projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4: cl_count is unused
[linux-2.6-block.git] / fs / nfsd / nfs4state.c
CommitLineData
1da177e4 1/*
1da177e4
LT		 2* Copyright (c) 2001 The Regents of the University of Michigan.
3* All rights reserved.
4*
5* Kendrick Smith <kmsmith@umich.edu>
6* Andy Adamson <kandros@umich.edu>
7*
8* Redistribution and use in source and binary forms, with or without
9* modification, are permitted provided that the following conditions
10* are met:
11*
12* 1. Redistributions of source code must retain the above copyright
13* notice, this list of conditions and the following disclaimer.
14* 2. Redistributions in binary form must reproduce the above copyright
15* notice, this list of conditions and the following disclaimer in the
16* documentation and/or other materials provided with the distribution.
17* 3. Neither the name of the University nor the names of its
18* contributors may be used to endorse or promote products derived
19* from this software without specific prior written permission.
20*
21* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24* DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32*
33*/
34
aceaf78d 35#include <linux/file.h>
1da177e4 36#include <linux/smp_lock.h>
0964a3d3 37#include <linux/namei.h>
c2f1a551 38#include <linux/swap.h>
68e76ad0 39#include <linux/sunrpc/svcauth_gss.h>
363168b4 40#include <linux/sunrpc/clnt.h>
9a74af21 41#include "xdr4.h"
0a3adade 42#include "vfs.h"
1da177e4
LT		 43
44#define NFSDDBG_FACILITY NFSDDBG_PROC
45
46/* Globals */
cf07d2ea 47time_t nfsd4_lease = 90; /* default lease time */
efc4bb4f 48time_t nfsd4_grace = 90;
fd39ca9a 49static time_t boot_time;
1da177e4
LT		 50static u32 current_ownerid = 1;
51static u32 current_fileid = 1;
52static u32 current_delegid = 1;
53static u32 nfs4_init;
fd39ca9a
N		 54static stateid_t zerostateid; /* bits all 0 */
55static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 56static u64 current_sessionid = 1;
fd39ca9a
N		 57
58#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
59#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 60
1da177e4 61/* forward declarations */
fd39ca9a 62static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 63static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 64static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
65static void nfs4_set_recdir(char *recdir);
1da177e4 66
8b671b80
BF		 67/* Locking: */
68
69/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 70static DEFINE_MUTEX(client_mutex);
1da177e4 71
8b671b80
BF		 72/*
73 * Currently used for the del_recall_lru and file hash table. In an
74 * effort to decrease the scope of the client_mutex, this spinlock may
75 * eventually cover more:
76 */
77static DEFINE_SPINLOCK(recall_lock);
78
e18b890b
CL		 79static struct kmem_cache *stateowner_slab = NULL;
80static struct kmem_cache *file_slab = NULL;
81static struct kmem_cache *stateid_slab = NULL;
82static struct kmem_cache *deleg_slab = NULL;
e60d4398 83
1da177e4
LT		 84void
85nfs4_lock_state(void)
86{
353ab6e9 87 mutex_lock(&client_mutex);
1da177e4
LT		 88}
89
90void
91nfs4_unlock_state(void)
92{
353ab6e9 93 mutex_unlock(&client_mutex);
1da177e4
LT		 94}
95
96static inline u32
97opaque_hashval(const void *ptr, int nbytes)
98{
99 unsigned char *cptr = (unsigned char *) ptr;
100
101 u32 x = 0;
102 while (nbytes--) {
103 x *= 37;
104 x += *cptr++;
105 }
106 return x;
107}
108
1da177e4
LT		 109static struct list_head del_recall_lru;
110
13cd2184
N		 111static inline void
112put_nfs4_file(struct nfs4_file *fi)
113{
8b671b80
BF		 114 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
115 list_del(&fi->fi_hash);
116 spin_unlock(&recall_lock);
117 iput(fi->fi_inode);
118 kmem_cache_free(file_slab, fi);
119 }
13cd2184
N		 120}
121
122static inline void
123get_nfs4_file(struct nfs4_file *fi)
124{
8b671b80 125 atomic_inc(&fi->fi_ref);
13cd2184
N		 126}
127
ef0f3390 128static int num_delegations;
c2f1a551 129unsigned int max_delegations;
ef0f3390
N		 130
131/*
132 * Open owner state (share locks)
133 */
134
135/* hash tables for nfs4_stateowner */
136#define OWNER_HASH_BITS 8
137#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
138#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
139
140#define ownerid_hashval(id) \
141 ((id) & OWNER_HASH_MASK)
142#define ownerstr_hashval(clientid, ownername) \
143 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
144
145static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
146static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
147
148/* hash table for nfs4_file */
149#define FILE_HASH_BITS 8
150#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
151#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
152/* hash table for (open)nfs4_stateid */
153#define STATEID_HASH_BITS 10
154#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
155#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
156
157#define file_hashval(x) \
158 hash_ptr(x, FILE_HASH_BITS)
159#define stateid_hashval(owner_id, file_id) \
160 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
161
162static struct list_head file_hashtbl[FILE_HASH_SIZE];
163static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
164
1da177e4
LT		 165static struct nfs4_delegation *
166alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
167{
168 struct nfs4_delegation *dp;
169 struct nfs4_file *fp = stp->st_file;
c237dc03 170 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 171
172 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 173 if (fp->fi_had_conflict)
174 return NULL;
c2f1a551 175 if (num_delegations > max_delegations)
ef0f3390 176 return NULL;
5b2d21c1
N		 177 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
178 if (dp == NULL)
1da177e4 179 return dp;
ef0f3390 180 num_delegations++;
ea1da636
N		 181 INIT_LIST_HEAD(&dp->dl_perfile);
182 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 183 INIT_LIST_HEAD(&dp->dl_recall_lru);
184 dp->dl_client = clp;
13cd2184 185 get_nfs4_file(fp);
1da177e4
LT		 186 dp->dl_file = fp;
187 dp->dl_flock = NULL;
188 get_file(stp->st_vfs_file);
189 dp->dl_vfs_file = stp->st_vfs_file;
190 dp->dl_type = type;
b53d40c5 191 dp->dl_ident = cb->cb_ident;
78155ed7 192 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 193 dp->dl_stateid.si_stateownerid = current_delegid++;
194 dp->dl_stateid.si_fileid = 0;
195 dp->dl_stateid.si_generation = 0;
6c02eaa1 196 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 197 dp->dl_time = 0;
198 atomic_set(&dp->dl_count, 1);
ea1da636
N		 199 list_add(&dp->dl_perfile, &fp->fi_delegations);
200 list_add(&dp->dl_perclnt, &clp->cl_delegations);
b5a1a81e 201 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
1da177e4
LT		 202 return dp;
203}
204
205void
206nfs4_put_delegation(struct nfs4_delegation *dp)
207{
208 if (atomic_dec_and_test(&dp->dl_count)) {
209 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 210 put_nfs4_file(dp->dl_file);
5b2d21c1 211 kmem_cache_free(deleg_slab, dp);
ef0f3390 212 num_delegations--;
1da177e4
LT		 213 }
214}
215
216/* Remove the associated file_lock first, then remove the delegation.
217 * lease_modify() is called to remove the FS_LEASE file_lock from
218 * the i_flock list, eventually calling nfsd's lock_manager
219 * fl_release_callback.
220 */
221static void
222nfs4_close_delegation(struct nfs4_delegation *dp)
223{
224 struct file *filp = dp->dl_vfs_file;
225
226 dprintk("NFSD: close_delegation dp %p\n",dp);
227 dp->dl_vfs_file = NULL;
228 /* The following nfsd_close may not actually close the file,
229 * but we want to remove the lease in any case. */
c907132d 230 if (dp->dl_flock)
a9933cea 231 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 232 nfsd_close(filp);
1da177e4
LT		 233}
234
235/* Called under the state lock. */
236static void
237unhash_delegation(struct nfs4_delegation *dp)
238{
ea1da636
N		 239 list_del_init(&dp->dl_perfile);
240 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 241 spin_lock(&recall_lock);
242 list_del_init(&dp->dl_recall_lru);
243 spin_unlock(&recall_lock);
244 nfs4_close_delegation(dp);
245 nfs4_put_delegation(dp);
246}
247
248/*
249 * SETCLIENTID state
250 */
251
252/* Hash tables for nfs4_clientid state */
253#define CLIENT_HASH_BITS 4
254#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
255#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
256
257#define clientid_hashval(id) \
258 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 259#define clientstr_hashval(name) \
260 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 261/*
262 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
263 * used in reboot/reset lease grace period processing
264 *
265 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
266 * setclientid_confirmed info.
267 *
268 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
269 * setclientid info.
270 *
271 * client_lru holds client queue ordered by nfs4_client.cl_time
272 * for lease renewal.
273 *
274 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
275 * for last close replay.
276 */
277static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
278static int reclaim_str_hashtbl_size = 0;
279static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
280static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
281static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
282static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
283static struct list_head client_lru;
284static struct list_head close_lru;
285
2283963f
BF		 286static void unhash_generic_stateid(struct nfs4_stateid *stp)
287{
288 list_del(&stp->st_hash);
289 list_del(&stp->st_perfile);
290 list_del(&stp->st_perstateowner);
291}
292
293static void free_generic_stateid(struct nfs4_stateid *stp)
294{
295 put_nfs4_file(stp->st_file);
296 kmem_cache_free(stateid_slab, stp);
297}
298
299static void release_lock_stateid(struct nfs4_stateid *stp)
300{
301 unhash_generic_stateid(stp);
302 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
303 free_generic_stateid(stp);
304}
305
f044ff83
BF		 306static void unhash_lockowner(struct nfs4_stateowner *sop)
307{
308 struct nfs4_stateid *stp;
309
310 list_del(&sop->so_idhash);
311 list_del(&sop->so_strhash);
312 list_del(&sop->so_perstateid);
313 while (!list_empty(&sop->so_stateids)) {
314 stp = list_first_entry(&sop->so_stateids,
315 struct nfs4_stateid, st_perstateowner);
316 release_lock_stateid(stp);
317 }
318}
319
320static void release_lockowner(struct nfs4_stateowner *sop)
321{
322 unhash_lockowner(sop);
323 nfs4_put_stateowner(sop);
324}
325
f1d110ca
BF		 326static void
327release_stateid_lockowners(struct nfs4_stateid *open_stp)
328{
329 struct nfs4_stateowner *lock_sop;
330
331 while (!list_empty(&open_stp->st_lockowners)) {
332 lock_sop = list_entry(open_stp->st_lockowners.next,
333 struct nfs4_stateowner, so_perstateid);
334 /* list_del(&open_stp->st_lockowners); */
335 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 336 release_lockowner(lock_sop);
f1d110ca
BF		 337 }
338}
339
2283963f
BF		 340static void release_open_stateid(struct nfs4_stateid *stp)
341{
342 unhash_generic_stateid(stp);
343 release_stateid_lockowners(stp);
344 nfsd_close(stp->st_vfs_file);
345 free_generic_stateid(stp);
346}
347
f044ff83 348static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 349{
350 struct nfs4_stateid *stp;
351
352 list_del(&sop->so_idhash);
353 list_del(&sop->so_strhash);
f044ff83
BF		 354 list_del(&sop->so_perclient);
355 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 356 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 357 stp = list_first_entry(&sop->so_stateids,
358 struct nfs4_stateid, st_perstateowner);
359 release_open_stateid(stp);
f1d110ca
BF		 360 }
361}
362
f044ff83 363static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 364{
f044ff83 365 unhash_openowner(sop);
f1d110ca
BF		 366 list_del(&sop->so_close_lru);
367 nfs4_put_stateowner(sop);
368}
369
5282fd72
ME		 370static DEFINE_SPINLOCK(sessionid_lock);
371#define SESSION_HASH_SIZE 512
372static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
373
374static inline int
375hash_sessionid(struct nfs4_sessionid *sessionid)
376{
377 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
378
379 return sid->sequence % SESSION_HASH_SIZE;
380}
381
382static inline void
383dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
384{
385 u32 *ptr = (u32 *)(&sessionid->data[0]);
386 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
387}
388
ec6b5d7b
AA		 389static void
390gen_sessionid(struct nfsd4_session *ses)
391{
392 struct nfs4_client *clp = ses->se_client;
393 struct nfsd4_sessionid *sid;
394
395 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
396 sid->clientid = clp->cl_clientid;
397 sid->sequence = current_sessionid++;
398 sid->reserved = 0;
399}
400
401/*
a649637c
AA		 402 * The protocol defines ca_maxresponssize_cached to include the size of
403 * the rpc header, but all we need to cache is the data starting after
404 * the end of the initial SEQUENCE operation--the rest we regenerate
405 * each time. Therefore we can advertise a ca_maxresponssize_cached
406 * value that is the number of bytes in our cache plus a few additional
407 * bytes. In order to stay on the safe side, and not promise more than
408 * we can cache, those additional bytes must be the minimum possible: 24
409 * bytes of rpc header (xid through accept state, with AUTH_NULL
410 * verifier), 12 for the compound header (with zero-length tag), and 44
411 * for the SEQUENCE op response:
412 */
413#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
414
415/*
416 * Give the client the number of ca_maxresponsesize_cached slots it
417 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
418 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
419 * than NFSD_MAX_SLOTS_PER_SESSION.
ec6b5d7b 420 *
a649637c
AA		 421 * If we run out of reserved DRC memory we should (up to a point)
422 * re-negotiate active sessions and reduce their slot usage to make
423 * rooom for new connections. For now we just fail the create session.
ec6b5d7b 424 */
a649637c 425static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
ec6b5d7b 426{
a649637c 427 int mem, size = fchan->maxresp_cached;
ec6b5d7b 428
5d77ddfb
AA		 429 if (fchan->maxreqs < 1)
430 return nfserr_inval;
5d77ddfb 431
a649637c
AA		 432 if (size < NFSD_MIN_HDR_SEQ_SZ)
433 size = NFSD_MIN_HDR_SEQ_SZ;
434 size -= NFSD_MIN_HDR_SEQ_SZ;
435 if (size > NFSD_SLOT_CACHE_SIZE)
436 size = NFSD_SLOT_CACHE_SIZE;
ec6b5d7b 437
a649637c
AA		 438 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
439 mem = fchan->maxreqs * size;
440 if (mem > NFSD_MAX_MEM_PER_SESSION) {
441 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
442 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
443 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
444 mem = fchan->maxreqs * size;
445 }
ec6b5d7b 446
4bd9b0f4 447 spin_lock(&nfsd_drc_lock);
a649637c
AA		 448 /* bound the total session drc memory ussage */
449 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
450 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
451 mem = fchan->maxreqs * size;
452 }
0c193054 453 nfsd_drc_mem_used += mem;
4bd9b0f4 454 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 455
b101ebbc 456 if (fchan->maxreqs == 0)
bdac86e2 457 return nfserr_serverfault;
a649637c
AA		 458
459 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
b101ebbc 460 return 0;
ec6b5d7b
AA		 461}
462
463/*
464 * fchan holds the client values on input, and the server values on output
ddc04fd4 465 * sv_max_mesg is the maximum payload plus one page for overhead.
ec6b5d7b
AA		 466 */
467static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 468 struct nfsd4_channel_attrs *session_fchan,
469 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 470{
471 int status = 0;
ddc04fd4 472 __u32 maxcount = nfsd_serv->sv_max_mesg;
ec6b5d7b
AA		 473
474 /* headerpadsz set to zero in encode routine */
475
476 /* Use the client's max request and max response size if possible */
477 if (fchan->maxreq_sz > maxcount)
478 fchan->maxreq_sz = maxcount;
6c18ba9f 479 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 480
481 if (fchan->maxresp_sz > maxcount)
482 fchan->maxresp_sz = maxcount;
6c18ba9f 483 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 484
ec6b5d7b
AA		 485 /* Use the client's maxops if possible */
486 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
487 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 488 session_fchan->maxops = fchan->maxops;
ec6b5d7b 489
ec6b5d7b
AA		 490 /* FIXME: Error means no more DRC pages so the server should
491 * recover pages from existing sessions. For now fail session
492 * creation.
