projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4: fix bare destroy_session null dereference
[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>
5a0e3ad6 37#include <linux/slab.h>
0964a3d3 38#include <linux/namei.h>
c2f1a551 39#include <linux/swap.h>
68e76ad0 40#include <linux/sunrpc/svcauth_gss.h>
363168b4 41#include <linux/sunrpc/clnt.h>
9a74af21 42#include "xdr4.h"
0a3adade 43#include "vfs.h"
1da177e4
LT		 44
45#define NFSDDBG_FACILITY NFSDDBG_PROC
46
47/* Globals */
cf07d2ea 48time_t nfsd4_lease = 90; /* default lease time */
efc4bb4f 49time_t nfsd4_grace = 90;
fd39ca9a 50static time_t boot_time;
1da177e4
LT		 51static u32 current_ownerid = 1;
52static u32 current_fileid = 1;
53static u32 current_delegid = 1;
54static u32 nfs4_init;
fd39ca9a
N		 55static stateid_t zerostateid; /* bits all 0 */
56static stateid_t onestateid; /* bits all 1 */
ec6b5d7b 57static u64 current_sessionid = 1;
fd39ca9a
N		 58
59#define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
60#define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
1da177e4 61
1da177e4 62/* forward declarations */
fd39ca9a 63static struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
1da177e4 64static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
0964a3d3
N		 65static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
66static void nfs4_set_recdir(char *recdir);
1da177e4 67
8b671b80
BF		 68/* Locking: */
69
70/* Currently used for almost all code touching nfsv4 state: */
353ab6e9 71static DEFINE_MUTEX(client_mutex);
1da177e4 72
8b671b80
BF		 73/*
74 * Currently used for the del_recall_lru and file hash table. In an
75 * effort to decrease the scope of the client_mutex, this spinlock may
76 * eventually cover more:
77 */
78static DEFINE_SPINLOCK(recall_lock);
79
e18b890b
CL		 80static struct kmem_cache *stateowner_slab = NULL;
81static struct kmem_cache *file_slab = NULL;
82static struct kmem_cache *stateid_slab = NULL;
83static struct kmem_cache *deleg_slab = NULL;
e60d4398 84
1da177e4
LT		 85void
86nfs4_lock_state(void)
87{
353ab6e9 88 mutex_lock(&client_mutex);
1da177e4
LT		 89}
90
91void
92nfs4_unlock_state(void)
93{
353ab6e9 94 mutex_unlock(&client_mutex);
1da177e4
LT		 95}
96
97static inline u32
98opaque_hashval(const void *ptr, int nbytes)
99{
100 unsigned char *cptr = (unsigned char *) ptr;
101
102 u32 x = 0;
103 while (nbytes--) {
104 x *= 37;
105 x += *cptr++;
106 }
107 return x;
108}
109
1da177e4
LT		 110static struct list_head del_recall_lru;
111
13cd2184
N		 112static inline void
113put_nfs4_file(struct nfs4_file *fi)
114{
8b671b80
BF		 115 if (atomic_dec_and_lock(&fi->fi_ref, &recall_lock)) {
116 list_del(&fi->fi_hash);
117 spin_unlock(&recall_lock);
118 iput(fi->fi_inode);
119 kmem_cache_free(file_slab, fi);
120 }
13cd2184
N		 121}
122
123static inline void
124get_nfs4_file(struct nfs4_file *fi)
125{
8b671b80 126 atomic_inc(&fi->fi_ref);
13cd2184
N		 127}
128
ef0f3390 129static int num_delegations;
c2f1a551 130unsigned int max_delegations;
ef0f3390
N		 131
132/*
133 * Open owner state (share locks)
134 */
135
136/* hash tables for nfs4_stateowner */
137#define OWNER_HASH_BITS 8
138#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
139#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
140
141#define ownerid_hashval(id) \
142 ((id) & OWNER_HASH_MASK)
143#define ownerstr_hashval(clientid, ownername) \
144 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
145
146static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
147static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
148
149/* hash table for nfs4_file */
150#define FILE_HASH_BITS 8
151#define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
152#define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
153/* hash table for (open)nfs4_stateid */
154#define STATEID_HASH_BITS 10
155#define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
156#define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
157
158#define file_hashval(x) \
159 hash_ptr(x, FILE_HASH_BITS)
160#define stateid_hashval(owner_id, file_id) \
161 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
162
163static struct list_head file_hashtbl[FILE_HASH_SIZE];
164static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
165
1da177e4
LT		 166static struct nfs4_delegation *
167alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
168{
169 struct nfs4_delegation *dp;
170 struct nfs4_file *fp = stp->st_file;
c237dc03 171 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
1da177e4
LT		 172
173 dprintk("NFSD alloc_init_deleg\n");
47f9940c
MS		 174 if (fp->fi_had_conflict)
175 return NULL;
c2f1a551 176 if (num_delegations > max_delegations)
ef0f3390 177 return NULL;
5b2d21c1
N		 178 dp = kmem_cache_alloc(deleg_slab, GFP_KERNEL);
179 if (dp == NULL)
1da177e4 180 return dp;
ef0f3390 181 num_delegations++;
ea1da636
N		 182 INIT_LIST_HEAD(&dp->dl_perfile);
183 INIT_LIST_HEAD(&dp->dl_perclnt);
1da177e4
LT		 184 INIT_LIST_HEAD(&dp->dl_recall_lru);
185 dp->dl_client = clp;
13cd2184 186 get_nfs4_file(fp);
1da177e4
LT		 187 dp->dl_file = fp;
188 dp->dl_flock = NULL;
189 get_file(stp->st_vfs_file);
190 dp->dl_vfs_file = stp->st_vfs_file;
191 dp->dl_type = type;
b53d40c5 192 dp->dl_ident = cb->cb_ident;
78155ed7 193 dp->dl_stateid.si_boot = get_seconds();
1da177e4
LT		 194 dp->dl_stateid.si_stateownerid = current_delegid++;
195 dp->dl_stateid.si_fileid = 0;
196 dp->dl_stateid.si_generation = 0;
6c02eaa1 197 fh_copy_shallow(&dp->dl_fh, &current_fh->fh_handle);
1da177e4
LT		 198 dp->dl_time = 0;
199 atomic_set(&dp->dl_count, 1);
ea1da636
N		 200 list_add(&dp->dl_perfile, &fp->fi_delegations);
201 list_add(&dp->dl_perclnt, &clp->cl_delegations);
b5a1a81e 202 INIT_WORK(&dp->dl_recall.cb_work, nfsd4_do_callback_rpc);
1da177e4
LT		 203 return dp;
204}
205
206void
207nfs4_put_delegation(struct nfs4_delegation *dp)
208{
209 if (atomic_dec_and_test(&dp->dl_count)) {
210 dprintk("NFSD: freeing dp %p\n",dp);
13cd2184 211 put_nfs4_file(dp->dl_file);
5b2d21c1 212 kmem_cache_free(deleg_slab, dp);
ef0f3390 213 num_delegations--;
1da177e4
LT		 214 }
215}
216
217/* Remove the associated file_lock first, then remove the delegation.
218 * lease_modify() is called to remove the FS_LEASE file_lock from
219 * the i_flock list, eventually calling nfsd's lock_manager
220 * fl_release_callback.
221 */
222static void
223nfs4_close_delegation(struct nfs4_delegation *dp)
224{
225 struct file *filp = dp->dl_vfs_file;
226
227 dprintk("NFSD: close_delegation dp %p\n",dp);
228 dp->dl_vfs_file = NULL;
229 /* The following nfsd_close may not actually close the file,
230 * but we want to remove the lease in any case. */
c907132d 231 if (dp->dl_flock)
a9933cea 232 vfs_setlease(filp, F_UNLCK, &dp->dl_flock);
1da177e4 233 nfsd_close(filp);
1da177e4
LT		 234}
235
236/* Called under the state lock. */
237static void
238unhash_delegation(struct nfs4_delegation *dp)
239{
ea1da636
N		 240 list_del_init(&dp->dl_perfile);
241 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 242 spin_lock(&recall_lock);
243 list_del_init(&dp->dl_recall_lru);
244 spin_unlock(&recall_lock);
245 nfs4_close_delegation(dp);
246 nfs4_put_delegation(dp);
247}
248
249/*
250 * SETCLIENTID state
251 */
252
253/* Hash tables for nfs4_clientid state */
254#define CLIENT_HASH_BITS 4
255#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
256#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
257
258#define clientid_hashval(id) \
259 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 260#define clientstr_hashval(name) \
261 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 262/*
263 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
264 * used in reboot/reset lease grace period processing
265 *
266 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
267 * setclientid_confirmed info.
268 *
269 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
270 * setclientid info.
271 *
272 * client_lru holds client queue ordered by nfs4_client.cl_time
273 * for lease renewal.
274 *
275 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
276 * for last close replay.
277 */
278static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
279static int reclaim_str_hashtbl_size = 0;
280static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
281static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
282static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
283static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
284static struct list_head client_lru;
285static struct list_head close_lru;
286
2283963f
BF		 287static void unhash_generic_stateid(struct nfs4_stateid *stp)
288{
289 list_del(&stp->st_hash);
290 list_del(&stp->st_perfile);
291 list_del(&stp->st_perstateowner);
292}
293
294static void free_generic_stateid(struct nfs4_stateid *stp)
295{
296 put_nfs4_file(stp->st_file);
297 kmem_cache_free(stateid_slab, stp);
298}
299
300static void release_lock_stateid(struct nfs4_stateid *stp)
301{
302 unhash_generic_stateid(stp);
303 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
304 free_generic_stateid(stp);
305}
306
f044ff83
BF		 307static void unhash_lockowner(struct nfs4_stateowner *sop)
308{
309 struct nfs4_stateid *stp;
310
311 list_del(&sop->so_idhash);
312 list_del(&sop->so_strhash);
313 list_del(&sop->so_perstateid);
314 while (!list_empty(&sop->so_stateids)) {
315 stp = list_first_entry(&sop->so_stateids,
316 struct nfs4_stateid, st_perstateowner);
317 release_lock_stateid(stp);
318 }
319}
320
321static void release_lockowner(struct nfs4_stateowner *sop)
322{
323 unhash_lockowner(sop);
324 nfs4_put_stateowner(sop);
325}
326
f1d110ca
BF		 327static void
328release_stateid_lockowners(struct nfs4_stateid *open_stp)
329{
330 struct nfs4_stateowner *lock_sop;
331
332 while (!list_empty(&open_stp->st_lockowners)) {
333 lock_sop = list_entry(open_stp->st_lockowners.next,
334 struct nfs4_stateowner, so_perstateid);
335 /* list_del(&open_stp->st_lockowners); */
336 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 337 release_lockowner(lock_sop);
f1d110ca
BF		 338 }
339}
340
2283963f
BF		 341static void release_open_stateid(struct nfs4_stateid *stp)
342{
343 unhash_generic_stateid(stp);
344 release_stateid_lockowners(stp);
345 nfsd_close(stp->st_vfs_file);
346 free_generic_stateid(stp);
347}
348
f044ff83 349static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 350{
351 struct nfs4_stateid *stp;
352
353 list_del(&sop->so_idhash);
354 list_del(&sop->so_strhash);
f044ff83
BF		 355 list_del(&sop->so_perclient);
356 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 357 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 358 stp = list_first_entry(&sop->so_stateids,
359 struct nfs4_stateid, st_perstateowner);
360 release_open_stateid(stp);
f1d110ca
BF		 361 }
362}
363
f044ff83 364static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 365{
f044ff83 366 unhash_openowner(sop);
f1d110ca
BF		 367 list_del(&sop->so_close_lru);
368 nfs4_put_stateowner(sop);
369}
370
5282fd72
ME		 371static DEFINE_SPINLOCK(sessionid_lock);
372#define SESSION_HASH_SIZE 512
373static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
374
375static inline int
376hash_sessionid(struct nfs4_sessionid *sessionid)
377{
378 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
379
380 return sid->sequence % SESSION_HASH_SIZE;
381}
382
383static inline void
384dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
385{
386 u32 *ptr = (u32 *)(&sessionid->data[0]);
387 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
388}
389
ec6b5d7b
AA		 390static void
391gen_sessionid(struct nfsd4_session *ses)
392{
393 struct nfs4_client *clp = ses->se_client;
394 struct nfsd4_sessionid *sid;
395
396 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
397 sid->clientid = clp->cl_clientid;
398 sid->sequence = current_sessionid++;
399 sid->reserved = 0;
400}
401
402/*
a649637c
AA		 403 * The protocol defines ca_maxresponssize_cached to include the size of
404 * the rpc header, but all we need to cache is the data starting after
405 * the end of the initial SEQUENCE operation--the rest we regenerate
406 * each time. Therefore we can advertise a ca_maxresponssize_cached
407 * value that is the number of bytes in our cache plus a few additional
408 * bytes. In order to stay on the safe side, and not promise more than
409 * we can cache, those additional bytes must be the minimum possible: 24
410 * bytes of rpc header (xid through accept state, with AUTH_NULL
411 * verifier), 12 for the compound header (with zero-length tag), and 44
412 * for the SEQUENCE op response:
413 */
414#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
415
416/*
417 * Give the client the number of ca_maxresponsesize_cached slots it
418 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
419 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
420 * than NFSD_MAX_SLOTS_PER_SESSION.
ec6b5d7b 421 *
a649637c
AA		 422 * If we run out of reserved DRC memory we should (up to a point)
423 * re-negotiate active sessions and reduce their slot usage to make
424 * rooom for new connections. For now we just fail the create session.
ec6b5d7b 425 */
a649637c 426static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
ec6b5d7b 427{
a649637c 428 int mem, size = fchan->maxresp_cached;
ec6b5d7b 429
5d77ddfb
AA		 430 if (fchan->maxreqs < 1)
431 return nfserr_inval;
5d77ddfb 432
a649637c
AA		 433 if (size < NFSD_MIN_HDR_SEQ_SZ)
434 size = NFSD_MIN_HDR_SEQ_SZ;
435 size -= NFSD_MIN_HDR_SEQ_SZ;
436 if (size > NFSD_SLOT_CACHE_SIZE)
437 size = NFSD_SLOT_CACHE_SIZE;
ec6b5d7b 438
a649637c
AA		 439 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
440 mem = fchan->maxreqs * size;
441 if (mem > NFSD_MAX_MEM_PER_SESSION) {
442 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
443 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
444 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
445 mem = fchan->maxreqs * size;
446 }
ec6b5d7b 447
4bd9b0f4 448 spin_lock(&nfsd_drc_lock);
a649637c
AA		 449 /* bound the total session drc memory ussage */
450 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
451 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
452 mem = fchan->maxreqs * size;
453 }
0c193054 454 nfsd_drc_mem_used += mem;
4bd9b0f4 455 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 456
b101ebbc 457 if (fchan->maxreqs == 0)
bdac86e2 458 return nfserr_serverfault;
a649637c
AA		 459
460 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
b101ebbc 461 return 0;
ec6b5d7b
AA		 462}
463
464/*
465 * fchan holds the client values on input, and the server values on output
ddc04fd4 466 * sv_max_mesg is the maximum payload plus one page for overhead.
ec6b5d7b
AA		 467 */
468static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 469 struct nfsd4_channel_attrs *session_fchan,
470 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 471{
472 int status = 0;
ddc04fd4 473 __u32 maxcount = nfsd_serv->sv_max_mesg;
ec6b5d7b
AA		 474
475 /* headerpadsz set to zero in encode routine */
476
477 /* Use the client's max request and max response size if possible */
478 if (fchan->maxreq_sz > maxcount)
479 fchan->maxreq_sz = maxcount;
6c18ba9f 480 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 481
482 if (fchan->maxresp_sz > maxcount)
483 fchan->maxresp_sz = maxcount;
6c18ba9f 484 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 485
ec6b5d7b
AA		 486 /* Use the client's maxops if possible */
487 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
488 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 489 session_fchan->maxops = fchan->maxops;
ec6b5d7b 490
ec6b5d7b
AA		 491 /* FIXME: Error means no more DRC pages so the server should
492 * recover pages from existing sessions. For now fail session
493 * creation.
