateid(stp);
361}
362
f044ff83 363static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 364{
365 struct nfs4_stateid *stp;
366
367 list_del(&sop->so_idhash);
368 list_del(&sop->so_strhash);
f044ff83
BF		 369 list_del(&sop->so_perclient);
370 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 371 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 372 stp = list_first_entry(&sop->so_stateids,
373 struct nfs4_stateid, st_perstateowner);
374 release_open_stateid(stp);
f1d110ca
BF		 375 }
376}
377
f044ff83 378static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 379{
f044ff83 380 unhash_openowner(sop);
f1d110ca
BF		 381 list_del(&sop->so_close_lru);
382 nfs4_put_stateowner(sop);
383}
384
5282fd72
ME		 385static DEFINE_SPINLOCK(sessionid_lock);
386#define SESSION_HASH_SIZE 512
387static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
388
389static inline int
390hash_sessionid(struct nfs4_sessionid *sessionid)
391{
392 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
393
394 return sid->sequence % SESSION_HASH_SIZE;
395}
396
397static inline void
398dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
399{
400 u32 *ptr = (u32 *)(&sessionid->data[0]);
401 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
402}
403
ec6b5d7b
AA		 404static void
405gen_sessionid(struct nfsd4_session *ses)
406{
407 struct nfs4_client *clp = ses->se_client;
408 struct nfsd4_sessionid *sid;
409
410 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
411 sid->clientid = clp->cl_clientid;
412 sid->sequence = current_sessionid++;
413 sid->reserved = 0;
414}
415
416/*
417 * Give the client the number of slots it requests bound by
0c193054 418 * NFSD_MAX_SLOTS_PER_SESSION and by nfsd_drc_max_mem.
ec6b5d7b 419 *
0c193054
AA		 420 * If we run out of reserved DRC memory we should (up to a point) re-negotiate
421 * active sessions and reduce their slot usage to make rooom for new
422 * connections. For now we just fail the create session.
ec6b5d7b
AA		 423 */
424static int set_forechannel_maxreqs(struct nfsd4_channel_attrs *fchan)
425{
0c193054 426 int mem;
ec6b5d7b 427
5d77ddfb
AA		 428 if (fchan->maxreqs < 1)
429 return nfserr_inval;
430 else if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
431 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
432
0c193054 433 mem = fchan->maxreqs * NFSD_SLOT_CACHE_SIZE;
b101ebbc 434
4bd9b0f4 435 spin_lock(&nfsd_drc_lock);
0c193054
AA		 436 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem)
437 mem = ((nfsd_drc_max_mem - nfsd_drc_mem_used) /
438 NFSD_SLOT_CACHE_SIZE) * NFSD_SLOT_CACHE_SIZE;
439 nfsd_drc_mem_used += mem;
4bd9b0f4 440 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 441
0c193054 442 fchan->maxreqs = mem / NFSD_SLOT_CACHE_SIZE;
b101ebbc
AA		 443 if (fchan->maxreqs == 0)
444 return nfserr_resource;
445 return 0;
ec6b5d7b
AA		 446}
447
448/*
449 * fchan holds the client values on input, and the server values on output
450 */
451static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 452 struct nfsd4_channel_attrs *session_fchan,
453 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 454{
455 int status = 0;
456 __u32 maxcount = svc_max_payload(rqstp);
457
458 /* headerpadsz set to zero in encode routine */
459
460 /* Use the client's max request and max response size if possible */
461 if (fchan->maxreq_sz > maxcount)
462 fchan->maxreq_sz = maxcount;
6c18ba9f 463 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 464
465 if (fchan->maxresp_sz > maxcount)
466 fchan->maxresp_sz = maxcount;
6c18ba9f 467 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 468
0c193054 469 session_fchan->maxresp_cached = NFSD_SLOT_CACHE_SIZE;
6c18ba9f 470 fchan->maxresp_cached = session_fchan->maxresp_cached;
ec6b5d7b
AA		 471
472 /* Use the client's maxops if possible */
473 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
474 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 475 session_fchan->maxops = fchan->maxops;
ec6b5d7b 476
ec6b5d7b
AA		 477 /* FIXME: Error means no more DRC pages so the server should
478 * recover pages from existing sessions. For now fail session
479 * creation.
480 */
481 status = set_forechannel_maxreqs(fchan);
482
6c18ba9f 483 session_fchan->maxreqs = fchan->maxreqs;
ec6b5d7b
AA		 484 return status;
485}
486
487static int
488alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
489 struct nfsd4_create_session *cses)
490{
491 struct nfsd4_session *new, tmp;
492 int idx, status = nfserr_resource, slotsize;
493
494 memset(&tmp, 0, sizeof(tmp));
495
496 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 497 tmp.se_bchannel = cses->back_channel;
498 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
499 &cses->fore_channel);
ec6b5d7b
AA		 500 if (status)
501 goto out;
502
503 /* allocate struct nfsd4_session and slot table in one piece */
6c18ba9f 504 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot);
ec6b5d7b
AA		 505 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
506 if (!new)
507 goto out;
508
509 memcpy(new, &tmp, sizeof(*new));
510
511 new->se_client = clp;
512 gen_sessionid(new);
513 idx = hash_sessionid(&new->se_sessionid);
514 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
515 NFS4_MAX_SESSIONID_LEN);
516
517 new->se_flags = cses->flags;
518 kref_init(&new->se_ref);
519 spin_lock(&sessionid_lock);
520 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
521 list_add(&new->se_perclnt, &clp->cl_sessions);
522 spin_unlock(&sessionid_lock);
523
524 status = nfs_ok;
525out:
526 return status;
527}
528
5282fd72
ME		 529/* caller must hold sessionid_lock */
530static struct nfsd4_session *
531find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
532{
533 struct nfsd4_session *elem;
534 int idx;
535
536 dump_sessionid(__func__, sessionid);
537 idx = hash_sessionid(sessionid);
538 dprintk("%s: idx is %d\n", __func__, idx);
539 /* Search in the appropriate list */
540 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
541 dump_sessionid("list traversal", &elem->se_sessionid);
542 if (!memcmp(elem->se_sessionid.data, sessionid->data,
543 NFS4_MAX_SESSIONID_LEN)) {
544 return elem;
545 }
546 }
547
548 dprintk("%s: session not found\n", __func__);
549 return NULL;
550}
551
552/* caller must hold sessionid_lock */
7116ed6b 553static void
5282fd72 554unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 555{
556 list_del(&ses->se_hash);
557 list_del(&ses->se_perclnt);
5282fd72
ME		 558}
559
560static void
561release_session(struct nfsd4_session *ses)
562{
563 spin_lock(&sessionid_lock);
564 unhash_session(ses);
565 spin_unlock(&sessionid_lock);
7116ed6b
AA		 566 nfsd4_put_session(ses);
567}
568
14778a13
AA		 569static void nfsd4_release_respages(struct page **respages, short resused);
570
7116ed6b
AA		 571void
572free_session(struct kref *kref)
573{
574 struct nfsd4_session *ses;
14778a13 575 int i;
7116ed6b
AA		 576
577 ses = container_of(kref, struct nfsd4_session, se_ref);
6c18ba9f 578 for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
14778a13
AA		 579 struct nfsd4_cache_entry *e = &ses->se_slots[i].sl_cache_entry;
580 nfsd4_release_respages(e->ce_respages, e->ce_resused);
581 }
be98d1bb 582 spin_lock(&nfsd_drc_lock);
0c193054 583 nfsd_drc_mem_used -= ses->se_fchannel.maxreqs * NFSD_SLOT_CACHE_SIZE;
be98d1bb 584 spin_unlock(&nfsd_drc_lock);
7116ed6b
AA		 585 kfree(ses);
586}
587
1da177e4
LT		 588static inline void
589renew_client(struct nfs4_client *clp)
590{
591 /*
592 * Move client to the end to the LRU list.
593 */
594 dprintk("renewing client (clientid %08x/%08x)\n",
595 clp->cl_clientid.cl_boot,
596 clp->cl_clientid.cl_id);
597 list_move_tail(&clp->cl_lru, &client_lru);
598 clp->cl_time = get_seconds();
599}
600
601/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
602static int
603STALE_CLIENTID(clientid_t *clid)
604{
605 if (clid->cl_boot == boot_time)
606 return 0;
60adfc50
AA		 607 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
608 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 609 return 1;
610}
611
612/*
613 * XXX Should we use a slab cache ?
614 * This type of memory management is somewhat inefficient, but we use it
615 * anyway since SETCLIENTID is not a common operation.
616 */
35bba9a3 617static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 618{
619 struct nfs4_client *clp;
620
35bba9a3
BF		 621 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
622 if (clp == NULL)
623 return NULL;
624 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
625 if (clp->cl_name.data == NULL) {
626 kfree(clp);
627 return NULL;
1da177e4 628 }
35bba9a3
BF		 629 memcpy(clp->cl_name.data, name.data, name.len);
630 clp->cl_name.len = name.len;
1da177e4
LT		 631 return clp;
632}
633
1b1a9b31
BF		 634static void
635shutdown_callback_client(struct nfs4_client *clp)
636{
c237dc03 637 struct rpc_clnt *clnt = clp->cl_cb_conn.cb_client;
1b1a9b31 638
1b1a9b31 639 if (clnt) {
404ec117
BF		 640 /*
641 * Callback threads take a reference on the client, so there
642 * should be no outstanding callbacks at this point.
643 */
c237dc03 644 clp->cl_cb_conn.cb_client = NULL;
1b1a9b31
BF		 645 rpc_shutdown_client(clnt);
646 }
c237dc03
BF		 647 if (clp->cl_cb_conn.cb_cred) {
648 put_rpccred(clp->cl_cb_conn.cb_cred);
649 clp->cl_cb_conn.cb_cred = NULL;
3cef9ab2 650 }
1b1a9b31
BF		 651}
652
1da177e4
LT		 653static inline void
654free_client(struct nfs4_client *clp)
655{
1b1a9b31 656 shutdown_callback_client(clp);
1da177e4
LT		 657 if (clp->cl_cred.cr_group_info)
658 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 659 kfree(clp->cl_principal);
1da177e4
LT		 660 kfree(clp->cl_name.data);
661 kfree(clp);
662}
663
664void
665put_nfs4_client(struct nfs4_client *clp)
666{
667 if (atomic_dec_and_test(&clp->cl_count))
668 free_client(clp);
669}
670
671static void
672expire_client(struct nfs4_client *clp)
673{
674 struct nfs4_stateowner *sop;
675 struct nfs4_delegation *dp;
1da177e4
LT		 676 struct list_head reaplist;
677
678 dprintk("NFSD: expire_client cl_count %d\n",
679 atomic_read(&clp->cl_count));
680
1da177e4
LT		 681 INIT_LIST_HEAD(&reaplist);
682 spin_lock(&recall_lock);
ea1da636
N		 683 while (!list_empty(&clp->cl_delegations)) {
684 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 685 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
686 dp->dl_flock);
ea1da636 687 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 688 list_move(&dp->dl_recall_lru, &reaplist);
689 }
690 spin_unlock(&recall_lock);
691 while (!list_empty(&reaplist)) {
692 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
693 list_del_init(&dp->dl_recall_lru);
694 unhash_delegation(dp);
695 }
696 list_del(&clp->cl_idhash);
697 list_del(&clp->cl_strhash);
698 list_del(&clp->cl_lru);
ea1da636
N		 699 while (!list_empty(&clp->cl_openowners)) {
700 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 701 release_openowner(sop);
1da177e4 702 }
c4bf7868
ME		 703 while (!list_empty(&clp->cl_sessions)) {
704 struct nfsd4_session *ses;
705 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
706 se_perclnt);
707 release_session(ses);
708 }
1da177e4
LT		 709 put_nfs4_client(clp);
710}
711
35bba9a3
BF		 712static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
713{
1da177e4
LT		 714 struct nfs4_client *clp;
715
35bba9a3
BF		 716 clp = alloc_client(name);
717 if (clp == NULL)
718 return NULL;
a55370a3 719 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4 720 atomic_set(&clp->cl_count, 1);
c237dc03 721 atomic_set(&clp->cl_cb_conn.cb_set, 0);
1da177e4
LT		 722 INIT_LIST_HEAD(&clp->cl_idhash);
723 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N		 724 INIT_LIST_HEAD(&clp->cl_openowners);
725 INIT_LIST_HEAD(&clp->cl_delegations);
9fb87070 726 INIT_LIST_HEAD(&clp->cl_sessions);
1da177e4 727 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT		 728 return clp;
729}
730
35bba9a3
BF		 731static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
732{
733 memcpy(target->cl_verifier.data, source->data,
734 sizeof(target->cl_verifier.data));
1da177e4
LT		 735}
736
35bba9a3
BF		 737static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
738{
1da177e4
LT		 739 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
740 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
741}
742
35bba9a3
BF		 743static void copy_cred(struct svc_cred *target, struct svc_cred *source)
744{
1da177e4
LT		 745 target->cr_uid = source->cr_uid;
746 target->cr_gid = source->cr_gid;
747 target->cr_group_info = source->cr_group_info;
748 get_group_info(target->cr_group_info);
749}
750
35bba9a3 751static int same_name(const char *n1, const char *n2)
599e0a22 752{
a55370a3 753 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 754}
755
756static int
599e0a22
BF		 757same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
758{
759 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 760}
761
762static int
599e0a22
BF		 763same_clid(clientid_t *cl1, clientid_t *cl2)
764{
765 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 766}
767
768/* XXX what about NGROUP */
769static int
599e0a22
BF		 770same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
771{
772 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 773}
774
5ec7b46c
BF		 775static void gen_clid(struct nfs4_client *clp)
776{
777 static u32 current_clientid = 1;
778
1da177e4
LT		 779 clp->cl_clientid.cl_boot = boot_time;
780 clp->cl_clientid.cl_id = current_clientid++;
781}
782
deda2faa
BF		 783static void gen_confirm(struct nfs4_client *clp)
784{
785 static u32 i;
786 u32 *p;
1da177e4 787
1da177e4 788 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 789 *p++ = get_seconds();
790 *p++ = i++;
1da177e4
LT		 791}
792
35bba9a3
BF		 793static int check_name(struct xdr_netobj name)
794{
1da177e4
LT		 795 if (name.len == 0)
796 return 0;
797 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 798 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 799 return 0;
800 }
801 return 1;
802}
803
fd39ca9a 804static void
1da177e4
LT		 805add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
806{
807 unsigned int idhashval;
808
809 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
810 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
811 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
812 list_add_tail(&clp->cl_lru, &client_lru);
813 clp->cl_time = get_seconds();
814}
815
fd39ca9a 816static void
1da177e4
LT		 817move_to_confirmed(struct nfs4_client *clp)
818{
819 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
820 unsigned int strhashval;
821
822 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
823 list_del_init(&clp->cl_strhash);
f116629d 824 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 825 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 826 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
827 renew_client(clp);
828}
829
830static struct nfs4_client *
831find_confirmed_client(clientid_t *clid)
832{
833 struct nfs4_client *clp;
834 unsigned int idhashval = clientid_hashval(clid->cl_id);
835
836 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 837 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 838 return clp;
839 }
840 return NULL;
841}
842
843static struct nfs4_client *
844find_unconfirmed_client(clientid_t *clid)
845{
846 struct nfs4_client *clp;
847 unsigned int idhashval = clientid_hashval(clid->cl_id);
848
849 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 850 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 851 return clp;
852 }
853 return NULL;
854}
855
a1bcecd2
AA		 856/*
857 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
858 * parameter. Matching is based on the fact the at least one of the
859 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
860 *
861 * FIXME: we need to unify the clientid namespaces for nfsv4.x
862 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
863 * and SET_CLIENTID{,_CONFIRM}
864 */
865static inline int
866match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
867{
868 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
869 return use_exchange_id == has_exchange_flags;
870}
871
28ce6054 872static struct nfs4_client *
a1bcecd2
AA		 873find_confirmed_client_by_str(const char *dname, unsigned int hashval,
874 bool use_exchange_id)
28ce6054
N		 875{
876 struct nfs4_client *clp;
877
878 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 879 if (same_name(clp->cl_recdir, dname) &&
880 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 881 return clp;
882 }
883 return NULL;
884}
885
886static struct nfs4_client *
a1bcecd2
AA		 887find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
888 bool use_exchange_id)
28ce6054
N		 889{
890 struct nfs4_client *clp;
891
892 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 893 if (same_name(clp->cl_recdir, dname) &&
894 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 895 return clp;
896 }
897 return NULL;
898}
899
fd39ca9a 900static void
1da177e4
LT		 901gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
902{
c237dc03 903 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
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
ab52ae6d 924 cb->cb_minorversion = 0;
1da177e4
LT		 925 cb->cb_prog = se->se_callback_prog;
926 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 927 return;
928out_err:
aa9a4ec7
JL		 929 cb->cb_addr.ss_family = AF_UNSPEC;
930 cb->cb_addrlen = 0;
849823c5 931 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 932 "will not receive delegations\n",
933 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
934
1da177e4
LT		 935 return;
936}
937
074fe897
AA		 938void
939nfsd4_set_statp(struct svc_rqst *rqstp, __be32 *statp)
940{
941 struct nfsd4_compoundres *resp = rqstp->rq_resp;
942
943 resp->cstate.statp = statp;
944}
945
946/*
947 * Dereference the result pages.