493 */
a649637c 494 status = set_forechannel_drc_size(fchan);
ec6b5d7b 495
a649637c 496 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 497 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA		 498
499 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA		 500 return status;
501}
502
557ce264
AA		 503static void
504free_session_slots(struct nfsd4_session *ses)
505{
506 int i;
507
508 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
509 kfree(ses->se_slots[i]);
510}
511
efe0cb6d
BF		 512/*
513 * We don't actually need to cache the rpc and session headers, so we
514 * can allocate a little less for each slot:
515 */
516static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
517{
518 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
519}
520
ec6b5d7b
AA		 521static int
522alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
523 struct nfsd4_create_session *cses)
524{
525 struct nfsd4_session *new, tmp;
557ce264
AA		 526 struct nfsd4_slot *sp;
527 int idx, slotsize, cachesize, i;
528 int status;
ec6b5d7b
AA		 529
530 memset(&tmp, 0, sizeof(tmp));
531
532 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 533 tmp.se_bchannel = cses->back_channel;
534 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
535 &cses->fore_channel);
ec6b5d7b
AA		 536 if (status)
537 goto out;
538
557ce264
AA		 539 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
540 + sizeof(struct nfsd4_session) > PAGE_SIZE);
541
542 status = nfserr_serverfault;
543 /* allocate struct nfsd4_session and slot table pointers in one piece */
544 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA		 545 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
546 if (!new)
547 goto out;
548
549 memcpy(new, &tmp, sizeof(*new));
550
557ce264 551 /* allocate each struct nfsd4_slot and data cache in one piece */
efe0cb6d 552 cachesize = slot_bytes(&new->se_fchannel);
557ce264
AA		 553 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
554 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
555 if (!sp)
556 goto out_free;
557 new->se_slots[i] = sp;
558 }
559
ec6b5d7b
AA		 560 new->se_client = clp;
561 gen_sessionid(new);
562 idx = hash_sessionid(&new->se_sessionid);
563 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
564 NFS4_MAX_SESSIONID_LEN);
565
566 new->se_flags = cses->flags;
567 kref_init(&new->se_ref);
568 spin_lock(&sessionid_lock);
569 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
570 list_add(&new->se_perclnt, &clp->cl_sessions);
571 spin_unlock(&sessionid_lock);
572
573 status = nfs_ok;
574out:
575 return status;
557ce264
AA		 576out_free:
577 free_session_slots(new);
578 kfree(new);
579 goto out;
ec6b5d7b
AA		 580}
581
5282fd72
ME		 582/* caller must hold sessionid_lock */
583static struct nfsd4_session *
584find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
585{
586 struct nfsd4_session *elem;
587 int idx;
588
589 dump_sessionid(__func__, sessionid);
590 idx = hash_sessionid(sessionid);
591 dprintk("%s: idx is %d\n", __func__, idx);
592 /* Search in the appropriate list */
593 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
594 dump_sessionid("list traversal", &elem->se_sessionid);
595 if (!memcmp(elem->se_sessionid.data, sessionid->data,
596 NFS4_MAX_SESSIONID_LEN)) {
597 return elem;
598 }
599 }
600
601 dprintk("%s: session not found\n", __func__);
602 return NULL;
603}
604
605/* caller must hold sessionid_lock */
7116ed6b 606static void
5282fd72 607unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 608{
609 list_del(&ses->se_hash);
610 list_del(&ses->se_perclnt);
5282fd72
ME		 611}
612
613static void
614release_session(struct nfsd4_session *ses)
615{
616 spin_lock(&sessionid_lock);
617 unhash_session(ses);
618 spin_unlock(&sessionid_lock);
7116ed6b
AA		 619 nfsd4_put_session(ses);
620}
621
622void
623free_session(struct kref *kref)
624{
625 struct nfsd4_session *ses;
dd829c45 626 int mem;
7116ed6b
AA		 627
628 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 629 spin_lock(&nfsd_drc_lock);
efe0cb6d 630 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
dd829c45 631 nfsd_drc_mem_used -= mem;
be98d1bb 632 spin_unlock(&nfsd_drc_lock);
557ce264 633 free_session_slots(ses);
7116ed6b
AA		 634 kfree(ses);
635}
636
1da177e4
LT		 637static inline void
638renew_client(struct nfs4_client *clp)
639{
640 /*
641 * Move client to the end to the LRU list.
642 */
643 dprintk("renewing client (clientid %08x/%08x)\n",
644 clp->cl_clientid.cl_boot,
645 clp->cl_clientid.cl_id);
646 list_move_tail(&clp->cl_lru, &client_lru);
647 clp->cl_time = get_seconds();
648}
649
650/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
651static int
652STALE_CLIENTID(clientid_t *clid)
653{
654 if (clid->cl_boot == boot_time)
655 return 0;
60adfc50
AA		 656 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
657 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 658 return 1;
659}
660
661/*
662 * XXX Should we use a slab cache ?
663 * This type of memory management is somewhat inefficient, but we use it
664 * anyway since SETCLIENTID is not a common operation.
665 */
35bba9a3 666static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 667{
668 struct nfs4_client *clp;
669
35bba9a3
BF		 670 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
671 if (clp == NULL)
672 return NULL;
673 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
674 if (clp->cl_name.data == NULL) {
675 kfree(clp);
676 return NULL;
1da177e4 677 }
35bba9a3
BF		 678 memcpy(clp->cl_name.data, name.data, name.len);
679 clp->cl_name.len = name.len;
1da177e4
LT		 680 return clp;
681}
682
683static inline void
684free_client(struct nfs4_client *clp)
685{
686 if (clp->cl_cred.cr_group_info)
687 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 688 kfree(clp->cl_principal);
1da177e4
LT		 689 kfree(clp->cl_name.data);
690 kfree(clp);
691}
692
1da177e4
LT		 693static void
694expire_client(struct nfs4_client *clp)
695{
696 struct nfs4_stateowner *sop;
697 struct nfs4_delegation *dp;
1da177e4
LT		 698 struct list_head reaplist;
699
1da177e4
LT		 700 INIT_LIST_HEAD(&reaplist);
701 spin_lock(&recall_lock);
ea1da636
N		 702 while (!list_empty(&clp->cl_delegations)) {
703 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 704 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
705 dp->dl_flock);
ea1da636 706 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 707 list_move(&dp->dl_recall_lru, &reaplist);
708 }
709 spin_unlock(&recall_lock);
710 while (!list_empty(&reaplist)) {
711 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
712 list_del_init(&dp->dl_recall_lru);
713 unhash_delegation(dp);
714 }
715 list_del(&clp->cl_idhash);
716 list_del(&clp->cl_strhash);
717 list_del(&clp->cl_lru);
ea1da636
N		 718 while (!list_empty(&clp->cl_openowners)) {
719 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 720 release_openowner(sop);
1da177e4 721 }
c4bf7868
ME		 722 while (!list_empty(&clp->cl_sessions)) {
723 struct nfsd4_session *ses;
724 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
725 se_perclnt);
726 release_session(ses);
727 }
b5a1a81e 728 nfsd4_set_callback_client(clp, NULL);
147efd0d
BF		 729 if (clp->cl_cb_xprt)
730 svc_xprt_put(clp->cl_cb_xprt);
b12a05cb 731 free_client(clp);
1da177e4
LT		 732}
733
35bba9a3
BF		 734static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
735{
736 memcpy(target->cl_verifier.data, source->data,
737 sizeof(target->cl_verifier.data));
1da177e4
LT		 738}
739
35bba9a3
BF		 740static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
741{
1da177e4
LT		 742 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
743 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
744}
745
35bba9a3
BF		 746static void copy_cred(struct svc_cred *target, struct svc_cred *source)
747{
1da177e4
LT		 748 target->cr_uid = source->cr_uid;
749 target->cr_gid = source->cr_gid;
750 target->cr_group_info = source->cr_group_info;
751 get_group_info(target->cr_group_info);
752}
753
35bba9a3 754static int same_name(const char *n1, const char *n2)
599e0a22 755{
a55370a3 756 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 757}
758
759static int
599e0a22
BF		 760same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
761{
762 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 763}
764
765static int
599e0a22
BF		 766same_clid(clientid_t *cl1, clientid_t *cl2)
767{
768 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 769}
770
771/* XXX what about NGROUP */
772static int
599e0a22
BF		 773same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
774{
775 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 776}
777
5ec7b46c
BF		 778static void gen_clid(struct nfs4_client *clp)
779{
780 static u32 current_clientid = 1;
781
1da177e4
LT		 782 clp->cl_clientid.cl_boot = boot_time;
783 clp->cl_clientid.cl_id = current_clientid++;
784}
785
deda2faa
BF		 786static void gen_confirm(struct nfs4_client *clp)
787{
788 static u32 i;
789 u32 *p;
1da177e4 790
1da177e4 791 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 792 *p++ = get_seconds();
793 *p++ = i++;
1da177e4
LT		 794}
795
b09333c4
RL		 796static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
797 struct svc_rqst *rqstp, nfs4_verifier *verf)
798{
799 struct nfs4_client *clp;
800 struct sockaddr *sa = svc_addr(rqstp);
801 char *princ;
802
803 clp = alloc_client(name);
804 if (clp == NULL)
805 return NULL;
806
807 princ = svc_gss_principal(rqstp);
808 if (princ) {
809 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
810 if (clp->cl_principal == NULL) {
811 free_client(clp);
812 return NULL;
813 }
814 }
815
816 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
b09333c4
RL		 817 atomic_set(&clp->cl_cb_conn.cb_set, 0);
818 INIT_LIST_HEAD(&clp->cl_idhash);
819 INIT_LIST_HEAD(&clp->cl_strhash);
820 INIT_LIST_HEAD(&clp->cl_openowners);
821 INIT_LIST_HEAD(&clp->cl_delegations);
822 INIT_LIST_HEAD(&clp->cl_sessions);
823 INIT_LIST_HEAD(&clp->cl_lru);
824 clear_bit(0, &clp->cl_cb_slot_busy);
825 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
826 copy_verf(clp, verf);
827 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
828 clp->cl_flavor = rqstp->rq_flavor;
829 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
830 gen_confirm(clp);
831
832 return clp;
833}
834
35bba9a3
BF		 835static int check_name(struct xdr_netobj name)
836{
1da177e4
LT		 837 if (name.len == 0)
838 return 0;
839 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 840 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 841 return 0;
842 }
843 return 1;
844}
845
fd39ca9a 846static void
1da177e4
LT		 847add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
848{
849 unsigned int idhashval;
850
851 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
852 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
853 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
854 list_add_tail(&clp->cl_lru, &client_lru);
855 clp->cl_time = get_seconds();
856}
857
fd39ca9a 858static void
1da177e4
LT		 859move_to_confirmed(struct nfs4_client *clp)
860{
861 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
862 unsigned int strhashval;
863
864 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
865 list_del_init(&clp->cl_strhash);
f116629d 866 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 867 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 868 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
869 renew_client(clp);
870}
871
872static struct nfs4_client *
873find_confirmed_client(clientid_t *clid)
874{
875 struct nfs4_client *clp;
876 unsigned int idhashval = clientid_hashval(clid->cl_id);
877
878 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 879 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 880 return clp;
881 }
882 return NULL;
883}
884
885static struct nfs4_client *
886find_unconfirmed_client(clientid_t *clid)
887{
888 struct nfs4_client *clp;
889 unsigned int idhashval = clientid_hashval(clid->cl_id);
890
891 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 892 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 893 return clp;
894 }
895 return NULL;
896}
897
a1bcecd2
AA		 898/*
899 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
900 * parameter. Matching is based on the fact the at least one of the
901 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
902 *
903 * FIXME: we need to unify the clientid namespaces for nfsv4.x
904 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
905 * and SET_CLIENTID{,_CONFIRM}
906 */
907static inline int
908match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
909{
910 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
911 return use_exchange_id == has_exchange_flags;
912}
913
28ce6054 914static struct nfs4_client *
a1bcecd2
AA		 915find_confirmed_client_by_str(const char *dname, unsigned int hashval,
916 bool use_exchange_id)
28ce6054
N		 917{
918 struct nfs4_client *clp;
919
920 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 921 if (same_name(clp->cl_recdir, dname) &&
922 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 923 return clp;
924 }
925 return NULL;
926}
927
928static struct nfs4_client *
a1bcecd2
AA		 929find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
930 bool use_exchange_id)
28ce6054
N		 931{
932 struct nfs4_client *clp;
933
934 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 935 if (same_name(clp->cl_recdir, dname) &&
936 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 937 return clp;
938 }
939 return NULL;
940}
941
fd39ca9a 942static void
fbf4665f 943gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 944{
c237dc03 945 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 946 unsigned short expected_family;
947
948 /* Currently, we only support tcp and tcp6 for the callback channel */
949 if (se->se_callback_netid_len == 3 &&
950 !memcmp(se->se_callback_netid_val, "tcp", 3))
951 expected_family = AF_INET;
952 else if (se->se_callback_netid_len == 4 &&
953 !memcmp(se->se_callback_netid_val, "tcp6", 4))
954 expected_family = AF_INET6;
955 else
1da177e4
LT		 956 goto out_err;
957
aa9a4ec7
JL		 958 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
959 se->se_callback_addr_len,
960 (struct sockaddr *) &cb->cb_addr,
961 sizeof(cb->cb_addr));
962
7077ecba 963 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 964 goto out_err;
aa9a4ec7 965
fbf4665f
JL		 966 if (cb->cb_addr.ss_family == AF_INET6)
967 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
968
ab52ae6d 969 cb->cb_minorversion = 0;
1da177e4
LT		 970 cb->cb_prog = se->se_callback_prog;
971 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 972 return;
973out_err:
aa9a4ec7
JL		 974 cb->cb_addr.ss_family = AF_UNSPEC;
975 cb->cb_addrlen = 0;
849823c5 976 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 977 "will not receive delegations\n",
978 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
979
1da177e4
LT		 980 return;
981}
982
074fe897 983/*
557ce264 984 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897 985 */
074fe897
AA		 986void
987nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
074fe897 988{
557ce264
AA		 989 struct nfsd4_slot *slot = resp->cstate.slot;
990 unsigned int base;
074fe897 991
557ce264 992 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 993
557ce264
AA		 994 slot->sl_opcnt = resp->opcnt;
995 slot->sl_status = resp->cstate.status;
074fe897 996
bf864a31 997 if (nfsd4_not_cached(resp)) {
557ce264 998 slot->sl_datalen = 0;
bf864a31 999 return;
074fe897 1000 }
557ce264
AA		 1001 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1002 base = (char *)resp->cstate.datap -
1003 (char *)resp->xbuf->head[0].iov_base;
1004 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1005 slot->sl_datalen))
1006 WARN("%s: sessions DRC could not cache compound\n", __func__);
1007 return;
074fe897
AA		 1008}
1009
1010/*
abfabf8c
AA		 1011 * Encode the replay sequence operation from the slot values.
1012 * If cachethis is FALSE encode the uncached rep error on the next
1013 * operation which sets resp->p and increments resp->opcnt for
1014 * nfs4svc_encode_compoundres.
074fe897 1015 *
074fe897 1016 */
abfabf8c
AA		 1017static __be32
1018nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1019 struct nfsd4_compoundres *resp)
074fe897 1020{
abfabf8c
AA		 1021 struct nfsd4_op *op;
1022 struct nfsd4_slot *slot = resp->cstate.slot;
bf864a31 1023
abfabf8c 1024 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1025 resp->opcnt, resp->cstate.slot->sl_cachethis);
bf864a31 1026
abfabf8c
AA		 1027 /* Encode the replayed sequence operation */
1028 op = &args->ops[resp->opcnt - 1];
1029 nfsd4_encode_operation(resp, op);
bf864a31 1030
abfabf8c 1031 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1032 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA		 1033 op = &args->ops[resp->opcnt++];
1034 op->status = nfserr_retry_uncached_rep;
1035 nfsd4_encode_operation(resp, op);
074fe897 1036 }
abfabf8c 1037 return op->status;
074fe897
AA		 1038}
1039
1040/*
557ce264
AA		 1041 * The sequence operation is not cached because we can use the slot and
1042 * session values.
