nfsd: document open share bit tracking
[linux-2.6-block.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4
LT
1/*
2* linux/fs/nfsd/nfs4state.c
3*
4* Copyright (c) 2001 The Regents of the University of Michigan.
5* All rights reserved.
6*
7* Kendrick Smith <kmsmith@umich.edu>
8* Andy Adamson <kandros@umich.edu>
9*
10* Redistribution and use in source and binary forms, with or without
11* modification, are permitted provided that the following conditions
12* are met:
13*
14* 1. Redistributions of source code must retain the above copyright
15* notice, this list of conditions and the following disclaimer.
16* 2. Redistributions in binary form must reproduce the above copyright
17* notice, this list of conditions and the following disclaimer in the
18* documentation and/or other materials provided with the distribution.
19* 3. Neither the name of the University nor the names of its
20* contributors may be used to endorse or promote products derived
21* from this software without specific prior written permission.
22*
23* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34*
35*/
36
37#include <linux/param.h>
38#include <linux/major.h>
39#include <linux/slab.h>
40
41#include <linux/sunrpc/svc.h>
42#include <linux/nfsd/nfsd.h>
43#include <linux/nfsd/cache.h>
aceaf78d 44#include <linux/file.h>
1da177e4
LT
45#include <linux/mount.h>
46#include <linux/workqueue.h>
47#include <linux/smp_lock.h>
48#include <linux/kthread.h>
49#include <linux/nfs4.h>
50#include <linux/nfsd/state.h>
51#include <linux/nfsd/xdr4.h>
0964a3d3 52#include <linux/namei.h>
c2f1a551 53#include <linux/swap.h>
353ab6e9 54#include <linux/mutex.h>
fd85b817 55#include <linux/lockd/bind.h>
9a8db97e 56#include <linux/module.h>
1da177e4
LT
57
58#define NFSDDBG_FACILITY NFSDDBG_PROC
59
60/* Globals */
61static time_t lease_time = 90; /* default lease time */
d99a05ad 62static time_t user_lease_time = 90;
fd39ca9a 63static time_t boot_time;
a76b4319 64static int in_grace = 1;
1da177e4
LT
65static u32 current_ownerid = 1;
66static u32 current_fileid = 1;
67static u32 current_delegid = 1;
68static u32 nfs4_init;
fd39ca9a
N
69static stateid_t zerostateid; /* bits all 0 */
70static stateid_t onestateid; /* bits all 1 */
71
72#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
73#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 74
1da177e4 75/* forward declarations */
fd39ca9a 76static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4
LT
77static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
78static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
0964a3d3
N
79static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
80static void nfs4_set_recdir(char *recdir);
1da177e4
LT
81
82/* Locking:
83 *
353ab6e9 84 * client_mutex:
1da177e4
LT
85 * protects clientid_hashtbl[], clientstr_hashtbl[],
86 * unconfstr_hashtbl[], uncofid_hashtbl[].
87 */
353ab6e9 88static DEFINE_MUTEX(client_mutex);
1da177e4 89
e18b890b
CL
90static struct kmem_cache *stateowner_slab = NULL;
91static struct kmem_cache *file_slab = NULL;
92static struct kmem_cache *stateid_slab = NULL;
93static struct kmem_cache *deleg_slab = NULL;
e60d4398 94
1da177e4
LT
95void
96nfs4_lock_state(void)
97{
353ab6e9 98 mutex_lock(&client_mutex);
1da177e4
LT
99}
100
101void
102nfs4_unlock_state(void)
103{
353ab6e9 104 mutex_unlock(&client_mutex);
1da177e4
LT
105}
106
107static inline u32
108opaque_hashval(const void *ptr, int nbytes)
109{
110 unsigned char *cptr = (unsigned char *) ptr;
111
112 u32 x = 0;
113 while (nbytes--) {
114 x *= 37;
115 x += *cptr++;
116 }
117 return x;
118}
119
120/* forward declarations */
121static void release_stateowner(struct nfs4_stateowner *sop);
122static void release_stateid(struct nfs4_stateid *stp, int flags);
1da177e4
LT
123
124/*
125 * Delegation state
126 */
127
128/* recall_lock protects the del_recall_lru */
34af946a 129static DEFINE_SPINLOCK(recall_lock);
1da177e4
LT
130static struct list_head del_recall_lru;
131
13cd2184
N
132static void
133free_nfs4_file(struct kref *kref)
134{
135 struct nfs4_file *fp = container_of(kref, struct nfs4_file, fi_ref);
136 list_del(&fp->fi_hash);
137 iput(fp->fi_inode);
138 kmem_cache_free(file_slab, fp);
139}
140
141static inline void
142put_nfs4_file(struct nfs4_file *fi)
143{
144 kref_put(&fi->fi_ref, free_nfs4_file);
145}
146
147static inline void
148get_nfs4_file(struct nfs4_file *fi)
149{
150 kref_get(&fi->fi_ref);
151}
152
ef0f3390 153static int num_delegations;
c2f1a551 154unsigned int max_delegations;
ef0f3390
N
155
156/*
157 * Open owner state (share locks)
158 */
159
160/* hash tables for nfs4_stateowner */
161#define OWNER_HASH_BITS 8
162#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
163#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
164
165#define ownerid_hashval(id) \
166 ((id) & OWNER_HASH_MASK)
167#define ownerstr_hashval(clientid, ownername) \
168 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
169
170static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
171static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
172
173/* hash table for nfs4_file */
174#define FILE_HASH_BITS 8
175#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
176#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
177/* hash table for (open)nfs4_stateid */
178#define STATEID_HASH_BITS 10
179#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
180#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
181
182#define file_hashval(x) \
183 hash_ptr(x, FILE_HASH_BITS)
184#define stateid_hashval(owner_id, file_id) \
185 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
186
187static struct list_head file_hashtbl[FILE_HASH_SIZE];
188static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
189
1da177e4
LT
190static struct nfs4_delegation *
191alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
192{
193 struct nfs4_delegation *dp;
194 struct nfs4_file *fp = stp->st_file;
195 struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
196
197 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS
198 if (fp->fi_had_conflict)
199 return NULL;
c2f1a551 200 if (num_delegations > max_delegations)
ef0f3390 201 return NULL;
5b2d21c1
N
202 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
203 if (dp == NULL)
1da177e4 204 return dp;
ef0f3390 205 num_delegations++;
ea1da636
N
206 INIT_LIST_HEAD(&dp->dl_perfile);
207 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT
208 INIT_LIST_HEAD(&dp->dl_recall_lru);
209 dp->dl_client = clp;
13cd2184 210 get_nfs4_file(fp);
1da177e4
LT
211 dp->dl_file = fp;
212 dp->dl_flock = NULL;
213 get_file(stp->st_vfs_file);
214 dp->dl_vfs_file = stp->st_vfs_file;
215 dp->dl_type = type;
216 dp->dl_recall.cbr_dp = NULL;
217 dp->dl_recall.cbr_ident = cb->cb_ident;
218 dp->dl_recall.cbr_trunc = 0;
219 dp->dl_stateid.si_boot = boot_time;
220 dp->dl_stateid.si_stateownerid = current_delegid++;
221 dp->dl_stateid.si_fileid = 0;
222 dp->dl_stateid.si_generation = 0;
223 dp->dl_fhlen = current_fh->fh_handle.fh_size;
224 memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
225 current_fh->fh_handle.fh_size);
226 dp->dl_time = 0;
227 atomic_set(&dp->dl_count, 1);
ea1da636
N
228 list_add(&dp->dl_perfile, &fp->fi_delegations);
229 list_add(&dp->dl_perclnt, &clp->cl_delegations);
1da177e4
LT
230 return dp;
231}
232
233void
234nfs4_put_delegation(struct nfs4_delegation *dp)
235{
236 if (atomic_dec_and_test(&dp->dl_count)) {
237 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 238 put_nfs4_file(dp->dl_file);
5b2d21c1 239 kmem_cache_free(deleg_slab, dp);
ef0f3390 240 num_delegations--;
1da177e4
LT
241 }
242}
243
244/* Remove the associated file_lock first, then remove the delegation.
245 * lease_modify() is called to remove the FS_LEASE file_lock from
246 * the i_flock list, eventually calling nfsd's lock_manager
247 * fl_release_callback.
248 */
249static void
250nfs4_close_delegation(struct nfs4_delegation *dp)
251{
252 struct file *filp = dp->dl_vfs_file;
253
254 dprintk("NFSD: close_delegation dp %p\n",dp);
255 dp->dl_vfs_file = NULL;
256 /* The following nfsd_close may not actually close the file,
257 * but we want to remove the lease in any case. */
c907132d 258 if (dp->dl_flock)
a9933cea 259 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 260 nfsd_close(filp);
1da177e4
LT
261}
262
263/* Called under the state lock. */
264static void
265unhash_delegation(struct nfs4_delegation *dp)
266{
ea1da636
N
267 list_del_init(&dp->dl_perfile);
268 list_del_init(&dp->dl_perclnt);
1da177e4
LT
269 spin_lock(&recall_lock);
270 list_del_init(&dp->dl_recall_lru);
271 spin_unlock(&recall_lock);
272 nfs4_close_delegation(dp);
273 nfs4_put_delegation(dp);
274}
275
276/*
277 * SETCLIENTID state
278 */
279
280/* Hash tables for nfs4_clientid state */
281#define CLIENT_HASH_BITS 4
282#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
283#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
284
285#define clientid_hashval(id) \
286 ((id) & CLIENT_HASH_MASK)
a55370a3
N
287#define clientstr_hashval(name) \
288 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT
289/*
290 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
291 * used in reboot/reset lease grace period processing
292 *
293 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
294 * setclientid_confirmed info.
295 *
296 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
297 * setclientid info.
298 *
299 * client_lru holds client queue ordered by nfs4_client.cl_time
300 * for lease renewal.
301 *
302 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
303 * for last close replay.
304 */
305static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
306static int reclaim_str_hashtbl_size = 0;
307static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
308static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
309static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
310static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
311static struct list_head client_lru;
312static struct list_head close_lru;
313
314static inline void
315renew_client(struct nfs4_client *clp)
316{
317 /*
318 * Move client to the end to the LRU list.
319 */
320 dprintk("renewing client (clientid %08x/%08x)\n",
321 clp->cl_clientid.cl_boot,
322 clp->cl_clientid.cl_id);
323 list_move_tail(&clp->cl_lru, &client_lru);
324 clp->cl_time = get_seconds();
325}
326
327/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
328static int
329STALE_CLIENTID(clientid_t *clid)
330{
331 if (clid->cl_boot == boot_time)
332 return 0;
333 dprintk("NFSD stale clientid (%08x/%08x)\n",
334 clid->cl_boot, clid->cl_id);
335 return 1;
336}
337
338/*
339 * XXX Should we use a slab cache ?
340 * This type of memory management is somewhat inefficient, but we use it
341 * anyway since SETCLIENTID is not a common operation.
342 */
35bba9a3 343static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT
344{
345 struct nfs4_client *clp;
346
35bba9a3
BF
347 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
348 if (clp == NULL)
349 return NULL;
350 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
351 if (clp->cl_name.data == NULL) {
352 kfree(clp);
353 return NULL;
1da177e4 354 }
35bba9a3
BF
355 memcpy(clp->cl_name.data, name.data, name.len);
356 clp->cl_name.len = name.len;
1da177e4
LT
357 return clp;
358}
359
1b1a9b31
BF
360static void
361shutdown_callback_client(struct nfs4_client *clp)
362{
363 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
364
1b1a9b31 365 if (clnt) {
404ec117
BF
366 /*
367 * Callback threads take a reference on the client, so there
368 * should be no outstanding callbacks at this point.