494 */
a649637c 495 status = set_forechannel_drc_size(fchan);
ec6b5d7b 496
a649637c 497 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 498 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA		 499
500 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA		 501 return status;
502}
503
557ce264
AA		 504static void
505free_session_slots(struct nfsd4_session *ses)
506{
507 int i;
508
509 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
510 kfree(ses->se_slots[i]);
511}
512
efe0cb6d
BF		 513/*
514 * We don't actually need to cache the rpc and session headers, so we
515 * can allocate a little less for each slot:
516 */
517static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
518{
519 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
520}
521
ec6b5d7b
AA		 522static int
523alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
524 struct nfsd4_create_session *cses)
525{
526 struct nfsd4_session *new, tmp;
557ce264
AA		 527 struct nfsd4_slot *sp;
528 int idx, slotsize, cachesize, i;
529 int status;
ec6b5d7b
AA		 530
531 memset(&tmp, 0, sizeof(tmp));
532
533 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 534 tmp.se_bchannel = cses->back_channel;
535 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
536 &cses->fore_channel);
ec6b5d7b
AA		 537 if (status)
538 goto out;
539
557ce264
AA		 540 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
541 + sizeof(struct nfsd4_session) > PAGE_SIZE);
542
543 status = nfserr_serverfault;
544 /* allocate struct nfsd4_session and slot table pointers in one piece */
545 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA		 546 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
547 if (!new)
548 goto out;
549
550 memcpy(new, &tmp, sizeof(*new));
551
557ce264 552 /* allocate each struct nfsd4_slot and data cache in one piece */
efe0cb6d 553 cachesize = slot_bytes(&new->se_fchannel);
557ce264
AA		 554 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
555 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
556 if (!sp)
557 goto out_free;
558 new->se_slots[i] = sp;
559 }
560
ec6b5d7b
AA		 561 new->se_client = clp;
562 gen_sessionid(new);
563 idx = hash_sessionid(&new->se_sessionid);
564 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
565 NFS4_MAX_SESSIONID_LEN);
566
567 new->se_flags = cses->flags;
568 kref_init(&new->se_ref);
569 spin_lock(&sessionid_lock);
570 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
571 list_add(&new->se_perclnt, &clp->cl_sessions);
572 spin_unlock(&sessionid_lock);
573
574 status = nfs_ok;
575out:
576 return status;
557ce264
AA		 577out_free:
578 free_session_slots(new);
579 kfree(new);
580 goto out;
ec6b5d7b
AA		 581}
582
5282fd72
ME		 583/* caller must hold sessionid_lock */
584static struct nfsd4_session *
585find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
586{
587 struct nfsd4_session *elem;
588 int idx;
589
590 dump_sessionid(__func__, sessionid);
591 idx = hash_sessionid(sessionid);
592 dprintk("%s: idx is %d\n", __func__, idx);
593 /* Search in the appropriate list */
594 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
595 dump_sessionid("list traversal", &elem->se_sessionid);
596 if (!memcmp(elem->se_sessionid.data, sessionid->data,
597 NFS4_MAX_SESSIONID_LEN)) {
598 return elem;
599 }
600 }
601
602 dprintk("%s: session not found\n", __func__);
603 return NULL;
604}
605
606/* caller must hold sessionid_lock */
7116ed6b 607static void
5282fd72 608unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 609{
610 list_del(&ses->se_hash);
611 list_del(&ses->se_perclnt);
5282fd72
ME		 612}
613
614static void
615release_session(struct nfsd4_session *ses)
616{
617 spin_lock(&sessionid_lock);
618 unhash_session(ses);
619 spin_unlock(&sessionid_lock);
7116ed6b
AA		 620 nfsd4_put_session(ses);
621}
622
623void
624free_session(struct kref *kref)
625{
626 struct nfsd4_session *ses;
dd829c45 627 int mem;
7116ed6b
AA		 628
629 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 630 spin_lock(&nfsd_drc_lock);
efe0cb6d 631 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
dd829c45 632 nfsd_drc_mem_used -= mem;
be98d1bb 633 spin_unlock(&nfsd_drc_lock);
557ce264 634 free_session_slots(ses);
7116ed6b
AA		 635 kfree(ses);
636}
637
1da177e4
LT		 638static inline void
639renew_client(struct nfs4_client *clp)
640{
641 /*
642 * Move client to the end to the LRU list.
643 */
644 dprintk("renewing client (clientid %08x/%08x)\n",
645 clp->cl_clientid.cl_boot,
646 clp->cl_clientid.cl_id);
647 list_move_tail(&clp->cl_lru, &client_lru);
648 clp->cl_time = get_seconds();
649}
650
651/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
652static int
653STALE_CLIENTID(clientid_t *clid)
654{
655 if (clid->cl_boot == boot_time)
656 return 0;
60adfc50
AA		 657 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
658 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 659 return 1;
660}
661
662/*
663 * XXX Should we use a slab cache ?
664 * This type of memory management is somewhat inefficient, but we use it
665 * anyway since SETCLIENTID is not a common operation.
666 */
35bba9a3 667static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 668{
669 struct nfs4_client *clp;
670
35bba9a3
BF		 671 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
672 if (clp == NULL)
673 return NULL;
674 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
675 if (clp->cl_name.data == NULL) {
676 kfree(clp);
677 return NULL;
1da177e4 678 }
35bba9a3
BF		 679 memcpy(clp->cl_name.data, name.data, name.len);
680 clp->cl_name.len = name.len;
1da177e4
LT		 681 return clp;
682}
683
684static inline void
685free_client(struct nfs4_client *clp)
686{
687 if (clp->cl_cred.cr_group_info)
688 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 689 kfree(clp->cl_principal);
1da177e4
LT		 690 kfree(clp->cl_name.data);
691 kfree(clp);
692}
693
1da177e4
LT		 694static void
695expire_client(struct nfs4_client *clp)
696{
697 struct nfs4_stateowner *sop;
698 struct nfs4_delegation *dp;
1da177e4
LT		 699 struct list_head reaplist;
700
1da177e4
LT		 701 INIT_LIST_HEAD(&reaplist);
702 spin_lock(&recall_lock);
ea1da636
N		 703 while (!list_empty(&clp->cl_delegations)) {
704 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 705 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
706 dp->dl_flock);
ea1da636 707 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 708 list_move(&dp->dl_recall_lru, &reaplist);
709 }
710 spin_unlock(&recall_lock);
711 while (!list_empty(&reaplist)) {
712 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
713 list_del_init(&dp->dl_recall_lru);
714 unhash_delegation(dp);
715 }
716 list_del(&clp->cl_idhash);
717 list_del(&clp->cl_strhash);
718 list_del(&clp->cl_lru);
ea1da636
N		 719 while (!list_empty(&clp->cl_openowners)) {
720 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 721 release_openowner(sop);
1da177e4 722 }
c4bf7868
ME		 723 while (!list_empty(&clp->cl_sessions)) {
724 struct nfsd4_session *ses;
725 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
726 se_perclnt);
727 release_session(ses);
728 }
b5a1a81e 729 nfsd4_set_callback_client(clp, NULL);
2bf23875
BF		 730 if (clp->cl_cb_conn.cb_xprt)
731 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
b12a05cb 732 free_client(clp);
1da177e4
LT		 733}
734
35bba9a3
BF		 735static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
736{
737 memcpy(target->cl_verifier.data, source->data,
738 sizeof(target->cl_verifier.data));
1da177e4
LT		 739}
740
35bba9a3
BF		 741static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
742{
1da177e4
LT		 743 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
744 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
745}
746
35bba9a3
BF		 747static void copy_cred(struct svc_cred *target, struct svc_cred *source)
748{
1da177e4
LT		 749 target->cr_uid = source->cr_uid;
750 target->cr_gid = source->cr_gid;
751 target->cr_group_info = source->cr_group_info;
752 get_group_info(target->cr_group_info);
753}
754
35bba9a3 755static int same_name(const char *n1, const char *n2)
599e0a22 756{
a55370a3 757 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 758}
759
760static int
599e0a22
BF		 761same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
762{
763 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 764}
765
766static int
599e0a22
BF		 767same_clid(clientid_t *cl1, clientid_t *cl2)
768{
769 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 770}
771
772/* XXX what about NGROUP */
773static int
599e0a22
BF		 774same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
775{
776 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 777}
778
5ec7b46c
BF		 779static void gen_clid(struct nfs4_client *clp)
780{
781 static u32 current_clientid = 1;
782
1da177e4
LT		 783 clp->cl_clientid.cl_boot = boot_time;
784 clp->cl_clientid.cl_id = current_clientid++;
785}
786
deda2faa
BF		 787static void gen_confirm(struct nfs4_client *clp)
788{
789 static u32 i;
790 u32 *p;
1da177e4 791
1da177e4 792 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 793 *p++ = get_seconds();
794 *p++ = i++;
1da177e4
LT		 795}
796
b09333c4
RL		 797static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
798 struct svc_rqst *rqstp, nfs4_verifier *verf)
799{
800 struct nfs4_client *clp;
801 struct sockaddr *sa = svc_addr(rqstp);
802 char *princ;
803
804 clp = alloc_client(name);
805 if (clp == NULL)
806 return NULL;
807
808 princ = svc_gss_principal(rqstp);
809 if (princ) {
810 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
811 if (clp->cl_principal == NULL) {
812 free_client(clp);
813 return NULL;
814 }
815 }
816
817 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
2bf23875 818 atomic_set(&clp->cl_cb_set, 0);
b09333c4
RL		 819 INIT_LIST_HEAD(&clp->cl_idhash);
820 INIT_LIST_HEAD(&clp->cl_strhash);
821 INIT_LIST_HEAD(&clp->cl_openowners);
822 INIT_LIST_HEAD(&clp->cl_delegations);
823 INIT_LIST_HEAD(&clp->cl_sessions);
824 INIT_LIST_HEAD(&clp->cl_lru);
825 clear_bit(0, &clp->cl_cb_slot_busy);
826 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
827 copy_verf(clp, verf);
828 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
829 clp->cl_flavor = rqstp->rq_flavor;
830 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
831 gen_confirm(clp);
832
833 return clp;
834}
835
35bba9a3
BF		 836static int check_name(struct xdr_netobj name)
837{
1da177e4
LT		 838 if (name.len == 0)
839 return 0;
840 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 841 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 842 return 0;
843 }
844 return 1;
845}
846
fd39ca9a 847static void
1da177e4
LT		 848add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
849{
850 unsigned int idhashval;
851
852 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
853 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
854 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
855 list_add_tail(&clp->cl_lru, &client_lru);
856 clp->cl_time = get_seconds();
857}
858
fd39ca9a 859static void
1da177e4
LT		 860move_to_confirmed(struct nfs4_client *clp)
861{
862 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
863 unsigned int strhashval;
864
865 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
866 list_del_init(&clp->cl_strhash);
f116629d 867 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 868 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4
LT		 869 list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
870 renew_client(clp);
871}
872
873static struct nfs4_client *
874find_confirmed_client(clientid_t *clid)
875{
876 struct nfs4_client *clp;
877 unsigned int idhashval = clientid_hashval(clid->cl_id);
878
879 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 880 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 881 return clp;
882 }
883 return NULL;
884}
885
886static struct nfs4_client *
887find_unconfirmed_client(clientid_t *clid)
888{
889 struct nfs4_client *clp;
890 unsigned int idhashval = clientid_hashval(clid->cl_id);
891
892 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 893 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 894 return clp;
895 }
896 return NULL;
897}
898
a1bcecd2
AA		 899/*
900 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
901 * parameter. Matching is based on the fact the at least one of the
902 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
903 *
904 * FIXME: we need to unify the clientid namespaces for nfsv4.x
905 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
906 * and SET_CLIENTID{,_CONFIRM}
907 */
908static inline int
909match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
910{
911 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
912 return use_exchange_id == has_exchange_flags;
913}
914
28ce6054 915static struct nfs4_client *
a1bcecd2
AA		 916find_confirmed_client_by_str(const char *dname, unsigned int hashval,
917 bool use_exchange_id)
28ce6054
N		 918{
919 struct nfs4_client *clp;
920
921 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 922 if (same_name(clp->cl_recdir, dname) &&
923 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 924 return clp;
925 }
926 return NULL;
927}
928
929static struct nfs4_client *
a1bcecd2
AA		 930find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
931 bool use_exchange_id)
28ce6054
N		 932{
933 struct nfs4_client *clp;
934
935 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 936 if (same_name(clp->cl_recdir, dname) &&
937 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 938 return clp;
939 }
940 return NULL;
941}
942
fd39ca9a 943static void
fbf4665f 944gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 945{
c237dc03 946 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 947 unsigned short expected_family;
948
949 /* Currently, we only support tcp and tcp6 for the callback channel */
950 if (se->se_callback_netid_len == 3 &&
951 !memcmp(se->se_callback_netid_val, "tcp", 3))
952 expected_family = AF_INET;
953 else if (se->se_callback_netid_len == 4 &&
954 !memcmp(se->se_callback_netid_val, "tcp6", 4))
955 expected_family = AF_INET6;
956 else
1da177e4
LT		 957 goto out_err;
958
aa9a4ec7
JL		 959 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
960 se->se_callback_addr_len,
961 (struct sockaddr *) &cb->cb_addr,
962 sizeof(cb->cb_addr));
963
7077ecba 964 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 965 goto out_err;
aa9a4ec7 966
fbf4665f
JL		 967 if (cb->cb_addr.ss_family == AF_INET6)
968 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
969
ab52ae6d 970 cb->cb_minorversion = 0;
1da177e4
LT		 971 cb->cb_prog = se->se_callback_prog;
972 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 973 return;
974out_err:
aa9a4ec7
JL		 975 cb->cb_addr.ss_family = AF_UNSPEC;
976 cb->cb_addrlen = 0;
849823c5 977 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 978 "will not receive delegations\n",
979 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
980
1da177e4
LT		 981 return;
982}
983
074fe897 984/*
557ce264 985 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897 986 */
074fe897
AA		 987void
988nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
074fe897 989{
557ce264
AA		 990 struct nfsd4_slot *slot = resp->cstate.slot;
991 unsigned int base;
074fe897 992
557ce264 993 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 994
557ce264
AA		 995 slot->sl_opcnt = resp->opcnt;
996 slot->sl_status = resp->cstate.status;
074fe897 997
bf864a31 998 if (nfsd4_not_cached(resp)) {
557ce264 999 slot->sl_datalen = 0;
bf864a31 1000 return;
074fe897 1001 }
557ce264
AA		 1002 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1003 base = (char *)resp->cstate.datap -
1004 (char *)resp->xbuf->head[0].iov_base;
1005 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1006 slot->sl_datalen))
1007 WARN("%s: sessions DRC could not cache compound\n", __func__);
1008 return;
074fe897
AA		 1009}
1010
1011/*
abfabf8c
AA		 1012 * Encode the replay sequence operation from the slot values.
1013 * If cachethis is FALSE encode the uncached rep error on the next
1014 * operation which sets resp->p and increments resp->opcnt for
1015 * nfs4svc_encode_compoundres.
074fe897 1016 *
074fe897 1017 */
abfabf8c
AA		 1018static __be32
1019nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1020 struct nfsd4_compoundres *resp)
074fe897 1021{
abfabf8c
AA		 1022 struct nfsd4_op *op;
1023 struct nfsd4_slot *slot = resp->cstate.slot;
bf864a31 1024
abfabf8c 1025 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1026 resp->opcnt, resp->cstate.slot->sl_cachethis);
bf864a31 1027
abfabf8c
AA		 1028 /* Encode the replayed sequence operation */
1029 op = &args->ops[resp->opcnt - 1];
1030 nfsd4_encode_operation(resp, op);
bf864a31 1031
abfabf8c 1032 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1033 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA		 1034 op = &args->ops[resp->opcnt++];
1035 op->status = nfserr_retry_uncached_rep;
1036 nfsd4_encode_operation(resp, op);
074fe897 1037 }
abfabf8c 1038 return op->status;
074fe897
AA		 1039}
1040
1041/*
557ce264
AA		 1042 * The sequence operation is not cached because we can use the slot and
1043 * session values.
074fe897
AA		 1044 */
1045__be32
bf864a31
AA		 1046nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1047 struct nfsd4_sequence *seq)
074fe897 1048{
557ce264 1049 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA		 1050 __be32 status;
1051
557ce264 1052 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1053
abfabf8c
AA		 1054 /* Either returns 0 or nfserr_retry_uncached */
1055 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1056 if (status == nfserr_retry_uncached_rep)
1057 return status;
074fe897 1058
557ce264
AA		 1059 /* The sequence operation has been encoded, cstate->datap set. */
1060 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1061
557ce264
AA		 1062 resp->opcnt = slot->sl_opcnt;
1063 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1064 status = slot->sl_status;
074fe897
AA		 1065
1066 return status;
1067}
1068
0733d213
AA		 1069/*
1070 * Set the exchange_id flags returned by the server.
1071 */
1072static void
1073nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1074{
1075 /* pNFS is not supported */
1076 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1077
1078 /* Referrals are supported, Migration is not. */
1079 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1080
1081 /* set the wire flags to return to client. */
1082 clid->flags = new->cl_exchange_flags;
1083}
1084
069b6ad4
AA		 1085__be32
1086nfsd4_exchange_id(struct svc_rqst *rqstp,
1087 struct nfsd4_compound_state *cstate,
1088 struct nfsd4_exchange_id *exid)
1089{
0733d213
AA		 1090 struct nfs4_client *unconf, *conf, *new;
1091 int status;
1092 unsigned int strhashval;
1093 char dname[HEXDIR_LEN];
363168b4 1094 char addr_str[INET6_ADDRSTRLEN];
0733d213 1095 nfs4_verifier verf = exid->verifier;
363168b4 1096 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1097
363168b4 1098 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1099 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1100 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1101 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1102 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1103
1104 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1105 return nfserr_inval;
1106
1107 /* Currently only support SP4_NONE */
1108 switch (exid->spa_how) {
1109 case SP4_NONE:
1110 break;
1111 case SP4_SSV:
1112 return nfserr_encr_alg_unsupp;
1113 default:
1114 BUG(); /* checked by xdr code */
1115 case SP4_MACH_CRED:
1116 return nfserr_serverfault; /* no excuse :-/ */
1117 }
1118
1119 status = nfs4_make_rec_clidname(dname, &exid->clname);
1120
1121 if (status)
1122 goto error;
1123
1124 strhashval = clientstr_hashval(dname);
1125
1126 nfs4_lock_state();
1127 status = nfs_ok;
1128
a1bcecd2 1129 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1130 if (conf) {
1131 if (!same_verf(&verf, &conf->cl_verifier)) {
1132 /* 18.35.4 case 8 */
1133 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1134 status = nfserr_not_same;
1135 goto out;
1136 }
1137 /* Client reboot: destroy old state */
1138 expire_client(conf);
1139 goto out_new;
1140 }
1141 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1142 /* 18.35.4 case 9 */
1143 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1144 status = nfserr_perm;
1145 goto out;
1146 }
1147 expire_client(conf);
1148 goto out_new;
1149 }
0733d213
AA		 1150 /*
1151 * Set bit when the owner id and verifier map to an already
1152 * confirmed client id (18.35.3).
