projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
nfsd4: extend the client_lock to cover cl_lru
[deliverable/linux.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
36acb66b 253/* client_lock protects the client lru list and session hash table */
9089f1b4
BH		 254static DEFINE_SPINLOCK(client_lock);
255
1da177e4
LT		 256/* Hash tables for nfs4_clientid state */
257#define CLIENT_HASH_BITS 4
258#define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
259#define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
260
261#define clientid_hashval(id) \
262 ((id) & CLIENT_HASH_MASK)
a55370a3
N		 263#define clientstr_hashval(name) \
264 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
1da177e4
LT		 265/*
266 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
267 * used in reboot/reset lease grace period processing
268 *
269 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
270 * setclientid_confirmed info.
271 *
272 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
273 * setclientid info.
274 *
275 * client_lru holds client queue ordered by nfs4_client.cl_time
276 * for lease renewal.
277 *
278 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
279 * for last close replay.
280 */
281static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
282static int reclaim_str_hashtbl_size = 0;
283static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
284static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
285static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
286static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
287static struct list_head client_lru;
288static struct list_head close_lru;
289
2283963f
BF		 290static void unhash_generic_stateid(struct nfs4_stateid *stp)
291{
292 list_del(&stp->st_hash);
293 list_del(&stp->st_perfile);
294 list_del(&stp->st_perstateowner);
295}
296
297static void free_generic_stateid(struct nfs4_stateid *stp)
298{
299 put_nfs4_file(stp->st_file);
300 kmem_cache_free(stateid_slab, stp);
301}
302
303static void release_lock_stateid(struct nfs4_stateid *stp)
304{
305 unhash_generic_stateid(stp);
306 locks_remove_posix(stp->st_vfs_file, (fl_owner_t)stp->st_stateowner);
307 free_generic_stateid(stp);
308}
309
f044ff83
BF		 310static void unhash_lockowner(struct nfs4_stateowner *sop)
311{
312 struct nfs4_stateid *stp;
313
314 list_del(&sop->so_idhash);
315 list_del(&sop->so_strhash);
316 list_del(&sop->so_perstateid);
317 while (!list_empty(&sop->so_stateids)) {
318 stp = list_first_entry(&sop->so_stateids,
319 struct nfs4_stateid, st_perstateowner);
320 release_lock_stateid(stp);
321 }
322}
323
324static void release_lockowner(struct nfs4_stateowner *sop)
325{
326 unhash_lockowner(sop);
327 nfs4_put_stateowner(sop);
328}
329
f1d110ca
BF		 330static void
331release_stateid_lockowners(struct nfs4_stateid *open_stp)
332{
333 struct nfs4_stateowner *lock_sop;
334
335 while (!list_empty(&open_stp->st_lockowners)) {
336 lock_sop = list_entry(open_stp->st_lockowners.next,
337 struct nfs4_stateowner, so_perstateid);
338 /* list_del(&open_stp->st_lockowners); */
339 BUG_ON(lock_sop->so_is_open_owner);
f044ff83 340 release_lockowner(lock_sop);
f1d110ca
BF		 341 }
342}
343
2283963f
BF		 344static void release_open_stateid(struct nfs4_stateid *stp)
345{
346 unhash_generic_stateid(stp);
347 release_stateid_lockowners(stp);
348 nfsd_close(stp->st_vfs_file);
349 free_generic_stateid(stp);
350}
351
f044ff83 352static void unhash_openowner(struct nfs4_stateowner *sop)
f1d110ca
BF		 353{
354 struct nfs4_stateid *stp;
355
356 list_del(&sop->so_idhash);
357 list_del(&sop->so_strhash);
f044ff83
BF		 358 list_del(&sop->so_perclient);
359 list_del(&sop->so_perstateid); /* XXX: necessary? */
f1d110ca 360 while (!list_empty(&sop->so_stateids)) {
f044ff83
BF		 361 stp = list_first_entry(&sop->so_stateids,
362 struct nfs4_stateid, st_perstateowner);
363 release_open_stateid(stp);
f1d110ca
BF		 364 }
365}
366
f044ff83 367static void release_openowner(struct nfs4_stateowner *sop)
f1d110ca 368{
f044ff83 369 unhash_openowner(sop);
f1d110ca
BF		 370 list_del(&sop->so_close_lru);
371 nfs4_put_stateowner(sop);
372}
373
5282fd72
ME		 374#define SESSION_HASH_SIZE 512
375static struct list_head sessionid_hashtbl[SESSION_HASH_SIZE];
376
377static inline int
378hash_sessionid(struct nfs4_sessionid *sessionid)
379{
380 struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
381
382 return sid->sequence % SESSION_HASH_SIZE;
383}
384
385static inline void
386dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
387{
388 u32 *ptr = (u32 *)(&sessionid->data[0]);
389 dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
390}
391
ec6b5d7b
AA		 392static void
393gen_sessionid(struct nfsd4_session *ses)
394{
395 struct nfs4_client *clp = ses->se_client;
396 struct nfsd4_sessionid *sid;
397
398 sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
399 sid->clientid = clp->cl_clientid;
400 sid->sequence = current_sessionid++;
401 sid->reserved = 0;
402}
403
404/*
a649637c
AA		 405 * The protocol defines ca_maxresponssize_cached to include the size of
406 * the rpc header, but all we need to cache is the data starting after
407 * the end of the initial SEQUENCE operation--the rest we regenerate
408 * each time. Therefore we can advertise a ca_maxresponssize_cached
409 * value that is the number of bytes in our cache plus a few additional
410 * bytes. In order to stay on the safe side, and not promise more than
411 * we can cache, those additional bytes must be the minimum possible: 24
412 * bytes of rpc header (xid through accept state, with AUTH_NULL
413 * verifier), 12 for the compound header (with zero-length tag), and 44
414 * for the SEQUENCE op response:
415 */
416#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
417
418/*
419 * Give the client the number of ca_maxresponsesize_cached slots it
420 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE,
421 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more
422 * than NFSD_MAX_SLOTS_PER_SESSION.
ec6b5d7b 423 *
a649637c
AA		 424 * If we run out of reserved DRC memory we should (up to a point)
425 * re-negotiate active sessions and reduce their slot usage to make
426 * rooom for new connections. For now we just fail the create session.
ec6b5d7b 427 */
a649637c 428static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan)
ec6b5d7b 429{
a649637c 430 int mem, size = fchan->maxresp_cached;
ec6b5d7b 431
5d77ddfb
AA		 432 if (fchan->maxreqs < 1)
433 return nfserr_inval;
5d77ddfb 434
a649637c
AA		 435 if (size < NFSD_MIN_HDR_SEQ_SZ)
436 size = NFSD_MIN_HDR_SEQ_SZ;
437 size -= NFSD_MIN_HDR_SEQ_SZ;
438 if (size > NFSD_SLOT_CACHE_SIZE)
439 size = NFSD_SLOT_CACHE_SIZE;
ec6b5d7b 440
a649637c
AA		 441 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
442 mem = fchan->maxreqs * size;
443 if (mem > NFSD_MAX_MEM_PER_SESSION) {
444 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
445 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
446 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
447 mem = fchan->maxreqs * size;
448 }
ec6b5d7b 449
4bd9b0f4 450 spin_lock(&nfsd_drc_lock);
a649637c
AA		 451 /* bound the total session drc memory ussage */
452 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
453 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
454 mem = fchan->maxreqs * size;
455 }
0c193054 456 nfsd_drc_mem_used += mem;
4bd9b0f4 457 spin_unlock(&nfsd_drc_lock);
ec6b5d7b 458
b101ebbc 459 if (fchan->maxreqs == 0)
bdac86e2 460 return nfserr_serverfault;
a649637c
AA		 461
462 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ;
b101ebbc 463 return 0;
ec6b5d7b
AA		 464}
465
466/*
467 * fchan holds the client values on input, and the server values on output
ddc04fd4 468 * sv_max_mesg is the maximum payload plus one page for overhead.
ec6b5d7b
AA		 469 */
470static int init_forechannel_attrs(struct svc_rqst *rqstp,
6c18ba9f
AB		 471 struct nfsd4_channel_attrs *session_fchan,
472 struct nfsd4_channel_attrs *fchan)
ec6b5d7b
AA		 473{
474 int status = 0;
ddc04fd4 475 __u32 maxcount = nfsd_serv->sv_max_mesg;
ec6b5d7b
AA		 476
477 /* headerpadsz set to zero in encode routine */
478
479 /* Use the client's max request and max response size if possible */
480 if (fchan->maxreq_sz > maxcount)
481 fchan->maxreq_sz = maxcount;
6c18ba9f 482 session_fchan->maxreq_sz = fchan->maxreq_sz;
ec6b5d7b
AA		 483
484 if (fchan->maxresp_sz > maxcount)
485 fchan->maxresp_sz = maxcount;
6c18ba9f 486 session_fchan->maxresp_sz = fchan->maxresp_sz;
ec6b5d7b 487
ec6b5d7b
AA		 488 /* Use the client's maxops if possible */
489 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND)
490 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND;
6c18ba9f 491 session_fchan->maxops = fchan->maxops;
ec6b5d7b 492
ec6b5d7b
AA		 493 /* FIXME: Error means no more DRC pages so the server should
494 * recover pages from existing sessions. For now fail session
495 * creation.
496 */
a649637c 497 status = set_forechannel_drc_size(fchan);
ec6b5d7b 498
a649637c 499 session_fchan->maxresp_cached = fchan->maxresp_cached;
6c18ba9f 500 session_fchan->maxreqs = fchan->maxreqs;
a649637c
AA		 501
502 dprintk("%s status %d\n", __func__, status);
ec6b5d7b
AA		 503 return status;
504}
505
557ce264
AA		 506static void
507free_session_slots(struct nfsd4_session *ses)
508{
509 int i;
510
511 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
512 kfree(ses->se_slots[i]);
513}
514
efe0cb6d
BF		 515/*
516 * We don't actually need to cache the rpc and session headers, so we
517 * can allocate a little less for each slot:
518 */
519static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
520{
521 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
522}
523
ec6b5d7b
AA		 524static int
525alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
526 struct nfsd4_create_session *cses)
527{
528 struct nfsd4_session *new, tmp;
557ce264
AA		 529 struct nfsd4_slot *sp;
530 int idx, slotsize, cachesize, i;
531 int status;
ec6b5d7b
AA		 532
533 memset(&tmp, 0, sizeof(tmp));
534
535 /* FIXME: For now, we just accept the client back channel attributes. */
6c18ba9f
AB		 536 tmp.se_bchannel = cses->back_channel;
537 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel,
538 &cses->fore_channel);
ec6b5d7b
AA		 539 if (status)
540 goto out;
541
557ce264
AA		 542 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot)
543 + sizeof(struct nfsd4_session) > PAGE_SIZE);
544
545 status = nfserr_serverfault;
546 /* allocate struct nfsd4_session and slot table pointers in one piece */
547 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
ec6b5d7b
AA		 548 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
549 if (!new)
550 goto out;
551
552 memcpy(new, &tmp, sizeof(*new));
553
557ce264 554 /* allocate each struct nfsd4_slot and data cache in one piece */
efe0cb6d 555 cachesize = slot_bytes(&new->se_fchannel);
557ce264
AA		 556 for (i = 0; i < new->se_fchannel.maxreqs; i++) {
557 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL);
558 if (!sp)
559 goto out_free;
560 new->se_slots[i] = sp;
561 }
562
ec6b5d7b
AA		 563 new->se_client = clp;
564 gen_sessionid(new);
565 idx = hash_sessionid(&new->se_sessionid);
566 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
567 NFS4_MAX_SESSIONID_LEN);
568
569 new->se_flags = cses->flags;
570 kref_init(&new->se_ref);
9089f1b4 571 spin_lock(&client_lock);
ec6b5d7b
AA		 572 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
573 list_add(&new->se_perclnt, &clp->cl_sessions);
9089f1b4 574 spin_unlock(&client_lock);
ec6b5d7b
AA		 575
576 status = nfs_ok;
577out:
578 return status;
557ce264
AA		 579out_free:
580 free_session_slots(new);
581 kfree(new);
582 goto out;
ec6b5d7b
AA		 583}
584
9089f1b4 585/* caller must hold client_lock */
5282fd72
ME		 586static struct nfsd4_session *
587find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid)
588{
589 struct nfsd4_session *elem;
590 int idx;
591
592 dump_sessionid(__func__, sessionid);
593 idx = hash_sessionid(sessionid);
594 dprintk("%s: idx is %d\n", __func__, idx);
595 /* Search in the appropriate list */
596 list_for_each_entry(elem, &sessionid_hashtbl[idx], se_hash) {
597 dump_sessionid("list traversal", &elem->se_sessionid);
598 if (!memcmp(elem->se_sessionid.data, sessionid->data,
599 NFS4_MAX_SESSIONID_LEN)) {
600 return elem;
601 }
602 }
603
604 dprintk("%s: session not found\n", __func__);
605 return NULL;
606}
607
9089f1b4 608/* caller must hold client_lock */
7116ed6b 609static void
5282fd72 610unhash_session(struct nfsd4_session *ses)
7116ed6b
AA		 611{
612 list_del(&ses->se_hash);
613 list_del(&ses->se_perclnt);
5282fd72
ME		 614}
615
7116ed6b
AA		 616void
617free_session(struct kref *kref)
618{
619 struct nfsd4_session *ses;
dd829c45 620 int mem;
7116ed6b
AA		 621
622 ses = container_of(kref, struct nfsd4_session, se_ref);
be98d1bb 623 spin_lock(&nfsd_drc_lock);
efe0cb6d 624 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
dd829c45 625 nfsd_drc_mem_used -= mem;
be98d1bb 626 spin_unlock(&nfsd_drc_lock);
557ce264 627 free_session_slots(ses);
7116ed6b
AA		 628 kfree(ses);
629}
630
36acb66b 631/* must be called under the client_lock */
1da177e4 632static inline void
36acb66b 633renew_client_locked(struct nfs4_client *clp)
1da177e4
LT		 634{
635 /*
636 * Move client to the end to the LRU list.