948 */
949static void
950nfsd4_release_respages(struct page **respages, short resused)
951{
952 int i;
953
954 dprintk("--> %s\n", __func__);
955 for (i = 0; i < resused; i++) {
956 if (!respages[i])
957 continue;
958 put_page(respages[i]);
959 respages[i] = NULL;
960 }
961}
962
963static void
964nfsd4_copy_pages(struct page **topages, struct page **frompages, short count)
965{
966 int i;
967
968 for (i = 0; i < count; i++) {
969 topages[i] = frompages[i];
970 if (!topages[i])
971 continue;
972 get_page(topages[i]);
973 }
974}
975
976/*
977 * Cache the reply pages up to NFSD_PAGES_PER_SLOT + 1, clearing the previous
978 * pages. We add a page to NFSD_PAGES_PER_SLOT for the case where the total
979 * length of the XDR response is less than se_fmaxresp_cached
980 * (NFSD_PAGES_PER_SLOT * PAGE_SIZE) but the xdr_buf pages is used for a
981 * of the reply (e.g. readdir).
982 *
983 * Store the base and length of the rq_req.head[0] page
984 * of the NFSv4.1 data, just past the rpc header.
985 */
986void
987nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
988{
989 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
990 struct svc_rqst *rqstp = resp->rqstp;
991 struct nfsd4_compoundargs *args = rqstp->rq_argp;
992 struct nfsd4_op *op = &args->ops[resp->opcnt];
993 struct kvec *resv = &rqstp->rq_res.head[0];
994
995 dprintk("--> %s entry %p\n", __func__, entry);
996
997 /* Don't cache a failed OP_SEQUENCE. */
998 if (resp->opcnt == 1 && op->opnum == OP_SEQUENCE && resp->cstate.status)
999 return;
bf864a31 1000
074fe897 1001 nfsd4_release_respages(entry->ce_respages, entry->ce_resused);
bf864a31
AA		 1002 entry->ce_opcnt = resp->opcnt;
1003 entry->ce_status = resp->cstate.status;
1004
1005 /*
1006 * Don't need a page to cache just the sequence operation - the slot
1007 * does this for us!
1008 */
1009
1010 if (nfsd4_not_cached(resp)) {
1011 entry->ce_resused = 0;
1012 entry->ce_rpchdrlen = 0;
1013 dprintk("%s Just cache SEQUENCE. ce_cachethis %d\n", __func__,
1014 resp->cstate.slot->sl_cache_entry.ce_cachethis);
1015 return;
1016 }
074fe897
AA		 1017 entry->ce_resused = rqstp->rq_resused;
1018 if (entry->ce_resused > NFSD_PAGES_PER_SLOT + 1)
1019 entry->ce_resused = NFSD_PAGES_PER_SLOT + 1;
1020 nfsd4_copy_pages(entry->ce_respages, rqstp->rq_respages,
1021 entry->ce_resused);
074fe897
AA		 1022 entry->ce_datav.iov_base = resp->cstate.statp;
1023 entry->ce_datav.iov_len = resv->iov_len - ((char *)resp->cstate.statp -
1024 (char *)page_address(rqstp->rq_respages[0]));
074fe897
AA		 1025 /* Current request rpc header length*/
1026 entry->ce_rpchdrlen = (char *)resp->cstate.statp -
1027 (char *)page_address(rqstp->rq_respages[0]);
1028}
1029
1030/*
1031 * We keep the rpc header, but take the nfs reply from the replycache.
1032 */
1033static int
1034nfsd41_copy_replay_data(struct nfsd4_compoundres *resp,
1035 struct nfsd4_cache_entry *entry)
1036{
1037 struct svc_rqst *rqstp = resp->rqstp;
1038 struct kvec *resv = &resp->rqstp->rq_res.head[0];
1039 int len;
1040
1041 /* Current request rpc header length*/
1042 len = (char *)resp->cstate.statp -
1043 (char *)page_address(rqstp->rq_respages[0]);
1044 if (entry->ce_datav.iov_len + len > PAGE_SIZE) {
1045 dprintk("%s v41 cached reply too large (%Zd).\n", __func__,
1046 entry->ce_datav.iov_len);
1047 return 0;
1048 }
1049 /* copy the cached reply nfsd data past the current rpc header */
1050 memcpy((char *)resv->iov_base + len, entry->ce_datav.iov_base,
1051 entry->ce_datav.iov_len);
1052 resv->iov_len = len + entry->ce_datav.iov_len;
1053 return 1;
1054}
1055
abfabf8c
AA		 1056/*
1057 * Encode the replay sequence operation from the slot values.
1058 * If cachethis is FALSE encode the uncached rep error on the next
1059 * operation which sets resp->p and increments resp->opcnt for
1060 * nfs4svc_encode_compoundres.
1061 *
1062 */
1063static __be32
1064nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1065 struct nfsd4_compoundres *resp)
1066{
1067 struct nfsd4_op *op;
1068 struct nfsd4_slot *slot = resp->cstate.slot;
1069
1070 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
1071 resp->opcnt, resp->cstate.slot->sl_cache_entry.ce_cachethis);
1072
1073 /* Encode the replayed sequence operation */
1074 op = &args->ops[resp->opcnt - 1];
1075 nfsd4_encode_operation(resp, op);
1076
1077 /* Return nfserr_retry_uncached_rep in next operation. */
1078 if (args->opcnt > 1 && slot->sl_cache_entry.ce_cachethis == 0) {
1079 op = &args->ops[resp->opcnt++];
1080 op->status = nfserr_retry_uncached_rep;
1081 nfsd4_encode_operation(resp, op);
1082 }
1083 return op->status;
1084}
1085
074fe897
AA		 1086/*
1087 * Keep the first page of the replay. Copy the NFSv4.1 data from the first
1088 * cached page. Replace any futher replay pages from the cache.
1089 */
1090__be32
bf864a31
AA		 1091nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1092 struct nfsd4_sequence *seq)
074fe897
AA		 1093{
1094 struct nfsd4_cache_entry *entry = &resp->cstate.slot->sl_cache_entry;
1095 __be32 status;
1096
1097 dprintk("--> %s entry %p\n", __func__, entry);
1098
bf864a31
AA		 1099 /*
1100 * If this is just the sequence operation, we did not keep
1101 * a page in the cache entry because we can just use the
1102 * slot info stored in struct nfsd4_sequence that was checked
1103 * against the slot in nfsd4_sequence().
1104 *
1105 * This occurs when seq->cachethis is FALSE, or when the client
1106 * session inactivity timer fires and a solo sequence operation
1107 * is sent (lease renewal).
1108 */
abfabf8c
AA		 1109 seq->maxslots = resp->cstate.session->se_fchannel.maxreqs;
1110
1111 /* Either returns 0 or nfserr_retry_uncached */
1112 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1113 if (status == nfserr_retry_uncached_rep)
1114 return status;
074fe897
AA		 1115
1116 if (!nfsd41_copy_replay_data(resp, entry)) {
1117 /*
1118 * Not enough room to use the replay rpc header, send the
1119 * cached header. Release all the allocated result pages.
1120 */
1121 svc_free_res_pages(resp->rqstp);
1122 nfsd4_copy_pages(resp->rqstp->rq_respages, entry->ce_respages,
1123 entry->ce_resused);
1124 } else {
1125 /* Release all but the first allocated result page */
1126
1127 resp->rqstp->rq_resused--;
1128 svc_free_res_pages(resp->rqstp);
1129
1130 nfsd4_copy_pages(&resp->rqstp->rq_respages[1],
1131 &entry->ce_respages[1],
1132 entry->ce_resused - 1);
1133 }
1134
1135 resp->rqstp->rq_resused = entry->ce_resused;
da3846a2
AA		 1136 resp->opcnt = entry->ce_opcnt;
1137 resp->cstate.iovlen = entry->ce_datav.iov_len + entry->ce_rpchdrlen;
074fe897
AA		 1138 status = entry->ce_status;
1139
1140 return status;
1141}
1142
0733d213
AA		 1143/*
1144 * Set the exchange_id flags returned by the server.
1145 */
1146static void
1147nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1148{
1149 /* pNFS is not supported */
1150 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1151
1152 /* Referrals are supported, Migration is not. */
1153 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1154
1155 /* set the wire flags to return to client. */
1156 clid->flags = new->cl_exchange_flags;
1157}
1158
069b6ad4
AA		 1159__be32
1160nfsd4_exchange_id(struct svc_rqst *rqstp,
1161 struct nfsd4_compound_state *cstate,
1162 struct nfsd4_exchange_id *exid)
1163{
0733d213
AA		 1164 struct nfs4_client *unconf, *conf, *new;
1165 int status;
1166 unsigned int strhashval;
1167 char dname[HEXDIR_LEN];
363168b4 1168 char addr_str[INET6_ADDRSTRLEN];
0733d213 1169 nfs4_verifier verf = exid->verifier;
363168b4 1170 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1171
363168b4 1172 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1173 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1174 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1175 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1176 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1177
1178 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1179 return nfserr_inval;
1180
1181 /* Currently only support SP4_NONE */
1182 switch (exid->spa_how) {
1183 case SP4_NONE:
1184 break;
1185 case SP4_SSV:
1186 return nfserr_encr_alg_unsupp;
1187 default:
1188 BUG(); /* checked by xdr code */
1189 case SP4_MACH_CRED:
1190 return nfserr_serverfault; /* no excuse :-/ */
1191 }
1192
1193 status = nfs4_make_rec_clidname(dname, &exid->clname);
1194
1195 if (status)
1196 goto error;
1197
1198 strhashval = clientstr_hashval(dname);
1199
1200 nfs4_lock_state();
1201 status = nfs_ok;
1202
a1bcecd2 1203 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1204 if (conf) {
1205 if (!same_verf(&verf, &conf->cl_verifier)) {
1206 /* 18.35.4 case 8 */
1207 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1208 status = nfserr_not_same;
1209 goto out;
1210 }
1211 /* Client reboot: destroy old state */
1212 expire_client(conf);
1213 goto out_new;
1214 }
1215 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1216 /* 18.35.4 case 9 */
1217 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1218 status = nfserr_perm;
1219 goto out;
1220 }
1221 expire_client(conf);
1222 goto out_new;
1223 }
0733d213
AA		 1224 /*
1225 * Set bit when the owner id and verifier map to an already
1226 * confirmed client id (18.35.3).
1227 */
1228 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1229
1230 /*
1231 * Falling into 18.35.4 case 2, possible router replay.
1232 * Leave confirmed record intact and return same result.
1233 */
1234 copy_verf(conf, &verf);
1235 new = conf;
1236 goto out_copy;
6ddbbbfe
MS		 1237 }
1238
1239 /* 18.35.4 case 7 */
1240 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1241 status = nfserr_noent;
1242 goto out;
0733d213
AA		 1243 }
1244
a1bcecd2 1245 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1246 if (unconf) {
1247 /*
1248 * Possible retry or client restart. Per 18.35.4 case 4,
1249 * a new unconfirmed record should be generated regardless
1250 * of whether any properties have changed.
1251 */
1252 expire_client(unconf);
1253 }
1254
1255out_new:
1256 /* Normal case */
1257 new = create_client(exid->clname, dname);
1258 if (new == NULL) {
1259 status = nfserr_resource;
1260 goto out;
1261 }
1262
1263 copy_verf(new, &verf);
1264 copy_cred(&new->cl_cred, &rqstp->rq_cred);
363168b4 1265 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
0733d213
AA		 1266 gen_clid(new);
1267 gen_confirm(new);
1268 add_to_unconfirmed(new, strhashval);
1269out_copy:
1270 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1271 exid->clientid.cl_id = new->cl_clientid.cl_id;
1272
38eb76a5 1273 exid->seqid = 1;
0733d213
AA		 1274 nfsd4_set_ex_flags(new, exid);
1275
1276 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1277 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1278 status = nfs_ok;
1279
1280out:
1281 nfs4_unlock_state();
1282error:
1283 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1284 return status;
069b6ad4
AA		 1285}
1286
b85d4c01 1287static int
88e588d5 1288check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1289{
88e588d5
AA		 1290 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1291 slot_seqid);
b85d4c01
BH		 1292
1293 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1294 if (slot_inuse) {
1295 if (seqid == slot_seqid)
b85d4c01
BH		 1296 return nfserr_jukebox;
1297 else
1298 return nfserr_seq_misordered;
1299 }
1300 /* Normal */
88e588d5 1301 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1302 return nfs_ok;
1303 /* Replay */
88e588d5 1304 if (seqid == slot_seqid)
b85d4c01
BH		 1305 return nfserr_replay_cache;
1306 /* Wraparound */
88e588d5 1307 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1308 return nfs_ok;
1309 /* Misordered replay or misordered new request */
1310 return nfserr_seq_misordered;
1311}
1312
49557cc7
AA		 1313/*
1314 * Cache the create session result into the create session single DRC
1315 * slot cache by saving the xdr structure. sl_seqid has been set.
1316 * Do this for solo or embedded create session operations.