1153 */
1154 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1155
1156 /*
1157 * Falling into 18.35.4 case 2, possible router replay.
1158 * Leave confirmed record intact and return same result.
1159 */
1160 copy_verf(conf, &verf);
1161 new = conf;
1162 goto out_copy;
6ddbbbfe
MS		 1163 }
1164
1165 /* 18.35.4 case 7 */
1166 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1167 status = nfserr_noent;
1168 goto out;
0733d213
AA		 1169 }
1170
a1bcecd2 1171 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1172 if (unconf) {
1173 /*
1174 * Possible retry or client restart. Per 18.35.4 case 4,
1175 * a new unconfirmed record should be generated regardless
1176 * of whether any properties have changed.
1177 */
1178 expire_client(unconf);
1179 }
1180
1181out_new:
1182 /* Normal case */
b09333c4 1183 new = create_client(exid->clname, dname, rqstp, &verf);
0733d213 1184 if (new == NULL) {
bdac86e2 1185 status = nfserr_serverfault;
0733d213
AA		 1186 goto out;
1187 }
1188
0733d213 1189 gen_clid(new);
0733d213
AA		 1190 add_to_unconfirmed(new, strhashval);
1191out_copy:
1192 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1193 exid->clientid.cl_id = new->cl_clientid.cl_id;
1194
38eb76a5 1195 exid->seqid = 1;
0733d213
AA		 1196 nfsd4_set_ex_flags(new, exid);
1197
1198 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1199 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1200 status = nfs_ok;
1201
1202out:
1203 nfs4_unlock_state();
1204error:
1205 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1206 return status;
069b6ad4
AA		 1207}
1208
b85d4c01 1209static int
88e588d5 1210check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1211{
88e588d5
AA		 1212 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1213 slot_seqid);
b85d4c01
BH		 1214
1215 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1216 if (slot_inuse) {
1217 if (seqid == slot_seqid)
b85d4c01
BH		 1218 return nfserr_jukebox;
1219 else
1220 return nfserr_seq_misordered;
1221 }
1222 /* Normal */
88e588d5 1223 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1224 return nfs_ok;
1225 /* Replay */
88e588d5 1226 if (seqid == slot_seqid)
b85d4c01
BH		 1227 return nfserr_replay_cache;
1228 /* Wraparound */
88e588d5 1229 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1230 return nfs_ok;
1231 /* Misordered replay or misordered new request */
1232 return nfserr_seq_misordered;
1233}
1234
49557cc7
AA		 1235/*
1236 * Cache the create session result into the create session single DRC
1237 * slot cache by saving the xdr structure. sl_seqid has been set.
1238 * Do this for solo or embedded create session operations.
1239 */
1240static void
1241nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1242 struct nfsd4_clid_slot *slot, int nfserr)
1243{
1244 slot->sl_status = nfserr;
1245 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1246}
1247
1248static __be32
1249nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1250 struct nfsd4_clid_slot *slot)
1251{
1252 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1253 return slot->sl_status;
1254}
1255
069b6ad4
AA		 1256__be32
1257nfsd4_create_session(struct svc_rqst *rqstp,
1258 struct nfsd4_compound_state *cstate,
1259 struct nfsd4_create_session *cr_ses)
1260{
363168b4 1261 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1262 struct nfs4_client *conf, *unconf;
49557cc7 1263 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1264 int status = 0;
1265
1266 nfs4_lock_state();
1267 unconf = find_unconfirmed_client(&cr_ses->clientid);
1268 conf = find_confirmed_client(&cr_ses->clientid);
1269
1270 if (conf) {
49557cc7
AA		 1271 cs_slot = &conf->cl_cs_slot;
1272 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1273 if (status == nfserr_replay_cache) {
ec6b5d7b 1274 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1275 cs_slot->sl_seqid);
38eb76a5 1276 /* Return the cached reply status */
49557cc7 1277 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1278 goto out;
49557cc7 1279 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1280 status = nfserr_seq_misordered;
1281 dprintk("Sequence misordered!\n");
1282 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1283 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1284 goto out;
1285 }
49557cc7 1286 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1287 } else if (unconf) {
1288 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1289 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1290 status = nfserr_clid_inuse;
1291 goto out;
1292 }
1293
49557cc7
AA		 1294 cs_slot = &unconf->cl_cs_slot;
1295 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1296 if (status) {
1297 /* an unconfirmed replay returns misordered */
ec6b5d7b 1298 status = nfserr_seq_misordered;
49557cc7 1299 goto out_cache;
ec6b5d7b
AA		 1300 }
1301
49557cc7 1302 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1303 move_to_confirmed(unconf);
1304
38524ab3 1305 if (cr_ses->flags & SESSION4_BACK_CHAN) {
2bf23875
BF		 1306 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1307 svc_xprt_get(rqstp->rq_xprt);
38524ab3
AA		 1308 rpc_copy_addr(
1309 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1310 sa);
1311 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1312 unconf->cl_cb_conn.cb_minorversion =
1313 cstate->minorversion;
1314 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1315 unconf->cl_cb_seq_nr = 1;
4b21d0de 1316 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
38524ab3 1317 }
ec6b5d7b
AA		 1318 conf = unconf;
1319 } else {
1320 status = nfserr_stale_clientid;
1321 goto out;
1322 }
1323
408b79bc
BF		 1324 /*
1325 * We do not support RDMA or persistent sessions
1326 */
1327 cr_ses->flags &= ~SESSION4_PERSIST;
1328 cr_ses->flags &= ~SESSION4_RDMA;
1329
ec6b5d7b
AA		 1330 status = alloc_init_session(rqstp, conf, cr_ses);
1331 if (status)
1332 goto out;
1333
ec6b5d7b
AA		 1334 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1335 NFS4_MAX_SESSIONID_LEN);
49557cc7 1336 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1337
49557cc7
AA		 1338out_cache:
1339 /* cache solo and embedded create sessions under the state lock */
1340 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1341out:
1342 nfs4_unlock_state();
1343 dprintk("%s returns %d\n", __func__, ntohl(status));
1344 return status;
069b6ad4
AA		 1345}
1346
57716355
BF		 1347static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1348{
1349 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1350 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1351
1352 return argp->opcnt == resp->opcnt;
1353}
1354
5d4cec2f
BF		 1355static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1356{
1357 if (!session)
1358 return 0;
1359 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1360}
1361
069b6ad4
AA		 1362__be32
1363nfsd4_destroy_session(struct svc_rqst *r,
1364 struct nfsd4_compound_state *cstate,
1365 struct nfsd4_destroy_session *sessionid)
1366{
e10e0cfc
BH		 1367 struct nfsd4_session *ses;
1368 u32 status = nfserr_badsession;
1369
1370 /* Notes:
1371 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1372 * - Should we return nfserr_back_chan_busy if waiting for
1373 * callbacks on to-be-destroyed session?
1374 * - Do we need to clear any callback info from previous session?
1375 */
1376
5d4cec2f 1377 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
57716355
BF		 1378 if (!nfsd4_last_compound_op(r))
1379 return nfserr_not_only_op;
1380 }
e10e0cfc
BH		 1381 dump_sessionid(__func__, &sessionid->sessionid);
1382 spin_lock(&sessionid_lock);
1383 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1384 if (!ses) {
1385 spin_unlock(&sessionid_lock);
1386 goto out;
1387 }
1388
1389 unhash_session(ses);
1390 spin_unlock(&sessionid_lock);
1391
1392 /* wait for callbacks */
b5a1a81e 1393 nfsd4_set_callback_client(ses->se_client, NULL);
e10e0cfc
BH		 1394 nfsd4_put_session(ses);
1395 status = nfs_ok;
1396out:
1397 dprintk("%s returns %d\n", __func__, ntohl(status));
1398 return status;
069b6ad4
AA		 1399}
1400
1401__be32
b85d4c01 1402nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1403 struct nfsd4_compound_state *cstate,
1404 struct nfsd4_sequence *seq)
1405{
f9bb94c4 1406 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1407 struct nfsd4_session *session;
1408 struct nfsd4_slot *slot;
1409 int status;
1410
f9bb94c4
AA		 1411 if (resp->opcnt != 1)
1412 return nfserr_sequence_pos;
1413
b85d4c01
BH		 1414 spin_lock(&sessionid_lock);
1415 status = nfserr_badsession;
1416 session = find_in_sessionid_hashtbl(&seq->sessionid);
1417 if (!session)
1418 goto out;
1419
1420 status = nfserr_badslot;
6c18ba9f 1421 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1422 goto out;
1423
557ce264 1424 slot = session->se_slots[seq->slotid];
b85d4c01
BH		 1425 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1426
a8dfdaeb
AA		 1427 /* We do not negotiate the number of slots yet, so set the
1428 * maxslots to the session maxreqs which is used to encode
1429 * sr_highest_slotid and the sr_target_slot id to maxslots */
1430 seq->maxslots = session->se_fchannel.maxreqs;
1431
88e588d5 1432 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1433 if (status == nfserr_replay_cache) {
1434 cstate->slot = slot;
1435 cstate->session = session;
da3846a2 1436 /* Return the cached reply status and set cstate->status
557ce264 1437 * for nfsd4_proc_compound processing */
bf864a31 1438 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1439 cstate->status = nfserr_replay_cache;
aaf84eb9 1440 goto out;
b85d4c01
BH		 1441 }
1442 if (status)
1443 goto out;
1444
1445 /* Success! bump slot seqid */
1446 slot->sl_inuse = true;
1447 slot->sl_seqid = seq->seqid;
557ce264 1448 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH		 1449
1450 cstate->slot = slot;
1451 cstate->session = session;
1452
b85d4c01 1453out:
26c0c75e
BF		 1454 /* Hold a session reference until done processing the compound. */
1455 if (cstate->session)
1456 nfsd4_get_session(cstate->session);
b85d4c01 1457 spin_unlock(&sessionid_lock);
aaf84eb9
BH		 1458 /* Renew the clientid on success and on replay */
1459 if (cstate->session) {
1460 nfs4_lock_state();
1461 renew_client(session->se_client);
1462 nfs4_unlock_state();
1463 }
b85d4c01
BH		 1464 dprintk("%s: return %d\n", __func__, ntohl(status));
1465 return status;
069b6ad4
AA		 1466}
1467
b37ad28b 1468__be32
b591480b
BF		 1469nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1470 struct nfsd4_setclientid *setclid)
1da177e4 1471{
363168b4 1472 struct sockaddr *sa = svc_addr(rqstp);
1da177e4
LT		 1473 struct xdr_netobj clname = {
1474 .len = setclid->se_namelen,
1475 .data = setclid->se_name,
1476 };
1477 nfs4_verifier clverifier = setclid->se_verf;
1478 unsigned int strhashval;
28ce6054 1479 struct nfs4_client *conf, *unconf, *new;
b37ad28b 1480 __be32 status;
a55370a3 1481 char dname[HEXDIR_LEN];
1da177e4 1482
1da177e4 1483 if (!check_name(clname))
73aea4ec 1484 return nfserr_inval;
1da177e4 1485
a55370a3
N		 1486 status = nfs4_make_rec_clidname(dname, &clname);
1487 if (status)
73aea4ec 1488 return status;
a55370a3 1489
1da177e4
LT		 1490 /*
1491 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1492 * We get here on a DRC miss.
1493 */
1494
a55370a3 1495 strhashval = clientstr_hashval(dname);
1da177e4 1496
1da177e4 1497 nfs4_lock_state();
a1bcecd2 1498 conf = find_confirmed_client_by_str(dname, strhashval, false);
28ce6054 1499 if (conf) {
a186e767 1500 /* RFC 3530 14.2.33 CASE 0: */
1da177e4 1501 status = nfserr_clid_inuse;
026722c2 1502 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
363168b4
JL		 1503 char addr_str[INET6_ADDRSTRLEN];
1504 rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
1505 sizeof(addr_str));
1506 dprintk("NFSD: setclientid: string in use by client "
1507 "at %s\n", addr_str);
1da177e4
LT		 1508 goto out;
1509 }
1da177e4 1510 }
a186e767
BF		 1511 /*
1512 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
1513 * has a description of SETCLIENTID request processing consisting
1514 * of 5 bullet points, labeled as CASE0 - CASE4 below.
1515 */
a1bcecd2 1516 unconf = find_unconfirmed_client_by_str(dname, strhashval, false);
1da177e4
LT		 1517 status = nfserr_resource;
1518 if (!conf) {
a186e767
BF		 1519 /*
1520 * RFC 3530 14.2.33 CASE 4:
1521 * placed first, because it is the normal case
1da177e4
LT		 1522 */
1523 if (unconf)
1524 expire_client(unconf);
b09333c4 1525 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1526 if (new == NULL)
1da177e4 1527 goto out;
1da177e4 1528 gen_clid(new);
599e0a22 1529 } else if (same_verf(&conf->cl_verifier, &clverifier)) {
1da177e4 1530 /*
a186e767
BF		 1531 * RFC 3530 14.2.33 CASE 1:
1532 * probable callback update
1da177e4 1533 */
31f4a6c1
N		 1534 if (unconf) {
1535 /* Note this is removing unconfirmed {*x***},
1536 * which is stronger than RFC recommended {vxc**}.
1537 * This has the advantage that there is at most
1538 * one {*x***} in either list at any time.
1539 */
1540 expire_client(unconf);
1da177e4 1541 }
b09333c4 1542 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1543 if (new == NULL)
1da177e4 1544 goto out;
1da177e4 1545 copy_clid(new, conf);
1da177e4
LT		 1546 } else if (!unconf) {
1547 /*
a186e767
BF		 1548 * RFC 3530 14.2.33 CASE 2:
1549 * probable client reboot; state will be removed if
1550 * confirmed.
1da177e4 1551 */
b09333c4 1552 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1553 if (new == NULL)
1da177e4 1554 goto out;
1da177e4 1555 gen_clid(new);
49ba8781 1556 } else {
a186e767
BF		 1557 /*
1558 * RFC 3530 14.2.33 CASE 3:
1559 * probable client reboot; state will be removed if
1560 * confirmed.
1da177e4
LT		 1561 */
1562 expire_client(unconf);
b09333c4 1563 new = create_client(clname, dname, rqstp, &clverifier);
a55370a3 1564 if (new == NULL)
1da177e4 1565 goto out;
1da177e4 1566 gen_clid(new);
1da177e4 1567 }
fbf4665f 1568 gen_callback(new, setclid, rpc_get_scope_id(sa));
c175b83c 1569 add_to_unconfirmed(new, strhashval);
1da177e4
LT		 1570 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
1571 setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
1572 memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
1573 status = nfs_ok;
1574out:
1575 nfs4_unlock_state();
1576 return status;
1577}
1578
1579
1580/*
a186e767
BF		 1581 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
1582 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
1583 * bullets, labeled as CASE1 - CASE4 below.
1da177e4 1584 */
b37ad28b 1585__be32
b591480b
BF		 1586nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1587 struct nfsd4_compound_state *cstate,
1588 struct nfsd4_setclientid_confirm *setclientid_confirm)
1da177e4 1589{
363168b4 1590 struct sockaddr *sa = svc_addr(rqstp);
21ab45a4 1591 struct nfs4_client *conf, *unconf;
1da177e4
LT		 1592 nfs4_verifier confirm = setclientid_confirm->sc_confirm;
1593 clientid_t * clid = &setclientid_confirm->sc_clientid;
b37ad28b 1594 __be32 status;
1da177e4
LT		 1595
1596 if (STALE_CLIENTID(clid))
1597 return nfserr_stale_clientid;
1598 /*
1599 * XXX The Duplicate Request Cache (DRC) has been checked (??)
1600 * We get here on a DRC miss.
1601 */
1602
1603 nfs4_lock_state();
21ab45a4
N		 1604
1605 conf = find_confirmed_client(clid);
1606 unconf = find_unconfirmed_client(clid);
1607
1608 status = nfserr_clid_inuse;
363168b4 1609 if (conf && !rpc_cmp_addr((struct sockaddr *) &conf->cl_addr, sa))
21ab45a4 1610 goto out;
363168b4 1611 if (unconf && !rpc_cmp_addr((struct sockaddr *) &unconf->cl_addr, sa))
21ab45a4
N		 1612 goto out;
1613
a186e767
BF		 1614 /*
1615 * section 14.2.34 of RFC 3530 has a description of
1616 * SETCLIENTID_CONFIRM request processing consisting
1617 * of 4 bullet points, labeled as CASE1 - CASE4 below.