637 */
638 dprintk("renewing client (clientid %08x/%08x)\n",
639 clp->cl_clientid.cl_boot,
640 clp->cl_clientid.cl_id);
641 list_move_tail(&clp->cl_lru, &client_lru);
642 clp->cl_time = get_seconds();
643}
644
36acb66b
BH		 645static inline void
646renew_client(struct nfs4_client *clp)
647{
648 spin_lock(&client_lock);
649 renew_client_locked(clp);
650 spin_unlock(&client_lock);
651}
652
1da177e4
LT		 653/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
654static int
655STALE_CLIENTID(clientid_t *clid)
656{
657 if (clid->cl_boot == boot_time)
658 return 0;
60adfc50
AA		 659 dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
660 clid->cl_boot, clid->cl_id, boot_time);
1da177e4
LT		 661 return 1;
662}
663
664/*
665 * XXX Should we use a slab cache ?
666 * This type of memory management is somewhat inefficient, but we use it
667 * anyway since SETCLIENTID is not a common operation.
668 */
35bba9a3 669static struct nfs4_client *alloc_client(struct xdr_netobj name)
1da177e4
LT		 670{
671 struct nfs4_client *clp;
672
35bba9a3
BF		 673 clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
674 if (clp == NULL)
675 return NULL;
676 clp->cl_name.data = kmalloc(name.len, GFP_KERNEL);
677 if (clp->cl_name.data == NULL) {
678 kfree(clp);
679 return NULL;
1da177e4 680 }
35bba9a3
BF		 681 memcpy(clp->cl_name.data, name.data, name.len);
682 clp->cl_name.len = name.len;
1da177e4
LT		 683 return clp;
684}
685
686static inline void
687free_client(struct nfs4_client *clp)
688{
689 if (clp->cl_cred.cr_group_info)
690 put_group_info(clp->cl_cred.cr_group_info);
68e76ad0 691 kfree(clp->cl_principal);
1da177e4
LT		 692 kfree(clp->cl_name.data);
693 kfree(clp);
694}
695
1da177e4
LT		 696static void
697expire_client(struct nfs4_client *clp)
698{
699 struct nfs4_stateowner *sop;
700 struct nfs4_delegation *dp;
1da177e4
LT		 701 struct list_head reaplist;
702
1da177e4
LT		 703 INIT_LIST_HEAD(&reaplist);
704 spin_lock(&recall_lock);
ea1da636
N		 705 while (!list_empty(&clp->cl_delegations)) {
706 dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1da177e4
LT		 707 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
708 dp->dl_flock);
ea1da636 709 list_del_init(&dp->dl_perclnt);
1da177e4
LT		 710 list_move(&dp->dl_recall_lru, &reaplist);
711 }
712 spin_unlock(&recall_lock);
713 while (!list_empty(&reaplist)) {
714 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
715 list_del_init(&dp->dl_recall_lru);
716 unhash_delegation(dp);
717 }
ea1da636
N		 718 while (!list_empty(&clp->cl_openowners)) {
719 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
f044ff83 720 release_openowner(sop);
1da177e4 721 }
36acb66b
BH		 722 list_del(&clp->cl_idhash);
723 list_del(&clp->cl_strhash);
be1fdf6c 724 spin_lock(&client_lock);
36acb66b 725 list_del(&clp->cl_lru);
c4bf7868
ME		 726 while (!list_empty(&clp->cl_sessions)) {
727 struct nfsd4_session *ses;
728 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
729 se_perclnt);
be1fdf6c
BH		 730 unhash_session(ses);
731 nfsd4_put_session(ses);
c4bf7868 732 }
be1fdf6c 733 spin_unlock(&client_lock);
b5a1a81e 734 nfsd4_set_callback_client(clp, NULL);
2bf23875
BF		 735 if (clp->cl_cb_conn.cb_xprt)
736 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
b12a05cb 737 free_client(clp);
1da177e4
LT		 738}
739
35bba9a3
BF		 740static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
741{
742 memcpy(target->cl_verifier.data, source->data,
743 sizeof(target->cl_verifier.data));
1da177e4
LT		 744}
745
35bba9a3
BF		 746static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
747{
1da177e4
LT		 748 target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
749 target->cl_clientid.cl_id = source->cl_clientid.cl_id;
750}
751
35bba9a3
BF		 752static void copy_cred(struct svc_cred *target, struct svc_cred *source)
753{
1da177e4
LT		 754 target->cr_uid = source->cr_uid;
755 target->cr_gid = source->cr_gid;
756 target->cr_group_info = source->cr_group_info;
757 get_group_info(target->cr_group_info);
758}
759
35bba9a3 760static int same_name(const char *n1, const char *n2)
599e0a22 761{
a55370a3 762 return 0 == memcmp(n1, n2, HEXDIR_LEN);
1da177e4
LT		 763}
764
765static int
599e0a22
BF		 766same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
767{
768 return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1da177e4
LT		 769}
770
771static int
599e0a22
BF		 772same_clid(clientid_t *cl1, clientid_t *cl2)
773{
774 return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1da177e4
LT		 775}
776
777/* XXX what about NGROUP */
778static int
599e0a22
BF		 779same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
780{
781 return cr1->cr_uid == cr2->cr_uid;
1da177e4
LT		 782}
783
5ec7b46c
BF		 784static void gen_clid(struct nfs4_client *clp)
785{
786 static u32 current_clientid = 1;
787
1da177e4
LT		 788 clp->cl_clientid.cl_boot = boot_time;
789 clp->cl_clientid.cl_id = current_clientid++;
790}
791
deda2faa
BF		 792static void gen_confirm(struct nfs4_client *clp)
793{
794 static u32 i;
795 u32 *p;
1da177e4 796
1da177e4 797 p = (u32 *)clp->cl_confirm.data;
deda2faa
BF		 798 *p++ = get_seconds();
799 *p++ = i++;
1da177e4
LT		 800}
801
b09333c4
RL		 802static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
803 struct svc_rqst *rqstp, nfs4_verifier *verf)
804{
805 struct nfs4_client *clp;
806 struct sockaddr *sa = svc_addr(rqstp);
807 char *princ;
808
809 clp = alloc_client(name);
810 if (clp == NULL)
811 return NULL;
812
813 princ = svc_gss_principal(rqstp);
814 if (princ) {
815 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
816 if (clp->cl_principal == NULL) {
817 free_client(clp);
818 return NULL;
819 }
820 }
821
822 memcpy(clp->cl_recdir, recdir, HEXDIR_LEN);
2bf23875 823 atomic_set(&clp->cl_cb_set, 0);
b09333c4
RL		 824 INIT_LIST_HEAD(&clp->cl_idhash);
825 INIT_LIST_HEAD(&clp->cl_strhash);
826 INIT_LIST_HEAD(&clp->cl_openowners);
827 INIT_LIST_HEAD(&clp->cl_delegations);
828 INIT_LIST_HEAD(&clp->cl_sessions);
829 INIT_LIST_HEAD(&clp->cl_lru);
830 clear_bit(0, &clp->cl_cb_slot_busy);
831 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
832 copy_verf(clp, verf);
833 rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
834 clp->cl_flavor = rqstp->rq_flavor;
835 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
836 gen_confirm(clp);
837
838 return clp;
839}
840
35bba9a3
BF		 841static int check_name(struct xdr_netobj name)
842{
1da177e4
LT		 843 if (name.len == 0)
844 return 0;
845 if (name.len > NFS4_OPAQUE_LIMIT) {
2fdada03 846 dprintk("NFSD: check_name: name too long(%d)!\n", name.len);
1da177e4
LT		 847 return 0;
848 }
849 return 1;
850}
851
fd39ca9a 852static void
1da177e4
LT		 853add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
854{
855 unsigned int idhashval;
856
857 list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
858 idhashval = clientid_hashval(clp->cl_clientid.cl_id);
859 list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
36acb66b 860 renew_client(clp);
1da177e4
LT		 861}
862
fd39ca9a 863static void
1da177e4
LT		 864move_to_confirmed(struct nfs4_client *clp)
865{
866 unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
867 unsigned int strhashval;
868
869 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
f116629d 870 list_move(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
a55370a3 871 strhashval = clientstr_hashval(clp->cl_recdir);
328efbab 872 list_move(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
1da177e4
LT		 873 renew_client(clp);
874}
875
876static struct nfs4_client *
877find_confirmed_client(clientid_t *clid)
878{
879 struct nfs4_client *clp;
880 unsigned int idhashval = clientid_hashval(clid->cl_id);
881
882 list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 883 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 884 return clp;
885 }
886 return NULL;
887}
888
889static struct nfs4_client *
890find_unconfirmed_client(clientid_t *clid)
891{
892 struct nfs4_client *clp;
893 unsigned int idhashval = clientid_hashval(clid->cl_id);
894
895 list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
599e0a22 896 if (same_clid(&clp->cl_clientid, clid))
1da177e4
LT		 897 return clp;
898 }
899 return NULL;
900}
901
a1bcecd2
AA		 902/*
903 * Return 1 iff clp's clientid establishment method matches the use_exchange_id
904 * parameter. Matching is based on the fact the at least one of the
905 * EXCHGID4_FLAG_USE_{NON_PNFS,PNFS_MDS,PNFS_DS} flags must be set for v4.1
906 *
907 * FIXME: we need to unify the clientid namespaces for nfsv4.x
908 * and correctly deal with client upgrade/downgrade in EXCHANGE_ID
909 * and SET_CLIENTID{,_CONFIRM}
910 */
911static inline int
912match_clientid_establishment(struct nfs4_client *clp, bool use_exchange_id)
913{
914 bool has_exchange_flags = (clp->cl_exchange_flags != 0);
915 return use_exchange_id == has_exchange_flags;
916}
917
28ce6054 918static struct nfs4_client *
a1bcecd2
AA		 919find_confirmed_client_by_str(const char *dname, unsigned int hashval,
920 bool use_exchange_id)
28ce6054
N		 921{
922 struct nfs4_client *clp;
923
924 list_for_each_entry(clp, &conf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 925 if (same_name(clp->cl_recdir, dname) &&
926 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 927 return clp;
928 }
929 return NULL;
930}
931
932static struct nfs4_client *
a1bcecd2
AA		 933find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
934 bool use_exchange_id)
28ce6054
N		 935{
936 struct nfs4_client *clp;
937
938 list_for_each_entry(clp, &unconf_str_hashtbl[hashval], cl_strhash) {
a1bcecd2
AA		 939 if (same_name(clp->cl_recdir, dname) &&
940 match_clientid_establishment(clp, use_exchange_id))
28ce6054
N		 941 return clp;
942 }
943 return NULL;
944}
945
fd39ca9a 946static void
fbf4665f 947gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1da177e4 948{
c237dc03 949 struct nfs4_cb_conn *cb = &clp->cl_cb_conn;
7077ecba
JL		 950 unsigned short expected_family;
951
952 /* Currently, we only support tcp and tcp6 for the callback channel */
953 if (se->se_callback_netid_len == 3 &&
954 !memcmp(se->se_callback_netid_val, "tcp", 3))
955 expected_family = AF_INET;
956 else if (se->se_callback_netid_len == 4 &&
957 !memcmp(se->se_callback_netid_val, "tcp6", 4))
958 expected_family = AF_INET6;
959 else
1da177e4
LT		 960 goto out_err;
961
aa9a4ec7
JL		 962 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
963 se->se_callback_addr_len,
964 (struct sockaddr *) &cb->cb_addr,
965 sizeof(cb->cb_addr));
966
7077ecba 967 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family)
1da177e4 968 goto out_err;
aa9a4ec7 969
fbf4665f
JL		 970 if (cb->cb_addr.ss_family == AF_INET6)
971 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid;
972
ab52ae6d 973 cb->cb_minorversion = 0;
1da177e4
LT		 974 cb->cb_prog = se->se_callback_prog;
975 cb->cb_ident = se->se_callback_ident;
1da177e4
LT		 976 return;
977out_err:
aa9a4ec7
JL		 978 cb->cb_addr.ss_family = AF_UNSPEC;
979 cb->cb_addrlen = 0;
849823c5 980 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1da177e4
LT		 981 "will not receive delegations\n",
982 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
983
1da177e4
LT		 984 return;
985}
986
074fe897 987/*
557ce264 988 * Cache a reply. nfsd4_check_drc_limit() has bounded the cache size.