369 */
1b1a9b31
BF
370 clp->cl_callback.cb_client = NULL;
371 rpc_shutdown_client(clnt);
372 }
373}
374
1da177e4
LT
375static inline void
376free_client(struct nfs4_client *clp)
377{
1b1a9b31 378 shutdown_callback_client(clp);
1da177e4
LT
379 if (clp->cl_cred.cr_group_info)
380 put_group_info(clp->cl_cred.cr_group_info);
381 kfree(clp->cl_name.data);
382 kfree(clp);
383}
384
385void
386put_nfs4_client(struct nfs4_client *clp)
387{
388 if (atomic_dec_and_test(&clp->cl_count))
389 free_client(clp);
390}
391
392static void
393expire_client(struct nfs4_client *clp)
394{
395 struct nfs4_stateowner *sop;
396 struct nfs4_delegation *dp;
1da177e4
LT
397 struct list_head reaplist;
398
399 dprintk("NFSD: expire_client cl_count %d\n",
400 atomic_read(&clp->cl_count));
401
1da177e4
LT
402 INIT_LIST_HEAD(&reaplist);
403 spin_lock(&recall_lock);
ea1da636
N
404 while (!list_empty(&clp->cl_delegations)) {
405 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT
406 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
407 dp->dl_flock);
ea1da636 408 list_del_init(&dp->dl_perclnt);
1da177e4
LT
409 list_move(&dp->dl_recall_lru, &reaplist);
410 }
411 spin_unlock(&recall_lock);
412 while (!list_empty(&reaplist)) {
413 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
414 list_del_init(&dp->dl_recall_lru);
415 unhash_delegation(dp);
416 }
417 list_del(&clp->cl_idhash);
418 list_del(&clp->cl_strhash);
419 list_del(&clp->cl_lru);
ea1da636
N
420 while (!list_empty(&clp->cl_openowners)) {
421 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
1da177e4
LT
422 release_stateowner(sop);
423 }
424 put_nfs4_client(clp);
425}
426
35bba9a3
BF
427static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir)
428{
1da177e4
LT
429 struct nfs4_client *clp;
430
35bba9a3
BF
431 clp = alloc_client(name);
432 if (clp == NULL)
433 return NULL;
a55370a3 434 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
1da177e4
LT
435 atomic_set(&clp->cl_count, 1);
436 atomic_set(&clp->cl_callback.cb_set, 0);
1da177e4
LT
437 INIT_LIST_HEAD(&clp->cl_idhash);
438 INIT_LIST_HEAD(&clp->cl_strhash);
ea1da636
N
439 INIT_LIST_HEAD(&clp->cl_openowners);
440 INIT_LIST_HEAD(&clp->cl_delegations);
1da177e4 441 INIT_LIST_HEAD(&clp->cl_lru);
1da177e4
LT
442 return clp;
443}
444
35bba9a3
BF
445static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
446{
447 memcpy(target->cl_verifier.data, source->data,
448 sizeof(target->cl_verifier.data));
1da177e4
LT
449}
450
35bba9a3
BF
451static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
452{
1da177e4
LT
453 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
454 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
455}
456
35bba9a3
BF
457static void copy_cred(struct svc_cred *target, struct svc_cred *source)
458{
1da177e4
LT
459 target->cr_uid = source->cr_uid;
460 target->cr_gid = source->cr_gid;
461 target->cr_group_info = source->cr_group_info;
462 get_group_info(target->cr_group_info);
463}
464
35bba9a3 465static int same_name(const char *n1, const char *n2)
599e0a22 466{
a55370a3 467 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT
468}
469
470static int
599e0a22
BF
471same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
472{
473 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT
474}
475
476static int
599e0a22
BF
477same_clid(clientid_t *cl1, clientid_t *cl2)
478{
479 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT
480}
481
482/* XXX what about NGROUP */
483static int
599e0a22
BF
484same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
485{
486 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT
487}
488
5ec7b46c
BF
489static void gen_clid(struct nfs4_client *clp)
490{
491 static u32 current_clientid = 1;
492
1da177e4
LT
493 clp->cl_clientid.cl_boot = boot_time;
494 clp->cl_clientid.cl_id = current_clientid++;
495}
496
deda2faa
BF
497static void gen_confirm(struct nfs4_client *clp)
498{
499 static u32 i;
500 u32 *p;
1da177e4 501
1da177e4 502 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF
503 *p++ = get_seconds();
504 *p++ = i++;
1da177e4
LT
505}
506
35bba9a3
BF
507static int check_name(struct xdr_netobj name)
508{
1da177e4
LT
509 if (name.len == 0)
510 return 0;
511 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 512 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT
513 return 0;
514 }
515 return 1;
516}
517
fd39ca9a 518static void
1da177e4
LT
519add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
520{
521 unsigned int idhashval;
522
523 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
524 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
525 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
526 list_add_tail(&clp->cl_lru, &client_lru);
527 clp->cl_time = get_seconds();
528}
529
fd39ca9a 530static void
1da177e4
LT
531move_to_confirmed(struct nfs4_client *clp)
532{
533 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
534 unsigned int strhashval;
535
536 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
537 list_del_init(&clp->cl_strhash);
f116629d 538 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 539 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT
540 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
541 renew_client(clp);
542}
543
544static struct nfs4_client *
545find_confirmed_client(clientid_t *clid)
546{
547 struct nfs4_client *clp;
548 unsigned int idhashval = clientid_hashval(clid->cl_id);
549
550 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 551 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
552 return clp;
553 }
554 return NULL;
555}
556
557static struct nfs4_client *
558find_unconfirmed_client(clientid_t *clid)
559{
560 struct nfs4_client *clp;
561 unsigned int idhashval = clientid_hashval(clid->cl_id);
562
563 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 564 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT
565 return clp;
566 }
567 return NULL;
568}
569
28ce6054
N
570static struct nfs4_client *
571find_confirmed_client_by_str(const char *dname, unsigned int hashval)
572{
573 struct nfs4_client *clp;
574
575 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
576 if (same_name(clp->cl_recdir, dname))
577 return clp;
578 }
579 return NULL;
580}
581
582static struct nfs4_client *
583find_unconfirmed_client_by_str(const char *dname, unsigned int hashval)
584{
585 struct nfs4_client *clp;
586
587 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
588 if (same_name(clp->cl_recdir, dname))
589 return clp;
590 }
591 return NULL;
592}
593
1da177e4
LT
594/* a helper function for parse_callback */
595static int
596parse_octet(unsigned int *lenp, char **addrp)
597{
598 unsigned int len = *lenp;
599 char *p = *addrp;
600 int n = -1;
601 char c;
602
603 for (;;) {
604 if (!len)
605 break;
606 len--;
607 c = *p++;
608 if (c == '.')
609 break;
610 if ((c < '0') || (c > '9')) {
611 n = -1;
612 break;
613 }
614 if (n < 0)
615 n = 0;
616 n = (n * 10) + (c - '0');
617 if (n > 255) {
618 n = -1;
619 break;
620 }
621 }
622 *lenp = len;
623 *addrp = p;
624 return n;
625}
626
627/* parse and set the setclientid ipv4 callback address */
fd39ca9a 628static int
1da177e4
LT
629parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
630{
631 int temp = 0;
632 u32 cbaddr = 0;
633 u16 cbport = 0;
634 u32 addrlen = addr_len;
635 char *addr = addr_val;
636 int i, shift;
637
638 /* ipaddress */
639 shift = 24;
640 for(i = 4; i > 0 ; i--) {
641 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
642 return 0;
643 }
644 cbaddr |= (temp << shift);
645 if (shift > 0)
646 shift -= 8;
647 }
648 *cbaddrp = cbaddr;
649
650 /* port */
651 shift = 8;
652 for(i = 2; i > 0 ; i--) {
653 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
654 return 0;
655 }
656 cbport |= (temp << shift);
657 if (shift > 0)
658 shift -= 8;
659 }
660 *cbportp = cbport;
661 return 1;
662}
663
fd39ca9a 664static void
1da177e4
LT
665gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
666{
667 struct nfs4_callback *cb = &clp->cl_callback;
668
669 /* Currently, we only support tcp for the callback channel */
670 if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
671 goto out_err;
672
673 if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
674 &cb->cb_addr, &cb->cb_port)))
675 goto out_err;
676 cb->cb_prog = se->se_callback_prog;
677 cb->cb_ident = se->se_callback_ident;
1da177e4
LT
678 return;
679out_err:
849823c5 680 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT
681 "will not receive delegations\n",
682 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
683
1da177e4
LT
684 return;
685}
686
b37ad28b 687__be32
b591480b
BF
688nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
689 struct nfsd4_setclientid *setclid)
1da177e4 690{
27459f09 691 struct sockaddr_in *sin = svc_addr_in(rqstp);
1da177e4
LT
692 struct xdr_netobj clname = {
693 .len = setclid->se_namelen,
694 .data = setclid->se_name,
695 };
696 nfs4_verifier clverifier = setclid->se_verf;
697 unsigned int strhashval;
28ce6054 698 struct nfs4_client *conf, *unconf, *new;
b37ad28b 699 __be32 status;
a55370a3 700 char dname[HEXDIR_LEN];
1da177e4 701
1da177e4 702 if (!check_name(clname))
73aea4ec 703 return nfserr_inval;
1da177e4 704
a55370a3
N
705 status = nfs4_make_rec_clidname(dname, &clname);
706 if (status)
73aea4ec 707 return status;
a55370a3 708
1da177e4
LT
709 /*
710 * XXX The Duplicate Request Cache (DRC) has been checked (??)
711 * We get here on a DRC miss.
712 */
713
a55370a3 714 strhashval = clientstr_hashval(dname);
1da177e4 715
1da177e4 716 nfs4_lock_state();
28ce6054
N
717 conf = find_confirmed_client_by_str(dname, strhashval);
718 if (conf) {
a186e767 719 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 720 status = nfserr_clid_inuse;
599e0a22 721 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)
27459f09 722 || conf->cl_addr != sin->sin_addr.s_addr) {
21315edd
BF
723 dprintk("NFSD: setclientid: string in use by client"
724 "at %u.%u.%u.%u\n", NIPQUAD(conf->cl_addr));
1da177e4
LT
725 goto out;
726 }
1da177e4 727 }
a186e767
BF
728 /*
729 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
730 * has a description of SETCLIENTID request processing consisting
731 * of 5 bullet points, labeled as CASE0 - CASE4 below.
732 */
28ce6054 733 unconf = find_unconfirmed_client_by_str(dname, strhashval);
1da177e4
LT
734 status = nfserr_resource;
735 if (!conf) {
a186e767
BF
736 /*
737 * RFC 3530 14.2.33 CASE 4:
738 * placed first, because it is the normal case
1da177e4
LT
739 */
740 if (unconf)
741 expire_client(unconf);
a55370a3
N
742 new = create_client(clname, dname);
743 if (new == NULL)
1da177e4 744 goto out;
1da177e4 745 gen_clid(new);
599e0a22 746 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 747 /*
a186e767
BF
748 * RFC 3530 14.2.33 CASE 1:
749 * probable callback update
1da177e4 750 */
31f4a6c1
N
751 if (unconf) {
752 /* Note this is removing unconfirmed {*x***},
753 * which is stronger than RFC recommended {vxc**}.
754 * This has the advantage that there is at most
755 * one {*x***} in either list at any time.
756 */
757 expire_client(unconf);
1da177e4 758 }
a55370a3
N
759 new = create_client(clname, dname);
760 if (new == NULL)
1da177e4 761 goto out;
1da177e4 762 copy_clid(new, conf);
1da177e4
LT
763 } else if (!unconf) {
764 /*
a186e767
BF
765 * RFC 3530 14.2.33 CASE 2:
766 * probable client reboot; state will be removed if
767 * confirmed.
1da177e4 768 */
a55370a3
N
769 new = create_client(clname, dname);
770 if (new == NULL)
1da177e4 771 goto out;
1da177e4 772 gen_clid(new);
49ba8781 773 } else {
a186e767
BF
774 /*
775 * RFC 3530 14.2.33 CASE 3:
776 * probable client reboot; state will be removed if
777 * confirmed.
1da177e4
LT
778 */
779 expire_client(unconf);
a55370a3
N
780 new = create_client(clname, dname);
781 if (new == NULL)
1da177e4 782 goto out;
1da177e4 783 gen_clid(new);
1da177e4 784 }
c175b83c
BF
785 copy_verf(new, &clverifier);
786 new->cl_addr = sin->sin_addr.s_addr;
787 copy_cred(&new->cl_cred, &rqstp->rq_cred);
788 gen_confirm(new);
789 gen_callback(new, setclid);
790 add_to_unconfirmed(new, strhashval);
1da177e4
LT
791 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
792 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
793 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
794 status = nfs_ok;
795out:
796 nfs4_unlock_state();
797 return status;
798}
799
800
801/*
a186e767
BF
802 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
803 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
804 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 805 */
b37ad28b 806__be32
b591480b
BF
807nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
808 struct nfsd4_compound_state *cstate,
809 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 810{
27459f09 811 struct sockaddr_in *sin = svc_addr_in(rqstp);
21ab45a4 812 struct nfs4_client *conf, *unconf;
1da177e4
LT
813 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
814 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 815 __be32 status;
1da177e4
LT
816
817 if (STALE_CLIENTID(clid))
818 return nfserr_stale_clientid;
819 /*
820 * XXX The Duplicate Request Cache (DRC) has been checked (??)
821 * We get here on a DRC miss.
822 */
823
824 nfs4_lock_state();
21ab45a4
N
825
826 conf = find_confirmed_client(clid);
827 unconf = find_unconfirmed_client(clid);
828
829 status = nfserr_clid_inuse;
27459f09 830 if (conf && conf->cl_addr != sin->sin_addr.s_addr)
21ab45a4 831 goto out;
27459f09 832 if (unconf && unconf->cl_addr != sin->sin_addr.s_addr)
21ab45a4
N
833 goto out;
834
a186e767
BF
835 /*
836 * section 14.2.34 of RFC 3530 has a description of
837 * SETCLIENTID_CONFIRM request processing consisting
838 * of 4 bullet points, labeled as CASE1 - CASE4 below.
839 */
366e0c1d 840 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF
841 /*
842 * RFC 3530 14.2.34 CASE 1:
843 * callback update
844 */
599e0a22 845 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT
846 status = nfserr_clid_inuse;
847 else {
1a69c179
N
848 /* XXX: We just turn off callbacks until we can handle
849 * change request correctly. */
cb36d634 850 atomic_set(&conf->cl_callback.cb_set, 0);
21ab45a4 851 gen_confirm(conf);
46309029 852 nfsd4_remove_clid_dir(unconf);
1a69c179 853 expire_client(unconf);
1da177e4 854 status = nfs_ok;
1a69c179 855
1da177e4 856 }
f3aba4e5 857 } else if (conf && !unconf) {
a186e767
BF
858 /*
859 * RFC 3530 14.2.34 CASE 2:
860 * probable retransmitted request; play it safe and
861 * do nothing.