1618 */
366e0c1d 1619 if (conf && unconf && same_verf(&confirm, &unconf->cl_confirm)) {
a186e767
BF		 1620 /*
1621 * RFC 3530 14.2.34 CASE 1:
1622 * callback update
1623 */
599e0a22 1624 if (!same_creds(&conf->cl_cred, &unconf->cl_cred))
1da177e4
LT		 1625 status = nfserr_clid_inuse;
1626 else {
2bf23875 1627 atomic_set(&conf->cl_cb_set, 0);
4b21d0de 1628 nfsd4_probe_callback(conf, &unconf->cl_cb_conn);
1a69c179 1629 expire_client(unconf);
1da177e4 1630 status = nfs_ok;
1a69c179 1631
1da177e4 1632 }
f3aba4e5 1633 } else if (conf && !unconf) {
a186e767
BF		 1634 /*
1635 * RFC 3530 14.2.34 CASE 2:
1636 * probable retransmitted request; play it safe and
1637 * do nothing.
7c79f737 1638 */
599e0a22 1639 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred))
1da177e4 1640 status = nfserr_clid_inuse;
21ab45a4 1641 else
1da177e4 1642 status = nfs_ok;
7c79f737 1643 } else if (!conf && unconf
599e0a22 1644 && same_verf(&unconf->cl_confirm, &confirm)) {
a186e767
BF		 1645 /*
1646 * RFC 3530 14.2.34 CASE 3:
1647 * Normal case; new or rebooted client:
7c79f737 1648 */
599e0a22 1649 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
1da177e4
LT		 1650 status = nfserr_clid_inuse;
1651 } else {
1a69c179
N		 1652 unsigned int hash =
1653 clientstr_hashval(unconf->cl_recdir);
1654 conf = find_confirmed_client_by_str(unconf->cl_recdir,
a1bcecd2 1655 hash, false);
1a69c179 1656 if (conf) {
c7b9a459 1657 nfsd4_remove_clid_dir(conf);
1a69c179
N		 1658 expire_client(conf);
1659 }
1da177e4 1660 move_to_confirmed(unconf);
21ab45a4 1661 conf = unconf;
4b21d0de 1662 nfsd4_probe_callback(conf, &conf->cl_cb_conn);
1a69c179 1663 status = nfs_ok;
1da177e4 1664 }
599e0a22
BF		 1665 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
1666 && (!unconf || (unconf && !same_verf(&unconf->cl_confirm,
7c79f737 1667 &confirm)))) {
a186e767
BF		 1668 /*
1669 * RFC 3530 14.2.34 CASE 4:
1670 * Client probably hasn't noticed that we rebooted yet.
7c79f737 1671 */
1da177e4 1672 status = nfserr_stale_clientid;
7c79f737 1673 } else {
08e8987c
N		 1674 /* check that we have hit one of the cases...*/
1675 status = nfserr_clid_inuse;
1676 }
1da177e4 1677out:
1da177e4
LT		 1678 nfs4_unlock_state();
1679 return status;
1680}
1681
1da177e4
LT		 1682/* OPEN Share state helper functions */
1683static inline struct nfs4_file *
1684alloc_init_file(struct inode *ino)
1685{
1686 struct nfs4_file *fp;
1687 unsigned int hashval = file_hashval(ino);
1688
e60d4398
N		 1689 fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
1690 if (fp) {
8b671b80 1691 atomic_set(&fp->fi_ref, 1);
1da177e4 1692 INIT_LIST_HEAD(&fp->fi_hash);
8beefa24
N		 1693 INIT_LIST_HEAD(&fp->fi_stateids);
1694 INIT_LIST_HEAD(&fp->fi_delegations);
8b671b80 1695 spin_lock(&recall_lock);
1da177e4 1696 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
8b671b80 1697 spin_unlock(&recall_lock);
1da177e4
LT		 1698 fp->fi_inode = igrab(ino);
1699 fp->fi_id = current_fileid++;
47f9940c 1700 fp->fi_had_conflict = false;
1da177e4
LT		 1701 return fp;
1702 }
1703 return NULL;
1704}
1705
e60d4398 1706static void
e18b890b 1707nfsd4_free_slab(struct kmem_cache **slab)
1da177e4 1708{
e60d4398
N		 1709 if (*slab == NULL)
1710 return;
1a1d92c1 1711 kmem_cache_destroy(*slab);
e60d4398 1712 *slab = NULL;
1da177e4
LT		 1713}
1714
e8ff2a84 1715void
1da177e4
LT		 1716nfsd4_free_slabs(void)
1717{
e60d4398
N		 1718 nfsd4_free_slab(&stateowner_slab);
1719 nfsd4_free_slab(&file_slab);
5ac049ac 1720 nfsd4_free_slab(&stateid_slab);
5b2d21c1 1721 nfsd4_free_slab(&deleg_slab);
e60d4398 1722}
1da177e4 1723
e60d4398
N		 1724static int
1725nfsd4_init_slabs(void)
1726{
1727 stateowner_slab = kmem_cache_create("nfsd4_stateowners",
20c2df83 1728 sizeof(struct nfs4_stateowner), 0, 0, NULL);
e60d4398
N		 1729 if (stateowner_slab == NULL)
1730 goto out_nomem;
1731 file_slab = kmem_cache_create("nfsd4_files",
20c2df83 1732 sizeof(struct nfs4_file), 0, 0, NULL);
e60d4398
N		 1733 if (file_slab == NULL)
1734 goto out_nomem;
5ac049ac 1735 stateid_slab = kmem_cache_create("nfsd4_stateids",
20c2df83 1736 sizeof(struct nfs4_stateid), 0, 0, NULL);
5ac049ac
N		 1737 if (stateid_slab == NULL)
1738 goto out_nomem;
5b2d21c1 1739 deleg_slab = kmem_cache_create("nfsd4_delegations",
20c2df83 1740 sizeof(struct nfs4_delegation), 0, 0, NULL);
5b2d21c1
N		 1741 if (deleg_slab == NULL)
1742 goto out_nomem;
e60d4398
N		 1743 return 0;
1744out_nomem:
1745 nfsd4_free_slabs();
1746 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1747 return -ENOMEM;
1da177e4
LT		 1748}
1749
1750void
1751nfs4_free_stateowner(struct kref *kref)
1752{
1753 struct nfs4_stateowner *sop =
1754 container_of(kref, struct nfs4_stateowner, so_ref);
1755 kfree(sop->so_owner.data);
1756 kmem_cache_free(stateowner_slab, sop);
1757}
1758
1759static inline struct nfs4_stateowner *
1760alloc_stateowner(struct xdr_netobj *owner)
1761{
1762 struct nfs4_stateowner *sop;
1763
1764 if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1765 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1766 memcpy(sop->so_owner.data, owner->data, owner->len);
1767 sop->so_owner.len = owner->len;
1768 kref_init(&sop->so_ref);
1769 return sop;
1770 }
1771 kmem_cache_free(stateowner_slab, sop);
1772 }
1773 return NULL;
1774}
1775
1776static struct nfs4_stateowner *
1777alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1778 struct nfs4_stateowner *sop;
1779 struct nfs4_replay *rp;
1780 unsigned int idhashval;
1781
1782 if (!(sop = alloc_stateowner(&open->op_owner)))
1783 return NULL;
1784 idhashval = ownerid_hashval(current_ownerid);
1785 INIT_LIST_HEAD(&sop->so_idhash);
1786 INIT_LIST_HEAD(&sop->so_strhash);
1787 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 1788 INIT_LIST_HEAD(&sop->so_stateids);
1789 INIT_LIST_HEAD(&sop->so_perstateid); /* not used */
1da177e4
LT		 1790 INIT_LIST_HEAD(&sop->so_close_lru);
1791 sop->so_time = 0;
1792 list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1793 list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
ea1da636 1794 list_add(&sop->so_perclient, &clp->cl_openowners);
1da177e4
LT		 1795 sop->so_is_open_owner = 1;
1796 sop->so_id = current_ownerid++;
1797 sop->so_client = clp;
1798 sop->so_seqid = open->op_seqid;
1799 sop->so_confirmed = 0;
1800 rp = &sop->so_replay;
de1ae286 1801 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 1802 rp->rp_buflen = 0;
1803 rp->rp_buf = rp->rp_ibuf;
1804 return sop;
1805}
1806
1da177e4
LT		 1807static inline void
1808init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1809 struct nfs4_stateowner *sop = open->op_stateowner;
1810 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1811
1812 INIT_LIST_HEAD(&stp->st_hash);
ea1da636
N		 1813 INIT_LIST_HEAD(&stp->st_perstateowner);
1814 INIT_LIST_HEAD(&stp->st_lockowners);
1da177e4
LT		 1815 INIT_LIST_HEAD(&stp->st_perfile);
1816 list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
ea1da636 1817 list_add(&stp->st_perstateowner, &sop->so_stateids);
8beefa24 1818 list_add(&stp->st_perfile, &fp->fi_stateids);
1da177e4 1819 stp->st_stateowner = sop;
13cd2184 1820 get_nfs4_file(fp);
1da177e4 1821 stp->st_file = fp;
78155ed7 1822 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 1823 stp->st_stateid.si_stateownerid = sop->so_id;
1824 stp->st_stateid.si_fileid = fp->fi_id;
1825 stp->st_stateid.si_generation = 0;
1826 stp->st_access_bmap = 0;
1827 stp->st_deny_bmap = 0;
84459a11
AA		 1828 __set_bit(open->op_share_access & ~NFS4_SHARE_WANT_MASK,
1829 &stp->st_access_bmap);
1da177e4 1830 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
4c4cd222 1831 stp->st_openstp = NULL;
1da177e4
LT		 1832}
1833
fd39ca9a 1834static void
1da177e4
LT		 1835move_to_close_lru(struct nfs4_stateowner *sop)
1836{
1837 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1838
358dd55a 1839 list_move_tail(&sop->so_close_lru, &close_lru);
1da177e4
LT		 1840 sop->so_time = get_seconds();
1841}
1842
1da177e4 1843static int
599e0a22
BF		 1844same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner,
1845 clientid_t *clid)
1846{
1847 return (sop->so_owner.len == owner->len) &&
1848 0 == memcmp(sop->so_owner.data, owner->data, owner->len) &&
1849 (sop->so_client->cl_clientid.cl_id == clid->cl_id);
1da177e4
LT		 1850}
1851
1852static struct nfs4_stateowner *
1853find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1854{
1855 struct nfs4_stateowner *so = NULL;
1856
1857 list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 1858 if (same_owner_str(so, &open->op_owner, &open->op_clientid))
1da177e4
LT		 1859 return so;
1860 }
1861 return NULL;
1862}
1863
1864/* search file_hashtbl[] for file */
1865static struct nfs4_file *
1866find_file(struct inode *ino)
1867{
1868 unsigned int hashval = file_hashval(ino);
1869 struct nfs4_file *fp;
1870
8b671b80 1871 spin_lock(&recall_lock);
1da177e4 1872 list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
13cd2184
N		 1873 if (fp->fi_inode == ino) {
1874 get_nfs4_file(fp);
8b671b80 1875 spin_unlock(&recall_lock);
1da177e4 1876 return fp;
13cd2184 1877 }
1da177e4 1878 }
8b671b80 1879 spin_unlock(&recall_lock);
1da177e4
LT		 1880 return NULL;
1881}
1882
d87a8ade 1883static inline int access_valid(u32 x, u32 minorversion)
ba5a6a19 1884{
d87a8ade 1885 if ((x & NFS4_SHARE_ACCESS_MASK) < NFS4_SHARE_ACCESS_READ)
8838dc43 1886 return 0;
d87a8ade
AA		 1887 if ((x & NFS4_SHARE_ACCESS_MASK) > NFS4_SHARE_ACCESS_BOTH)
1888 return 0;
1889 x &= ~NFS4_SHARE_ACCESS_MASK;
1890 if (minorversion && x) {
1891 if ((x & NFS4_SHARE_WANT_MASK) > NFS4_SHARE_WANT_CANCEL)
1892 return 0;
1893 if ((x & NFS4_SHARE_WHEN_MASK) > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED)
1894 return 0;
1895 x &= ~(NFS4_SHARE_WANT_MASK | NFS4_SHARE_WHEN_MASK);
1896 }
1897 if (x)
8838dc43
BF		 1898 return 0;
1899 return 1;
ba5a6a19
BF		 1900}
1901
8838dc43 1902static inline int deny_valid(u32 x)
ba5a6a19 1903{
8838dc43
BF		 1904 /* Note: unlike access bits, deny bits may be zero. */
1905 return x <= NFS4_SHARE_DENY_BOTH;
ba5a6a19 1906}
1da177e4 1907
4f83aa30
BF		 1908/*
1909 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1910 * st_{access,deny}_bmap field of the stateid, in order to track not
1911 * only what share bits are currently in force, but also what
1912 * combinations of share bits previous opens have used. This allows us
1913 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1914 * return an error if the client attempt to downgrade to a combination
1915 * of share bits not explicable by closing some of its previous opens.
1916 *
1917 * XXX: This enforcement is actually incomplete, since we don't keep
1918 * track of access/deny bit combinations; so, e.g., we allow:
1919 *
1920 * OPEN allow read, deny write
1921 * OPEN allow both, deny none
1922 * DOWNGRADE allow read, deny none
1923 *
1924 * which we should reject.
1925 */
fd39ca9a 1926static void
1da177e4
LT		 1927set_access(unsigned int *access, unsigned long bmap) {
1928 int i;
1929
1930 *access = 0;
1931 for (i = 1; i < 4; i++) {
1932 if (test_bit(i, &bmap))
1933 *access |= i;
1934 }
1935}
1936
fd39ca9a 1937static void
1da177e4
LT		 1938set_deny(unsigned int *deny, unsigned long bmap) {
1939 int i;
1940
1941 *deny = 0;
1942 for (i = 0; i < 4; i++) {
1943 if (test_bit(i, &bmap))
1944 *deny |= i ;
1945 }
1946}
1947
1948static int
1949test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1950 unsigned int access, deny;
1951
1952 set_access(&access, stp->st_access_bmap);
1953 set_deny(&deny, stp->st_deny_bmap);
1954 if ((access & open->op_share_deny) || (deny & open->op_share_access))
1955 return 0;
1956 return 1;
1957}
1958
1959/*
1960 * Called to check deny when READ with all zero stateid or
1961 * WRITE with all zero or all one stateid
1962 */
b37ad28b 1963static __be32
1da177e4
LT		 1964nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1965{
1966 struct inode *ino = current_fh->fh_dentry->d_inode;
1967 struct nfs4_file *fp;
1968 struct nfs4_stateid *stp;
b37ad28b 1969 __be32 ret;
1da177e4
LT		 1970
1971 dprintk("NFSD: nfs4_share_conflict\n");
1972
1973 fp = find_file(ino);
13cd2184
N		 1974 if (!fp)
1975 return nfs_ok;
b700949b 1976 ret = nfserr_locked;
1da177e4 1977 /* Search for conflicting share reservations */
13cd2184
N		 1978 list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1979 if (test_bit(deny_type, &stp->st_deny_bmap) ||
1980 test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1981 goto out;
1da177e4 1982 }
13cd2184
N		 1983 ret = nfs_ok;
1984out:
1985 put_nfs4_file(fp);
1986 return ret;
1da177e4
LT		 1987}
1988
1989static inline void
1990nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1991{
1992 if (share_access & NFS4_SHARE_ACCESS_WRITE) {
aceaf78d 1993 drop_file_write_access(filp);
42e49608 1994 spin_lock(&filp->f_lock);
1da177e4 1995 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
42e49608 1996 spin_unlock(&filp->f_lock);
1da177e4
LT		 1997 }
1998}
1999
1da177e4
LT		 2000/*
2001 * Spawn a thread to perform a recall on the delegation represented
2002 * by the lease (file_lock)
2003 *
2004 * Called from break_lease() with lock_kernel() held.
2005 * Note: we assume break_lease will only call this *once* for any given
2006 * lease.
2007 */
2008static
2009void nfsd_break_deleg_cb(struct file_lock *fl)
2010{
63e4863f 2011 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1da177e4
LT		 2012
2013 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
2014 if (!dp)
2015 return;
2016
2017 /* We're assuming the state code never drops its reference
2018 * without first removing the lease. Since we're in this lease
2019 * callback (and since the lease code is serialized by the kernel
2020 * lock) we know the server hasn't removed the lease yet, we know
2021 * it's safe to take a reference: */
2022 atomic_inc(&dp->dl_count);
2023
2024 spin_lock(&recall_lock);
2025 list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
2026 spin_unlock(&recall_lock);
2027
2028 /* only place dl_time is set. protected by lock_kernel*/
2029 dp->dl_time = get_seconds();
2030
0272e1fd
BF		 2031 /*
2032 * We don't want the locks code to timeout the lease for us;
2033 * we'll remove it ourself if the delegation isn't returned
2034 * in time.