074fe897 989 */
074fe897
AA		 990void
991nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
074fe897 992{
557ce264
AA		 993 struct nfsd4_slot *slot = resp->cstate.slot;
994 unsigned int base;
074fe897 995
557ce264 996 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 997
557ce264
AA		 998 slot->sl_opcnt = resp->opcnt;
999 slot->sl_status = resp->cstate.status;
074fe897 1000
bf864a31 1001 if (nfsd4_not_cached(resp)) {
557ce264 1002 slot->sl_datalen = 0;
bf864a31 1003 return;
074fe897 1004 }
557ce264
AA		 1005 slot->sl_datalen = (char *)resp->p - (char *)resp->cstate.datap;
1006 base = (char *)resp->cstate.datap -
1007 (char *)resp->xbuf->head[0].iov_base;
1008 if (read_bytes_from_xdr_buf(resp->xbuf, base, slot->sl_data,
1009 slot->sl_datalen))
1010 WARN("%s: sessions DRC could not cache compound\n", __func__);
1011 return;
074fe897
AA		 1012}
1013
1014/*
abfabf8c
AA		 1015 * Encode the replay sequence operation from the slot values.
1016 * If cachethis is FALSE encode the uncached rep error on the next
1017 * operation which sets resp->p and increments resp->opcnt for
1018 * nfs4svc_encode_compoundres.
074fe897 1019 *
074fe897 1020 */
abfabf8c
AA		 1021static __be32
1022nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
1023 struct nfsd4_compoundres *resp)
074fe897 1024{
abfabf8c
AA		 1025 struct nfsd4_op *op;
1026 struct nfsd4_slot *slot = resp->cstate.slot;
bf864a31 1027
abfabf8c 1028 dprintk("--> %s resp->opcnt %d cachethis %u \n", __func__,
557ce264 1029 resp->opcnt, resp->cstate.slot->sl_cachethis);
bf864a31 1030
abfabf8c
AA		 1031 /* Encode the replayed sequence operation */
1032 op = &args->ops[resp->opcnt - 1];
1033 nfsd4_encode_operation(resp, op);
bf864a31 1034
abfabf8c 1035 /* Return nfserr_retry_uncached_rep in next operation. */
557ce264 1036 if (args->opcnt > 1 && slot->sl_cachethis == 0) {
abfabf8c
AA		 1037 op = &args->ops[resp->opcnt++];
1038 op->status = nfserr_retry_uncached_rep;
1039 nfsd4_encode_operation(resp, op);
074fe897 1040 }
abfabf8c 1041 return op->status;
074fe897
AA		 1042}
1043
1044/*
557ce264
AA		 1045 * The sequence operation is not cached because we can use the slot and
1046 * session values.
074fe897
AA		 1047 */
1048__be32
bf864a31
AA		 1049nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
1050 struct nfsd4_sequence *seq)
074fe897 1051{
557ce264 1052 struct nfsd4_slot *slot = resp->cstate.slot;
074fe897
AA		 1053 __be32 status;
1054
557ce264 1055 dprintk("--> %s slot %p\n", __func__, slot);
074fe897 1056
abfabf8c
AA		 1057 /* Either returns 0 or nfserr_retry_uncached */
1058 status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
1059 if (status == nfserr_retry_uncached_rep)
1060 return status;
074fe897 1061
557ce264
AA		 1062 /* The sequence operation has been encoded, cstate->datap set. */
1063 memcpy(resp->cstate.datap, slot->sl_data, slot->sl_datalen);
074fe897 1064
557ce264
AA		 1065 resp->opcnt = slot->sl_opcnt;
1066 resp->p = resp->cstate.datap + XDR_QUADLEN(slot->sl_datalen);
1067 status = slot->sl_status;
074fe897
AA		 1068
1069 return status;
1070}
1071
0733d213
AA		 1072/*
1073 * Set the exchange_id flags returned by the server.
1074 */
1075static void
1076nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
1077{
1078 /* pNFS is not supported */
1079 new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
1080
1081 /* Referrals are supported, Migration is not. */
1082 new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
1083
1084 /* set the wire flags to return to client. */
1085 clid->flags = new->cl_exchange_flags;
1086}
1087
069b6ad4
AA		 1088__be32
1089nfsd4_exchange_id(struct svc_rqst *rqstp,
1090 struct nfsd4_compound_state *cstate,
1091 struct nfsd4_exchange_id *exid)
1092{
0733d213
AA		 1093 struct nfs4_client *unconf, *conf, *new;
1094 int status;
1095 unsigned int strhashval;
1096 char dname[HEXDIR_LEN];
363168b4 1097 char addr_str[INET6_ADDRSTRLEN];
0733d213 1098 nfs4_verifier verf = exid->verifier;
363168b4 1099 struct sockaddr *sa = svc_addr(rqstp);
0733d213 1100
363168b4 1101 rpc_ntop(sa, addr_str, sizeof(addr_str));
0733d213 1102 dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
363168b4 1103 "ip_addr=%s flags %x, spa_how %d\n",
0733d213 1104 __func__, rqstp, exid, exid->clname.len, exid->clname.data,
363168b4 1105 addr_str, exid->flags, exid->spa_how);
0733d213
AA		 1106
1107 if (!check_name(exid->clname) || (exid->flags & ~EXCHGID4_FLAG_MASK_A))
1108 return nfserr_inval;
1109
1110 /* Currently only support SP4_NONE */
1111 switch (exid->spa_how) {
1112 case SP4_NONE:
1113 break;
1114 case SP4_SSV:
1115 return nfserr_encr_alg_unsupp;
1116 default:
1117 BUG(); /* checked by xdr code */
1118 case SP4_MACH_CRED:
1119 return nfserr_serverfault; /* no excuse :-/ */
1120 }
1121
1122 status = nfs4_make_rec_clidname(dname, &exid->clname);
1123
1124 if (status)
1125 goto error;
1126
1127 strhashval = clientstr_hashval(dname);
1128
1129 nfs4_lock_state();
1130 status = nfs_ok;
1131
a1bcecd2 1132 conf = find_confirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1133 if (conf) {
1134 if (!same_verf(&verf, &conf->cl_verifier)) {
1135 /* 18.35.4 case 8 */
1136 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1137 status = nfserr_not_same;
1138 goto out;
1139 }
1140 /* Client reboot: destroy old state */
1141 expire_client(conf);
1142 goto out_new;
1143 }
1144 if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
1145 /* 18.35.4 case 9 */
1146 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1147 status = nfserr_perm;
1148 goto out;
1149 }
1150 expire_client(conf);
1151 goto out_new;
1152 }
0733d213
AA		 1153 /*
1154 * Set bit when the owner id and verifier map to an already
1155 * confirmed client id (18.35.3).
1156 */
1157 exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
1158
1159 /*
1160 * Falling into 18.35.4 case 2, possible router replay.
1161 * Leave confirmed record intact and return same result.
1162 */
1163 copy_verf(conf, &verf);
1164 new = conf;
1165 goto out_copy;
6ddbbbfe
MS		 1166 }
1167
1168 /* 18.35.4 case 7 */
1169 if (exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A) {
1170 status = nfserr_noent;
1171 goto out;
0733d213
AA		 1172 }
1173
a1bcecd2 1174 unconf = find_unconfirmed_client_by_str(dname, strhashval, true);
0733d213
AA		 1175 if (unconf) {
1176 /*
1177 * Possible retry or client restart. Per 18.35.4 case 4,
1178 * a new unconfirmed record should be generated regardless
1179 * of whether any properties have changed.
1180 */
1181 expire_client(unconf);
1182 }
1183
1184out_new:
1185 /* Normal case */
b09333c4 1186 new = create_client(exid->clname, dname, rqstp, &verf);
0733d213 1187 if (new == NULL) {
bdac86e2 1188 status = nfserr_serverfault;
0733d213
AA		 1189 goto out;
1190 }
1191
0733d213 1192 gen_clid(new);
0733d213
AA		 1193 add_to_unconfirmed(new, strhashval);
1194out_copy:
1195 exid->clientid.cl_boot = new->cl_clientid.cl_boot;
1196 exid->clientid.cl_id = new->cl_clientid.cl_id;
1197
38eb76a5 1198 exid->seqid = 1;
0733d213
AA		 1199 nfsd4_set_ex_flags(new, exid);
1200
1201 dprintk("nfsd4_exchange_id seqid %d flags %x\n",
49557cc7 1202 new->cl_cs_slot.sl_seqid, new->cl_exchange_flags);
0733d213
AA		 1203 status = nfs_ok;
1204
1205out:
1206 nfs4_unlock_state();
1207error:
1208 dprintk("nfsd4_exchange_id returns %d\n", ntohl(status));
1209 return status;
069b6ad4
AA		 1210}
1211
b85d4c01 1212static int
88e588d5 1213check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
b85d4c01 1214{
88e588d5
AA		 1215 dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
1216 slot_seqid);
b85d4c01
BH		 1217
1218 /* The slot is in use, and no response has been sent. */
88e588d5
AA		 1219 if (slot_inuse) {
1220 if (seqid == slot_seqid)
b85d4c01
BH		 1221 return nfserr_jukebox;
1222 else
1223 return nfserr_seq_misordered;
1224 }
1225 /* Normal */
88e588d5 1226 if (likely(seqid == slot_seqid + 1))
b85d4c01
BH		 1227 return nfs_ok;
1228 /* Replay */
88e588d5 1229 if (seqid == slot_seqid)
b85d4c01
BH		 1230 return nfserr_replay_cache;
1231 /* Wraparound */
88e588d5 1232 if (seqid == 1 && (slot_seqid + 1) == 0)
b85d4c01
BH		 1233 return nfs_ok;
1234 /* Misordered replay or misordered new request */
1235 return nfserr_seq_misordered;
1236}
1237
49557cc7
AA		 1238/*
1239 * Cache the create session result into the create session single DRC
1240 * slot cache by saving the xdr structure. sl_seqid has been set.
1241 * Do this for solo or embedded create session operations.
1242 */
1243static void
1244nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
1245 struct nfsd4_clid_slot *slot, int nfserr)
1246{
1247 slot->sl_status = nfserr;
1248 memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
1249}
1250
1251static __be32
1252nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
1253 struct nfsd4_clid_slot *slot)
1254{
1255 memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
1256 return slot->sl_status;
1257}
1258
069b6ad4
AA		 1259__be32
1260nfsd4_create_session(struct svc_rqst *rqstp,
1261 struct nfsd4_compound_state *cstate,
1262 struct nfsd4_create_session *cr_ses)
1263{
363168b4 1264 struct sockaddr *sa = svc_addr(rqstp);
ec6b5d7b 1265 struct nfs4_client *conf, *unconf;
49557cc7 1266 struct nfsd4_clid_slot *cs_slot = NULL;
ec6b5d7b
AA		 1267 int status = 0;
1268
1269 nfs4_lock_state();
1270 unconf = find_unconfirmed_client(&cr_ses->clientid);
1271 conf = find_confirmed_client(&cr_ses->clientid);
1272
1273 if (conf) {
49557cc7
AA		 1274 cs_slot = &conf->cl_cs_slot;
1275 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5 1276 if (status == nfserr_replay_cache) {
ec6b5d7b 1277 dprintk("Got a create_session replay! seqid= %d\n",
49557cc7 1278 cs_slot->sl_seqid);
38eb76a5 1279 /* Return the cached reply status */
49557cc7 1280 status = nfsd4_replay_create_session(cr_ses, cs_slot);
38eb76a5 1281 goto out;
49557cc7 1282 } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
ec6b5d7b
AA		 1283 status = nfserr_seq_misordered;
1284 dprintk("Sequence misordered!\n");
1285 dprintk("Expected seqid= %d but got seqid= %d\n",
49557cc7 1286 cs_slot->sl_seqid, cr_ses->seqid);
ec6b5d7b
AA		 1287 goto out;
1288 }
49557cc7 1289 cs_slot->sl_seqid++;
ec6b5d7b
AA		 1290 } else if (unconf) {
1291 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
363168b4 1292 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
ec6b5d7b
AA		 1293 status = nfserr_clid_inuse;
1294 goto out;
1295 }
1296
49557cc7
AA		 1297 cs_slot = &unconf->cl_cs_slot;
1298 status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
38eb76a5
AA		 1299 if (status) {
1300 /* an unconfirmed replay returns misordered */
ec6b5d7b 1301 status = nfserr_seq_misordered;
49557cc7 1302 goto out_cache;
ec6b5d7b
AA		 1303 }
1304
49557cc7 1305 cs_slot->sl_seqid++; /* from 0 to 1 */
ec6b5d7b
AA		 1306 move_to_confirmed(unconf);
1307
38524ab3 1308 if (cr_ses->flags & SESSION4_BACK_CHAN) {
2bf23875
BF		 1309 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1310 svc_xprt_get(rqstp->rq_xprt);
38524ab3
AA		 1311 rpc_copy_addr(
1312 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1313 sa);
1314 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1315 unconf->cl_cb_conn.cb_minorversion =
1316 cstate->minorversion;
1317 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
2a1d1b59 1318 unconf->cl_cb_seq_nr = 1;
4b21d0de 1319 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
38524ab3 1320 }
ec6b5d7b
AA		 1321 conf = unconf;
1322 } else {
1323 status = nfserr_stale_clientid;
1324 goto out;
1325 }
1326
408b79bc
BF		 1327 /*
1328 * We do not support RDMA or persistent sessions
1329 */
1330 cr_ses->flags &= ~SESSION4_PERSIST;
1331 cr_ses->flags &= ~SESSION4_RDMA;
1332
ec6b5d7b
AA		 1333 status = alloc_init_session(rqstp, conf, cr_ses);
1334 if (status)
1335 goto out;
1336
ec6b5d7b
AA		 1337 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data,
1338 NFS4_MAX_SESSIONID_LEN);
49557cc7 1339 cr_ses->seqid = cs_slot->sl_seqid;
ec6b5d7b 1340
49557cc7
AA		 1341out_cache:
1342 /* cache solo and embedded create sessions under the state lock */
1343 nfsd4_cache_create_session(cr_ses, cs_slot, status);
ec6b5d7b
AA		 1344out:
1345 nfs4_unlock_state();
1346 dprintk("%s returns %d\n", __func__, ntohl(status));
1347 return status;
069b6ad4
AA		 1348}
1349
57716355
BF		 1350static bool nfsd4_last_compound_op(struct svc_rqst *rqstp)
1351{
1352 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1353 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1354
1355 return argp->opcnt == resp->opcnt;
1356}
1357
5d4cec2f
BF		 1358static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
1359{
1360 if (!session)
1361 return 0;
1362 return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
1363}
1364
069b6ad4
AA		 1365__be32
1366nfsd4_destroy_session(struct svc_rqst *r,
1367 struct nfsd4_compound_state *cstate,
1368 struct nfsd4_destroy_session *sessionid)
1369{
e10e0cfc
BH		 1370 struct nfsd4_session *ses;
1371 u32 status = nfserr_badsession;
1372
1373 /* Notes:
1374 * - The confirmed nfs4_client->cl_sessionid holds destroyed sessinid
1375 * - Should we return nfserr_back_chan_busy if waiting for
1376 * callbacks on to-be-destroyed session?