7c79f737 862 */
599e0a22 863 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 864 status = nfserr_clid_inuse;
21ab45a4 865 else
1da177e4 866 status = nfs_ok;
7c79f737 867 } else if (!conf && unconf
599e0a22 868 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF
869 /*
870 * RFC 3530 14.2.34 CASE 3:
871 * Normal case; new or rebooted client:
7c79f737 872 */
599e0a22 873 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT
874 status = nfserr_clid_inuse;
875 } else {
1a69c179
N
876 unsigned int hash =
877 clientstr_hashval(unconf->cl_recdir);
878 conf = find_confirmed_client_by_str(unconf->cl_recdir,
879 hash);
880 if (conf) {
c7b9a459 881 nfsd4_remove_clid_dir(conf);
1a69c179
N
882 expire_client(conf);
883 }
1da177e4 884 move_to_confirmed(unconf);
21ab45a4 885 conf = unconf;
46f8a64b 886 nfsd4_probe_callback(conf);
1a69c179 887 status = nfs_ok;
1da177e4 888 }
599e0a22
BF
889 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
890 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 891 &confirm)))) {
a186e767
BF
892 /*
893 * RFC 3530 14.2.34 CASE 4:
894 * Client probably hasn't noticed that we rebooted yet.
7c79f737 895 */
1da177e4 896 status = nfserr_stale_clientid;
7c79f737 897 } else {
08e8987c
N
898 /* check that we have hit one of the cases...*/
899 status = nfserr_clid_inuse;
900 }
1da177e4 901out:
1da177e4
LT
902 nfs4_unlock_state();
903 return status;
904}
905
1da177e4
LT
906/* OPEN Share state helper functions */
907static inline struct nfs4_file *
908alloc_init_file(struct inode *ino)
909{
910 struct nfs4_file *fp;
911 unsigned int hashval = file_hashval(ino);
912
e60d4398
N
913 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
914 if (fp) {
13cd2184 915 kref_init(&fp->fi_ref);
1da177e4 916 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N
917 INIT_LIST_HEAD(&fp->fi_stateids);
918 INIT_LIST_HEAD(&fp->fi_delegations);
1da177e4
LT
919 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
920 fp->fi_inode = igrab(ino);
921 fp->fi_id = current_fileid++;
47f9940c 922 fp->fi_had_conflict = false;
1da177e4
LT
923 return fp;
924 }
925 return NULL;
926}
927
e60d4398 928static void
e18b890b 929nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 930{
e60d4398
N
931 if (*slab == NULL)
932 return;
1a1d92c1 933 kmem_cache_destroy(*slab);
e60d4398 934 *slab = NULL;
1da177e4
LT
935}
936
e8ff2a84 937void
1da177e4
LT
938nfsd4_free_slabs(void)
939{
e60d4398
N
940 nfsd4_free_slab(&stateowner_slab);
941 nfsd4_free_slab(&file_slab);
5ac049ac 942 nfsd4_free_slab(&stateid_slab);
5b2d21c1 943 nfsd4_free_slab(&deleg_slab);
e60d4398 944}
1da177e4 945
e60d4398
N
946static int
947nfsd4_init_slabs(void)
948{
949 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 950 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N
951 if (stateowner_slab == NULL)
952 goto out_nomem;
953 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 954 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N
955 if (file_slab == NULL)
956 goto out_nomem;
5ac049ac 957 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 958 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N
959 if (stateid_slab == NULL)
960 goto out_nomem;
5b2d21c1 961 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 962 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N
963 if (deleg_slab == NULL)
964 goto out_nomem;
e60d4398
N
965 return 0;
966out_nomem:
967 nfsd4_free_slabs();
968 dprintk("nfsd4: out of memory while initializing nfsv4\n");
969 return -ENOMEM;
1da177e4
LT
970}
971
972void
973nfs4_free_stateowner(struct kref *kref)
974{
975 struct nfs4_stateowner *sop =
976 container_of(kref, struct nfs4_stateowner, so_ref);
977 kfree(sop->so_owner.data);
978 kmem_cache_free(stateowner_slab, sop);
979}
980
981static inline struct nfs4_stateowner *
982alloc_stateowner(struct xdr_netobj *owner)
983{
984 struct nfs4_stateowner *sop;
985
986 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
987 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
988 memcpy(sop->so_owner.data, owner->data, owner->len);
989 sop->so_owner.len = owner->len;
990 kref_init(&sop->so_ref);
991 return sop;
992 }
993 kmem_cache_free(stateowner_slab, sop);
994 }
995 return NULL;
996}
997
998static struct nfs4_stateowner *
999alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1000 struct nfs4_stateowner *sop;
1001 struct nfs4_replay *rp;
1002 unsigned int idhashval;
1003
1004 if (!(sop = alloc_stateowner(&open->op_owner)))
1005 return NULL;
1006 idhashval = ownerid_hashval(current_ownerid);
1007 INIT_LIST_HEAD(&sop->so_idhash);
1008 INIT_LIST_HEAD(&sop->so_strhash);
1009 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
1010 INIT_LIST_HEAD(&sop->so_stateids);
1011 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT
1012 INIT_LIST_HEAD(&sop->so_close_lru);
1013 sop->so_time = 0;
1014 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1015 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1016 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT
1017 sop->so_is_open_owner = 1;
1018 sop->so_id = current_ownerid++;
1019 sop->so_client = clp;
1020 sop->so_seqid = open->op_seqid;
1021 sop->so_confirmed = 0;
1022 rp = &sop->so_replay;
de1ae286 1023 rp->rp_status = nfserr_serverfault;
1da177e4
LT
1024 rp->rp_buflen = 0;
1025 rp->rp_buf = rp->rp_ibuf;
1026 return sop;
1027}
1028
1029static void
1030release_stateid_lockowners(struct nfs4_stateid *open_stp)
1031{
1032 struct nfs4_stateowner *lock_sop;
1033
ea1da636
N
1034 while (!list_empty(&open_stp->st_lockowners)) {
1035 lock_sop = list_entry(open_stp->st_lockowners.next,
1036 struct nfs4_stateowner, so_perstateid);
1037 /* list_del(&open_stp->st_lockowners); */
1da177e4
LT
1038 BUG_ON(lock_sop->so_is_open_owner);
1039 release_stateowner(lock_sop);
1040 }
1041}
1042
1043static void
1044unhash_stateowner(struct nfs4_stateowner *sop)
1045{
1046 struct nfs4_stateid *stp;
1047
1048 list_del(&sop->so_idhash);
1049 list_del(&sop->so_strhash);
6fa305de 1050 if (sop->so_is_open_owner)
1da177e4 1051 list_del(&sop->so_perclient);
ea1da636
N
1052 list_del(&sop->so_perstateid);
1053 while (!list_empty(&sop->so_stateids)) {
1054 stp = list_entry(sop->so_stateids.next,
1055 struct nfs4_stateid, st_perstateowner);
1da177e4
LT
1056 if (sop->so_is_open_owner)
1057 release_stateid(stp, OPEN_STATE);
1058 else
1059 release_stateid(stp, LOCK_STATE);
1060 }
1061}
1062
1063static void
1064release_stateowner(struct nfs4_stateowner *sop)
1065{
1066 unhash_stateowner(sop);
1067 list_del(&sop->so_close_lru);
1068 nfs4_put_stateowner(sop);
1069}
1070
1071static inline void
1072init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1073 struct nfs4_stateowner *sop = open->op_stateowner;
1074 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1075
1076 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N
1077 INIT_LIST_HEAD(&stp->st_perstateowner);
1078 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT
1079 INIT_LIST_HEAD(&stp->st_perfile);
1080 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1081 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1082 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1083 stp->st_stateowner = sop;
13cd2184 1084 get_nfs4_file(fp);
1da177e4
LT
1085 stp->st_file = fp;
1086 stp->st_stateid.si_boot = boot_time;
1087 stp->st_stateid.si_stateownerid = sop->so_id;
1088 stp->st_stateid.si_fileid = fp->fi_id;
1089 stp->st_stateid.si_generation = 0;
1090 stp->st_access_bmap = 0;
1091 stp->st_deny_bmap = 0;
1092 __set_bit(open->op_share_access, &stp->st_access_bmap);
1093 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1094 stp->st_openstp = NULL;
1da177e4
LT
1095}
1096
1097static void
1098release_stateid(struct nfs4_stateid *stp, int flags)
1099{
1100 struct file *filp = stp->st_vfs_file;
1101
1102 list_del(&stp->st_hash);
1da177e4 1103 list_del(&stp->st_perfile);
ea1da636 1104 list_del(&stp->st_perstateowner);
1da177e4
LT
1105 if (flags & OPEN_STATE) {
1106 release_stateid_lockowners(stp);
1107 stp->st_vfs_file = NULL;
1108 nfsd_close(filp);
1da177e4
LT
1109 } else if (flags & LOCK_STATE)
1110 locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
13cd2184 1111 put_nfs4_file(stp->st_file);
5ac049ac 1112 kmem_cache_free(stateid_slab, stp);
1da177e4
LT
1113}
1114
fd39ca9a 1115static void
1da177e4
LT
1116move_to_close_lru(struct nfs4_stateowner *sop)
1117{
1118 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1119
358dd55a 1120 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT
1121 sop->so_time = get_seconds();
1122}
1123
1da177e4 1124static int
599e0a22
BF
1125same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1126 clientid_t *clid)
1127{
1128 return (sop->so_owner.len == owner->len) &&
1129 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1130 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT
1131}
1132
1133static struct nfs4_stateowner *
1134find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1135{
1136 struct nfs4_stateowner *so = NULL;
1137
1138 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1139 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT
1140 return so;
1141 }
1142 return NULL;
1143}
1144
1145/* search file_hashtbl[] for file */
1146static struct nfs4_file *
1147find_file(struct inode *ino)
1148{
1149 unsigned int hashval = file_hashval(ino);
1150 struct nfs4_file *fp;
1151
1152 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N
1153 if (fp->fi_inode == ino) {
1154 get_nfs4_file(fp);
1da177e4 1155 return fp;
13cd2184 1156 }
1da177e4
LT
1157 }
1158 return NULL;
1159}
1160
8838dc43 1161static inline int access_valid(u32 x)
ba5a6a19 1162{
8838dc43
BF
1163 if (x < NFS4_SHARE_ACCESS_READ)
1164 return 0;
1165 if (x > NFS4_SHARE_ACCESS_BOTH)
1166 return 0;
1167 return 1;
ba5a6a19
BF
1168}
1169
8838dc43 1170static inline int deny_valid(u32 x)
ba5a6a19 1171{
8838dc43
BF
1172 /* Note: unlike access bits, deny bits may be zero. */
1173 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1174}
1da177e4 1175
4f83aa30
BF
1176/*
1177 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1178 * st_{access,deny}_bmap field of the stateid, in order to track not
1179 * only what share bits are currently in force, but also what
1180 * combinations of share bits previous opens have used. This allows us
1181 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1182 * return an error if the client attempt to downgrade to a combination
1183 * of share bits not explicable by closing some of its previous opens.
1184 *
1185 * XXX: This enforcement is actually incomplete, since we don't keep
1186 * track of access/deny bit combinations; so, e.g., we allow:
1187 *
1188 * OPEN allow read, deny write
1189 * OPEN allow both, deny none
1190 * DOWNGRADE allow read, deny none
1191 *
1192 * which we should reject.
1193 */
fd39ca9a 1194static void
1da177e4
LT
1195set_access(unsigned int *access, unsigned long bmap) {
1196 int i;
1197
1198 *access = 0;
1199 for (i = 1; i < 4; i++) {
1200 if (test_bit(i, &bmap))
1201 *access |= i;
1202 }
1203}
1204
fd39ca9a 1205static void
1da177e4
LT
1206set_deny(unsigned int *deny, unsigned long bmap) {
1207 int i;
1208
1209 *deny = 0;
1210 for (i = 0; i < 4; i++) {
1211 if (test_bit(i, &bmap))
1212 *deny |= i ;
1213 }
1214}
1215
1216static int
1217test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1218 unsigned int access, deny;
1219
1220 set_access(&access, stp->st_access_bmap);
1221 set_deny(&deny, stp->st_deny_bmap);
1222 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1223 return 0;
1224 return 1;
1225}
1226
1227/*
1228 * Called to check deny when READ with all zero stateid or
1229 * WRITE with all zero or all one stateid
1230 */
b37ad28b 1231static __be32
1da177e4
LT
1232nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1233{
1234 struct inode *ino = current_fh->fh_dentry->d_inode;
1235 struct nfs4_file *fp;
1236 struct nfs4_stateid *stp;
b37ad28b 1237 __be32 ret;
1da177e4
LT
1238
1239 dprintk("NFSD: nfs4_share_conflict\n");
1240
1241 fp = find_file(ino);
13cd2184
N
1242 if (!fp)
1243 return nfs_ok;
b700949b 1244 ret = nfserr_locked;
1da177e4 1245 /* Search for conflicting share reservations */
13cd2184
N
1246 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1247 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1248 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1249 goto out;
1da177e4 1250 }
13cd2184
N
1251 ret = nfs_ok;
1252out:
1253 put_nfs4_file(fp);
1254 return ret;
1da177e4
LT
1255}
1256
1257static inline void
1258nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1259{
1260 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 1261 drop_file_write_access(filp);
1da177e4
LT
1262 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1263 }
1264}
1265
1266/*
1267 * Recall a delegation
1268 */
1269static int
1270do_recall(void *__dp)
1271{
1272 struct nfs4_delegation *dp = __dp;
1273
47f9940c 1274 dp->dl_file->fi_had_conflict = true;
1da177e4
LT
1275 nfsd4_cb_recall(dp);
1276 return 0;
1277}
1278
1279/*
1280 * Spawn a thread to perform a recall on the delegation represented
1281 * by the lease (file_lock)
1282 *
1283 * Called from break_lease() with lock_kernel() held.
