fs/nfsd/nfs4state.c

   1 /*
   2 *  linux/fs/nfsd/nfs4state.c
   3 *
   4 *  Copyright (c) 2001 The Regents of the University of Michigan.
   5 *  All rights reserved.
   6 *
   7 *  Kendrick Smith <kmsmith@umich.edu>
   8 *  Andy Adamson <kandros@umich.edu>
   9 *
  10 *  Redistribution and use in source and binary forms, with or without
  11 *  modification, are permitted provided that the following conditions
  12 *  are met:
  13 *
  14 *  1. Redistributions of source code must retain the above copyright
  15 *     notice, this list of conditions and the following disclaimer.
  16 *  2. Redistributions in binary form must reproduce the above copyright
  17 *     notice, this list of conditions and the following disclaimer in the
  18 *     documentation and/or other materials provided with the distribution.
  19 *  3. Neither the name of the University nor the names of its
  20 *     contributors may be used to endorse or promote products derived
  21 *     from this software without specific prior written permission.
  22 *
  23 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  24 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  25 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  26 *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  27 *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  28 *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  29 *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  30 *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  31 *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  32 *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  33 *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  34 *
  35 */
  36
  37 #include <linux/param.h>
  38 #include <linux/major.h>
  39 #include <linux/slab.h>
  40
  41 #include <linux/sunrpc/svc.h>
  42 #include <linux/nfsd/nfsd.h>
  43 #include <linux/nfsd/cache.h>
  44 #include <linux/mount.h>
  45 #include <linux/workqueue.h>
  46 #include <linux/smp_lock.h>
  47 #include <linux/kthread.h>
  48 #include <linux/nfs4.h>
  49 #include <linux/nfsd/state.h>
  50 #include <linux/nfsd/xdr4.h>
  51
  52 #define NFSDDBG_FACILITY                NFSDDBG_PROC
  53
  54 /* Globals */
  55 static time_t lease_time = 90;     /* default lease time */
  56 static time_t old_lease_time = 90; /* past incarnation lease time */
  57 static u32 nfs4_reclaim_init = 0;
  58 time_t boot_time;
  59 static time_t grace_end = 0;
  60 static u32 current_clientid = 1;
  61 static u32 current_ownerid = 1;
  62 static u32 current_fileid = 1;
  63 static u32 current_delegid = 1;
  64 static u32 nfs4_init;
  65 stateid_t zerostateid;             /* bits all 0 */
  66 stateid_t onestateid;              /* bits all 1 */
  67
  68 /* debug counters */
  69 u32 list_add_perfile = 0;
  70 u32 list_del_perfile = 0;
  71 u32 add_perclient = 0;
  72 u32 del_perclient = 0;
  73 u32 vfsopen = 0;
  74 u32 vfsclose = 0;
  75 u32 alloc_delegation= 0;
  76 u32 free_delegation= 0;
  77
  78 /* forward declarations */
  79 struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
  80 static struct nfs4_delegation * find_delegation_stateid(struct inode *ino, stateid_t *stid);
  81 static void release_stateid_lockowners(struct nfs4_stateid *open_stp);
  82
  83 /* Locking:
  84  *
  85  * client_sema:
  86  *      protects clientid_hashtbl[], clientstr_hashtbl[],
  87  *      unconfstr_hashtbl[], uncofid_hashtbl[].
  88  */
  89 static DECLARE_MUTEX(client_sema);
  90
  91 kmem_cache_t *stateowner_slab = NULL;
  92 kmem_cache_t *file_slab = NULL;
  93 kmem_cache_t *stateid_slab = NULL;
  94
  95 void
  96 nfs4_lock_state(void)
  97 {
  98         down(&client_sema);
  99 }
 100
 101 void
 102 nfs4_unlock_state(void)
 103 {
 104         up(&client_sema);
 105 }
 106
 107 static inline u32
 108 opaque_hashval(const void *ptr, int nbytes)
 109 {
 110         unsigned char *cptr = (unsigned char *) ptr;
 111
 112         u32 x = 0;
 113         while (nbytes--) {
 114                 x *= 37;
 115                 x += *cptr++;
 116         }
 117         return x;
 118 }
 119
 120 /* forward declarations */
 121 static void release_stateowner(struct nfs4_stateowner *sop);
 122 static void release_stateid(struct nfs4_stateid *stp, int flags);
 123 static void release_file(struct nfs4_file *fp);
 124
 125 /*
 126  * Delegation state
 127  */
 128
 129 /* recall_lock protects the del_recall_lru */
 130 spinlock_t recall_lock;
 131 static struct list_head del_recall_lru;
 132
 133 static struct nfs4_delegation *
 134 alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_fh *current_fh, u32 type)
 135 {
 136         struct nfs4_delegation *dp;
 137         struct nfs4_file *fp = stp->st_file;
 138         struct nfs4_callback *cb = &stp->st_stateowner->so_client->cl_callback;
 139
 140         dprintk("NFSD alloc_init_deleg\n");
 141         if ((dp = kmalloc(sizeof(struct nfs4_delegation),
 142                 GFP_KERNEL)) == NULL)
 143                 return dp;
 144         INIT_LIST_HEAD(&dp->dl_del_perfile);
 145         INIT_LIST_HEAD(&dp->dl_del_perclnt);
 146         INIT_LIST_HEAD(&dp->dl_recall_lru);
 147         dp->dl_client = clp;
 148         dp->dl_file = fp;
 149         dp->dl_flock = NULL;
 150         get_file(stp->st_vfs_file);
 151         dp->dl_vfs_file = stp->st_vfs_file;
 152         dp->dl_type = type;
 153         dp->dl_recall.cbr_dp = NULL;
 154         dp->dl_recall.cbr_ident = cb->cb_ident;
 155         dp->dl_recall.cbr_trunc = 0;
 156         dp->dl_stateid.si_boot = boot_time;
 157         dp->dl_stateid.si_stateownerid = current_delegid++;
 158         dp->dl_stateid.si_fileid = 0;
 159         dp->dl_stateid.si_generation = 0;
 160         dp->dl_fhlen = current_fh->fh_handle.fh_size;
 161         memcpy(dp->dl_fhval, &current_fh->fh_handle.fh_base,
 162                         current_fh->fh_handle.fh_size);
 163         dp->dl_time = 0;
 164         atomic_set(&dp->dl_count, 1);
 165         list_add(&dp->dl_del_perfile, &fp->fi_del_perfile);
 166         list_add(&dp->dl_del_perclnt, &clp->cl_del_perclnt);
 167         alloc_delegation++;
 168         return dp;
 169 }
 170
 171 void
 172 nfs4_put_delegation(struct nfs4_delegation *dp)
 173 {
 174         if (atomic_dec_and_test(&dp->dl_count)) {
 175                 dprintk("NFSD: freeing dp %p\n",dp);
 176                 kfree(dp);
 177                 free_delegation++;
 178         }
 179 }
 180
 181 /* Remove the associated file_lock first, then remove the delegation.
 182  * lease_modify() is called to remove the FS_LEASE file_lock from
 183  * the i_flock list, eventually calling nfsd's lock_manager
 184  * fl_release_callback.
 185  */
 186 static void
 187 nfs4_close_delegation(struct nfs4_delegation *dp)
 188 {
 189         struct file *filp = dp->dl_vfs_file;
 190
 191         dprintk("NFSD: close_delegation dp %p\n",dp);
 192         dp->dl_vfs_file = NULL;
 193         /* The following nfsd_close may not actually close the file,
 194          * but we want to remove the lease in any case. */
 195         if (dp->dl_flock)
 196                 setlease(filp, F_UNLCK, &dp->dl_flock);
 197         nfsd_close(filp);
 198         vfsclose++;
 199 }
 200
 201 /* Called under the state lock. */
 202 static void
 203 unhash_delegation(struct nfs4_delegation *dp)
 204 {
 205         list_del_init(&dp->dl_del_perfile);
 206         list_del_init(&dp->dl_del_perclnt);
 207         spin_lock(&recall_lock);
 208         list_del_init(&dp->dl_recall_lru);
 209         spin_unlock(&recall_lock);
 210         nfs4_close_delegation(dp);
 211         nfs4_put_delegation(dp);
 212 }
 213
 214 /*
 215  * SETCLIENTID state
 216  */
 217
 218 /* Hash tables for nfs4_clientid state */
 219 #define CLIENT_HASH_BITS                 4
 220 #define CLIENT_HASH_SIZE                (1 << CLIENT_HASH_BITS)
 221 #define CLIENT_HASH_MASK                (CLIENT_HASH_SIZE - 1)
 222
 223 #define clientid_hashval(id) \
 224         ((id) & CLIENT_HASH_MASK)
 225 #define clientstr_hashval(name, namelen) \
 226         (opaque_hashval((name), (namelen)) & CLIENT_HASH_MASK)
 227 /*
 228  * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
 229  * used in reboot/reset lease grace period processing
 230  *
 231  * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
 232  * setclientid_confirmed info.
 233  *
 234  * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
 235  * setclientid info.
 236  *
 237  * client_lru holds client queue ordered by nfs4_client.cl_time
 238  * for lease renewal.
 239  *
 240  * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
 241  * for last close replay.
 242  */
 243 static struct list_head reclaim_str_hashtbl[CLIENT_HASH_SIZE];
 244 static int reclaim_str_hashtbl_size = 0;
 245 static struct list_head conf_id_hashtbl[CLIENT_HASH_SIZE];
 246 static struct list_head conf_str_hashtbl[CLIENT_HASH_SIZE];
 247 static struct list_head unconf_str_hashtbl[CLIENT_HASH_SIZE];
 248 static struct list_head unconf_id_hashtbl[CLIENT_HASH_SIZE];
 249 static struct list_head client_lru;
 250 static struct list_head close_lru;
 251
 252 static inline void
 253 renew_client(struct nfs4_client *clp)
 254 {
 255         /*
 256         * Move client to the end to the LRU list.
 257         */
 258         dprintk("renewing client (clientid %08x/%08x)\n",
 259                         clp->cl_clientid.cl_boot,
 260                         clp->cl_clientid.cl_id);
 261         list_move_tail(&clp->cl_lru, &client_lru);
 262         clp->cl_time = get_seconds();
 263 }
 264
 265 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
 266 static int
 267 STALE_CLIENTID(clientid_t *clid)
 268 {
 269         if (clid->cl_boot == boot_time)
 270                 return 0;
 271         dprintk("NFSD stale clientid (%08x/%08x)\n",
 272                         clid->cl_boot, clid->cl_id);
 273         return 1;
 274 }
 275
 276 /*
 277  * XXX Should we use a slab cache ?
 278  * This type of memory management is somewhat inefficient, but we use it
 279  * anyway since SETCLIENTID is not a common operation.
 280  */
 281 static inline struct nfs4_client *
 282 alloc_client(struct xdr_netobj name)
 283 {
 284         struct nfs4_client *clp;
 285
 286         if ((clp = kmalloc(sizeof(struct nfs4_client), GFP_KERNEL))!= NULL) {
 287                 memset(clp, 0, sizeof(*clp));
 288                 if ((clp->cl_name.data = kmalloc(name.len, GFP_KERNEL)) != NULL) {
 289                         memcpy(clp->cl_name.data, name.data, name.len);
 290                         clp->cl_name.len = name.len;
 291                 }
 292                 else {
 293                         kfree(clp);
 294                         clp = NULL;
 295                 }
 296         }
 297         return clp;
 298 }
 299
 300 static inline void
 301 free_client(struct nfs4_client *clp)
 302 {
 303         if (clp->cl_cred.cr_group_info)
 304                 put_group_info(clp->cl_cred.cr_group_info);
 305         kfree(clp->cl_name.data);
 306         kfree(clp);
 307 }
 308
 309 void
 310 put_nfs4_client(struct nfs4_client *clp)
 311 {
 312         if (atomic_dec_and_test(&clp->cl_count))
 313                 free_client(clp);
 314 }
 315
 316 static void
 317 expire_client(struct nfs4_client *clp)
 318 {
 319         struct nfs4_stateowner *sop;
 320         struct nfs4_delegation *dp;
 321         struct nfs4_callback *cb = &clp->cl_callback;
 322         struct rpc_clnt *clnt = clp->cl_callback.cb_client;
 323         struct list_head reaplist;
 324
 325         dprintk("NFSD: expire_client cl_count %d\n",
 326                             atomic_read(&clp->cl_count));
 327
 328         /* shutdown rpc client, ending any outstanding recall rpcs */
 329         if (atomic_read(&cb->cb_set) == 1 && clnt) {
 330                 rpc_shutdown_client(clnt);
 331                 clnt = clp->cl_callback.cb_client = NULL;
 332         }
 333
 334         INIT_LIST_HEAD(&reaplist);
 335         spin_lock(&recall_lock);
 336         while (!list_empty(&clp->cl_del_perclnt)) {
 337                 dp = list_entry(clp->cl_del_perclnt.next, struct nfs4_delegation, dl_del_perclnt);
 338                 dprintk("NFSD: expire client. dp %p, fp %p\n", dp,
 339                                 dp->dl_flock);
 340                 list_del_init(&dp->dl_del_perclnt);
 341                 list_move(&dp->dl_recall_lru, &reaplist);
 342         }
 343         spin_unlock(&recall_lock);
 344         while (!list_empty(&reaplist)) {
 345                 dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
 346                 list_del_init(&dp->dl_recall_lru);
 347                 unhash_delegation(dp);
 348         }
 349         list_del(&clp->cl_idhash);
 350         list_del(&clp->cl_strhash);
 351         list_del(&clp->cl_lru);
 352         while (!list_empty(&clp->cl_perclient)) {
 353                 sop = list_entry(clp->cl_perclient.next, struct nfs4_stateowner, so_perclient);
 354                 release_stateowner(sop);
 355         }
 356         put_nfs4_client(clp);
 357 }
 358
 359 static struct nfs4_client *
 360 create_client(struct xdr_netobj name) {
 361         struct nfs4_client *clp;
 362
 363         if (!(clp = alloc_client(name)))
 364                 goto out;
 365         atomic_set(&clp->cl_count, 1);
 366         atomic_set(&clp->cl_callback.cb_set, 0);
 367         clp->cl_callback.cb_parsed = 0;
 368         INIT_LIST_HEAD(&clp->cl_idhash);
 369         INIT_LIST_HEAD(&clp->cl_strhash);
 370         INIT_LIST_HEAD(&clp->cl_perclient);
 371         INIT_LIST_HEAD(&clp->cl_del_perclnt);
 372         INIT_LIST_HEAD(&clp->cl_lru);
 373 out:
 374         return clp;
 375 }
 376
 377 static void
 378 copy_verf(struct nfs4_client *target, nfs4_verifier *source) {
 379         memcpy(target->cl_verifier.data, source->data, sizeof(target->cl_verifier.data));
 380 }
 381
 382 static void
 383 copy_clid(struct nfs4_client *target, struct nfs4_client *source) {
 384         target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
 385         target->cl_clientid.cl_id = source->cl_clientid.cl_id;
 386 }
 387
 388 static void
 389 copy_cred(struct svc_cred *target, struct svc_cred *source) {
 390
 391         target->cr_uid = source->cr_uid;
 392         target->cr_gid = source->cr_gid;
 393         target->cr_group_info = source->cr_group_info;
 394         get_group_info(target->cr_group_info);
 395 }
 396
 397 static int
 398 cmp_name(struct xdr_netobj *n1, struct xdr_netobj *n2) {
 399         if (!n1 || !n2)
 400                 return 0;
 401         return((n1->len == n2->len) && !memcmp(n1->data, n2->data, n2->len));
 402 }
 403
 404 static int
 405 cmp_verf(nfs4_verifier *v1, nfs4_verifier *v2) {
 406         return(!memcmp(v1->data,v2->data,sizeof(v1->data)));
 407 }
 408
 409 static int
 410 cmp_clid(clientid_t * cl1, clientid_t * cl2) {
 411         return((cl1->cl_boot == cl2->cl_boot) &&
 412                 (cl1->cl_id == cl2->cl_id));
 413 }
 414
 415 /* XXX what about NGROUP */
 416 static int
 417 cmp_creds(struct svc_cred *cr1, struct svc_cred *cr2){
 418         return(cr1->cr_uid == cr2->cr_uid);
 419
 420 }
 421
 422 static void
 423 gen_clid(struct nfs4_client *clp) {
 424         clp->cl_clientid.cl_boot = boot_time;
 425         clp->cl_clientid.cl_id = current_clientid++;
 426 }
 427
 428 static void
 429 gen_confirm(struct nfs4_client *clp) {
 430         struct timespec         tv;
 431         u32 *                   p;
 432
 433         tv = CURRENT_TIME;
 434         p = (u32 *)clp->cl_confirm.data;
 435         *p++ = tv.tv_sec;
 436         *p++ = tv.tv_nsec;
 437 }
 438
 439 static int
 440 check_name(struct xdr_netobj name) {
 441
 442         if (name.len == 0)
 443                 return 0;
 444         if (name.len > NFS4_OPAQUE_LIMIT) {
 445                 printk("NFSD: check_name: name too long(%d)!\n", name.len);
 446                 return 0;
 447         }
 448         return 1;
 449 }
 450
 451 void
 452 add_to_unconfirmed(struct nfs4_client *clp, unsigned int strhashval)
 453 {
 454         unsigned int idhashval;
 455
 456         list_add(&clp->cl_strhash, &unconf_str_hashtbl[strhashval]);
 457         idhashval = clientid_hashval(clp->cl_clientid.cl_id);
 458         list_add(&clp->cl_idhash, &unconf_id_hashtbl[idhashval]);
 459         list_add_tail(&clp->cl_lru, &client_lru);
 460         clp->cl_time = get_seconds();
 461 }
 462
 463 void
 464 move_to_confirmed(struct nfs4_client *clp)
 465 {
 466         unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
 467         unsigned int strhashval;
 468
 469         dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
 470         list_del_init(&clp->cl_strhash);
 471         list_del_init(&clp->cl_idhash);
 472         list_add(&clp->cl_idhash, &conf_id_hashtbl[idhashval]);
 473         strhashval = clientstr_hashval(clp->cl_name.data,
 474                         clp->cl_name.len);
 475         list_add(&clp->cl_strhash, &conf_str_hashtbl[strhashval]);
 476         renew_client(clp);
 477 }
 478
 479 static struct nfs4_client *
 480 find_confirmed_client(clientid_t *clid)
 481 {
 482         struct nfs4_client *clp;
 483         unsigned int idhashval = clientid_hashval(clid->cl_id);
 484
 485         list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
 486                 if (cmp_clid(&clp->cl_clientid, clid))
 487                         return clp;
 488         }
 489         return NULL;
 490 }
 491
 492 static struct nfs4_client *
 493 find_unconfirmed_client(clientid_t *clid)
 494 {
 495         struct nfs4_client *clp;
 496         unsigned int idhashval = clientid_hashval(clid->cl_id);
 497
 498         list_for_each_entry(clp, &unconf_id_hashtbl[idhashval], cl_idhash) {
 499                 if (cmp_clid(&clp->cl_clientid, clid))
 500                         return clp;
 501         }
 502         return NULL;
 503 }
 504
 505 /* a helper function for parse_callback */
 506 static int
 507 parse_octet(unsigned int *lenp, char **addrp)
 508 {
 509         unsigned int len = *lenp;
 510         char *p = *addrp;
 511         int n = -1;
 512         char c;
 513
 514         for (;;) {
 515                 if (!len)
 516                         break;
 517                 len--;
 518                 c = *p++;
 519                 if (c == '.')