1377 * - Do we need to clear any callback info from previous session?
1378 */
1379
5d4cec2f 1380 if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
57716355
BF		 1381 if (!nfsd4_last_compound_op(r))
1382 return nfserr_not_only_op;
1383 }
e10e0cfc 1384 dump_sessionid(__func__, &sessionid->sessionid);
9089f1b4 1385 spin_lock(&client_lock);
e10e0cfc
BH		 1386 ses = find_in_sessionid_hashtbl(&sessionid->sessionid);
1387 if (!ses) {
9089f1b4 1388 spin_unlock(&client_lock);
e10e0cfc
BH		 1389 goto out;
1390 }
1391
1392 unhash_session(ses);
9089f1b4 1393 spin_unlock(&client_lock);
e10e0cfc
BH		 1394
1395 /* wait for callbacks */
b5a1a81e 1396 nfsd4_set_callback_client(ses->se_client, NULL);
e10e0cfc
BH		 1397 nfsd4_put_session(ses);
1398 status = nfs_ok;
1399out:
1400 dprintk("%s returns %d\n", __func__, ntohl(status));
1401 return status;
069b6ad4
AA		 1402}
1403
1404__be32
b85d4c01 1405nfsd4_sequence(struct svc_rqst *rqstp,
069b6ad4
AA		 1406 struct nfsd4_compound_state *cstate,
1407 struct nfsd4_sequence *seq)
1408{
f9bb94c4 1409 struct nfsd4_compoundres *resp = rqstp->rq_resp;
b85d4c01
BH		 1410 struct nfsd4_session *session;
1411 struct nfsd4_slot *slot;
1412 int status;
1413
f9bb94c4
AA		 1414 if (resp->opcnt != 1)
1415 return nfserr_sequence_pos;
1416
9089f1b4 1417 spin_lock(&client_lock);
b85d4c01
BH		 1418 status = nfserr_badsession;
1419 session = find_in_sessionid_hashtbl(&seq->sessionid);
1420 if (!session)
1421 goto out;
1422
1423 status = nfserr_badslot;
6c18ba9f 1424 if (seq->slotid >= session->se_fchannel.maxreqs)
b85d4c01
BH		 1425 goto out;
1426
557ce264 1427 slot = session->se_slots[seq->slotid];
b85d4c01
BH		 1428 dprintk("%s: slotid %d\n", __func__, seq->slotid);
1429
a8dfdaeb
AA		 1430 /* We do not negotiate the number of slots yet, so set the
1431 * maxslots to the session maxreqs which is used to encode
1432 * sr_highest_slotid and the sr_target_slot id to maxslots */
1433 seq->maxslots = session->se_fchannel.maxreqs;
1434
88e588d5 1435 status = check_slot_seqid(seq->seqid, slot->sl_seqid, slot->sl_inuse);
b85d4c01
BH		 1436 if (status == nfserr_replay_cache) {
1437 cstate->slot = slot;
1438 cstate->session = session;
da3846a2 1439 /* Return the cached reply status and set cstate->status
557ce264 1440 * for nfsd4_proc_compound processing */
bf864a31 1441 status = nfsd4_replay_cache_entry(resp, seq);
da3846a2 1442 cstate->status = nfserr_replay_cache;
aaf84eb9 1443 goto out;
b85d4c01
BH		 1444 }
1445 if (status)
1446 goto out;
1447
1448 /* Success! bump slot seqid */
1449 slot->sl_inuse = true;
1450 slot->sl_seqid = seq->seqid;
557ce264 1451 slot->sl_cachethis = seq->cachethis;
b85d4c01
BH		 1452
1453 cstate->slot = slot;
1454 cstate->session = session;
1455
b85d4c01 1456out:
26c0c75e 1457 /* Hold a session reference until done processing the compound. */
aaf84eb9 1458 if (cstate->session) {
36acb66b
BH		 1459 nfsd4_get_session(cstate->session);
1460 /* Renew the clientid on success and on replay */
1461 renew_client_locked(session->se_client);
aaf84eb9 1462 }
36acb66b 1463 spin_unlock(&client_lock);
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();
36acb66b
BH		 2572 INIT_LIST_HEAD(&reaplist);
2573 spin_lock(&client_lock);
1da177e4
LT		 2574 list_for_each_safe(pos, next, &client_lru) {
2575 clp = list_entry(pos, struct nfs4_client, cl_lru);
2576 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
2577 t = clp->cl_time - cutoff;
2578 if (clientid_val > t)
2579 clientid_val = t;
2580 break;
2581 }
36acb66b
BH		 2582 list_move(&clp->cl_lru, &reaplist);
2583 }
2584 spin_unlock(&client_lock);
2585 list_for_each_safe(pos, next, &reaplist) {
2586 clp = list_entry(pos, struct nfs4_client, cl_lru);
1da177e4
LT		 2587 dprintk("NFSD: purging unused client (clientid %08x)\n",
2588 clp->cl_clientid.cl_id);
c7b9a459 2589 nfsd4_remove_clid_dir(clp);
1da177e4
LT		 2590 expire_client(clp);
2591 }
1da177e4
LT		 2592 spin_lock(&recall_lock);
2593 list_for_each_safe(pos, next, &del_recall_lru) {
2594 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2595 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
2596 u = dp->dl_time - cutoff;
2597 if (test_val > u)
2598 test_val = u;
2599 break;
2600 }
2601 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
2602 dp, dp->dl_flock);
2603 list_move(&dp->dl_recall_lru, &reaplist);
2604 }
2605 spin_unlock(&recall_lock);
2606 list_for_each_safe(pos, next, &reaplist) {
2607 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
2608 list_del_init(&dp->dl_recall_lru);
2609 unhash_delegation(dp);
2610 }
cf07d2ea 2611 test_val = nfsd4_lease;
1da177e4
LT		 2612 list_for_each_safe(pos, next, &close_lru) {
2613 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
2614 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
2615 u = sop->so_time - cutoff;
2616 if (test_val > u)
2617 test_val = u;
2618 break;
2619 }
2620 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
2621 sop->so_id);
f044ff83 2622 release_openowner(sop);
1da177e4
LT		 2623 }
2624 if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
2625 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
2626 nfs4_unlock_state();
2627 return clientid_val;
2628}
2629
a254b246
HH		 2630static struct workqueue_struct *laundry_wq;
2631static void laundromat_main(struct work_struct *);
2632static DECLARE_DELAYED_WORK(laundromat_work, laundromat_main);
2633
2634static void
c4028958 2635laundromat_main(struct work_struct *not_used)
1da177e4
LT		 2636{
2637 time_t t;
2638
2639 t = nfs4_laundromat();
2640 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
58da282b 2641 queue_delayed_work(laundry_wq, &laundromat_work, t*HZ);
1da177e4
LT		 2642}
2643
fd39ca9a 2644static struct nfs4_stateowner *
f8816512
N		 2645search_close_lru(u32 st_id, int flags)
2646{
1da177e4
LT		 2647 struct nfs4_stateowner *local = NULL;
2648
1da177e4
LT		 2649 if (flags & CLOSE_STATE) {
2650 list_for_each_entry(local, &close_lru, so_close_lru) {
2651 if (local->so_id == st_id)
2652 return local;
2653 }
2654 }
2655 return NULL;
2656}
2657
2658static inline int
2659nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
2660{
7eaa36e2 2661 return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_path.dentry->d_inode;
1da177e4
LT		 2662}
2663
2664static int
2665STALE_STATEID(stateid_t *stateid)
2666{
78155ed7
BN		 2667 if (time_after((unsigned long)boot_time,
2668 (unsigned long)stateid->si_boot)) {
8c10cbdb
BH		 2669 dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
2670 STATEID_VAL(stateid));
78155ed7
BN		 2671 return 1;
2672 }
2673 return 0;
2674}
2675
2676static int
2677EXPIRED_STATEID(stateid_t *stateid)
2678{
2679 if (time_before((unsigned long)boot_time,
2680 ((unsigned long)stateid->si_boot)) &&
cf07d2ea 2681 time_before((unsigned long)(stateid->si_boot + nfsd4_lease), get_seconds())) {
8c10cbdb
BH		 2682 dprintk("NFSD: expired stateid " STATEID_FMT "!\n",
2683 STATEID_VAL(stateid));
78155ed7
BN		 2684 return 1;
2685 }
2686 return 0;
2687}
2688
2689static __be32
2690stateid_error_map(stateid_t *stateid)
2691{
2692 if (STALE_STATEID(stateid))
2693 return nfserr_stale_stateid;
2694 if (EXPIRED_STATEID(stateid))
2695 return nfserr_expired;
2696
8c10cbdb
BH		 2697 dprintk("NFSD: bad stateid " STATEID_FMT "!\n",
2698 STATEID_VAL(stateid));
78155ed7 2699 return nfserr_bad_stateid;
1da177e4
LT		 2700}
2701
2702static inline int
2703access_permit_read(unsigned long access_bmap)
2704{
2705 return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2706 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2707 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2708}
2709
2710static inline int
2711access_permit_write(unsigned long access_bmap)
2712{
2713 return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2714 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2715}
2716
2717static
b37ad28b 2718__be32 nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
1da177e4 2719{
b37ad28b 2720 __be32 status = nfserr_openmode;
1da177e4
LT		 2721
2722 if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2723 goto out;
2724 if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2725 goto out;
2726 status = nfs_ok;
2727out:
2728 return status;
2729}
2730
b37ad28b 2731static inline __be32
1da177e4
LT		 2732check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2733{
203a8c8e 2734 if (ONE_STATEID(stateid) && (flags & RD_STATE))
1da177e4 2735 return nfs_ok;
af558e33 2736 else if (locks_in_grace()) {
1da177e4
LT		 2737 /* Answer in remaining cases depends on existance of
2738 * conflicting state; so we must wait out the grace period. */
2739 return nfserr_grace;
2740 } else if (flags & WR_STATE)
2741 return nfs4_share_conflict(current_fh,
2742 NFS4_SHARE_DENY_WRITE);
2743 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2744 return nfs4_share_conflict(current_fh,
2745 NFS4_SHARE_DENY_READ);
2746}
2747
2748/*
2749 * Allow READ/WRITE during grace period on recovered state only for files
2750 * that are not able to provide mandatory locking.
2751 */
2752static inline int
18f82731 2753grace_disallows_io(struct inode *inode)
1da177e4 2754{
203a8c8e 2755 return locks_in_grace() && mandatory_lock(inode);
1da177e4
LT		 2756}
2757
6668958f 2758static int check_stateid_generation(stateid_t *in, stateid_t *ref, int flags)
0836f587 2759{
6668958f
AA		 2760 /*
2761 * When sessions are used the stateid generation number is ignored
2762 * when it is zero.
2763 */
2764 if ((flags & HAS_SESSION) && in->si_generation == 0)
2765 goto out;
2766
0836f587
BF		 2767 /* If the client sends us a stateid from the future, it's buggy: */
2768 if (in->si_generation > ref->si_generation)
2769 return nfserr_bad_stateid;
2770 /*
2771 * The following, however, can happen. For example, if the
2772 * client sends an open and some IO at the same time, the open
2773 * may bump si_generation while the IO is still in flight.