1284 * Note: we assume break_lease will only call this *once* for any given
1285 * lease.
1286 */
1287static
1288void nfsd_break_deleg_cb(struct file_lock *fl)
1289{
1290 struct nfs4_delegation *dp= (struct nfs4_delegation *)fl->fl_owner;
1291 struct task_struct *t;
1292
1293 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1294 if (!dp)
1295 return;
1296
1297 /* We're assuming the state code never drops its reference
1298 * without first removing the lease. Since we're in this lease
1299 * callback (and since the lease code is serialized by the kernel
1300 * lock) we know the server hasn't removed the lease yet, we know
1301 * it's safe to take a reference: */
1302 atomic_inc(&dp->dl_count);
cfdcad4d 1303 atomic_inc(&dp->dl_client->cl_count);
1da177e4
LT
1304
1305 spin_lock(&recall_lock);
1306 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1307 spin_unlock(&recall_lock);
1308
1309 /* only place dl_time is set. protected by lock_kernel*/
1310 dp->dl_time = get_seconds();
1311
0272e1fd
BF
1312 /*
1313 * We don't want the locks code to timeout the lease for us;
1314 * we'll remove it ourself if the delegation isn't returned
1315 * in time.
1316 */
1317 fl->fl_break_time = 0;
1da177e4
LT
1318
1319 t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1320 if (IS_ERR(t)) {
1321 struct nfs4_client *clp = dp->dl_client;
1322
1323 printk(KERN_INFO "NFSD: Callback thread failed for "
1324 "for client (clientid %08x/%08x)\n",
1325 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
cfdcad4d 1326 put_nfs4_client(dp->dl_client);
1da177e4
LT
1327 nfs4_put_delegation(dp);
1328 }
1329}
1330
1331/*
1332 * The file_lock is being reapd.
1333 *
1334 * Called by locks_free_lock() with lock_kernel() held.
1335 */
1336static
1337void nfsd_release_deleg_cb(struct file_lock *fl)
1338{
1339 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1340
1341 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1342
1343 if (!(fl->fl_flags & FL_LEASE) || !dp)
1344 return;
1345 dp->dl_flock = NULL;
1346}
1347
1348/*
1349 * Set the delegation file_lock back pointer.
1350 *
a9933cea 1351 * Called from setlease() with lock_kernel() held.
1da177e4
LT
1352 */
1353static
1354void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1355{
1356 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1357
1358 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1359 if (!dp)
1360 return;
1361 dp->dl_flock = new;
1362}
1363
1364/*
a9933cea 1365 * Called from setlease() with lock_kernel() held
1da177e4
LT
1366 */
1367static
1368int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1369{
1370 struct nfs4_delegation *onlistd =
1371 (struct nfs4_delegation *)onlist->fl_owner;
1372 struct nfs4_delegation *tryd =
1373 (struct nfs4_delegation *)try->fl_owner;
1374
1375 if (onlist->fl_lmops != try->fl_lmops)
1376 return 0;
1377
1378 return onlistd->dl_client == tryd->dl_client;
1379}
1380
1381
1382static
1383int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1384{
1385 if (arg & F_UNLCK)
1386 return lease_modify(onlist, arg);
1387 else
1388 return -EAGAIN;
1389}
1390
fd39ca9a 1391static struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT
1392 .fl_break = nfsd_break_deleg_cb,
1393 .fl_release_private = nfsd_release_deleg_cb,
1394 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1395 .fl_mylease = nfsd_same_client_deleg_cb,
1396 .fl_change = nfsd_change_deleg_cb,
1397};
1398
1399
b37ad28b 1400__be32
1da177e4
LT
1401nfsd4_process_open1(struct nfsd4_open *open)
1402{
1da177e4
LT
1403 clientid_t *clientid = &open->op_clientid;
1404 struct nfs4_client *clp = NULL;
1405 unsigned int strhashval;
1406 struct nfs4_stateowner *sop = NULL;
1407
1da177e4 1408 if (!check_name(open->op_owner))
0f442aa2 1409 return nfserr_inval;
1da177e4
LT
1410
1411 if (STALE_CLIENTID(&open->op_clientid))
1412 return nfserr_stale_clientid;
1413
1414 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1415 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF
1416 open->op_stateowner = sop;
1417 if (!sop) {
1418 /* Make sure the client's lease hasn't expired. */
1da177e4
LT
1419 clp = find_confirmed_client(clientid);
1420 if (clp == NULL)
0f442aa2
BF
1421 return nfserr_expired;
1422 goto renew;
1da177e4 1423 }
0f442aa2
BF
1424 if (!sop->so_confirmed) {
1425 /* Replace unconfirmed owners without checking for replay. */
1426 clp = sop->so_client;
1427 release_stateowner(sop);
1428 open->op_stateowner = NULL;
1429 goto renew;
1430 }
1431 if (open->op_seqid == sop->so_seqid - 1) {
1432 if (sop->so_replay.rp_buflen)
a90b061c 1433 return nfserr_replay_me;
0f442aa2
BF
1434 /* The original OPEN failed so spectacularly
1435 * that we don't even have replay data saved!
1436 * Therefore, we have no choice but to continue
1437 * processing this OPEN; presumably, we'll
1438 * fail again for the same reason.
1439 */
1440 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1441 goto renew;
1442 }
1443 if (open->op_seqid != sop->so_seqid)
1444 return nfserr_bad_seqid;
1da177e4 1445renew:
0f442aa2
BF
1446 if (open->op_stateowner == NULL) {
1447 sop = alloc_init_open_stateowner(strhashval, clp, open);
1448 if (sop == NULL)
1449 return nfserr_resource;
1450 open->op_stateowner = sop;
1451 }
1452 list_del_init(&sop->so_close_lru);
1da177e4 1453 renew_client(sop->so_client);
0f442aa2 1454 return nfs_ok;
1da177e4
LT
1455}
1456
b37ad28b 1457static inline __be32
4a6e43e6
N
1458nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1459{
1460 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1461 return nfserr_openmode;
1462 else
1463 return nfs_ok;
1464}
1465
52f4fb43
N
1466static struct nfs4_delegation *
1467find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1468{
1469 struct nfs4_delegation *dp;
1470
ea1da636 1471 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N
1472 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1473 return dp;
1474 }
1475 return NULL;
1476}
1477
b37ad28b 1478static __be32
567d9829
N
1479nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1480 struct nfs4_delegation **dp)
1481{
1482 int flags;
b37ad28b 1483 __be32 status = nfserr_bad_stateid;
567d9829
N
1484
1485 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1486 if (*dp == NULL)
c44c5eeb 1487 goto out;
567d9829
N
1488 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1489 RD_STATE : WR_STATE;
1490 status = nfs4_check_delegmode(*dp, flags);
1491 if (status)
1492 *dp = NULL;
c44c5eeb
N
1493out:
1494 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1495 return nfs_ok;
1496 if (status)
1497 return status;
1498 open->op_stateowner->so_confirmed = 1;
1499 return nfs_ok;
567d9829
N
1500}
1501
b37ad28b 1502static __be32
1da177e4
LT
1503nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1504{
1505 struct nfs4_stateid *local;
b37ad28b 1506 __be32 status = nfserr_share_denied;
1da177e4
LT
1507 struct nfs4_stateowner *sop = open->op_stateowner;
1508
8beefa24 1509 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT
1510 /* ignore lock owners */
1511 if (local->st_stateowner->so_is_open_owner == 0)
1512 continue;
1513 /* remember if we have seen this open owner */
1514 if (local->st_stateowner == sop)
1515 *stpp = local;
1516 /* check for conflicting share reservations */
1517 if (!test_share(local, open))
1518 goto out;
1519 }
1520 status = 0;
1521out:
1522 return status;
1523}
1524
5ac049ac
N
1525static inline struct nfs4_stateid *
1526nfs4_alloc_stateid(void)
1527{
1528 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1529}
1530
b37ad28b 1531static __be32
1da177e4 1532nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 1533 struct nfs4_delegation *dp,
1da177e4
LT
1534 struct svc_fh *cur_fh, int flags)
1535{
1536 struct nfs4_stateid *stp;
1da177e4 1537
5ac049ac 1538 stp = nfs4_alloc_stateid();
1da177e4
LT
1539 if (stp == NULL)
1540 return nfserr_resource;
1541
567d9829
N
1542 if (dp) {
1543 get_file(dp->dl_vfs_file);
1544 stp->st_vfs_file = dp->dl_vfs_file;
1545 } else {
b37ad28b 1546 __be32 status;
567d9829
N
1547 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1548 &stp->st_vfs_file);
1549 if (status) {
1550 if (status == nfserr_dropit)
1551 status = nfserr_jukebox;
5ac049ac 1552 kmem_cache_free(stateid_slab, stp);
567d9829
N
1553 return status;
1554 }
1da177e4 1555 }
1da177e4
LT
1556 *stpp = stp;
1557 return 0;
1558}
1559
b37ad28b 1560static inline __be32
1da177e4
LT
1561nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1562 struct nfsd4_open *open)
1563{
1564 struct iattr iattr = {
1565 .ia_valid = ATTR_SIZE,
1566 .ia_size = 0,
1567 };
1568 if (!open->op_truncate)
1569 return 0;
1570 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 1571 return nfserr_inval;
1da177e4
LT
1572 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1573}
1574
b37ad28b 1575static __be32
1da177e4
LT
1576nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1577{
1578 struct file *filp = stp->st_vfs_file;
7eaa36e2 1579 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 1580 unsigned int share_access, new_writer;
b37ad28b 1581 __be32 status;
1da177e4
LT
1582
1583 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF
1584 new_writer = (~share_access) & open->op_share_access
1585 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 1586
6c26d08f 1587 if (new_writer) {
b37ad28b
AV
1588 int err = get_write_access(inode);
1589 if (err)
1590 return nfserrno(err);
6c26d08f 1591 }
1da177e4
LT
1592 status = nfsd4_truncate(rqstp, cur_fh, open);
1593 if (status) {
6c26d08f
BF
1594 if (new_writer)
1595 put_write_access(inode);
1da177e4
LT
1596 return status;
1597 }
1598 /* remember the open */
6c26d08f 1599 filp->f_mode |= open->op_share_access;
b55e0ba1
BF
1600 __set_bit(open->op_share_access, &stp->st_access_bmap);
1601 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT
1602
1603 return nfs_ok;
1604}
1605
1606
1da177e4 1607static void
37515177 1608nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 1609{
37515177
N
1610 open->op_stateowner->so_confirmed = 1;
1611 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT
1612}
1613
1614/*
1615 * Attempt to hand out a delegation.
1616 */
1617static void
1618nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1619{
1620 struct nfs4_delegation *dp;
1621 struct nfs4_stateowner *sop = stp->st_stateowner;
1622 struct nfs4_callback *cb = &sop->so_client->cl_callback;
1623 struct file_lock fl, *flp = &fl;
1624 int status, flag = 0;
1625
1626 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N
1627 open->op_recall = 0;
1628 switch (open->op_claim_type) {
1629 case NFS4_OPEN_CLAIM_PREVIOUS:
1630 if (!atomic_read(&cb->cb_set))
1631 open->op_recall = 1;
1632 flag = open->op_delegate_type;
1633 if (flag == NFS4_OPEN_DELEGATE_NONE)
1634 goto out;
1635 break;
1636 case NFS4_OPEN_CLAIM_NULL:
1637 /* Let's not give out any delegations till everyone's
1638 * had the chance to reclaim theirs.... */
1639 if (nfs4_in_grace())
1640 goto out;
1641 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
1642 goto out;
1643 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1644 flag = NFS4_OPEN_DELEGATE_WRITE;
1645 else
1646 flag = NFS4_OPEN_DELEGATE_READ;
1647 break;
1648 default:
1649 goto out;
1650 }
1da177e4
LT
1651
1652 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1653 if (dp == NULL) {
1654 flag = NFS4_OPEN_DELEGATE_NONE;
1655 goto out;
1656 }
1657 locks_init_lock(&fl);
1658 fl.fl_lmops = &nfsd_lease_mng_ops;
1659 fl.fl_flags = FL_LEASE;
9167f501 1660 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT
1661 fl.fl_end = OFFSET_MAX;
1662 fl.fl_owner = (fl_owner_t)dp;
1663 fl.fl_file = stp->st_vfs_file;
1664 fl.fl_pid = current->tgid;
1665
a9933cea 1666 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT
1667 * the lock_manager callbacks fl_mylease and fl_change are used
1668 */
9167f501 1669 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 1670 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 1671 unhash_delegation(dp);
1da177e4
LT
1672 flag = NFS4_OPEN_DELEGATE_NONE;
1673 goto out;
1674 }
1675
1676 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1677
1678 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1679 dp->dl_stateid.si_boot,
1680 dp->dl_stateid.si_stateownerid,
1681 dp->dl_stateid.si_fileid,
1682 dp->dl_stateid.si_generation);
1683out:
7b190fec
N
1684 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
1685 && flag == NFS4_OPEN_DELEGATE_NONE
1686 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 1687 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT
1688 open->op_delegate_type = flag;
1689}
1690
1691/*
1692 * called with nfs4_lock_state() held.