 520                         break;
 521                 if ((c < '0') || (c > '9')) {
 522                         n = -1;
 523                         break;
 524                 }
 525                 if (n < 0)
 526                         n = 0;
 527                 n = (n * 10) + (c - '0');
 528                 if (n > 255) {
 529                         n = -1;
 530                         break;
 531                 }
 532         }
 533         *lenp = len;
 534         *addrp = p;
 535         return n;
 536 }
 537
 538 /* parse and set the setclientid ipv4 callback address */
 539 int
 540 parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
 541 {
 542         int temp = 0;
 543         u32 cbaddr = 0;
 544         u16 cbport = 0;
 545         u32 addrlen = addr_len;
 546         char *addr = addr_val;
 547         int i, shift;
 548
 549         /* ipaddress */
 550         shift = 24;
 551         for(i = 4; i > 0  ; i--) {
 552                 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
 553                         return 0;
 554                 }
 555                 cbaddr |= (temp << shift);
 556                 if (shift > 0)
 557                 shift -= 8;
 558         }
 559         *cbaddrp = cbaddr;
 560
 561         /* port */
 562         shift = 8;
 563         for(i = 2; i > 0  ; i--) {
 564                 if ((temp = parse_octet(&addrlen, &addr)) < 0) {
 565                         return 0;
 566                 }
 567                 cbport |= (temp << shift);
 568                 if (shift > 0)
 569                         shift -= 8;
 570         }
 571         *cbportp = cbport;
 572         return 1;
 573 }
 574
 575 void
 576 gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
 577 {
 578         struct nfs4_callback *cb = &clp->cl_callback;
 579
 580         /* Currently, we only support tcp for the callback channel */
 581         if ((se->se_callback_netid_len != 3) || memcmp((char *)se->se_callback_netid_val, "tcp", 3))
 582                 goto out_err;
 583
 584         if ( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
 585                          &cb->cb_addr, &cb->cb_port)))
 586                 goto out_err;
 587         cb->cb_prog = se->se_callback_prog;
 588         cb->cb_ident = se->se_callback_ident;
 589         cb->cb_parsed = 1;
 590         return;
 591 out_err:
 592         printk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
 593                 "will not receive delegations\n",
 594                 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
 595
 596         cb->cb_parsed = 0;
 597         return;
 598 }
 599
 600 /*
 601  * RFC 3010 has a complex implmentation description of processing a
 602  * SETCLIENTID request consisting of 5 bullets, labeled as
 603  * CASE0 - CASE4 below.
 604  *
 605  * NOTES:
 606  *      callback information will be processed in a future patch
 607  *
 608  *      an unconfirmed record is added when:
 609  *      NORMAL (part of CASE 4): there is no confirmed nor unconfirmed record.
 610  *      CASE 1: confirmed record found with matching name, principal,
 611  *              verifier, and clientid.
 612  *      CASE 2: confirmed record found with matching name, principal,
 613  *              and there is no unconfirmed record with matching
 614  *              name and principal
 615  *
 616  *      an unconfirmed record is replaced when:
 617  *      CASE 3: confirmed record found with matching name, principal,
 618  *              and an unconfirmed record is found with matching
 619  *              name, principal, and with clientid and
 620  *              confirm that does not match the confirmed record.
 621  *      CASE 4: there is no confirmed record with matching name and
 622  *              principal. there is an unconfirmed record with
 623  *              matching name, principal.
 624  *
 625  *      an unconfirmed record is deleted when:
 626  *      CASE 1: an unconfirmed record that matches input name, verifier,
 627  *              and confirmed clientid.
 628  *      CASE 4: any unconfirmed records with matching name and principal
 629  *              that exist after an unconfirmed record has been replaced
 630  *              as described above.
 631  *
 632  */
 633 int
 634 nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
 635 {
 636         u32                     ip_addr = rqstp->rq_addr.sin_addr.s_addr;
 637         struct xdr_netobj       clname = {
 638                 .len = setclid->se_namelen,
 639                 .data = setclid->se_name,
 640         };
 641         nfs4_verifier           clverifier = setclid->se_verf;
 642         unsigned int            strhashval;
 643         struct nfs4_client *    conf, * unconf, * new, * clp;
 644         int                     status;
 645
 646         status = nfserr_inval;
 647         if (!check_name(clname))
 648                 goto out;
 649
 650         /*
 651          * XXX The Duplicate Request Cache (DRC) has been checked (??)
 652          * We get here on a DRC miss.
 653          */
 654
 655         strhashval = clientstr_hashval(clname.data, clname.len);
 656
 657         conf = NULL;
 658         nfs4_lock_state();
 659         list_for_each_entry(clp, &conf_str_hashtbl[strhashval], cl_strhash) {
 660                 if (!cmp_name(&clp->cl_name, &clname))
 661                         continue;
 662                 /*
 663                  * CASE 0:
 664                  * clname match, confirmed, different principal
 665                  * or different ip_address
 666                  */
 667                 status = nfserr_clid_inuse;
 668                 if (!cmp_creds(&clp->cl_cred,&rqstp->rq_cred)) {
 669                         printk("NFSD: setclientid: string in use by client"
 670                         "(clientid %08x/%08x)\n",
 671                         clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
 672                         goto out;
 673                 }
 674                 if (clp->cl_addr != ip_addr) {
 675                         printk("NFSD: setclientid: string in use by client"
 676                         "(clientid %08x/%08x)\n",
 677                         clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
 678                         goto out;
 679                 }
 680
 681                 /*
 682                  * cl_name match from a previous SETCLIENTID operation
 683                  * XXX check for additional matches?
 684                  */
 685                 conf = clp;
 686                 break;
 687         }
 688         unconf = NULL;
 689         list_for_each_entry(clp, &unconf_str_hashtbl[strhashval], cl_strhash) {
 690                 if (!cmp_name(&clp->cl_name, &clname))
 691                         continue;
 692                 /* cl_name match from a previous SETCLIENTID operation */
 693                 unconf = clp;
 694                 break;
 695         }
 696         status = nfserr_resource;
 697         if (!conf) {
 698                 /*
 699                  * CASE 4:
 700                  * placed first, because it is the normal case.
 701                  */
 702                 if (unconf)
 703                         expire_client(unconf);
 704                 if (!(new = create_client(clname)))
 705                         goto out;
 706                 copy_verf(new, &clverifier);
 707                 new->cl_addr = ip_addr;
 708                 copy_cred(&new->cl_cred,&rqstp->rq_cred);
 709                 gen_clid(new);
 710                 gen_confirm(new);
 711                 gen_callback(new, setclid);
 712                 add_to_unconfirmed(new, strhashval);
 713         } else if (cmp_verf(&conf->cl_verifier, &clverifier)) {
 714                 /*
 715                  * CASE 1:
 716                  * cl_name match, confirmed, principal match
 717                  * verifier match: probable callback update
 718                  *
 719                  * remove any unconfirmed nfs4_client with
 720                  * matching cl_name, cl_verifier, and cl_clientid
 721                  *
 722                  * create and insert an unconfirmed nfs4_client with same
 723                  * cl_name, cl_verifier, and cl_clientid as existing
 724                  * nfs4_client,  but with the new callback info and a
 725                  * new cl_confirm
 726                  */
 727                 if ((unconf) &&
 728                     cmp_verf(&unconf->cl_verifier, &conf->cl_verifier) &&
 729                      cmp_clid(&unconf->cl_clientid, &conf->cl_clientid)) {
 730                                 expire_client(unconf);
 731                 }
 732                 if (!(new = create_client(clname)))
 733                         goto out;
 734                 copy_verf(new,&conf->cl_verifier);
 735                 new->cl_addr = ip_addr;
 736                 copy_cred(&new->cl_cred,&rqstp->rq_cred);
 737                 copy_clid(new, conf);
 738                 gen_confirm(new);
 739                 gen_callback(new, setclid);
 740                 add_to_unconfirmed(new,strhashval);
 741         } else if (!unconf) {
 742                 /*
 743                  * CASE 2:
 744                  * clname match, confirmed, principal match
 745                  * verfier does not match
 746                  * no unconfirmed. create a new unconfirmed nfs4_client
 747                  * using input clverifier, clname, and callback info
 748                  * and generate a new cl_clientid and cl_confirm.
 749                  */
 750                 if (!(new = create_client(clname)))
 751                         goto out;
 752                 copy_verf(new,&clverifier);
 753                 new->cl_addr = ip_addr;
 754                 copy_cred(&new->cl_cred,&rqstp->rq_cred);
 755                 gen_clid(new);
 756                 gen_confirm(new);
 757                 gen_callback(new, setclid);
 758                 add_to_unconfirmed(new, strhashval);
 759         } else if (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm)) {
 760                 /*
 761                  * CASE3:
 762                  * confirmed found (name, principal match)
 763                  * confirmed verifier does not match input clverifier
 764                  *
 765                  * unconfirmed found (name match)
 766                  * confirmed->cl_confirm != unconfirmed->cl_confirm
 767                  *
 768                  * remove unconfirmed.
 769                  *
 770                  * create an unconfirmed nfs4_client
 771                  * with same cl_name as existing confirmed nfs4_client,
 772                  * but with new callback info, new cl_clientid,
 773                  * new cl_verifier and a new cl_confirm
 774                  */
 775                 expire_client(unconf);
 776                 if (!(new = create_client(clname)))
 777                         goto out;
 778                 copy_verf(new,&clverifier);
 779                 new->cl_addr = ip_addr;
 780                 copy_cred(&new->cl_cred,&rqstp->rq_cred);
 781                 gen_clid(new);
 782                 gen_confirm(new);
 783                 gen_callback(new, setclid);
 784                 add_to_unconfirmed(new, strhashval);
 785         } else {
 786                 /* No cases hit !!! */
 787                 status = nfserr_inval;
 788                 goto out;
 789
 790         }
 791         setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
 792         setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
 793         memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
 794         status = nfs_ok;
 795 out:
 796         nfs4_unlock_state();
 797         return status;
 798 }
 799
 800
 801 /*
 802  * RFC 3010 has a complex implmentation description of processing a
 803  * SETCLIENTID_CONFIRM request consisting of 4 bullets describing
 804  * processing on a DRC miss, labeled as CASE1 - CASE4 below.
 805  *
 806  * NOTE: callback information will be processed here in a future patch
 807  */
 808 int
 809 nfsd4_setclientid_confirm(struct svc_rqst *rqstp, struct nfsd4_setclientid_confirm *setclientid_confirm)
 810 {
 811         u32 ip_addr = rqstp->rq_addr.sin_addr.s_addr;
 812         struct nfs4_client *clp, *conf = NULL, *unconf = NULL;
 813         nfs4_verifier confirm = setclientid_confirm->sc_confirm;
 814         clientid_t * clid = &setclientid_confirm->sc_clientid;
 815         int status;
 816
 817         if (STALE_CLIENTID(clid))
 818                 return nfserr_stale_clientid;
 819         /*
 820          * XXX The Duplicate Request Cache (DRC) has been checked (??)
 821          * We get here on a DRC miss.
 822          */
 823
 824         nfs4_lock_state();
 825         clp = find_confirmed_client(clid);
 826         if (clp) {
 827                 status = nfserr_inval;
 828                 /*
 829                  * Found a record for this clientid. If the IP addresses
 830                  * don't match, return ERR_INVAL just as if the record had
 831                  * not been found.
 832                  */
 833                 if (clp->cl_addr != ip_addr) {
 834                         printk("NFSD: setclientid: string in use by client"
 835                         "(clientid %08x/%08x)\n",
 836                         clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
 837                         goto out;
 838                 }
 839                 conf = clp;
 840         }
 841         clp = find_unconfirmed_client(clid);
 842         if (clp) {
 843                 status = nfserr_inval;
 844                 if (clp->cl_addr != ip_addr) {
 845                         printk("NFSD: setclientid: string in use by client"
 846                         "(clientid %08x/%08x)\n",
 847                         clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
 848                         goto out;
 849                 }
 850                 unconf = clp;
 851         }
 852         /* CASE 1:
 853         * unconf record that matches input clientid and input confirm.