2774 * Thanks to hard links and renames, the client never knows what
2775 * file an open will affect. So it could avoid that situation
2776 * only by serializing all opens and IO from the same open
2777 * owner. To recover from the old_stateid error, the client
2778 * will just have to retry the IO:
2779 */
2780 if (in->si_generation < ref->si_generation)
2781 return nfserr_old_stateid;
6668958f 2782out:
0836f587
BF		 2783 return nfs_ok;
2784}
2785
3e633079
BF		 2786static int is_delegation_stateid(stateid_t *stateid)
2787{
2788 return stateid->si_fileid == 0;
2789}
2790
1da177e4
LT		 2791/*
2792* Checks for stateid operations
2793*/
b37ad28b 2794__be32
dd453dfd
BH		 2795nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2796 stateid_t *stateid, int flags, struct file **filpp)
1da177e4
LT		 2797{
2798 struct nfs4_stateid *stp = NULL;
2799 struct nfs4_delegation *dp = NULL;
dd453dfd 2800 struct svc_fh *current_fh = &cstate->current_fh;
1da177e4 2801 struct inode *ino = current_fh->fh_dentry->d_inode;
b37ad28b 2802 __be32 status;
1da177e4 2803
1da177e4
LT		 2804 if (filpp)
2805 *filpp = NULL;
2806
18f82731 2807 if (grace_disallows_io(ino))
1da177e4
LT		 2808 return nfserr_grace;
2809
6668958f
AA		 2810 if (nfsd4_has_session(cstate))
2811 flags |= HAS_SESSION;
2812
1da177e4
LT		 2813 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2814 return check_special_stateids(current_fh, stateid, flags);
2815
1da177e4
LT		 2816 status = nfserr_stale_stateid;
2817 if (STALE_STATEID(stateid))
2818 goto out;
2819
1da177e4 2820 status = nfserr_bad_stateid;
3e633079 2821 if (is_delegation_stateid(stateid)) {
a4455be0 2822 dp = find_delegation_stateid(ino, stateid);
78155ed7
BN		 2823 if (!dp) {
2824 status = stateid_error_map(stateid);
1da177e4 2825 goto out;
78155ed7 2826 }
6668958f
AA		 2827 status = check_stateid_generation(stateid, &dp->dl_stateid,
2828 flags);
0c2a498f
BF		 2829 if (status)
2830 goto out;
dc9bf700
BF		 2831 status = nfs4_check_delegmode(dp, flags);
2832 if (status)
2833 goto out;
2834 renew_client(dp->dl_client);
dc9bf700
BF		 2835 if (filpp)
2836 *filpp = dp->dl_vfs_file;
1da177e4 2837 } else { /* open or lock stateid */
a4455be0 2838 stp = find_stateid(stateid, flags);
78155ed7
BN		 2839 if (!stp) {
2840 status = stateid_error_map(stateid);
1da177e4 2841 goto out;
78155ed7 2842 }
6150ef0d 2843 if (nfs4_check_fh(current_fh, stp))
1da177e4
LT		 2844 goto out;
2845 if (!stp->st_stateowner->so_confirmed)
2846 goto out;
6668958f
AA		 2847 status = check_stateid_generation(stateid, &stp->st_stateid,
2848 flags);
0c2a498f
BF		 2849 if (status)
2850 goto out;
a4455be0
BF		 2851 status = nfs4_check_openmode(stp, flags);
2852 if (status)
1da177e4
LT		 2853 goto out;
2854 renew_client(stp->st_stateowner->so_client);
2855 if (filpp)
2856 *filpp = stp->st_vfs_file;
1da177e4
LT		 2857 }
2858 status = nfs_ok;
2859out:
2860 return status;
2861}
2862
4c4cd222
N		 2863static inline int
2864setlkflg (int type)
2865{
2866 return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
2867 RD_STATE : WR_STATE;
2868}
1da177e4
LT		 2869
2870/*
2871 * Checks for sequence id mutating operations.
2872 */
b37ad28b 2873static __be32
dd453dfd
BH		 2874nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
2875 stateid_t *stateid, int flags,
2876 struct nfs4_stateowner **sopp,
2877 struct nfs4_stateid **stpp, struct nfsd4_lock *lock)
1da177e4 2878{
1da177e4
LT		 2879 struct nfs4_stateid *stp;
2880 struct nfs4_stateowner *sop;
dd453dfd 2881 struct svc_fh *current_fh = &cstate->current_fh;
0836f587 2882 __be32 status;
1da177e4 2883
8c10cbdb
BH		 2884 dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
2885 seqid, STATEID_VAL(stateid));
3a4f98bb 2886
1da177e4
LT		 2887 *stpp = NULL;
2888 *sopp = NULL;
2889
1da177e4 2890 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2fdada03 2891 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
3a4f98bb 2892 return nfserr_bad_stateid;
1da177e4
LT		 2893 }
2894
1da177e4 2895 if (STALE_STATEID(stateid))
3a4f98bb 2896 return nfserr_stale_stateid;
6668958f
AA		 2897
2898 if (nfsd4_has_session(cstate))
2899 flags |= HAS_SESSION;
2900
1da177e4
LT		 2901 /*
2902 * We return BAD_STATEID if filehandle doesn't match stateid,
2903 * the confirmed flag is incorrecly set, or the generation
2904 * number is incorrect.
1da177e4 2905 */
f8816512
N		 2906 stp = find_stateid(stateid, flags);
2907 if (stp == NULL) {
2908 /*
2909 * Also, we should make sure this isn't just the result of
2910 * a replayed close:
2911 */
2912 sop = search_close_lru(stateid->si_stateownerid, flags);
2913 if (sop == NULL)
78155ed7 2914 return stateid_error_map(stateid);
f8816512
N		 2915 *sopp = sop;
2916 goto check_replay;
2917 }
1da177e4 2918
39325bd0
BF		 2919 *stpp = stp;
2920 *sopp = sop = stp->st_stateowner;
2921
4c4cd222 2922 if (lock) {
4c4cd222 2923 clientid_t *lockclid = &lock->v.new.clientid;
1da177e4 2924 struct nfs4_client *clp = sop->so_client;
4c4cd222 2925 int lkflg = 0;
b37ad28b 2926 __be32 status;
4c4cd222
N		 2927
2928 lkflg = setlkflg(lock->lk_type);
2929
2930 if (lock->lk_is_new) {
599e0a22
BF		 2931 if (!sop->so_is_open_owner)
2932 return nfserr_bad_stateid;
60adfc50
AA		 2933 if (!(flags & HAS_SESSION) &&
2934 !same_clid(&clp->cl_clientid, lockclid))
2935 return nfserr_bad_stateid;
599e0a22
BF		 2936 /* stp is the open stateid */
2937 status = nfs4_check_openmode(stp, lkflg);
2938 if (status)
2939 return status;
2940 } else {
2941 /* stp is the lock stateid */
2942 status = nfs4_check_openmode(stp->st_openstp, lkflg);
2943 if (status)
2944 return status;
4c4cd222 2945 }
1da177e4
LT		 2946 }
2947
f3362737 2948 if (nfs4_check_fh(current_fh, stp)) {
2fdada03 2949 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
3a4f98bb 2950 return nfserr_bad_stateid;
1da177e4
LT		 2951 }
2952
1da177e4
LT		 2953 /*
2954 * We now validate the seqid and stateid generation numbers.
2955 * For the moment, we ignore the possibility of
2956 * generation number wraparound.
2957 */
6668958f 2958 if (!(flags & HAS_SESSION) && seqid != sop->so_seqid)
1da177e4
LT		 2959 goto check_replay;
2960
3a4f98bb 2961 if (sop->so_confirmed && flags & CONFIRM) {
2fdada03 2962 dprintk("NFSD: preprocess_seqid_op: expected"
3a4f98bb
N		 2963 " unconfirmed stateowner!\n");
2964 return nfserr_bad_stateid;
1da177e4 2965 }
3a4f98bb 2966 if (!sop->so_confirmed && !(flags & CONFIRM)) {
2fdada03 2967 dprintk("NFSD: preprocess_seqid_op: stateowner not"
3a4f98bb
N		 2968 " confirmed yet!\n");
2969 return nfserr_bad_stateid;
1da177e4 2970 }
6668958f 2971 status = check_stateid_generation(stateid, &stp->st_stateid, flags);
0836f587
BF		 2972 if (status)
2973 return status;
52fd004e 2974 renew_client(sop->so_client);
3a4f98bb 2975 return nfs_ok;
1da177e4 2976
1da177e4 2977check_replay:
bd9aac52 2978 if (seqid == sop->so_seqid - 1) {
849823c5 2979 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
1da177e4 2980 /* indicate replay to calling function */
a90b061c 2981 return nfserr_replay_me;
1da177e4 2982 }
2fdada03 2983 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
3a4f98bb
N		 2984 sop->so_seqid, seqid);
2985 *sopp = NULL;
2986 return nfserr_bad_seqid;
1da177e4
LT		 2987}
2988
b37ad28b 2989__be32
ca364317 2990nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 2991 struct nfsd4_open_confirm *oc)
1da177e4 2992{
b37ad28b 2993 __be32 status;
1da177e4
LT		 2994 struct nfs4_stateowner *sop;
2995 struct nfs4_stateid *stp;
2996
2997 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
ca364317
BF		 2998 (int)cstate->current_fh.fh_dentry->d_name.len,
2999 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3000
ca364317 3001 status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
a8cddc5d
BF		 3002 if (status)
3003 return status;
1da177e4
LT		 3004
3005 nfs4_lock_state();
3006
dd453dfd 3007 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3008 oc->oc_seqid, &oc->oc_req_stateid,
f3362737 3009 CONFIRM | OPEN_STATE,
1da177e4
LT		 3010 &oc->oc_stateowner, &stp, NULL)))
3011 goto out;
3012
3013 sop = oc->oc_stateowner;
3014 sop->so_confirmed = 1;
3015 update_stateid(&stp->st_stateid);
3016 memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
8c10cbdb
BH		 3017 dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
3018 __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stateid));
c7b9a459
N		 3019
3020 nfsd4_create_clid_dir(sop->so_client);
1da177e4 3021out:
f2327d9a 3022 if (oc->oc_stateowner) {
1da177e4 3023 nfs4_get_stateowner(oc->oc_stateowner);
a4f1706a 3024 cstate->replay_owner = oc->oc_stateowner;
f2327d9a 3025 }
1da177e4
LT		 3026 nfs4_unlock_state();
3027 return status;
3028}
3029
3030
3031/*
3032 * unset all bits in union bitmap (bmap) that
3033 * do not exist in share (from successful OPEN_DOWNGRADE)
3034 */
3035static void
3036reset_union_bmap_access(unsigned long access, unsigned long *bmap)
3037{
3038 int i;
3039 for (i = 1; i < 4; i++) {
3040 if ((i & access) != i)
3041 __clear_bit(i, bmap);
3042 }
3043}
3044
3045static void
3046reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
3047{
3048 int i;
3049 for (i = 0; i < 4; i++) {
3050 if ((i & deny) != i)
3051 __clear_bit(i, bmap);
3052 }
3053}
3054
b37ad28b 3055__be32
ca364317
BF		 3056nfsd4_open_downgrade(struct svc_rqst *rqstp,
3057 struct nfsd4_compound_state *cstate,
a4f1706a 3058 struct nfsd4_open_downgrade *od)
1da177e4 3059{
b37ad28b 3060 __be32 status;
1da177e4
LT		 3061 struct nfs4_stateid *stp;
3062 unsigned int share_access;
3063
3064 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
ca364317
BF		 3065 (int)cstate->current_fh.fh_dentry->d_name.len,
3066 cstate->current_fh.fh_dentry->d_name.name);
1da177e4 3067
d87a8ade 3068 if (!access_valid(od->od_share_access, cstate->minorversion)
ba5a6a19 3069 || !deny_valid(od->od_share_deny))
1da177e4
LT		 3070 return nfserr_inval;
3071
3072 nfs4_lock_state();
dd453dfd 3073 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3074 od->od_seqid,
1da177e4 3075 &od->od_stateid,
f3362737 3076 OPEN_STATE,
1da177e4
LT		 3077 &od->od_stateowner, &stp, NULL)))
3078 goto out;
3079
3080 status = nfserr_inval;
3081 if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
3082 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
3083 stp->st_access_bmap, od->od_share_access);
3084 goto out;
3085 }
3086 if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
3087 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
3088 stp->st_deny_bmap, od->od_share_deny);
3089 goto out;
3090 }
3091 set_access(&share_access, stp->st_access_bmap);
3092 nfs4_file_downgrade(stp->st_vfs_file,
3093 share_access & ~od->od_share_access);
3094
3095 reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
3096 reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
3097
3098 update_stateid(&stp->st_stateid);
3099 memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
3100 status = nfs_ok;
3101out:
f2327d9a 3102 if (od->od_stateowner) {
1da177e4 3103 nfs4_get_stateowner(od->od_stateowner);
a4f1706a 3104 cstate->replay_owner = od->od_stateowner;
f2327d9a 3105 }
1da177e4
LT		 3106 nfs4_unlock_state();
3107 return status;
3108}
3109
3110/*
3111 * nfs4_unlock_state() called after encode
3112 */
b37ad28b 3113__be32
ca364317 3114nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3115 struct nfsd4_close *close)
1da177e4 3116{
b37ad28b 3117 __be32 status;
1da177e4
LT		 3118 struct nfs4_stateid *stp;
3119
3120 dprintk("NFSD: nfsd4_close on file %.*s\n",
ca364317
BF		 3121 (int)cstate->current_fh.fh_dentry->d_name.len,
3122 cstate->current_fh.fh_dentry->d_name.name);
1da177e4
LT		 3123
3124 nfs4_lock_state();
3125 /* check close_lru for replay */
dd453dfd 3126 if ((status = nfs4_preprocess_seqid_op(cstate,