1317 */
1318static void
1319nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1320 struct nfsd4_clid_slot *slot, int nfserr)
1321{
1322 slot->sl_status = nfserr;
1323 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1324}
1325
1326static __be32
1327nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1328 struct nfsd4_clid_slot *slot)
1329{
1330 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1331 return slot->sl_status;
1332}
1333
069b6ad4
AA		 1334__be32
1335nfsd4_create_session(struct svc_rqst *rqstp,
1336 struct nfsd4_compound_state *cstate,
1337 struct nfsd4_create_session *cr_ses)
1338{
363168b4 1339 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1340 struct nfs4_client *conf, *unconf;
49557cc7 1341 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1342 int status = 0;
1343
1344 nfs4_lock_state();
1345 unconf = find_unconfirmed_client(&cr_ses->clientid);
1346 conf = find_confirmed_client(&cr_ses->clientid);
1347
1348 if (conf) {
49557cc7
AA		 1349 cs_slot = &conf->cl_cs_slot;
1350 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1351 if (status == nfserr_replay_cache) {
ec6b5d7b 1352 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1353 cs_slot->sl_seqid);
38eb76a5 1354 /* Return the cached reply status */
49557cc7 1355 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1356 goto out;
49557cc7 1357 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1358 status = nfserr_seq_misordered;
1359 dprintk("Sequence misordered!\n");
1360 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1361 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1362 goto out;
1363 }
49557cc7 1364 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1365 } else if (unconf) {
1366 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1367 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1368 status = nfserr_clid_inuse;
1369 goto out;
1370 }
1371
49557cc7
AA		 1372 cs_slot = &unconf->cl_cs_slot;
1373 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1374 if (status) {
1375 /* an unconfirmed replay returns misordered */
ec6b5d7b 1376 status = nfserr_seq_misordered;
49557cc7 1377 goto out_cache;
ec6b5d7b
AA		 1378 }
1379
49557cc7 1380 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1381 move_to_confirmed(unconf);
1382
1383 /*
1384 * We do not support RDMA or persistent sessions
1385 */
1386 cr_ses->flags &= ~SESSION4_PERSIST;
1387 cr_ses->flags &= ~SESSION4_RDMA;
1388
1389 conf = unconf;
1390 } else {
1391 status = nfserr_stale_clientid;
1392 goto out;
1393 }
1394
1395 status = alloc_init_session(rqstp, conf, cr_ses);
1396 if (status)
1397 goto out;
1398
ec6b5d7b
AA		 1399 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1400 NFS4_MAX_SESSIONID_LEN);
49557cc7 1401 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1402
49557cc7
AA		 1403out_cache:
1404 /* cache solo and embedded create sessions under the state lock */
1405 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1406out:
1407 nfs4_unlock_state();
1408 dprintk("%s returns %d\n", __func__, ntohl(status));
1409 return status;
069b6ad4
AA		 1410}
1411
1412__be32
1413nfsd4_destroy_session(struct svc_rqst *r,
1414 struct nfsd4_compound_state *cstate,
1415 struct nfsd4_destroy_session *sessionid)
1416{
e10e0cfc
BH		 1417 struct nfsd4_session *ses;
1418 u32 status = nfserr_badsession;
1419
1420 /* Notes:
1421 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1422 * - Should we return nfserr_back_chan_busy if waiting for
1423 * callbacks on to-be-destroyed session?
1424 * - Do we need to clear any callback info from previous session?
1425 */
1426
1427 dump_sessionid(__func__, &sessionid->sessionid);
1428 spin_lock(&sessionid_lock);
1429 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1430 if (!ses) {
1431 spin_unlock(&sessionid_lock);
1432 goto out;
1433 }
1434
1435 unhash_session(ses);
1436 spin_unlock(&sessionid_lock);
1437
1438 /* wait for callbacks */
1439 shutdown_callback_client(ses->se_client);
1440 nfsd4_put_session(ses);
1441 status = nfs_ok;
1442out:
1443 dprintk("%s returns %d\n", __func__, ntohl(status));
1444 return status;
069b6ad4
AA		 1445}
1446
1447__be32
b85d4c01 1448nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1449 struct nfsd4_compound_state *cstate,
1450 struct nfsd4_sequence *seq)
1451{
f9bb94c4 1452 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1453 struct nfsd4_session *session;
1454 struct nfsd4_slot *slot;
1455 int status;
1456
f9bb94c4
AA		 1457 if (resp->opcnt != 1)
1458 return nfserr_sequence_pos;
1459
b85d4c01
BH		 1460 spin_lock(&sessionid_lock);
1461 status = nfserr_badsession;
1462 session = find_in_sessionid_hashtbl(&seq->sessionid);
1463 if (!session)
1464 goto out;
1465
1466 status = nfserr_badslot;
6c18ba9f 1467 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1468 goto out;
1469
1470 slot = &session->se_slots[seq->slotid];
1471 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1472
88e588d5 1473 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1474 if (status == nfserr_replay_cache) {
1475 cstate->slot = slot;
1476 cstate->session = session;
da3846a2 1477 /* Return the cached reply status and set cstate->status
bf864a31
AA		 1478 * for nfsd4_svc_encode_compoundres processing */
1479 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1480 cstate->status = nfserr_replay_cache;
b85d4c01
BH		 1481 goto replay_cache;
1482 }
1483 if (status)
1484 goto out;
1485
1486 /* Success! bump slot seqid */
1487 slot->sl_inuse = true;
1488 slot->sl_seqid = seq->seqid;
bf864a31
AA		 1489 slot->sl_cache_entry.ce_cachethis = seq->cachethis;
1490 /* Always set the cache entry cachethis for solo sequence */
1491 if (nfsd4_is_solo_sequence(resp))
1492 slot->sl_cache_entry.ce_cachethis = 1;
b85d4c01
BH		 1493
1494 cstate->slot = slot;
1495 cstate->session = session;
1496
1497replay_cache:
1498 /* Renew the clientid on success and on replay.
1499 * Hold a session reference until done processing the compound:
1500 * nfsd4_put_session called only if the cstate slot is set.
1501 */
1502 renew_client(session->se_client);
1503 nfsd4_get_session(session);
1504out:
1505 spin_unlock(&sessionid_lock);
1506 dprintk("%s: return %d\n", __func__, ntohl(status));
1507 return status;
069b6ad4
AA		 1508}
1509
b37ad28b 1510__be32
b591480b
BF		 1511nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1512 struct nfsd4_setclientid *setclid)
1da177e4 1513{
363168b4 1514 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1515 struct xdr_netobj clname = {
1516 .len = setclid->se_namelen,
1517 .data = setclid->se_name,
1518 };
1519 nfs4_verifier clverifier = setclid->se_verf;
1520 unsigned int strhashval;
28ce6054 1521 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1522 __be32 status;
68e76ad0 1523 char *princ;
a55370a3 1524 char dname[HEXDIR_LEN];
1da177e4 1525
1da177e4 1526 if (!check_name(clname))
73aea4ec 1527 return nfserr_inval;
1da177e4 1528
a55370a3
N		 1529 status = nfs4_make_rec_clidname(dname, &clname);
1530 if (status)
73aea4ec 1531 return status;
a55370a3 1532
1da177e4
LT		 1533 /*
1534 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1535 * We get here on a DRC miss.
1536 */
1537
a55370a3 1538 strhashval = clientstr_hashval(dname);
1da177e4 1539
1da177e4 1540 nfs4_lock_state();
a1bcecd2 1541 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1542 if (conf) {
a186e767 1543 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1544 status = nfserr_clid_inuse;
026722c2 1545 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1546 char addr_str[INET6_ADDRSTRLEN];
1547 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1548 sizeof(addr_str));
1549 dprintk("NFSD: setclientid: string in use by client "
1550 "at %s\n", addr_str);
1da177e4
LT		 1551 goto out;
1552 }
1da177e4 1553 }
a186e767
BF		 1554 /*
1555 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1556 * has a description of SETCLIENTID request processing consisting
1557 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1558 */
a1bcecd2 1559 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1560 status = nfserr_resource;
1561 if (!conf) {
a186e767
BF		 1562 /*
1563 * RFC 3530 14.2.33 CASE 4:
1564 * placed first, because it is the normal case
1da177e4
LT		 1565 */
1566 if (unconf)
1567 expire_client(unconf);
a55370a3
N		 1568 new = create_client(clname, dname);
1569 if (new == NULL)
1da177e4 1570 goto out;
1da177e4 1571 gen_clid(new);
599e0a22 1572 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1573 /*
a186e767
BF		 1574 * RFC 3530 14.2.33 CASE 1:
1575 * probable callback update
1da177e4 1576 */
31f4a6c1
N		 1577 if (unconf) {
1578 /* Note this is removing unconfirmed {*x***},
1579 * which is stronger than RFC recommended {vxc**}.
1580 * This has the advantage that there is at most
1581 * one {*x***} in either list at any time.
1582 */
1583 expire_client(unconf);
1da177e4 1584 }
a55370a3
N		 1585 new = create_client(clname, dname);
1586 if (new == NULL)
1da177e4 1587 goto out;
1da177e4 1588 copy_clid(new, conf);
1da177e4
LT		 1589 } else if (!unconf) {
1590 /*
a186e767
BF		 1591 * RFC 3530 14.2.33 CASE 2:
1592 * probable client reboot; state will be removed if
1593 * confirmed.
1da177e4 1594 */
a55370a3
N		 1595 new = create_client(clname, dname);
1596 if (new == NULL)
1da177e4 1597 goto out;
1da177e4 1598 gen_clid(new);
49ba8781 1599 } else {
a186e767
BF		 1600 /*
1601 * RFC 3530 14.2.33 CASE 3:
1602 * probable client reboot; state will be removed if
1603 * confirmed.
1da177e4
LT		 1604 */
1605 expire_client(unconf);
a55370a3
N		 1606 new = create_client(clname, dname);
1607 if (new == NULL)
1da177e4 1608 goto out;
1da177e4 1609 gen_clid(new);
1da177e4 1610 }
c175b83c 1611 copy_verf(new, &clverifier);
363168b4 1612 rpc_copy_addr((struct sockaddr *) &new->cl_addr, sa);
61054b14 1613 new->cl_flavor = rqstp->rq_flavor;
68e76ad0
OK		 1614 princ = svc_gss_principal(rqstp);
1615 if (princ) {
1616 new->cl_principal = kstrdup(princ, GFP_KERNEL);
1617 if (new->cl_principal == NULL) {
1618 free_client(new);
1619 goto out;
1620 }
1621 }
c175b83c
BF		 1622 copy_cred(&new->cl_cred, &rqstp->rq_cred);
1623 gen_confirm(new);
1624 gen_callback(new, setclid);
1625 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1626 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1627 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1628 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1629 status = nfs_ok;
1630out:
1631 nfs4_unlock_state();
1632 return status;
1633}
1634
1635
1636/*
a186e767
BF		 1637 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1638 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1639 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1640 */
b37ad28b 1641__be32
b591480b
BF		 1642nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1643 struct nfsd4_compound_state *cstate,
1644 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1645{
363168b4 1646 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1647 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1648 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1649 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1650 __be32 status;
1da177e4
LT		 1651
1652 if (STALE_CLIENTID(clid))
1653 return nfserr_stale_clientid;
1654 /*
1655 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1656 * We get here on a DRC miss.
1657 */
1658
1659 nfs4_lock_state();
21ab45a4
N		 1660
1661 conf = find_confirmed_client(clid);
1662 unconf = find_unconfirmed_client(clid);
1663
1664 status = nfserr_clid_inuse;
363168b4 1665 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1666 goto out;
363168b4 1667 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1668 goto out;
1669
a186e767
BF		 1670 /*
1671 * section 14.2.34 of RFC 3530 has a description of
1672 * SETCLIENTID_CONFIRM request processing consisting
1673 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1674 */
366e0c1d 1675 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1676 /*
1677 * RFC 3530 14.2.34 CASE 1:
1678 * callback update
1679 */
599e0a22 1680 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1681 status = nfserr_clid_inuse;
1682 else {
1a69c179
N		 1683 /* XXX: We just turn off callbacks until we can handle
1684 * change request correctly. */
c237dc03 1685 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1686 expire_client(unconf);
1da177e4 1687 status = nfs_ok;
1a69c179 1688
1da177e4 1689 }
f3aba4e5 1690 } else if (conf && !unconf) {
a186e767
BF		 1691 /*
1692 * RFC 3530 14.2.34 CASE 2:
1693 * probable retransmitted request; play it safe and
1694 * do nothing.