1693 */
b37ad28b 1694__be32
1da177e4
LT
1695nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1696{
1697 struct nfs4_file *fp = NULL;
1698 struct inode *ino = current_fh->fh_dentry->d_inode;
1699 struct nfs4_stateid *stp = NULL;
567d9829 1700 struct nfs4_delegation *dp = NULL;
b37ad28b 1701 __be32 status;
1da177e4
LT
1702
1703 status = nfserr_inval;
ba5a6a19
BF
1704 if (!access_valid(open->op_share_access)
1705 || !deny_valid(open->op_share_deny))
1da177e4
LT
1706 goto out;
1707 /*
1708 * Lookup file; if found, lookup stateid and check open request,
1709 * and check for delegations in the process of being recalled.
1710 * If not found, create the nfs4_file struct
1711 */
1712 fp = find_file(ino);
1713 if (fp) {
1714 if ((status = nfs4_check_open(fp, open, &stp)))
1715 goto out;
c44c5eeb
N
1716 status = nfs4_check_deleg(fp, open, &dp);
1717 if (status)
1718 goto out;
1da177e4 1719 } else {
c44c5eeb
N
1720 status = nfserr_bad_stateid;
1721 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1722 goto out;
1da177e4
LT
1723 status = nfserr_resource;
1724 fp = alloc_init_file(ino);
1725 if (fp == NULL)
1726 goto out;
1727 }
1728
1729 /*
1730 * OPEN the file, or upgrade an existing OPEN.
1731 * If truncate fails, the OPEN fails.
1732 */
1733 if (stp) {
1734 /* Stateid was found, this is an OPEN upgrade */
1735 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1736 if (status)
1737 goto out;
444c2c07 1738 update_stateid(&stp->st_stateid);
1da177e4
LT
1739 } else {
1740 /* Stateid was not found, this is a new OPEN */
1741 int flags = 0;
9ecb6a08 1742 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 1743 flags |= NFSD_MAY_READ;
1da177e4 1744 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 1745 flags |= NFSD_MAY_WRITE;
567d9829
N
1746 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1747 if (status)
1da177e4
LT
1748 goto out;
1749 init_stateid(stp, fp, open);
1750 status = nfsd4_truncate(rqstp, current_fh, open);
1751 if (status) {
1752 release_stateid(stp, OPEN_STATE);
1753 goto out;
1754 }
1755 }
1756 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1757
1758 /*
1759 * Attempt to hand out a delegation. No error return, because the
1760 * OPEN succeeds even if we fail.
1761 */
1762 nfs4_open_delegation(current_fh, open, stp);
1763
1764 status = nfs_ok;
1765
1766 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1767 stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1768 stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1769out:
13cd2184
N
1770 if (fp)
1771 put_nfs4_file(fp);
37515177
N
1772 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1773 nfs4_set_claim_prev(open);
1da177e4
LT
1774 /*
1775 * To finish the open response, we just need to set the rflags.
1776 */
1777 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1778 if (!open->op_stateowner->so_confirmed)
1779 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1780
1781 return status;
1782}
1783
b37ad28b 1784__be32
b591480b
BF
1785nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1786 clientid_t *clid)
1da177e4
LT
1787{
1788 struct nfs4_client *clp;
b37ad28b 1789 __be32 status;
1da177e4
LT
1790
1791 nfs4_lock_state();
1792 dprintk("process_renew(%08x/%08x): starting\n",
1793 clid->cl_boot, clid->cl_id);
1794 status = nfserr_stale_clientid;
1795 if (STALE_CLIENTID(clid))
1796 goto out;
1797 clp = find_confirmed_client(clid);
1798 status = nfserr_expired;
1799 if (clp == NULL) {
1800 /* We assume the client took too long to RENEW. */
1801 dprintk("nfsd4_renew: clientid not found!\n");
1802 goto out;
1803 }
1804 renew_client(clp);
1805 status = nfserr_cb_path_down;
ea1da636 1806 if (!list_empty(&clp->cl_delegations)
1da177e4
LT
1807 && !atomic_read(&clp->cl_callback.cb_set))
1808 goto out;
1809 status = nfs_ok;
1810out:
1811 nfs4_unlock_state();
1812 return status;
1813}
1814
a76b4319
N
1815static void
1816end_grace(void)
1817{
1818 dprintk("NFSD: end of grace period\n");
c7b9a459 1819 nfsd4_recdir_purge_old();
a76b4319
N
1820 in_grace = 0;
1821}
1822
fd39ca9a 1823static time_t
1da177e4
LT
1824nfs4_laundromat(void)
1825{
1826 struct nfs4_client *clp;
1827 struct nfs4_stateowner *sop;
1828 struct nfs4_delegation *dp;
1829 struct list_head *pos, *next, reaplist;
1830 time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1831 time_t t, clientid_val = NFSD_LEASE_TIME;
1832 time_t u, test_val = NFSD_LEASE_TIME;
1833
1834 nfs4_lock_state();
1835
1836 dprintk("NFSD: laundromat service - starting\n");
a76b4319
N
1837 if (in_grace)
1838 end_grace();
1da177e4
LT
1839 list_for_each_safe(pos, next, &client_lru) {
1840 clp = list_entry(pos, struct nfs4_client, cl_lru);
1841 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1842 t = clp->cl_time - cutoff;
1843 if (clientid_val > t)
1844 clientid_val = t;
1845 break;
1846 }
1847 dprintk("NFSD: purging unused client (clientid %08x)\n",
1848 clp->cl_clientid.cl_id);
c7b9a459 1849 nfsd4_remove_clid_dir(clp);
1da177e4
LT
1850 expire_client(clp);
1851 }
1852 INIT_LIST_HEAD(&reaplist);
1853 spin_lock(&recall_lock);
1854 list_for_each_safe(pos, next, &del_recall_lru) {
1855 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1856 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1857 u = dp->dl_time - cutoff;
1858 if (test_val > u)
1859 test_val = u;
1860 break;
1861 }
1862 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1863 dp, dp->dl_flock);
1864 list_move(&dp->dl_recall_lru, &reaplist);
1865 }
1866 spin_unlock(&recall_lock);
1867 list_for_each_safe(pos, next, &reaplist) {
1868 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1869 list_del_init(&dp->dl_recall_lru);
1870 unhash_delegation(dp);
1871 }
1872 test_val = NFSD_LEASE_TIME;
1873 list_for_each_safe(pos, next, &close_lru) {
1874 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1875 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1876 u = sop->so_time - cutoff;
1877 if (test_val > u)
1878 test_val = u;
1879 break;
1880 }
1881 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1882 sop->so_id);
358dd55a 1883 release_stateowner(sop);
1da177e4
LT
1884 }
1885 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1886 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1887 nfs4_unlock_state();
1888 return clientid_val;
1889}
1890
a254b246
HH
1891static struct workqueue_struct *laundry_wq;
1892static void laundromat_main(struct work_struct *);
1893static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
1894
1895static void
c4028958 1896laundromat_main(struct work_struct *not_used)
1da177e4
LT
1897{
1898 time_t t;
1899
1900 t = nfs4_laundromat();
1901 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 1902 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT
1903}
1904
fd39ca9a 1905static struct nfs4_stateowner *
f8816512
N
1906search_close_lru(u32 st_id, int flags)
1907{
1da177e4
LT
1908 struct nfs4_stateowner *local = NULL;
1909
1da177e4
LT
1910 if (flags & CLOSE_STATE) {
1911 list_for_each_entry(local, &close_lru, so_close_lru) {
1912 if (local->so_id == st_id)
1913 return local;
1914 }
1915 }
1916 return NULL;
1917}
1918
1919static inline int
1920nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1921{
7eaa36e2 1922 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT
1923}
1924
1925static int
1926STALE_STATEID(stateid_t *stateid)
1927{
1928 if (stateid->si_boot == boot_time)
1929 return 0;
849823c5 1930 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1da177e4
LT
1931 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
1932 stateid->si_generation);
1933 return 1;
1934}
1935
1936static inline int
1937access_permit_read(unsigned long access_bmap)
1938{
1939 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
1940 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
1941 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
1942}
1943
1944static inline int
1945access_permit_write(unsigned long access_bmap)
1946{
1947 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
1948 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
1949}
1950
1951static
b37ad28b 1952__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 1953{
b37ad28b 1954 __be32 status = nfserr_openmode;
1da177e4
LT
1955
1956 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
1957 goto out;
1958 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
1959 goto out;
1960 status = nfs_ok;
1961out:
1962 return status;
1963}
1964
b37ad28b 1965static inline __be32
1da177e4
LT
1966check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
1967{
1968 /* Trying to call delegreturn with a special stateid? Yuch: */
1969 if (!(flags & (RD_STATE | WR_STATE)))
1970 return nfserr_bad_stateid;
1971 else if (ONE_STATEID(stateid) && (flags & RD_STATE))
1972 return nfs_ok;
1973 else if (nfs4_in_grace()) {
1974 /* Answer in remaining cases depends on existance of
1975 * conflicting state; so we must wait out the grace period. */
1976 return nfserr_grace;
1977 } else if (flags & WR_STATE)
1978 return nfs4_share_conflict(current_fh,
1979 NFS4_SHARE_DENY_WRITE);
1980 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
1981 return nfs4_share_conflict(current_fh,
1982 NFS4_SHARE_DENY_READ);
1983}
1984
1985/*
1986 * Allow READ/WRITE during grace period on recovered state only for files
1987 * that are not able to provide mandatory locking.
1988 */
1989static inline int
1990io_during_grace_disallowed(struct inode *inode, int flags)
1991{
1992 return nfs4_in_grace() && (flags & (RD_STATE | WR_STATE))
a16877ca 1993 && mandatory_lock(inode);
1da177e4
LT
1994}
1995
0836f587
BF
1996static int check_stateid_generation(stateid_t *in, stateid_t *ref)
1997{
1998 /* If the client sends us a stateid from the future, it's buggy: */
1999 if (in->si_generation > ref->si_generation)
2000 return nfserr_bad_stateid;
2001 /*
2002 * The following, however, can happen. For example, if the
2003 * client sends an open and some IO at the same time, the open
2004 * may bump si_generation while the IO is still in flight.
2005 * Thanks to hard links and renames, the client never knows what
2006 * file an open will affect. So it could avoid that situation
2007 * only by serializing all opens and IO from the same open
2008 * owner. To recover from the old_stateid error, the client
2009 * will just have to retry the IO:
2010 */
2011 if (in->si_generation < ref->si_generation)
2012 return nfserr_old_stateid;
2013 return nfs_ok;
2014}
2015
1da177e4
LT
2016/*
2017* Checks for stateid operations
2018*/
b37ad28b 2019__be32
1da177e4
LT
2020nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2021{
2022 struct nfs4_stateid *stp = NULL;
2023 struct nfs4_delegation *dp = NULL;
2024 stateid_t *stidp;
2025 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2026 __be32 status;
1da177e4
LT
2027
2028 dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2029 stateid->si_boot, stateid->si_stateownerid,
2030 stateid->si_fileid, stateid->si_generation);
2031 if (filpp)
2032 *filpp = NULL;
2033
2034 if (io_during_grace_disallowed(ino, flags))
2035 return nfserr_grace;
2036
2037 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2038 return check_special_stateids(current_fh, stateid, flags);
2039
2040 /* STALE STATEID */
2041 status = nfserr_stale_stateid;
2042 if (STALE_STATEID(stateid))
2043 goto out;
2044
2045 /* BAD STATEID */
2046 status = nfserr_bad_stateid;
2047 if (!stateid->si_fileid) { /* delegation stateid */
2048 if(!(dp = find_delegation_stateid(ino, stateid))) {
2049 dprintk("NFSD: delegation stateid not found\n");
1da177e4
LT
2050 goto out;
2051 }
2052 stidp = &dp->dl_stateid;
2053 } else { /* open or lock stateid */
2054 if (!(stp = find_stateid(stateid, flags))) {
2055 dprintk("NFSD: open or lock stateid not found\n");
1da177e4
LT
2056 goto out;
2057 }
2058 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2059 goto out;
2060 if (!stp->st_stateowner->so_confirmed)
2061 goto out;
2062 stidp = &stp->st_stateid;
2063 }
0836f587
BF
2064 status = check_stateid_generation(stateid, stidp);
2065 if (status)
1da177e4
LT
2066 goto out;
2067 if (stp) {
2068 if ((status = nfs4_check_openmode(stp,flags)))
2069 goto out;
2070 renew_client(stp->st_stateowner->so_client);
2071 if (filpp)
2072 *filpp = stp->st_vfs_file;
6a85fa3a 2073 } else {
1da177e4
LT
2074 if ((status = nfs4_check_delegmode(dp, flags)))
2075 goto out;
2076 renew_client(dp->dl_client);
2077 if (flags & DELEG_RET)
2078 unhash_delegation(dp);
2079 if (filpp)
2080 *filpp = dp->dl_vfs_file;
2081 }
2082 status = nfs_ok;
2083out:
2084 return status;
2085}
2086
4c4cd222
N
2087static inline int
2088setlkflg (int type)
2089{
2090 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2091 RD_STATE : WR_STATE;
2092}
1da177e4
LT
2093
2094/*
2095 * Checks for sequence id mutating operations.