 854         * conf record that matches input clientid.
 855         * conf  and unconf records match names, verifiers
 856         */
 857         if ((conf && unconf) &&
 858             (cmp_verf(&unconf->cl_confirm, &confirm)) &&
 859             (cmp_verf(&conf->cl_verifier, &unconf->cl_verifier)) &&
 860             (cmp_name(&conf->cl_name,&unconf->cl_name))  &&
 861             (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm))) {
 862                 if (!cmp_creds(&conf->cl_cred, &unconf->cl_cred))
 863                         status = nfserr_clid_inuse;
 864                 else {
 865                         expire_client(conf);
 866                         clp = unconf;
 867                         move_to_confirmed(unconf);
 868                         status = nfs_ok;
 869                 }
 870                 goto out;
 871         }
 872         /* CASE 2:
 873          * conf record that matches input clientid.
 874          * if unconf record that matches input clientid, then unconf->cl_name
 875          * or unconf->cl_verifier don't match the conf record.
 876          */
 877         if ((conf && !unconf) ||
 878             ((conf && unconf) &&
 879              (!cmp_verf(&conf->cl_verifier, &unconf->cl_verifier) ||
 880               !cmp_name(&conf->cl_name, &unconf->cl_name)))) {
 881                 if (!cmp_creds(&conf->cl_cred,&rqstp->rq_cred)) {
 882                         status = nfserr_clid_inuse;
 883                 } else {
 884                         clp = conf;
 885                         status = nfs_ok;
 886                 }
 887                 goto out;
 888         }
 889         /* CASE 3:
 890          * conf record not found.
 891          * unconf record found.
 892          * unconf->cl_confirm matches input confirm
 893          */
 894         if (!conf && unconf && cmp_verf(&unconf->cl_confirm, &confirm)) {
 895                 if (!cmp_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
 896                         status = nfserr_clid_inuse;
 897                 } else {
 898                         status = nfs_ok;
 899                         clp = unconf;
 900                         move_to_confirmed(unconf);
 901                 }
 902                 goto out;
 903         }
 904         /* CASE 4:
 905          * conf record not found, or if conf, then conf->cl_confirm does not
 906          * match input confirm.
 907          * unconf record not found, or if unconf, then unconf->cl_confirm
 908          * does not match input confirm.
 909          */
 910         if ((!conf || (conf && !cmp_verf(&conf->cl_confirm, &confirm))) &&
 911             (!unconf || (unconf && !cmp_verf(&unconf->cl_confirm, &confirm)))) {
 912                 status = nfserr_stale_clientid;
 913                 goto out;
 914         }
 915         /* check that we have hit one of the cases...*/
 916         status = nfserr_inval;
 917         goto out;
 918 out:
 919         if (!status)
 920                 nfsd4_probe_callback(clp);
 921         nfs4_unlock_state();
 922         return status;
 923 }
 924
 925 /*
 926  * Open owner state (share locks)
 927  */
 928
 929 /* hash tables for nfs4_stateowner */
 930 #define OWNER_HASH_BITS              8
 931 #define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
 932 #define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
 933
 934 #define ownerid_hashval(id) \
 935         ((id) & OWNER_HASH_MASK)
 936 #define ownerstr_hashval(clientid, ownername) \
 937         (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
 938
 939 static struct list_head ownerid_hashtbl[OWNER_HASH_SIZE];
 940 static struct list_head ownerstr_hashtbl[OWNER_HASH_SIZE];
 941
 942 /* hash table for nfs4_file */
 943 #define FILE_HASH_BITS                   8
 944 #define FILE_HASH_SIZE                  (1 << FILE_HASH_BITS)
 945 #define FILE_HASH_MASK                  (FILE_HASH_SIZE - 1)
 946 /* hash table for (open)nfs4_stateid */
 947 #define STATEID_HASH_BITS              10
 948 #define STATEID_HASH_SIZE              (1 << STATEID_HASH_BITS)
 949 #define STATEID_HASH_MASK              (STATEID_HASH_SIZE - 1)
 950
 951 #define file_hashval(x) \
 952         hash_ptr(x, FILE_HASH_BITS)
 953 #define stateid_hashval(owner_id, file_id)  \
 954         (((owner_id) + (file_id)) & STATEID_HASH_MASK)
 955
 956 static struct list_head file_hashtbl[FILE_HASH_SIZE];
 957 static struct list_head stateid_hashtbl[STATEID_HASH_SIZE];
 958
 959 /* OPEN Share state helper functions */
 960 static inline struct nfs4_file *
 961 alloc_init_file(struct inode *ino)
 962 {
 963         struct nfs4_file *fp;
 964         unsigned int hashval = file_hashval(ino);
 965
 966         fp = kmem_cache_alloc(file_slab, GFP_KERNEL);
 967         if (fp) {
 968                 INIT_LIST_HEAD(&fp->fi_hash);
 969                 INIT_LIST_HEAD(&fp->fi_perfile);
 970                 INIT_LIST_HEAD(&fp->fi_del_perfile);
 971                 list_add(&fp->fi_hash, &file_hashtbl[hashval]);
 972                 fp->fi_inode = igrab(ino);
 973                 fp->fi_id = current_fileid++;
 974                 return fp;
 975         }
 976         return NULL;
 977 }
 978
 979 static void
 980 release_all_files(void)
 981 {
 982         int i;
 983         struct nfs4_file *fp;
 984
 985         for (i=0;i<FILE_HASH_SIZE;i++) {
 986                 while (!list_empty(&file_hashtbl[i])) {
 987                         fp = list_entry(file_hashtbl[i].next, struct nfs4_file, fi_hash);
 988                         /* this should never be more than once... */
 989                         if (!list_empty(&fp->fi_perfile) || !list_empty(&fp->fi_del_perfile)) {
 990                                 printk("ERROR: release_all_files: file %p is open, creating dangling state !!!\n",fp);
 991                         }
 992                         release_file(fp);
 993                 }
 994         }
 995 }
 996
 997 static void
 998 nfsd4_free_slab(kmem_cache_t **slab)
 999 {
1000         int status;
1001
1002         if (*slab == NULL)
1003                 return;
1004         status = kmem_cache_destroy(*slab);
1005         *slab = NULL;
1006         WARN_ON(status);
1007 }
1008
1009 static void
1010 nfsd4_free_slabs(void)
1011 {
1012         nfsd4_free_slab(&stateowner_slab);
1013         nfsd4_free_slab(&file_slab);
1014         nfsd4_free_slab(&stateid_slab);
1015 }
1016
1017 static int
1018 nfsd4_init_slabs(void)
1019 {
1020         stateowner_slab = kmem_cache_create("nfsd4_stateowners",
1021                         sizeof(struct nfs4_stateowner), 0, 0, NULL, NULL);
1022         if (stateowner_slab == NULL)
1023                 goto out_nomem;
1024         file_slab = kmem_cache_create("nfsd4_files",
1025                         sizeof(struct nfs4_file), 0, 0, NULL, NULL);
1026         if (file_slab == NULL)
1027                 goto out_nomem;
1028         stateid_slab = kmem_cache_create("nfsd4_stateids",
1029                         sizeof(struct nfs4_stateid), 0, 0, NULL, NULL);
1030         if (stateid_slab == NULL)
1031                 goto out_nomem;
1032         return 0;
1033 out_nomem:
1034         nfsd4_free_slabs();
1035         dprintk("nfsd4: out of memory while initializing nfsv4\n");
1036         return -ENOMEM;
1037 }
1038
1039 void
1040 nfs4_free_stateowner(struct kref *kref)
1041 {
1042         struct nfs4_stateowner *sop =
1043                 container_of(kref, struct nfs4_stateowner, so_ref);
1044         kfree(sop->so_owner.data);
1045         kmem_cache_free(stateowner_slab, sop);
1046 }
1047
1048 static inline struct nfs4_stateowner *
1049 alloc_stateowner(struct xdr_netobj *owner)
1050 {
1051         struct nfs4_stateowner *sop;
1052
1053         if ((sop = kmem_cache_alloc(stateowner_slab, GFP_KERNEL))) {
1054                 if ((sop->so_owner.data = kmalloc(owner->len, GFP_KERNEL))) {
1055                         memcpy(sop->so_owner.data, owner->data, owner->len);
1056                         sop->so_owner.len = owner->len;
1057                         kref_init(&sop->so_ref);
1058                         return sop;
1059                 }
1060                 kmem_cache_free(stateowner_slab, sop);
1061         }
1062         return NULL;
1063 }
1064
1065 static struct nfs4_stateowner *
1066 alloc_init_open_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfsd4_open *open) {
1067         struct nfs4_stateowner *sop;
1068         struct nfs4_replay *rp;
1069         unsigned int idhashval;
1070
1071         if (!(sop = alloc_stateowner(&open->op_owner)))
1072                 return NULL;
1073         idhashval = ownerid_hashval(current_ownerid);
1074         INIT_LIST_HEAD(&sop->so_idhash);
1075         INIT_LIST_HEAD(&sop->so_strhash);
1076         INIT_LIST_HEAD(&sop->so_perclient);
1077         INIT_LIST_HEAD(&sop->so_perfilestate);
1078         INIT_LIST_HEAD(&sop->so_perlockowner);  /* not used */
1079         INIT_LIST_HEAD(&sop->so_close_lru);
1080         sop->so_time = 0;
1081         list_add(&sop->so_idhash, &ownerid_hashtbl[idhashval]);
1082         list_add(&sop->so_strhash, &ownerstr_hashtbl[strhashval]);
1083         list_add(&sop->so_perclient, &clp->cl_perclient);
1084         add_perclient++;
1085         sop->so_is_open_owner = 1;
1086         sop->so_id = current_ownerid++;
1087         sop->so_client = clp;
1088         sop->so_seqid = open->op_seqid;
1089         sop->so_confirmed = 0;
1090         rp = &sop->so_replay;
1091         rp->rp_status = NFSERR_SERVERFAULT;
1092         rp->rp_buflen = 0;
1093         rp->rp_buf = rp->rp_ibuf;
1094         return sop;
1095 }
1096
1097 static void
1098 release_stateid_lockowners(struct nfs4_stateid *open_stp)
1099 {
1100         struct nfs4_stateowner *lock_sop;
1101
1102         while (!list_empty(&open_stp->st_perlockowner)) {
1103                 lock_sop = list_entry(open_stp->st_perlockowner.next,
1104                                 struct nfs4_stateowner, so_perlockowner);
1105                 /* list_del(&open_stp->st_perlockowner);  */
1106                 BUG_ON(lock_sop->so_is_open_owner);
1107                 release_stateowner(lock_sop);
1108         }
1109 }
1110
1111 static void
1112 unhash_stateowner(struct nfs4_stateowner *sop)
1113 {
1114         struct nfs4_stateid *stp;
1115
1116         list_del(&sop->so_idhash);
1117         list_del(&sop->so_strhash);
1118         if (sop->so_is_open_owner) {
1119                 list_del(&sop->so_perclient);
1120                 del_perclient++;
1121         }
1122         list_del(&sop->so_perlockowner);
1123         while (!list_empty(&sop->so_perfilestate)) {
1124                 stp = list_entry(sop->so_perfilestate.next,
1125                         struct nfs4_stateid, st_perfilestate);
1126                 if (sop->so_is_open_owner)
1127                         release_stateid(stp, OPEN_STATE);
1128                 else
1129                         release_stateid(stp, LOCK_STATE);
1130         }
1131 }
1132
1133 static void
1134 release_stateowner(struct nfs4_stateowner *sop)
1135 {
1136         unhash_stateowner(sop);
1137         list_del(&sop->so_close_lru);
1138         nfs4_put_stateowner(sop);
1139 }
1140
1141 static inline void
1142 init_stateid(struct nfs4_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
1143         struct nfs4_stateowner *sop = open->op_stateowner;
1144         unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
1145
1146         INIT_LIST_HEAD(&stp->st_hash);
1147         INIT_LIST_HEAD(&stp->st_perfilestate);
1148         INIT_LIST_HEAD(&stp->st_perlockowner);
1149         INIT_LIST_HEAD(&stp->st_perfile);
1150         list_add(&stp->st_hash, &stateid_hashtbl[hashval]);
1151         list_add(&stp->st_perfilestate, &sop->so_perfilestate);
1152         list_add_perfile++;
1153         list_add(&stp->st_perfile, &fp->fi_perfile);
1154         stp->st_stateowner = sop;
1155         stp->st_file = fp;
1156         stp->st_stateid.si_boot = boot_time;
1157         stp->st_stateid.si_stateownerid = sop->so_id;
1158         stp->st_stateid.si_fileid = fp->fi_id;
1159         stp->st_stateid.si_generation = 0;
1160         stp->st_access_bmap = 0;
1161         stp->st_deny_bmap = 0;
1162         __set_bit(open->op_share_access, &stp->st_access_bmap);
1163         __set_bit(open->op_share_deny, &stp->st_deny_bmap);
1164 }
1165
1166 static void
1167 release_stateid(struct nfs4_stateid *stp, int flags)
1168 {
1169         struct file *filp = stp->st_vfs_file;
1170
1171         list_del(&stp->st_hash);
1172         list_del_perfile++;
1173         list_del(&stp->st_perfile);
1174         list_del(&stp->st_perfilestate);
1175         if (flags & OPEN_STATE) {
1176                 release_stateid_lockowners(stp);
1177                 stp->st_vfs_file = NULL;
1178                 nfsd_close(filp);
1179                 vfsclose++;
1180         } else if (flags & LOCK_STATE)
1181                 locks_remove_posix(filp, (fl_owner_t) stp->st_stateowner);
1182         kmem_cache_free(stateid_slab, stp);
1183         stp = NULL;
1184 }
1185
1186 static void
1187 release_file(struct nfs4_file *fp)
1188 {
1189         list_del(&fp->fi_hash);
1190         iput(fp->fi_inode);
1191         kmem_cache_free(file_slab, fp);
1192 }
1193
1194 void
1195 move_to_close_lru(struct nfs4_stateowner *sop)
1196 {
1197         dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop);
1198
1199         unhash_stateowner(sop);
1200         list_add_tail(&sop->so_close_lru, &close_lru);
1201         sop->so_time = get_seconds();
1202 }
1203
1204 void
1205 release_state_owner(struct nfs4_stateid *stp, int flag)
1206 {
1207         struct nfs4_stateowner *sop = stp->st_stateowner;
1208         struct nfs4_file *fp = stp->st_file;
1209
1210         dprintk("NFSD: release_state_owner\n");
1211         release_stateid(stp, flag);
1212
1213         /* place unused nfs4_stateowners on so_close_lru list to be
1214          * released by the laundromat service after the lease period
1215          * to enable us to handle CLOSE replay
1216          */
1217         if (sop->so_confirmed && list_empty(&sop->so_perfilestate))
1218                 move_to_close_lru(sop);
1219         /* unused nfs4_file's are releseed. XXX slab cache? */
1220         if (list_empty(&fp->fi_perfile) && list_empty(&fp->fi_del_perfile)) {
1221                 release_file(fp);
1222         }
1223 }
1224
1225 static int
1226 cmp_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner, clientid_t *clid) {
1227         return ((sop->so_owner.len == owner->len) &&
1228          !memcmp(sop->so_owner.data, owner->data, owner->len) &&
1229           (sop->so_client->cl_clientid.cl_id == clid->cl_id));
1230 }
1231
1232 static struct nfs4_stateowner *
1233 find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open)
1234 {
1235         struct nfs4_stateowner *so = NULL;
1236
1237         list_for_each_entry(so, &ownerstr_hashtbl[hashval], so_strhash) {
1238                 if (cmp_owner_str(so, &open->op_owner, &open->op_clientid))
1239                         return so;
1240         }
1241         return NULL;
1242 }
1243
1244 /* search file_hashtbl[] for file */
1245 static struct nfs4_file *
1246 find_file(struct inode *ino)
1247 {
1248         unsigned int hashval = file_hashval(ino);
1249         struct nfs4_file *fp;
1250
1251         list_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
1252                 if (fp->fi_inode == ino)
1253                         return fp;
1254         }
1255         return NULL;
1256 }
1257
1258 #define TEST_ACCESS(x) ((x > 0 || x < 4)?1:0)
1259 #define TEST_DENY(x) ((x >= 0 || x < 5)?1:0)
1260
1261 void
1262 set_access(unsigned int *access, unsigned long bmap) {
1263         int i;
1264
1265         *access = 0;
1266         for (i = 1; i < 4; i++) {
1267                 if (test_bit(i, &bmap))
1268                         *access |= i;
1269         }
1270 }
1271
1272 void
1273 set_deny(unsigned int *deny, unsigned long bmap) {
1274         int i;
1275
1276         *deny = 0;
1277         for (i = 0; i < 4; i++) {
1278                 if (test_bit(i, &bmap))
1279                         *deny |= i ;
1280         }
1281 }
1282
1283 static int
1284 test_share(struct nfs4_stateid *stp, struct nfsd4_open *open) {
1285         unsigned int access, deny;
1286
1287         set_access(&access, stp->st_access_bmap);
1288         set_deny(&deny, stp->st_deny_bmap);
1289         if ((access & open->op_share_deny) || (deny & open->op_share_access))
1290                 return 0;
1291         return 1;
1292 }
1293
1294 /*
1295  * Called to check deny when READ with all zero stateid or
1296  * WRITE with all zero or all one stateid
1297  */
1298 int
1299 nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
1300 {
1301         struct inode *ino = current_fh->fh_dentry->d_inode;
1302         struct nfs4_file *fp;
1303         struct nfs4_stateid *stp;
1304
1305         dprintk("NFSD: nfs4_share_conflict\n");
1306
1307         fp = find_file(ino);
1308         if (fp) {
1309         /* Search for conflicting share reservations */
1310                 list_for_each_entry(stp, &fp->fi_perfile, st_perfile) {
1311                         if (test_bit(deny_type, &stp->st_deny_bmap) ||
1312                             test_bit(NFS4_SHARE_DENY_BOTH, &stp->st_deny_bmap))
1313                                 return nfserr_share_denied;
1314                 }
1315         }
1316         return nfs_ok;
1317 }
1318
1319 static inline void
1320 nfs4_file_downgrade(struct file *filp, unsigned int share_access)
1321 {
1322         if (share_access & NFS4_SHARE_ACCESS_WRITE) {
1323                 put_write_access(filp->f_dentry->d_inode);
1324                 filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
1325         }
1326 }
1327
1328 /*
1329  * Recall a delegation
1330  */
1331 static int
1332 do_recall(void *__dp)
1333 {
1334         struct nfs4_delegation *dp = __dp;
1335
1336         daemonize("nfsv4-recall");
1337
1338         nfsd4_cb_recall(dp);
1339         return 0;
1340 }
1341
1342 /*
1343  * Spawn a thread to perform a recall on the delegation represented
1344  * by the lease (file_lock)
1345  *
1346  * Called from break_lease() with lock_kernel() held.