7c79f737 1695 */
599e0a22 1696 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1697 status = nfserr_clid_inuse;
21ab45a4 1698 else
1da177e4 1699 status = nfs_ok;
7c79f737 1700 } else if (!conf && unconf
599e0a22 1701 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1702 /*
1703 * RFC 3530 14.2.34 CASE 3:
1704 * Normal case; new or rebooted client:
7c79f737 1705 */
599e0a22 1706 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1707 status = nfserr_clid_inuse;
1708 } else {
1a69c179
N		 1709 unsigned int hash =
1710 clientstr_hashval(unconf->cl_recdir);
1711 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1712 hash, false);
1a69c179 1713 if (conf) {
c7b9a459 1714 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1715 expire_client(conf);
1716 }
1da177e4 1717 move_to_confirmed(unconf);
21ab45a4 1718 conf = unconf;
46f8a64b 1719 nfsd4_probe_callback(conf);
1a69c179 1720 status = nfs_ok;
1da177e4 1721 }
599e0a22
BF		 1722 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1723 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1724 &confirm)))) {
a186e767
BF		 1725 /*
1726 * RFC 3530 14.2.34 CASE 4:
1727 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1728 */
1da177e4 1729 status = nfserr_stale_clientid;
7c79f737 1730 } else {
08e8987c
N		 1731 /* check that we have hit one of the cases...*/
1732 status = nfserr_clid_inuse;
1733 }
1da177e4 1734out:
1da177e4
LT		 1735 nfs4_unlock_state();
1736 return status;
1737}
1738
1da177e4
LT		 1739/* OPEN Share state helper functions */
1740static inline struct nfs4_file *
1741alloc_init_file(struct inode *ino)
1742{
1743 struct nfs4_file *fp;
1744 unsigned int hashval = file_hashval(ino);
1745
e60d4398
N		 1746 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1747 if (fp) {
8b671b80 1748 atomic_set(&fp->fi_ref, 1);
1da177e4 1749 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1750 INIT_LIST_HEAD(&fp->fi_stateids);
1751 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1752 spin_lock(&recall_lock);
1da177e4 1753 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1754 spin_unlock(&recall_lock);
1da177e4
LT		 1755 fp->fi_inode = igrab(ino);
1756 fp->fi_id = current_fileid++;
47f9940c 1757 fp->fi_had_conflict = false;
1da177e4
LT		 1758 return fp;
1759 }
1760 return NULL;
1761}
1762
e60d4398 1763static void
e18b890b 1764nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1765{
e60d4398
N		 1766 if (*slab == NULL)
1767 return;
1a1d92c1 1768 kmem_cache_destroy(*slab);
e60d4398 1769 *slab = NULL;
1da177e4
LT		 1770}
1771
e8ff2a84 1772void
1da177e4
LT		 1773nfsd4_free_slabs(void)
1774{
e60d4398
N		 1775 nfsd4_free_slab(&stateowner_slab);
1776 nfsd4_free_slab(&file_slab);
5ac049ac 1777 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1778 nfsd4_free_slab(&deleg_slab);
e60d4398 1779}
1da177e4 1780
e60d4398
N		 1781static int
1782nfsd4_init_slabs(void)
1783{
1784 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1785 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1786 if (stateowner_slab == NULL)
1787 goto out_nomem;
1788 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1789 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1790 if (file_slab == NULL)
1791 goto out_nomem;
5ac049ac 1792 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1793 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1794 if (stateid_slab == NULL)
1795 goto out_nomem;
5b2d21c1 1796 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1797 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1798 if (deleg_slab == NULL)
1799 goto out_nomem;
e60d4398
N		 1800 return 0;
1801out_nomem:
1802 nfsd4_free_slabs();
1803 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1804 return -ENOMEM;
1da177e4
LT		 1805}
1806
1807void
1808nfs4_free_stateowner(struct kref *kref)
1809{
1810 struct nfs4_stateowner *sop =
1811 container_of(kref, struct nfs4_stateowner, so_ref);
1812 kfree(sop->so_owner.data);
1813 kmem_cache_free(stateowner_slab, sop);
1814}
1815
1816static inline struct nfs4_stateowner *
1817alloc_stateowner(struct xdr_netobj *owner)
1818{
1819 struct nfs4_stateowner *sop;
1820
1821 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1822 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1823 memcpy(sop->so_owner.data, owner->data, owner->len);
1824 sop->so_owner.len = owner->len;
1825 kref_init(&sop->so_ref);
1826 return sop;
1827 }
1828 kmem_cache_free(stateowner_slab, sop);
1829 }
1830 return NULL;
1831}
1832
1833static struct nfs4_stateowner *
1834alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1835 struct nfs4_stateowner *sop;
1836 struct nfs4_replay *rp;
1837 unsigned int idhashval;
1838
1839 if (!(sop = alloc_stateowner(&open->op_owner)))
1840 return NULL;
1841 idhashval = ownerid_hashval(current_ownerid);
1842 INIT_LIST_HEAD(&sop->so_idhash);
1843 INIT_LIST_HEAD(&sop->so_strhash);
1844 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1845 INIT_LIST_HEAD(&sop->so_stateids);
1846 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1847 INIT_LIST_HEAD(&sop->so_close_lru);
1848 sop->so_time = 0;
1849 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1850 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1851 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1852 sop->so_is_open_owner = 1;
1853 sop->so_id = current_ownerid++;
1854 sop->so_client = clp;
1855 sop->so_seqid = open->op_seqid;
1856 sop->so_confirmed = 0;
1857 rp = &sop->so_replay;
de1ae286 1858 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1859 rp->rp_buflen = 0;
1860 rp->rp_buf = rp->rp_ibuf;
1861 return sop;
1862}
1863
1da177e4
LT		 1864static inline void
1865init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1866 struct nfs4_stateowner *sop = open->op_stateowner;
1867 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1868
1869 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1870 INIT_LIST_HEAD(&stp->st_perstateowner);
1871 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1872 INIT_LIST_HEAD(&stp->st_perfile);
1873 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1874 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1875 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1876 stp->st_stateowner = sop;
13cd2184 1877 get_nfs4_file(fp);
1da177e4 1878 stp->st_file = fp;
78155ed7 1879 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1880 stp->st_stateid.si_stateownerid = sop->so_id;
1881 stp->st_stateid.si_fileid = fp->fi_id;
1882 stp->st_stateid.si_generation = 0;
1883 stp->st_access_bmap = 0;
1884 stp->st_deny_bmap = 0;
84459a11
AA		 1885 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1886 &stp->st_access_bmap);
1da177e4 1887 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1888 stp->st_openstp = NULL;
1da177e4
LT		 1889}
1890
fd39ca9a 1891static void
1da177e4
LT		 1892move_to_close_lru(struct nfs4_stateowner *sop)
1893{
1894 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1895
358dd55a 1896 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1897 sop->so_time = get_seconds();
1898}
1899
1da177e4 1900static int
599e0a22
BF		 1901same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1902 clientid_t *clid)
1903{
1904 return (sop->so_owner.len == owner->len) &&
1905 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1906 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1907}
1908
1909static struct nfs4_stateowner *
1910find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1911{
1912 struct nfs4_stateowner *so = NULL;
1913
1914 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1915 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1916 return so;
1917 }
1918 return NULL;
1919}
1920
1921/* search file_hashtbl[] for file */
1922static struct nfs4_file *
1923find_file(struct inode *ino)
1924{
1925 unsigned int hashval = file_hashval(ino);
1926 struct nfs4_file *fp;
1927
8b671b80 1928 spin_lock(&recall_lock);
1da177e4 1929 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1930 if (fp->fi_inode == ino) {
1931 get_nfs4_file(fp);
8b671b80 1932 spin_unlock(&recall_lock);
1da177e4 1933 return fp;
13cd2184 1934 }
1da177e4 1935 }
8b671b80 1936 spin_unlock(&recall_lock);
1da177e4
LT		 1937 return NULL;
1938}
1939
d87a8ade 1940static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1941{
d87a8ade 1942 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1943 return 0;
d87a8ade
AA		 1944 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1945 return 0;
1946 x &= ~NFS4_SHARE_ACCESS_MASK;
1947 if (minorversion && x) {
1948 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1949 return 0;
1950 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1951 return 0;
1952 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1953 }
1954 if (x)
8838dc43
BF		 1955 return 0;
1956 return 1;
ba5a6a19
BF		 1957}
1958
8838dc43 1959static inline int deny_valid(u32 x)
ba5a6a19 1960{
8838dc43
BF		 1961 /* Note: unlike access bits, deny bits may be zero. */
1962 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1963}
1da177e4 1964
4f83aa30
BF		 1965/*
1966 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1967 * st_{access,deny}_bmap field of the stateid, in order to track not
1968 * only what share bits are currently in force, but also what
1969 * combinations of share bits previous opens have used. This allows us
1970 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1971 * return an error if the client attempt to downgrade to a combination
1972 * of share bits not explicable by closing some of its previous opens.
1973 *
1974 * XXX: This enforcement is actually incomplete, since we don't keep
1975 * track of access/deny bit combinations; so, e.g., we allow:
1976 *
1977 * OPEN allow read, deny write
1978 * OPEN allow both, deny none
1979 * DOWNGRADE allow read, deny none
1980 *
1981 * which we should reject.
1982 */
fd39ca9a 1983static void
1da177e4
LT		 1984set_access(unsigned int *access, unsigned long bmap) {
1985 int i;
1986
1987 *access = 0;
1988 for (i = 1; i < 4; i++) {
1989 if (test_bit(i, &bmap))
1990 *access |= i;
1991 }
1992}
1993
fd39ca9a 1994static void
1da177e4
LT		 1995set_deny(unsigned int *deny, unsigned long bmap) {
1996 int i;
1997
1998 *deny = 0;
1999 for (i = 0; i < 4; i++) {
2000 if (test_bit(i, &bmap))
2001 *deny |= i ;
2002 }
2003}
2004
2005static int
2006test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
2007 unsigned int access, deny;
2008
2009 set_access(&access, stp->st_access_bmap);
2010 set_deny(&deny, stp->st_deny_bmap);
2011 if ((access & open->op_share_deny) || (deny & open->op_share_access))
2012 return 0;
2013 return 1;
2014}
2015
2016/*
2017 * Called to check deny when READ with all zero stateid or
2018 * WRITE with all zero or all one stateid
2019 */
b37ad28b 2020static __be32
1da177e4
LT		 2021nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
2022{
2023 struct inode *ino = current_fh->fh_dentry->d_inode;
2024 struct nfs4_file *fp;
2025 struct nfs4_stateid *stp;
b37ad28b 2026 __be32 ret;
1da177e4
LT		 2027
2028 dprintk("NFSD: nfs4_share_conflict\n");
2029
2030 fp = find_file(ino);
13cd2184
N		 2031 if (!fp)
2032 return nfs_ok;
b700949b 2033 ret = nfserr_locked;
1da177e4 2034 /* Search for conflicting share reservations */
13cd2184
N		 2035 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
2036 if (test_bit(deny_type, &stp->st_deny_bmap) ||
2037 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
2038 goto out;
1da177e4 2039 }
13cd2184
N		 2040 ret = nfs_ok;
2041out:
2042 put_nfs4_file(fp);
2043 return ret;
1da177e4
LT		 2044}
2045
2046static inline void
2047nfs4_file_downgrade(struct file *filp, unsigned int share_access)
2048{
2049 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 2050 drop_file_write_access(filp);
1da177e4
LT		 2051 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
2052 }
2053}
2054
1da177e4
LT		 2055/*
2056 * Spawn a thread to perform a recall on the delegation represented
2057 * by the lease (file_lock)
2058 *
2059 * Called from break_lease() with lock_kernel() held.
2060 * Note: we assume break_lease will only call this *once* for any given
2061 * lease.
2062 */
2063static
2064void nfsd_break_deleg_cb(struct file_lock *fl)
2065{
63e4863f 2066 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2067
2068 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2069 if (!dp)
2070 return;
2071
2072 /* We're assuming the state code never drops its reference
2073 * without first removing the lease. Since we're in this lease
2074 * callback (and since the lease code is serialized by the kernel
2075 * lock) we know the server hasn't removed the lease yet, we know
2076 * it's safe to take a reference: */
2077 atomic_inc(&dp->dl_count);
cfdcad4d 2078 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT		 2079
2080 spin_lock(&recall_lock);
2081 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2082 spin_unlock(&recall_lock);
2083
2084 /* only place dl_time is set. protected by lock_kernel*/
2085 dp->dl_time = get_seconds();
2086
0272e1fd
BF		 2087 /*
2088 * We don't want the locks code to timeout the lease for us;
2089 * we'll remove it ourself if the delegation isn't returned
2090 * in time.
2091 */
2092 fl->fl_break_time = 0;
1da177e4 2093
63e4863f
BF		 2094 dp->dl_file->fi_had_conflict = true;
2095 nfsd4_cb_recall(dp);
1da177e4
LT		 2096}
2097
2098/*
2099 * The file_lock is being reapd.
2100 *
2101 * Called by locks_free_lock() with lock_kernel() held.
2102 */
2103static
2104void nfsd_release_deleg_cb(struct file_lock *fl)
2105{
2106 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2107
2108 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2109
2110 if (!(fl->fl_flags & FL_LEASE) || !dp)
2111 return;
2112 dp->dl_flock = NULL;
2113}
2114
2115/*
2116 * Set the delegation file_lock back pointer.
2117 *
a9933cea 2118 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2119 */
2120static
2121void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2122{
2123 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2124
2125 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2126 if (!dp)
2127 return;
2128 dp->dl_flock = new;
2129}
2130
2131/*
a9933cea 2132 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2133 */
2134static
2135int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2136{
2137 struct nfs4_delegation *onlistd =
2138 (struct nfs4_delegation *)onlist->fl_owner;
2139 struct nfs4_delegation *tryd =
2140 (struct nfs4_delegation *)try->fl_owner;
2141
2142 if (onlist->fl_lmops != try->fl_lmops)
2143 return 0;
2144
2145 return onlistd->dl_client == tryd->dl_client;
2146}
2147
2148
2149static
2150int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2151{
2152 if (arg & F_UNLCK)
2153 return lease_modify(onlist, arg);
2154 else
2155 return -EAGAIN;
2156}
2157
fd39ca9a 2158static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2159 .fl_break = nfsd_break_deleg_cb,
2160 .fl_release_private = nfsd_release_deleg_cb,
2161 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2162 .fl_mylease = nfsd_same_client_deleg_cb,
2163 .fl_change = nfsd_change_deleg_cb,
2164};
2165
2166
b37ad28b 2167__be32
6668958f
AA		 2168nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2169 struct nfsd4_open *open)
1da177e4 2170{
1da177e4
LT		 2171 clientid_t *clientid = &open->op_clientid;
2172 struct nfs4_client *clp = NULL;
2173 unsigned int strhashval;
2174 struct nfs4_stateowner *sop = NULL;
2175
1da177e4 2176 if (!check_name(open->op_owner))
0f442aa2 2177 return nfserr_inval;
1da177e4
LT		 2178
2179 if (STALE_CLIENTID(&open->op_clientid))
2180 return nfserr_stale_clientid;
2181
2182 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2183 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2184 open->op_stateowner = sop;
2185 if (!sop) {
2186 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2187 clp = find_confirmed_client(clientid);
2188 if (clp == NULL)
0f442aa2
BF		 2189 return nfserr_expired;
2190 goto renew;
1da177e4 2191 }
6668958f
AA		 2192 /* When sessions are used, skip open sequenceid processing */
2193 if (nfsd4_has_session(cstate))
2194 goto renew;
0f442aa2
BF		 2195 if (!sop->so_confirmed) {
2196 /* Replace unconfirmed owners without checking for replay. */
2197 clp = sop->so_client;
f044ff83 2198 release_openowner(sop);
0f442aa2
BF		 2199 open->op_stateowner = NULL;
2200 goto renew;
2201 }
2202 if (open->op_seqid == sop->so_seqid - 1) {
2203 if (sop->so_replay.rp_buflen)
a90b061c 2204 return nfserr_replay_me;
0f442aa2
BF		 2205 /* The original OPEN failed so spectacularly
2206 * that we don't even have replay data saved!
2207 * Therefore, we have no choice but to continue
2208 * processing this OPEN; presumably, we'll
2209 * fail again for the same reason.