074fe897
AA		 1043 */
1044__be32
bf864a31
AA		 1045nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1046 struct nfsd4_sequence *seq)
074fe897 1047{
557ce264 1048 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA		 1049 __be32 status;
1050
557ce264 1051 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1052
abfabf8c
AA		 1053 /* Either returns 0 or nfserr_retry_uncached */
1054 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1055 if (status == nfserr_retry_uncached_rep)
1056 return status;
074fe897 1057
557ce264
AA		 1058 /* The sequence operation has been encoded, cstate->datap set. */
1059 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1060
557ce264
AA		 1061 resp->opcnt = slot->sl_opcnt;
1062 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1063 status = slot->sl_status;
074fe897
AA		 1064
1065 return status;
1066}
1067
0733d213
AA		 1068/*
1069 * Set the exchange_id flags returned by the server.
1070 */
1071static void
1072nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1073{
1074 /* pNFS is not supported */
1075 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1076
1077 /* Referrals are supported, Migration is not. */
1078 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1079
1080 /* set the wire flags to return to client. */
1081 clid->flags = new->cl_exchange_flags;
1082}
1083
069b6ad4
AA		 1084__be32
1085nfsd4_exchange_id(struct svc_rqst *rqstp,
1086 struct nfsd4_compound_state *cstate,
1087 struct nfsd4_exchange_id *exid)
1088{
0733d213
AA		 1089 struct nfs4_client *unconf, *conf, *new;
1090 int status;
1091 unsigned int strhashval;
1092 char dname[HEXDIR_LEN];
363168b4 1093 char addr_str[INET6_ADDRSTRLEN];
0733d213 1094 nfs4_verifier verf = exid->verifier;
363168b4 1095 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1096
363168b4 1097 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1098 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1099 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1100 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1101 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1102
1103 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1104 return nfserr_inval;
1105
1106 /* Currently only support SP4_NONE */
1107 switch (exid->spa_how) {
1108 case SP4_NONE:
1109 break;
1110 case SP4_SSV:
1111 return nfserr_encr_alg_unsupp;
1112 default:
1113 BUG(); /* checked by xdr code */
1114 case SP4_MACH_CRED:
1115 return nfserr_serverfault; /* no excuse :-/ */
1116 }
1117
1118 status = nfs4_make_rec_clidname(dname, &exid->clname);
1119
1120 if (status)
1121 goto error;
1122
1123 strhashval = clientstr_hashval(dname);
1124
1125 nfs4_lock_state();
1126 status = nfs_ok;
1127
a1bcecd2 1128 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1129 if (conf) {
1130 if (!same_verf(&verf, &conf->cl_verifier)) {
1131 /* 18.35.4 case 8 */
1132 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1133 status = nfserr_not_same;
1134 goto out;
1135 }
1136 /* Client reboot: destroy old state */
1137 expire_client(conf);
1138 goto out_new;
1139 }
1140 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1141 /* 18.35.4 case 9 */
1142 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1143 status = nfserr_perm;
1144 goto out;
1145 }
1146 expire_client(conf);
1147 goto out_new;
1148 }
0733d213
AA		 1149 /*
1150 * Set bit when the owner id and verifier map to an already
1151 * confirmed client id (18.35.3).
1152 */
1153 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1154
1155 /*
1156 * Falling into 18.35.4 case 2, possible router replay.
1157 * Leave confirmed record intact and return same result.
1158 */
1159 copy_verf(conf, &verf);
1160 new = conf;
1161 goto out_copy;
6ddbbbfe
MS		 1162 }
1163
1164 /* 18.35.4 case 7 */
1165 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1166 status = nfserr_noent;
1167 goto out;
0733d213
AA		 1168 }
1169
a1bcecd2 1170 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1171 if (unconf) {
1172 /*
1173 * Possible retry or client restart. Per 18.35.4 case 4,
1174 * a new unconfirmed record should be generated regardless
1175 * of whether any properties have changed.
1176 */
1177 expire_client(unconf);
1178 }
1179
1180out_new:
1181 /* Normal case */
b09333c4 1182 new = create_client(exid->clname, dname, rqstp, &verf);
0733d213 1183 if (new == NULL) {
bdac86e2 1184 status = nfserr_serverfault;
0733d213
AA		 1185 goto out;
1186 }
1187
0733d213 1188 gen_clid(new);
0733d213
AA		 1189 add_to_unconfirmed(new, strhashval);
1190out_copy:
1191 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1192 exid->clientid.cl_id = new->cl_clientid.cl_id;
1193
38eb76a5 1194 exid->seqid = 1;
0733d213
AA		 1195 nfsd4_set_ex_flags(new, exid);
1196
1197 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1198 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1199 status = nfs_ok;
1200
1201out:
1202 nfs4_unlock_state();
1203error:
1204 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1205 return status;
069b6ad4
AA		 1206}
1207
b85d4c01 1208static int
88e588d5 1209check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1210{
88e588d5
AA		 1211 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1212 slot_seqid);
b85d4c01
BH		 1213
1214 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1215 if (slot_inuse) {
1216 if (seqid == slot_seqid)
b85d4c01
BH		 1217 return nfserr_jukebox;
1218 else
1219 return nfserr_seq_misordered;
1220 }
1221 /* Normal */
88e588d5 1222 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1223 return nfs_ok;
1224 /* Replay */
88e588d5 1225 if (seqid == slot_seqid)
b85d4c01
BH		 1226 return nfserr_replay_cache;
1227 /* Wraparound */
88e588d5 1228 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1229 return nfs_ok;
1230 /* Misordered replay or misordered new request */
1231 return nfserr_seq_misordered;
1232}
1233
49557cc7
AA		 1234/*
1235 * Cache the create session result into the create session single DRC
1236 * slot cache by saving the xdr structure. sl_seqid has been set.
1237 * Do this for solo or embedded create session operations.
1238 */
1239static void
1240nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1241 struct nfsd4_clid_slot *slot, int nfserr)
1242{
1243 slot->sl_status = nfserr;
1244 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1245}
1246
1247static __be32
1248nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1249 struct nfsd4_clid_slot *slot)
1250{
1251 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1252 return slot->sl_status;
1253}
1254
069b6ad4
AA		 1255__be32
1256nfsd4_create_session(struct svc_rqst *rqstp,
1257 struct nfsd4_compound_state *cstate,
1258 struct nfsd4_create_session *cr_ses)
1259{
363168b4 1260 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1261 struct nfs4_client *conf, *unconf;
49557cc7 1262 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1263 int status = 0;
1264
1265 nfs4_lock_state();
1266 unconf = find_unconfirmed_client(&cr_ses->clientid);
1267 conf = find_confirmed_client(&cr_ses->clientid);
1268
1269 if (conf) {
49557cc7
AA		 1270 cs_slot = &conf->cl_cs_slot;
1271 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1272 if (status == nfserr_replay_cache) {
ec6b5d7b 1273 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1274 cs_slot->sl_seqid);
38eb76a5 1275 /* Return the cached reply status */
49557cc7 1276 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1277 goto out;
49557cc7 1278 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1279 status = nfserr_seq_misordered;
1280 dprintk("Sequence misordered!\n");
1281 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1282 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1283 goto out;
1284 }
49557cc7 1285 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1286 } else if (unconf) {
1287 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1288 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1289 status = nfserr_clid_inuse;
1290 goto out;
1291 }
1292
49557cc7
AA		 1293 cs_slot = &unconf->cl_cs_slot;
1294 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1295 if (status) {
1296 /* an unconfirmed replay returns misordered */
ec6b5d7b 1297 status = nfserr_seq_misordered;
49557cc7 1298 goto out_cache;
ec6b5d7b
AA		 1299 }
1300
49557cc7 1301 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1302 move_to_confirmed(unconf);
1303
38524ab3
AA		 1304 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1305 unconf->cl_cb_xprt = rqstp->rq_xprt;
1306 svc_xprt_get(unconf->cl_cb_xprt);
1307 rpc_copy_addr(
1308 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1309 sa);
1310 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1311 unconf->cl_cb_conn.cb_minorversion =
1312 cstate->minorversion;
1313 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1314 unconf->cl_cb_seq_nr = 1;
38524ab3
AA		 1315 nfsd4_probe_callback(unconf);
1316 }
ec6b5d7b
AA		 1317 conf = unconf;
1318 } else {
1319 status = nfserr_stale_clientid;
1320 goto out;
1321 }
1322
408b79bc
BF		 1323 /*
1324 * We do not support RDMA or persistent sessions
1325 */
1326 cr_ses->flags &= ~SESSION4_PERSIST;
1327 cr_ses->flags &= ~SESSION4_RDMA;
1328
ec6b5d7b
AA		 1329 status = alloc_init_session(rqstp, conf, cr_ses);
1330 if (status)
1331 goto out;
1332
ec6b5d7b
AA		 1333 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1334 NFS4_MAX_SESSIONID_LEN);
49557cc7 1335 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1336
49557cc7
AA		 1337out_cache:
1338 /* cache solo and embedded create sessions under the state lock */
1339 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1340out:
1341 nfs4_unlock_state();
1342 dprintk("%s returns %d\n", __func__, ntohl(status));
1343 return status;
069b6ad4
AA		 1344}
1345
1346__be32
1347nfsd4_destroy_session(struct svc_rqst *r,
1348 struct nfsd4_compound_state *cstate,
1349 struct nfsd4_destroy_session *sessionid)
1350{
e10e0cfc
BH		 1351 struct nfsd4_session *ses;
1352 u32 status = nfserr_badsession;
1353
1354 /* Notes:
1355 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1356 * - Should we return nfserr_back_chan_busy if waiting for
1357 * callbacks on to-be-destroyed session?
1358 * - Do we need to clear any callback info from previous session?
1359 */
1360
1361 dump_sessionid(__func__, &sessionid->sessionid);
1362 spin_lock(&sessionid_lock);
1363 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1364 if (!ses) {
1365 spin_unlock(&sessionid_lock);
1366 goto out;
1367 }
1368
1369 unhash_session(ses);
1370 spin_unlock(&sessionid_lock);
1371
1372 /* wait for callbacks */
b5a1a81e 1373 nfsd4_set_callback_client(ses->se_client, NULL);
e10e0cfc
BH		 1374 nfsd4_put_session(ses);
1375 status = nfs_ok;
1376out:
1377 dprintk("%s returns %d\n", __func__, ntohl(status));
1378 return status;
069b6ad4
AA		 1379}
1380
1381__be32
b85d4c01 1382nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1383 struct nfsd4_compound_state *cstate,
1384 struct nfsd4_sequence *seq)
1385{
f9bb94c4 1386 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1387 struct nfsd4_session *session;
1388 struct nfsd4_slot *slot;
1389 int status;
1390
f9bb94c4
AA		 1391 if (resp->opcnt != 1)
1392 return nfserr_sequence_pos;
1393
b85d4c01
BH		 1394 spin_lock(&sessionid_lock);
1395 status = nfserr_badsession;
1396 session = find_in_sessionid_hashtbl(&seq->sessionid);
1397 if (!session)
1398 goto out;
1399
1400 status = nfserr_badslot;
6c18ba9f 1401 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1402 goto out;
1403
557ce264 1404 slot = session->se_slots[seq->slotid];
b85d4c01
BH		 1405 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1406
a8dfdaeb
AA		 1407 /* We do not negotiate the number of slots yet, so set the
1408 * maxslots to the session maxreqs which is used to encode
1409 * sr_highest_slotid and the sr_target_slot id to maxslots */
1410 seq->maxslots = session->se_fchannel.maxreqs;
1411
88e588d5 1412 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1413 if (status == nfserr_replay_cache) {
1414 cstate->slot = slot;
1415 cstate->session = session;
da3846a2 1416 /* Return the cached reply status and set cstate->status
557ce264 1417 * for nfsd4_proc_compound processing */
bf864a31 1418 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1419 cstate->status = nfserr_replay_cache;
aaf84eb9 1420 goto out;
b85d4c01
BH		 1421 }
1422 if (status)
1423 goto out;
1424
1425 /* Success! bump slot seqid */
1426 slot->sl_inuse = true;
1427 slot->sl_seqid = seq->seqid;
557ce264 1428 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH		 1429
1430 cstate->slot = slot;
1431 cstate->session = session;
1432
aaf84eb9 1433 /* Hold a session reference until done processing the compound:
b85d4c01
BH		 1434 * nfsd4_put_session called only if the cstate slot is set.
1435 */
b85d4c01
BH		 1436 nfsd4_get_session(session);
1437out:
1438 spin_unlock(&sessionid_lock);
aaf84eb9
BH		 1439 /* Renew the clientid on success and on replay */
1440 if (cstate->session) {
1441 nfs4_lock_state();
1442 renew_client(session->se_client);
1443 nfs4_unlock_state();
1444 }
b85d4c01
BH		 1445 dprintk("%s: return %d\n", __func__, ntohl(status));
1446 return status;
069b6ad4
AA		 1447}
1448
b37ad28b 1449__be32
b591480b
BF		 1450nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1451 struct nfsd4_setclientid *setclid)
1da177e4 1452{
363168b4 1453 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1454 struct xdr_netobj clname = {
1455 .len = setclid->se_namelen,
1456 .data = setclid->se_name,
1457 };
1458 nfs4_verifier clverifier = setclid->se_verf;
1459 unsigned int strhashval;
28ce6054 1460 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1461 __be32 status;
a55370a3 1462 char dname[HEXDIR_LEN];
1da177e4 1463
1da177e4 1464 if (!check_name(clname))
73aea4ec 1465 return nfserr_inval;
1da177e4 1466
a55370a3
N		 1467 status = nfs4_make_rec_clidname(dname, &clname);
1468 if (status)
73aea4ec 1469 return status;
a55370a3 1470
1da177e4
LT		 1471 /*
1472 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1473 * We get here on a DRC miss.
1474 */
1475
a55370a3 1476 strhashval = clientstr_hashval(dname);
1da177e4 1477
1da177e4 1478 nfs4_lock_state();
a1bcecd2 1479 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1480 if (conf) {
a186e767 1481 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1482 status = nfserr_clid_inuse;
026722c2 1483 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1484 char addr_str[INET6_ADDRSTRLEN];
1485 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1486 sizeof(addr_str));
1487 dprintk("NFSD: setclientid: string in use by client "
1488 "at %s\n", addr_str);
1da177e4
LT		 1489 goto out;
1490 }
1da177e4 1491 }
a186e767
BF		 1492 /*
1493 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1494 * has a description of SETCLIENTID request processing consisting
1495 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1496 */
a1bcecd2 1497 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1498 status = nfserr_resource;
1499 if (!conf) {
a186e767
BF		 1500 /*
1501 * RFC 3530 14.2.33 CASE 4:
1502 * placed first, because it is the normal case
1da177e4
LT		 1503 */
1504 if (unconf)
1505 expire_client(unconf);
b09333c4 1506 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1507 if (new == NULL)
1da177e4 1508 goto out;
1da177e4 1509 gen_clid(new);
599e0a22 1510 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1511 /*
a186e767
BF		 1512 * RFC 3530 14.2.33 CASE 1:
1513 * probable callback update
1da177e4 1514 */
31f4a6c1
N		 1515 if (unconf) {
1516 /* Note this is removing unconfirmed {*x***},
1517 * which is stronger than RFC recommended {vxc**}.
1518 * This has the advantage that there is at most
1519 * one {*x***} in either list at any time.
1520 */
1521 expire_client(unconf);
1da177e4 1522 }
b09333c4 1523 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1524 if (new == NULL)
1da177e4 1525 goto out;
1da177e4 1526 copy_clid(new, conf);
1da177e4
LT		 1527 } else if (!unconf) {
1528 /*
a186e767
BF		 1529 * RFC 3530 14.2.33 CASE 2:
1530 * probable client reboot; state will be removed if
1531 * confirmed.
1da177e4 1532 */
b09333c4 1533 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1534 if (new == NULL)
1da177e4 1535 goto out;
1da177e4 1536 gen_clid(new);
49ba8781 1537 } else {
a186e767
BF		 1538 /*
1539 * RFC 3530 14.2.33 CASE 3:
1540 * probable client reboot; state will be removed if
1541 * confirmed.