1347  * Note: we assume break_lease will only call this *once* for any given
1348  * lease.
1349  */
1350 static
1351 void nfsd_break_deleg_cb(struct file_lock *fl)
1352 {
1353         struct nfs4_delegation *dp=  (struct nfs4_delegation *)fl->fl_owner;
1354         struct task_struct *t;
1355
1356         dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp,fl);
1357         if (!dp)
1358                 return;
1359
1360         /* We're assuming the state code never drops its reference
1361          * without first removing the lease.  Since we're in this lease
1362          * callback (and since the lease code is serialized by the kernel
1363          * lock) we know the server hasn't removed the lease yet, we know
1364          * it's safe to take a reference: */
1365         atomic_inc(&dp->dl_count);
1366
1367         spin_lock(&recall_lock);
1368         list_add_tail(&dp->dl_recall_lru, &del_recall_lru);
1369         spin_unlock(&recall_lock);
1370
1371         /* only place dl_time is set. protected by lock_kernel*/
1372         dp->dl_time = get_seconds();
1373
1374         /* XXX need to merge NFSD_LEASE_TIME with fs/locks.c:lease_break_time */
1375         fl->fl_break_time = jiffies + NFSD_LEASE_TIME * HZ;
1376
1377         t = kthread_run(do_recall, dp, "%s", "nfs4_cb_recall");
1378         if (IS_ERR(t)) {
1379                 struct nfs4_client *clp = dp->dl_client;
1380
1381                 printk(KERN_INFO "NFSD: Callback thread failed for "
1382                         "for client (clientid %08x/%08x)\n",
1383                         clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
1384                 nfs4_put_delegation(dp);
1385         }
1386 }
1387
1388 /*
1389  * The file_lock is being reapd.
1390  *
1391  * Called by locks_free_lock() with lock_kernel() held.
1392  */
1393 static
1394 void nfsd_release_deleg_cb(struct file_lock *fl)
1395 {
1396         struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
1397
1398         dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl,dp, atomic_read(&dp->dl_count));
1399
1400         if (!(fl->fl_flags & FL_LEASE) || !dp)
1401                 return;
1402         dp->dl_flock = NULL;
1403 }
1404
1405 /*
1406  * Set the delegation file_lock back pointer.
1407  *
1408  * Called from __setlease() with lock_kernel() held.
1409  */
1410 static
1411 void nfsd_copy_lock_deleg_cb(struct file_lock *new, struct file_lock *fl)
1412 {
1413         struct nfs4_delegation *dp = (struct nfs4_delegation *)new->fl_owner;
1414
1415         dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp);
1416         if (!dp)
1417                 return;
1418         dp->dl_flock = new;
1419 }
1420
1421 /*
1422  * Called from __setlease() with lock_kernel() held
1423  */
1424 static
1425 int nfsd_same_client_deleg_cb(struct file_lock *onlist, struct file_lock *try)
1426 {
1427         struct nfs4_delegation *onlistd =
1428                 (struct nfs4_delegation *)onlist->fl_owner;
1429         struct nfs4_delegation *tryd =
1430                 (struct nfs4_delegation *)try->fl_owner;
1431
1432         if (onlist->fl_lmops != try->fl_lmops)
1433                 return 0;
1434
1435         return onlistd->dl_client == tryd->dl_client;
1436 }
1437
1438
1439 static
1440 int nfsd_change_deleg_cb(struct file_lock **onlist, int arg)
1441 {
1442         if (arg & F_UNLCK)
1443                 return lease_modify(onlist, arg);
1444         else
1445                 return -EAGAIN;
1446 }
1447
1448 struct lock_manager_operations nfsd_lease_mng_ops = {
1449         .fl_break = nfsd_break_deleg_cb,
1450         .fl_release_private = nfsd_release_deleg_cb,
1451         .fl_copy_lock = nfsd_copy_lock_deleg_cb,
1452         .fl_mylease = nfsd_same_client_deleg_cb,
1453         .fl_change = nfsd_change_deleg_cb,
1454 };
1455
1456
1457 /*
1458  * nfsd4_process_open1()
1459  *      lookup stateowner.
1460  *              found:
1461  *                      check confirmed
1462  *                              confirmed:
1463  *                                      check seqid
1464  *                              not confirmed:
1465  *                                      delete owner
1466  *                                      create new owner
1467  *              notfound:
1468  *                      verify clientid
1469  *                      create new owner
1470  *
1471  * called with nfs4_lock_state() held.
1472  */
1473 int
1474 nfsd4_process_open1(struct nfsd4_open *open)
1475 {
1476         int status;
1477         clientid_t *clientid = &open->op_clientid;
1478         struct nfs4_client *clp = NULL;
1479         unsigned int strhashval;
1480         struct nfs4_stateowner *sop = NULL;
1481
1482         status = nfserr_inval;
1483         if (!check_name(open->op_owner))
1484                 goto out;
1485
1486         if (STALE_CLIENTID(&open->op_clientid))
1487                 return nfserr_stale_clientid;
1488
1489         strhashval = ownerstr_hashval(clientid->cl_id, open->op_owner);
1490         sop = find_openstateowner_str(strhashval, open);
1491         if (sop) {
1492                 open->op_stateowner = sop;
1493                 /* check for replay */
1494                 if (open->op_seqid == sop->so_seqid){
1495                         if (sop->so_replay.rp_buflen)
1496                                 return NFSERR_REPLAY_ME;
1497                         else {
1498                                 /* The original OPEN failed so spectacularly
1499                                  * that we don't even have replay data saved!
1500                                  * Therefore, we have no choice but to continue
1501                                  * processing this OPEN; presumably, we'll
1502                                  * fail again for the same reason.
1503                                  */
1504                                 dprintk("nfsd4_process_open1:"
1505                                         " replay with no replay cache\n");
1506                                 goto renew;
1507                         }
1508                 } else if (sop->so_confirmed) {
1509                         if (open->op_seqid == sop->so_seqid + 1)
1510                                 goto renew;
1511                         status = nfserr_bad_seqid;
1512                         goto out;
1513                 } else {
1514                         /* If we get here, we received an OPEN for an
1515                          * unconfirmed nfs4_stateowner. Since the seqid's are
1516                          * different, purge the existing nfs4_stateowner, and
1517                          * instantiate a new one.
1518                          */
1519                         clp = sop->so_client;
1520                         release_stateowner(sop);
1521                 }
1522         } else {
1523                 /* nfs4_stateowner not found.
1524                  * Verify clientid and instantiate new nfs4_stateowner.
1525                  * If verify fails this is presumably the result of the
1526                  * client's lease expiring.
1527                  */
1528                 status = nfserr_expired;
1529                 clp = find_confirmed_client(clientid);
1530                 if (clp == NULL)
1531                         goto out;
1532         }
1533         status = nfserr_resource;
1534         sop = alloc_init_open_stateowner(strhashval, clp, open);
1535         if (sop == NULL)
1536                 goto out;
1537         open->op_stateowner = sop;
1538 renew:
1539         status = nfs_ok;
1540         renew_client(sop->so_client);
1541 out:
1542         if (status && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
1543                 status = nfserr_reclaim_bad;
1544         return status;
1545 }
1546
1547 static inline int
1548 nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
1549 {
1550         if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
1551                 return nfserr_openmode;
1552         else
1553                 return nfs_ok;
1554 }
1555
1556 static struct nfs4_delegation *
1557 find_delegation_file(struct nfs4_file *fp, stateid_t *stid)
1558 {
1559         struct nfs4_delegation *dp;
1560
1561         list_for_each_entry(dp, &fp->fi_del_perfile, dl_del_perfile) {
1562                 if (dp->dl_stateid.si_stateownerid == stid->si_stateownerid)
1563                         return dp;
1564         }
1565         return NULL;
1566 }
1567
1568 static int
1569 nfs4_check_deleg(struct nfs4_file *fp, struct nfsd4_open *open,
1570                 struct nfs4_delegation **dp)
1571 {
1572         int flags;
1573         int status = nfserr_bad_stateid;
1574
1575         *dp = find_delegation_file(fp, &open->op_delegate_stateid);
1576         if (*dp == NULL)
1577                 goto out;
1578         flags = open->op_share_access == NFS4_SHARE_ACCESS_READ ?
1579                                                 RD_STATE : WR_STATE;
1580         status = nfs4_check_delegmode(*dp, flags);
1581         if (status)
1582                 *dp = NULL;
1583 out:
1584         if (open->op_claim_type != NFS4_OPEN_CLAIM_DELEGATE_CUR)
1585                 return nfs_ok;
1586         if (status)
1587                 return status;
1588         open->op_stateowner->so_confirmed = 1;
1589         return nfs_ok;
1590 }
1591
1592 static int
1593 nfs4_check_open(struct nfs4_file *fp, struct nfsd4_open *open, struct nfs4_stateid **stpp)
1594 {
1595         struct nfs4_stateid *local;
1596         int status = nfserr_share_denied;
1597         struct nfs4_stateowner *sop = open->op_stateowner;
1598
1599         list_for_each_entry(local, &fp->fi_perfile, st_perfile) {
1600                 /* ignore lock owners */
1601                 if (local->st_stateowner->so_is_open_owner == 0)
1602                         continue;
1603                 /* remember if we have seen this open owner */
1604                 if (local->st_stateowner == sop)
1605                         *stpp = local;
1606                 /* check for conflicting share reservations */
1607                 if (!test_share(local, open))
1608                         goto out;
1609         }
1610         status = 0;
1611 out:
1612         return status;
1613 }
1614
1615 static inline struct nfs4_stateid *
1616 nfs4_alloc_stateid(void)
1617 {
1618         return kmem_cache_alloc(stateid_slab, GFP_KERNEL);
1619 }
1620
1621 static int
1622 nfs4_new_open(struct svc_rqst *rqstp, struct nfs4_stateid **stpp,
1623                 struct nfs4_delegation *dp,
1624                 struct svc_fh *cur_fh, int flags)
1625 {
1626         struct nfs4_stateid *stp;
1627
1628         stp = nfs4_alloc_stateid();
1629         if (stp == NULL)
1630                 return nfserr_resource;
1631
1632         if (dp) {
1633                 get_file(dp->dl_vfs_file);
1634                 stp->st_vfs_file = dp->dl_vfs_file;
1635         } else {
1636                 int status;
1637                 status = nfsd_open(rqstp, cur_fh, S_IFREG, flags,
1638                                 &stp->st_vfs_file);
1639                 if (status) {
1640                         if (status == nfserr_dropit)
1641                                 status = nfserr_jukebox;
1642                         kmem_cache_free(stateid_slab, stp);
1643                         return status;
1644                 }
1645         }
1646         vfsopen++;
1647         *stpp = stp;
1648         return 0;
1649 }
1650
1651 static inline int
1652 nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
1653                 struct nfsd4_open *open)
1654 {
1655         struct iattr iattr = {
1656                 .ia_valid = ATTR_SIZE,
1657                 .ia_size = 0,
1658         };
1659         if (!open->op_truncate)
1660                 return 0;
1661         if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
1662                 return -EINVAL;
1663         return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
1664 }
1665
1666 static int
1667 nfs4_upgrade_open(struct svc_rqst *rqstp, struct svc_fh *cur_fh, struct nfs4_stateid *stp, struct nfsd4_open *open)
1668 {
1669         struct file *filp = stp->st_vfs_file;
1670         struct inode *inode = filp->f_dentry->d_inode;
1671         unsigned int share_access;
1672         int status;
1673
1674         set_access(&share_access, stp->st_access_bmap);
1675         share_access = ~share_access;
1676         share_access &= open->op_share_access;
1677
1678         if (!(share_access & NFS4_SHARE_ACCESS_WRITE))
1679                 return nfsd4_truncate(rqstp, cur_fh, open);
1680
1681         status = get_write_access(inode);
1682         if (status)
1683                 return nfserrno(status);
1684         status = nfsd4_truncate(rqstp, cur_fh, open);
1685         if (status) {
1686                 put_write_access(inode);
1687                 return status;
1688         }
1689         /* remember the open */
1690         filp->f_mode = (filp->f_mode | FMODE_WRITE) & ~FMODE_READ;
1691         set_bit(open->op_share_access, &stp->st_access_bmap);
1692         set_bit(open->op_share_deny, &stp->st_deny_bmap);
1693
1694         return nfs_ok;
1695 }
1696
1697
1698 /* decrement seqid on successful reclaim, it will be bumped in encode_open */
1699 static void
1700 nfs4_set_claim_prev(struct nfsd4_open *open, int *status)
1701 {
1702         if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS) {
1703                 if (*status)
1704                         *status = nfserr_reclaim_bad;
1705                 else {
1706                         open->op_stateowner->so_confirmed = 1;
1707                         open->op_stateowner->so_seqid--;
1708                 }
1709         }
1710 }
1711
1712 /*
1713  * Attempt to hand out a delegation.