2210 */
2211 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2212 goto renew;
2213 }
2214 if (open->op_seqid != sop->so_seqid)
2215 return nfserr_bad_seqid;
1da177e4 2216renew:
0f442aa2
BF		 2217 if (open->op_stateowner == NULL) {
2218 sop = alloc_init_open_stateowner(strhashval, clp, open);
2219 if (sop == NULL)
2220 return nfserr_resource;
2221 open->op_stateowner = sop;
2222 }
2223 list_del_init(&sop->so_close_lru);
1da177e4 2224 renew_client(sop->so_client);
0f442aa2 2225 return nfs_ok;
1da177e4
LT		 2226}
2227
b37ad28b 2228static inline __be32
4a6e43e6
N		 2229nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2230{
2231 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2232 return nfserr_openmode;
2233 else
2234 return nfs_ok;
2235}
2236
52f4fb43
N		 2237static struct nfs4_delegation *
2238find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2239{
2240 struct nfs4_delegation *dp;
2241
ea1da636 2242 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2243 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2244 return dp;
2245 }
2246 return NULL;
2247}
2248
b37ad28b 2249static __be32
567d9829
N		 2250nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2251 struct nfs4_delegation **dp)
2252{
2253 int flags;
b37ad28b 2254 __be32 status = nfserr_bad_stateid;
567d9829
N		 2255
2256 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2257 if (*dp == NULL)
c44c5eeb 2258 goto out;
567d9829
N		 2259 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2260 RD_STATE : WR_STATE;
2261 status = nfs4_check_delegmode(*dp, flags);
2262 if (status)
2263 *dp = NULL;
c44c5eeb
N		 2264out:
2265 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2266 return nfs_ok;
2267 if (status)
2268 return status;
2269 open->op_stateowner->so_confirmed = 1;
2270 return nfs_ok;
567d9829
N		 2271}
2272
b37ad28b 2273static __be32
1da177e4
LT		 2274nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2275{
2276 struct nfs4_stateid *local;
b37ad28b 2277 __be32 status = nfserr_share_denied;
1da177e4
LT		 2278 struct nfs4_stateowner *sop = open->op_stateowner;
2279
8beefa24 2280 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2281 /* ignore lock owners */
2282 if (local->st_stateowner->so_is_open_owner == 0)
2283 continue;
2284 /* remember if we have seen this open owner */
2285 if (local->st_stateowner == sop)
2286 *stpp = local;
2287 /* check for conflicting share reservations */
2288 if (!test_share(local, open))
2289 goto out;
2290 }
2291 status = 0;
2292out:
2293 return status;
2294}
2295
5ac049ac
N		 2296static inline struct nfs4_stateid *
2297nfs4_alloc_stateid(void)
2298{
2299 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2300}
2301
b37ad28b 2302static __be32
1da177e4 2303nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2304 struct nfs4_delegation *dp,
1da177e4
LT		 2305 struct svc_fh *cur_fh, int flags)
2306{
2307 struct nfs4_stateid *stp;
1da177e4 2308
5ac049ac 2309 stp = nfs4_alloc_stateid();
1da177e4
LT		 2310 if (stp == NULL)
2311 return nfserr_resource;
2312
567d9829
N		 2313 if (dp) {
2314 get_file(dp->dl_vfs_file);
2315 stp->st_vfs_file = dp->dl_vfs_file;
2316 } else {
b37ad28b 2317 __be32 status;
567d9829
N		 2318 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2319 &stp->st_vfs_file);
2320 if (status) {
2321 if (status == nfserr_dropit)
2322 status = nfserr_jukebox;
5ac049ac 2323 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2324 return status;
2325 }
1da177e4 2326 }
1da177e4
LT		 2327 *stpp = stp;
2328 return 0;
2329}
2330
b37ad28b 2331static inline __be32
1da177e4
LT		 2332nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2333 struct nfsd4_open *open)
2334{
2335 struct iattr iattr = {
2336 .ia_valid = ATTR_SIZE,
2337 .ia_size = 0,
2338 };
2339 if (!open->op_truncate)
2340 return 0;
2341 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2342 return nfserr_inval;
1da177e4
LT		 2343 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2344}
2345
b37ad28b 2346static __be32
1da177e4
LT		 2347nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2348{
2349 struct file *filp = stp->st_vfs_file;
7eaa36e2 2350 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2351 unsigned int share_access, new_writer;
b37ad28b 2352 __be32 status;
1da177e4
LT		 2353
2354 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2355 new_writer = (~share_access) & open->op_share_access
2356 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2357
6c26d08f 2358 if (new_writer) {
b37ad28b
AV		 2359 int err = get_write_access(inode);
2360 if (err)
2361 return nfserrno(err);
e518f056
BH		 2362 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2363 if (err)
2364 return nfserrno(err);
2365 file_take_write(filp);
6c26d08f 2366 }
1da177e4
LT		 2367 status = nfsd4_truncate(rqstp, cur_fh, open);
2368 if (status) {
6c26d08f
BF		 2369 if (new_writer)
2370 put_write_access(inode);
1da177e4
LT		 2371 return status;
2372 }
2373 /* remember the open */
6c26d08f 2374 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2375 __set_bit(open->op_share_access, &stp->st_access_bmap);
2376 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2377
2378 return nfs_ok;
2379}
2380
2381
1da177e4 2382static void
37515177 2383nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2384{
37515177
N		 2385 open->op_stateowner->so_confirmed = 1;
2386 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2387}
2388
2389/*
2390 * Attempt to hand out a delegation.
2391 */
2392static void
2393nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2394{
2395 struct nfs4_delegation *dp;
2396 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2397 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2398 struct file_lock fl, *flp = &fl;
2399 int status, flag = 0;
2400
2401 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2402 open->op_recall = 0;
2403 switch (open->op_claim_type) {
2404 case NFS4_OPEN_CLAIM_PREVIOUS:
2405 if (!atomic_read(&cb->cb_set))
2406 open->op_recall = 1;
2407 flag = open->op_delegate_type;
2408 if (flag == NFS4_OPEN_DELEGATE_NONE)
2409 goto out;
2410 break;
2411 case NFS4_OPEN_CLAIM_NULL:
2412 /* Let's not give out any delegations till everyone's
2413 * had the chance to reclaim theirs.... */
af558e33 2414 if (locks_in_grace())
7b190fec
N		 2415 goto out;
2416 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2417 goto out;
2418 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2419 flag = NFS4_OPEN_DELEGATE_WRITE;
2420 else
2421 flag = NFS4_OPEN_DELEGATE_READ;
2422 break;
2423 default:
2424 goto out;
2425 }
1da177e4
LT		 2426
2427 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2428 if (dp == NULL) {
2429 flag = NFS4_OPEN_DELEGATE_NONE;
2430 goto out;
2431 }
2432 locks_init_lock(&fl);
2433 fl.fl_lmops = &nfsd_lease_mng_ops;
2434 fl.fl_flags = FL_LEASE;
9167f501 2435 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2436 fl.fl_end = OFFSET_MAX;
2437 fl.fl_owner = (fl_owner_t)dp;
2438 fl.fl_file = stp->st_vfs_file;
2439 fl.fl_pid = current->tgid;
2440
a9933cea 2441 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2442 * the lock_manager callbacks fl_mylease and fl_change are used
2443 */
9167f501 2444 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2445 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2446 unhash_delegation(dp);
1da177e4
LT		 2447 flag = NFS4_OPEN_DELEGATE_NONE;
2448 goto out;
2449 }
2450
2451 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2452
2453 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
2454 dp->dl_stateid.si_boot,
2455 dp->dl_stateid.si_stateownerid,
2456 dp->dl_stateid.si_fileid,
2457 dp->dl_stateid.si_generation);
2458out:
7b190fec
N		 2459 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2460 && flag == NFS4_OPEN_DELEGATE_NONE
2461 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2462 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2463 open->op_delegate_type = flag;
2464}
2465
2466/*
2467 * called with nfs4_lock_state() held.
2468 */
b37ad28b 2469__be32
1da177e4
LT		 2470nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2471{
6668958f 2472 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2473 struct nfs4_file *fp = NULL;
2474 struct inode *ino = current_fh->fh_dentry->d_inode;
2475 struct nfs4_stateid *stp = NULL;
567d9829 2476 struct nfs4_delegation *dp = NULL;
b37ad28b 2477 __be32 status;
1da177e4
LT		 2478
2479 status = nfserr_inval;
d87a8ade 2480 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2481 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2482 goto out;
2483 /*
2484 * Lookup file; if found, lookup stateid and check open request,
2485 * and check for delegations in the process of being recalled.
2486 * If not found, create the nfs4_file struct
2487 */
2488 fp = find_file(ino);
2489 if (fp) {
2490 if ((status = nfs4_check_open(fp, open, &stp)))
2491 goto out;
c44c5eeb
N		 2492 status = nfs4_check_deleg(fp, open, &dp);
2493 if (status)
2494 goto out;
1da177e4 2495 } else {
c44c5eeb
N		 2496 status = nfserr_bad_stateid;
2497 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2498 goto out;
1da177e4
LT		 2499 status = nfserr_resource;
2500 fp = alloc_init_file(ino);
2501 if (fp == NULL)
2502 goto out;
2503 }
2504
2505 /*
2506 * OPEN the file, or upgrade an existing OPEN.
2507 * If truncate fails, the OPEN fails.
2508 */
2509 if (stp) {
2510 /* Stateid was found, this is an OPEN upgrade */
2511 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2512 if (status)
2513 goto out;
444c2c07 2514 update_stateid(&stp->st_stateid);
1da177e4
LT		 2515 } else {
2516 /* Stateid was not found, this is a new OPEN */
2517 int flags = 0;
9ecb6a08 2518 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2519 flags |= NFSD_MAY_READ;
1da177e4 2520 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2521 flags |= NFSD_MAY_WRITE;
567d9829
N		 2522 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2523 if (status)
1da177e4
LT		 2524 goto out;
2525 init_stateid(stp, fp, open);
2526 status = nfsd4_truncate(rqstp, current_fh, open);
2527 if (status) {
2283963f 2528 release_open_stateid(stp);
1da177e4
LT		 2529 goto out;
2530 }
6668958f
AA		 2531 if (nfsd4_has_session(&resp->cstate))
2532 update_stateid(&stp->st_stateid);
1da177e4
LT		 2533 }
2534 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2535
6668958f
AA		 2536 if (nfsd4_has_session(&resp->cstate))
2537 open->op_stateowner->so_confirmed = 1;
2538
1da177e4
LT		 2539 /*
2540 * Attempt to hand out a delegation. No error return, because the
2541 * OPEN succeeds even if we fail.
2542 */
2543 nfs4_open_delegation(current_fh, open, stp);
2544
2545 status = nfs_ok;
2546
2547 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
2548 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
2549 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
2550out:
13cd2184
N		 2551 if (fp)
2552 put_nfs4_file(fp);
37515177
N		 2553 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2554 nfs4_set_claim_prev(open);
1da177e4
LT		 2555 /*
2556 * To finish the open response, we just need to set the rflags.
2557 */
2558 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2559 if (!open->op_stateowner->so_confirmed &&
2560 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2561 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2562
2563 return status;
2564}
2565
b37ad28b 2566__be32
b591480b
BF		 2567nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2568 clientid_t *clid)
1da177e4
LT		 2569{
2570 struct nfs4_client *clp;
b37ad28b 2571 __be32 status;
1da177e4
LT		 2572
2573 nfs4_lock_state();
2574 dprintk("process_renew(%08x/%08x): starting\n",
2575 clid->cl_boot, clid->cl_id);
2576 status = nfserr_stale_clientid;
2577 if (STALE_CLIENTID(clid))
2578 goto out;
2579 clp = find_confirmed_client(clid);
2580 status = nfserr_expired;
2581 if (clp == NULL) {
2582 /* We assume the client took too long to RENEW. */
2583 dprintk("nfsd4_renew: clientid not found!\n");
2584 goto out;
2585 }
2586 renew_client(clp);
2587 status = nfserr_cb_path_down;
ea1da636 2588 if (!list_empty(&clp->cl_delegations)
c237dc03 2589 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2590 goto out;
2591 status = nfs_ok;
2592out:
2593 nfs4_unlock_state();
2594 return status;
2595}
2596
af558e33
BF		 2597struct lock_manager nfsd4_manager = {
2598};
2599
a76b4319 2600static void
af558e33 2601nfsd4_end_grace(void)
a76b4319
N		 2602{
2603 dprintk("NFSD: end of grace period\n");
c7b9a459 2604 nfsd4_recdir_purge_old();
af558e33 2605 locks_end_grace(&nfsd4_manager);
a76b4319
N		 2606}
2607
fd39ca9a 2608static time_t
1da177e4
LT		 2609nfs4_laundromat(void)
2610{
2611 struct nfs4_client *clp;
2612 struct nfs4_stateowner *sop;
2613 struct nfs4_delegation *dp;
2614 struct list_head *pos, *next, reaplist;
2615 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
2616 time_t t, clientid_val = NFSD_LEASE_TIME;
2617 time_t u, test_val = NFSD_LEASE_TIME;
2618
2619 nfs4_lock_state();
2620
2621 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2622 if (locks_in_grace())
2623 nfsd4_end_grace();
1da177e4
LT		 2624 list_for_each_safe(pos, next, &client_lru) {
2625 clp = list_entry(pos, struct nfs4_client, cl_lru);
2626 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2627 t = clp->cl_time - cutoff;
2628 if (clientid_val > t)
2629 clientid_val = t;
2630 break;
2631 }
2632 dprintk("NFSD: purging unused client (clientid %08x)\n",
2633 clp->cl_clientid.cl_id);
c7b9a459 2634 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2635 expire_client(clp);
2636 }
2637 INIT_LIST_HEAD(&reaplist);
2638 spin_lock(&recall_lock);
2639 list_for_each_safe(pos, next, &del_recall_lru) {
2640 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2641 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2642 u = dp->dl_time - cutoff;
2643 if (test_val > u)
2644 test_val = u;
2645 break;
2646 }
2647 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2648 dp, dp->dl_flock);
2649 list_move(&dp->dl_recall_lru, &reaplist);
2650 }
2651 spin_unlock(&recall_lock);
2652 list_for_each_safe(pos, next, &reaplist) {
2653 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2654 list_del_init(&dp->dl_recall_lru);
2655 unhash_delegation(dp);
2656 }
2657 test_val = NFSD_LEASE_TIME;
2658 list_for_each_safe(pos, next, &close_lru) {
2659 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2660 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2661 u = sop->so_time - cutoff;
2662 if (test_val > u)
2663 test_val = u;
2664 break;
2665 }
2666 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2667 sop->so_id);
f044ff83 2668 release_openowner(sop);
1da177e4
LT		 2669 }
2670 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2671 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2672 nfs4_unlock_state();
2673 return clientid_val;
2674}
2675
a254b246
HH		 2676static struct workqueue_struct *laundry_wq;
2677static void laundromat_main(struct work_struct *);
2678static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2679
2680static void
c4028958 2681laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2682{
2683 time_t t;
2684
2685 t = nfs4_laundromat();
2686 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2687 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2688}
2689
fd39ca9a 2690static struct nfs4_stateowner *
f8816512
N		 2691search_close_lru(u32 st_id, int flags)
2692{
1da177e4
LT		 2693 struct nfs4_stateowner *local = NULL;
2694
1da177e4
LT		 2695 if (flags & CLOSE_STATE) {
2696 list_for_each_entry(local, &close_lru, so_close_lru) {
2697 if (local->so_id == st_id)
2698 return local;
2699 }
2700 }
2701 return NULL;
2702}
2703
2704static inline int
2705nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2706{
7eaa36e2 2707 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2708}
2709
2710static int
2711STALE_STATEID(stateid_t *stateid)
2712{
78155ed7
BN		 2713 if (time_after((unsigned long)boot_time,
2714 (unsigned long)stateid->si_boot)) {
2715 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2716 stateid->si_boot, stateid->si_stateownerid,
2717 stateid->si_fileid, stateid->si_generation);
2718 return 1;
2719 }
2720 return 0;
2721}
2722
2723static int
2724EXPIRED_STATEID(stateid_t *stateid)
2725{
2726 if (time_before((unsigned long)boot_time,
2727 ((unsigned long)stateid->si_boot)) &&
b8fd47ae 2728 time_before((unsigned long)(stateid->si_boot + lease_time), get_seconds())) {
78155ed7
BN		 2729 dprintk("NFSD: expired stateid (%08x/%08x/%08x/%08x)!\n",
2730 stateid->si_boot, stateid->si_stateownerid,
2731 stateid->si_fileid, stateid->si_generation);
2732 return 1;
2733 }
2734 return 0;
2735}
2736
2737static __be32
2738stateid_error_map(stateid_t *stateid)
2739{
2740 if (STALE_STATEID(stateid))
2741 return nfserr_stale_stateid;
2742 if (EXPIRED_STATEID(stateid))
2743 return nfserr_expired;
2744
2745 dprintk("NFSD: bad stateid (%08x/%08x/%08x/%08x)!\n",
2746 stateid->si_boot, stateid->si_stateownerid,
2747 stateid->si_fileid, stateid->si_generation);
2748 return nfserr_bad_stateid;
1da177e4
LT		 2749}
2750
2751static inline int
2752access_permit_read(unsigned long access_bmap)
2753{
2754 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2755 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2756 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2757}
2758
2759static inline int
2760access_permit_write(unsigned long access_bmap)
2761{
2762 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2763 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2764}
2765
2766static
b37ad28b 2767__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2768{
b37ad28b 2769 __be32 status = nfserr_openmode;
1da177e4
LT		 2770
2771 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2772 goto out;
2773 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2774 goto out;
2775 status = nfs_ok;
2776out:
2777 return status;
2778}
2779
b37ad28b 2780static inline __be32
1da177e4
LT		 2781check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2782{
203a8c8e 2783 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2784 return nfs_ok;
af558e33 2785 else if (locks_in_grace()) {
1da177e4
LT		 2786 /* Answer in remaining cases depends on existance of
2787 * conflicting state; so we must wait out the grace period. */
2788 return nfserr_grace;
2789 } else if (flags & WR_STATE)
2790 return nfs4_share_conflict(current_fh,
2791 NFS4_SHARE_DENY_WRITE);
2792 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2793 return nfs4_share_conflict(current_fh,
2794 NFS4_SHARE_DENY_READ);
2795}
2796
2797/*
2798 * Allow READ/WRITE during grace period on recovered state only for files
2799 * that are not able to provide mandatory locking.