ca364317 3127 close->cl_seqid,
1da177e4 3128 &close->cl_stateid,
f3362737 3129 OPEN_STATE | CLOSE_STATE,
1da177e4
LT		 3130 &close->cl_stateowner, &stp, NULL)))
3131 goto out;
1da177e4
LT		 3132 status = nfs_ok;
3133 update_stateid(&stp->st_stateid);
3134 memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
3135
04ef5954 3136 /* release_stateid() calls nfsd_close() if needed */
2283963f 3137 release_open_stateid(stp);
04ef5954
BF		 3138
3139 /* place unused nfs4_stateowners on so_close_lru list to be
3140 * released by the laundromat service after the lease period
3141 * to enable us to handle CLOSE replay
3142 */
3143 if (list_empty(&close->cl_stateowner->so_stateids))
3144 move_to_close_lru(close->cl_stateowner);
1da177e4 3145out:
f2327d9a 3146 if (close->cl_stateowner) {
1da177e4 3147 nfs4_get_stateowner(close->cl_stateowner);
a4f1706a 3148 cstate->replay_owner = close->cl_stateowner;
f2327d9a 3149 }
1da177e4
LT		 3150 nfs4_unlock_state();
3151 return status;
3152}
3153
b37ad28b 3154__be32
ca364317
BF		 3155nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3156 struct nfsd4_delegreturn *dr)
1da177e4 3157{
203a8c8e
BF		 3158 struct nfs4_delegation *dp;
3159 stateid_t *stateid = &dr->dr_stateid;
3160 struct inode *inode;
b37ad28b 3161 __be32 status;
6668958f 3162 int flags = 0;
1da177e4 3163
ca364317 3164 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
203a8c8e
BF		 3165 return status;
3166 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4 3167
6668958f
AA		 3168 if (nfsd4_has_session(cstate))
3169 flags |= HAS_SESSION;
1da177e4 3170 nfs4_lock_state();
203a8c8e
BF		 3171 status = nfserr_bad_stateid;
3172 if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
3173 goto out;
3174 status = nfserr_stale_stateid;
3175 if (STALE_STATEID(stateid))
3176 goto out;
7e0f7cf5 3177 status = nfserr_bad_stateid;
203a8c8e
BF		 3178 if (!is_delegation_stateid(stateid))
3179 goto out;
203a8c8e 3180 dp = find_delegation_stateid(inode, stateid);
78155ed7
BN		 3181 if (!dp) {
3182 status = stateid_error_map(stateid);
203a8c8e 3183 goto out;
78155ed7 3184 }
6668958f 3185 status = check_stateid_generation(stateid, &dp->dl_stateid, flags);
203a8c8e
BF		 3186 if (status)
3187 goto out;
3188 renew_client(dp->dl_client);
3189
3190 unhash_delegation(dp);
1da177e4 3191out:
203a8c8e
BF		 3192 nfs4_unlock_state();
3193
1da177e4
LT		 3194 return status;
3195}
3196
3197
3198/*
3199 * Lock owner state (byte-range locks)
3200 */
3201#define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
3202#define LOCK_HASH_BITS 8
3203#define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
3204#define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
3205
87df4de8
BH		 3206static inline u64
3207end_offset(u64 start, u64 len)
3208{
3209 u64 end;
3210
3211 end = start + len;
3212 return end >= start ? end: NFS4_MAX_UINT64;
3213}
3214
3215/* last octet in a range */
3216static inline u64
3217last_byte_offset(u64 start, u64 len)
3218{
3219 u64 end;
3220
3221 BUG_ON(!len);
3222 end = start + len;
3223 return end > start ? end - 1: NFS4_MAX_UINT64;
3224}
3225
1da177e4
LT		 3226#define lockownerid_hashval(id) \
3227 ((id) & LOCK_HASH_MASK)
3228
3229static inline unsigned int
3230lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
3231 struct xdr_netobj *ownername)
3232{
3233 return (file_hashval(inode) + cl_id
3234 + opaque_hashval(ownername->data, ownername->len))
3235 & LOCK_HASH_MASK;
3236}
3237
3238static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
3239static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
3240static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
3241
fd39ca9a 3242static struct nfs4_stateid *
1da177e4
LT		 3243find_stateid(stateid_t *stid, int flags)
3244{
9346eff0 3245 struct nfs4_stateid *local;
1da177e4
LT		 3246 u32 st_id = stid->si_stateownerid;
3247 u32 f_id = stid->si_fileid;
3248 unsigned int hashval;
3249
3250 dprintk("NFSD: find_stateid flags 0x%x\n",flags);
9346eff0 3251 if (flags & (LOCK_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3252 hashval = stateid_hashval(st_id, f_id);
3253 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
3254 if ((local->st_stateid.si_stateownerid == st_id) &&
3255 (local->st_stateid.si_fileid == f_id))
3256 return local;
3257 }
3258 }
9346eff0
KK		 3259
3260 if (flags & (OPEN_STATE | RD_STATE | WR_STATE)) {
1da177e4
LT		 3261 hashval = stateid_hashval(st_id, f_id);
3262 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
3263 if ((local->st_stateid.si_stateownerid == st_id) &&
3264 (local->st_stateid.si_fileid == f_id))
3265 return local;
3266 }
849823c5 3267 }
1da177e4
LT		 3268 return NULL;
3269}
3270
3271static struct nfs4_delegation *
3272find_delegation_stateid(struct inode *ino, stateid_t *stid)
3273{
13cd2184
N		 3274 struct nfs4_file *fp;
3275 struct nfs4_delegation *dl;
1da177e4 3276
8c10cbdb
BH		 3277 dprintk("NFSD: %s: stateid=" STATEID_FMT "\n", __func__,
3278 STATEID_VAL(stid));
1da177e4 3279
1da177e4 3280 fp = find_file(ino);
13cd2184
N		 3281 if (!fp)
3282 return NULL;
3283 dl = find_delegation_file(fp, stid);
3284 put_nfs4_file(fp);
3285 return dl;
1da177e4
LT		 3286}
3287
3288/*
3289 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
3290 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
3291 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
3292 * locking, this prevents us from being completely protocol-compliant. The
3293 * real solution to this problem is to start using unsigned file offsets in
3294 * the VFS, but this is a very deep change!
3295 */
3296static inline void
3297nfs4_transform_lock_offset(struct file_lock *lock)
3298{
3299 if (lock->fl_start < 0)
3300 lock->fl_start = OFFSET_MAX;
3301 if (lock->fl_end < 0)
3302 lock->fl_end = OFFSET_MAX;
3303}
3304
d5b9026a
N		 3305/* Hack!: For now, we're defining this just so we can use a pointer to it
3306 * as a unique cookie to identify our (NFSv4's) posix locks. */
7b021967 3307static const struct lock_manager_operations nfsd_posix_mng_ops = {
d5b9026a 3308};
1da177e4
LT		 3309
3310static inline void
3311nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
3312{
d5b9026a
N		 3313 struct nfs4_stateowner *sop;
3314 unsigned int hval;
1da177e4 3315
d5b9026a
N		 3316 if (fl->fl_lmops == &nfsd_posix_mng_ops) {
3317 sop = (struct nfs4_stateowner *) fl->fl_owner;
3318 hval = lockownerid_hashval(sop->so_id);
1da177e4
LT		 3319 kref_get(&sop->so_ref);
3320 deny->ld_sop = sop;
3321 deny->ld_clientid = sop->so_client->cl_clientid;
d5b9026a
N		 3322 } else {
3323 deny->ld_sop = NULL;
3324 deny->ld_clientid.cl_boot = 0;
3325 deny->ld_clientid.cl_id = 0;
1da177e4
LT		 3326 }
3327 deny->ld_start = fl->fl_start;
87df4de8
BH		 3328 deny->ld_length = NFS4_MAX_UINT64;
3329 if (fl->fl_end != NFS4_MAX_UINT64)
1da177e4
LT		 3330 deny->ld_length = fl->fl_end - fl->fl_start + 1;
3331 deny->ld_type = NFS4_READ_LT;
3332 if (fl->fl_type != F_RDLCK)
3333 deny->ld_type = NFS4_WRITE_LT;
3334}
3335
1da177e4
LT		 3336static struct nfs4_stateowner *
3337find_lockstateowner_str(struct inode *inode, clientid_t *clid,
3338 struct xdr_netobj *owner)
3339{
3340 unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
3341 struct nfs4_stateowner *op;
3342
3343 list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
599e0a22 3344 if (same_owner_str(op, owner, clid))
1da177e4
LT		 3345 return op;
3346 }
3347 return NULL;
3348}
3349
3350/*
3351 * Alloc a lock owner structure.
3352 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
3353 * occured.
3354 *
3355 * strhashval = lock_ownerstr_hashval
1da177e4
LT		 3356 */
3357
3358static struct nfs4_stateowner *
3359alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
3360 struct nfs4_stateowner *sop;
3361 struct nfs4_replay *rp;
3362 unsigned int idhashval;
3363
3364 if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
3365 return NULL;
3366 idhashval = lockownerid_hashval(current_ownerid);
3367 INIT_LIST_HEAD(&sop->so_idhash);
3368 INIT_LIST_HEAD(&sop->so_strhash);
3369 INIT_LIST_HEAD(&sop->so_perclient);
ea1da636
N		 3370 INIT_LIST_HEAD(&sop->so_stateids);
3371 INIT_LIST_HEAD(&sop->so_perstateid);
1da177e4
LT		 3372 INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
3373 sop->so_time = 0;
3374 list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
3375 list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
ea1da636 3376 list_add(&sop->so_perstateid, &open_stp->st_lockowners);
1da177e4
LT		 3377 sop->so_is_open_owner = 0;
3378 sop->so_id = current_ownerid++;
3379 sop->so_client = clp;
b59e3c0e
NB		 3380 /* It is the openowner seqid that will be incremented in encode in the
3381 * case of new lockowners; so increment the lock seqid manually: */
3382 sop->so_seqid = lock->lk_new_lock_seqid + 1;
1da177e4
LT		 3383 sop->so_confirmed = 1;
3384 rp = &sop->so_replay;
de1ae286 3385 rp->rp_status = nfserr_serverfault;
1da177e4
LT		 3386 rp->rp_buflen = 0;
3387 rp->rp_buf = rp->rp_ibuf;
3388 return sop;
3389}
3390
fd39ca9a 3391static struct nfs4_stateid *
1da177e4
LT		 3392alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
3393{
3394 struct nfs4_stateid *stp;
3395 unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
3396
5ac049ac
N		 3397 stp = nfs4_alloc_stateid();
3398 if (stp == NULL)
1da177e4
LT		 3399 goto out;
3400 INIT_LIST_HEAD(&stp->st_hash);
3401 INIT_LIST_HEAD(&stp->st_perfile);
ea1da636
N		 3402 INIT_LIST_HEAD(&stp->st_perstateowner);
3403 INIT_LIST_HEAD(&stp->st_lockowners); /* not used */
1da177e4 3404 list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
8beefa24 3405 list_add(&stp->st_perfile, &fp->fi_stateids);
ea1da636 3406 list_add(&stp->st_perstateowner, &sop->so_stateids);
1da177e4 3407 stp->st_stateowner = sop;
13cd2184 3408 get_nfs4_file(fp);
1da177e4 3409 stp->st_file = fp;
78155ed7 3410 stp->st_stateid.si_boot = get_seconds();
1da177e4
LT		 3411 stp->st_stateid.si_stateownerid = sop->so_id;
3412 stp->st_stateid.si_fileid = fp->fi_id;
3413 stp->st_stateid.si_generation = 0;
3414 stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
3415 stp->st_access_bmap = open_stp->st_access_bmap;
3416 stp->st_deny_bmap = open_stp->st_deny_bmap;
4c4cd222 3417 stp->st_openstp = open_stp;
1da177e4
LT		 3418
3419out:
3420 return stp;
3421}
3422
fd39ca9a 3423static int
1da177e4
LT		 3424check_lock_length(u64 offset, u64 length)
3425{
87df4de8 3426 return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
1da177e4
LT		 3427 LOFF_OVERFLOW(offset, length)));
3428}
3429
3430/*
3431 * LOCK operation
3432 */
b37ad28b 3433__be32
ca364317 3434nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3435 struct nfsd4_lock *lock)
1da177e4 3436{
3e9e3dbe 3437 struct nfs4_stateowner *open_sop = NULL;
b59e3c0e 3438 struct nfs4_stateowner *lock_sop = NULL;
1da177e4
LT		 3439 struct nfs4_stateid *lock_stp;
3440 struct file *filp;
3441 struct file_lock file_lock;
8dc7c311 3442 struct file_lock conflock;
b37ad28b 3443 __be32 status = 0;
1da177e4 3444 unsigned int strhashval;
150b3934 3445 unsigned int cmd;
b8dd7b9a 3446 int err;
1da177e4
LT		 3447
3448 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
3449 (long long) lock->lk_offset,
3450 (long long) lock->lk_length);
3451
1da177e4
LT		 3452 if (check_lock_length(lock->lk_offset, lock->lk_length))
3453 return nfserr_inval;
3454
ca364317 3455 if ((status = fh_verify(rqstp, &cstate->current_fh,
8837abca 3456 S_IFREG, NFSD_MAY_LOCK))) {
a6f6ef2f
AA		 3457 dprintk("NFSD: nfsd4_lock: permission denied!\n");
3458 return status;
3459 }
3460
1da177e4
LT		 3461 nfs4_lock_state();
3462
3463 if (lock->lk_is_new) {
893f8770
N		 3464 /*
3465 * Client indicates that this is a new lockowner.