1714  */
1715 static void
1716 nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_stateid *stp)
1717 {
1718         struct nfs4_delegation *dp;
1719         struct nfs4_stateowner *sop = stp->st_stateowner;
1720         struct nfs4_callback *cb = &sop->so_client->cl_callback;
1721         struct file_lock fl, *flp = &fl;
1722         int status, flag = 0;
1723
1724         flag = NFS4_OPEN_DELEGATE_NONE;
1725         if (open->op_claim_type != NFS4_OPEN_CLAIM_NULL
1726              || !atomic_read(&cb->cb_set) || !sop->so_confirmed)
1727                 goto out;
1728
1729         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1730                 flag = NFS4_OPEN_DELEGATE_WRITE;
1731         else
1732                 flag = NFS4_OPEN_DELEGATE_READ;
1733
1734         dp = alloc_init_deleg(sop->so_client, stp, fh, flag);
1735         if (dp == NULL) {
1736                 flag = NFS4_OPEN_DELEGATE_NONE;
1737                 goto out;
1738         }
1739         locks_init_lock(&fl);
1740         fl.fl_lmops = &nfsd_lease_mng_ops;
1741         fl.fl_flags = FL_LEASE;
1742         fl.fl_end = OFFSET_MAX;
1743         fl.fl_owner =  (fl_owner_t)dp;
1744         fl.fl_file = stp->st_vfs_file;
1745         fl.fl_pid = current->tgid;
1746
1747         /* setlease checks to see if delegation should be handed out.
1748          * the lock_manager callbacks fl_mylease and fl_change are used
1749          */
1750         if ((status = setlease(stp->st_vfs_file,
1751                 flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK, &flp))) {
1752                 dprintk("NFSD: setlease failed [%d], no delegation\n", status);
1753                 unhash_delegation(dp);
1754                 flag = NFS4_OPEN_DELEGATE_NONE;
1755                 goto out;
1756         }
1757
1758         memcpy(&open->op_delegate_stateid, &dp->dl_stateid, sizeof(dp->dl_stateid));
1759
1760         dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1761                      dp->dl_stateid.si_boot,
1762                      dp->dl_stateid.si_stateownerid,
1763                      dp->dl_stateid.si_fileid,
1764                      dp->dl_stateid.si_generation);
1765 out:
1766         open->op_delegate_type = flag;
1767 }
1768
1769 /*
1770  * called with nfs4_lock_state() held.
1771  */
1772 int
1773 nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
1774 {
1775         struct nfs4_file *fp = NULL;
1776         struct inode *ino = current_fh->fh_dentry->d_inode;
1777         struct nfs4_stateid *stp = NULL;
1778         struct nfs4_delegation *dp = NULL;
1779         int status;
1780
1781         status = nfserr_inval;
1782         if (!TEST_ACCESS(open->op_share_access) || !TEST_DENY(open->op_share_deny))
1783                 goto out;
1784         /*
1785          * Lookup file; if found, lookup stateid and check open request,
1786          * and check for delegations in the process of being recalled.
1787          * If not found, create the nfs4_file struct
1788          */
1789         fp = find_file(ino);
1790         if (fp) {
1791                 if ((status = nfs4_check_open(fp, open, &stp)))
1792                         goto out;
1793                 status = nfs4_check_deleg(fp, open, &dp);
1794                 if (status)
1795                         goto out;
1796         } else {
1797                 status = nfserr_bad_stateid;
1798                 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
1799                         goto out;
1800                 status = nfserr_resource;
1801                 fp = alloc_init_file(ino);
1802                 if (fp == NULL)
1803                         goto out;
1804         }
1805
1806         /*
1807          * OPEN the file, or upgrade an existing OPEN.
1808          * If truncate fails, the OPEN fails.
1809          */
1810         if (stp) {
1811                 /* Stateid was found, this is an OPEN upgrade */
1812                 status = nfs4_upgrade_open(rqstp, current_fh, stp, open);
1813                 if (status)
1814                         goto out;
1815         } else {
1816                 /* Stateid was not found, this is a new OPEN */
1817                 int flags = 0;
1818                 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
1819                         flags = MAY_WRITE;
1820                 else
1821                         flags = MAY_READ;
1822                 status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
1823                 if (status)
1824                         goto out;
1825                 init_stateid(stp, fp, open);
1826                 status = nfsd4_truncate(rqstp, current_fh, open);
1827                 if (status) {
1828                         release_stateid(stp, OPEN_STATE);
1829                         goto out;
1830                 }
1831         }
1832         memcpy(&open->op_stateid, &stp->st_stateid, sizeof(stateid_t));
1833
1834         /*
1835         * Attempt to hand out a delegation. No error return, because the
1836         * OPEN succeeds even if we fail.
1837         */
1838         nfs4_open_delegation(current_fh, open, stp);
1839
1840         status = nfs_ok;
1841
1842         dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1843                     stp->st_stateid.si_boot, stp->st_stateid.si_stateownerid,
1844                     stp->st_stateid.si_fileid, stp->st_stateid.si_generation);
1845 out:
1846         /* take the opportunity to clean up unused state */
1847         if (fp && list_empty(&fp->fi_perfile) && list_empty(&fp->fi_del_perfile))
1848                 release_file(fp);
1849
1850         /* CLAIM_PREVIOUS has different error returns */
1851         nfs4_set_claim_prev(open, &status);
1852         /*
1853         * To finish the open response, we just need to set the rflags.
1854         */
1855         open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
1856         if (!open->op_stateowner->so_confirmed)
1857                 open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
1858
1859         return status;
1860 }
1861
1862 static struct work_struct laundromat_work;
1863 static void laundromat_main(void *);
1864 static DECLARE_WORK(laundromat_work, laundromat_main, NULL);
1865
1866 int
1867 nfsd4_renew(clientid_t *clid)
1868 {
1869         struct nfs4_client *clp;
1870         int status;
1871
1872         nfs4_lock_state();
1873         dprintk("process_renew(%08x/%08x): starting\n",
1874                         clid->cl_boot, clid->cl_id);
1875         status = nfserr_stale_clientid;
1876         if (STALE_CLIENTID(clid))
1877                 goto out;
1878         clp = find_confirmed_client(clid);
1879         status = nfserr_expired;
1880         if (clp == NULL) {
1881                 /* We assume the client took too long to RENEW. */
1882                 dprintk("nfsd4_renew: clientid not found!\n");
1883                 goto out;
1884         }
1885         renew_client(clp);
1886         status = nfserr_cb_path_down;
1887         if (!list_empty(&clp->cl_del_perclnt)
1888                         && !atomic_read(&clp->cl_callback.cb_set))
1889                 goto out;
1890         status = nfs_ok;
1891 out:
1892         nfs4_unlock_state();
1893         return status;
1894 }
1895
1896 time_t
1897 nfs4_laundromat(void)
1898 {
1899         struct nfs4_client *clp;
1900         struct nfs4_stateowner *sop;
1901         struct nfs4_delegation *dp;
1902         struct list_head *pos, *next, reaplist;
1903         time_t cutoff = get_seconds() - NFSD_LEASE_TIME;
1904         time_t t, clientid_val = NFSD_LEASE_TIME;
1905         time_t u, test_val = NFSD_LEASE_TIME;
1906
1907         nfs4_lock_state();
1908
1909         dprintk("NFSD: laundromat service - starting\n");
1910         list_for_each_safe(pos, next, &client_lru) {
1911                 clp = list_entry(pos, struct nfs4_client, cl_lru);
1912                 if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
1913                         t = clp->cl_time - cutoff;
1914                         if (clientid_val > t)
1915                                 clientid_val = t;
1916                         break;
1917                 }
1918                 dprintk("NFSD: purging unused client (clientid %08x)\n",
1919                         clp->cl_clientid.cl_id);
1920                 expire_client(clp);
1921         }
1922         INIT_LIST_HEAD(&reaplist);
1923         spin_lock(&recall_lock);
1924         list_for_each_safe(pos, next, &del_recall_lru) {
1925                 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1926                 if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
1927                         u = dp->dl_time - cutoff;
1928                         if (test_val > u)
1929                                 test_val = u;
1930                         break;
1931                 }
1932                 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1933                                     dp, dp->dl_flock);
1934                 list_move(&dp->dl_recall_lru, &reaplist);
1935         }
1936         spin_unlock(&recall_lock);
1937         list_for_each_safe(pos, next, &reaplist) {
1938                 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
1939                 list_del_init(&dp->dl_recall_lru);
1940                 unhash_delegation(dp);
1941         }
1942         test_val = NFSD_LEASE_TIME;
1943         list_for_each_safe(pos, next, &close_lru) {
1944                 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
1945                 if (time_after((unsigned long)sop->so_time, (unsigned long)cutoff)) {
1946                         u = sop->so_time - cutoff;
1947                         if (test_val > u)
1948                                 test_val = u;
1949                         break;
1950                 }
1951                 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1952                         sop->so_id);
1953                 list_del(&sop->so_close_lru);
1954                 nfs4_put_stateowner(sop);
1955         }
1956         if (clientid_val < NFSD_LAUNDROMAT_MINTIMEOUT)
1957                 clientid_val = NFSD_LAUNDROMAT_MINTIMEOUT;
1958         nfs4_unlock_state();
1959         return clientid_val;
1960 }
1961
1962 void
1963 laundromat_main(void *not_used)
1964 {
1965         time_t t;
1966
1967         t = nfs4_laundromat();
1968         dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
1969         schedule_delayed_work(&laundromat_work, t*HZ);
1970 }
1971
1972 /* search ownerid_hashtbl[] and close_lru for stateid owner
1973  * (stateid->si_stateownerid)
1974  */
1975 struct nfs4_stateowner *
1976 find_openstateowner_id(u32 st_id, int flags) {
1977         struct nfs4_stateowner *local = NULL;
1978
1979         dprintk("NFSD: find_openstateowner_id %d\n", st_id);
1980         if (flags & CLOSE_STATE) {
1981                 list_for_each_entry(local, &close_lru, so_close_lru) {
1982                         if (local->so_id == st_id)
1983                                 return local;
1984                 }
1985         }
1986         return NULL;
1987 }
1988
1989 static inline int
1990 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stateid *stp)
1991 {
1992         return fhp->fh_dentry->d_inode != stp->st_vfs_file->f_dentry->d_inode;
1993 }
1994
1995 static int
1996 STALE_STATEID(stateid_t *stateid)
1997 {
1998         if (stateid->si_boot == boot_time)
1999                 return 0;
2000         printk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
2001                 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2002                 stateid->si_generation);
2003         return 1;
2004 }
2005
2006 static inline int
2007 access_permit_read(unsigned long access_bmap)
2008 {
2009         return test_bit(NFS4_SHARE_ACCESS_READ, &access_bmap) ||
2010                 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap) ||
2011                 test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap);
2012 }
2013
2014 static inline int
2015 access_permit_write(unsigned long access_bmap)
2016 {
2017         return test_bit(NFS4_SHARE_ACCESS_WRITE, &access_bmap) ||
2018                 test_bit(NFS4_SHARE_ACCESS_BOTH, &access_bmap);
2019 }
2020
2021 static
2022 int nfs4_check_openmode(struct nfs4_stateid *stp, int flags)
2023 {
2024         int status = nfserr_openmode;
2025
2026         if ((flags & WR_STATE) && (!access_permit_write(stp->st_access_bmap)))
2027                 goto out;
2028         if ((flags & RD_STATE) && (!access_permit_read(stp->st_access_bmap)))
2029                 goto out;
2030         status = nfs_ok;
2031 out:
2032         return status;
2033 }
2034
2035 static inline int
2036 check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
2037 {
2038         /* Trying to call delegreturn with a special stateid? Yuch: */
2039         if (!(flags & (RD_STATE | WR_STATE)))
2040                 return nfserr_bad_stateid;
2041         else if (ONE_STATEID(stateid) && (flags & RD_STATE))
2042                 return nfs_ok;
2043         else if (nfs4_in_grace()) {
2044                 /* Answer in remaining cases depends on existance of
2045                  * conflicting state; so we must wait out the grace period. */
2046                 return nfserr_grace;
2047         } else if (flags & WR_STATE)
2048                 return nfs4_share_conflict(current_fh,
2049                                 NFS4_SHARE_DENY_WRITE);
2050         else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2051                 return nfs4_share_conflict(current_fh,
2052                                 NFS4_SHARE_DENY_READ);
2053 }
2054
2055 /*
2056  * Allow READ/WRITE during grace period on recovered state only for files
2057  * that are not able to provide mandatory locking.
2058  */
2059 static inline int
2060 io_during_grace_disallowed(struct inode *inode, int flags)
2061 {
2062         return nfs4_in_grace() && (flags & (RD_STATE | WR_STATE))
2063                 && MANDATORY_LOCK(inode);
2064 }
2065
2066 /*
2067 * Checks for stateid operations
2068 */
2069 int
2070 nfs4_preprocess_stateid_op(struct svc_fh *current_fh, stateid_t *stateid, int flags, struct file **filpp)
2071 {
2072         struct nfs4_stateid *stp = NULL;
2073         struct nfs4_delegation *dp = NULL;
2074         stateid_t *stidp;
2075         struct inode *ino = current_fh->fh_dentry->d_inode;
2076         int status;
2077
2078         dprintk("NFSD: preprocess_stateid_op: stateid = (%08x/%08x/%08x/%08x)\n",
2079                 stateid->si_boot, stateid->si_stateownerid,
2080                 stateid->si_fileid, stateid->si_generation);
2081         if (filpp)
2082                 *filpp = NULL;
2083
2084         if (io_during_grace_disallowed(ino, flags))
2085                 return nfserr_grace;
2086
2087         if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
2088                 return check_special_stateids(current_fh, stateid, flags);
2089
2090         /* STALE STATEID */
2091         status = nfserr_stale_stateid;
2092         if (STALE_STATEID(stateid))
2093                 goto out;
2094
2095         /* BAD STATEID */
2096         status = nfserr_bad_stateid;
2097         if (!stateid->si_fileid) { /* delegation stateid */
2098                 if(!(dp = find_delegation_stateid(ino, stateid))) {
2099                         dprintk("NFSD: delegation stateid not found\n");
2100                         if (nfs4_in_grace())
2101                                 status = nfserr_grace;
2102                         goto out;
2103                 }
2104                 stidp = &dp->dl_stateid;
2105         } else { /* open or lock stateid */
2106                 if (!(stp = find_stateid(stateid, flags))) {
2107                         dprintk("NFSD: open or lock stateid not found\n");
2108                         if (nfs4_in_grace())
2109                                 status = nfserr_grace;
2110                         goto out;
2111                 }
2112                 if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))
2113                         goto out;
2114                 if (!stp->st_stateowner->so_confirmed)
2115                         goto out;
2116                 stidp = &stp->st_stateid;
2117         }
2118         if (stateid->si_generation > stidp->si_generation)
2119                 goto out;
2120
2121         /* OLD STATEID */
2122         status = nfserr_old_stateid;
2123         if (stateid->si_generation < stidp->si_generation)
2124                 goto out;
2125         if (stp) {
2126                 if ((status = nfs4_check_openmode(stp,flags)))
2127                         goto out;
2128                 renew_client(stp->st_stateowner->so_client);
2129                 if (filpp)
2130                         *filpp = stp->st_vfs_file;
2131         } else if (dp) {
2132                 if ((status = nfs4_check_delegmode(dp, flags)))
2133                         goto out;
2134                 renew_client(dp->dl_client);
2135                 if (flags & DELEG_RET)
2136                         unhash_delegation(dp);
2137                 if (filpp)
2138                         *filpp = dp->dl_vfs_file;
2139         }
2140         status = nfs_ok;
2141 out:
2142         return status;
2143 }
2144
2145
2146 /*
2147  * Checks for sequence id mutating operations.