2096 */
b37ad28b 2097static __be32
4c4cd222 2098nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *stateid, int flags, struct nfs4_stateowner **sopp, struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2099{
1da177e4
LT
2100 struct nfs4_stateid *stp;
2101 struct nfs4_stateowner *sop;
0836f587 2102 __be32 status;
1da177e4
LT
2103
2104 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2105 "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2106 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2107 stateid->si_generation);
3a4f98bb 2108
1da177e4
LT
2109 *stpp = NULL;
2110 *sopp = NULL;
2111
1da177e4 2112 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2113 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2114 return nfserr_bad_stateid;
1da177e4
LT
2115 }
2116
1da177e4 2117 if (STALE_STATEID(stateid))
3a4f98bb 2118 return nfserr_stale_stateid;
1da177e4
LT
2119 /*
2120 * We return BAD_STATEID if filehandle doesn't match stateid,
2121 * the confirmed flag is incorrecly set, or the generation
2122 * number is incorrect.
1da177e4 2123 */
f8816512
N
2124 stp = find_stateid(stateid, flags);
2125 if (stp == NULL) {
2126 /*
2127 * Also, we should make sure this isn't just the result of
2128 * a replayed close:
2129 */
2130 sop = search_close_lru(stateid->si_stateownerid, flags);
2131 if (sop == NULL)
2132 return nfserr_bad_stateid;
2133 *sopp = sop;
2134 goto check_replay;
2135 }
1da177e4 2136
39325bd0
BF
2137 *stpp = stp;
2138 *sopp = sop = stp->st_stateowner;
2139
4c4cd222 2140 if (lock) {
4c4cd222 2141 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2142 struct nfs4_client *clp = sop->so_client;
4c4cd222 2143 int lkflg = 0;
b37ad28b 2144 __be32 status;
4c4cd222
N
2145
2146 lkflg = setlkflg(lock->lk_type);
2147
2148 if (lock->lk_is_new) {
599e0a22
BF
2149 if (!sop->so_is_open_owner)
2150 return nfserr_bad_stateid;
2151 if (!same_clid(&clp->cl_clientid, lockclid))
4c4cd222 2152 return nfserr_bad_stateid;
599e0a22
BF
2153 /* stp is the open stateid */
2154 status = nfs4_check_openmode(stp, lkflg);
2155 if (status)
2156 return status;
2157 } else {
2158 /* stp is the lock stateid */
2159 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2160 if (status)
2161 return status;
4c4cd222 2162 }
1da177e4
LT
2163 }
2164
f3362737 2165 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2166 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2167 return nfserr_bad_stateid;
1da177e4
LT
2168 }
2169
1da177e4
LT
2170 /*
2171 * We now validate the seqid and stateid generation numbers.
2172 * For the moment, we ignore the possibility of
2173 * generation number wraparound.
2174 */
bd9aac52 2175 if (seqid != sop->so_seqid)
1da177e4
LT
2176 goto check_replay;
2177
3a4f98bb 2178 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2179 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N
2180 " unconfirmed stateowner!\n");
2181 return nfserr_bad_stateid;
1da177e4 2182 }
3a4f98bb 2183 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2184 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N
2185 " confirmed yet!\n");
2186 return nfserr_bad_stateid;
1da177e4 2187 }
0836f587
BF
2188 status = check_stateid_generation(stateid, &stp->st_stateid);
2189 if (status)
2190 return status;
52fd004e 2191 renew_client(sop->so_client);
3a4f98bb 2192 return nfs_ok;
1da177e4 2193
1da177e4 2194check_replay:
bd9aac52 2195 if (seqid == sop->so_seqid - 1) {
849823c5 2196 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2197 /* indicate replay to calling function */
a90b061c 2198 return nfserr_replay_me;
1da177e4 2199 }
2fdada03 2200 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N
2201 sop->so_seqid, seqid);
2202 *sopp = NULL;
2203 return nfserr_bad_seqid;
1da177e4
LT
2204}
2205
b37ad28b 2206__be32
ca364317 2207nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2208 struct nfsd4_open_confirm *oc)
1da177e4 2209{
b37ad28b 2210 __be32 status;
1da177e4
LT
2211 struct nfs4_stateowner *sop;
2212 struct nfs4_stateid *stp;
2213
2214 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF
2215 (int)cstate->current_fh.fh_dentry->d_name.len,
2216 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2217
ca364317 2218 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF
2219 if (status)
2220 return status;
1da177e4
LT
2221
2222 nfs4_lock_state();
2223
ca364317
BF
2224 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2225 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 2226 CONFIRM | OPEN_STATE,
1da177e4
LT
2227 &oc->oc_stateowner, &stp, NULL)))
2228 goto out;
2229
2230 sop = oc->oc_stateowner;
2231 sop->so_confirmed = 1;
2232 update_stateid(&stp->st_stateid);
2233 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2234 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2235 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2236 stp->st_stateid.si_boot,
2237 stp->st_stateid.si_stateownerid,
2238 stp->st_stateid.si_fileid,
2239 stp->st_stateid.si_generation);
c7b9a459
N
2240
2241 nfsd4_create_clid_dir(sop->so_client);
1da177e4 2242out:
f2327d9a 2243 if (oc->oc_stateowner) {
1da177e4 2244 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 2245 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 2246 }
1da177e4
LT
2247 nfs4_unlock_state();
2248 return status;
2249}
2250
2251
2252/*
2253 * unset all bits in union bitmap (bmap) that
2254 * do not exist in share (from successful OPEN_DOWNGRADE)
2255 */
2256static void
2257reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2258{
2259 int i;
2260 for (i = 1; i < 4; i++) {
2261 if ((i & access) != i)
2262 __clear_bit(i, bmap);
2263 }
2264}
2265
2266static void
2267reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2268{
2269 int i;
2270 for (i = 0; i < 4; i++) {
2271 if ((i & deny) != i)
2272 __clear_bit(i, bmap);
2273 }
2274}
2275
b37ad28b 2276__be32
ca364317
BF
2277nfsd4_open_downgrade(struct svc_rqst *rqstp,
2278 struct nfsd4_compound_state *cstate,
a4f1706a 2279 struct nfsd4_open_downgrade *od)
1da177e4 2280{
b37ad28b 2281 __be32 status;
1da177e4
LT
2282 struct nfs4_stateid *stp;
2283 unsigned int share_access;
2284
2285 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF
2286 (int)cstate->current_fh.fh_dentry->d_name.len,
2287 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2288
ba5a6a19
BF
2289 if (!access_valid(od->od_share_access)
2290 || !deny_valid(od->od_share_deny))
1da177e4
LT
2291 return nfserr_inval;
2292
2293 nfs4_lock_state();
ca364317
BF
2294 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2295 od->od_seqid,
1da177e4 2296 &od->od_stateid,
f3362737 2297 OPEN_STATE,
1da177e4
LT
2298 &od->od_stateowner, &stp, NULL)))
2299 goto out;
2300
2301 status = nfserr_inval;
2302 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2303 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2304 stp->st_access_bmap, od->od_share_access);
2305 goto out;
2306 }
2307 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2308 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2309 stp->st_deny_bmap, od->od_share_deny);
2310 goto out;
2311 }
2312 set_access(&share_access, stp->st_access_bmap);
2313 nfs4_file_downgrade(stp->st_vfs_file,
2314 share_access & ~od->od_share_access);
2315
2316 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2317 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2318
2319 update_stateid(&stp->st_stateid);
2320 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2321 status = nfs_ok;
2322out:
f2327d9a 2323 if (od->od_stateowner) {
1da177e4 2324 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 2325 cstate->replay_owner = od->od_stateowner;
f2327d9a 2326 }
1da177e4
LT
2327 nfs4_unlock_state();
2328 return status;
2329}
2330
2331/*
2332 * nfs4_unlock_state() called after encode
2333 */
b37ad28b 2334__be32
ca364317 2335nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2336 struct nfsd4_close *close)
1da177e4 2337{
b37ad28b 2338 __be32 status;
1da177e4
LT
2339 struct nfs4_stateid *stp;
2340
2341 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF
2342 (int)cstate->current_fh.fh_dentry->d_name.len,
2343 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT
2344
2345 nfs4_lock_state();
2346 /* check close_lru for replay */
ca364317
BF
2347 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
2348 close->cl_seqid,
1da177e4 2349 &close->cl_stateid,
f3362737 2350 OPEN_STATE | CLOSE_STATE,
1da177e4
LT
2351 &close->cl_stateowner, &stp, NULL)))
2352 goto out;
1da177e4
LT
2353 status = nfs_ok;
2354 update_stateid(&stp->st_stateid);
2355 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2356
04ef5954
BF
2357 /* release_stateid() calls nfsd_close() if needed */
2358 release_stateid(stp, OPEN_STATE);
2359
2360 /* place unused nfs4_stateowners on so_close_lru list to be
2361 * released by the laundromat service after the lease period
2362 * to enable us to handle CLOSE replay
2363 */
2364 if (list_empty(&close->cl_stateowner->so_stateids))
2365 move_to_close_lru(close->cl_stateowner);
1da177e4 2366out:
f2327d9a 2367 if (close->cl_stateowner) {
1da177e4 2368 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 2369 cstate->replay_owner = close->cl_stateowner;
f2327d9a 2370 }
1da177e4
LT
2371 nfs4_unlock_state();
2372 return status;
2373}
2374
b37ad28b 2375__be32
ca364317
BF
2376nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2377 struct nfsd4_delegreturn *dr)
1da177e4 2378{
b37ad28b 2379 __be32 status;
1da177e4 2380
ca364317 2381 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
1da177e4
LT
2382 goto out;
2383
2384 nfs4_lock_state();
ca364317
BF
2385 status = nfs4_preprocess_stateid_op(&cstate->current_fh,
2386 &dr->dr_stateid, DELEG_RET, NULL);
1da177e4
LT
2387 nfs4_unlock_state();
2388out:
2389 return status;
2390}
2391
2392
2393/*
2394 * Lock owner state (byte-range locks)
2395 */
2396#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2397#define LOCK_HASH_BITS 8
2398#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2399#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2400
2401#define lockownerid_hashval(id) \
2402 ((id) & LOCK_HASH_MASK)
2403
2404static inline unsigned int
2405lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2406 struct xdr_netobj *ownername)
2407{
2408 return (file_hashval(inode) + cl_id
2409 + opaque_hashval(ownername->data, ownername->len))
2410 & LOCK_HASH_MASK;
2411}
2412
2413static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2414static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2415static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2416
fd39ca9a 2417static struct nfs4_stateid *
1da177e4
LT
2418find_stateid(stateid_t *stid, int flags)
2419{
2420 struct nfs4_stateid *local = NULL;
2421 u32 st_id = stid->si_stateownerid;
2422 u32 f_id = stid->si_fileid;
2423 unsigned int hashval;
2424
2425 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2426 if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2427 hashval = stateid_hashval(st_id, f_id);
2428 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2429 if ((local->st_stateid.si_stateownerid == st_id) &&
2430 (local->st_stateid.si_fileid == f_id))
2431 return local;
2432 }
2433 }
2434 if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2435 hashval = stateid_hashval(st_id, f_id);
2436 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2437 if ((local->st_stateid.si_stateownerid == st_id) &&
2438 (local->st_stateid.si_fileid == f_id))
2439 return local;
2440 }
849823c5 2441 }
1da177e4
LT
2442 return NULL;
2443}
2444
2445static struct nfs4_delegation *
2446find_delegation_stateid(struct inode *ino, stateid_t *stid)
2447{
13cd2184
N
2448 struct nfs4_file *fp;
2449 struct nfs4_delegation *dl;
1da177e4
LT
2450
2451 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2452 stid->si_boot, stid->si_stateownerid,
2453 stid->si_fileid, stid->si_generation);
2454
1da177e4 2455 fp = find_file(ino);
13cd2184
N
2456 if (!fp)
2457 return NULL;
2458 dl = find_delegation_file(fp, stid);
2459 put_nfs4_file(fp);
2460 return dl;
1da177e4
LT
2461}
2462
2463/*
2464 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2465 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2466 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2467 * locking, this prevents us from being completely protocol-compliant. The
2468 * real solution to this problem is to start using unsigned file offsets in
2469 * the VFS, but this is a very deep change!
2470 */
2471static inline void
2472nfs4_transform_lock_offset(struct file_lock *lock)
2473{
2474 if (lock->fl_start < 0)
2475 lock->fl_start = OFFSET_MAX;
2476 if (lock->fl_end < 0)
2477 lock->fl_end = OFFSET_MAX;
2478}
2479
d5b9026a
N
2480/* Hack!: For now, we're defining this just so we can use a pointer to it
2481 * as a unique cookie to identify our (NFSv4's) posix locks. */
e465a77f 2482static struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 2483};
1da177e4
LT
2484
2485static inline void
2486nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2487{
d5b9026a
N
2488 struct nfs4_stateowner *sop;
2489 unsigned int hval;
1da177e4 2490
d5b9026a
N
2491 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
2492 sop = (struct nfs4_stateowner *) fl->fl_owner;
2493 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT
2494 kref_get(&sop->so_ref);
2495 deny->ld_sop = sop;
2496 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N
2497 } else {
2498 deny->ld_sop = NULL;
2499 deny->ld_clientid.cl_boot = 0;
2500 deny->ld_clientid.cl_id = 0;
1da177e4
LT
2501 }
2502 deny->ld_start = fl->fl_start;
2503 deny->ld_length = ~(u64)0;
2504 if (fl->fl_end != ~(u64)0)
2505 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2506 deny->ld_type = NFS4_READ_LT;
2507 if (fl->fl_type != F_RDLCK)
2508 deny->ld_type = NFS4_WRITE_LT;
2509}
2510
1da177e4
LT
2511static struct nfs4_stateowner *
2512find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2513 struct xdr_netobj *owner)
2514{
2515 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2516 struct nfs4_stateowner *op;
2517
2518 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 2519 if (same_owner_str(op, owner, clid))
1da177e4
LT
2520 return op;
2521 }
2522 return NULL;
2523}
2524
2525/*
2526 * Alloc a lock owner structure.