1da177e4
LT		 1542 */
1543 expire_client(unconf);
b09333c4 1544 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1545 if (new == NULL)
1da177e4 1546 goto out;
1da177e4 1547 gen_clid(new);
1da177e4 1548 }
fbf4665f 1549 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1550 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1551 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1552 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1553 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1554 status = nfs_ok;
1555out:
1556 nfs4_unlock_state();
1557 return status;
1558}
1559
1560
1561/*
a186e767
BF		 1562 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1563 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1564 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1565 */
b37ad28b 1566__be32
b591480b
BF		 1567nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1568 struct nfsd4_compound_state *cstate,
1569 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1570{
363168b4 1571 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1572 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1573 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1574 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1575 __be32 status;
1da177e4
LT		 1576
1577 if (STALE_CLIENTID(clid))
1578 return nfserr_stale_clientid;
1579 /*
1580 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1581 * We get here on a DRC miss.
1582 */
1583
1584 nfs4_lock_state();
21ab45a4
N		 1585
1586 conf = find_confirmed_client(clid);
1587 unconf = find_unconfirmed_client(clid);
1588
1589 status = nfserr_clid_inuse;
363168b4 1590 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1591 goto out;
363168b4 1592 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1593 goto out;
1594
a186e767
BF		 1595 /*
1596 * section 14.2.34 of RFC 3530 has a description of
1597 * SETCLIENTID_CONFIRM request processing consisting
1598 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1599 */
366e0c1d 1600 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1601 /*
1602 * RFC 3530 14.2.34 CASE 1:
1603 * callback update
1604 */
599e0a22 1605 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1606 status = nfserr_clid_inuse;
1607 else {
1a69c179
N		 1608 /* XXX: We just turn off callbacks until we can handle
1609 * change request correctly. */
c237dc03 1610 atomic_set(&conf->cl_cb_conn.cb_set, 0);
1a69c179 1611 expire_client(unconf);
1da177e4 1612 status = nfs_ok;
1a69c179 1613
1da177e4 1614 }
f3aba4e5 1615 } else if (conf && !unconf) {
a186e767
BF		 1616 /*
1617 * RFC 3530 14.2.34 CASE 2:
1618 * probable retransmitted request; play it safe and
1619 * do nothing.
7c79f737 1620 */
599e0a22 1621 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1622 status = nfserr_clid_inuse;
21ab45a4 1623 else
1da177e4 1624 status = nfs_ok;
7c79f737 1625 } else if (!conf && unconf
599e0a22 1626 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1627 /*
1628 * RFC 3530 14.2.34 CASE 3:
1629 * Normal case; new or rebooted client:
7c79f737 1630 */
599e0a22 1631 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1632 status = nfserr_clid_inuse;
1633 } else {
1a69c179
N		 1634 unsigned int hash =
1635 clientstr_hashval(unconf->cl_recdir);
1636 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1637 hash, false);
1a69c179 1638 if (conf) {
c7b9a459 1639 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1640 expire_client(conf);
1641 }
1da177e4 1642 move_to_confirmed(unconf);
21ab45a4 1643 conf = unconf;
46f8a64b 1644 nfsd4_probe_callback(conf);
1a69c179 1645 status = nfs_ok;
1da177e4 1646 }
599e0a22
BF		 1647 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1648 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1649 &confirm)))) {
a186e767
BF		 1650 /*
1651 * RFC 3530 14.2.34 CASE 4:
1652 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1653 */
1da177e4 1654 status = nfserr_stale_clientid;
7c79f737 1655 } else {
08e8987c
N		 1656 /* check that we have hit one of the cases...*/
1657 status = nfserr_clid_inuse;
1658 }
1da177e4 1659out:
1da177e4
LT		 1660 nfs4_unlock_state();
1661 return status;
1662}
1663
1da177e4
LT		 1664/* OPEN Share state helper functions */
1665static inline struct nfs4_file *
1666alloc_init_file(struct inode *ino)
1667{
1668 struct nfs4_file *fp;
1669 unsigned int hashval = file_hashval(ino);
1670
e60d4398
N		 1671 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1672 if (fp) {
8b671b80 1673 atomic_set(&fp->fi_ref, 1);
1da177e4 1674 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1675 INIT_LIST_HEAD(&fp->fi_stateids);
1676 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1677 spin_lock(&recall_lock);
1da177e4 1678 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1679 spin_unlock(&recall_lock);
1da177e4
LT		 1680 fp->fi_inode = igrab(ino);
1681 fp->fi_id = current_fileid++;
47f9940c 1682 fp->fi_had_conflict = false;
1da177e4
LT		 1683 return fp;
1684 }
1685 return NULL;
1686}
1687
e60d4398 1688static void
e18b890b 1689nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1690{
e60d4398
N		 1691 if (*slab == NULL)
1692 return;
1a1d92c1 1693 kmem_cache_destroy(*slab);
e60d4398 1694 *slab = NULL;
1da177e4
LT		 1695}
1696
e8ff2a84 1697void
1da177e4
LT		 1698nfsd4_free_slabs(void)
1699{
e60d4398
N		 1700 nfsd4_free_slab(&stateowner_slab);
1701 nfsd4_free_slab(&file_slab);
5ac049ac 1702 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1703 nfsd4_free_slab(&deleg_slab);
e60d4398 1704}
1da177e4 1705
e60d4398
N		 1706static int
1707nfsd4_init_slabs(void)
1708{
1709 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1710 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1711 if (stateowner_slab == NULL)
1712 goto out_nomem;
1713 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1714 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1715 if (file_slab == NULL)
1716 goto out_nomem;
5ac049ac 1717 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1718 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1719 if (stateid_slab == NULL)
1720 goto out_nomem;
5b2d21c1 1721 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1722 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1723 if (deleg_slab == NULL)
1724 goto out_nomem;
e60d4398
N		 1725 return 0;
1726out_nomem:
1727 nfsd4_free_slabs();
1728 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1729 return -ENOMEM;
1da177e4
LT		 1730}
1731
1732void
1733nfs4_free_stateowner(struct kref *kref)
1734{
1735 struct nfs4_stateowner *sop =
1736 container_of(kref, struct nfs4_stateowner, so_ref);
1737 kfree(sop->so_owner.data);
1738 kmem_cache_free(stateowner_slab, sop);
1739}
1740
1741static inline struct nfs4_stateowner *
1742alloc_stateowner(struct xdr_netobj *owner)
1743{
1744 struct nfs4_stateowner *sop;
1745
1746 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1747 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1748 memcpy(sop->so_owner.data, owner->data, owner->len);
1749 sop->so_owner.len = owner->len;
1750 kref_init(&sop->so_ref);
1751 return sop;
1752 }
1753 kmem_cache_free(stateowner_slab, sop);
1754 }
1755 return NULL;
1756}
1757
1758static struct nfs4_stateowner *
1759alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1760 struct nfs4_stateowner *sop;
1761 struct nfs4_replay *rp;
1762 unsigned int idhashval;
1763
1764 if (!(sop = alloc_stateowner(&open->op_owner)))
1765 return NULL;
1766 idhashval = ownerid_hashval(current_ownerid);
1767 INIT_LIST_HEAD(&sop->so_idhash);
1768 INIT_LIST_HEAD(&sop->so_strhash);
1769 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1770 INIT_LIST_HEAD(&sop->so_stateids);
1771 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1772 INIT_LIST_HEAD(&sop->so_close_lru);
1773 sop->so_time = 0;
1774 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1775 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1776 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1777 sop->so_is_open_owner = 1;
1778 sop->so_id = current_ownerid++;
1779 sop->so_client = clp;
1780 sop->so_seqid = open->op_seqid;
1781 sop->so_confirmed = 0;
1782 rp = &sop->so_replay;
de1ae286 1783 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1784 rp->rp_buflen = 0;
1785 rp->rp_buf = rp->rp_ibuf;
1786 return sop;
1787}
1788
1da177e4
LT		 1789static inline void
1790init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1791 struct nfs4_stateowner *sop = open->op_stateowner;
1792 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1793
1794 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1795 INIT_LIST_HEAD(&stp->st_perstateowner);
1796 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1797 INIT_LIST_HEAD(&stp->st_perfile);
1798 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1799 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1800 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1801 stp->st_stateowner = sop;
13cd2184 1802 get_nfs4_file(fp);
1da177e4 1803 stp->st_file = fp;
78155ed7 1804 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1805 stp->st_stateid.si_stateownerid = sop->so_id;
1806 stp->st_stateid.si_fileid = fp->fi_id;
1807 stp->st_stateid.si_generation = 0;
1808 stp->st_access_bmap = 0;
1809 stp->st_deny_bmap = 0;
84459a11
AA		 1810 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1811 &stp->st_access_bmap);
1da177e4 1812 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1813 stp->st_openstp = NULL;
1da177e4
LT		 1814}
1815
fd39ca9a 1816static void
1da177e4
LT		 1817move_to_close_lru(struct nfs4_stateowner *sop)
1818{
1819 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1820
358dd55a 1821 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1822 sop->so_time = get_seconds();
1823}
1824
1da177e4 1825static int
599e0a22
BF		 1826same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1827 clientid_t *clid)
1828{
1829 return (sop->so_owner.len == owner->len) &&
1830 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1831 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1832}
1833
1834static struct nfs4_stateowner *
1835find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1836{
1837 struct nfs4_stateowner *so = NULL;
1838
1839 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1840 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1841 return so;
1842 }
1843 return NULL;
1844}
1845
1846/* search file_hashtbl[] for file */
1847static struct nfs4_file *
1848find_file(struct inode *ino)
1849{
1850 unsigned int hashval = file_hashval(ino);
1851 struct nfs4_file *fp;
1852
8b671b80 1853 spin_lock(&recall_lock);
1da177e4 1854 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1855 if (fp->fi_inode == ino) {
1856 get_nfs4_file(fp);
8b671b80 1857 spin_unlock(&recall_lock);
1da177e4 1858 return fp;
13cd2184 1859 }
1da177e4 1860 }
8b671b80 1861 spin_unlock(&recall_lock);
1da177e4
LT		 1862 return NULL;
1863}
1864
d87a8ade 1865static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1866{
d87a8ade 1867 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1868 return 0;
d87a8ade
AA		 1869 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1870 return 0;
1871 x &= ~NFS4_SHARE_ACCESS_MASK;
1872 if (minorversion && x) {
1873 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1874 return 0;
1875 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1876 return 0;
1877 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1878 }
1879 if (x)
8838dc43
BF		 1880 return 0;
1881 return 1;
ba5a6a19
BF		 1882}
1883
8838dc43 1884static inline int deny_valid(u32 x)
ba5a6a19 1885{
8838dc43
BF		 1886 /* Note: unlike access bits, deny bits may be zero. */
1887 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1888}
1da177e4 1889
4f83aa30
BF		 1890/*
1891 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1892 * st_{access,deny}_bmap field of the stateid, in order to track not
1893 * only what share bits are currently in force, but also what
1894 * combinations of share bits previous opens have used. This allows us
1895 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1896 * return an error if the client attempt to downgrade to a combination
1897 * of share bits not explicable by closing some of its previous opens.
1898 *
1899 * XXX: This enforcement is actually incomplete, since we don't keep
1900 * track of access/deny bit combinations; so, e.g., we allow:
1901 *
1902 * OPEN allow read, deny write
1903 * OPEN allow both, deny none
1904 * DOWNGRADE allow read, deny none
1905 *
1906 * which we should reject.
1907 */
fd39ca9a 1908static void
1da177e4
LT		 1909set_access(unsigned int *access, unsigned long bmap) {
1910 int i;
1911
1912 *access = 0;
1913 for (i = 1; i < 4; i++) {
1914 if (test_bit(i, &bmap))
1915 *access |= i;
1916 }
1917}
1918
fd39ca9a 1919static void
1da177e4
LT		 1920set_deny(unsigned int *deny, unsigned long bmap) {
1921 int i;
1922
1923 *deny = 0;
1924 for (i = 0; i < 4; i++) {
1925 if (test_bit(i, &bmap))
1926 *deny |= i ;
1927 }
1928}
1929
1930static int
1931test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1932 unsigned int access, deny;
1933
1934 set_access(&access, stp->st_access_bmap);
1935 set_deny(&deny, stp->st_deny_bmap);
1936 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1937 return 0;
1938 return 1;
1939}
1940
1941/*
1942 * Called to check deny when READ with all zero stateid or
1943 * WRITE with all zero or all one stateid
1944 */
b37ad28b 1945static __be32
1da177e4
LT		 1946nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1947{
1948 struct inode *ino = current_fh->fh_dentry->d_inode;
1949 struct nfs4_file *fp;
1950 struct nfs4_stateid *stp;
b37ad28b 1951 __be32 ret;
1da177e4
LT		 1952
1953 dprintk("NFSD: nfs4_share_conflict\n");
1954
1955 fp = find_file(ino);
13cd2184
N		 1956 if (!fp)
1957 return nfs_ok;
b700949b 1958 ret = nfserr_locked;
1da177e4 1959 /* Search for conflicting share reservations */
13cd2184
N		 1960 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1961 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1962 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1963 goto out;
1da177e4 1964 }
13cd2184
N		 1965 ret = nfs_ok;
1966out:
1967 put_nfs4_file(fp);
1968 return ret;
1da177e4
LT		 1969}
1970
1971static inline void
1972nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1973{
1974 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 1975 drop_file_write_access(filp);
42e49608 1976 spin_lock(&filp->f_lock);
1da177e4 1977 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
42e49608 1978 spin_unlock(&filp->f_lock);
1da177e4
LT		 1979 }
1980}
1981
1da177e4
LT		 1982/*
1983 * Spawn a thread to perform a recall on the delegation represented
1984 * by the lease (file_lock)
1985 *
1986 * Called from break_lease() with lock_kernel() held.
1987 * Note: we assume break_lease will only call this *once* for any given
1988 * lease.
1989 */
1990static
1991void nfsd_break_deleg_cb(struct file_lock *fl)
1992{
63e4863f 1993 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 1994
1995 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1996 if (!dp)
1997 return;
1998
1999 /* We're assuming the state code never drops its reference
2000 * without first removing the lease. Since we're in this lease
2001 * callback (and since the lease code is serialized by the kernel
2002 * lock) we know the server hasn't removed the lease yet, we know
2003 * it's safe to take a reference: */
2004 atomic_inc(&dp->dl_count);
2005
2006 spin_lock(&recall_lock);
2007 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2008 spin_unlock(&recall_lock);
2009
2010 /* only place dl_time is set. protected by lock_kernel*/
2011 dp->dl_time = get_seconds();
2012
0272e1fd
BF		 2013 /*
2014 * We don't want the locks code to timeout the lease for us;
2015 * we'll remove it ourself if the delegation isn't returned
2016 * in time.
2017 */
2018 fl->fl_break_time = 0;
1da177e4 2019
63e4863f
BF		 2020 dp->dl_file->fi_had_conflict = true;
2021 nfsd4_cb_recall(dp);
1da177e4
LT		 2022}
2023
2024/*
2025 * The file_lock is being reapd.
2026 *
2027 * Called by locks_free_lock() with lock_kernel() held.
2028 */
2029static
2030void nfsd_release_deleg_cb(struct file_lock *fl)
2031{
2032 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2033
2034 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2035
2036 if (!(fl->fl_flags & FL_LEASE) || !dp)
2037 return;
2038 dp->dl_flock = NULL;
2039}
2040
2041/*
2042 * Set the delegation file_lock back pointer.