2035 */
2036 fl->fl_break_time = 0;
1da177e4 2037
63e4863f
BF		 2038 dp->dl_file->fi_had_conflict = true;
2039 nfsd4_cb_recall(dp);
1da177e4
LT		 2040}
2041
2042/*
2043 * The file_lock is being reapd.
2044 *
2045 * Called by locks_free_lock() with lock_kernel() held.
2046 */
2047static
2048void nfsd_release_deleg_cb(struct file_lock *fl)
2049{
2050 struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
2051
2052 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
2053
2054 if (!(fl->fl_flags & FL_LEASE) || !dp)
2055 return;
2056 dp->dl_flock = NULL;
2057}
2058
2059/*
2060 * Set the delegation file_lock back pointer.
2061 *
a9933cea 2062 * Called from setlease() with lock_kernel() held.
1da177e4
LT		 2063 */
2064static
2065void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
2066{
2067 struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
2068
2069 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
2070 if (!dp)
2071 return;
2072 dp->dl_flock = new;
2073}
2074
2075/*
a9933cea 2076 * Called from setlease() with lock_kernel() held
1da177e4
LT		 2077 */
2078static
2079int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
2080{
2081 struct nfs4_delegation *onlistd =
2082 (struct nfs4_delegation *)onlist->fl_owner;
2083 struct nfs4_delegation *tryd =
2084 (struct nfs4_delegation *)try->fl_owner;
2085
2086 if (onlist->fl_lmops != try->fl_lmops)
2087 return 0;
2088
2089 return onlistd->dl_client == tryd->dl_client;
2090}
2091
2092
2093static
2094int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
2095{
2096 if (arg & F_UNLCK)
2097 return lease_modify(onlist, arg);
2098 else
2099 return -EAGAIN;
2100}
2101
7b021967 2102static const struct lock_manager_operations nfsd_lease_mng_ops = {
1da177e4
LT		 2103 .fl_break = nfsd_break_deleg_cb,
2104 .fl_release_private = nfsd_release_deleg_cb,
2105 .fl_copy_lock = nfsd_copy_lock_deleg_cb,
2106 .fl_mylease = nfsd_same_client_deleg_cb,
2107 .fl_change = nfsd_change_deleg_cb,
2108};
2109
2110
b37ad28b 2111__be32
6668958f
AA		 2112nfsd4_process_open1(struct nfsd4_compound_state *cstate,
2113 struct nfsd4_open *open)
1da177e4 2114{
1da177e4
LT		 2115 clientid_t *clientid = &open->op_clientid;
2116 struct nfs4_client *clp = NULL;
2117 unsigned int strhashval;
2118 struct nfs4_stateowner *sop = NULL;
2119
1da177e4 2120 if (!check_name(open->op_owner))
0f442aa2 2121 return nfserr_inval;
1da177e4
LT		 2122
2123 if (STALE_CLIENTID(&open->op_clientid))
2124 return nfserr_stale_clientid;
2125
2126 strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
2127 sop = find_openstateowner_str(strhashval, open);
0f442aa2
BF		 2128 open->op_stateowner = sop;
2129 if (!sop) {
2130 /* Make sure the client's lease hasn't expired. */
1da177e4
LT		 2131 clp = find_confirmed_client(clientid);
2132 if (clp == NULL)
0f442aa2
BF		 2133 return nfserr_expired;
2134 goto renew;
1da177e4 2135 }
6668958f
AA		 2136 /* When sessions are used, skip open sequenceid processing */
2137 if (nfsd4_has_session(cstate))
2138 goto renew;
0f442aa2
BF		 2139 if (!sop->so_confirmed) {
2140 /* Replace unconfirmed owners without checking for replay. */
2141 clp = sop->so_client;
f044ff83 2142 release_openowner(sop);
0f442aa2
BF		 2143 open->op_stateowner = NULL;
2144 goto renew;
2145 }
2146 if (open->op_seqid == sop->so_seqid - 1) {
2147 if (sop->so_replay.rp_buflen)
a90b061c 2148 return nfserr_replay_me;
0f442aa2
BF		 2149 /* The original OPEN failed so spectacularly
2150 * that we don't even have replay data saved!
2151 * Therefore, we have no choice but to continue
2152 * processing this OPEN; presumably, we'll
2153 * fail again for the same reason.
2154 */
2155 dprintk("nfsd4_process_open1: replay with no replay cache\n");
2156 goto renew;
2157 }
2158 if (open->op_seqid != sop->so_seqid)
2159 return nfserr_bad_seqid;
1da177e4 2160renew:
0f442aa2
BF		 2161 if (open->op_stateowner == NULL) {
2162 sop = alloc_init_open_stateowner(strhashval, clp, open);
2163 if (sop == NULL)
2164 return nfserr_resource;
2165 open->op_stateowner = sop;
2166 }
2167 list_del_init(&sop->so_close_lru);
1da177e4 2168 renew_client(sop->so_client);
0f442aa2 2169 return nfs_ok;
1da177e4
LT		 2170}
2171
b37ad28b 2172static inline __be32
4a6e43e6
N		 2173nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
2174{
2175 if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
2176 return nfserr_openmode;
2177 else
2178 return nfs_ok;
2179}
2180
52f4fb43
N		 2181static struct nfs4_delegation *
2182find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
2183{
2184 struct nfs4_delegation *dp;
2185
ea1da636 2186 list_for_each_entry(dp, &fp->fi_delegations, dl_perfile) {
52f4fb43
N		 2187 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
2188 return dp;
2189 }
2190 return NULL;
2191}
2192
b37ad28b 2193static __be32
567d9829
N		 2194nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
2195 struct nfs4_delegation **dp)
2196{
2197 int flags;
b37ad28b 2198 __be32 status = nfserr_bad_stateid;
567d9829
N		 2199
2200 *dp = find_delegation_file(fp, &open->op_delegate_stateid);
2201 if (*dp == NULL)
c44c5eeb 2202 goto out;
567d9829
N		 2203 flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
2204 RD_STATE : WR_STATE;
2205 status = nfs4_check_delegmode(*dp, flags);
2206 if (status)
2207 *dp = NULL;
c44c5eeb
N		 2208out:
2209 if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
2210 return nfs_ok;
2211 if (status)
2212 return status;
2213 open->op_stateowner->so_confirmed = 1;
2214 return nfs_ok;
567d9829
N		 2215}
2216
b37ad28b 2217static __be32
1da177e4
LT		 2218nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
2219{
2220 struct nfs4_stateid *local;
b37ad28b 2221 __be32 status = nfserr_share_denied;
1da177e4
LT		 2222 struct nfs4_stateowner *sop = open->op_stateowner;
2223
8beefa24 2224 list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
1da177e4
LT		 2225 /* ignore lock owners */
2226 if (local->st_stateowner->so_is_open_owner == 0)
2227 continue;
2228 /* remember if we have seen this open owner */
2229 if (local->st_stateowner == sop)
2230 *stpp = local;
2231 /* check for conflicting share reservations */
2232 if (!test_share(local, open))
2233 goto out;
2234 }
2235 status = 0;
2236out:
2237 return status;
2238}
2239
5ac049ac
N		 2240static inline struct nfs4_stateid *
2241nfs4_alloc_stateid(void)
2242{
2243 return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
2244}
2245
b37ad28b 2246static __be32
1da177e4 2247nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
567d9829 2248 struct nfs4_delegation *dp,
1da177e4
LT		 2249 struct svc_fh *cur_fh, int flags)
2250{
2251 struct nfs4_stateid *stp;
1da177e4 2252
5ac049ac 2253 stp = nfs4_alloc_stateid();
1da177e4
LT		 2254 if (stp == NULL)
2255 return nfserr_resource;
2256
567d9829
N		 2257 if (dp) {
2258 get_file(dp->dl_vfs_file);
2259 stp->st_vfs_file = dp->dl_vfs_file;
2260 } else {
b37ad28b 2261 __be32 status;
567d9829
N		 2262 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
2263 &stp->st_vfs_file);
2264 if (status) {
2265 if (status == nfserr_dropit)
2266 status = nfserr_jukebox;
5ac049ac 2267 kmem_cache_free(stateid_slab, stp);
567d9829
N		 2268 return status;
2269 }
1da177e4 2270 }
1da177e4
LT		 2271 *stpp = stp;
2272 return 0;
2273}
2274
b37ad28b 2275static inline __be32
1da177e4
LT		 2276nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
2277 struct nfsd4_open *open)
2278{
2279 struct iattr iattr = {
2280 .ia_valid = ATTR_SIZE,
2281 .ia_size = 0,
2282 };
2283 if (!open->op_truncate)
2284 return 0;
2285 if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
9246585a 2286 return nfserr_inval;
1da177e4
LT		 2287 return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
2288}
2289
b37ad28b 2290static __be32
1da177e4
LT		 2291nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
2292{
2293 struct file *filp = stp->st_vfs_file;
7eaa36e2 2294 struct inode *inode = filp->f_path.dentry->d_inode;
6c26d08f 2295 unsigned int share_access, new_writer;
b37ad28b 2296 __be32 status;
1da177e4
LT		 2297
2298 set_access(&share_access, stp->st_access_bmap);
6c26d08f
BF		 2299 new_writer = (~share_access) & open->op_share_access
2300 & NFS4_SHARE_ACCESS_WRITE;
1da177e4 2301
6c26d08f 2302 if (new_writer) {
b37ad28b
AV		 2303 int err = get_write_access(inode);
2304 if (err)
2305 return nfserrno(err);
e518f056
BH		 2306 err = mnt_want_write(cur_fh->fh_export->ex_path.mnt);
2307 if (err)
2308 return nfserrno(err);
2309 file_take_write(filp);
6c26d08f 2310 }
1da177e4
LT		 2311 status = nfsd4_truncate(rqstp, cur_fh, open);
2312 if (status) {
6c26d08f
BF		 2313 if (new_writer)
2314 put_write_access(inode);
1da177e4
LT		 2315 return status;
2316 }
2317 /* remember the open */
6c26d08f 2318 filp->f_mode |= open->op_share_access;
b55e0ba1
BF		 2319 __set_bit(open->op_share_access, &stp->st_access_bmap);
2320 __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1da177e4
LT		 2321
2322 return nfs_ok;
2323}
2324
2325
1da177e4 2326static void
37515177 2327nfs4_set_claim_prev(struct nfsd4_open *open)
1da177e4 2328{
37515177
N		 2329 open->op_stateowner->so_confirmed = 1;
2330 open->op_stateowner->so_client->cl_firststate = 1;
1da177e4
LT		 2331}
2332
2333/*
2334 * Attempt to hand out a delegation.
2335 */
2336static void
2337nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
2338{
2339 struct nfs4_delegation *dp;
2340 struct nfs4_stateowner *sop = stp->st_stateowner;
2bf23875 2341 int cb_up = atomic_read(&sop->so_client->cl_cb_set);
1da177e4
LT		 2342 struct file_lock fl, *flp = &fl;
2343 int status, flag = 0;
2344
2345 flag = NFS4_OPEN_DELEGATE_NONE;
7b190fec
N		 2346 open->op_recall = 0;
2347 switch (open->op_claim_type) {
2348 case NFS4_OPEN_CLAIM_PREVIOUS:
2bf23875 2349 if (!cb_up)
7b190fec
N		 2350 open->op_recall = 1;
2351 flag = open->op_delegate_type;
2352 if (flag == NFS4_OPEN_DELEGATE_NONE)
2353 goto out;
2354 break;
2355 case NFS4_OPEN_CLAIM_NULL:
2356 /* Let's not give out any delegations till everyone's
2357 * had the chance to reclaim theirs.... */
af558e33 2358 if (locks_in_grace())
7b190fec 2359 goto out;
2bf23875 2360 if (!cb_up || !sop->so_confirmed)
7b190fec
N		 2361 goto out;
2362 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
2363 flag = NFS4_OPEN_DELEGATE_WRITE;
2364 else
2365 flag = NFS4_OPEN_DELEGATE_READ;
2366 break;
2367 default:
2368 goto out;
2369 }
1da177e4
LT		 2370
2371 dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
2372 if (dp == NULL) {
2373 flag = NFS4_OPEN_DELEGATE_NONE;
2374 goto out;
2375 }
2376 locks_init_lock(&fl);
2377 fl.fl_lmops = &nfsd_lease_mng_ops;
2378 fl.fl_flags = FL_LEASE;
9167f501 2379 fl.fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
1da177e4
LT		 2380 fl.fl_end = OFFSET_MAX;
2381 fl.fl_owner = (fl_owner_t)dp;
2382 fl.fl_file = stp->st_vfs_file;
2383 fl.fl_pid = current->tgid;
2384
a9933cea 2385 /* vfs_setlease checks to see if delegation should be handed out.
1da177e4
LT		 2386 * the lock_manager callbacks fl_mylease and fl_change are used
2387 */
9167f501 2388 if ((status = vfs_setlease(stp->st_vfs_file, fl.fl_type, &flp))) {
1da177e4 2389 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
c907132d 2390 unhash_delegation(dp);
1da177e4
LT		 2391 flag = NFS4_OPEN_DELEGATE_NONE;
2392 goto out;
2393 }
2394
2395 memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
2396
8c10cbdb
BH		 2397 dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
2398 STATEID_VAL(&dp->dl_stateid));
1da177e4 2399out:
7b190fec
N		 2400 if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS
2401 && flag == NFS4_OPEN_DELEGATE_NONE
2402 && open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE)
2fdada03 2403 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1da177e4
LT		 2404 open->op_delegate_type = flag;
2405}
2406
2407/*
2408 * called with nfs4_lock_state() held.
2409 */
b37ad28b 2410__be32
1da177e4
LT		 2411nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
2412{
6668958f 2413 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1da177e4
LT		 2414 struct nfs4_file *fp = NULL;
2415 struct inode *ino = current_fh->fh_dentry->d_inode;
2416 struct nfs4_stateid *stp = NULL;
567d9829 2417 struct nfs4_delegation *dp = NULL;
b37ad28b 2418 __be32 status;
1da177e4
LT		 2419
2420 status = nfserr_inval;
d87a8ade 2421 if (!access_valid(open->op_share_access, resp->cstate.minorversion)
ba5a6a19 2422 || !deny_valid(open->op_share_deny))
1da177e4
LT		 2423 goto out;
2424 /*
2425 * Lookup file; if found, lookup stateid and check open request,
2426 * and check for delegations in the process of being recalled.
2427 * If not found, create the nfs4_file struct
2428 */
2429 fp = find_file(ino);
2430 if (fp) {
2431 if ((status = nfs4_check_open(fp, open, &stp)))
2432 goto out;
c44c5eeb
N		 2433 status = nfs4_check_deleg(fp, open, &dp);
2434 if (status)
2435 goto out;
1da177e4 2436 } else {
c44c5eeb
N		 2437 status = nfserr_bad_stateid;
2438 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
2439 goto out;
1da177e4
LT		 2440 status = nfserr_resource;
2441 fp = alloc_init_file(ino);
2442 if (fp == NULL)
2443 goto out;
2444 }
2445
2446 /*
2447 * OPEN the file, or upgrade an existing OPEN.
2448 * If truncate fails, the OPEN fails.
2449 */
2450 if (stp) {
2451 /* Stateid was found, this is an OPEN upgrade */
2452 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
2453 if (status)
2454 goto out;
444c2c07 2455 update_stateid(&stp->st_stateid);
1da177e4
LT		 2456 } else {
2457 /* Stateid was not found, this is a new OPEN */
2458 int flags = 0;
9ecb6a08 2459 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
8837abca 2460 flags |= NFSD_MAY_READ;
1da177e4 2461 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
8837abca 2462 flags |= NFSD_MAY_WRITE;
567d9829
N		 2463 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
2464 if (status)
1da177e4
LT		 2465 goto out;
2466 init_stateid(stp, fp, open);
2467 status = nfsd4_truncate(rqstp, current_fh, open);
2468 if (status) {
2283963f 2469 release_open_stateid(stp);
1da177e4
LT		 2470 goto out;
2471 }
6668958f
AA		 2472 if (nfsd4_has_session(&resp->cstate))
2473 update_stateid(&stp->st_stateid);
1da177e4
LT		 2474 }
2475 memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
2476
8b8aae40 2477 if (nfsd4_has_session(&resp->cstate)) {
6668958f 2478 open->op_stateowner->so_confirmed = 1;
8b8aae40
RL		 2479 nfsd4_create_clid_dir(open->op_stateowner->so_client);
2480 }
6668958f 2481
1da177e4
LT		 2482 /*
2483 * Attempt to hand out a delegation. No error return, because the
2484 * OPEN succeeds even if we fail.
2485 */
2486 nfs4_open_delegation(current_fh, open, stp);
2487
2488 status = nfs_ok;
2489
8c10cbdb
BH		 2490 dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
2491 STATEID_VAL(&stp->st_stateid));
1da177e4 2492out:
13cd2184
N		 2493 if (fp)
2494 put_nfs4_file(fp);
37515177
N		 2495 if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
2496 nfs4_set_claim_prev(open);
1da177e4
LT		 2497 /*
2498 * To finish the open response, we just need to set the rflags.