2148  */
2149 int
2150 nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *stateid, int flags, struct nfs4_stateowner **sopp, struct nfs4_stateid **stpp, clientid_t *lockclid)
2151 {
2152         int status;
2153         struct nfs4_stateid *stp;
2154         struct nfs4_stateowner *sop;
2155
2156         dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2157                         "stateid = (%08x/%08x/%08x/%08x)\n", seqid,
2158                 stateid->si_boot, stateid->si_stateownerid, stateid->si_fileid,
2159                 stateid->si_generation);
2160
2161         *stpp = NULL;
2162         *sopp = NULL;
2163
2164         status = nfserr_bad_stateid;
2165         if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
2166                 printk("NFSD: preprocess_seqid_op: magic stateid!\n");
2167                 goto out;
2168         }
2169
2170         status = nfserr_stale_stateid;
2171         if (STALE_STATEID(stateid))
2172                 goto out;
2173         /*
2174         * We return BAD_STATEID if filehandle doesn't match stateid,
2175         * the confirmed flag is incorrecly set, or the generation
2176         * number is incorrect.
2177         * If there is no entry in the openfile table for this id,
2178         * we can't always return BAD_STATEID;
2179         * this might be a retransmitted CLOSE which has arrived after
2180         * the openfile has been released.
2181         */
2182         if (!(stp = find_stateid(stateid, flags)))
2183                 goto no_nfs4_stateid;
2184
2185         status = nfserr_bad_stateid;
2186
2187         /* for new lock stateowners:
2188          * check that the lock->v.new.open_stateid
2189          * refers to an open stateowner
2190          *
2191          * check that the lockclid (nfs4_lock->v.new.clientid) is the same
2192          * as the open_stateid->st_stateowner->so_client->clientid
2193          */
2194         if (lockclid) {
2195                 struct nfs4_stateowner *sop = stp->st_stateowner;
2196                 struct nfs4_client *clp = sop->so_client;
2197
2198                 if (!sop->so_is_open_owner)
2199                         goto out;
2200                 if (!cmp_clid(&clp->cl_clientid, lockclid))
2201                         goto out;
2202         }
2203
2204         if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp)) {
2205                 printk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2206                 goto out;
2207         }
2208
2209         *stpp = stp;
2210         *sopp = sop = stp->st_stateowner;
2211
2212         /*
2213         *  We now validate the seqid and stateid generation numbers.
2214         *  For the moment, we ignore the possibility of
2215         *  generation number wraparound.
2216         */
2217         if (seqid != sop->so_seqid + 1)
2218                 goto check_replay;
2219
2220         if (sop->so_confirmed) {
2221                 if (flags & CONFIRM) {
2222                         printk("NFSD: preprocess_seqid_op: expected unconfirmed stateowner!\n");
2223                         goto out;
2224                 }
2225         }
2226         else {
2227                 if (!(flags & CONFIRM)) {
2228                         printk("NFSD: preprocess_seqid_op: stateowner not confirmed yet!\n");
2229                         goto out;
2230                 }
2231         }
2232         if (stateid->si_generation > stp->st_stateid.si_generation) {
2233                 printk("NFSD: preprocess_seqid_op: future stateid?!\n");
2234                 goto out;
2235         }
2236
2237         status = nfserr_old_stateid;
2238         if (stateid->si_generation < stp->st_stateid.si_generation) {
2239                 printk("NFSD: preprocess_seqid_op: old stateid!\n");
2240                 goto out;
2241         }
2242         /* XXX renew the client lease here */
2243         status = nfs_ok;
2244
2245 out:
2246         return status;
2247
2248 no_nfs4_stateid:
2249
2250         /*
2251         * We determine whether this is a bad stateid or a replay,
2252         * starting by trying to look up the stateowner.
2253         * If stateowner is not found - stateid is bad.
2254         */
2255         if (!(sop = find_openstateowner_id(stateid->si_stateownerid, flags))) {
2256                 printk("NFSD: preprocess_seqid_op: no stateowner or nfs4_stateid!\n");
2257                 status = nfserr_bad_stateid;
2258                 goto out;
2259         }
2260         *sopp = sop;
2261
2262 check_replay:
2263         if (seqid == sop->so_seqid) {
2264                 printk("NFSD: preprocess_seqid_op: retransmission?\n");
2265                 /* indicate replay to calling function */
2266                 status = NFSERR_REPLAY_ME;
2267         } else  {
2268                 printk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d\n", sop->so_seqid +1, seqid);
2269
2270                 *sopp = NULL;
2271                 status = nfserr_bad_seqid;
2272         }
2273         goto out;
2274 }
2275
2276 int
2277 nfsd4_open_confirm(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_confirm *oc)
2278 {
2279         int status;
2280         struct nfs4_stateowner *sop;
2281         struct nfs4_stateid *stp;
2282
2283         dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2284                         (int)current_fh->fh_dentry->d_name.len,
2285                         current_fh->fh_dentry->d_name.name);
2286
2287         if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0)))
2288                 goto out;
2289
2290         nfs4_lock_state();
2291
2292         if ((status = nfs4_preprocess_seqid_op(current_fh, oc->oc_seqid,
2293                                         &oc->oc_req_stateid,
2294                                         CHECK_FH | CONFIRM | OPEN_STATE,
2295                                         &oc->oc_stateowner, &stp, NULL)))
2296                 goto out;
2297
2298         sop = oc->oc_stateowner;
2299         sop->so_confirmed = 1;
2300         update_stateid(&stp->st_stateid);
2301         memcpy(&oc->oc_resp_stateid, &stp->st_stateid, sizeof(stateid_t));
2302         dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2303                 "stateid=(%08x/%08x/%08x/%08x)\n", oc->oc_seqid,
2304                          stp->st_stateid.si_boot,
2305                          stp->st_stateid.si_stateownerid,
2306                          stp->st_stateid.si_fileid,
2307                          stp->st_stateid.si_generation);
2308 out:
2309         if (oc->oc_stateowner)
2310                 nfs4_get_stateowner(oc->oc_stateowner);
2311         nfs4_unlock_state();
2312         return status;
2313 }
2314
2315
2316 /*
2317  * unset all bits in union bitmap (bmap) that
2318  * do not exist in share (from successful OPEN_DOWNGRADE)
2319  */
2320 static void
2321 reset_union_bmap_access(unsigned long access, unsigned long *bmap)
2322 {
2323         int i;
2324         for (i = 1; i < 4; i++) {
2325                 if ((i & access) != i)
2326                         __clear_bit(i, bmap);
2327         }
2328 }
2329
2330 static void
2331 reset_union_bmap_deny(unsigned long deny, unsigned long *bmap)
2332 {
2333         int i;
2334         for (i = 0; i < 4; i++) {
2335                 if ((i & deny) != i)
2336                         __clear_bit(i, bmap);
2337         }
2338 }
2339
2340 int
2341 nfsd4_open_downgrade(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open_downgrade *od)
2342 {
2343         int status;
2344         struct nfs4_stateid *stp;
2345         unsigned int share_access;
2346
2347         dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2348                         (int)current_fh->fh_dentry->d_name.len,
2349                         current_fh->fh_dentry->d_name.name);
2350
2351         if (!TEST_ACCESS(od->od_share_access) || !TEST_DENY(od->od_share_deny))
2352                 return nfserr_inval;
2353
2354         nfs4_lock_state();
2355         if ((status = nfs4_preprocess_seqid_op(current_fh, od->od_seqid,
2356                                         &od->od_stateid,
2357                                         CHECK_FH | OPEN_STATE,
2358                                         &od->od_stateowner, &stp, NULL)))
2359                 goto out;
2360
2361         status = nfserr_inval;
2362         if (!test_bit(od->od_share_access, &stp->st_access_bmap)) {
2363                 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2364                         stp->st_access_bmap, od->od_share_access);
2365                 goto out;
2366         }
2367         if (!test_bit(od->od_share_deny, &stp->st_deny_bmap)) {
2368                 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2369                         stp->st_deny_bmap, od->od_share_deny);
2370                 goto out;
2371         }
2372         set_access(&share_access, stp->st_access_bmap);
2373         nfs4_file_downgrade(stp->st_vfs_file,
2374                             share_access & ~od->od_share_access);
2375
2376         reset_union_bmap_access(od->od_share_access, &stp->st_access_bmap);
2377         reset_union_bmap_deny(od->od_share_deny, &stp->st_deny_bmap);
2378
2379         update_stateid(&stp->st_stateid);
2380         memcpy(&od->od_stateid, &stp->st_stateid, sizeof(stateid_t));
2381         status = nfs_ok;
2382 out:
2383         if (od->od_stateowner)
2384                 nfs4_get_stateowner(od->od_stateowner);
2385         nfs4_unlock_state();
2386         return status;
2387 }
2388
2389 /*
2390  * nfs4_unlock_state() called after encode
2391  */
2392 int
2393 nfsd4_close(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_close *close)
2394 {
2395         int status;
2396         struct nfs4_stateid *stp;
2397
2398         dprintk("NFSD: nfsd4_close on file %.*s\n",
2399                         (int)current_fh->fh_dentry->d_name.len,
2400                         current_fh->fh_dentry->d_name.name);
2401
2402         nfs4_lock_state();
2403         /* check close_lru for replay */
2404         if ((status = nfs4_preprocess_seqid_op(current_fh, close->cl_seqid,
2405                                         &close->cl_stateid,
2406                                         CHECK_FH | OPEN_STATE | CLOSE_STATE,
2407                                         &close->cl_stateowner, &stp, NULL)))
2408                 goto out;
2409         /*
2410         *  Return success, but first update the stateid.
2411         */
2412         status = nfs_ok;
2413         update_stateid(&stp->st_stateid);
2414         memcpy(&close->cl_stateid, &stp->st_stateid, sizeof(stateid_t));
2415
2416         /* release_state_owner() calls nfsd_close() if needed */
2417         release_state_owner(stp, OPEN_STATE);
2418 out:
2419         if (close->cl_stateowner)
2420                 nfs4_get_stateowner(close->cl_stateowner);
2421         nfs4_unlock_state();
2422         return status;
2423 }
2424
2425 int
2426 nfsd4_delegreturn(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_delegreturn *dr)
2427 {
2428         int status;
2429
2430         if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0)))
2431                 goto out;
2432
2433         nfs4_lock_state();
2434         status = nfs4_preprocess_stateid_op(current_fh, &dr->dr_stateid, DELEG_RET, NULL);
2435         nfs4_unlock_state();
2436 out:
2437         return status;
2438 }
2439
2440
2441 /*
2442  * Lock owner state (byte-range locks)
2443  */
2444 #define LOFF_OVERFLOW(start, len)      ((u64)(len) > ~(u64)(start))
2445 #define LOCK_HASH_BITS              8
2446 #define LOCK_HASH_SIZE             (1 << LOCK_HASH_BITS)
2447 #define LOCK_HASH_MASK             (LOCK_HASH_SIZE - 1)
2448
2449 #define lockownerid_hashval(id) \
2450         ((id) & LOCK_HASH_MASK)
2451
2452 static inline unsigned int
2453 lock_ownerstr_hashval(struct inode *inode, u32 cl_id,
2454                 struct xdr_netobj *ownername)
2455 {
2456         return (file_hashval(inode) + cl_id
2457                         + opaque_hashval(ownername->data, ownername->len))
2458                 & LOCK_HASH_MASK;
2459 }
2460
2461 static struct list_head lock_ownerid_hashtbl[LOCK_HASH_SIZE];
2462 static struct list_head lock_ownerstr_hashtbl[LOCK_HASH_SIZE];
2463 static struct list_head lockstateid_hashtbl[STATEID_HASH_SIZE];
2464
2465 struct nfs4_stateid *
2466 find_stateid(stateid_t *stid, int flags)
2467 {
2468         struct nfs4_stateid *local = NULL;
2469         u32 st_id = stid->si_stateownerid;
2470         u32 f_id = stid->si_fileid;
2471         unsigned int hashval;
2472
2473         dprintk("NFSD: find_stateid flags 0x%x\n",flags);
2474         if ((flags & LOCK_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2475                 hashval = stateid_hashval(st_id, f_id);
2476                 list_for_each_entry(local, &lockstateid_hashtbl[hashval], st_hash) {
2477                         if ((local->st_stateid.si_stateownerid == st_id) &&
2478                             (local->st_stateid.si_fileid == f_id))
2479                                 return local;
2480                 }
2481         }
2482         if ((flags & OPEN_STATE) || (flags & RD_STATE) || (flags & WR_STATE)) {
2483                 hashval = stateid_hashval(st_id, f_id);
2484                 list_for_each_entry(local, &stateid_hashtbl[hashval], st_hash) {
2485                         if ((local->st_stateid.si_stateownerid == st_id) &&
2486                             (local->st_stateid.si_fileid == f_id))
2487                                 return local;
2488                 }
2489         } else
2490                 printk("NFSD: find_stateid: ERROR: no state flag\n");
2491         return NULL;
2492 }
2493
2494 static struct nfs4_delegation *
2495 find_delegation_stateid(struct inode *ino, stateid_t *stid)
2496 {
2497         struct nfs4_file *fp = NULL;
2498
2499         dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2500                     stid->si_boot, stid->si_stateownerid,
2501                     stid->si_fileid, stid->si_generation);
2502
2503         fp = find_file(ino);
2504         if (fp)
2505                 return find_delegation_file(fp, stid);
2506         return NULL;
2507 }
2508
2509 /*
2510  * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2511  * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2512  * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
2513  * locking, this prevents us from being completely protocol-compliant.  The
2514  * real solution to this problem is to start using unsigned file offsets in
2515  * the VFS, but this is a very deep change!
2516  */
2517 static inline void
2518 nfs4_transform_lock_offset(struct file_lock *lock)
2519 {
2520         if (lock->fl_start < 0)
2521                 lock->fl_start = OFFSET_MAX;
2522         if (lock->fl_end < 0)
2523                 lock->fl_end = OFFSET_MAX;
2524 }
2525
2526 int
2527 nfs4_verify_lock_stateowner(struct nfs4_stateowner *sop, unsigned int hashval)
2528 {
2529         struct nfs4_stateowner *local = NULL;
2530         int status = 0;
2531
2532         if (hashval >= LOCK_HASH_SIZE)
2533                 goto out;
2534         list_for_each_entry(local, &lock_ownerid_hashtbl[hashval], so_idhash) {
2535                 if (local == sop) {
2536                         status = 1;
2537                         goto out;
2538                 }
2539         }
2540 out:
2541         return status;
2542 }
2543
2544
2545 static inline void
2546 nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
2547 {
2548         struct nfs4_stateowner *sop = (struct nfs4_stateowner *) fl->fl_owner;
2549         unsigned int hval = lockownerid_hashval(sop->so_id);
2550
2551         deny->ld_sop = NULL;
2552         if (nfs4_verify_lock_stateowner(sop, hval)) {
2553                 kref_get(&sop->so_ref);
2554                 deny->ld_sop = sop;
2555                 deny->ld_clientid = sop->so_client->cl_clientid;
2556         }
2557         deny->ld_start = fl->fl_start;
2558         deny->ld_length = ~(u64)0;
2559         if (fl->fl_end != ~(u64)0)
2560                 deny->ld_length = fl->fl_end - fl->fl_start + 1;
2561         deny->ld_type = NFS4_READ_LT;
2562         if (fl->fl_type != F_RDLCK)
2563                 deny->ld_type = NFS4_WRITE_LT;
2564 }
2565
2566 static struct nfs4_stateowner *
2567 find_lockstateowner(struct xdr_netobj *owner, clientid_t *clid)
2568 {
2569         struct nfs4_stateowner *local = NULL;
2570         int i;
2571
2572         for (i = 0; i < LOCK_HASH_SIZE; i++) {
2573                 list_for_each_entry(local, &lock_ownerid_hashtbl[i], so_idhash) {
2574                         if (!cmp_owner_str(local, owner, clid))
2575                                 continue;
2576                         return local;
2577                 }
2578         }
2579         return NULL;
2580 }
2581
2582 static struct nfs4_stateowner *
2583 find_lockstateowner_str(struct inode *inode, clientid_t *clid,
2584                 struct xdr_netobj *owner)
2585 {
2586         unsigned int hashval = lock_ownerstr_hashval(inode, clid->cl_id, owner);
2587         struct nfs4_stateowner *op;
2588
2589         list_for_each_entry(op, &lock_ownerstr_hashtbl[hashval], so_strhash) {
2590                 if (cmp_owner_str(op, owner, clid))
2591                         return op;
2592         }
2593         return NULL;
2594 }
2595
2596 /*
2597  * Alloc a lock owner structure.