2800 */
2801static inline int
18f82731 2802grace_disallows_io(struct inode *inode)
1da177e4 2803{
203a8c8e 2804 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2805}
2806
6668958f 2807static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2808{
6668958f
AA		 2809 /*
2810 * When sessions are used the stateid generation number is ignored
2811 * when it is zero.
2812 */
2813 if ((flags & HAS_SESSION) && in->si_generation == 0)
2814 goto out;
2815
0836f587
BF		 2816 /* If the client sends us a stateid from the future, it's buggy: */
2817 if (in->si_generation > ref->si_generation)
2818 return nfserr_bad_stateid;
2819 /*
2820 * The following, however, can happen. For example, if the
2821 * client sends an open and some IO at the same time, the open
2822 * may bump si_generation while the IO is still in flight.
2823 * Thanks to hard links and renames, the client never knows what
2824 * file an open will affect. So it could avoid that situation
2825 * only by serializing all opens and IO from the same open
2826 * owner. To recover from the old_stateid error, the client
2827 * will just have to retry the IO:
2828 */
2829 if (in->si_generation < ref->si_generation)
2830 return nfserr_old_stateid;
6668958f 2831out:
0836f587
BF		 2832 return nfs_ok;
2833}
2834
3e633079
BF		 2835static int is_delegation_stateid(stateid_t *stateid)
2836{
2837 return stateid->si_fileid == 0;
2838}
2839
1da177e4
LT		 2840/*
2841* Checks for stateid operations
2842*/
b37ad28b 2843__be32
dd453dfd
BH		 2844nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2845 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2846{
2847 struct nfs4_stateid *stp = NULL;
2848 struct nfs4_delegation *dp = NULL;
dd453dfd 2849 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2850 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2851 __be32 status;
1da177e4 2852
1da177e4
LT		 2853 if (filpp)
2854 *filpp = NULL;
2855
18f82731 2856 if (grace_disallows_io(ino))
1da177e4
LT		 2857 return nfserr_grace;
2858
6668958f
AA		 2859 if (nfsd4_has_session(cstate))
2860 flags |= HAS_SESSION;
2861
1da177e4
LT		 2862 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2863 return check_special_stateids(current_fh, stateid, flags);
2864
1da177e4
LT		 2865 status = nfserr_stale_stateid;
2866 if (STALE_STATEID(stateid))
2867 goto out;
2868
1da177e4 2869 status = nfserr_bad_stateid;
3e633079 2870 if (is_delegation_stateid(stateid)) {
a4455be0 2871 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2872 if (!dp) {
2873 status = stateid_error_map(stateid);
1da177e4 2874 goto out;
78155ed7 2875 }
6668958f
AA		 2876 status = check_stateid_generation(stateid, &dp->dl_stateid,
2877 flags);
0c2a498f
BF		 2878 if (status)
2879 goto out;
dc9bf700
BF		 2880 status = nfs4_check_delegmode(dp, flags);
2881 if (status)
2882 goto out;
2883 renew_client(dp->dl_client);
dc9bf700
BF		 2884 if (filpp)
2885 *filpp = dp->dl_vfs_file;
1da177e4 2886 } else { /* open or lock stateid */
a4455be0 2887 stp = find_stateid(stateid, flags);
78155ed7
BN		 2888 if (!stp) {
2889 status = stateid_error_map(stateid);
1da177e4 2890 goto out;
78155ed7 2891 }
6150ef0d 2892 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2893 goto out;
2894 if (!stp->st_stateowner->so_confirmed)
2895 goto out;
6668958f
AA		 2896 status = check_stateid_generation(stateid, &stp->st_stateid,
2897 flags);
0c2a498f
BF		 2898 if (status)
2899 goto out;
a4455be0
BF		 2900 status = nfs4_check_openmode(stp, flags);
2901 if (status)
1da177e4
LT		 2902 goto out;
2903 renew_client(stp->st_stateowner->so_client);
2904 if (filpp)
2905 *filpp = stp->st_vfs_file;
1da177e4
LT		 2906 }
2907 status = nfs_ok;
2908out:
2909 return status;
2910}
2911
4c4cd222
N		 2912static inline int
2913setlkflg (int type)
2914{
2915 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2916 RD_STATE : WR_STATE;
2917}
1da177e4
LT		 2918
2919/*
2920 * Checks for sequence id mutating operations.
2921 */
b37ad28b 2922static __be32
dd453dfd
BH		 2923nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2924 stateid_t *stateid, int flags,
2925 struct nfs4_stateowner **sopp,
2926 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2927{
1da177e4
LT		 2928 struct nfs4_stateid *stp;
2929 struct nfs4_stateowner *sop;
dd453dfd 2930 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2931 __be32 status;
1da177e4
LT		 2932
2933 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2934 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2935 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2936 stateid->si_generation);
3a4f98bb 2937
1da177e4
LT		 2938 *stpp = NULL;
2939 *sopp = NULL;
2940
1da177e4 2941 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2942 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2943 return nfserr_bad_stateid;
1da177e4
LT		 2944 }
2945
1da177e4 2946 if (STALE_STATEID(stateid))
3a4f98bb 2947 return nfserr_stale_stateid;
6668958f
AA		 2948
2949 if (nfsd4_has_session(cstate))
2950 flags |= HAS_SESSION;
2951
1da177e4
LT		 2952 /*
2953 * We return BAD_STATEID if filehandle doesn't match stateid,
2954 * the confirmed flag is incorrecly set, or the generation
2955 * number is incorrect.
1da177e4 2956 */
f8816512
N		 2957 stp = find_stateid(stateid, flags);
2958 if (stp == NULL) {
2959 /*
2960 * Also, we should make sure this isn't just the result of
2961 * a replayed close:
2962 */
2963 sop = search_close_lru(stateid->si_stateownerid, flags);
2964 if (sop == NULL)
78155ed7 2965 return stateid_error_map(stateid);
f8816512
N		 2966 *sopp = sop;
2967 goto check_replay;
2968 }
1da177e4 2969
39325bd0
BF		 2970 *stpp = stp;
2971 *sopp = sop = stp->st_stateowner;
2972
4c4cd222 2973 if (lock) {
4c4cd222 2974 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2975 struct nfs4_client *clp = sop->so_client;
4c4cd222 2976 int lkflg = 0;
b37ad28b 2977 __be32 status;
4c4cd222
N		 2978
2979 lkflg = setlkflg(lock->lk_type);
2980
2981 if (lock->lk_is_new) {
599e0a22
BF		 2982 if (!sop->so_is_open_owner)
2983 return nfserr_bad_stateid;
60adfc50
AA		 2984 if (!(flags & HAS_SESSION) &&
2985 !same_clid(&clp->cl_clientid, lockclid))
2986 return nfserr_bad_stateid;
599e0a22
BF		 2987 /* stp is the open stateid */
2988 status = nfs4_check_openmode(stp, lkflg);
2989 if (status)
2990 return status;
2991 } else {
2992 /* stp is the lock stateid */
2993 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2994 if (status)
2995 return status;
4c4cd222 2996 }
1da177e4
LT		 2997 }
2998
f3362737 2999 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 3000 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 3001 return nfserr_bad_stateid;
1da177e4
LT		 3002 }
3003
1da177e4
LT		 3004 /*
3005 * We now validate the seqid and stateid generation numbers.
3006 * For the moment, we ignore the possibility of
3007 * generation number wraparound.
3008 */
6668958f 3009 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 3010 goto check_replay;
3011
3a4f98bb 3012 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 3013 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 3014 " unconfirmed stateowner!\n");
3015 return nfserr_bad_stateid;
1da177e4 3016 }
3a4f98bb 3017 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 3018 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 3019 " confirmed yet!\n");
3020 return nfserr_bad_stateid;
1da177e4 3021 }
6668958f 3022 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 3023 if (status)
3024 return status;
52fd004e 3025 renew_client(sop->so_client);
3a4f98bb 3026 return nfs_ok;
1da177e4 3027
1da177e4 3028check_replay:
bd9aac52 3029 if (seqid == sop->so_seqid - 1) {
849823c5 3030 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 3031 /* indicate replay to calling function */
a90b061c 3032 return nfserr_replay_me;
1da177e4 3033 }
2fdada03 3034 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 3035 sop->so_seqid, seqid);
3036 *sopp = NULL;
3037 return nfserr_bad_seqid;
1da177e4
LT		 3038}
3039
b37ad28b 3040__be32
ca364317 3041nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3042 struct nfsd4_open_confirm *oc)
1da177e4 3043{
b37ad28b 3044 __be32 status;
1da177e4
LT		 3045 struct nfs4_stateowner *sop;
3046 struct nfs4_stateid *stp;
3047
3048 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 3049 (int)cstate->current_fh.fh_dentry->d_name.len,
3050 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3051
ca364317 3052 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3053 if (status)
3054 return status;
1da177e4
LT		 3055
3056 nfs4_lock_state();
3057
dd453dfd 3058 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3059 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3060 CONFIRM | OPEN_STATE,
1da177e4
LT		 3061 &oc->oc_stateowner, &stp, NULL)))
3062 goto out;
3063
3064 sop = oc->oc_stateowner;
3065 sop->so_confirmed = 1;
3066 update_stateid(&stp->st_stateid);
3067 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
3068 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
3069 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
3070 stp->st_stateid.si_boot,
3071 stp->st_stateid.si_stateownerid,
3072 stp->st_stateid.si_fileid,
3073 stp->st_stateid.si_generation);
c7b9a459
N		 3074
3075 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3076out:
f2327d9a 3077 if (oc->oc_stateowner) {
1da177e4 3078 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3079 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3080 }
1da177e4
LT		 3081 nfs4_unlock_state();
3082 return status;
3083}
3084
3085
3086/*
3087 * unset all bits in union bitmap (bmap) that
3088 * do not exist in share (from successful OPEN_DOWNGRADE)
3089 */
3090static void
3091reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3092{
3093 int i;
3094 for (i = 1; i < 4; i++) {
3095 if ((i & access) != i)
3096 __clear_bit(i, bmap);
3097 }
3098}
3099
3100static void
3101reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3102{
3103 int i;
3104 for (i = 0; i < 4; i++) {
3105 if ((i & deny) != i)
3106 __clear_bit(i, bmap);
3107 }
3108}
3109
b37ad28b 3110__be32
ca364317
BF		 3111nfsd4_open_downgrade(struct svc_rqst *rqstp,
3112 struct nfsd4_compound_state *cstate,
a4f1706a 3113 struct nfsd4_open_downgrade *od)
1da177e4 3114{
b37ad28b 3115 __be32 status;
1da177e4
LT		 3116 struct nfs4_stateid *stp;
3117 unsigned int share_access;
3118
3119 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3120 (int)cstate->current_fh.fh_dentry->d_name.len,
3121 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3122
d87a8ade 3123 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3124 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3125 return nfserr_inval;
3126
3127 nfs4_lock_state();
dd453dfd 3128 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3129 od->od_seqid,
1da177e4 3130 &od->od_stateid,
f3362737 3131 OPEN_STATE,
1da177e4
LT		 3132 &od->od_stateowner, &stp, NULL)))
3133 goto out;
3134
3135 status = nfserr_inval;
3136 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3137 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3138 stp->st_access_bmap, od->od_share_access);
3139 goto out;
3140 }
3141 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3142 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3143 stp->st_deny_bmap, od->od_share_deny);
3144 goto out;
3145 }
3146 set_access(&share_access, stp->st_access_bmap);
3147 nfs4_file_downgrade(stp->st_vfs_file,
3148 share_access & ~od->od_share_access);
3149
3150 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3151 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3152
3153 update_stateid(&stp->st_stateid);
3154 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3155 status = nfs_ok;
3156out:
f2327d9a 3157 if (od->od_stateowner) {
1da177e4 3158 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3159 cstate->replay_owner = od->od_stateowner;
f2327d9a 3160 }
1da177e4
LT		 3161 nfs4_unlock_state();
3162 return status;
3163}
3164
3165/*
3166 * nfs4_unlock_state() called after encode
3167 */
b37ad28b 3168__be32
ca364317 3169nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3170 struct nfsd4_close *close)
1da177e4 3171{
b37ad28b 3172 __be32 status;
1da177e4
LT		 3173 struct nfs4_stateid *stp;
3174
3175 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3176 (int)cstate->current_fh.fh_dentry->d_name.len,
3177 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3178
3179 nfs4_lock_state();
3180 /* check close_lru for replay */
dd453dfd 3181 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3182 close->cl_seqid,
1da177e4 3183 &close->cl_stateid,
f3362737 3184 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3185 &close->cl_stateowner, &stp, NULL)))
3186 goto out;
1da177e4
LT		 3187 status = nfs_ok;
3188 update_stateid(&stp->st_stateid);
3189 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3190
04ef5954 3191 /* release_stateid() calls nfsd_close() if needed */
2283963f 3192 release_open_stateid(stp);
04ef5954
BF		 3193
3194 /* place unused nfs4_stateowners on so_close_lru list to be
3195 * released by the laundromat service after the lease period
3196 * to enable us to handle CLOSE replay
3197 */
3198 if (list_empty(&close->cl_stateowner->so_stateids))
3199 move_to_close_lru(close->cl_stateowner);
1da177e4 3200out:
f2327d9a 3201 if (close->cl_stateowner) {
1da177e4 3202 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3203 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3204 }
1da177e4
LT		 3205 nfs4_unlock_state();
3206 return status;
3207}
3208
b37ad28b 3209__be32
ca364317
BF		 3210nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3211 struct nfsd4_delegreturn *dr)
1da177e4 3212{
203a8c8e
BF		 3213 struct nfs4_delegation *dp;
3214 stateid_t *stateid = &dr->dr_stateid;
3215 struct inode *inode;
b37ad28b 3216 __be32 status;
6668958f 3217 int flags = 0;
1da177e4 3218
ca364317 3219 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3220 return status;
3221 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3222
6668958f
AA		 3223 if (nfsd4_has_session(cstate))
3224 flags |= HAS_SESSION;
1da177e4 3225 nfs4_lock_state();
203a8c8e
BF		 3226 status = nfserr_bad_stateid;
3227 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3228 goto out;
3229 status = nfserr_stale_stateid;
3230 if (STALE_STATEID(stateid))
3231 goto out;
7e0f7cf5 3232 status = nfserr_bad_stateid;
203a8c8e
BF		 3233 if (!is_delegation_stateid(stateid))
3234 goto out;
203a8c8e 3235 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3236 if (!dp) {
3237 status = stateid_error_map(stateid);
203a8c8e 3238 goto out;
78155ed7 3239 }
6668958f 3240 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3241 if (status)
3242 goto out;
3243 renew_client(dp->dl_client);
3244
3245 unhash_delegation(dp);
1da177e4 3246out:
203a8c8e
BF		 3247 nfs4_unlock_state();
3248
1da177e4
LT		 3249 return status;
3250}
3251
3252
3253/*
3254 * Lock owner state (byte-range locks)
3255 */
3256#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3257#define LOCK_HASH_BITS 8
3258#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3259#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3260
87df4de8
BH		 3261static inline u64
3262end_offset(u64 start, u64 len)
3263{
3264 u64 end;
3265
3266 end = start + len;
3267 return end >= start ? end: NFS4_MAX_UINT64;
3268}
3269
3270/* last octet in a range */
3271static inline u64
3272last_byte_offset(u64 start, u64 len)
3273{
3274 u64 end;
3275
3276 BUG_ON(!len);
3277 end = start + len;
3278 return end > start ? end - 1: NFS4_MAX_UINT64;
3279}
3280
1da177e4
LT		 3281#define lockownerid_hashval(id) \
3282 ((id) & LOCK_HASH_MASK)
3283
3284static inline unsigned int
3285lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3286 struct xdr_netobj *ownername)
3287{
3288 return (file_hashval(inode) + cl_id
3289 + opaque_hashval(ownername->data, ownername->len))
3290 & LOCK_HASH_MASK;
3291}
3292
3293static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3294static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3295static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3296
fd39ca9a 3297static struct nfs4_stateid *
1da177e4
LT		 3298find_stateid(stateid_t *stid, int flags)
3299{
9346eff0 3300 struct nfs4_stateid *local;
1da177e4
LT		 3301 u32 st_id = stid->si_stateownerid;
3302 u32 f_id = stid->si_fileid;
3303 unsigned int hashval;
3304
3305 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3306 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3307 hashval = stateid_hashval(st_id, f_id);
3308 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3309 if ((local->st_stateid.si_stateownerid == st_id) &&
3310 (local->st_stateid.si_fileid == f_id))
3311 return local;
3312 }
3313 }
9346eff0
KK		 3314
3315 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3316 hashval = stateid_hashval(st_id, f_id);
3317 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3318 if ((local->st_stateid.si_stateownerid == st_id) &&
3319 (local->st_stateid.si_fileid == f_id))
3320 return local;
3321 }
849823c5 3322 }
1da177e4
LT		 3323 return NULL;
3324}
3325
3326static struct nfs4_delegation *
3327find_delegation_stateid(struct inode *ino, stateid_t *stid)
3328{
13cd2184
N		 3329 struct nfs4_file *fp;
3330 struct nfs4_delegation *dl;
1da177e4
LT		 3331
3332 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
3333 stid->si_boot, stid->si_stateownerid,
3334 stid->si_fileid, stid->si_generation);
3335
1da177e4 3336 fp = find_file(ino);
13cd2184
N		 3337 if (!fp)
3338 return NULL;
3339 dl = find_delegation_file(fp, stid);
3340 put_nfs4_file(fp);
3341 return dl;
1da177e4
LT		 3342}
3343
3344/*
3345 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3346 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3347 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3348 * locking, this prevents us from being completely protocol-compliant. The
3349 * real solution to this problem is to start using unsigned file offsets in
3350 * the VFS, but this is a very deep change!