3466 * Use open owner and open stateid to create lock owner and
3467 * lock stateid.
3468 */
1da177e4
LT		 3469 struct nfs4_stateid *open_stp = NULL;
3470 struct nfs4_file *fp;
3471
3472 status = nfserr_stale_clientid;
60adfc50
AA		 3473 if (!nfsd4_has_session(cstate) &&
3474 STALE_CLIENTID(&lock->lk_new_clientid))
1da177e4 3475 goto out;
1da177e4 3476
1da177e4 3477 /* validate and update open stateid and open seqid */
dd453dfd 3478 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3479 lock->lk_new_open_seqid,
3480 &lock->lk_new_open_stateid,
f3362737 3481 OPEN_STATE,
3a65588a 3482 &lock->lk_replay_owner, &open_stp,
b59e3c0e 3483 lock);
37515177 3484 if (status)
1da177e4 3485 goto out;
3a65588a 3486 open_sop = lock->lk_replay_owner;
1da177e4
LT		 3487 /* create lockowner and lock stateid */
3488 fp = open_stp->st_file;
3489 strhashval = lock_ownerstr_hashval(fp->fi_inode,
3490 open_sop->so_client->cl_clientid.cl_id,
3491 &lock->v.new.owner);
3e9e3dbe
N		 3492 /* XXX: Do we need to check for duplicate stateowners on
3493 * the same file, or should they just be allowed (and
3494 * create new stateids)? */
1da177e4 3495 status = nfserr_resource;
b59e3c0e
NB		 3496 lock_sop = alloc_init_lock_stateowner(strhashval,
3497 open_sop->so_client, open_stp, lock);
3498 if (lock_sop == NULL)
1da177e4 3499 goto out;
b59e3c0e 3500 lock_stp = alloc_init_lock_stateid(lock_sop, fp, open_stp);
8a280510 3501 if (lock_stp == NULL)
1da177e4 3502 goto out;
1da177e4
LT		 3503 } else {
3504 /* lock (lock owner + lock stateid) already exists */
dd453dfd 3505 status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3506 lock->lk_old_lock_seqid,
3507 &lock->lk_old_lock_stateid,
f3362737 3508 LOCK_STATE,
3a65588a 3509 &lock->lk_replay_owner, &lock_stp, lock);
1da177e4
LT		 3510 if (status)
3511 goto out;
3a65588a 3512 lock_sop = lock->lk_replay_owner;
1da177e4 3513 }
3a65588a 3514 /* lock->lk_replay_owner and lock_stp have been created or found */
1da177e4
LT		 3515 filp = lock_stp->st_vfs_file;
3516
0dd395dc 3517 status = nfserr_grace;
af558e33 3518 if (locks_in_grace() && !lock->lk_reclaim)
0dd395dc
N		 3519 goto out;
3520 status = nfserr_no_grace;
af558e33 3521 if (!locks_in_grace() && lock->lk_reclaim)
0dd395dc
N		 3522 goto out;
3523
1da177e4
LT		 3524 locks_init_lock(&file_lock);
3525 switch (lock->lk_type) {
3526 case NFS4_READ_LT:
3527 case NFS4_READW_LT:
3528 file_lock.fl_type = F_RDLCK;
150b3934 3529 cmd = F_SETLK;
1da177e4
LT		 3530 break;
3531 case NFS4_WRITE_LT:
3532 case NFS4_WRITEW_LT:
3533 file_lock.fl_type = F_WRLCK;
150b3934 3534 cmd = F_SETLK;
1da177e4
LT		 3535 break;
3536 default:
3537 status = nfserr_inval;
3538 goto out;
3539 }
b59e3c0e 3540 file_lock.fl_owner = (fl_owner_t)lock_sop;
1da177e4
LT		 3541 file_lock.fl_pid = current->tgid;
3542 file_lock.fl_file = filp;
3543 file_lock.fl_flags = FL_POSIX;
d5b9026a 3544 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3545
3546 file_lock.fl_start = lock->lk_offset;
87df4de8 3547 file_lock.fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
1da177e4
LT		 3548 nfs4_transform_lock_offset(&file_lock);
3549
3550 /*
3551 * Try to lock the file in the VFS.
3552 * Note: locks.c uses the BKL to protect the inode's lock list.
3553 */
3554
fd85b817 3555 err = vfs_lock_file(filp, cmd, &file_lock, &conflock);
b8dd7b9a 3556 switch (-err) {
1da177e4
LT		 3557 case 0: /* success! */
3558 update_stateid(&lock_stp->st_stateid);
3559 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
3560 sizeof(stateid_t));
b8dd7b9a 3561 status = 0;
eb76b3fd
AA		 3562 break;
3563 case (EAGAIN): /* conflock holds conflicting lock */
3564 status = nfserr_denied;
3565 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
3566 nfs4_set_lock_denied(&conflock, &lock->lk_denied);
3567 break;
1da177e4
LT		 3568 case (EDEADLK):
3569 status = nfserr_deadlock;
eb76b3fd 3570 break;
1da177e4 3571 default:
fd85b817 3572 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
eb76b3fd
AA		 3573 status = nfserr_resource;
3574 break;
1da177e4 3575 }
1da177e4 3576out:
8a280510 3577 if (status && lock->lk_is_new && lock_sop)
f044ff83 3578 release_lockowner(lock_sop);
3a65588a
BF		 3579 if (lock->lk_replay_owner) {
3580 nfs4_get_stateowner(lock->lk_replay_owner);
a4f1706a 3581 cstate->replay_owner = lock->lk_replay_owner;
f2327d9a 3582 }
1da177e4
LT		 3583 nfs4_unlock_state();
3584 return status;
3585}
3586
55ef1274
BF		 3587/*
3588 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
3589 * so we do a temporary open here just to get an open file to pass to
3590 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
3591 * inode operation.)
3592 */
3593static int nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
3594{
3595 struct file *file;
3596 int err;
3597
3598 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
3599 if (err)
3600 return err;
3601 err = vfs_test_lock(file, lock);
3602 nfsd_close(file);
3603 return err;
3604}
3605
1da177e4
LT		 3606/*
3607 * LOCKT operation
3608 */
b37ad28b 3609__be32
ca364317
BF		 3610nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3611 struct nfsd4_lockt *lockt)
1da177e4
LT		 3612{
3613 struct inode *inode;
1da177e4 3614 struct file_lock file_lock;
fd85b817 3615 int error;
b37ad28b 3616 __be32 status;
1da177e4 3617
af558e33 3618 if (locks_in_grace())
1da177e4
LT		 3619 return nfserr_grace;
3620
3621 if (check_lock_length(lockt->lt_offset, lockt->lt_length))
3622 return nfserr_inval;
3623
3624 lockt->lt_stateowner = NULL;
3625 nfs4_lock_state();
3626
3627 status = nfserr_stale_clientid;
60adfc50 3628 if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
1da177e4 3629 goto out;
1da177e4 3630
ca364317 3631 if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0))) {
849823c5 3632 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
1da177e4
LT		 3633 if (status == nfserr_symlink)
3634 status = nfserr_inval;
3635 goto out;
3636 }
3637
ca364317 3638 inode = cstate->current_fh.fh_dentry->d_inode;
1da177e4
LT		 3639 locks_init_lock(&file_lock);
3640 switch (lockt->lt_type) {
3641 case NFS4_READ_LT:
3642 case NFS4_READW_LT:
3643 file_lock.fl_type = F_RDLCK;
3644 break;
3645 case NFS4_WRITE_LT:
3646 case NFS4_WRITEW_LT:
3647 file_lock.fl_type = F_WRLCK;
3648 break;
3649 default:
2fdada03 3650 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
1da177e4
LT		 3651 status = nfserr_inval;
3652 goto out;
3653 }
3654
3655 lockt->lt_stateowner = find_lockstateowner_str(inode,
3656 &lockt->lt_clientid, &lockt->lt_owner);
3657 if (lockt->lt_stateowner)
3658 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
3659 file_lock.fl_pid = current->tgid;
3660 file_lock.fl_flags = FL_POSIX;
3661
3662 file_lock.fl_start = lockt->lt_offset;
87df4de8 3663 file_lock.fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
1da177e4
LT		 3664
3665 nfs4_transform_lock_offset(&file_lock);
3666
1da177e4 3667 status = nfs_ok;
55ef1274 3668 error = nfsd_test_lock(rqstp, &cstate->current_fh, &file_lock);
fd85b817
ME		 3669 if (error) {
3670 status = nfserrno(error);
3671 goto out;
3672 }
9d6a8c5c 3673 if (file_lock.fl_type != F_UNLCK) {
1da177e4 3674 status = nfserr_denied;
9d6a8c5c 3675 nfs4_set_lock_denied(&file_lock, &lockt->lt_denied);
1da177e4
LT		 3676 }
3677out:
3678 nfs4_unlock_state();
3679 return status;
3680}
3681
b37ad28b 3682__be32
ca364317 3683nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
a4f1706a 3684 struct nfsd4_locku *locku)
1da177e4
LT		 3685{
3686 struct nfs4_stateid *stp;
3687 struct file *filp = NULL;
3688 struct file_lock file_lock;
b37ad28b 3689 __be32 status;
b8dd7b9a 3690 int err;
1da177e4
LT		 3691
3692 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
3693 (long long) locku->lu_offset,
3694 (long long) locku->lu_length);
3695
3696 if (check_lock_length(locku->lu_offset, locku->lu_length))
3697 return nfserr_inval;
3698
3699 nfs4_lock_state();
3700
dd453dfd 3701 if ((status = nfs4_preprocess_seqid_op(cstate,
1da177e4
LT		 3702 locku->lu_seqid,
3703 &locku->lu_stateid,
f3362737 3704 LOCK_STATE,
1da177e4
LT		 3705 &locku->lu_stateowner, &stp, NULL)))
3706 goto out;
3707
3708 filp = stp->st_vfs_file;
3709 BUG_ON(!filp);
3710 locks_init_lock(&file_lock);
3711 file_lock.fl_type = F_UNLCK;
3712 file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
3713 file_lock.fl_pid = current->tgid;
3714 file_lock.fl_file = filp;
3715 file_lock.fl_flags = FL_POSIX;
d5b9026a 3716 file_lock.fl_lmops = &nfsd_posix_mng_ops;
1da177e4
LT		 3717 file_lock.fl_start = locku->lu_offset;
3718
87df4de8 3719 file_lock.fl_end = last_byte_offset(locku->lu_offset, locku->lu_length);
1da177e4
LT		 3720 nfs4_transform_lock_offset(&file_lock);
3721
3722 /*
3723 * Try to unlock the file in the VFS.
3724 */
fd85b817 3725 err = vfs_lock_file(filp, F_SETLK, &file_lock, NULL);
b8dd7b9a 3726 if (err) {
fd85b817 3727 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
1da177e4
LT		 3728 goto out_nfserr;
3729 }
3730 /*
3731 * OK, unlock succeeded; the only thing left to do is update the stateid.
3732 */
3733 update_stateid(&stp->st_stateid);
3734 memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3735
3736out:
f2327d9a 3737 if (locku->lu_stateowner) {
1da177e4 3738 nfs4_get_stateowner(locku->lu_stateowner);
a4f1706a 3739 cstate->replay_owner = locku->lu_stateowner;
f2327d9a 3740 }
1da177e4
LT		 3741 nfs4_unlock_state();
3742 return status;
3743
3744out_nfserr:
b8dd7b9a 3745 status = nfserrno(err);
1da177e4
LT		 3746 goto out;
3747}
3748
3749/*
3750 * returns
3751 * 1: locks held by lockowner
3752 * 0: no locks held by lockowner
3753 */
3754static int
3755check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3756{
3757 struct file_lock **flpp;
7eaa36e2 3758 struct inode *inode = filp->f_path.dentry->d_inode;
1da177e4
LT		 3759 int status = 0;
3760
3761 lock_kernel();
3762 for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
796dadfd 3763 if ((*flpp)->fl_owner == (fl_owner_t)lowner) {
1da177e4
LT		 3764 status = 1;
3765 goto out;
796dadfd 3766 }
1da177e4
LT		 3767 }
3768out:
3769 unlock_kernel();
3770 return status;
3771}
3772
b37ad28b 3773__be32
b591480b
BF		 3774nfsd4_release_lockowner(struct svc_rqst *rqstp,
3775 struct nfsd4_compound_state *cstate,
3776 struct nfsd4_release_lockowner *rlockowner)
1da177e4
LT		 3777{
3778 clientid_t *clid = &rlockowner->rl_clientid;
3e9e3dbe
N		 3779 struct nfs4_stateowner *sop;
3780 struct nfs4_stateid *stp;
1da177e4 3781 struct xdr_netobj *owner = &rlockowner->rl_owner;
3e9e3dbe
N		 3782 struct list_head matches;
3783 int i;
b37ad28b 3784 __be32 status;
1da177e4
LT		 3785
3786 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3787 clid->cl_boot, clid->cl_id);
3788
3789 /* XXX check for lease expiration */
3790
3791 status = nfserr_stale_clientid;
849823c5 3792 if (STALE_CLIENTID(clid))
1da177e4 3793 return status;
1da177e4
LT		 3794
3795 nfs4_lock_state();
3796
3e9e3dbe
N		 3797 status = nfserr_locks_held;
3798 /* XXX: we're doing a linear search through all the lockowners.