2598  * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2599  * occured.
2600  *
2601  * strhashval = lock_ownerstr_hashval
2602  * so_seqid = lock->lk_new_lock_seqid - 1: it gets bumped in encode
2603  */
2604
2605 static struct nfs4_stateowner *
2606 alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp, struct nfs4_stateid *open_stp, struct nfsd4_lock *lock) {
2607         struct nfs4_stateowner *sop;
2608         struct nfs4_replay *rp;
2609         unsigned int idhashval;
2610
2611         if (!(sop = alloc_stateowner(&lock->lk_new_owner)))
2612                 return NULL;
2613         idhashval = lockownerid_hashval(current_ownerid);
2614         INIT_LIST_HEAD(&sop->so_idhash);
2615         INIT_LIST_HEAD(&sop->so_strhash);
2616         INIT_LIST_HEAD(&sop->so_perclient);
2617         INIT_LIST_HEAD(&sop->so_perfilestate);
2618         INIT_LIST_HEAD(&sop->so_perlockowner);
2619         INIT_LIST_HEAD(&sop->so_close_lru); /* not used */
2620         sop->so_time = 0;
2621         list_add(&sop->so_idhash, &lock_ownerid_hashtbl[idhashval]);
2622         list_add(&sop->so_strhash, &lock_ownerstr_hashtbl[strhashval]);
2623         list_add(&sop->so_perlockowner, &open_stp->st_perlockowner);
2624         sop->so_is_open_owner = 0;
2625         sop->so_id = current_ownerid++;
2626         sop->so_client = clp;
2627         sop->so_seqid = lock->lk_new_lock_seqid - 1;
2628         sop->so_confirmed = 1;
2629         rp = &sop->so_replay;
2630         rp->rp_status = NFSERR_SERVERFAULT;
2631         rp->rp_buflen = 0;
2632         rp->rp_buf = rp->rp_ibuf;
2633         return sop;
2634 }
2635
2636 struct nfs4_stateid *
2637 alloc_init_lock_stateid(struct nfs4_stateowner *sop, struct nfs4_file *fp, struct nfs4_stateid *open_stp)
2638 {
2639         struct nfs4_stateid *stp;
2640         unsigned int hashval = stateid_hashval(sop->so_id, fp->fi_id);
2641
2642         stp = nfs4_alloc_stateid();
2643         if (stp == NULL)
2644                 goto out;
2645         INIT_LIST_HEAD(&stp->st_hash);
2646         INIT_LIST_HEAD(&stp->st_perfile);
2647         INIT_LIST_HEAD(&stp->st_perfilestate);
2648         INIT_LIST_HEAD(&stp->st_perlockowner); /* not used */
2649         list_add(&stp->st_hash, &lockstateid_hashtbl[hashval]);
2650         list_add(&stp->st_perfile, &fp->fi_perfile);
2651         list_add_perfile++;
2652         list_add(&stp->st_perfilestate, &sop->so_perfilestate);
2653         stp->st_stateowner = sop;
2654         stp->st_file = fp;
2655         stp->st_stateid.si_boot = boot_time;
2656         stp->st_stateid.si_stateownerid = sop->so_id;
2657         stp->st_stateid.si_fileid = fp->fi_id;
2658         stp->st_stateid.si_generation = 0;
2659         stp->st_vfs_file = open_stp->st_vfs_file; /* FIXME refcount?? */
2660         stp->st_access_bmap = open_stp->st_access_bmap;
2661         stp->st_deny_bmap = open_stp->st_deny_bmap;
2662
2663 out:
2664         return stp;
2665 }
2666
2667 int
2668 check_lock_length(u64 offset, u64 length)
2669 {
2670         return ((length == 0)  || ((length != ~(u64)0) &&
2671              LOFF_OVERFLOW(offset, length)));
2672 }
2673
2674 /*
2675  *  LOCK operation
2676  */
2677 int
2678 nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock *lock)
2679 {
2680         struct nfs4_stateowner *lock_sop = NULL, *open_sop = NULL;
2681         struct nfs4_stateid *lock_stp;
2682         struct file *filp;
2683         struct file_lock file_lock;
2684         struct file_lock *conflock;
2685         int status = 0;
2686         unsigned int strhashval;
2687
2688         dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2689                 (long long) lock->lk_offset,
2690                 (long long) lock->lk_length);
2691
2692         if (nfs4_in_grace() && !lock->lk_reclaim)
2693                 return nfserr_grace;
2694         if (!nfs4_in_grace() && lock->lk_reclaim)
2695                 return nfserr_no_grace;
2696
2697         if (check_lock_length(lock->lk_offset, lock->lk_length))
2698                  return nfserr_inval;
2699
2700         nfs4_lock_state();
2701
2702         if (lock->lk_is_new) {
2703         /*
2704          * Client indicates that this is a new lockowner.
2705          * Use open owner and open stateid to create lock owner and lock
2706          * stateid.
2707          */
2708                 struct nfs4_stateid *open_stp = NULL;
2709                 struct nfs4_file *fp;
2710
2711                 status = nfserr_stale_clientid;
2712                 if (STALE_CLIENTID(&lock->lk_new_clientid)) {
2713                         printk("NFSD: nfsd4_lock: clientid is stale!\n");
2714                         goto out;
2715                 }
2716
2717                 /* is the new lock seqid presented by the client zero? */
2718                 status = nfserr_bad_seqid;
2719                 if (lock->v.new.lock_seqid != 0)
2720                         goto out;
2721
2722                 /* validate and update open stateid and open seqid */
2723                 status = nfs4_preprocess_seqid_op(current_fh,
2724                                         lock->lk_new_open_seqid,
2725                                         &lock->lk_new_open_stateid,
2726                                         CHECK_FH | OPEN_STATE,
2727                                         &open_sop, &open_stp,
2728                                         &lock->v.new.clientid);
2729                 if (status) {
2730                         if (lock->lk_reclaim)
2731                                 status = nfserr_reclaim_bad;
2732                         goto out;
2733                 }
2734                 /* create lockowner and lock stateid */
2735                 fp = open_stp->st_file;
2736                 strhashval = lock_ownerstr_hashval(fp->fi_inode,
2737                                 open_sop->so_client->cl_clientid.cl_id,
2738                                 &lock->v.new.owner);
2739                 /*
2740                  * If we already have this lock owner, the client is in
2741                  * error (or our bookeeping is wrong!)
2742                  * for asking for a 'new lock'.
2743                  */
2744                 status = nfserr_bad_stateid;
2745                 lock_sop = find_lockstateowner(&lock->v.new.owner,
2746                                                 &lock->v.new.clientid);
2747                 if (lock_sop)
2748                         goto out;
2749                 status = nfserr_resource;
2750                 if (!(lock->lk_stateowner = alloc_init_lock_stateowner(strhashval, open_sop->so_client, open_stp, lock)))
2751                         goto out;
2752                 if ((lock_stp = alloc_init_lock_stateid(lock->lk_stateowner,
2753                                                 fp, open_stp)) == NULL) {
2754                         release_stateowner(lock->lk_stateowner);
2755                         lock->lk_stateowner = NULL;
2756                         goto out;
2757                 }
2758                 /* bump the open seqid used to create the lock */
2759                 open_sop->so_seqid++;
2760         } else {
2761                 /* lock (lock owner + lock stateid) already exists */
2762                 status = nfs4_preprocess_seqid_op(current_fh,
2763                                        lock->lk_old_lock_seqid,
2764                                        &lock->lk_old_lock_stateid,
2765                                        CHECK_FH | LOCK_STATE,
2766                                        &lock->lk_stateowner, &lock_stp, NULL);
2767                 if (status)
2768                         goto out;
2769         }
2770         /* lock->lk_stateowner and lock_stp have been created or found */
2771         filp = lock_stp->st_vfs_file;
2772
2773         if ((status = fh_verify(rqstp, current_fh, S_IFREG, MAY_LOCK))) {
2774                 printk("NFSD: nfsd4_lock: permission denied!\n");
2775                 goto out;
2776         }
2777
2778         locks_init_lock(&file_lock);
2779         switch (lock->lk_type) {
2780                 case NFS4_READ_LT:
2781                 case NFS4_READW_LT:
2782                         file_lock.fl_type = F_RDLCK;
2783                 break;
2784                 case NFS4_WRITE_LT:
2785                 case NFS4_WRITEW_LT:
2786                         file_lock.fl_type = F_WRLCK;
2787                 break;
2788                 default:
2789                         status = nfserr_inval;
2790                 goto out;
2791         }
2792         file_lock.fl_owner = (fl_owner_t) lock->lk_stateowner;
2793         file_lock.fl_pid = current->tgid;
2794         file_lock.fl_file = filp;
2795         file_lock.fl_flags = FL_POSIX;
2796
2797         file_lock.fl_start = lock->lk_offset;
2798         if ((lock->lk_length == ~(u64)0) ||
2799                         LOFF_OVERFLOW(lock->lk_offset, lock->lk_length))
2800                 file_lock.fl_end = ~(u64)0;
2801         else
2802                 file_lock.fl_end = lock->lk_offset + lock->lk_length - 1;
2803         nfs4_transform_lock_offset(&file_lock);
2804
2805         /*
2806         * Try to lock the file in the VFS.
2807         * Note: locks.c uses the BKL to protect the inode's lock list.
2808         */
2809
2810         status = posix_lock_file(filp, &file_lock);
2811         if (file_lock.fl_ops && file_lock.fl_ops->fl_release_private)
2812                 file_lock.fl_ops->fl_release_private(&file_lock);
2813         dprintk("NFSD: nfsd4_lock: posix_lock_file status %d\n",status);
2814         switch (-status) {
2815         case 0: /* success! */
2816                 update_stateid(&lock_stp->st_stateid);
2817                 memcpy(&lock->lk_resp_stateid, &lock_stp->st_stateid,
2818                                 sizeof(stateid_t));
2819                 goto out;
2820         case (EAGAIN):
2821                 goto conflicting_lock;
2822         case (EDEADLK):
2823                 status = nfserr_deadlock;
2824         default:
2825                 dprintk("NFSD: nfsd4_lock: posix_lock_file() failed! status %d\n",status);
2826                 goto out_destroy_new_stateid;
2827         }
2828
2829 conflicting_lock:
2830         dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2831         status = nfserr_denied;
2832         /* XXX There is a race here. Future patch needed to provide
2833          * an atomic posix_lock_and_test_file
2834          */
2835         if (!(conflock = posix_test_lock(filp, &file_lock))) {
2836                 status = nfserr_serverfault;
2837                 goto out;
2838         }
2839         nfs4_set_lock_denied(conflock, &lock->lk_denied);
2840
2841 out_destroy_new_stateid:
2842         if (lock->lk_is_new) {
2843                 dprintk("NFSD: nfsd4_lock: destroy new stateid!\n");
2844         /*
2845         * An error encountered after instantiation of the new
2846         * stateid has forced us to destroy it.
2847         */
2848                 if (!seqid_mutating_err(status))
2849                         open_sop->so_seqid--;
2850
2851                 release_state_owner(lock_stp, LOCK_STATE);
2852         }
2853 out:
2854         if (lock->lk_stateowner)
2855                 nfs4_get_stateowner(lock->lk_stateowner);
2856         nfs4_unlock_state();
2857         return status;
2858 }
2859
2860 /*
2861  * LOCKT operation
2862  */
2863 int
2864 nfsd4_lockt(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lockt *lockt)
2865 {
2866         struct inode *inode;
2867         struct file file;
2868         struct file_lock file_lock;
2869         struct file_lock *conflicting_lock;
2870         int status;
2871
2872         if (nfs4_in_grace())
2873                 return nfserr_grace;
2874
2875         if (check_lock_length(lockt->lt_offset, lockt->lt_length))
2876                  return nfserr_inval;
2877
2878         lockt->lt_stateowner = NULL;
2879         nfs4_lock_state();
2880
2881         status = nfserr_stale_clientid;
2882         if (STALE_CLIENTID(&lockt->lt_clientid)) {
2883                 printk("NFSD: nfsd4_lockt: clientid is stale!\n");
2884                 goto out;
2885         }
2886
2887         if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0))) {
2888                 printk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2889                 if (status == nfserr_symlink)
2890                         status = nfserr_inval;
2891                 goto out;
2892         }
2893
2894         inode = current_fh->fh_dentry->d_inode;
2895         locks_init_lock(&file_lock);
2896         switch (lockt->lt_type) {
2897                 case NFS4_READ_LT:
2898                 case NFS4_READW_LT:
2899                         file_lock.fl_type = F_RDLCK;
2900                 break;
2901                 case NFS4_WRITE_LT:
2902                 case NFS4_WRITEW_LT:
2903                         file_lock.fl_type = F_WRLCK;
2904                 break;
2905                 default:
2906                         printk("NFSD: nfs4_lockt: bad lock type!\n");
2907                         status = nfserr_inval;
2908                 goto out;
2909         }
2910
2911         lockt->lt_stateowner = find_lockstateowner_str(inode,
2912                         &lockt->lt_clientid, &lockt->lt_owner);
2913         if (lockt->lt_stateowner)
2914                 file_lock.fl_owner = (fl_owner_t)lockt->lt_stateowner;
2915         file_lock.fl_pid = current->tgid;
2916         file_lock.fl_flags = FL_POSIX;
2917
2918         file_lock.fl_start = lockt->lt_offset;
2919         if ((lockt->lt_length == ~(u64)0) || LOFF_OVERFLOW(lockt->lt_offset, lockt->lt_length))
2920                 file_lock.fl_end = ~(u64)0;
2921         else
2922                 file_lock.fl_end = lockt->lt_offset + lockt->lt_length - 1;
2923
2924         nfs4_transform_lock_offset(&file_lock);
2925
2926         /* posix_test_lock uses the struct file _only_ to resolve the inode.
2927          * since LOCKT doesn't require an OPEN, and therefore a struct
2928          * file may not exist, pass posix_test_lock a struct file with
2929          * only the dentry:inode set.