2499 */
2500 open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6668958f
AA		 2501 if (!open->op_stateowner->so_confirmed &&
2502 !nfsd4_has_session(&resp->cstate))
1da177e4
LT		 2503 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
2504
2505 return status;
2506}
2507
b37ad28b 2508__be32
b591480b
BF		 2509nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2510 clientid_t *clid)
1da177e4
LT		 2511{
2512 struct nfs4_client *clp;
b37ad28b 2513 __be32 status;
1da177e4
LT		 2514
2515 nfs4_lock_state();
2516 dprintk("process_renew(%08x/%08x): starting\n",
2517 clid->cl_boot, clid->cl_id);
2518 status = nfserr_stale_clientid;
2519 if (STALE_CLIENTID(clid))
2520 goto out;
2521 clp = find_confirmed_client(clid);
2522 status = nfserr_expired;
2523 if (clp == NULL) {
2524 /* We assume the client took too long to RENEW. */
2525 dprintk("nfsd4_renew: clientid not found!\n");
2526 goto out;
2527 }
2528 renew_client(clp);
2529 status = nfserr_cb_path_down;
ea1da636 2530 if (!list_empty(&clp->cl_delegations)
2bf23875 2531 && !atomic_read(&clp->cl_cb_set))
1da177e4
LT		 2532 goto out;
2533 status = nfs_ok;
2534out:
2535 nfs4_unlock_state();
2536 return status;
2537}
2538
af558e33
BF		 2539struct lock_manager nfsd4_manager = {
2540};
2541
a76b4319 2542static void
af558e33 2543nfsd4_end_grace(void)
a76b4319
N		 2544{
2545 dprintk("NFSD: end of grace period\n");
c7b9a459 2546 nfsd4_recdir_purge_old();
af558e33 2547 locks_end_grace(&nfsd4_manager);
e46b498c
BF		 2548 /*
2549 * Now that every NFSv4 client has had the chance to recover and
2550 * to see the (possibly new, possibly shorter) lease time, we
2551 * can safely set the next grace time to the current lease time:
2552 */
2553 nfsd4_grace = nfsd4_lease;
a76b4319
N		 2554}
2555
fd39ca9a 2556static time_t
1da177e4
LT		 2557nfs4_laundromat(void)
2558{
2559 struct nfs4_client *clp;
2560 struct nfs4_stateowner *sop;
2561 struct nfs4_delegation *dp;
2562 struct list_head *pos, *next, reaplist;
cf07d2ea
BF		 2563 time_t cutoff = get_seconds() - nfsd4_lease;
2564 time_t t, clientid_val = nfsd4_lease;
2565 time_t u, test_val = nfsd4_lease;
1da177e4
LT		 2566
2567 nfs4_lock_state();
2568
2569 dprintk("NFSD: laundromat service - starting\n");
af558e33
BF		 2570 if (locks_in_grace())
2571 nfsd4_end_grace();
1da177e4
LT		 2572 list_for_each_safe(pos, next, &client_lru) {
2573 clp = list_entry(pos, struct nfs4_client, cl_lru);
2574 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2575 t = clp->cl_time - cutoff;
2576 if (clientid_val > t)
2577 clientid_val = t;
2578 break;
2579 }
2580 dprintk("NFSD: purging unused client (clientid %08x)\n",
2581 clp->cl_clientid.cl_id);
c7b9a459 2582 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2583 expire_client(clp);
2584 }
2585 INIT_LIST_HEAD(&reaplist);
2586 spin_lock(&recall_lock);
2587 list_for_each_safe(pos, next, &del_recall_lru) {
2588 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2589 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2590 u = dp->dl_time - cutoff;
2591 if (test_val > u)
2592 test_val = u;
2593 break;
2594 }
2595 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2596 dp, dp->dl_flock);
2597 list_move(&dp->dl_recall_lru, &reaplist);
2598 }
2599 spin_unlock(&recall_lock);
2600 list_for_each_safe(pos, next, &reaplist) {
2601 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2602 list_del_init(&dp->dl_recall_lru);
2603 unhash_delegation(dp);
2604 }
cf07d2ea 2605 test_val = nfsd4_lease;
1da177e4
LT		 2606 list_for_each_safe(pos, next, &close_lru) {
2607 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2608 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2609 u = sop->so_time - cutoff;
2610 if (test_val > u)
2611 test_val = u;
2612 break;
2613 }
2614 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2615 sop->so_id);
f044ff83 2616 release_openowner(sop);
1da177e4
LT		 2617 }
2618 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2619 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2620 nfs4_unlock_state();
2621 return clientid_val;
2622}
2623
a254b246
HH		 2624static struct workqueue_struct *laundry_wq;
2625static void laundromat_main(struct work_struct *);
2626static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2627
2628static void
c4028958 2629laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2630{
2631 time_t t;
2632
2633 t = nfs4_laundromat();
2634 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2635 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2636}
2637
fd39ca9a 2638static struct nfs4_stateowner *
f8816512
N		 2639search_close_lru(u32 st_id, int flags)
2640{
1da177e4
LT		 2641 struct nfs4_stateowner *local = NULL;
2642
1da177e4
LT		 2643 if (flags & CLOSE_STATE) {
2644 list_for_each_entry(local, &close_lru, so_close_lru) {
2645 if (local->so_id == st_id)
2646 return local;
2647 }
2648 }
2649 return NULL;
2650}
2651
2652static inline int
2653nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2654{
7eaa36e2 2655 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2656}
2657
2658static int
2659STALE_STATEID(stateid_t *stateid)
2660{
78155ed7
BN		 2661 if (time_after((unsigned long)boot_time,
2662 (unsigned long)stateid->si_boot)) {
8c10cbdb
BH		 2663 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2664 STATEID_VAL(stateid));
78155ed7
BN		 2665 return 1;
2666 }
2667 return 0;
2668}
2669
2670static int
2671EXPIRED_STATEID(stateid_t *stateid)
2672{
2673 if (time_before((unsigned long)boot_time,
2674 ((unsigned long)stateid->si_boot)) &&
cf07d2ea 2675 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
8c10cbdb
BH		 2676 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2677 STATEID_VAL(stateid));
78155ed7
BN		 2678 return 1;
2679 }
2680 return 0;
2681}
2682
2683static __be32
2684stateid_error_map(stateid_t *stateid)
2685{
2686 if (STALE_STATEID(stateid))
2687 return nfserr_stale_stateid;
2688 if (EXPIRED_STATEID(stateid))
2689 return nfserr_expired;
2690
8c10cbdb
BH		 2691 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2692 STATEID_VAL(stateid));
78155ed7 2693 return nfserr_bad_stateid;
1da177e4
LT		 2694}
2695
2696static inline int
2697access_permit_read(unsigned long access_bmap)
2698{
2699 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2700 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2701 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2702}
2703
2704static inline int
2705access_permit_write(unsigned long access_bmap)
2706{
2707 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2708 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2709}
2710
2711static
b37ad28b 2712__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2713{
b37ad28b 2714 __be32 status = nfserr_openmode;
1da177e4
LT		 2715
2716 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2717 goto out;
2718 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2719 goto out;
2720 status = nfs_ok;
2721out:
2722 return status;
2723}
2724
b37ad28b 2725static inline __be32
1da177e4
LT		 2726check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2727{
203a8c8e 2728 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2729 return nfs_ok;
af558e33 2730 else if (locks_in_grace()) {
1da177e4
LT		 2731 /* Answer in remaining cases depends on existance of
2732 * conflicting state; so we must wait out the grace period. */
2733 return nfserr_grace;
2734 } else if (flags & WR_STATE)
2735 return nfs4_share_conflict(current_fh,
2736 NFS4_SHARE_DENY_WRITE);
2737 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2738 return nfs4_share_conflict(current_fh,
2739 NFS4_SHARE_DENY_READ);
2740}
2741
2742/*
2743 * Allow READ/WRITE during grace period on recovered state only for files
2744 * that are not able to provide mandatory locking.
2745 */
2746static inline int
18f82731 2747grace_disallows_io(struct inode *inode)
1da177e4 2748{
203a8c8e 2749 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2750}
2751
6668958f 2752static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2753{
6668958f
AA		 2754 /*
2755 * When sessions are used the stateid generation number is ignored
2756 * when it is zero.
2757 */
2758 if ((flags & HAS_SESSION) && in->si_generation == 0)
2759 goto out;
2760
0836f587
BF		 2761 /* If the client sends us a stateid from the future, it's buggy: */
2762 if (in->si_generation > ref->si_generation)
2763 return nfserr_bad_stateid;
2764 /*
2765 * The following, however, can happen. For example, if the
2766 * client sends an open and some IO at the same time, the open
2767 * may bump si_generation while the IO is still in flight.
2768 * Thanks to hard links and renames, the client never knows what
2769 * file an open will affect. So it could avoid that situation
2770 * only by serializing all opens and IO from the same open
2771 * owner. To recover from the old_stateid error, the client
2772 * will just have to retry the IO:
2773 */
2774 if (in->si_generation < ref->si_generation)
2775 return nfserr_old_stateid;
6668958f 2776out:
0836f587
BF		 2777 return nfs_ok;
2778}
2779
3e633079
BF		 2780static int is_delegation_stateid(stateid_t *stateid)
2781{
2782 return stateid->si_fileid == 0;
2783}
2784
1da177e4
LT		 2785/*
2786* Checks for stateid operations
2787*/
b37ad28b 2788__be32
dd453dfd
BH		 2789nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2790 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2791{
2792 struct nfs4_stateid *stp = NULL;
2793 struct nfs4_delegation *dp = NULL;
dd453dfd 2794 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2795 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2796 __be32 status;
1da177e4 2797
1da177e4
LT		 2798 if (filpp)
2799 *filpp = NULL;
2800
18f82731 2801 if (grace_disallows_io(ino))
1da177e4
LT		 2802 return nfserr_grace;
2803
6668958f
AA		 2804 if (nfsd4_has_session(cstate))
2805 flags |= HAS_SESSION;
2806
1da177e4
LT		 2807 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2808 return check_special_stateids(current_fh, stateid, flags);
2809
1da177e4
LT		 2810 status = nfserr_stale_stateid;
2811 if (STALE_STATEID(stateid))
2812 goto out;
2813
1da177e4 2814 status = nfserr_bad_stateid;
3e633079 2815 if (is_delegation_stateid(stateid)) {
a4455be0 2816 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2817 if (!dp) {
2818 status = stateid_error_map(stateid);
1da177e4 2819 goto out;
78155ed7 2820 }
6668958f
AA		 2821 status = check_stateid_generation(stateid, &dp->dl_stateid,
2822 flags);
0c2a498f
BF		 2823 if (status)
2824 goto out;
dc9bf700
BF		 2825 status = nfs4_check_delegmode(dp, flags);
2826 if (status)
2827 goto out;
2828 renew_client(dp->dl_client);
dc9bf700
BF		 2829 if (filpp)
2830 *filpp = dp->dl_vfs_file;
1da177e4 2831 } else { /* open or lock stateid */
a4455be0 2832 stp = find_stateid(stateid, flags);
78155ed7
BN		 2833 if (!stp) {
2834 status = stateid_error_map(stateid);
1da177e4 2835 goto out;
78155ed7 2836 }
6150ef0d 2837 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2838 goto out;
2839 if (!stp->st_stateowner->so_confirmed)
2840 goto out;
6668958f
AA		 2841 status = check_stateid_generation(stateid, &stp->st_stateid,
2842 flags);
0c2a498f
BF		 2843 if (status)
2844 goto out;
a4455be0
BF		 2845 status = nfs4_check_openmode(stp, flags);
2846 if (status)
1da177e4
LT		 2847 goto out;
2848 renew_client(stp->st_stateowner->so_client);
2849 if (filpp)
2850 *filpp = stp->st_vfs_file;
1da177e4
LT		 2851 }
2852 status = nfs_ok;
2853out:
2854 return status;
2855}
2856
4c4cd222
N		 2857static inline int
2858setlkflg (int type)
2859{
2860 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2861 RD_STATE : WR_STATE;
2862}
1da177e4
LT		 2863
2864/*
2865 * Checks for sequence id mutating operations.
2866 */
b37ad28b 2867static __be32
dd453dfd
BH		 2868nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2869 stateid_t *stateid, int flags,
2870 struct nfs4_stateowner **sopp,
2871 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2872{
1da177e4
LT		 2873 struct nfs4_stateid *stp;
2874 struct nfs4_stateowner *sop;
dd453dfd 2875 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2876 __be32 status;
1da177e4 2877
8c10cbdb
BH		 2878 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2879 seqid, STATEID_VAL(stateid));
3a4f98bb 2880
1da177e4
LT		 2881 *stpp = NULL;
2882 *sopp = NULL;
2883
1da177e4 2884 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2885 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2886 return nfserr_bad_stateid;
1da177e4
LT		 2887 }
2888
1da177e4 2889 if (STALE_STATEID(stateid))
3a4f98bb 2890 return nfserr_stale_stateid;
6668958f
AA		 2891
2892 if (nfsd4_has_session(cstate))
2893 flags |= HAS_SESSION;
2894
1da177e4
LT		 2895 /*
2896 * We return BAD_STATEID if filehandle doesn't match stateid,
2897 * the confirmed flag is incorrecly set, or the generation
2898 * number is incorrect.
1da177e4 2899 */
f8816512
N		 2900 stp = find_stateid(stateid, flags);
2901 if (stp == NULL) {
2902 /*
2903 * Also, we should make sure this isn't just the result of
2904 * a replayed close:
2905 */
2906 sop = search_close_lru(stateid->si_stateownerid, flags);
2907 if (sop == NULL)
78155ed7 2908 return stateid_error_map(stateid);
f8816512
N		 2909 *sopp = sop;
2910 goto check_replay;
2911 }
1da177e4 2912
39325bd0
BF		 2913 *stpp = stp;
2914 *sopp = sop = stp->st_stateowner;
2915
4c4cd222 2916 if (lock) {
4c4cd222 2917 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2918 struct nfs4_client *clp = sop->so_client;
4c4cd222 2919 int lkflg = 0;
b37ad28b 2920 __be32 status;
4c4cd222
N		 2921
2922 lkflg = setlkflg(lock->lk_type);
2923
2924 if (lock->lk_is_new) {
599e0a22
BF		 2925 if (!sop->so_is_open_owner)
2926 return nfserr_bad_stateid;
60adfc50
AA		 2927 if (!(flags & HAS_SESSION) &&
2928 !same_clid(&clp->cl_clientid, lockclid))
2929 return nfserr_bad_stateid;
599e0a22
BF		 2930 /* stp is the open stateid */
2931 status = nfs4_check_openmode(stp, lkflg);
2932 if (status)
2933 return status;
2934 } else {
2935 /* stp is the lock stateid */
2936 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2937 if (status)
2938 return status;
4c4cd222 2939 }
1da177e4
LT		 2940 }
2941
f3362737 2942 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2943 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2944 return nfserr_bad_stateid;
1da177e4
LT		 2945 }
2946
1da177e4
LT		 2947 /*
2948 * We now validate the seqid and stateid generation numbers.
2949 * For the moment, we ignore the possibility of
2950 * generation number wraparound.