3799 * Yipes! For now we'll just hope clients aren't really using
3800 * release_lockowner much, but eventually we have to fix these
3801 * data structures. */
3802 INIT_LIST_HEAD(&matches);
3803 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3804 list_for_each_entry(sop, &lock_ownerid_hashtbl[i], so_idhash) {
599e0a22 3805 if (!same_owner_str(sop, owner, clid))
3e9e3dbe
N		 3806 continue;
3807 list_for_each_entry(stp, &sop->so_stateids,
3808 st_perstateowner) {
3809 if (check_for_locks(stp->st_vfs_file, sop))
3810 goto out;
3811 /* Note: so_perclient unused for lockowners,
3812 * so it's OK to fool with here. */
3813 list_add(&sop->so_perclient, &matches);
3814 }
1da177e4 3815 }
3e9e3dbe
N		 3816 }
3817 /* Clients probably won't expect us to return with some (but not all)
3818 * of the lockowner state released; so don't release any until all
3819 * have been checked. */
3820 status = nfs_ok;
0fa822e4
N		 3821 while (!list_empty(&matches)) {
3822 sop = list_entry(matches.next, struct nfs4_stateowner,
3823 so_perclient);
3824 /* unhash_stateowner deletes so_perclient only
3825 * for openowners. */
3826 list_del(&sop->so_perclient);
f044ff83 3827 release_lockowner(sop);
1da177e4
LT		 3828 }
3829out:
3830 nfs4_unlock_state();
3831 return status;
3832}
3833
3834static inline struct nfs4_client_reclaim *
a55370a3 3835alloc_reclaim(void)
1da177e4 3836{
a55370a3 3837 return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
1da177e4
LT		 3838}
3839
c7b9a459 3840int
a1bcecd2 3841nfs4_has_reclaimed_state(const char *name, bool use_exchange_id)
c7b9a459
N		 3842{
3843 unsigned int strhashval = clientstr_hashval(name);
3844 struct nfs4_client *clp;
3845
a1bcecd2 3846 clp = find_confirmed_client_by_str(name, strhashval, use_exchange_id);
c7b9a459
N		 3847 return clp ? 1 : 0;
3848}
3849
1da177e4
LT		 3850/*
3851 * failure => all reset bets are off, nfserr_no_grace...
3852 */
190e4fbf
N		 3853int
3854nfs4_client_to_reclaim(const char *name)
1da177e4
LT		 3855{
3856 unsigned int strhashval;
3857 struct nfs4_client_reclaim *crp = NULL;
3858
a55370a3
N		 3859 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
3860 crp = alloc_reclaim();
1da177e4
LT		 3861 if (!crp)
3862 return 0;
a55370a3 3863 strhashval = clientstr_hashval(name);
1da177e4
LT		 3864 INIT_LIST_HEAD(&crp->cr_strhash);
3865 list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
a55370a3 3866 memcpy(crp->cr_recdir, name, HEXDIR_LEN);
1da177e4
LT		 3867 reclaim_str_hashtbl_size++;
3868 return 1;
3869}
3870
3871static void
3872nfs4_release_reclaim(void)
3873{
3874 struct nfs4_client_reclaim *crp = NULL;
3875 int i;
3876
1da177e4
LT		 3877 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3878 while (!list_empty(&reclaim_str_hashtbl[i])) {
3879 crp = list_entry(reclaim_str_hashtbl[i].next,
3880 struct nfs4_client_reclaim, cr_strhash);
3881 list_del(&crp->cr_strhash);
1da177e4
LT		 3882 kfree(crp);
3883 reclaim_str_hashtbl_size--;
3884 }
3885 }
3886 BUG_ON(reclaim_str_hashtbl_size);
3887}
3888
3889/*
3890 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
fd39ca9a 3891static struct nfs4_client_reclaim *
1da177e4
LT		 3892nfs4_find_reclaim_client(clientid_t *clid)
3893{
3894 unsigned int strhashval;
3895 struct nfs4_client *clp;
3896 struct nfs4_client_reclaim *crp = NULL;
3897
3898
3899 /* find clientid in conf_id_hashtbl */
3900 clp = find_confirmed_client(clid);
3901 if (clp == NULL)
3902 return NULL;
3903
a55370a3
N		 3904 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3905 clp->cl_name.len, clp->cl_name.data,
3906 clp->cl_recdir);
1da177e4
LT		 3907
3908 /* find clp->cl_name in reclaim_str_hashtbl */
a55370a3 3909 strhashval = clientstr_hashval(clp->cl_recdir);
1da177e4 3910 list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
a55370a3 3911 if (same_name(crp->cr_recdir, clp->cl_recdir)) {
1da177e4
LT		 3912 return crp;
3913 }
3914 }
3915 return NULL;
3916}
3917
3918/*
3919* Called from OPEN. Look for clientid in reclaim list.
3920*/
b37ad28b 3921__be32
1da177e4
LT		 3922nfs4_check_open_reclaim(clientid_t *clid)
3923{
dfc83565 3924 return nfs4_find_reclaim_client(clid) ? nfs_ok : nfserr_reclaim_bad;
1da177e4
LT		 3925}
3926
ac4d8ff2 3927/* initialization to perform at module load time: */
1da177e4 3928
e8ff2a84 3929int
ac4d8ff2 3930nfs4_state_init(void)
1da177e4 3931{
e8ff2a84 3932 int i, status;
1da177e4 3933
e8ff2a84
BF		 3934 status = nfsd4_init_slabs();
3935 if (status)
3936 return status;
1da177e4
LT		 3937 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3938 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3939 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3940 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3941 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
02cb2858 3942 INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
1da177e4 3943 }
5282fd72
ME		 3944 for (i = 0; i < SESSION_HASH_SIZE; i++)
3945 INIT_LIST_HEAD(&sessionid_hashtbl[i]);
1da177e4
LT		 3946 for (i = 0; i < FILE_HASH_SIZE; i++) {
3947 INIT_LIST_HEAD(&file_hashtbl[i]);
3948 }
3949 for (i = 0; i < OWNER_HASH_SIZE; i++) {
3950 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3951 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3952 }
3953 for (i = 0; i < STATEID_HASH_SIZE; i++) {
3954 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3955 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3956 }
3957 for (i = 0; i < LOCK_HASH_SIZE; i++) {
3958 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3959 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3960 }
1da177e4 3961 memset(&onestateid, ~0, sizeof(stateid_t));
1da177e4
LT		 3962 INIT_LIST_HEAD(&close_lru);
3963 INIT_LIST_HEAD(&client_lru);
3964 INIT_LIST_HEAD(&del_recall_lru);
ac4d8ff2 3965 reclaim_str_hashtbl_size = 0;
e8ff2a84 3966 return 0;
ac4d8ff2
N		 3967}
3968
190e4fbf
N		 3969static void
3970nfsd4_load_reboot_recovery_data(void)
3971{
3972 int status;
3973
0964a3d3
N		 3974 nfs4_lock_state();
3975 nfsd4_init_recdir(user_recovery_dirname);
190e4fbf 3976 status = nfsd4_recdir_load();
0964a3d3 3977 nfs4_unlock_state();
190e4fbf
N		 3978 if (status)
3979 printk("NFSD: Failure reading reboot recovery data\n");
3980}
3981
c2f1a551
MS		 3982/*
3983 * Since the lifetime of a delegation isn't limited to that of an open, a
3984 * client may quite reasonably hang on to a delegation as long as it has
3985 * the inode cached. This becomes an obvious problem the first time a
3986 * client's inode cache approaches the size of the server's total memory.
3987 *
3988 * For now we avoid this problem by imposing a hard limit on the number
3989 * of delegations, which varies according to the server's memory size.
3990 */
3991static void
3992set_max_delegations(void)
3993{
3994 /*
3995 * Allow at most 4 delegations per megabyte of RAM. Quick
3996 * estimates suggest that in the worst case (where every delegation
3997 * is for a different inode), a delegation could take about 1.5K,
3998 * giving a worst case usage of about 6% of memory.
3999 */
4000 max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
4001}
4002
ac4d8ff2
N		 4003/* initialization to perform when the nfsd service is started: */
4004
29ab23cc 4005static int
ac4d8ff2
N		 4006__nfs4_state_start(void)
4007{
b5a1a81e
BF		 4008 int ret;
4009
1da177e4 4010 boot_time = get_seconds();
af558e33 4011 locks_start_grace(&nfsd4_manager);
9a8db97e 4012 printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
e46b498c 4013 nfsd4_grace);
b5a1a81e
BF		 4014 ret = set_callback_cred();
4015 if (ret)
4016 return -ENOMEM;
58da282b 4017 laundry_wq = create_singlethread_workqueue("nfsd4");
29ab23cc
BF		 4018 if (laundry_wq == NULL)
4019 return -ENOMEM;
b5a1a81e
BF		 4020 ret = nfsd4_create_callback_queue();
4021 if (ret)
4022 goto out_free_laundry;
e46b498c 4023 queue_delayed_work(laundry_wq, &laundromat_work, nfsd4_grace * HZ);
c2f1a551 4024 set_max_delegations();
b5a1a81e
BF		 4025 return 0;
4026out_free_laundry:
4027 destroy_workqueue(laundry_wq);
4028 return ret;
1da177e4
LT		 4029}
4030
29ab23cc 4031int
76a3550e 4032nfs4_state_start(void)
1da177e4 4033{
29ab23cc
BF		 4034 int ret;
4035
1da177e4 4036 if (nfs4_init)
29ab23cc 4037 return 0;
190e4fbf 4038 nfsd4_load_reboot_recovery_data();
29ab23cc
BF		 4039 ret = __nfs4_state_start();
4040 if (ret)
4041 return ret;
1da177e4 4042 nfs4_init = 1;
29ab23cc 4043 return 0;
1da177e4
LT		 4044}
4045
1da177e4
LT		 4046static void
4047__nfs4_state_shutdown(void)
4048{
4049 int i;
4050 struct nfs4_client *clp = NULL;
4051 struct nfs4_delegation *dp = NULL;
1da177e4
LT		 4052 struct list_head *pos, *next, reaplist;
4053
1da177e4
LT		 4054 for (i = 0; i < CLIENT_HASH_SIZE; i++) {
4055 while (!list_empty(&conf_id_hashtbl[i])) {
4056 clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
4057 expire_client(clp);
4058 }
4059 while (!list_empty(&unconf_str_hashtbl[i])) {
4060 clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
4061 expire_client(clp);
4062 }
4063 }
4064 INIT_LIST_HEAD(&reaplist);
4065 spin_lock(&recall_lock);
4066 list_for_each_safe(pos, next, &del_recall_lru) {
4067 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4068 list_move(&dp->dl_recall_lru, &reaplist);
4069 }
4070 spin_unlock(&recall_lock);
4071 list_for_each_safe(pos, next, &reaplist) {
4072 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4073 list_del_init(&dp->dl_recall_lru);
4074 unhash_delegation(dp);
4075 }
4076
190e4fbf 4077 nfsd4_shutdown_recdir();
1da177e4 4078 nfs4_init = 0;
1da177e4
LT		 4079}
4080
4081void
4082nfs4_state_shutdown(void)
4083{
5e8d5c29
N		 4084 cancel_rearming_delayed_workqueue(laundry_wq, &laundromat_work);
4085 destroy_workqueue(laundry_wq);
2c5e7615 4086 locks_end_grace(&nfsd4_manager);
1da177e4
LT		 4087 nfs4_lock_state();
4088 nfs4_release_reclaim();
4089 __nfs4_state_shutdown();
b5a1a81e 4090 nfsd4_destroy_callback_queue();
1da177e4
LT		 4091 nfs4_unlock_state();
4092}
4093
3dd98a3b
JL		 4094/*
4095 * user_recovery_dirname is protected by the nfsd_mutex since it's only
4096 * accessed when nfsd is starting.
4097 */
0964a3d3
N		 4098static void
4099nfs4_set_recdir(char *recdir)
4100{
0964a3d3 4101 strcpy(user_recovery_dirname, recdir);
0964a3d3
N		 4102}
4103
4104/*
4105 * Change the NFSv4 recovery directory to recdir.
4106 */
4107int
4108nfs4_reset_recoverydir(char *recdir)
4109{
4110 int status;
a63bb996 4111 struct path path;
0964a3d3 4112
a63bb996 4113 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
0964a3d3
N		 4114 if (status)
4115 return status;
4116 status = -ENOTDIR;
a63bb996 4117 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
0964a3d3
N		 4118 nfs4_set_recdir(recdir);
4119 status = 0;
4120 }
a63bb996 4121 path_put(&path);
0964a3d3
N		 4122 return status;
4123}
4124
3dd98a3b
JL		 4125char *
4126nfs4_recoverydir(void)
4127{
4128 return user_recovery_dirname;
4129}
Linux kernel - Deliverables
This page took 0.75426 seconds and 5 git commands to generate.