3351 */
3352static inline void
3353nfs4_transform_lock_offset(struct file_lock *lock)
3354{
3355 if (lock->fl_start < 0)
3356 lock->fl_start = OFFSET_MAX;
3357 if (lock->fl_end < 0)
3358 lock->fl_end = OFFSET_MAX;
3359}
3360
d5b9026a
N		 3361/* Hack!: For now, we're defining this just so we can use a pointer to it
3362 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 3363static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3364};
1da177e4
LT		 3365
3366static inline void
3367nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3368{
d5b9026a
N		 3369 struct nfs4_stateowner *sop;
3370 unsigned int hval;
1da177e4 3371
d5b9026a
N		 3372 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3373 sop = (struct nfs4_stateowner *) fl->fl_owner;
3374 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3375 kref_get(&sop->so_ref);
3376 deny->ld_sop = sop;
3377 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3378 } else {
3379 deny->ld_sop = NULL;
3380 deny->ld_clientid.cl_boot = 0;
3381 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3382 }
3383 deny->ld_start = fl->fl_start;
87df4de8
BH		 3384 deny->ld_length = NFS4_MAX_UINT64;
3385 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3386 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3387 deny->ld_type = NFS4_READ_LT;
3388 if (fl->fl_type != F_RDLCK)
3389 deny->ld_type = NFS4_WRITE_LT;
3390}
3391
1da177e4
LT		 3392static struct nfs4_stateowner *
3393find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3394 struct xdr_netobj *owner)
3395{
3396 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3397 struct nfs4_stateowner *op;
3398
3399 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3400 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3401 return op;
3402 }
3403 return NULL;
3404}
3405
3406/*
3407 * Alloc a lock owner structure.
3408 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3409 * occured.
3410 *
3411 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3412 */
3413
3414static struct nfs4_stateowner *
3415alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3416 struct nfs4_stateowner *sop;
3417 struct nfs4_replay *rp;
3418 unsigned int idhashval;
3419
3420 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3421 return NULL;
3422 idhashval = lockownerid_hashval(current_ownerid);
3423 INIT_LIST_HEAD(&sop->so_idhash);
3424 INIT_LIST_HEAD(&sop->so_strhash);
3425 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3426 INIT_LIST_HEAD(&sop->so_stateids);
3427 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3428 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3429 sop->so_time = 0;
3430 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3431 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3432 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3433 sop->so_is_open_owner = 0;
3434 sop->so_id = current_ownerid++;
3435 sop->so_client = clp;
b59e3c0e
NB		 3436 /* It is the openowner seqid that will be incremented in encode in the
3437 * case of new lockowners; so increment the lock seqid manually: */
3438 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3439 sop->so_confirmed = 1;
3440 rp = &sop->so_replay;
de1ae286 3441 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3442 rp->rp_buflen = 0;
3443 rp->rp_buf = rp->rp_ibuf;
3444 return sop;
3445}
3446
fd39ca9a 3447static struct nfs4_stateid *
1da177e4
LT		 3448alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3449{
3450 struct nfs4_stateid *stp;
3451 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3452
5ac049ac
N		 3453 stp = nfs4_alloc_stateid();
3454 if (stp == NULL)
1da177e4
LT		 3455 goto out;
3456 INIT_LIST_HEAD(&stp->st_hash);
3457 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3458 INIT_LIST_HEAD(&stp->st_perstateowner);
3459 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3460 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3461 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3462 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3463 stp->st_stateowner = sop;
13cd2184 3464 get_nfs4_file(fp);
1da177e4 3465 stp->st_file = fp;
78155ed7 3466 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3467 stp->st_stateid.si_stateownerid = sop->so_id;
3468 stp->st_stateid.si_fileid = fp->fi_id;
3469 stp->st_stateid.si_generation = 0;
3470 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3471 stp->st_access_bmap = open_stp->st_access_bmap;
3472 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3473 stp->st_openstp = open_stp;
1da177e4
LT		 3474
3475out:
3476 return stp;
3477}
3478
fd39ca9a 3479static int
1da177e4
LT		 3480check_lock_length(u64 offset, u64 length)
3481{
87df4de8 3482 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3483 LOFF_OVERFLOW(offset, length)));
3484}
3485
3486/*
3487 * LOCK operation
3488 */
b37ad28b 3489__be32
ca364317 3490nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3491 struct nfsd4_lock *lock)
1da177e4 3492{
3e9e3dbe 3493 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3494 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3495 struct nfs4_stateid *lock_stp;
3496 struct file *filp;
3497 struct file_lock file_lock;
8dc7c311 3498 struct file_lock conflock;
b37ad28b 3499 __be32 status = 0;
1da177e4 3500 unsigned int strhashval;
150b3934 3501 unsigned int cmd;
b8dd7b9a 3502 int err;
1da177e4
LT		 3503
3504 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3505 (long long) lock->lk_offset,
3506 (long long) lock->lk_length);
3507
1da177e4
LT		 3508 if (check_lock_length(lock->lk_offset, lock->lk_length))
3509 return nfserr_inval;
3510
ca364317 3511 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3512 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3513 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3514 return status;
3515 }
3516
1da177e4
LT		 3517 nfs4_lock_state();
3518
3519 if (lock->lk_is_new) {
893f8770
N		 3520 /*
3521 * Client indicates that this is a new lockowner.
3522 * Use open owner and open stateid to create lock owner and
3523 * lock stateid.
3524 */
1da177e4
LT		 3525 struct nfs4_stateid *open_stp = NULL;
3526 struct nfs4_file *fp;
3527
3528 status = nfserr_stale_clientid;
60adfc50
AA		 3529 if (!nfsd4_has_session(cstate) &&
3530 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3531 goto out;
1da177e4 3532
1da177e4 3533 /* validate and update open stateid and open seqid */
dd453dfd 3534 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3535 lock->lk_new_open_seqid,
3536 &lock->lk_new_open_stateid,
f3362737 3537 OPEN_STATE,
3a65588a 3538 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3539 lock);
37515177 3540 if (status)
1da177e4 3541 goto out;
3a65588a 3542 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3543 /* create lockowner and lock stateid */
3544 fp = open_stp->st_file;
3545 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3546 open_sop->so_client->cl_clientid.cl_id,
3547 &lock->v.new.owner);
3e9e3dbe
N		 3548 /* XXX: Do we need to check for duplicate stateowners on
3549 * the same file, or should they just be allowed (and
3550 * create new stateids)? */
1da177e4 3551 status = nfserr_resource;
b59e3c0e
NB		 3552 lock_sop = alloc_init_lock_stateowner(strhashval,
3553 open_sop->so_client, open_stp, lock);
3554 if (lock_sop == NULL)
1da177e4 3555 goto out;
b59e3c0e 3556 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3557 if (lock_stp == NULL)
1da177e4 3558 goto out;
1da177e4
LT		 3559 } else {
3560 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3561 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3562 lock->lk_old_lock_seqid,
3563 &lock->lk_old_lock_stateid,
f3362737 3564 LOCK_STATE,
3a65588a 3565 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3566 if (status)
3567 goto out;
3a65588a 3568 lock_sop = lock->lk_replay_owner;
1da177e4 3569 }
3a65588a 3570 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3571 filp = lock_stp->st_vfs_file;
3572
0dd395dc 3573 status = nfserr_grace;
af558e33 3574 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3575 goto out;
3576 status = nfserr_no_grace;
af558e33 3577 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3578 goto out;
3579
1da177e4
LT		 3580 locks_init_lock(&file_lock);
3581 switch (lock->lk_type) {
3582 case NFS4_READ_LT:
3583 case NFS4_READW_LT:
3584 file_lock.fl_type = F_RDLCK;
150b3934 3585 cmd = F_SETLK;
1da177e4
LT		 3586 break;
3587 case NFS4_WRITE_LT:
3588 case NFS4_WRITEW_LT:
3589 file_lock.fl_type = F_WRLCK;
150b3934 3590 cmd = F_SETLK;
1da177e4
LT		 3591 break;
3592 default:
3593 status = nfserr_inval;
3594 goto out;
3595 }
b59e3c0e 3596 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3597 file_lock.fl_pid = current->tgid;
3598 file_lock.fl_file = filp;
3599 file_lock.fl_flags = FL_POSIX;
d5b9026a 3600 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3601
3602 file_lock.fl_start = lock->lk_offset;
87df4de8 3603 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3604 nfs4_transform_lock_offset(&file_lock);
3605
3606 /*
3607 * Try to lock the file in the VFS.
3608 * Note: locks.c uses the BKL to protect the inode's lock list.
3609 */
3610
fd85b817 3611 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3612 switch (-err) {
1da177e4
LT		 3613 case 0: /* success! */
3614 update_stateid(&lock_stp->st_stateid);
3615 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3616 sizeof(stateid_t));
b8dd7b9a 3617 status = 0;
eb76b3fd
AA		 3618 break;
3619 case (EAGAIN): /* conflock holds conflicting lock */
3620 status = nfserr_denied;
3621 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3622 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3623 break;
1da177e4
LT		 3624 case (EDEADLK):
3625 status = nfserr_deadlock;
eb76b3fd 3626 break;
1da177e4 3627 default:
fd85b817 3628 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3629 status = nfserr_resource;
3630 break;
1da177e4 3631 }
1da177e4 3632out:
8a280510 3633 if (status && lock->lk_is_new && lock_sop)
f044ff83 3634 release_lockowner(lock_sop);
3a65588a
BF		 3635 if (lock->lk_replay_owner) {
3636 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3637 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3638 }
1da177e4
LT		 3639 nfs4_unlock_state();
3640 return status;
3641}
3642
55ef1274
BF		 3643/*
3644 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3645 * so we do a temporary open here just to get an open file to pass to
3646 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3647 * inode operation.)