2527 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2528 * occured.
2529 *
2530 * strhashval = lock_ownerstr_hashval
1da177e4
LT
2531 */
2532
2533static struct nfs4_stateowner *
2534alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2535 struct nfs4_stateowner *sop;
2536 struct nfs4_replay *rp;
2537 unsigned int idhashval;
2538
2539 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2540 return NULL;
2541 idhashval = lockownerid_hashval(current_ownerid);
2542 INIT_LIST_HEAD(&sop->so_idhash);
2543 INIT_LIST_HEAD(&sop->so_strhash);
2544 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N
2545 INIT_LIST_HEAD(&sop->so_stateids);
2546 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT
2547 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2548 sop->so_time = 0;
2549 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2550 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 2551 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT
2552 sop->so_is_open_owner = 0;
2553 sop->so_id = current_ownerid++;
2554 sop->so_client = clp;
b59e3c0e
NB
2555 /* It is the openowner seqid that will be incremented in encode in the
2556 * case of new lockowners; so increment the lock seqid manually: */
2557 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT
2558 sop->so_confirmed = 1;
2559 rp = &sop->so_replay;
de1ae286 2560 rp->rp_status = nfserr_serverfault;
1da177e4
LT
2561 rp->rp_buflen = 0;
2562 rp->rp_buf = rp->rp_ibuf;
2563 return sop;
2564}
2565
fd39ca9a 2566static struct nfs4_stateid *
1da177e4
LT
2567alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2568{
2569 struct nfs4_stateid *stp;
2570 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2571
5ac049ac
N
2572 stp = nfs4_alloc_stateid();
2573 if (stp == NULL)
1da177e4
LT
2574 goto out;
2575 INIT_LIST_HEAD(&stp->st_hash);
2576 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N
2577 INIT_LIST_HEAD(&stp->st_perstateowner);
2578 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 2579 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 2580 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 2581 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 2582 stp->st_stateowner = sop;
13cd2184 2583 get_nfs4_file(fp);
1da177e4
LT
2584 stp->st_file = fp;
2585 stp->st_stateid.si_boot = boot_time;
2586 stp->st_stateid.si_stateownerid = sop->so_id;
2587 stp->st_stateid.si_fileid = fp->fi_id;
2588 stp->st_stateid.si_generation = 0;
2589 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2590 stp->st_access_bmap = open_stp->st_access_bmap;
2591 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 2592 stp->st_openstp = open_stp;
1da177e4
LT
2593
2594out:
2595 return stp;
2596}
2597
fd39ca9a 2598static int
1da177e4
LT
2599check_lock_length(u64 offset, u64 length)
2600{
2601 return ((length == 0) || ((length != ~(u64)0) &&
2602 LOFF_OVERFLOW(offset, length)));
2603}
2604
2605/*
2606 * LOCK operation
2607 */
b37ad28b 2608__be32
ca364317 2609nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2610 struct nfsd4_lock *lock)
1da177e4 2611{
3e9e3dbe 2612 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 2613 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT
2614 struct nfs4_stateid *lock_stp;
2615 struct file *filp;
2616 struct file_lock file_lock;
8dc7c311 2617 struct file_lock conflock;
b37ad28b 2618 __be32 status = 0;
1da177e4 2619 unsigned int strhashval;
150b3934 2620 unsigned int cmd;
b8dd7b9a 2621 int err;
1da177e4
LT
2622
2623 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2624 (long long) lock->lk_offset,
2625 (long long) lock->lk_length);
2626
1da177e4
LT
2627 if (check_lock_length(lock->lk_offset, lock->lk_length))
2628 return nfserr_inval;
2629
ca364317 2630 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 2631 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA
2632 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2633 return status;
2634 }
2635
1da177e4
LT
2636 nfs4_lock_state();
2637
2638 if (lock->lk_is_new) {
893f8770
N
2639 /*
2640 * Client indicates that this is a new lockowner.
2641 * Use open owner and open stateid to create lock owner and
2642 * lock stateid.
2643 */
1da177e4
LT
2644 struct nfs4_stateid *open_stp = NULL;
2645 struct nfs4_file *fp;
2646
2647 status = nfserr_stale_clientid;
849823c5 2648 if (STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 2649 goto out;
1da177e4 2650
1da177e4 2651 /* validate and update open stateid and open seqid */
ca364317 2652 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT
2653 lock->lk_new_open_seqid,
2654 &lock->lk_new_open_stateid,
f3362737 2655 OPEN_STATE,
3a65588a 2656 &lock->lk_replay_owner, &open_stp,
b59e3c0e 2657 lock);
37515177 2658 if (status)
1da177e4 2659 goto out;
3a65588a 2660 open_sop = lock->lk_replay_owner;
1da177e4
LT
2661 /* create lockowner and lock stateid */
2662 fp = open_stp->st_file;
2663 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2664 open_sop->so_client->cl_clientid.cl_id,
2665 &lock->v.new.owner);
3e9e3dbe
N
2666 /* XXX: Do we need to check for duplicate stateowners on
2667 * the same file, or should they just be allowed (and
2668 * create new stateids)? */
1da177e4 2669 status = nfserr_resource;
b59e3c0e
NB
2670 lock_sop = alloc_init_lock_stateowner(strhashval,
2671 open_sop->so_client, open_stp, lock);
2672 if (lock_sop == NULL)
1da177e4 2673 goto out;
b59e3c0e 2674 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 2675 if (lock_stp == NULL)
1da177e4 2676 goto out;
1da177e4
LT
2677 } else {
2678 /* lock (lock owner + lock stateid) already exists */
ca364317 2679 status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT
2680 lock->lk_old_lock_seqid,
2681 &lock->lk_old_lock_stateid,
f3362737 2682 LOCK_STATE,
3a65588a 2683 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT
2684 if (status)
2685 goto out;
3a65588a 2686 lock_sop = lock->lk_replay_owner;
1da177e4 2687 }
3a65588a 2688 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT
2689 filp = lock_stp->st_vfs_file;
2690
0dd395dc
N
2691 status = nfserr_grace;
2692 if (nfs4_in_grace() && !lock->lk_reclaim)
2693 goto out;
2694 status = nfserr_no_grace;
2695 if (!nfs4_in_grace() && lock->lk_reclaim)
2696 goto out;
2697
1da177e4
LT
2698 locks_init_lock(&file_lock);
2699 switch (lock->lk_type) {
2700 case NFS4_READ_LT:
2701 case NFS4_READW_LT:
2702 file_lock.fl_type = F_RDLCK;
150b3934 2703 cmd = F_SETLK;
1da177e4
LT
2704 break;
2705 case NFS4_WRITE_LT:
2706 case NFS4_WRITEW_LT:
2707 file_lock.fl_type = F_WRLCK;
150b3934 2708 cmd = F_SETLK;
1da177e4
LT
2709 break;
2710 default:
2711 status = nfserr_inval;
2712 goto out;
2713 }
b59e3c0e 2714 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT
2715 file_lock.fl_pid = current->tgid;
2716 file_lock.fl_file = filp;
2717 file_lock.fl_flags = FL_POSIX;
d5b9026a 2718 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
2719
2720 file_lock.fl_start = lock->lk_offset;
2721 if ((lock->lk_length == ~(u64)0) ||
2722 LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2723 file_lock.fl_end = ~(u64)0;
2724 else
2725 file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2726 nfs4_transform_lock_offset(&file_lock);
2727
2728 /*
2729 * Try to lock the file in the VFS.
2730 * Note: locks.c uses the BKL to protect the inode's lock list.
2731 */
2732
fd85b817 2733 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 2734 switch (-err) {
1da177e4
LT
2735 case 0: /* success! */
2736 update_stateid(&lock_stp->st_stateid);
2737 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2738 sizeof(stateid_t));
b8dd7b9a 2739 status = 0;
eb76b3fd
AA
2740 break;
2741 case (EAGAIN): /* conflock holds conflicting lock */
2742 status = nfserr_denied;
2743 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2744 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
2745 break;
1da177e4
LT
2746 case (EDEADLK):
2747 status = nfserr_deadlock;
eb76b3fd 2748 break;
1da177e4 2749 default:
fd85b817 2750 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA
2751 status = nfserr_resource;
2752 break;
1da177e4 2753 }
1da177e4 2754out:
8a280510
BF
2755 if (status && lock->lk_is_new && lock_sop)
2756 release_stateowner(lock_sop);
3a65588a
BF
2757 if (lock->lk_replay_owner) {
2758 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 2759 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 2760 }
1da177e4
LT
2761 nfs4_unlock_state();
2762 return status;
2763}
2764
2765/*
2766 * LOCKT operation
2767 */
b37ad28b 2768__be32
ca364317
BF
2769nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2770 struct nfsd4_lockt *lockt)
1da177e4
LT
2771{
2772 struct inode *inode;
2773 struct file file;
2774 struct file_lock file_lock;
fd85b817 2775 int error;
b37ad28b 2776 __be32 status;
1da177e4
LT
2777
2778 if (nfs4_in_grace())
2779 return nfserr_grace;
2780
2781 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2782 return nfserr_inval;
2783
2784 lockt->lt_stateowner = NULL;
2785 nfs4_lock_state();
2786
2787 status = nfserr_stale_clientid;
849823c5 2788 if (STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 2789 goto out;
1da177e4 2790
ca364317 2791 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 2792 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT
2793 if (status == nfserr_symlink)
2794 status = nfserr_inval;
2795 goto out;
2796 }
2797
ca364317 2798 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT
2799 locks_init_lock(&file_lock);
2800 switch (lockt->lt_type) {
2801 case NFS4_READ_LT:
2802 case NFS4_READW_LT:
2803 file_lock.fl_type = F_RDLCK;
2804 break;
2805 case NFS4_WRITE_LT:
2806 case NFS4_WRITEW_LT:
2807 file_lock.fl_type = F_WRLCK;
2808 break;
2809 default:
2fdada03 2810 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT
2811 status = nfserr_inval;
2812 goto out;
2813 }
2814
2815 lockt->lt_stateowner = find_lockstateowner_str(inode,
2816 &lockt->lt_clientid, &lockt->lt_owner);
2817 if (lockt->lt_stateowner)
2818 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2819 file_lock.fl_pid = current->tgid;
2820 file_lock.fl_flags = FL_POSIX;
d5b9026a 2821 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
2822
2823 file_lock.fl_start = lockt->lt_offset;
2824 if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2825 file_lock.fl_end = ~(u64)0;
2826 else
2827 file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2828
2829 nfs4_transform_lock_offset(&file_lock);
2830
fd85b817 2831 /* vfs_test_lock uses the struct file _only_ to resolve the inode.
1da177e4 2832 * since LOCKT doesn't require an OPEN, and therefore a struct
fd85b817 2833 * file may not exist, pass vfs_test_lock a struct file with
1da177e4
LT
2834 * only the dentry:inode set.
2835 */
2836 memset(&file, 0, sizeof (struct file));
ca364317 2837 file.f_path.dentry = cstate->current_fh.fh_dentry;
1da177e4
LT
2838
2839 status = nfs_ok;
fd85b817
ME
2840 error = vfs_test_lock(&file, &file_lock);
2841 if (error) {
2842 status = nfserrno(error);
2843 goto out;
2844 }
9d6a8c5c 2845 if (file_lock.fl_type != F_UNLCK) {
1da177e4 2846 status = nfserr_denied;
9d6a8c5c 2847 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT
2848 }
2849out:
2850 nfs4_unlock_state();
2851 return status;
2852}
2853
b37ad28b 2854__be32
ca364317 2855nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2856 struct nfsd4_locku *locku)
1da177e4
LT
2857{
2858 struct nfs4_stateid *stp;
2859 struct file *filp = NULL;
2860 struct file_lock file_lock;
b37ad28b 2861 __be32 status;
b8dd7b9a 2862 int err;
1da177e4
LT
2863
2864 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2865 (long long) locku->lu_offset,
2866 (long long) locku->lu_length);
2867
2868 if (check_lock_length(locku->lu_offset, locku->lu_length))
2869 return nfserr_inval;
2870
2871 nfs4_lock_state();
2872
ca364317 2873 if ((status = nfs4_preprocess_seqid_op(&cstate->current_fh,
1da177e4
LT
2874 locku->lu_seqid,
2875 &locku->lu_stateid,
f3362737 2876 LOCK_STATE,
1da177e4
LT
2877 &locku->lu_stateowner, &stp, NULL)))
2878 goto out;
2879
2880 filp = stp->st_vfs_file;
2881 BUG_ON(!filp);
2882 locks_init_lock(&file_lock);
2883 file_lock.fl_type = F_UNLCK;
2884 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2885 file_lock.fl_pid = current->tgid;
2886 file_lock.fl_file = filp;
2887 file_lock.fl_flags = FL_POSIX;
d5b9026a 2888 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT
2889 file_lock.fl_start = locku->lu_offset;
2890
2891 if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2892 file_lock.fl_end = ~(u64)0;
2893 else
2894 file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2895 nfs4_transform_lock_offset(&file_lock);
2896
2897 /*
2898 * Try to unlock the file in the VFS.
2899 */
fd85b817 2900 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 2901 if (err) {
fd85b817 2902 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT
2903 goto out_nfserr;
2904 }
2905 /*
2906 * OK, unlock succeeded; the only thing left to do is update the stateid.