2951 */
6668958f 2952 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 2953 goto check_replay;
2954
3a4f98bb 2955 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2956 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2957 " unconfirmed stateowner!\n");
2958 return nfserr_bad_stateid;
1da177e4 2959 }
3a4f98bb 2960 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2961 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2962 " confirmed yet!\n");
2963 return nfserr_bad_stateid;
1da177e4 2964 }
6668958f 2965 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 2966 if (status)
2967 return status;
52fd004e 2968 renew_client(sop->so_client);
3a4f98bb 2969 return nfs_ok;
1da177e4 2970
1da177e4 2971check_replay:
bd9aac52 2972 if (seqid == sop->so_seqid - 1) {
849823c5 2973 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2974 /* indicate replay to calling function */
a90b061c 2975 return nfserr_replay_me;
1da177e4 2976 }
2fdada03 2977 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2978 sop->so_seqid, seqid);
2979 *sopp = NULL;
2980 return nfserr_bad_seqid;
1da177e4
LT		 2981}
2982
b37ad28b 2983__be32
ca364317 2984nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2985 struct nfsd4_open_confirm *oc)
1da177e4 2986{
b37ad28b 2987 __be32 status;
1da177e4
LT		 2988 struct nfs4_stateowner *sop;
2989 struct nfs4_stateid *stp;
2990
2991 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 2992 (int)cstate->current_fh.fh_dentry->d_name.len,
2993 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 2994
ca364317 2995 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 2996 if (status)
2997 return status;
1da177e4
LT		 2998
2999 nfs4_lock_state();
3000
dd453dfd 3001 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3002 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3003 CONFIRM | OPEN_STATE,
1da177e4
LT		 3004 &oc->oc_stateowner, &stp, NULL)))
3005 goto out;
3006
3007 sop = oc->oc_stateowner;
3008 sop->so_confirmed = 1;
3009 update_stateid(&stp->st_stateid);
3010 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
8c10cbdb
BH		 3011 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3012 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
c7b9a459
N		 3013
3014 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3015out:
f2327d9a 3016 if (oc->oc_stateowner) {
1da177e4 3017 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3018 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3019 }
1da177e4
LT		 3020 nfs4_unlock_state();
3021 return status;
3022}
3023
3024
3025/*
3026 * unset all bits in union bitmap (bmap) that
3027 * do not exist in share (from successful OPEN_DOWNGRADE)
3028 */
3029static void
3030reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3031{
3032 int i;
3033 for (i = 1; i < 4; i++) {
3034 if ((i & access) != i)
3035 __clear_bit(i, bmap);
3036 }
3037}
3038
3039static void
3040reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3041{
3042 int i;
3043 for (i = 0; i < 4; i++) {
3044 if ((i & deny) != i)
3045 __clear_bit(i, bmap);
3046 }
3047}
3048
b37ad28b 3049__be32
ca364317
BF		 3050nfsd4_open_downgrade(struct svc_rqst *rqstp,
3051 struct nfsd4_compound_state *cstate,
a4f1706a 3052 struct nfsd4_open_downgrade *od)
1da177e4 3053{
b37ad28b 3054 __be32 status;
1da177e4
LT		 3055 struct nfs4_stateid *stp;
3056 unsigned int share_access;
3057
3058 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3059 (int)cstate->current_fh.fh_dentry->d_name.len,
3060 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3061
d87a8ade 3062 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3063 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3064 return nfserr_inval;
3065
3066 nfs4_lock_state();
dd453dfd 3067 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3068 od->od_seqid,
1da177e4 3069 &od->od_stateid,
f3362737 3070 OPEN_STATE,
1da177e4
LT		 3071 &od->od_stateowner, &stp, NULL)))
3072 goto out;
3073
3074 status = nfserr_inval;
3075 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3076 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3077 stp->st_access_bmap, od->od_share_access);
3078 goto out;
3079 }
3080 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3081 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3082 stp->st_deny_bmap, od->od_share_deny);
3083 goto out;
3084 }
3085 set_access(&share_access, stp->st_access_bmap);
3086 nfs4_file_downgrade(stp->st_vfs_file,
3087 share_access & ~od->od_share_access);
3088
3089 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3090 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3091
3092 update_stateid(&stp->st_stateid);
3093 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3094 status = nfs_ok;
3095out:
f2327d9a 3096 if (od->od_stateowner) {
1da177e4 3097 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3098 cstate->replay_owner = od->od_stateowner;
f2327d9a 3099 }
1da177e4
LT		 3100 nfs4_unlock_state();
3101 return status;
3102}
3103
3104/*
3105 * nfs4_unlock_state() called after encode
3106 */
b37ad28b 3107__be32
ca364317 3108nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3109 struct nfsd4_close *close)
1da177e4 3110{
b37ad28b 3111 __be32 status;
1da177e4
LT		 3112 struct nfs4_stateid *stp;
3113
3114 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3115 (int)cstate->current_fh.fh_dentry->d_name.len,
3116 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3117
3118 nfs4_lock_state();
3119 /* check close_lru for replay */
dd453dfd 3120 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3121 close->cl_seqid,
1da177e4 3122 &close->cl_stateid,
f3362737 3123 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3124 &close->cl_stateowner, &stp, NULL)))
3125 goto out;
1da177e4
LT		 3126 status = nfs_ok;
3127 update_stateid(&stp->st_stateid);
3128 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3129
04ef5954 3130 /* release_stateid() calls nfsd_close() if needed */
2283963f 3131 release_open_stateid(stp);
04ef5954
BF		 3132
3133 /* place unused nfs4_stateowners on so_close_lru list to be
3134 * released by the laundromat service after the lease period
3135 * to enable us to handle CLOSE replay
3136 */
3137 if (list_empty(&close->cl_stateowner->so_stateids))
3138 move_to_close_lru(close->cl_stateowner);
1da177e4 3139out:
f2327d9a 3140 if (close->cl_stateowner) {
1da177e4 3141 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3142 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3143 }
1da177e4
LT		 3144 nfs4_unlock_state();
3145 return status;
3146}
3147
b37ad28b 3148__be32
ca364317
BF		 3149nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3150 struct nfsd4_delegreturn *dr)
1da177e4 3151{
203a8c8e
BF		 3152 struct nfs4_delegation *dp;
3153 stateid_t *stateid = &dr->dr_stateid;
3154 struct inode *inode;
b37ad28b 3155 __be32 status;
6668958f 3156 int flags = 0;
1da177e4 3157
ca364317 3158 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3159 return status;
3160 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3161
6668958f
AA		 3162 if (nfsd4_has_session(cstate))
3163 flags |= HAS_SESSION;
1da177e4 3164 nfs4_lock_state();
203a8c8e
BF		 3165 status = nfserr_bad_stateid;
3166 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3167 goto out;
3168 status = nfserr_stale_stateid;
3169 if (STALE_STATEID(stateid))
3170 goto out;
7e0f7cf5 3171 status = nfserr_bad_stateid;
203a8c8e
BF		 3172 if (!is_delegation_stateid(stateid))
3173 goto out;
203a8c8e 3174 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3175 if (!dp) {
3176 status = stateid_error_map(stateid);
203a8c8e 3177 goto out;
78155ed7 3178 }
6668958f 3179 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3180 if (status)
3181 goto out;
3182 renew_client(dp->dl_client);
3183
3184 unhash_delegation(dp);
1da177e4 3185out:
203a8c8e
BF		 3186 nfs4_unlock_state();
3187
1da177e4
LT		 3188 return status;
3189}
3190
3191
3192/*
3193 * Lock owner state (byte-range locks)
3194 */
3195#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3196#define LOCK_HASH_BITS 8
3197#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3198#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3199
87df4de8
BH		 3200static inline u64
3201end_offset(u64 start, u64 len)
3202{
3203 u64 end;
3204
3205 end = start + len;
3206 return end >= start ? end: NFS4_MAX_UINT64;
3207}
3208
3209/* last octet in a range */
3210static inline u64
3211last_byte_offset(u64 start, u64 len)
3212{
3213 u64 end;
3214
3215 BUG_ON(!len);
3216 end = start + len;
3217 return end > start ? end - 1: NFS4_MAX_UINT64;
3218}
3219
1da177e4
LT		 3220#define lockownerid_hashval(id) \
3221 ((id) & LOCK_HASH_MASK)
3222
3223static inline unsigned int
3224lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3225 struct xdr_netobj *ownername)
3226{
3227 return (file_hashval(inode) + cl_id
3228 + opaque_hashval(ownername->data, ownername->len))
3229 & LOCK_HASH_MASK;
3230}
3231
3232static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3233static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3234static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3235
fd39ca9a 3236static struct nfs4_stateid *
1da177e4
LT		 3237find_stateid(stateid_t *stid, int flags)
3238{
9346eff0 3239 struct nfs4_stateid *local;
1da177e4
LT		 3240 u32 st_id = stid->si_stateownerid;
3241 u32 f_id = stid->si_fileid;
3242 unsigned int hashval;
3243
3244 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3245 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3246 hashval = stateid_hashval(st_id, f_id);
3247 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3248 if ((local->st_stateid.si_stateownerid == st_id) &&
3249 (local->st_stateid.si_fileid == f_id))
3250 return local;
3251 }
3252 }
9346eff0
KK		 3253
3254 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3255 hashval = stateid_hashval(st_id, f_id);
3256 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3257 if ((local->st_stateid.si_stateownerid == st_id) &&
3258 (local->st_stateid.si_fileid == f_id))
3259 return local;
3260 }
849823c5 3261 }
1da177e4
LT		 3262 return NULL;
3263}
3264
3265static struct nfs4_delegation *
3266find_delegation_stateid(struct inode *ino, stateid_t *stid)
3267{
13cd2184
N		 3268 struct nfs4_file *fp;
3269 struct nfs4_delegation *dl;
1da177e4 3270
8c10cbdb
BH		 3271 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3272 STATEID_VAL(stid));
1da177e4 3273
1da177e4 3274 fp = find_file(ino);
13cd2184
N		 3275 if (!fp)
3276 return NULL;
3277 dl = find_delegation_file(fp, stid);
3278 put_nfs4_file(fp);
3279 return dl;
1da177e4
LT		 3280}
3281
3282/*
3283 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3284 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3285 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3286 * locking, this prevents us from being completely protocol-compliant. The
3287 * real solution to this problem is to start using unsigned file offsets in
3288 * the VFS, but this is a very deep change!
3289 */
3290static inline void
3291nfs4_transform_lock_offset(struct file_lock *lock)
3292{
3293 if (lock->fl_start < 0)
3294 lock->fl_start = OFFSET_MAX;
3295 if (lock->fl_end < 0)
3296 lock->fl_end = OFFSET_MAX;
3297}
3298
d5b9026a
N		 3299/* Hack!: For now, we're defining this just so we can use a pointer to it
3300 * as a unique cookie to identify our (NFSv4's) posix locks. */
7b021967 3301static const struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3302};
1da177e4
LT		 3303
3304static inline void
3305nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3306{
d5b9026a
N		 3307 struct nfs4_stateowner *sop;
3308 unsigned int hval;
1da177e4 3309
d5b9026a
N		 3310 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3311 sop = (struct nfs4_stateowner *) fl->fl_owner;
3312 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3313 kref_get(&sop->so_ref);
3314 deny->ld_sop = sop;
3315 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3316 } else {
3317 deny->ld_sop = NULL;
3318 deny->ld_clientid.cl_boot = 0;
3319 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3320 }
3321 deny->ld_start = fl->fl_start;
87df4de8
BH		 3322 deny->ld_length = NFS4_MAX_UINT64;
3323 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3324 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3325 deny->ld_type = NFS4_READ_LT;
3326 if (fl->fl_type != F_RDLCK)
3327 deny->ld_type = NFS4_WRITE_LT;
3328}
3329
1da177e4
LT		 3330static struct nfs4_stateowner *
3331find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3332 struct xdr_netobj *owner)
3333{
3334 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3335 struct nfs4_stateowner *op;
3336
3337 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3338 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3339 return op;
3340 }
3341 return NULL;
3342}
3343
3344/*
3345 * Alloc a lock owner structure.
3346 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3347 * occured.
3348 *
3349 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3350 */
3351
3352static struct nfs4_stateowner *
3353alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3354 struct nfs4_stateowner *sop;
3355 struct nfs4_replay *rp;
3356 unsigned int idhashval;
3357
3358 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3359 return NULL;
3360 idhashval = lockownerid_hashval(current_ownerid);
3361 INIT_LIST_HEAD(&sop->so_idhash);
3362 INIT_LIST_HEAD(&sop->so_strhash);
3363 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3364 INIT_LIST_HEAD(&sop->so_stateids);
3365 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3366 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3367 sop->so_time = 0;
3368 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3369 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3370 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3371 sop->so_is_open_owner = 0;
3372 sop->so_id = current_ownerid++;
3373 sop->so_client = clp;
b59e3c0e
NB		 3374 /* It is the openowner seqid that will be incremented in encode in the
3375 * case of new lockowners; so increment the lock seqid manually: */
3376 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3377 sop->so_confirmed = 1;
3378 rp = &sop->so_replay;
de1ae286 3379 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3380 rp->rp_buflen = 0;
3381 rp->rp_buf = rp->rp_ibuf;
3382 return sop;
3383}
3384
fd39ca9a 3385static struct nfs4_stateid *
1da177e4
LT		 3386alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3387{
3388 struct nfs4_stateid *stp;
3389 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3390
5ac049ac
N		 3391 stp = nfs4_alloc_stateid();
3392 if (stp == NULL)
1da177e4
LT		 3393 goto out;
3394 INIT_LIST_HEAD(&stp->st_hash);
3395 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3396 INIT_LIST_HEAD(&stp->st_perstateowner);
3397 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3398 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3399 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3400 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3401 stp->st_stateowner = sop;
13cd2184 3402 get_nfs4_file(fp);
1da177e4 3403 stp->st_file = fp;
78155ed7 3404 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3405 stp->st_stateid.si_stateownerid = sop->so_id;
3406 stp->st_stateid.si_fileid = fp->fi_id;
3407 stp->st_stateid.si_generation = 0;
3408 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3409 stp->st_access_bmap = open_stp->st_access_bmap;
3410 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3411 stp->st_openstp = open_stp;
1da177e4
LT		 3412
3413out:
3414 return stp;
3415}
3416
fd39ca9a 3417static int
1da177e4
LT		 3418check_lock_length(u64 offset, u64 length)
3419{
87df4de8 3420 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3421 LOFF_OVERFLOW(offset, length)));
3422}
3423
3424/*
3425 * LOCK operation
3426 */
b37ad28b 3427__be32
ca364317 3428nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3429 struct nfsd4_lock *lock)
1da177e4 3430{
3e9e3dbe 3431 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3432 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3433 struct nfs4_stateid *lock_stp;
3434 struct file *filp;
3435 struct file_lock file_lock;
8dc7c311 3436 struct file_lock conflock;
b37ad28b 3437 __be32 status = 0;
1da177e4 3438 unsigned int strhashval;
150b3934 3439 unsigned int cmd;
b8dd7b9a 3440 int err;
1da177e4
LT		 3441
3442 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3443 (long long) lock->lk_offset,
3444 (long long) lock->lk_length);
3445
1da177e4
LT		 3446 if (check_lock_length(lock->lk_offset, lock->lk_length))
3447 return nfserr_inval;
3448
ca364317 3449 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3450 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3451 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3452 return status;
3453 }
3454
1da177e4
LT		 3455 nfs4_lock_state();
3456
3457 if (lock->lk_is_new) {
893f8770
N		 3458 /*
3459 * Client indicates that this is a new lockowner.
3460 * Use open owner and open stateid to create lock owner and
3461 * lock stateid.
3462 */
1da177e4
LT		 3463 struct nfs4_stateid *open_stp = NULL;
3464 struct nfs4_file *fp;
3465
3466 status = nfserr_stale_clientid;
60adfc50
AA		 3467 if (!nfsd4_has_session(cstate) &&
3468 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3469 goto out;
1da177e4 3470
1da177e4 3471 /* validate and update open stateid and open seqid */
dd453dfd 3472 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3473 lock->lk_new_open_seqid,
3474 &lock->lk_new_open_stateid,
f3362737 3475 OPEN_STATE,
3a65588a 3476 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3477 lock);
37515177 3478 if (status)
1da177e4 3479 goto out;
3a65588a 3480 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3481 /* create lockowner and lock stateid */
3482 fp = open_stp->st_file;
3483 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3484 open_sop->so_client->cl_clientid.cl_id,
3485 &lock->v.new.owner);
3e9e3dbe
N		 3486 /* XXX: Do we need to check for duplicate stateowners on
3487 * the same file, or should they just be allowed (and
3488 * create new stateids)? */
1da177e4 3489 status = nfserr_resource;
b59e3c0e
NB		 3490 lock_sop = alloc_init_lock_stateowner(strhashval,
3491 open_sop->so_client, open_stp, lock);
3492 if (lock_sop == NULL)
1da177e4 3493 goto out;
b59e3c0e 3494 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3495 if (lock_stp == NULL)
1da177e4 3496 goto out;
1da177e4
LT		 3497 } else {
3498 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3499 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3500 lock->lk_old_lock_seqid,
3501 &lock->lk_old_lock_stateid,
f3362737 3502 LOCK_STATE,
3a65588a 3503 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3504 if (status)
3505 goto out;
3a65588a 3506 lock_sop = lock->lk_replay_owner;
1da177e4 3507 }
3a65588a 3508 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3509 filp = lock_stp->st_vfs_file;
3510
0dd395dc 3511 status = nfserr_grace;
af558e33 3512 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3513 goto out;
3514 status = nfserr_no_grace;
af558e33 3515 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3516 goto out;
3517
1da177e4
LT		 3518 locks_init_lock(&file_lock);
3519 switch (lock->lk_type) {
3520 case NFS4_READ_LT:
3521 case NFS4_READW_LT:
3522 file_lock.fl_type = F_RDLCK;
150b3934 3523 cmd = F_SETLK;
1da177e4
LT		 3524 break;
3525 case NFS4_WRITE_LT:
3526 case NFS4_WRITEW_LT:
3527 file_lock.fl_type = F_WRLCK;
150b3934 3528 cmd = F_SETLK;
1da177e4
LT		 3529 break;
3530 default:
3531 status = nfserr_inval;
3532 goto out;
3533 }
b59e3c0e 3534 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3535 file_lock.fl_pid = current->tgid;
3536 file_lock.fl_file = filp;
3537 file_lock.fl_flags = FL_POSIX;
d5b9026a 3538 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3539
3540 file_lock.fl_start = lock->lk_offset;
87df4de8 3541 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3542 nfs4_transform_lock_offset(&file_lock);
3543
3544 /*
3545 * Try to lock the file in the VFS.
3546 * Note: locks.c uses the BKL to protect the inode's lock list.