3648 */
3649static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3650{
3651 struct file *file;
3652 int err;
3653
3654 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3655 if (err)
3656 return err;
3657 err = vfs_test_lock(file, lock);
3658 nfsd_close(file);
3659 return err;
3660}
3661
1da177e4
LT		 3662/*
3663 * LOCKT operation
3664 */
b37ad28b 3665__be32
ca364317
BF		 3666nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3667 struct nfsd4_lockt *lockt)
1da177e4
LT		 3668{
3669 struct inode *inode;
1da177e4 3670 struct file_lock file_lock;
fd85b817 3671 int error;
b37ad28b 3672 __be32 status;
1da177e4 3673
af558e33 3674 if (locks_in_grace())
1da177e4
LT		 3675 return nfserr_grace;
3676
3677 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3678 return nfserr_inval;
3679
3680 lockt->lt_stateowner = NULL;
3681 nfs4_lock_state();
3682
3683 status = nfserr_stale_clientid;
60adfc50 3684 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3685 goto out;
1da177e4 3686
ca364317 3687 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3688 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3689 if (status == nfserr_symlink)
3690 status = nfserr_inval;
3691 goto out;
3692 }
3693
ca364317 3694 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3695 locks_init_lock(&file_lock);
3696 switch (lockt->lt_type) {
3697 case NFS4_READ_LT:
3698 case NFS4_READW_LT:
3699 file_lock.fl_type = F_RDLCK;
3700 break;
3701 case NFS4_WRITE_LT:
3702 case NFS4_WRITEW_LT:
3703 file_lock.fl_type = F_WRLCK;
3704 break;
3705 default:
2fdada03 3706 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3707 status = nfserr_inval;
3708 goto out;
3709 }
3710
3711 lockt->lt_stateowner = find_lockstateowner_str(inode,
3712 &lockt->lt_clientid, &lockt->lt_owner);
3713 if (lockt->lt_stateowner)
3714 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3715 file_lock.fl_pid = current->tgid;
3716 file_lock.fl_flags = FL_POSIX;
3717
3718 file_lock.fl_start = lockt->lt_offset;
87df4de8 3719 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3720
3721 nfs4_transform_lock_offset(&file_lock);
3722
1da177e4 3723 status = nfs_ok;
55ef1274 3724 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3725 if (error) {
3726 status = nfserrno(error);
3727 goto out;
3728 }
9d6a8c5c 3729 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3730 status = nfserr_denied;
9d6a8c5c 3731 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3732 }
3733out:
3734 nfs4_unlock_state();
3735 return status;
3736}
3737
b37ad28b 3738__be32
ca364317 3739nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3740 struct nfsd4_locku *locku)
1da177e4
LT		 3741{
3742 struct nfs4_stateid *stp;
3743 struct file *filp = NULL;
3744 struct file_lock file_lock;
b37ad28b 3745 __be32 status;
b8dd7b9a 3746 int err;
1da177e4
LT		 3747
3748 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3749 (long long) locku->lu_offset,
3750 (long long) locku->lu_length);
3751
3752 if (check_lock_length(locku->lu_offset, locku->lu_length))
3753 return nfserr_inval;
3754
3755 nfs4_lock_state();
3756
dd453dfd 3757 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3758 locku->lu_seqid,
3759 &locku->lu_stateid,
f3362737 3760 LOCK_STATE,
1da177e4
LT		 3761 &locku->lu_stateowner, &stp, NULL)))
3762 goto out;
3763
3764 filp = stp->st_vfs_file;
3765 BUG_ON(!filp);
3766 locks_init_lock(&file_lock);
3767 file_lock.fl_type = F_UNLCK;
3768 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3769 file_lock.fl_pid = current->tgid;
3770 file_lock.fl_file = filp;
3771 file_lock.fl_flags = FL_POSIX;
d5b9026a 3772 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3773 file_lock.fl_start = locku->lu_offset;
3774
87df4de8 3775 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3776 nfs4_transform_lock_offset(&file_lock);
3777
3778 /*
3779 * Try to unlock the file in the VFS.
3780 */
fd85b817 3781 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3782 if (err) {
fd85b817 3783 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3784 goto out_nfserr;
3785 }
3786 /*
3787 * OK, unlock succeeded; the only thing left to do is update the stateid.
3788 */
3789 update_stateid(&stp->st_stateid);
3790 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3791
3792out:
f2327d9a 3793 if (locku->lu_stateowner) {
1da177e4 3794 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3795 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3796 }
1da177e4
LT		 3797 nfs4_unlock_state();
3798 return status;
3799
3800out_nfserr:
b8dd7b9a 3801 status = nfserrno(err);
1da177e4
LT		 3802 goto out;
3803}
3804
3805/*
3806 * returns
3807 * 1: locks held by lockowner
3808 * 0: no locks held by lockowner
3809 */
3810static int
3811check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3812{
3813 struct file_lock **flpp;
7eaa36e2 3814 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3815 int status = 0;
3816
3817 lock_kernel();
3818 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3819 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3820 status = 1;
3821 goto out;
796dadfd 3822 }
1da177e4
LT		 3823 }
3824out:
3825 unlock_kernel();
3826 return status;
3827}
3828
b37ad28b 3829__be32
b591480b
BF		 3830nfsd4_release_lockowner(struct svc_rqst *rqstp,
3831 struct nfsd4_compound_state *cstate,
3832 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3833{
3834 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3835 struct nfs4_stateowner *sop;
3836 struct nfs4_stateid *stp;
1da177e4 3837 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3838 struct list_head matches;
3839 int i;
b37ad28b 3840 __be32 status;
1da177e4
LT		 3841
3842 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3843 clid->cl_boot, clid->cl_id);
3844
3845 /* XXX check for lease expiration */
3846
3847 status = nfserr_stale_clientid;
849823c5 3848 if (STALE_CLIENTID(clid))
1da177e4 3849 return status;
1da177e4
LT		 3850
3851 nfs4_lock_state();
3852
3e9e3dbe
N		 3853 status = nfserr_locks_held;
3854 /* XXX: we're doing a linear search through all the lockowners.
3855 * Yipes! For now we'll just hope clients aren't really using
3856 * release_lockowner much, but eventually we have to fix these
3857 * data structures. */
3858 INIT_LIST_HEAD(&matches);
3859 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3860 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3861 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3862 continue;
3863 list_for_each_entry(stp, &sop->so_stateids,
3864 st_perstateowner) {
3865 if (check_for_locks(stp->st_vfs_file, sop))
3866 goto out;
3867 /* Note: so_perclient unused for lockowners,
3868 * so it's OK to fool with here. */
3869 list_add(&sop->so_perclient, &matches);
3870 }
1da177e4 3871 }
3e9e3dbe
N		 3872 }
3873 /* Clients probably won't expect us to return with some (but not all)
3874 * of the lockowner state released; so don't release any until all
3875 * have been checked. */
3876 status = nfs_ok;
0fa822e4
N		 3877 while (!list_empty(&matches)) {
3878 sop = list_entry(matches.next, struct nfs4_stateowner,
3879 so_perclient);
3880 /* unhash_stateowner deletes so_perclient only
3881 * for openowners. */
3882 list_del(&sop->so_perclient);
f044ff83 3883 release_lockowner(sop);
1da177e4
LT		 3884 }
3885out:
3886 nfs4_unlock_state();
3887 return status;
3888}
3889
3890static inline struct nfs4_client_reclaim *
a55370a3 3891alloc_reclaim(void)
1da177e4 3892{
a55370a3 3893 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3894}
3895
c7b9a459 3896int
a1bcecd2 3897nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3898{
3899 unsigned int strhashval = clientstr_hashval(name);
3900 struct nfs4_client *clp;
3901
a1bcecd2 3902 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3903 return clp ? 1 : 0;
3904}
3905
1da177e4
LT		 3906/*
3907 * failure => all reset bets are off, nfserr_no_grace...
3908 */
190e4fbf
N		 3909int
3910nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3911{
3912 unsigned int strhashval;
3913 struct nfs4_client_reclaim *crp = NULL;
3914
a55370a3
N		 3915 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3916 crp = alloc_reclaim();
1da177e4
LT		 3917 if (!crp)
3918 return 0;
a55370a3 3919 strhashval = clientstr_hashval(name);
1da177e4
LT		 3920 INIT_LIST_HEAD(&crp->cr_strhash);
3921 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3922 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3923 reclaim_str_hashtbl_size++;
3924 return 1;
3925}
3926
3927static void
3928nfs4_release_reclaim(void)
3929{
3930 struct nfs4_client_reclaim *crp = NULL;
3931 int i;
3932
1da177e4
LT		 3933 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3934 while (!list_empty(&reclaim_str_hashtbl[i])) {
3935 crp = list_entry(reclaim_str_hashtbl[i].next,
3936 struct nfs4_client_reclaim, cr_strhash);
3937 list_del(&crp->cr_strhash);
1da177e4
LT		 3938 kfree(crp);
3939 reclaim_str_hashtbl_size--;
3940 }
3941 }
3942 BUG_ON(reclaim_str_hashtbl_size);
3943}
3944
3945/*
3946 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3947static struct nfs4_client_reclaim *
1da177e4
LT		 3948nfs4_find_reclaim_client(clientid_t *clid)
3949{
3950 unsigned int strhashval;
3951 struct nfs4_client *clp;
3952 struct nfs4_client_reclaim *crp = NULL;
3953
3954
3955 /* find clientid in conf_id_hashtbl */
3956 clp = find_confirmed_client(clid);
3957 if (clp == NULL)
3958 return NULL;
3959
a55370a3
N		 3960 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3961 clp->cl_name.len, clp->cl_name.data,
3962 clp->cl_recdir);
1da177e4
LT		 3963
3964 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3965 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3966 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3967 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3968 return crp;
3969 }
3970 }
3971 return NULL;
3972}
3973
3974/*
3975* Called from OPEN. Look for clientid in reclaim list.
3976*/
b37ad28b 3977__be32
1da177e4
LT		 3978nfs4_check_open_reclaim(clientid_t *clid)
3979{
dfc83565 3980 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3981}
3982
ac4d8ff2 3983/* initialization to perform at module load time: */
1da177e4 3984
e8ff2a84 3985int
ac4d8ff2 3986nfs4_state_init(void)
1da177e4 3987{
e8ff2a84 3988 int i, status;
1da177e4 3989
e8ff2a84
BF		 3990 status = nfsd4_init_slabs();
3991 if (status)
3992 return status;
1da177e4
LT		 3993 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3994 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3995 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3996 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3997 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3998 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3999 }
5282fd72
ME		 4000 for (i = 0; i < SESSION_HASH_SIZE; i++)
4001 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 4002 for (i = 0; i < FILE_HASH_SIZE; i++) {
4003 INIT_LIST_HEAD(&file_hashtbl[i]);
4004 }
4005 for (i = 0; i < OWNER_HASH_SIZE; i++) {
4006 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
4007 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
4008 }
4009 for (i = 0; i < STATEID_HASH_SIZE; i++) {
4010 INIT_LIST_HEAD(&stateid_hashtbl[i]);
4011 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
4012 }
4013 for (i = 0; i < LOCK_HASH_SIZE; i++) {
4014 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
4015 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
4016 }
1da177e4 4017 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 4018 INIT_LIST_HEAD(&close_lru);
4019 INIT_LIST_HEAD(&client_lru);
4020 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 4021 reclaim_str_hashtbl_size = 0;
e8ff2a84 4022 return 0;
ac4d8ff2
N		 4023}
4024
190e4fbf
N		 4025static void
4026nfsd4_load_reboot_recovery_data(void)
4027{
4028 int status;
4029
0964a3d3
N		 4030 nfs4_lock_state();
4031 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 4032 status = nfsd4_recdir_load();
0964a3d3 4033 nfs4_unlock_state();
190e4fbf
N		 4034 if (status)
4035 printk("NFSD: Failure reading reboot recovery data\n");
4036}
4037
9a8db97e
ME		 4038unsigned long
4039get_nfs4_grace_period(void)
4040{
4041 return max(user_lease_time, lease_time) * HZ;
4042}
4043
c2f1a551
MS		 4044/*
4045 * Since the lifetime of a delegation isn't limited to that of an open, a
4046 * client may quite reasonably hang on to a delegation as long as it has
4047 * the inode cached. This becomes an obvious problem the first time a
4048 * client's inode cache approaches the size of the server's total memory.
4049 *
4050 * For now we avoid this problem by imposing a hard limit on the number
4051 * of delegations, which varies according to the server's memory size.
4052 */
4053static void
4054set_max_delegations(void)
4055{
4056 /*
4057 * Allow at most 4 delegations per megabyte of RAM. Quick
4058 * estimates suggest that in the worst case (where every delegation
4059 * is for a different inode), a delegation could take about 1.5K,
4060 * giving a worst case usage of about 6% of memory.
4061 */
4062 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4063}
4064
ac4d8ff2
N		 4065/* initialization to perform when the nfsd service is started: */
4066
4067static void
4068__nfs4_state_start(void)
4069{
9a8db97e 4070 unsigned long grace_time;
ac4d8ff2 4071
1da177e4 4072 boot_time = get_seconds();
af558e33 4073 grace_time = get_nfs4_grace_period();
d99a05ad 4074 lease_time = user_lease_time;
af558e33 4075 locks_start_grace(&nfsd4_manager);
9a8db97e
ME		 4076 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
4077 grace_time/HZ);
58da282b 4078 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 4079 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 4080 set_max_delegations();
1da177e4
LT		 4081}
4082
e8ff2a84 4083void
76a3550e 4084nfs4_state_start(void)
1da177e4 4085{
1da177e4 4086 if (nfs4_init)
e8ff2a84 4087 return;
190e4fbf 4088 nfsd4_load_reboot_recovery_data();
76a3550e 4089 __nfs4_state_start();
1da177e4 4090 nfs4_init = 1;
e8ff2a84 4091 return;
1da177e4
LT		 4092}
4093
1da177e4
LT		 4094time_t
4095nfs4_lease_time(void)
4096{
4097 return lease_time;
4098}
4099
4100static void
4101__nfs4_state_shutdown(void)
4102{
4103 int i;
4104 struct nfs4_client *clp = NULL;
4105 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4106 struct list_head *pos, *next, reaplist;
4107
1da177e4
LT		 4108 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4109 while (!list_empty(&conf_id_hashtbl[i])) {
4110 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4111 expire_client(clp);
4112 }
4113 while (!list_empty(&unconf_str_hashtbl[i])) {
4114 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4115 expire_client(clp);
4116 }
4117 }
4118 INIT_LIST_HEAD(&reaplist);
4119 spin_lock(&recall_lock);
4120 list_for_each_safe(pos, next, &del_recall_lru) {
4121 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4122 list_move(&dp->dl_recall_lru, &reaplist);
4123 }
4124 spin_unlock(&recall_lock);
4125 list_for_each_safe(pos, next, &reaplist) {
4126 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4127 list_del_init(&dp->dl_recall_lru);
4128 unhash_delegation(dp);
4129 }
4130
190e4fbf 4131 nfsd4_shutdown_recdir();
1da177e4 4132 nfs4_init = 0;
1da177e4
LT		 4133}
4134
4135void
4136nfs4_state_shutdown(void)
4137{
5e8d5c29
N		 4138 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4139 destroy_workqueue(laundry_wq);
2c5e7615 4140 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4141 nfs4_lock_state();
4142 nfs4_release_reclaim();
4143 __nfs4_state_shutdown();
1da177e4
LT		 4144 nfs4_unlock_state();
4145}
4146
3dd98a3b
JL		 4147/*
4148 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4149 * accessed when nfsd is starting.
4150 */
0964a3d3
N		 4151static void
4152nfs4_set_recdir(char *recdir)
4153{
0964a3d3 4154 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4155}
4156
4157/*
4158 * Change the NFSv4 recovery directory to recdir.
4159 */
4160int
4161nfs4_reset_recoverydir(char *recdir)
4162{
4163 int status;
a63bb996 4164 struct path path;
0964a3d3 4165
a63bb996 4166 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4167 if (status)
4168 return status;
4169 status = -ENOTDIR;
a63bb996 4170 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4171 nfs4_set_recdir(recdir);
4172 status = 0;
4173 }
a63bb996 4174 path_put(&path);
0964a3d3
N		 4175 return status;
4176}
4177
3dd98a3b
JL		 4178char *
4179nfs4_recoverydir(void)
4180{
4181 return user_recovery_dirname;
4182}
4183
1da177e4
LT		 4184/*
4185 * Called when leasetime is changed.
4186 *
d99a05ad
N		 4187 * The only way the protocol gives us to handle on-the-fly lease changes is to
4188 * simulate a reboot. Instead of doing that, we just wait till the next time
4189 * we start to register any changes in lease time. If the administrator
4190 * really wants to change the lease time *now*, they can go ahead and bring
4191 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL		 4192 *
4193 * user_lease_time is protected by nfsd_mutex since it's only really accessed
4194 * when nfsd is starting
1da177e4
LT		 4195 */
4196void
4197nfs4_reset_lease(time_t leasetime)
4198{
d99a05ad 4199 user_lease_time = leasetime;
1da177e4 4200}
Ubuntu Artful kernel mirror