2043 *
a9933cea 2044 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2045 */
2046static
2047void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2048{
2049 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2050
2051 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2052 if (!dp)
2053 return;
2054 dp->dl_flock = new;
2055}
2056
2057/*
a9933cea 2058 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2059 */
2060static
2061int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2062{
2063 struct nfs4_delegation *onlistd =
2064 (struct nfs4_delegation *)onlist->fl_owner;
2065 struct nfs4_delegation *tryd =
2066 (struct nfs4_delegation *)try->fl_owner;
2067
2068 if (onlist->fl_lmops != try->fl_lmops)
2069 return 0;
2070
2071 return onlistd->dl_client == tryd->dl_client;
2072}
2073
2074
2075static
2076int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2077{
2078 if (arg & F_UNLCK)
2079 return lease_modify(onlist, arg);
2080 else
2081 return -EAGAIN;
2082}
2083
7b021967 2084static const struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2085 .fl_break = nfsd_break_deleg_cb,
2086 .fl_release_private = nfsd_release_deleg_cb,
2087 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2088 .fl_mylease = nfsd_same_client_deleg_cb,
2089 .fl_change = nfsd_change_deleg_cb,
2090};
2091
2092
b37ad28b 2093__be32
6668958f
AA		 2094nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2095 struct nfsd4_open *open)
1da177e4 2096{
1da177e4
LT		 2097 clientid_t *clientid = &open->op_clientid;
2098 struct nfs4_client *clp = NULL;
2099 unsigned int strhashval;
2100 struct nfs4_stateowner *sop = NULL;
2101
1da177e4 2102 if (!check_name(open->op_owner))
0f442aa2 2103 return nfserr_inval;
1da177e4
LT		 2104
2105 if (STALE_CLIENTID(&open->op_clientid))
2106 return nfserr_stale_clientid;
2107
2108 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2109 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2110 open->op_stateowner = sop;
2111 if (!sop) {
2112 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2113 clp = find_confirmed_client(clientid);
2114 if (clp == NULL)
0f442aa2
BF		 2115 return nfserr_expired;
2116 goto renew;
1da177e4 2117 }
6668958f
AA		 2118 /* When sessions are used, skip open sequenceid processing */
2119 if (nfsd4_has_session(cstate))
2120 goto renew;
0f442aa2
BF		 2121 if (!sop->so_confirmed) {
2122 /* Replace unconfirmed owners without checking for replay. */
2123 clp = sop->so_client;
f044ff83 2124 release_openowner(sop);
0f442aa2
BF		 2125 open->op_stateowner = NULL;
2126 goto renew;
2127 }
2128 if (open->op_seqid == sop->so_seqid - 1) {
2129 if (sop->so_replay.rp_buflen)
a90b061c 2130 return nfserr_replay_me;
0f442aa2
BF		 2131 /* The original OPEN failed so spectacularly
2132 * that we don't even have replay data saved!
2133 * Therefore, we have no choice but to continue
2134 * processing this OPEN; presumably, we'll
2135 * fail again for the same reason.
2136 */
2137 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2138 goto renew;
2139 }
2140 if (open->op_seqid != sop->so_seqid)
2141 return nfserr_bad_seqid;
1da177e4 2142renew:
0f442aa2
BF		 2143 if (open->op_stateowner == NULL) {
2144 sop = alloc_init_open_stateowner(strhashval, clp, open);
2145 if (sop == NULL)
2146 return nfserr_resource;
2147 open->op_stateowner = sop;
2148 }
2149 list_del_init(&sop->so_close_lru);
1da177e4 2150 renew_client(sop->so_client);
0f442aa2 2151 return nfs_ok;
1da177e4
LT		 2152}
2153
b37ad28b 2154static inline __be32
4a6e43e6
N		 2155nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2156{
2157 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2158 return nfserr_openmode;
2159 else
2160 return nfs_ok;
2161}
2162
52f4fb43
N		 2163static struct nfs4_delegation *
2164find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2165{
2166 struct nfs4_delegation *dp;
2167
ea1da636 2168 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2169 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2170 return dp;
2171 }
2172 return NULL;
2173}
2174
b37ad28b 2175static __be32
567d9829
N		 2176nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2177 struct nfs4_delegation **dp)
2178{
2179 int flags;
b37ad28b 2180 __be32 status = nfserr_bad_stateid;
567d9829
N		 2181
2182 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2183 if (*dp == NULL)
c44c5eeb 2184 goto out;
567d9829
N		 2185 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2186 RD_STATE : WR_STATE;
2187 status = nfs4_check_delegmode(*dp, flags);
2188 if (status)
2189 *dp = NULL;
c44c5eeb
N		 2190out:
2191 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2192 return nfs_ok;
2193 if (status)
2194 return status;
2195 open->op_stateowner->so_confirmed = 1;
2196 return nfs_ok;
567d9829
N		 2197}
2198
b37ad28b 2199static __be32
1da177e4
LT		 2200nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2201{
2202 struct nfs4_stateid *local;
b37ad28b 2203 __be32 status = nfserr_share_denied;
1da177e4
LT		 2204 struct nfs4_stateowner *sop = open->op_stateowner;
2205
8beefa24 2206 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2207 /* ignore lock owners */
2208 if (local->st_stateowner->so_is_open_owner == 0)
2209 continue;
2210 /* remember if we have seen this open owner */
2211 if (local->st_stateowner == sop)
2212 *stpp = local;
2213 /* check for conflicting share reservations */
2214 if (!test_share(local, open))
2215 goto out;
2216 }
2217 status = 0;
2218out:
2219 return status;
2220}
2221
5ac049ac
N		 2222static inline struct nfs4_stateid *
2223nfs4_alloc_stateid(void)
2224{
2225 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2226}
2227
b37ad28b 2228static __be32
1da177e4 2229nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2230 struct nfs4_delegation *dp,
1da177e4
LT		 2231 struct svc_fh *cur_fh, int flags)
2232{
2233 struct nfs4_stateid *stp;
1da177e4 2234
5ac049ac 2235 stp = nfs4_alloc_stateid();
1da177e4
LT		 2236 if (stp == NULL)
2237 return nfserr_resource;
2238
567d9829
N		 2239 if (dp) {
2240 get_file(dp->dl_vfs_file);
2241 stp->st_vfs_file = dp->dl_vfs_file;
2242 } else {
b37ad28b 2243 __be32 status;
567d9829
N		 2244 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2245 &stp->st_vfs_file);
2246 if (status) {
2247 if (status == nfserr_dropit)
2248 status = nfserr_jukebox;
5ac049ac 2249 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2250 return status;
2251 }
1da177e4 2252 }
1da177e4
LT		 2253 *stpp = stp;
2254 return 0;
2255}
2256
b37ad28b 2257static inline __be32
1da177e4
LT		 2258nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2259 struct nfsd4_open *open)
2260{
2261 struct iattr iattr = {
2262 .ia_valid = ATTR_SIZE,
2263 .ia_size = 0,
2264 };
2265 if (!open->op_truncate)
2266 return 0;
2267 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2268 return nfserr_inval;
1da177e4
LT		 2269 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2270}
2271
b37ad28b 2272static __be32
1da177e4
LT		 2273nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2274{
2275 struct file *filp = stp->st_vfs_file;
7eaa36e2 2276 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2277 unsigned int share_access, new_writer;
b37ad28b 2278 __be32 status;
1da177e4
LT		 2279
2280 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2281 new_writer = (~share_access) & open->op_share_access
2282 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2283
6c26d08f 2284 if (new_writer) {
b37ad28b
AV		 2285 int err = get_write_access(inode);
2286 if (err)
2287 return nfserrno(err);
e518f056
BH		 2288 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2289 if (err)
2290 return nfserrno(err);
2291 file_take_write(filp);
6c26d08f 2292 }
1da177e4
LT		 2293 status = nfsd4_truncate(rqstp, cur_fh, open);
2294 if (status) {
6c26d08f
BF		 2295 if (new_writer)
2296 put_write_access(inode);
1da177e4
LT		 2297 return status;
2298 }
2299 /* remember the open */
6c26d08f 2300 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2301 __set_bit(open->op_share_access, &stp->st_access_bmap);
2302 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2303
2304 return nfs_ok;
2305}
2306
2307
1da177e4 2308static void
37515177 2309nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2310{
37515177
N		 2311 open->op_stateowner->so_confirmed = 1;
2312 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2313}
2314
2315/*
2316 * Attempt to hand out a delegation.
2317 */
2318static void
2319nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2320{
2321 struct nfs4_delegation *dp;
2322 struct nfs4_stateowner *sop = stp->st_stateowner;
c237dc03 2323 struct nfs4_cb_conn *cb = &sop->so_client->cl_cb_conn;
1da177e4
LT		 2324 struct file_lock fl, *flp = &fl;
2325 int status, flag = 0;
2326
2327 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2328 open->op_recall = 0;
2329 switch (open->op_claim_type) {
2330 case NFS4_OPEN_CLAIM_PREVIOUS:
2331 if (!atomic_read(&cb->cb_set))
2332 open->op_recall = 1;
2333 flag = open->op_delegate_type;
2334 if (flag == NFS4_OPEN_DELEGATE_NONE)
2335 goto out;
2336 break;
2337 case NFS4_OPEN_CLAIM_NULL:
2338 /* Let's not give out any delegations till everyone's
2339 * had the chance to reclaim theirs.... */
af558e33 2340 if (locks_in_grace())
7b190fec
N		 2341 goto out;
2342 if (!atomic_read(&cb->cb_set) || !sop->so_confirmed)
2343 goto out;
2344 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2345 flag = NFS4_OPEN_DELEGATE_WRITE;
2346 else
2347 flag = NFS4_OPEN_DELEGATE_READ;
2348 break;
2349 default:
2350 goto out;
2351 }
1da177e4
LT		 2352
2353 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2354 if (dp == NULL) {
2355 flag = NFS4_OPEN_DELEGATE_NONE;
2356 goto out;
2357 }
2358 locks_init_lock(&fl);
2359 fl.fl_lmops = &nfsd_lease_mng_ops;
2360 fl.fl_flags = FL_LEASE;
9167f501 2361 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2362 fl.fl_end = OFFSET_MAX;
2363 fl.fl_owner = (fl_owner_t)dp;
2364 fl.fl_file = stp->st_vfs_file;
2365 fl.fl_pid = current->tgid;
2366
a9933cea 2367 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2368 * the lock_manager callbacks fl_mylease and fl_change are used
2369 */
9167f501 2370 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2371 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2372 unhash_delegation(dp);
1da177e4
LT		 2373 flag = NFS4_OPEN_DELEGATE_NONE;
2374 goto out;
2375 }
2376
2377 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2378
8c10cbdb
BH		 2379 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2380 STATEID_VAL(&dp->dl_stateid));
1da177e4 2381out:
7b190fec
N		 2382 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2383 && flag == NFS4_OPEN_DELEGATE_NONE
2384 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2385 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2386 open->op_delegate_type = flag;
2387}
2388
2389/*
2390 * called with nfs4_lock_state() held.
2391 */
b37ad28b 2392__be32
1da177e4
LT		 2393nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2394{
6668958f 2395 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2396 struct nfs4_file *fp = NULL;
2397 struct inode *ino = current_fh->fh_dentry->d_inode;
2398 struct nfs4_stateid *stp = NULL;
567d9829 2399 struct nfs4_delegation *dp = NULL;
b37ad28b 2400 __be32 status;
1da177e4
LT		 2401
2402 status = nfserr_inval;
d87a8ade 2403 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2404 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2405 goto out;
2406 /*
2407 * Lookup file; if found, lookup stateid and check open request,
2408 * and check for delegations in the process of being recalled.
2409 * If not found, create the nfs4_file struct
2410 */
2411 fp = find_file(ino);
2412 if (fp) {
2413 if ((status = nfs4_check_open(fp, open, &stp)))
2414 goto out;
c44c5eeb
N		 2415 status = nfs4_check_deleg(fp, open, &dp);
2416 if (status)
2417 goto out;
1da177e4 2418 } else {
c44c5eeb
N		 2419 status = nfserr_bad_stateid;
2420 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2421 goto out;
1da177e4
LT		 2422 status = nfserr_resource;
2423 fp = alloc_init_file(ino);
2424 if (fp == NULL)
2425 goto out;
2426 }
2427
2428 /*
2429 * OPEN the file, or upgrade an existing OPEN.
2430 * If truncate fails, the OPEN fails.
2431 */
2432 if (stp) {
2433 /* Stateid was found, this is an OPEN upgrade */
2434 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2435 if (status)
2436 goto out;
444c2c07 2437 update_stateid(&stp->st_stateid);
1da177e4
LT		 2438 } else {
2439 /* Stateid was not found, this is a new OPEN */
2440 int flags = 0;
9ecb6a08 2441 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2442 flags |= NFSD_MAY_READ;
1da177e4 2443 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2444 flags |= NFSD_MAY_WRITE;
567d9829
N		 2445 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2446 if (status)
1da177e4
LT		 2447 goto out;
2448 init_stateid(stp, fp, open);
2449 status = nfsd4_truncate(rqstp, current_fh, open);
2450 if (status) {
2283963f 2451 release_open_stateid(stp);
1da177e4
LT		 2452 goto out;
2453 }
6668958f
AA		 2454 if (nfsd4_has_session(&resp->cstate))
2455 update_stateid(&stp->st_stateid);
1da177e4
LT		 2456 }
2457 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2458
8b8aae40 2459 if (nfsd4_has_session(&resp->cstate)) {
6668958f 2460 open->op_stateowner->so_confirmed = 1;
8b8aae40
RL		 2461 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2462 }
6668958f 2463
1da177e4
LT		 2464 /*
2465 * Attempt to hand out a delegation. No error return, because the
2466 * OPEN succeeds even if we fail.
2467 */
2468 nfs4_open_delegation(current_fh, open, stp);
2469
2470 status = nfs_ok;
2471
8c10cbdb
BH		 2472 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2473 STATEID_VAL(&stp->st_stateid));
1da177e4 2474out:
13cd2184
N		 2475 if (fp)
2476 put_nfs4_file(fp);
37515177
N		 2477 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2478 nfs4_set_claim_prev(open);
1da177e4
LT		 2479 /*
2480 * To finish the open response, we just need to set the rflags.