3547 */
3548
fd85b817 3549 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3550 switch (-err) {
1da177e4
LT		 3551 case 0: /* success! */
3552 update_stateid(&lock_stp->st_stateid);
3553 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3554 sizeof(stateid_t));
b8dd7b9a 3555 status = 0;
eb76b3fd
AA		 3556 break;
3557 case (EAGAIN): /* conflock holds conflicting lock */
3558 status = nfserr_denied;
3559 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3560 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3561 break;
1da177e4
LT		 3562 case (EDEADLK):
3563 status = nfserr_deadlock;
eb76b3fd 3564 break;
1da177e4 3565 default:
fd85b817 3566 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3567 status = nfserr_resource;
3568 break;
1da177e4 3569 }
1da177e4 3570out:
8a280510 3571 if (status && lock->lk_is_new && lock_sop)
f044ff83 3572 release_lockowner(lock_sop);
3a65588a
BF		 3573 if (lock->lk_replay_owner) {
3574 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3575 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3576 }
1da177e4
LT		 3577 nfs4_unlock_state();
3578 return status;
3579}
3580
55ef1274
BF		 3581/*
3582 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3583 * so we do a temporary open here just to get an open file to pass to
3584 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3585 * inode operation.)
3586 */
3587static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3588{
3589 struct file *file;
3590 int err;
3591
3592 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3593 if (err)
3594 return err;
3595 err = vfs_test_lock(file, lock);
3596 nfsd_close(file);
3597 return err;
3598}
3599
1da177e4
LT		 3600/*
3601 * LOCKT operation
3602 */
b37ad28b 3603__be32
ca364317
BF		 3604nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3605 struct nfsd4_lockt *lockt)
1da177e4
LT		 3606{
3607 struct inode *inode;
1da177e4 3608 struct file_lock file_lock;
fd85b817 3609 int error;
b37ad28b 3610 __be32 status;
1da177e4 3611
af558e33 3612 if (locks_in_grace())
1da177e4
LT		 3613 return nfserr_grace;
3614
3615 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3616 return nfserr_inval;
3617
3618 lockt->lt_stateowner = NULL;
3619 nfs4_lock_state();
3620
3621 status = nfserr_stale_clientid;
60adfc50 3622 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3623 goto out;
1da177e4 3624
ca364317 3625 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3626 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3627 if (status == nfserr_symlink)
3628 status = nfserr_inval;
3629 goto out;
3630 }
3631
ca364317 3632 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3633 locks_init_lock(&file_lock);
3634 switch (lockt->lt_type) {
3635 case NFS4_READ_LT:
3636 case NFS4_READW_LT:
3637 file_lock.fl_type = F_RDLCK;
3638 break;
3639 case NFS4_WRITE_LT:
3640 case NFS4_WRITEW_LT:
3641 file_lock.fl_type = F_WRLCK;
3642 break;
3643 default:
2fdada03 3644 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3645 status = nfserr_inval;
3646 goto out;
3647 }
3648
3649 lockt->lt_stateowner = find_lockstateowner_str(inode,
3650 &lockt->lt_clientid, &lockt->lt_owner);
3651 if (lockt->lt_stateowner)
3652 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3653 file_lock.fl_pid = current->tgid;
3654 file_lock.fl_flags = FL_POSIX;
3655
3656 file_lock.fl_start = lockt->lt_offset;
87df4de8 3657 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3658
3659 nfs4_transform_lock_offset(&file_lock);
3660
1da177e4 3661 status = nfs_ok;
55ef1274 3662 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3663 if (error) {
3664 status = nfserrno(error);
3665 goto out;
3666 }
9d6a8c5c 3667 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3668 status = nfserr_denied;
9d6a8c5c 3669 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3670 }
3671out:
3672 nfs4_unlock_state();
3673 return status;
3674}
3675
b37ad28b 3676__be32
ca364317 3677nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3678 struct nfsd4_locku *locku)
1da177e4
LT		 3679{
3680 struct nfs4_stateid *stp;
3681 struct file *filp = NULL;
3682 struct file_lock file_lock;
b37ad28b 3683 __be32 status;
b8dd7b9a 3684 int err;
1da177e4
LT		 3685
3686 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3687 (long long) locku->lu_offset,
3688 (long long) locku->lu_length);
3689
3690 if (check_lock_length(locku->lu_offset, locku->lu_length))
3691 return nfserr_inval;
3692
3693 nfs4_lock_state();
3694
dd453dfd 3695 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3696 locku->lu_seqid,
3697 &locku->lu_stateid,
f3362737 3698 LOCK_STATE,
1da177e4
LT		 3699 &locku->lu_stateowner, &stp, NULL)))
3700 goto out;
3701
3702 filp = stp->st_vfs_file;
3703 BUG_ON(!filp);
3704 locks_init_lock(&file_lock);
3705 file_lock.fl_type = F_UNLCK;
3706 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3707 file_lock.fl_pid = current->tgid;
3708 file_lock.fl_file = filp;
3709 file_lock.fl_flags = FL_POSIX;
d5b9026a 3710 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3711 file_lock.fl_start = locku->lu_offset;
3712
87df4de8 3713 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3714 nfs4_transform_lock_offset(&file_lock);
3715
3716 /*
3717 * Try to unlock the file in the VFS.
3718 */
fd85b817 3719 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3720 if (err) {
fd85b817 3721 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3722 goto out_nfserr;
3723 }
3724 /*
3725 * OK, unlock succeeded; the only thing left to do is update the stateid.
3726 */
3727 update_stateid(&stp->st_stateid);
3728 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3729
3730out:
f2327d9a 3731 if (locku->lu_stateowner) {
1da177e4 3732 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3733 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3734 }
1da177e4
LT		 3735 nfs4_unlock_state();
3736 return status;
3737
3738out_nfserr:
b8dd7b9a 3739 status = nfserrno(err);
1da177e4
LT		 3740 goto out;
3741}
3742
3743/*
3744 * returns
3745 * 1: locks held by lockowner
3746 * 0: no locks held by lockowner
3747 */
3748static int
3749check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3750{
3751 struct file_lock **flpp;
7eaa36e2 3752 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3753 int status = 0;
3754
3755 lock_kernel();
3756 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3757 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3758 status = 1;
3759 goto out;
796dadfd 3760 }
1da177e4
LT		 3761 }
3762out:
3763 unlock_kernel();
3764 return status;
3765}
3766
b37ad28b 3767__be32
b591480b
BF		 3768nfsd4_release_lockowner(struct svc_rqst *rqstp,
3769 struct nfsd4_compound_state *cstate,
3770 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3771{
3772 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3773 struct nfs4_stateowner *sop;
3774 struct nfs4_stateid *stp;
1da177e4 3775 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3776 struct list_head matches;
3777 int i;
b37ad28b 3778 __be32 status;
1da177e4
LT		 3779
3780 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3781 clid->cl_boot, clid->cl_id);
3782
3783 /* XXX check for lease expiration */
3784
3785 status = nfserr_stale_clientid;
849823c5 3786 if (STALE_CLIENTID(clid))
1da177e4 3787 return status;
1da177e4
LT		 3788
3789 nfs4_lock_state();
3790
3e9e3dbe
N		 3791 status = nfserr_locks_held;
3792 /* XXX: we're doing a linear search through all the lockowners.
3793 * Yipes! For now we'll just hope clients aren't really using
3794 * release_lockowner much, but eventually we have to fix these
3795 * data structures. */
3796 INIT_LIST_HEAD(&matches);
3797 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3798 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3799 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3800 continue;
3801 list_for_each_entry(stp, &sop->so_stateids,
3802 st_perstateowner) {
3803 if (check_for_locks(stp->st_vfs_file, sop))
3804 goto out;
3805 /* Note: so_perclient unused for lockowners,
3806 * so it's OK to fool with here. */
3807 list_add(&sop->so_perclient, &matches);
3808 }
1da177e4 3809 }
3e9e3dbe
N		 3810 }
3811 /* Clients probably won't expect us to return with some (but not all)
3812 * of the lockowner state released; so don't release any until all
3813 * have been checked. */
3814 status = nfs_ok;
0fa822e4
N		 3815 while (!list_empty(&matches)) {
3816 sop = list_entry(matches.next, struct nfs4_stateowner,
3817 so_perclient);
3818 /* unhash_stateowner deletes so_perclient only
3819 * for openowners. */
3820 list_del(&sop->so_perclient);
f044ff83 3821 release_lockowner(sop);
1da177e4
LT		 3822 }
3823out:
3824 nfs4_unlock_state();
3825 return status;
3826}
3827
3828static inline struct nfs4_client_reclaim *
a55370a3 3829alloc_reclaim(void)
1da177e4 3830{
a55370a3 3831 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3832}
3833
c7b9a459 3834int
a1bcecd2 3835nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3836{
3837 unsigned int strhashval = clientstr_hashval(name);
3838 struct nfs4_client *clp;
3839
a1bcecd2 3840 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3841 return clp ? 1 : 0;
3842}
3843
1da177e4
LT		 3844/*
3845 * failure => all reset bets are off, nfserr_no_grace...
3846 */
190e4fbf
N		 3847int
3848nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3849{
3850 unsigned int strhashval;
3851 struct nfs4_client_reclaim *crp = NULL;
3852
a55370a3
N		 3853 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3854 crp = alloc_reclaim();
1da177e4
LT		 3855 if (!crp)
3856 return 0;
a55370a3 3857 strhashval = clientstr_hashval(name);
1da177e4
LT		 3858 INIT_LIST_HEAD(&crp->cr_strhash);
3859 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3860 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3861 reclaim_str_hashtbl_size++;
3862 return 1;
3863}
3864
3865static void
3866nfs4_release_reclaim(void)
3867{
3868 struct nfs4_client_reclaim *crp = NULL;
3869 int i;
3870
1da177e4
LT		 3871 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3872 while (!list_empty(&reclaim_str_hashtbl[i])) {
3873 crp = list_entry(reclaim_str_hashtbl[i].next,
3874 struct nfs4_client_reclaim, cr_strhash);
3875 list_del(&crp->cr_strhash);
1da177e4
LT		 3876 kfree(crp);
3877 reclaim_str_hashtbl_size--;
3878 }
3879 }
3880 BUG_ON(reclaim_str_hashtbl_size);
3881}
3882
3883/*
3884 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3885static struct nfs4_client_reclaim *
1da177e4
LT		 3886nfs4_find_reclaim_client(clientid_t *clid)
3887{
3888 unsigned int strhashval;
3889 struct nfs4_client *clp;
3890 struct nfs4_client_reclaim *crp = NULL;
3891
3892
3893 /* find clientid in conf_id_hashtbl */
3894 clp = find_confirmed_client(clid);
3895 if (clp == NULL)
3896 return NULL;
3897
a55370a3
N		 3898 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3899 clp->cl_name.len, clp->cl_name.data,
3900 clp->cl_recdir);
1da177e4
LT		 3901
3902 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3903 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3904 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3905 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3906 return crp;
3907 }
3908 }
3909 return NULL;
3910}
3911
3912/*
3913* Called from OPEN. Look for clientid in reclaim list.
3914*/
b37ad28b 3915__be32
1da177e4
LT		 3916nfs4_check_open_reclaim(clientid_t *clid)
3917{
dfc83565 3918 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3919}
3920
ac4d8ff2 3921/* initialization to perform at module load time: */
1da177e4 3922
e8ff2a84 3923int
ac4d8ff2 3924nfs4_state_init(void)
1da177e4 3925{
e8ff2a84 3926 int i, status;
1da177e4 3927
e8ff2a84
BF		 3928 status = nfsd4_init_slabs();
3929 if (status)
3930 return status;
1da177e4
LT		 3931 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3932 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3933 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3934 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3935 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3936 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3937 }
5282fd72
ME		 3938 for (i = 0; i < SESSION_HASH_SIZE; i++)
3939 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3940 for (i = 0; i < FILE_HASH_SIZE; i++) {
3941 INIT_LIST_HEAD(&file_hashtbl[i]);
3942 }
3943 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3944 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3945 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3946 }
3947 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3948 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3949 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3950 }
3951 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3952 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3953 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3954 }
1da177e4 3955 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3956 INIT_LIST_HEAD(&close_lru);
3957 INIT_LIST_HEAD(&client_lru);
3958 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 3959 reclaim_str_hashtbl_size = 0;
e8ff2a84 3960 return 0;
ac4d8ff2
N		 3961}
3962
190e4fbf
N		 3963static void
3964nfsd4_load_reboot_recovery_data(void)
3965{
3966 int status;
3967
0964a3d3
N		 3968 nfs4_lock_state();
3969 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3970 status = nfsd4_recdir_load();
0964a3d3 3971 nfs4_unlock_state();
190e4fbf
N		 3972 if (status)
3973 printk("NFSD: Failure reading reboot recovery data\n");
3974}
3975
c2f1a551
MS		 3976/*
3977 * Since the lifetime of a delegation isn't limited to that of an open, a
3978 * client may quite reasonably hang on to a delegation as long as it has
3979 * the inode cached. This becomes an obvious problem the first time a
3980 * client's inode cache approaches the size of the server's total memory.
3981 *
3982 * For now we avoid this problem by imposing a hard limit on the number
3983 * of delegations, which varies according to the server's memory size.
3984 */
3985static void
3986set_max_delegations(void)
3987{
3988 /*
3989 * Allow at most 4 delegations per megabyte of RAM. Quick
3990 * estimates suggest that in the worst case (where every delegation
3991 * is for a different inode), a delegation could take about 1.5K,
3992 * giving a worst case usage of about 6% of memory.
3993 */
3994 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
3995}
3996
ac4d8ff2
N		 3997/* initialization to perform when the nfsd service is started: */
3998
29ab23cc 3999static int
ac4d8ff2
N		 4000__nfs4_state_start(void)
4001{
b5a1a81e
BF		 4002 int ret;
4003
1da177e4 4004 boot_time = get_seconds();
af558e33 4005 locks_start_grace(&nfsd4_manager);
9a8db97e 4006 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
e46b498c 4007 nfsd4_grace);
b5a1a81e
BF		 4008 ret = set_callback_cred();
4009 if (ret)
4010 return -ENOMEM;
58da282b 4011 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF		 4012 if (laundry_wq == NULL)
4013 return -ENOMEM;
b5a1a81e
BF		 4014 ret = nfsd4_create_callback_queue();
4015 if (ret)
4016 goto out_free_laundry;
e46b498c 4017 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
c2f1a551 4018 set_max_delegations();
b5a1a81e
BF		 4019 return 0;
4020out_free_laundry:
4021 destroy_workqueue(laundry_wq);
4022 return ret;
1da177e4
LT		 4023}
4024
29ab23cc 4025int
76a3550e 4026nfs4_state_start(void)
1da177e4 4027{
29ab23cc
BF		 4028 int ret;
4029
1da177e4 4030 if (nfs4_init)
29ab23cc 4031 return 0;
190e4fbf 4032 nfsd4_load_reboot_recovery_data();
29ab23cc
BF		 4033 ret = __nfs4_state_start();
4034 if (ret)
4035 return ret;
1da177e4 4036 nfs4_init = 1;
29ab23cc 4037 return 0;
1da177e4
LT		 4038}
4039
1da177e4
LT		 4040static void
4041__nfs4_state_shutdown(void)
4042{
4043 int i;
4044 struct nfs4_client *clp = NULL;
4045 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4046 struct list_head *pos, *next, reaplist;
4047
1da177e4
LT		 4048 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4049 while (!list_empty(&conf_id_hashtbl[i])) {
4050 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4051 expire_client(clp);
4052 }
4053 while (!list_empty(&unconf_str_hashtbl[i])) {
4054 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4055 expire_client(clp);
4056 }
4057 }
4058 INIT_LIST_HEAD(&reaplist);
4059 spin_lock(&recall_lock);
4060 list_for_each_safe(pos, next, &del_recall_lru) {
4061 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4062 list_move(&dp->dl_recall_lru, &reaplist);
4063 }
4064 spin_unlock(&recall_lock);
4065 list_for_each_safe(pos, next, &reaplist) {
4066 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4067 list_del_init(&dp->dl_recall_lru);
4068 unhash_delegation(dp);
4069 }
4070
190e4fbf 4071 nfsd4_shutdown_recdir();
1da177e4 4072 nfs4_init = 0;
1da177e4
LT		 4073}
4074
4075void
4076nfs4_state_shutdown(void)
4077{
5e8d5c29
N		 4078 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4079 destroy_workqueue(laundry_wq);
2c5e7615 4080 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4081 nfs4_lock_state();
4082 nfs4_release_reclaim();
4083 __nfs4_state_shutdown();
b5a1a81e 4084 nfsd4_destroy_callback_queue();
1da177e4
LT		 4085 nfs4_unlock_state();
4086}
4087
3dd98a3b
JL		 4088/*
4089 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4090 * accessed when nfsd is starting.
4091 */
0964a3d3
N		 4092static void
4093nfs4_set_recdir(char *recdir)
4094{
0964a3d3 4095 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4096}
4097
4098/*
4099 * Change the NFSv4 recovery directory to recdir.
4100 */
4101int
4102nfs4_reset_recoverydir(char *recdir)
4103{
4104 int status;
a63bb996 4105 struct path path;
0964a3d3 4106
a63bb996 4107 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4108 if (status)
4109 return status;
4110 status = -ENOTDIR;
a63bb996 4111 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4112 nfs4_set_recdir(recdir);
4113 status = 0;
4114 }
a63bb996 4115 path_put(&path);
0964a3d3
N		 4116 return status;
4117}
4118
3dd98a3b
JL		 4119char *
4120nfs4_recoverydir(void)
4121{
4122 return user_recovery_dirname;
4123}
Linux 6.x block layer and io_uring tree(s)