2907 */
2908 update_stateid(&stp->st_stateid);
2909 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
2910
2911out:
f2327d9a 2912 if (locku->lu_stateowner) {
1da177e4 2913 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 2914 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 2915 }
1da177e4
LT
2916 nfs4_unlock_state();
2917 return status;
2918
2919out_nfserr:
b8dd7b9a 2920 status = nfserrno(err);
1da177e4
LT
2921 goto out;
2922}
2923
2924/*
2925 * returns
2926 * 1: locks held by lockowner
2927 * 0: no locks held by lockowner
2928 */
2929static int
2930check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
2931{
2932 struct file_lock **flpp;
7eaa36e2 2933 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT
2934 int status = 0;
2935
2936 lock_kernel();
2937 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 2938 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT
2939 status = 1;
2940 goto out;
796dadfd 2941 }
1da177e4
LT
2942 }
2943out:
2944 unlock_kernel();
2945 return status;
2946}
2947
b37ad28b 2948__be32
b591480b
BF
2949nfsd4_release_lockowner(struct svc_rqst *rqstp,
2950 struct nfsd4_compound_state *cstate,
2951 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT
2952{
2953 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N
2954 struct nfs4_stateowner *sop;
2955 struct nfs4_stateid *stp;
1da177e4 2956 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N
2957 struct list_head matches;
2958 int i;
b37ad28b 2959 __be32 status;
1da177e4
LT
2960
2961 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
2962 clid->cl_boot, clid->cl_id);
2963
2964 /* XXX check for lease expiration */
2965
2966 status = nfserr_stale_clientid;
849823c5 2967 if (STALE_CLIENTID(clid))
1da177e4 2968 return status;
1da177e4
LT
2969
2970 nfs4_lock_state();
2971
3e9e3dbe
N
2972 status = nfserr_locks_held;
2973 /* XXX: we're doing a linear search through all the lockowners.
2974 * Yipes! For now we'll just hope clients aren't really using
2975 * release_lockowner much, but eventually we have to fix these
2976 * data structures. */
2977 INIT_LIST_HEAD(&matches);
2978 for (i = 0; i < LOCK_HASH_SIZE; i++) {
2979 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 2980 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N
2981 continue;
2982 list_for_each_entry(stp, &sop->so_stateids,
2983 st_perstateowner) {
2984 if (check_for_locks(stp->st_vfs_file, sop))
2985 goto out;
2986 /* Note: so_perclient unused for lockowners,
2987 * so it's OK to fool with here. */
2988 list_add(&sop->so_perclient, &matches);
2989 }
1da177e4 2990 }
3e9e3dbe
N
2991 }
2992 /* Clients probably won't expect us to return with some (but not all)
2993 * of the lockowner state released; so don't release any until all
2994 * have been checked. */
2995 status = nfs_ok;
0fa822e4
N
2996 while (!list_empty(&matches)) {
2997 sop = list_entry(matches.next, struct nfs4_stateowner,
2998 so_perclient);
2999 /* unhash_stateowner deletes so_perclient only
3000 * for openowners. */
3001 list_del(&sop->so_perclient);
3e9e3dbe 3002 release_stateowner(sop);
1da177e4
LT
3003 }
3004out:
3005 nfs4_unlock_state();
3006 return status;
3007}
3008
3009static inline struct nfs4_client_reclaim *
a55370a3 3010alloc_reclaim(void)
1da177e4 3011{
a55370a3 3012 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT
3013}
3014
c7b9a459
N
3015int
3016nfs4_has_reclaimed_state(const char *name)
3017{
3018 unsigned int strhashval = clientstr_hashval(name);
3019 struct nfs4_client *clp;
3020
3021 clp = find_confirmed_client_by_str(name, strhashval);
3022 return clp ? 1 : 0;
3023}
3024
1da177e4
LT
3025/*
3026 * failure => all reset bets are off, nfserr_no_grace...
3027 */
190e4fbf
N
3028int
3029nfs4_client_to_reclaim(const char *name)
1da177e4
LT
3030{
3031 unsigned int strhashval;
3032 struct nfs4_client_reclaim *crp = NULL;
3033
a55370a3
N
3034 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3035 crp = alloc_reclaim();
1da177e4
LT
3036 if (!crp)
3037 return 0;
a55370a3 3038 strhashval = clientstr_hashval(name);
1da177e4
LT
3039 INIT_LIST_HEAD(&crp->cr_strhash);
3040 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3041 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT
3042 reclaim_str_hashtbl_size++;
3043 return 1;
3044}
3045
3046static void
3047nfs4_release_reclaim(void)
3048{
3049 struct nfs4_client_reclaim *crp = NULL;
3050 int i;
3051
1da177e4
LT
3052 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3053 while (!list_empty(&reclaim_str_hashtbl[i])) {
3054 crp = list_entry(reclaim_str_hashtbl[i].next,
3055 struct nfs4_client_reclaim, cr_strhash);
3056 list_del(&crp->cr_strhash);
1da177e4
LT
3057 kfree(crp);
3058 reclaim_str_hashtbl_size--;
3059 }
3060 }
3061 BUG_ON(reclaim_str_hashtbl_size);
3062}
3063
3064/*
3065 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3066static struct nfs4_client_reclaim *
1da177e4
LT
3067nfs4_find_reclaim_client(clientid_t *clid)
3068{
3069 unsigned int strhashval;
3070 struct nfs4_client *clp;
3071 struct nfs4_client_reclaim *crp = NULL;
3072
3073
3074 /* find clientid in conf_id_hashtbl */
3075 clp = find_confirmed_client(clid);
3076 if (clp == NULL)
3077 return NULL;
3078
a55370a3
N
3079 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3080 clp->cl_name.len, clp->cl_name.data,
3081 clp->cl_recdir);
1da177e4
LT
3082
3083 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3084 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3085 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3086 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT
3087 return crp;
3088 }
3089 }
3090 return NULL;
3091}
3092
3093/*
3094* Called from OPEN. Look for clientid in reclaim list.
3095*/
b37ad28b 3096__be32
1da177e4
LT
3097nfs4_check_open_reclaim(clientid_t *clid)
3098{
dfc83565 3099 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT
3100}
3101
ac4d8ff2 3102/* initialization to perform at module load time: */
1da177e4 3103
e8ff2a84 3104int
ac4d8ff2 3105nfs4_state_init(void)
1da177e4 3106{
e8ff2a84 3107 int i, status;
1da177e4 3108
e8ff2a84
BF
3109 status = nfsd4_init_slabs();
3110 if (status)
3111 return status;
1da177e4
LT
3112 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3113 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3114 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3115 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3116 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3117 }
3118 for (i = 0; i < FILE_HASH_SIZE; i++) {
3119 INIT_LIST_HEAD(&file_hashtbl[i]);
3120 }
3121 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3122 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3123 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3124 }
3125 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3126 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3127 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3128 }
3129 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3130 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3131 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3132 }
1da177e4 3133 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT
3134 INIT_LIST_HEAD(&close_lru);
3135 INIT_LIST_HEAD(&client_lru);
3136 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2
N
3137 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3138 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3139 reclaim_str_hashtbl_size = 0;
e8ff2a84 3140 return 0;
ac4d8ff2
N
3141}
3142
190e4fbf
N
3143static void
3144nfsd4_load_reboot_recovery_data(void)
3145{
3146 int status;
3147
0964a3d3
N
3148 nfs4_lock_state();
3149 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3150 status = nfsd4_recdir_load();
0964a3d3 3151 nfs4_unlock_state();
190e4fbf
N
3152 if (status)
3153 printk("NFSD: Failure reading reboot recovery data\n");
3154}
3155
9a8db97e
ME
3156unsigned long
3157get_nfs4_grace_period(void)
3158{
3159 return max(user_lease_time, lease_time) * HZ;
3160}
3161
c2f1a551
MS
3162/*
3163 * Since the lifetime of a delegation isn't limited to that of an open, a
3164 * client may quite reasonably hang on to a delegation as long as it has
3165 * the inode cached. This becomes an obvious problem the first time a
3166 * client's inode cache approaches the size of the server's total memory.
3167 *
3168 * For now we avoid this problem by imposing a hard limit on the number
3169 * of delegations, which varies according to the server's memory size.
3170 */
3171static void
3172set_max_delegations(void)
3173{
3174 /*
3175 * Allow at most 4 delegations per megabyte of RAM. Quick
3176 * estimates suggest that in the worst case (where every delegation
3177 * is for a different inode), a delegation could take about 1.5K,
3178 * giving a worst case usage of about 6% of memory.
3179 */
3180 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3181}
3182
ac4d8ff2
N
3183/* initialization to perform when the nfsd service is started: */
3184
3185static void
3186__nfs4_state_start(void)
3187{
9a8db97e 3188 unsigned long grace_time;
ac4d8ff2 3189
1da177e4 3190 boot_time = get_seconds();
9a8db97e 3191 grace_time = get_nfs_grace_period();
d99a05ad 3192 lease_time = user_lease_time;
a76b4319 3193 in_grace = 1;
9a8db97e
ME
3194 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
3195 grace_time/HZ);
58da282b 3196 laundry_wq = create_singlethread_workqueue("nfsd4");
9a8db97e 3197 queue_delayed_work(laundry_wq, &laundromat_work, grace_time);
c2f1a551 3198 set_max_delegations();
1da177e4
LT
3199}
3200
e8ff2a84 3201void
76a3550e 3202nfs4_state_start(void)
1da177e4 3203{
1da177e4 3204 if (nfs4_init)
e8ff2a84 3205 return;
190e4fbf 3206 nfsd4_load_reboot_recovery_data();
76a3550e 3207 __nfs4_state_start();
1da177e4 3208 nfs4_init = 1;
e8ff2a84 3209 return;
1da177e4
LT
3210}
3211
3212int
3213nfs4_in_grace(void)
3214{
a76b4319 3215 return in_grace;
1da177e4
LT
3216}
3217
3218time_t
3219nfs4_lease_time(void)
3220{
3221 return lease_time;
3222}
3223
3224static void
3225__nfs4_state_shutdown(void)
3226{
3227 int i;
3228 struct nfs4_client *clp = NULL;
3229 struct nfs4_delegation *dp = NULL;
1da177e4
LT
3230 struct list_head *pos, *next, reaplist;
3231
1da177e4
LT
3232 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3233 while (!list_empty(&conf_id_hashtbl[i])) {
3234 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3235 expire_client(clp);
3236 }
3237 while (!list_empty(&unconf_str_hashtbl[i])) {
3238 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3239 expire_client(clp);
3240 }
3241 }
3242 INIT_LIST_HEAD(&reaplist);
3243 spin_lock(&recall_lock);
3244 list_for_each_safe(pos, next, &del_recall_lru) {
3245 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3246 list_move(&dp->dl_recall_lru, &reaplist);
3247 }
3248 spin_unlock(&recall_lock);
3249 list_for_each_safe(pos, next, &reaplist) {
3250 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3251 list_del_init(&dp->dl_recall_lru);
3252 unhash_delegation(dp);
3253 }
3254
190e4fbf 3255 nfsd4_shutdown_recdir();
1da177e4 3256 nfs4_init = 0;
1da177e4
LT
3257}
3258
3259void
3260nfs4_state_shutdown(void)
3261{
5e8d5c29
N
3262 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
3263 destroy_workqueue(laundry_wq);
1da177e4
LT
3264 nfs4_lock_state();
3265 nfs4_release_reclaim();
3266 __nfs4_state_shutdown();
1da177e4
LT
3267 nfs4_unlock_state();
3268}
3269
3dd98a3b
JL
3270/*
3271 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3272 * accessed when nfsd is starting.
3273 */
0964a3d3
N
3274static void
3275nfs4_set_recdir(char *recdir)
3276{
0964a3d3 3277 strcpy(user_recovery_dirname, recdir);
0964a3d3
N
3278}
3279
3280/*
3281 * Change the NFSv4 recovery directory to recdir.
3282 */
3283int
3284nfs4_reset_recoverydir(char *recdir)
3285{
3286 int status;
3287 struct nameidata nd;
3288
3289 status = path_lookup(recdir, LOOKUP_FOLLOW, &nd);
3290 if (status)
3291 return status;
3292 status = -ENOTDIR;
4ac91378 3293 if (S_ISDIR(nd.path.dentry->d_inode->i_mode)) {
0964a3d3
N
3294 nfs4_set_recdir(recdir);
3295 status = 0;
3296 }
1d957f9b 3297 path_put(&nd.path);
0964a3d3
N
3298 return status;
3299}
3300
3dd98a3b
JL
3301char *
3302nfs4_recoverydir(void)
3303{
3304 return user_recovery_dirname;
3305}
3306
1da177e4
LT
3307/*
3308 * Called when leasetime is changed.
3309 *
d99a05ad
N
3310 * The only way the protocol gives us to handle on-the-fly lease changes is to
3311 * simulate a reboot. Instead of doing that, we just wait till the next time
3312 * we start to register any changes in lease time. If the administrator
3313 * really wants to change the lease time *now*, they can go ahead and bring
3314 * nfsd down and then back up again after changing the lease time.
3dd98a3b
JL
3315 *
3316 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3317 * when nfsd is starting
1da177e4
LT
3318 */
3319void
3320nfs4_reset_lease(time_t leasetime)
3321{
d99a05ad 3322 user_lease_time = leasetime;
1da177e4 3323}