2481 */
2482 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2483 if (!open->op_stateowner->so_confirmed &&
2484 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2485 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2486
2487 return status;
2488}
2489
b37ad28b 2490__be32
b591480b
BF		 2491nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2492 clientid_t *clid)
1da177e4
LT		 2493{
2494 struct nfs4_client *clp;
b37ad28b 2495 __be32 status;
1da177e4
LT		 2496
2497 nfs4_lock_state();
2498 dprintk("process_renew(%08x/%08x): starting\n",
2499 clid->cl_boot, clid->cl_id);
2500 status = nfserr_stale_clientid;
2501 if (STALE_CLIENTID(clid))
2502 goto out;
2503 clp = find_confirmed_client(clid);
2504 status = nfserr_expired;
2505 if (clp == NULL) {
2506 /* We assume the client took too long to RENEW. */
2507 dprintk("nfsd4_renew: clientid not found!\n");
2508 goto out;
2509 }
2510 renew_client(clp);
2511 status = nfserr_cb_path_down;
ea1da636 2512 if (!list_empty(&clp->cl_delegations)
c237dc03 2513 && !atomic_read(&clp->cl_cb_conn.cb_set))
1da177e4
LT		 2514 goto out;
2515 status = nfs_ok;
2516out:
2517 nfs4_unlock_state();
2518 return status;
2519}
2520
af558e33
BF		 2521struct lock_manager nfsd4_manager = {
2522};
2523
a76b4319 2524static void
af558e33 2525nfsd4_end_grace(void)
a76b4319
N		 2526{
2527 dprintk("NFSD: end of grace period\n");
c7b9a459 2528 nfsd4_recdir_purge_old();
af558e33 2529 locks_end_grace(&nfsd4_manager);
e46b498c
BF		 2530 /*
2531 * Now that every NFSv4 client has had the chance to recover and
2532 * to see the (possibly new, possibly shorter) lease time, we
2533 * can safely set the next grace time to the current lease time:
2534 */
2535 nfsd4_grace = nfsd4_lease;
a76b4319
N		 2536}
2537
fd39ca9a 2538static time_t
1da177e4
LT		 2539nfs4_laundromat(void)
2540{
2541 struct nfs4_client *clp;
2542 struct nfs4_stateowner *sop;
2543 struct nfs4_delegation *dp;
2544 struct list_head *pos, *next, reaplist;
cf07d2ea
BF		 2545 time_t cutoff = get_seconds() - nfsd4_lease;
2546 time_t t, clientid_val = nfsd4_lease;
2547 time_t u, test_val = nfsd4_lease;
1da177e4
LT		 2548
2549 nfs4_lock_state();
2550
2551 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2552 if (locks_in_grace())
2553 nfsd4_end_grace();
1da177e4
LT		 2554 list_for_each_safe(pos, next, &client_lru) {
2555 clp = list_entry(pos, struct nfs4_client, cl_lru);
2556 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2557 t = clp->cl_time - cutoff;
2558 if (clientid_val > t)
2559 clientid_val = t;
2560 break;
2561 }
2562 dprintk("NFSD: purging unused client (clientid %08x)\n",
2563 clp->cl_clientid.cl_id);
c7b9a459 2564 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2565 expire_client(clp);
2566 }
2567 INIT_LIST_HEAD(&reaplist);
2568 spin_lock(&recall_lock);
2569 list_for_each_safe(pos, next, &del_recall_lru) {
2570 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2571 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2572 u = dp->dl_time - cutoff;
2573 if (test_val > u)
2574 test_val = u;
2575 break;
2576 }
2577 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2578 dp, dp->dl_flock);
2579 list_move(&dp->dl_recall_lru, &reaplist);
2580 }
2581 spin_unlock(&recall_lock);
2582 list_for_each_safe(pos, next, &reaplist) {
2583 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2584 list_del_init(&dp->dl_recall_lru);
2585 unhash_delegation(dp);
2586 }
cf07d2ea 2587 test_val = nfsd4_lease;
1da177e4
LT		 2588 list_for_each_safe(pos, next, &close_lru) {
2589 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2590 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2591 u = sop->so_time - cutoff;
2592 if (test_val > u)
2593 test_val = u;
2594 break;
2595 }
2596 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2597 sop->so_id);
f044ff83 2598 release_openowner(sop);
1da177e4
LT		 2599 }
2600 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2601 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2602 nfs4_unlock_state();
2603 return clientid_val;
2604}
2605
a254b246
HH		 2606static struct workqueue_struct *laundry_wq;
2607static void laundromat_main(struct work_struct *);
2608static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2609
2610static void
c4028958 2611laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2612{
2613 time_t t;
2614
2615 t = nfs4_laundromat();
2616 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2617 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2618}
2619
fd39ca9a 2620static struct nfs4_stateowner *
f8816512
N		 2621search_close_lru(u32 st_id, int flags)
2622{
1da177e4
LT		 2623 struct nfs4_stateowner *local = NULL;
2624
1da177e4
LT		 2625 if (flags & CLOSE_STATE) {
2626 list_for_each_entry(local, &close_lru, so_close_lru) {
2627 if (local->so_id == st_id)
2628 return local;
2629 }
2630 }
2631 return NULL;
2632}
2633
2634static inline int
2635nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2636{
7eaa36e2 2637 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2638}
2639
2640static int
2641STALE_STATEID(stateid_t *stateid)
2642{
78155ed7
BN		 2643 if (time_after((unsigned long)boot_time,
2644 (unsigned long)stateid->si_boot)) {
8c10cbdb
BH		 2645 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2646 STATEID_VAL(stateid));
78155ed7
BN		 2647 return 1;
2648 }
2649 return 0;
2650}
2651
2652static int
2653EXPIRED_STATEID(stateid_t *stateid)
2654{
2655 if (time_before((unsigned long)boot_time,
2656 ((unsigned long)stateid->si_boot)) &&
cf07d2ea 2657 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
8c10cbdb
BH		 2658 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2659 STATEID_VAL(stateid));
78155ed7
BN		 2660 return 1;
2661 }
2662 return 0;
2663}
2664
2665static __be32
2666stateid_error_map(stateid_t *stateid)
2667{
2668 if (STALE_STATEID(stateid))
2669 return nfserr_stale_stateid;
2670 if (EXPIRED_STATEID(stateid))
2671 return nfserr_expired;
2672
8c10cbdb
BH		 2673 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2674 STATEID_VAL(stateid));
78155ed7 2675 return nfserr_bad_stateid;
1da177e4
LT		 2676}
2677
2678static inline int
2679access_permit_read(unsigned long access_bmap)
2680{
2681 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2682 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2683 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2684}
2685
2686static inline int
2687access_permit_write(unsigned long access_bmap)
2688{
2689 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2690 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2691}
2692
2693static
b37ad28b 2694__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2695{
b37ad28b 2696 __be32 status = nfserr_openmode;
1da177e4
LT		 2697
2698 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2699 goto out;
2700 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2701 goto out;
2702 status = nfs_ok;
2703out:
2704 return status;
2705}
2706
b37ad28b 2707static inline __be32
1da177e4
LT		 2708check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2709{
203a8c8e 2710 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2711 return nfs_ok;
af558e33 2712 else if (locks_in_grace()) {
1da177e4
LT		 2713 /* Answer in remaining cases depends on existance of
2714 * conflicting state; so we must wait out the grace period. */
2715 return nfserr_grace;
2716 } else if (flags & WR_STATE)
2717 return nfs4_share_conflict(current_fh,
2718 NFS4_SHARE_DENY_WRITE);
2719 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2720 return nfs4_share_conflict(current_fh,
2721 NFS4_SHARE_DENY_READ);
2722}
2723
2724/*
2725 * Allow READ/WRITE during grace period on recovered state only for files
2726 * that are not able to provide mandatory locking.
2727 */
2728static inline int
18f82731 2729grace_disallows_io(struct inode *inode)
1da177e4 2730{
203a8c8e 2731 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2732}
2733
6668958f 2734static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2735{
6668958f
AA		 2736 /*
2737 * When sessions are used the stateid generation number is ignored
2738 * when it is zero.
2739 */
2740 if ((flags & HAS_SESSION) && in->si_generation == 0)
2741 goto out;
2742
0836f587
BF		 2743 /* If the client sends us a stateid from the future, it's buggy: */
2744 if (in->si_generation > ref->si_generation)
2745 return nfserr_bad_stateid;
2746 /*
2747 * The following, however, can happen. For example, if the
2748 * client sends an open and some IO at the same time, the open
2749 * may bump si_generation while the IO is still in flight.
2750 * Thanks to hard links and renames, the client never knows what
2751 * file an open will affect. So it could avoid that situation
2752 * only by serializing all opens and IO from the same open
2753 * owner. To recover from the old_stateid error, the client
2754 * will just have to retry the IO:
2755 */
2756 if (in->si_generation < ref->si_generation)
2757 return nfserr_old_stateid;
6668958f 2758out:
0836f587
BF		 2759 return nfs_ok;
2760}
2761
3e633079
BF		 2762static int is_delegation_stateid(stateid_t *stateid)
2763{
2764 return stateid->si_fileid == 0;
2765}
2766
1da177e4
LT		 2767/*
2768* Checks for stateid operations
2769*/
b37ad28b 2770__be32
dd453dfd
BH		 2771nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2772 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2773{
2774 struct nfs4_stateid *stp = NULL;
2775 struct nfs4_delegation *dp = NULL;
dd453dfd 2776 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2777 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2778 __be32 status;
1da177e4 2779
1da177e4
LT		 2780 if (filpp)
2781 *filpp = NULL;
2782
18f82731 2783 if (grace_disallows_io(ino))
1da177e4
LT		 2784 return nfserr_grace;
2785
6668958f
AA		 2786 if (nfsd4_has_session(cstate))
2787 flags |= HAS_SESSION;
2788
1da177e4
LT		 2789 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2790 return check_special_stateids(current_fh, stateid, flags);
2791
1da177e4
LT		 2792 status = nfserr_stale_stateid;
2793 if (STALE_STATEID(stateid))
2794 goto out;
2795
1da177e4 2796 status = nfserr_bad_stateid;
3e633079 2797 if (is_delegation_stateid(stateid)) {
a4455be0 2798 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2799 if (!dp) {
2800 status = stateid_error_map(stateid);
1da177e4 2801 goto out;
78155ed7 2802 }
6668958f
AA		 2803 status = check_stateid_generation(stateid, &dp->dl_stateid,
2804 flags);
0c2a498f
BF		 2805 if (status)
2806 goto out;
dc9bf700
BF		 2807 status = nfs4_check_delegmode(dp, flags);
2808 if (status)
2809 goto out;
2810 renew_client(dp->dl_client);
dc9bf700
BF		 2811 if (filpp)
2812 *filpp = dp->dl_vfs_file;
1da177e4 2813 } else { /* open or lock stateid */
a4455be0 2814 stp = find_stateid(stateid, flags);
78155ed7
BN		 2815 if (!stp) {
2816 status = stateid_error_map(stateid);
1da177e4 2817 goto out;
78155ed7 2818 }
6150ef0d 2819 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2820 goto out;
2821 if (!stp->st_stateowner->so_confirmed)
2822 goto out;
6668958f
AA		 2823 status = check_stateid_generation(stateid, &stp->st_stateid,
2824 flags);
0c2a498f
BF		 2825 if (status)
2826 goto out;
a4455be0
BF		 2827 status = nfs4_check_openmode(stp, flags);
2828 if (status)
1da177e4
LT		 2829 goto out;
2830 renew_client(stp->st_stateowner->so_client);
2831 if (filpp)
2832 *filpp = stp->st_vfs_file;
1da177e4
LT		 2833 }
2834 status = nfs_ok;
2835out:
2836 return status;
2837}
2838
4c4cd222
N		 2839static inline int
2840setlkflg (int type)
2841{
2842 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2843 RD_STATE : WR_STATE;
2844}
1da177e4
LT		 2845
2846/*
2847 * Checks for sequence id mutating operations.
2848 */
b37ad28b 2849static __be32
dd453dfd
BH		 2850nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2851 stateid_t *stateid, int flags,
2852 struct nfs4_stateowner **sopp,
2853 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2854{
1da177e4
LT		 2855 struct nfs4_stateid *stp;
2856 struct nfs4_stateowner *sop;
dd453dfd 2857 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2858 __be32 status;
1da177e4 2859
8c10cbdb
BH		 2860 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2861 seqid, STATEID_VAL(stateid));
3a4f98bb 2862
1da177e4
LT		 2863 *stpp = NULL;
2864 *sopp = NULL;
2865
1da177e4 2866 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2867 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2868 return nfserr_bad_stateid;
1da177e4
LT		 2869 }
2870
1da177e4 2871 if (STALE_STATEID(stateid))
3a4f98bb 2872 return nfserr_stale_stateid;
6668958f
AA		 2873
2874 if (nfsd4_has_session(cstate))
2875 flags |= HAS_SESSION;
2876
1da177e4
LT		 2877 /*
2878 * We return BAD_STATEID if filehandle doesn't match stateid,
2879 * the confirmed flag is incorrecly set, or the generation
2880 * number is incorrect.
1da177e4 2881 */
f8816512
N		 2882 stp = find_stateid(stateid, flags);
2883 if (stp == NULL) {
2884 /*
2885 * Also, we should make sure this isn't just the result of
2886 * a replayed close:
2887 */
2888 sop = search_close_lru(stateid->si_stateownerid, flags);
2889 if (sop == NULL)
78155ed7 2890 return stateid_error_map(stateid);
f8816512
N		 2891 *sopp = sop;
2892 goto check_replay;
2893 }
1da177e4 2894
39325bd0
BF		 2895 *stpp = stp;
2896 *sopp = sop = stp->st_stateowner;
2897
4c4cd222 2898 if (lock) {
4c4cd222 2899 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2900 struct nfs4_client *clp = sop->so_client;
4c4cd222 2901 int lkflg = 0;
b37ad28b 2902 __be32 status;
4c4cd222
N		 2903
2904 lkflg = setlkflg(lock->lk_type);
2905
2906 if (lock->lk_is_new) {
599e0a22
BF		 2907 if (!sop->so_is_open_owner)
2908 return nfserr_bad_stateid;
60adfc50
AA		 2909 if (!(flags & HAS_SESSION) &&
2910 !same_clid(&clp->cl_clientid, lockclid))
2911 return nfserr_bad_stateid;
599e0a22
BF		 2912 /* stp is the open stateid */
2913 status = nfs4_check_openmode(stp, lkflg);
2914 if (status)
2915 return status;
2916 } else {
2917 /* stp is the lock stateid */
2918 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2919 if (status)
2920 return status;
4c4cd222 2921 }
1da177e4
LT		 2922 }
2923
f3362737 2924 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2925 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2926 return nfserr_bad_stateid;
1da177e4
LT		 2927 }
2928
1da177e4
LT		 2929 /*
2930 * We now validate the seqid and stateid generation numbers.
2931 * For the moment, we ignore the possibility of
2932 * generation number wraparound.