2930          */
2931         memset(&file, 0, sizeof (struct file));
2932         file.f_dentry = current_fh->fh_dentry;
2933
2934         status = nfs_ok;
2935         conflicting_lock = posix_test_lock(&file, &file_lock);
2936         if (conflicting_lock) {
2937                 status = nfserr_denied;
2938                 nfs4_set_lock_denied(conflicting_lock, &lockt->lt_denied);
2939         }
2940 out:
2941         nfs4_unlock_state();
2942         return status;
2943 }
2944
2945 int
2946 nfsd4_locku(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_locku *locku)
2947 {
2948         struct nfs4_stateid *stp;
2949         struct file *filp = NULL;
2950         struct file_lock file_lock;
2951         int status;
2952
2953         dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2954                 (long long) locku->lu_offset,
2955                 (long long) locku->lu_length);
2956
2957         if (check_lock_length(locku->lu_offset, locku->lu_length))
2958                  return nfserr_inval;
2959
2960         nfs4_lock_state();
2961
2962         if ((status = nfs4_preprocess_seqid_op(current_fh,
2963                                         locku->lu_seqid,
2964                                         &locku->lu_stateid,
2965                                         CHECK_FH | LOCK_STATE,
2966                                         &locku->lu_stateowner, &stp, NULL)))
2967                 goto out;
2968
2969         filp = stp->st_vfs_file;
2970         BUG_ON(!filp);
2971         locks_init_lock(&file_lock);
2972         file_lock.fl_type = F_UNLCK;
2973         file_lock.fl_owner = (fl_owner_t) locku->lu_stateowner;
2974         file_lock.fl_pid = current->tgid;
2975         file_lock.fl_file = filp;
2976         file_lock.fl_flags = FL_POSIX;
2977         file_lock.fl_start = locku->lu_offset;
2978
2979         if ((locku->lu_length == ~(u64)0) || LOFF_OVERFLOW(locku->lu_offset, locku->lu_length))
2980                 file_lock.fl_end = ~(u64)0;
2981         else
2982                 file_lock.fl_end = locku->lu_offset + locku->lu_length - 1;
2983         nfs4_transform_lock_offset(&file_lock);
2984
2985         /*
2986         *  Try to unlock the file in the VFS.
2987         */
2988         status = posix_lock_file(filp, &file_lock);
2989         if (file_lock.fl_ops && file_lock.fl_ops->fl_release_private)
2990                 file_lock.fl_ops->fl_release_private(&file_lock);
2991         if (status) {
2992                 printk("NFSD: nfs4_locku: posix_lock_file failed!\n");
2993                 goto out_nfserr;
2994         }
2995         /*
2996         * OK, unlock succeeded; the only thing left to do is update the stateid.
2997         */
2998         update_stateid(&stp->st_stateid);
2999         memcpy(&locku->lu_stateid, &stp->st_stateid, sizeof(stateid_t));
3000
3001 out:
3002         if (locku->lu_stateowner)
3003                 nfs4_get_stateowner(locku->lu_stateowner);
3004         nfs4_unlock_state();
3005         return status;
3006
3007 out_nfserr:
3008         status = nfserrno(status);
3009         goto out;
3010 }
3011
3012 /*
3013  * returns
3014  *      1: locks held by lockowner
3015  *      0: no locks held by lockowner
3016  */
3017 static int
3018 check_for_locks(struct file *filp, struct nfs4_stateowner *lowner)
3019 {
3020         struct file_lock **flpp;
3021         struct inode *inode = filp->f_dentry->d_inode;
3022         int status = 0;
3023
3024         lock_kernel();
3025         for (flpp = &inode->i_flock; *flpp != NULL; flpp = &(*flpp)->fl_next) {
3026                 if ((*flpp)->fl_owner == (fl_owner_t)lowner)
3027                         status = 1;
3028                         goto out;
3029         }
3030 out:
3031         unlock_kernel();
3032         return status;
3033 }
3034
3035 int
3036 nfsd4_release_lockowner(struct svc_rqst *rqstp, struct nfsd4_release_lockowner *rlockowner)
3037 {
3038         clientid_t *clid = &rlockowner->rl_clientid;
3039         struct nfs4_stateowner *local = NULL;
3040         struct xdr_netobj *owner = &rlockowner->rl_owner;
3041         int status;
3042
3043         dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3044                 clid->cl_boot, clid->cl_id);
3045
3046         /* XXX check for lease expiration */
3047
3048         status = nfserr_stale_clientid;
3049         if (STALE_CLIENTID(clid)) {
3050                 printk("NFSD: nfsd4_release_lockowner: clientid is stale!\n");
3051                 return status;
3052         }
3053
3054         nfs4_lock_state();
3055
3056         status = nfs_ok;
3057         local = find_lockstateowner(owner, clid);
3058         if (local) {
3059                 struct nfs4_stateid *stp;
3060
3061                 /* check for any locks held by any stateid
3062                  * associated with the (lock) stateowner */
3063                 status = nfserr_locks_held;
3064                 list_for_each_entry(stp, &local->so_perfilestate,
3065                                 st_perfilestate) {
3066                         if (check_for_locks(stp->st_vfs_file, local))
3067                                 goto out;
3068                 }
3069                 /* no locks held by (lock) stateowner */
3070                 status = nfs_ok;
3071                 release_stateowner(local);
3072         }
3073 out:
3074         nfs4_unlock_state();
3075         return status;
3076 }
3077
3078 static inline struct nfs4_client_reclaim *
3079 alloc_reclaim(int namelen)
3080 {
3081         struct nfs4_client_reclaim *crp = NULL;
3082
3083         crp = kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
3084         if (!crp)
3085                 return NULL;
3086         crp->cr_name.data = kmalloc(namelen, GFP_KERNEL);
3087         if (!crp->cr_name.data) {
3088                 kfree(crp);
3089                 return NULL;
3090         }
3091         return crp;
3092 }
3093
3094 /*
3095  * failure => all reset bets are off, nfserr_no_grace...
3096  */
3097 static int
3098 nfs4_client_to_reclaim(char *name, int namlen)
3099 {
3100         unsigned int strhashval;
3101         struct nfs4_client_reclaim *crp = NULL;
3102
3103         dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", namlen, name);
3104         crp = alloc_reclaim(namlen);
3105         if (!crp)
3106                 return 0;
3107         strhashval = clientstr_hashval(name, namlen);
3108         INIT_LIST_HEAD(&crp->cr_strhash);
3109         list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
3110         memcpy(crp->cr_name.data, name, namlen);
3111         crp->cr_name.len = namlen;
3112         reclaim_str_hashtbl_size++;
3113         return 1;
3114 }
3115
3116 static void
3117 nfs4_release_reclaim(void)
3118 {
3119         struct nfs4_client_reclaim *crp = NULL;
3120         int i;
3121
3122         BUG_ON(!nfs4_reclaim_init);
3123         for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3124                 while (!list_empty(&reclaim_str_hashtbl[i])) {
3125                         crp = list_entry(reclaim_str_hashtbl[i].next,
3126                                         struct nfs4_client_reclaim, cr_strhash);
3127                         list_del(&crp->cr_strhash);
3128                         kfree(crp->cr_name.data);
3129                         kfree(crp);
3130                         reclaim_str_hashtbl_size--;
3131                 }
3132         }
3133         BUG_ON(reclaim_str_hashtbl_size);
3134 }
3135
3136 /*
3137  * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3138 struct nfs4_client_reclaim *
3139 nfs4_find_reclaim_client(clientid_t *clid)
3140 {
3141         unsigned int strhashval;
3142         struct nfs4_client *clp;
3143         struct nfs4_client_reclaim *crp = NULL;
3144
3145
3146         /* find clientid in conf_id_hashtbl */
3147         clp = find_confirmed_client(clid);
3148         if (clp == NULL)
3149                 return NULL;
3150
3151         dprintk("NFSD: nfs4_find_reclaim_client for %.*s\n",
3152                             clp->cl_name.len, clp->cl_name.data);
3153
3154         /* find clp->cl_name in reclaim_str_hashtbl */
3155         strhashval = clientstr_hashval(clp->cl_name.data, clp->cl_name.len);
3156         list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
3157                 if (cmp_name(&crp->cr_name, &clp->cl_name)) {
3158                         return crp;
3159                 }
3160         }
3161         return NULL;
3162 }
3163
3164 /*
3165 * Called from OPEN. Look for clientid in reclaim list.
3166 */
3167 int
3168 nfs4_check_open_reclaim(clientid_t *clid)
3169 {
3170         struct nfs4_client_reclaim *crp;
3171
3172         if ((crp = nfs4_find_reclaim_client(clid)) == NULL)
3173                 return nfserr_reclaim_bad;
3174         return nfs_ok;
3175 }
3176
3177
3178 /*
3179  * Start and stop routines
3180  */
3181
3182 static void
3183 __nfs4_state_init(void)
3184 {
3185         int i;
3186         time_t grace_time;
3187
3188         if (!nfs4_reclaim_init) {
3189                 for (i = 0; i < CLIENT_HASH_SIZE; i++)
3190                         INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
3191                 reclaim_str_hashtbl_size = 0;
3192                 nfs4_reclaim_init = 1;
3193         }
3194         for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3195                 INIT_LIST_HEAD(&conf_id_hashtbl[i]);
3196                 INIT_LIST_HEAD(&conf_str_hashtbl[i]);
3197                 INIT_LIST_HEAD(&unconf_str_hashtbl[i]);
3198                 INIT_LIST_HEAD(&unconf_id_hashtbl[i]);
3199         }
3200         for (i = 0; i < FILE_HASH_SIZE; i++) {
3201                 INIT_LIST_HEAD(&file_hashtbl[i]);
3202         }
3203         for (i = 0; i < OWNER_HASH_SIZE; i++) {
3204                 INIT_LIST_HEAD(&ownerstr_hashtbl[i]);
3205                 INIT_LIST_HEAD(&ownerid_hashtbl[i]);
3206         }
3207         for (i = 0; i < STATEID_HASH_SIZE; i++) {
3208                 INIT_LIST_HEAD(&stateid_hashtbl[i]);
3209                 INIT_LIST_HEAD(&lockstateid_hashtbl[i]);
3210         }
3211         for (i = 0; i < LOCK_HASH_SIZE; i++) {
3212                 INIT_LIST_HEAD(&lock_ownerid_hashtbl[i]);
3213                 INIT_LIST_HEAD(&lock_ownerstr_hashtbl[i]);
3214         }
3215         memset(&zerostateid, 0, sizeof(stateid_t));
3216         memset(&onestateid, ~0, sizeof(stateid_t));
3217
3218         INIT_LIST_HEAD(&close_lru);
3219         INIT_LIST_HEAD(&client_lru);
3220         INIT_LIST_HEAD(&del_recall_lru);
3221         spin_lock_init(&recall_lock);
3222         boot_time = get_seconds();
3223         grace_time = max(old_lease_time, lease_time);
3224         if (reclaim_str_hashtbl_size == 0)
3225                 grace_time = 0;
3226         if (grace_time)
3227                 printk("NFSD: starting %ld-second grace period\n", grace_time);
3228         grace_end = boot_time + grace_time;
3229         INIT_WORK(&laundromat_work,laundromat_main, NULL);
3230         schedule_delayed_work(&laundromat_work, NFSD_LEASE_TIME*HZ);
3231 }
3232
3233 int
3234 nfs4_state_init(void)
3235 {
3236         int status;
3237
3238         if (nfs4_init)
3239                 return 0;
3240         status = nfsd4_init_slabs();
3241         if (status)
3242                 return status;
3243         __nfs4_state_init();
3244         nfs4_init = 1;
3245         return 0;
3246 }
3247
3248 int
3249 nfs4_in_grace(void)
3250 {
3251         return get_seconds() < grace_end;
3252 }
3253
3254 void
3255 set_no_grace(void)
3256 {
3257         printk("NFSD: ERROR in reboot recovery.  State reclaims will fail.\n");
3258         grace_end = get_seconds();
3259 }
3260
3261 time_t
3262 nfs4_lease_time(void)
3263 {
3264         return lease_time;
3265 }
3266
3267 static void
3268 __nfs4_state_shutdown(void)
3269 {
3270         int i;
3271         struct nfs4_client *clp = NULL;
3272         struct nfs4_delegation *dp = NULL;
3273         struct nfs4_stateowner *sop = NULL;
3274         struct list_head *pos, *next, reaplist;
3275
3276         list_for_each_safe(pos, next, &close_lru) {
3277                 sop = list_entry(pos, struct nfs4_stateowner, so_close_lru);
3278                 list_del(&sop->so_close_lru);
3279                 nfs4_put_stateowner(sop);
3280         }
3281
3282         for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3283                 while (!list_empty(&conf_id_hashtbl[i])) {
3284                         clp = list_entry(conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
3285                         expire_client(clp);
3286                 }
3287                 while (!list_empty(&unconf_str_hashtbl[i])) {
3288                         clp = list_entry(unconf_str_hashtbl[i].next, struct nfs4_client, cl_strhash);
3289                         expire_client(clp);
3290                 }
3291         }
3292         INIT_LIST_HEAD(&reaplist);
3293         spin_lock(&recall_lock);
3294         list_for_each_safe(pos, next, &del_recall_lru) {
3295                 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3296                 list_move(&dp->dl_recall_lru, &reaplist);
3297         }
3298         spin_unlock(&recall_lock);
3299         list_for_each_safe(pos, next, &reaplist) {
3300                 dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
3301                 list_del_init(&dp->dl_recall_lru);
3302                 unhash_delegation(dp);
3303         }
3304
3305         release_all_files();
3306         cancel_delayed_work(&laundromat_work);
3307         flush_scheduled_work();
3308         nfs4_init = 0;
3309         dprintk("NFSD: list_add_perfile %d list_del_perfile %d\n",
3310                         list_add_perfile, list_del_perfile);
3311         dprintk("NFSD: add_perclient %d del_perclient %d\n",
3312                         add_perclient, del_perclient);
3313         dprintk("NFSD: vfsopen %d vfsclose %d\n",
3314                         vfsopen, vfsclose);
3315         dprintk("NFSD: alloc_delegation %d free_delegation %d\n",
3316                         alloc_delegation, free_delegation);
3317
3318 }
3319
3320 void
3321 nfs4_state_shutdown(void)
3322 {
3323         nfs4_lock_state();
3324         nfs4_release_reclaim();
3325         __nfs4_state_shutdown();
3326         nfsd4_free_slabs();
3327         nfs4_unlock_state();
3328 }
3329
3330 /*
3331  * Called when leasetime is changed.
3332  *
3333  * if nfsd is not started, simply set the global lease.
3334  *
3335  * if nfsd(s) are running, lease change requires nfsv4 state to be reset.
3336  * e.g: boot_time is reset, existing nfs4_client structs are
3337  * used to fill reclaim_str_hashtbl, then all state (except for the
3338  * reclaim_str_hashtbl) is re-initialized.
3339  *
3340  * if the old lease time is greater than the new lease time, the grace
3341  * period needs to be set to the old lease time to allow clients to reclaim
3342  * their state. XXX - we may want to set the grace period == lease time
3343  * after an initial grace period == old lease time
3344  *
3345  * if an error occurs in this process, the new lease is set, but the server
3346  * will not honor OPEN or LOCK reclaims, and will return nfserr_no_grace
3347  * which means OPEN/LOCK/READ/WRITE will fail during grace period.
3348  *
3349  * clients will attempt to reset all state with SETCLIENTID/CONFIRM, and
3350  * OPEN and LOCK reclaims.
3351  */
3352 void
3353 nfs4_reset_lease(time_t leasetime)
3354 {
3355         struct nfs4_client *clp;
3356         int i;
3357
3358         printk("NFSD: New leasetime %ld\n",leasetime);
3359         if (!nfs4_init)
3360                 return;
3361         nfs4_lock_state();
3362         old_lease_time = lease_time;
3363         lease_time = leasetime;
3364
3365         nfs4_release_reclaim();
3366
3367         /* populate reclaim_str_hashtbl with current confirmed nfs4_clientid */
3368         for (i = 0; i < CLIENT_HASH_SIZE; i++) {
3369                 list_for_each_entry(clp, &conf_id_hashtbl[i], cl_idhash) {
3370                         if (!nfs4_client_to_reclaim(clp->cl_name.data,
3371                                                 clp->cl_name.len)) {
3372                                 nfs4_release_reclaim();
3373                                 goto init_state;
3374                         }
3375                 }
3376         }
3377 init_state:
3378         __nfs4_state_shutdown();
3379         __nfs4_state_init();
3380         nfs4_unlock_state();
3381 }
3382