2933 */
6668958f 2934 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 2935 goto check_replay;
2936
3a4f98bb 2937 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2938 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2939 " unconfirmed stateowner!\n");
2940 return nfserr_bad_stateid;
1da177e4 2941 }
3a4f98bb 2942 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2943 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2944 " confirmed yet!\n");
2945 return nfserr_bad_stateid;
1da177e4 2946 }
6668958f 2947 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 2948 if (status)
2949 return status;
52fd004e 2950 renew_client(sop->so_client);
3a4f98bb 2951 return nfs_ok;
1da177e4 2952
1da177e4 2953check_replay:
bd9aac52 2954 if (seqid == sop->so_seqid - 1) {
849823c5 2955 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2956 /* indicate replay to calling function */
a90b061c 2957 return nfserr_replay_me;
1da177e4 2958 }
2fdada03 2959 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2960 sop->so_seqid, seqid);
2961 *sopp = NULL;
2962 return nfserr_bad_seqid;
1da177e4
LT		 2963}
2964
b37ad28b 2965__be32
ca364317 2966nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2967 struct nfsd4_open_confirm *oc)
1da177e4 2968{
b37ad28b 2969 __be32 status;
1da177e4
LT		 2970 struct nfs4_stateowner *sop;
2971 struct nfs4_stateid *stp;
2972
2973 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 2974 (int)cstate->current_fh.fh_dentry->d_name.len,
2975 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2976
ca364317 2977 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 2978 if (status)
2979 return status;
1da177e4
LT		 2980
2981 nfs4_lock_state();
2982
dd453dfd 2983 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 2984 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 2985 CONFIRM | OPEN_STATE,
1da177e4
LT		 2986 &oc->oc_stateowner, &stp, NULL)))
2987 goto out;
2988
2989 sop = oc->oc_stateowner;
2990 sop->so_confirmed = 1;
2991 update_stateid(&stp->st_stateid);
2992 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
8c10cbdb
BH		 2993 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
2994 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
c7b9a459
N		 2995
2996 nfsd4_create_clid_dir(sop->so_client);
1da177e4 2997out:
f2327d9a 2998 if (oc->oc_stateowner) {
1da177e4 2999 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3000 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3001 }
1da177e4
LT		 3002 nfs4_unlock_state();
3003 return status;
3004}
3005
3006
3007/*
3008 * unset all bits in union bitmap (bmap) that
3009 * do not exist in share (from successful OPEN_DOWNGRADE)
3010 */
3011static void
3012reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3013{
3014 int i;
3015 for (i = 1; i < 4; i++) {
3016 if ((i & access) != i)
3017 __clear_bit(i, bmap);
3018 }
3019}
3020
3021static void
3022reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3023{
3024 int i;
3025 for (i = 0; i < 4; i++) {
3026 if ((i & deny) != i)
3027 __clear_bit(i, bmap);
3028 }
3029}
3030
b37ad28b 3031__be32
ca364317
BF		 3032nfsd4_open_downgrade(struct svc_rqst *rqstp,
3033 struct nfsd4_compound_state *cstate,
a4f1706a 3034 struct nfsd4_open_downgrade *od)
1da177e4 3035{
b37ad28b 3036 __be32 status;
1da177e4
LT		 3037 struct nfs4_stateid *stp;
3038 unsigned int share_access;
3039
3040 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3041 (int)cstate->current_fh.fh_dentry->d_name.len,
3042 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3043
d87a8ade 3044 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3045 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3046 return nfserr_inval;
3047
3048 nfs4_lock_state();
dd453dfd 3049 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3050 od->od_seqid,
1da177e4 3051 &od->od_stateid,
f3362737 3052 OPEN_STATE,
1da177e4
LT		 3053 &od->od_stateowner, &stp, NULL)))
3054 goto out;
3055
3056 status = nfserr_inval;
3057 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3058 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3059 stp->st_access_bmap, od->od_share_access);
3060 goto out;
3061 }
3062 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3063 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3064 stp->st_deny_bmap, od->od_share_deny);
3065 goto out;
3066 }
3067 set_access(&share_access, stp->st_access_bmap);
3068 nfs4_file_downgrade(stp->st_vfs_file,
3069 share_access & ~od->od_share_access);
3070
3071 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3072 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3073
3074 update_stateid(&stp->st_stateid);
3075 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3076 status = nfs_ok;
3077out:
f2327d9a 3078 if (od->od_stateowner) {
1da177e4 3079 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3080 cstate->replay_owner = od->od_stateowner;
f2327d9a 3081 }
1da177e4
LT		 3082 nfs4_unlock_state();
3083 return status;
3084}
3085
3086/*
3087 * nfs4_unlock_state() called after encode
3088 */
b37ad28b 3089__be32
ca364317 3090nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3091 struct nfsd4_close *close)
1da177e4 3092{
b37ad28b 3093 __be32 status;
1da177e4
LT		 3094 struct nfs4_stateid *stp;
3095
3096 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3097 (int)cstate->current_fh.fh_dentry->d_name.len,
3098 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3099
3100 nfs4_lock_state();
3101 /* check close_lru for replay */
dd453dfd 3102 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3103 close->cl_seqid,
1da177e4 3104 &close->cl_stateid,
f3362737 3105 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3106 &close->cl_stateowner, &stp, NULL)))
3107 goto out;
1da177e4
LT		 3108 status = nfs_ok;
3109 update_stateid(&stp->st_stateid);
3110 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3111
04ef5954 3112 /* release_stateid() calls nfsd_close() if needed */
2283963f 3113 release_open_stateid(stp);
04ef5954
BF		 3114
3115 /* place unused nfs4_stateowners on so_close_lru list to be
3116 * released by the laundromat service after the lease period
3117 * to enable us to handle CLOSE replay
3118 */
3119 if (list_empty(&close->cl_stateowner->so_stateids))
3120 move_to_close_lru(close->cl_stateowner);
1da177e4 3121out:
f2327d9a 3122 if (close->cl_stateowner) {
1da177e4 3123 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3124 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3125 }
1da177e4
LT		 3126 nfs4_unlock_state();
3127 return status;
3128}
3129
b37ad28b 3130__be32
ca364317
BF		 3131nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3132 struct nfsd4_delegreturn *dr)
1da177e4 3133{
203a8c8e
BF		 3134 struct nfs4_delegation *dp;
3135 stateid_t *stateid = &dr->dr_stateid;
3136 struct inode *inode;
b37ad28b 3137 __be32 status;
6668958f 3138 int flags = 0;
1da177e4 3139
ca364317 3140 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3141 return status;
3142 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3143
6668958f
AA		 3144 if (nfsd4_has_session(cstate))
3145 flags |= HAS_SESSION;
1da177e4 3146 nfs4_lock_state();
203a8c8e
BF		 3147 status = nfserr_bad_stateid;
3148 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3149 goto out;
3150 status = nfserr_stale_stateid;
3151 if (STALE_STATEID(stateid))
3152 goto out;
7e0f7cf5 3153 status = nfserr_bad_stateid;
203a8c8e
BF		 3154 if (!is_delegation_stateid(stateid))
3155 goto out;
203a8c8e 3156 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3157 if (!dp) {
3158 status = stateid_error_map(stateid);
203a8c8e 3159 goto out;
78155ed7 3160 }
6668958f 3161 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3162 if (status)
3163 goto out;
3164 renew_client(dp->dl_client);
3165
3166 unhash_delegation(dp);
1da177e4 3167out:
203a8c8e
BF		 3168 nfs4_unlock_state();
3169
1da177e4
LT		 3170 return status;
3171}
3172
3173
3174/*
3175 * Lock owner state (byte-range locks)
3176 */
3177#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3178#define LOCK_HASH_BITS 8
3179#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3180#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3181
87df4de8
BH		 3182static inline u64
3183end_offset(u64 start, u64 len)
3184{
3185 u64 end;
3186
3187 end = start + len;
3188 return end >= start ? end: NFS4_MAX_UINT64;
3189}
3190
3191/* last octet in a range */
3192static inline u64
3193last_byte_offset(u64 start, u64 len)
3194{
3195 u64 end;
3196
3197 BUG_ON(!len);
3198 end = start + len;
3199 return end > start ? end - 1: NFS4_MAX_UINT64;
3200}
3201
1da177e4
LT		 3202#define lockownerid_hashval(id) \
3203 ((id) & LOCK_HASH_MASK)
3204
3205static inline unsigned int
3206lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3207 struct xdr_netobj *ownername)
3208{
3209 return (file_hashval(inode) + cl_id
3210 + opaque_hashval(ownername->data, ownername->len))
3211 & LOCK_HASH_MASK;
3212}
3213
3214static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3215static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3216static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3217
fd39ca9a 3218static struct nfs4_stateid *
1da177e4
LT		 3219find_stateid(stateid_t *stid, int flags)
3220{
9346eff0 3221 struct nfs4_stateid *local;
1da177e4
LT		 3222 u32 st_id = stid->si_stateownerid;
3223 u32 f_id = stid->si_fileid;
3224 unsigned int hashval;
3225
3226 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3227 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3228 hashval = stateid_hashval(st_id, f_id);
3229 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3230 if ((local->st_stateid.si_stateownerid == st_id) &&
3231 (local->st_stateid.si_fileid == f_id))
3232 return local;
3233 }
3234 }
9346eff0
KK		 3235
3236 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3237 hashval = stateid_hashval(st_id, f_id);
3238 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3239 if ((local->st_stateid.si_stateownerid == st_id) &&
3240 (local->st_stateid.si_fileid == f_id))
3241 return local;
3242 }
849823c5 3243 }
1da177e4
LT		 3244 return NULL;
3245}
3246
3247static struct nfs4_delegation *
3248find_delegation_stateid(struct inode *ino, stateid_t *stid)
3249{
13cd2184
N		 3250 struct nfs4_file *fp;
3251 struct nfs4_delegation *dl;
1da177e4 3252
8c10cbdb
BH		 3253 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3254 STATEID_VAL(stid));
1da177e4 3255
1da177e4 3256 fp = find_file(ino);
13cd2184
N		 3257 if (!fp)
3258 return NULL;
3259 dl = find_delegation_file(fp, stid);
3260 put_nfs4_file(fp);
3261 return dl;
1da177e4
LT		 3262}
3263
3264/*
3265 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3266 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3267 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3268 * locking, this prevents us from being completely protocol-compliant. The
3269 * real solution to this problem is to start using unsigned file offsets in
3270 * the VFS, but this is a very deep change!
3271 */
3272static inline void
3273nfs4_transform_lock_offset(struct file_lock *lock)
3274{
3275 if (lock->fl_start < 0)
3276 lock->fl_start = OFFSET_MAX;
3277 if (lock->fl_end < 0)
3278 lock->fl_end = OFFSET_MAX;
3279}
3280
d5b9026a
N		 3281/* Hack!: For now, we're defining this just so we can use a pointer to it
3282 * as a unique cookie to identify our (NFSv4's) posix locks. */
7b021967 3283static const struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3284};
1da177e4
LT		 3285
3286static inline void
3287nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3288{
d5b9026a
N		 3289 struct nfs4_stateowner *sop;
3290 unsigned int hval;
1da177e4 3291
d5b9026a
N		 3292 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3293 sop = (struct nfs4_stateowner *) fl->fl_owner;
3294 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3295 kref_get(&sop->so_ref);
3296 deny->ld_sop = sop;
3297 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3298 } else {
3299 deny->ld_sop = NULL;
3300 deny->ld_clientid.cl_boot = 0;
3301 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3302 }
3303 deny->ld_start = fl->fl_start;
87df4de8
BH		 3304 deny->ld_length = NFS4_MAX_UINT64;
3305 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3306 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3307 deny->ld_type = NFS4_READ_LT;
3308 if (fl->fl_type != F_RDLCK)
3309 deny->ld_type = NFS4_WRITE_LT;
3310}
3311
1da177e4
LT		 3312static struct nfs4_stateowner *
3313find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3314 struct xdr_netobj *owner)
3315{
3316 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3317 struct nfs4_stateowner *op;
3318
3319 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3320 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3321 return op;
3322 }
3323 return NULL;
3324}
3325
3326/*
3327 * Alloc a lock owner structure.
3328 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3329 * occured.
3330 *
3331 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3332 */
3333
3334static struct nfs4_stateowner *
3335alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3336 struct nfs4_stateowner *sop;
3337 struct nfs4_replay *rp;
3338 unsigned int idhashval;
3339
3340 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3341 return NULL;
3342 idhashval = lockownerid_hashval(current_ownerid);
3343 INIT_LIST_HEAD(&sop->so_idhash);
3344 INIT_LIST_HEAD(&sop->so_strhash);
3345 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3346 INIT_LIST_HEAD(&sop->so_stateids);
3347 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3348 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3349 sop->so_time = 0;
3350 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3351 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3352 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3353 sop->so_is_open_owner = 0;
3354 sop->so_id = current_ownerid++;
3355 sop->so_client = clp;
b59e3c0e
NB		 3356 /* It is the openowner seqid that will be incremented in encode in the
3357 * case of new lockowners; so increment the lock seqid manually: */
3358 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3359 sop->so_confirmed = 1;
3360 rp = &sop->so_replay;
de1ae286 3361 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3362 rp->rp_buflen = 0;
3363 rp->rp_buf = rp->rp_ibuf;
3364 return sop;
3365}
3366
fd39ca9a 3367static struct nfs4_stateid *
1da177e4
LT		 3368alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3369{
3370 struct nfs4_stateid *stp;
3371 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3372
5ac049ac
N		 3373 stp = nfs4_alloc_stateid();
3374 if (stp == NULL)
1da177e4
LT		 3375 goto out;
3376 INIT_LIST_HEAD(&stp->st_hash);
3377 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3378 INIT_LIST_HEAD(&stp->st_perstateowner);
3379 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3380 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3381 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3382 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3383 stp->st_stateowner = sop;
13cd2184 3384 get_nfs4_file(fp);
1da177e4 3385 stp->st_file = fp;
78155ed7 3386 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3387 stp->st_stateid.si_stateownerid = sop->so_id;
3388 stp->st_stateid.si_fileid = fp->fi_id;
3389 stp->st_stateid.si_generation = 0;
3390 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3391 stp->st_access_bmap = open_stp->st_access_bmap;
3392 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3393 stp->st_openstp = open_stp;
1da177e4
LT		 3394
3395out:
3396 return stp;
3397}
3398
fd39ca9a 3399static int
1da177e4
LT		 3400check_lock_length(u64 offset, u64 length)
3401{
87df4de8 3402 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3403 LOFF_OVERFLOW(offset, length)));
3404}
3405
3406/*
3407 * LOCK operation
3408 */
b37ad28b 3409__be32
ca364317 3410nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3411 struct nfsd4_lock *lock)
1da177e4 3412{
3e9e3dbe 3413 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3414 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3415 struct nfs4_stateid *lock_stp;
3416 struct file *filp;
3417 struct file_lock file_lock;
8dc7c311 3418 struct file_lock conflock;
b37ad28b 3419 __be32 status = 0;
1da177e4 3420 unsigned int strhashval;
150b3934 3421 unsigned int cmd;
b8dd7b9a 3422 int err;
1da177e4
LT		 3423
3424 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3425 (long long) lock->lk_offset,
3426 (long long) lock->lk_length);
3427
1da177e4
LT		 3428 if (check_lock_length(lock->lk_offset, lock->lk_length))
3429 return nfserr_inval;
3430
ca364317 3431 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3432 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3433 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3434 return status;
3435 }
3436
1da177e4
LT		 3437 nfs4_lock_state();
3438
3439 if (lock->lk_is_new) {
893f8770
N		 3440 /*
3441 * Client indicates that this is a new lockowner.
3442 * Use open owner and open stateid to create lock owner and
3443 * lock stateid.
3444 */
1da177e4
LT		 3445 struct nfs4_stateid *open_stp = NULL;
3446 struct nfs4_file *fp;
3447
3448 status = nfserr_stale_clientid;
60adfc50
AA		 3449 if (!nfsd4_has_session(cstate) &&
3450 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3451 goto out;
1da177e4 3452
1da177e4 3453 /* validate and update open stateid and open seqid */
dd453dfd 3454 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3455 lock->lk_new_open_seqid,
3456 &lock->lk_new_open_stateid,
f3362737 3457 OPEN_STATE,
3a65588a 3458 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3459 lock);
37515177 3460 if (status)
1da177e4 3461 goto out;
3a65588a 3462 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3463 /* create lockowner and lock stateid */
3464 fp = open_stp->st_file;
3465 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3466 open_sop->so_client->cl_clientid.cl_id,
3467 &lock->v.new.owner);
3e9e3dbe
N		 3468 /* XXX: Do we need to check for duplicate stateowners on
3469 * the same file, or should they just be allowed (and
3470 * create new stateids)? */
1da177e4 3471 status = nfserr_resource;
b59e3c0e
NB		 3472 lock_sop = alloc_init_lock_stateowner(strhashval,
3473 open_sop->so_client, open_stp, lock);
3474 if (lock_sop == NULL)
1da177e4 3475 goto out;
b59e3c0e 3476 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3477 if (lock_stp == NULL)
1da177e4 3478 goto out;
1da177e4
LT		 3479 } else {
3480 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3481 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3482 lock->lk_old_lock_seqid,
3483 &lock->lk_old_lock_stateid,
f3362737 3484 LOCK_STATE,
3a65588a 3485 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3486 if (status)
3487 goto out;
3a65588a 3488 lock_sop = lock->lk_replay_owner;
1da177e4 3489 }
3a65588a 3490 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3491 filp = lock_stp->st_vfs_file;
3492
0dd395dc 3493 status = nfserr_grace;
af558e33 3494 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3495 goto out;
3496 status = nfserr_no_grace;
af558e33 3497 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3498 goto out;
3499
1da177e4
LT		 3500 locks_init_lock(&file_lock);
3501 switch (lock->lk_type) {
3502 case NFS4_READ_LT:
3503 case NFS4_READW_LT:
3504 file_lock.fl_type = F_RDLCK;
150b3934 3505 cmd = F_SETLK;
1da177e4
LT		 3506 break;
3507 case NFS4_WRITE_LT:
3508 case NFS4_WRITEW_LT:
3509 file_lock.fl_type = F_WRLCK;
150b3934 3510 cmd = F_SETLK;
1da177e4
LT		 3511 break;
3512 default:
3513 status = nfserr_inval;
3514 goto out;
3515 }
b59e3c0e 3516 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3517 file_lock.fl_pid = current->tgid;
3518 file_lock.fl_file = filp;
3519 file_lock.fl_flags = FL_POSIX;
d5b9026a 3520 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3521
3522 file_lock.fl_start = lock->lk_offset;
87df4de8 3523 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3524 nfs4_transform_lock_offset(&file_lock);
3525
3526 /*
3527 * Try to lock the file in the VFS.
3528 * Note: locks.c uses the BKL to protect the inode's lock list.
3529 */
3530
fd85b817 3531 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3532 switch (-err) {
1da177e4
LT		 3533 case 0: /* success! */
3534 update_stateid(&lock_stp->st_stateid);
3535 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3536 sizeof(stateid_t));
b8dd7b9a 3537 status = 0;
eb76b3fd
AA		 3538 break;
3539 case (EAGAIN): /* conflock holds conflicting lock */
3540 status = nfserr_denied;
3541 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3542 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3543 break;
1da177e4
LT		 3544 case (EDEADLK):
3545 status = nfserr_deadlock;
eb76b3fd 3546 break;
1da177e4 3547 default:
fd85b817 3548 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3549 status = nfserr_resource;
3550 break;
1da177e4 3551 }
1da177e4 3552out:
8a280510 3553 if (status && lock->lk_is_new && lock_sop)
f044ff83 3554 release_lockowner(lock_sop);
3a65588a
BF		 3555 if (lock->lk_replay_owner) {
3556 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3557 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3558 }
1da177e4
LT		 3559 nfs4_unlock_state();
3560 return status;
3561}
3562
55ef1274
BF		 3563/*
3564 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3565 * so we do a temporary open here just to get an open file to pass to
3566 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3567 * inode operation.)
3568 */
3569static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3570{
3571 struct file *file;
3572 int err;
3573
3574 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3575 if (err)
3576 return err;
3577 err = vfs_test_lock(file, lock);
3578 nfsd_close(file);
3579 return err;
3580}
3581
1da177e4
LT		 3582/*
3583 * LOCKT operation
3584 */
b37ad28b 3585__be32
ca364317
BF		 3586nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3587 struct nfsd4_lockt *lockt)
1da177e4
LT		 3588{
3589 struct inode *inode;
1da177e4 3590 struct file_lock file_lock;
fd85b817 3591 int error;
b37ad28b 3592 __be32 status;
1da177e4 3593
af558e33 3594 if (locks_in_grace())
1da177e4
LT		 3595 return nfserr_grace;
3596
3597 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3598 return nfserr_inval;
3599
3600 lockt->lt_stateowner = NULL;
3601 nfs4_lock_state();
3602
3603 status = nfserr_stale_clientid;
60adfc50 3604 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3605 goto out;
1da177e4 3606
ca364317 3607 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3608 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3609 if (status == nfserr_symlink)
3610 status = nfserr_inval;
3611 goto out;
3612 }
3613
ca364317 3614 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3615 locks_init_lock(&file_lock);
3616 switch (lockt->lt_type) {
3617 case NFS4_READ_LT:
3618 case NFS4_READW_LT:
3619 file_lock.fl_type = F_RDLCK;
3620 break;
3621 case NFS4_WRITE_LT:
3622 case NFS4_WRITEW_LT:
3623 file_lock.fl_type = F_WRLCK;
3624 break;
3625 default:
2fdada03 3626 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3627 status = nfserr_inval;
3628 goto out;
3629 }
3630
3631 lockt->lt_stateowner = find_lockstateowner_str(inode,
3632 &lockt->lt_clientid, &lockt->lt_owner);
3633 if (lockt->lt_stateowner)
3634 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3635 file_lock.fl_pid = current->tgid;
3636 file_lock.fl_flags = FL_POSIX;
3637
3638 file_lock.fl_start = lockt->lt_offset;
87df4de8 3639 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3640
3641 nfs4_transform_lock_offset(&file_lock);
3642
1da177e4 3643 status = nfs_ok;
55ef1274 3644 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3645 if (error) {
3646 status = nfserrno(error);
3647 goto out;
3648 }
9d6a8c5c 3649 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3650 status = nfserr_denied;
9d6a8c5c 3651 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3652 }
3653out:
3654 nfs4_unlock_state();
3655 return status;
3656}
3657
b37ad28b 3658__be32
ca364317 3659nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3660 struct nfsd4_locku *locku)
1da177e4
LT		 3661{
3662 struct nfs4_stateid *stp;
3663 struct file *filp = NULL;
3664 struct file_lock file_lock;
b37ad28b 3665 __be32 status;
b8dd7b9a 3666 int err;
1da177e4
LT		 3667
3668 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3669 (long long) locku->lu_offset,
3670 (long long) locku->lu_length);
3671
3672 if (check_lock_length(locku->lu_offset, locku->lu_length))
3673 return nfserr_inval;
3674
3675 nfs4_lock_state();
3676
dd453dfd 3677 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3678 locku->lu_seqid,
3679 &locku->lu_stateid,
f3362737 3680 LOCK_STATE,
1da177e4
LT		 3681 &locku->lu_stateowner, &stp, NULL)))
3682 goto out;
3683
3684 filp = stp->st_vfs_file;
3685 BUG_ON(!filp);
3686 locks_init_lock(&file_lock);
3687 file_lock.fl_type = F_UNLCK;
3688 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3689 file_lock.fl_pid = current->tgid;
3690 file_lock.fl_file = filp;
3691 file_lock.fl_flags = FL_POSIX;
d5b9026a 3692 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3693 file_lock.fl_start = locku->lu_offset;
3694
87df4de8 3695 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3696 nfs4_transform_lock_offset(&file_lock);
3697
3698 /*
3699 * Try to unlock the file in the VFS.
3700 */
fd85b817 3701 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3702 if (err) {
fd85b817 3703 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3704 goto out_nfserr;
3705 }
3706 /*
3707 * OK, unlock succeeded; the only thing left to do is update the stateid.
3708 */
3709 update_stateid(&stp->st_stateid);
3710 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3711
3712out:
f2327d9a 3713 if (locku->lu_stateowner) {
1da177e4 3714 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3715 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3716 }
1da177e4
LT		 3717 nfs4_unlock_state();
3718 return status;
3719
3720out_nfserr:
b8dd7b9a 3721 status = nfserrno(err);
1da177e4
LT		 3722 goto out;
3723}
3724
3725/*
3726 * returns
3727 * 1: locks held by lockowner
3728 * 0: no locks held by lockowner
3729 */
3730static int
3731check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3732{
3733 struct file_lock **flpp;
7eaa36e2 3734 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3735 int status = 0;
3736
3737 lock_kernel();
3738 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3739 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3740 status = 1;
3741 goto out;
796dadfd 3742 }
1da177e4
LT		 3743 }
3744out:
3745 unlock_kernel();
3746 return status;
3747}
3748
b37ad28b 3749__be32
b591480b
BF		 3750nfsd4_release_lockowner(struct svc_rqst *rqstp,
3751 struct nfsd4_compound_state *cstate,
3752 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3753{
3754 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3755 struct nfs4_stateowner *sop;
3756 struct nfs4_stateid *stp;
1da177e4 3757 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3758 struct list_head matches;
3759 int i;
b37ad28b 3760 __be32 status;
1da177e4
LT		 3761
3762 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3763 clid->cl_boot, clid->cl_id);
3764
3765 /* XXX check for lease expiration */
3766
3767 status = nfserr_stale_clientid;
849823c5 3768 if (STALE_CLIENTID(clid))
1da177e4 3769 return status;
1da177e4
LT		 3770
3771 nfs4_lock_state();
3772
3e9e3dbe
N		 3773 status = nfserr_locks_held;
3774 /* XXX: we're doing a linear search through all the lockowners.
3775 * Yipes! For now we'll just hope clients aren't really using
3776 * release_lockowner much, but eventually we have to fix these
3777 * data structures. */
3778 INIT_LIST_HEAD(&matches);
3779 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3780 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3781 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3782 continue;
3783 list_for_each_entry(stp, &sop->so_stateids,
3784 st_perstateowner) {
3785 if (check_for_locks(stp->st_vfs_file, sop))
3786 goto out;
3787 /* Note: so_perclient unused for lockowners,
3788 * so it's OK to fool with here. */
3789 list_add(&sop->so_perclient, &matches);
3790 }
1da177e4 3791 }
3e9e3dbe
N		 3792 }
3793 /* Clients probably won't expect us to return with some (but not all)
3794 * of the lockowner state released; so don't release any until all
3795 * have been checked. */
3796 status = nfs_ok;
0fa822e4
N		 3797 while (!list_empty(&matches)) {
3798 sop = list_entry(matches.next, struct nfs4_stateowner,
3799 so_perclient);
3800 /* unhash_stateowner deletes so_perclient only
3801 * for openowners. */
3802 list_del(&sop->so_perclient);
f044ff83 3803 release_lockowner(sop);
1da177e4
LT		 3804 }
3805out:
3806 nfs4_unlock_state();
3807 return status;
3808}
3809
3810static inline struct nfs4_client_reclaim *
a55370a3 3811alloc_reclaim(void)
1da177e4 3812{
a55370a3 3813 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3814}
3815
c7b9a459 3816int
a1bcecd2 3817nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3818{
3819 unsigned int strhashval = clientstr_hashval(name);
3820 struct nfs4_client *clp;
3821
a1bcecd2 3822 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3823 return clp ? 1 : 0;
3824}
3825
1da177e4
LT		 3826/*
3827 * failure => all reset bets are off, nfserr_no_grace...
3828 */
190e4fbf
N		 3829int
3830nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3831{
3832 unsigned int strhashval;
3833 struct nfs4_client_reclaim *crp = NULL;
3834
a55370a3
N		 3835 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3836 crp = alloc_reclaim();
1da177e4
LT		 3837 if (!crp)
3838 return 0;
a55370a3 3839 strhashval = clientstr_hashval(name);
1da177e4
LT		 3840 INIT_LIST_HEAD(&crp->cr_strhash);
3841 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3842 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3843 reclaim_str_hashtbl_size++;
3844 return 1;
3845}
3846
3847static void
3848nfs4_release_reclaim(void)
3849{
3850 struct nfs4_client_reclaim *crp = NULL;
3851 int i;
3852
1da177e4
LT		 3853 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3854 while (!list_empty(&reclaim_str_hashtbl[i])) {
3855 crp = list_entry(reclaim_str_hashtbl[i].next,
3856 struct nfs4_client_reclaim, cr_strhash);
3857 list_del(&crp->cr_strhash);
1da177e4
LT		 3858 kfree(crp);
3859 reclaim_str_hashtbl_size--;
3860 }
3861 }
3862 BUG_ON(reclaim_str_hashtbl_size);
3863}
3864
3865/*
3866 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3867static struct nfs4_client_reclaim *
1da177e4
LT		 3868nfs4_find_reclaim_client(clientid_t *clid)
3869{
3870 unsigned int strhashval;
3871 struct nfs4_client *clp;
3872 struct nfs4_client_reclaim *crp = NULL;
3873
3874
3875 /* find clientid in conf_id_hashtbl */
3876 clp = find_confirmed_client(clid);
3877 if (clp == NULL)
3878 return NULL;
3879
a55370a3
N		 3880 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3881 clp->cl_name.len, clp->cl_name.data,
3882 clp->cl_recdir);
1da177e4
LT		 3883
3884 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3885 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3886 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3887 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3888 return crp;
3889 }
3890 }
3891 return NULL;
3892}
3893
3894/*
3895* Called from OPEN. Look for clientid in reclaim list.
3896*/
b37ad28b 3897__be32
1da177e4
LT		 3898nfs4_check_open_reclaim(clientid_t *clid)
3899{
dfc83565 3900 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3901}
3902
ac4d8ff2 3903/* initialization to perform at module load time: */
1da177e4 3904
e8ff2a84 3905int
ac4d8ff2 3906nfs4_state_init(void)
1da177e4 3907{
e8ff2a84 3908 int i, status;
1da177e4 3909
e8ff2a84
BF		 3910 status = nfsd4_init_slabs();
3911 if (status)
3912 return status;
1da177e4
LT		 3913 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3914 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3915 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3916 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3917 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3918 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3919 }
5282fd72
ME		 3920 for (i = 0; i < SESSION_HASH_SIZE; i++)
3921 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3922 for (i = 0; i < FILE_HASH_SIZE; i++) {
3923 INIT_LIST_HEAD(&file_hashtbl[i]);
3924 }
3925 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3926 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3927 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3928 }
3929 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3930 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3931 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3932 }
3933 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3934 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3935 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3936 }
1da177e4 3937 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3938 INIT_LIST_HEAD(&close_lru);
3939 INIT_LIST_HEAD(&client_lru);
3940 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 3941 reclaim_str_hashtbl_size = 0;
e8ff2a84 3942 return 0;
ac4d8ff2
N		 3943}
3944
190e4fbf
N		 3945static void
3946nfsd4_load_reboot_recovery_data(void)
3947{
3948 int status;
3949
0964a3d3
N		 3950 nfs4_lock_state();
3951 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3952 status = nfsd4_recdir_load();
0964a3d3 3953 nfs4_unlock_state();
190e4fbf
N		 3954 if (status)
3955 printk("NFSD: Failure reading reboot recovery data\n");
3956}
3957
c2f1a551
MS		 3958/*
3959 * Since the lifetime of a delegation isn't limited to that of an open, a
3960 * client may quite reasonably hang on to a delegation as long as it has
3961 * the inode cached. This becomes an obvious problem the first time a
3962 * client's inode cache approaches the size of the server's total memory.
3963 *
3964 * For now we avoid this problem by imposing a hard limit on the number
3965 * of delegations, which varies according to the server's memory size.
3966 */
3967static void
3968set_max_delegations(void)
3969{
3970 /*
3971 * Allow at most 4 delegations per megabyte of RAM. Quick
3972 * estimates suggest that in the worst case (where every delegation
3973 * is for a different inode), a delegation could take about 1.5K,
3974 * giving a worst case usage of about 6% of memory.
3975 */
3976 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3977}
3978
ac4d8ff2
N		 3979/* initialization to perform when the nfsd service is started: */
3980
29ab23cc 3981static int
ac4d8ff2
N		 3982__nfs4_state_start(void)
3983{
b5a1a81e
BF		 3984 int ret;
3985
1da177e4 3986 boot_time = get_seconds();
af558e33 3987 locks_start_grace(&nfsd4_manager);
9a8db97e 3988 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
e46b498c 3989 nfsd4_grace);
b5a1a81e
BF		 3990 ret = set_callback_cred();
3991 if (ret)
3992 return -ENOMEM;
58da282b 3993 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF		 3994 if (laundry_wq == NULL)
3995 return -ENOMEM;
b5a1a81e
BF		 3996 ret = nfsd4_create_callback_queue();
3997 if (ret)
3998 goto out_free_laundry;
e46b498c 3999 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
c2f1a551 4000 set_max_delegations();
b5a1a81e
BF		 4001 return 0;
4002out_free_laundry:
4003 destroy_workqueue(laundry_wq);
4004 return ret;
1da177e4
LT		 4005}
4006
29ab23cc 4007int
76a3550e 4008nfs4_state_start(void)
1da177e4 4009{
29ab23cc
BF		 4010 int ret;
4011
1da177e4 4012 if (nfs4_init)
29ab23cc 4013 return 0;
190e4fbf 4014 nfsd4_load_reboot_recovery_data();
29ab23cc
BF		 4015 ret = __nfs4_state_start();
4016 if (ret)
4017 return ret;
1da177e4 4018 nfs4_init = 1;
29ab23cc 4019 return 0;
1da177e4
LT		 4020}
4021
1da177e4
LT		 4022static void
4023__nfs4_state_shutdown(void)
4024{
4025 int i;
4026 struct nfs4_client *clp = NULL;
4027 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4028 struct list_head *pos, *next, reaplist;
4029
1da177e4
LT		 4030 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4031 while (!list_empty(&conf_id_hashtbl[i])) {
4032 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4033 expire_client(clp);
4034 }
4035 while (!list_empty(&unconf_str_hashtbl[i])) {
4036 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4037 expire_client(clp);
4038 }
4039 }
4040 INIT_LIST_HEAD(&reaplist);
4041 spin_lock(&recall_lock);
4042 list_for_each_safe(pos, next, &del_recall_lru) {
4043 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4044 list_move(&dp->dl_recall_lru, &reaplist);
4045 }
4046 spin_unlock(&recall_lock);
4047 list_for_each_safe(pos, next, &reaplist) {
4048 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4049 list_del_init(&dp->dl_recall_lru);
4050 unhash_delegation(dp);
4051 }
4052
190e4fbf 4053 nfsd4_shutdown_recdir();
1da177e4 4054 nfs4_init = 0;
1da177e4
LT		 4055}
4056
4057void
4058nfs4_state_shutdown(void)
4059{
5e8d5c29
N		 4060 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4061 destroy_workqueue(laundry_wq);
2c5e7615 4062 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4063 nfs4_lock_state();
4064 nfs4_release_reclaim();
4065 __nfs4_state_shutdown();
b5a1a81e 4066 nfsd4_destroy_callback_queue();
1da177e4
LT		 4067 nfs4_unlock_state();
4068}
4069
3dd98a3b
JL		 4070/*
4071 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4072 * accessed when nfsd is starting.
4073 */
0964a3d3
N		 4074static void
4075nfs4_set_recdir(char *recdir)
4076{
0964a3d3 4077 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4078}
4079
4080/*
4081 * Change the NFSv4 recovery directory to recdir.
4082 */
4083int
4084nfs4_reset_recoverydir(char *recdir)
4085{
4086 int status;
a63bb996 4087 struct path path;
0964a3d3 4088
a63bb996 4089 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4090 if (status)
4091 return status;
4092 status = -ENOTDIR;
a63bb996 4093 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4094 nfs4_set_recdir(recdir);
4095 status = 0;
4096 }
a63bb996 4097 path_put(&path);
0964a3d3
N		 4098 return status;
4099}
4100
3dd98a3b
JL		 4101char *
4102nfs4_recoverydir(void)
4103{
4104 return user_recovery_dirname;
4105}
Linux 6.x block layer and io_uring tree(s)