1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 
  22 /*
  23  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
  24  */
  25 
  26 /*
  27  * Copyright 2018 Nexenta Systems, Inc.
  28  * Copyright 2019 Nexenta by DDN, Inc.
  29  */
  30 
  31 #include <sys/systm.h>
  32 #include <sys/kmem.h>
  33 #include <sys/cmn_err.h>
  34 #include <sys/atomic.h>
  35 #include <sys/clconf.h>
  36 #include <sys/cladm.h>
  37 #include <sys/flock.h>
  38 #include <nfs/export.h>
  39 #include <nfs/nfs.h>
  40 #include <nfs/nfs4.h>
  41 #include <nfs/nfssys.h>
  42 #include <nfs/lm.h>
  43 #include <sys/pathname.h>
  44 #include <sys/sdt.h>
  45 #include <sys/nvpair.h>
  46 
  47 extern u_longlong_t nfs4_srv_caller_id;
  48 
  49 extern uint_t nfs4_srv_vkey;
  50 
  51 stateid4 special0 = {
  52         0,
  53         { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }
  54 };
  55 
  56 stateid4 special1 = {
  57         0xffffffff,
  58         {
  59                 (char)0xff, (char)0xff, (char)0xff, (char)0xff,
  60                 (char)0xff, (char)0xff, (char)0xff, (char)0xff,
  61                 (char)0xff, (char)0xff, (char)0xff, (char)0xff
  62         }
  63 };
  64 
  65 
  66 #define ISSPECIAL(id)  (stateid4_cmp(id, &special0) || \
  67                         stateid4_cmp(id, &special1))
  68 
  69 /* For embedding the cluster nodeid into our clientid */
  70 #define CLUSTER_NODEID_SHIFT    24
  71 #define CLUSTER_MAX_NODEID      255
  72 
  73 #ifdef DEBUG
  74 int rfs4_debug;
  75 #endif
  76 
  77 static uint32_t rfs4_database_debug = 0x00;
  78 
  79 /* CSTYLED */
  80 static void rfs4_ss_clid_write(nfs4_srv_t *nsrv4, rfs4_client_t *cp, char *leaf);
  81 static void rfs4_ss_clid_write_one(rfs4_client_t *cp, char *dir, char *leaf);
  82 static void rfs4_dss_clear_oldstate(rfs4_servinst_t *sip);
  83 static void rfs4_ss_chkclid_sip(rfs4_client_t *cp, rfs4_servinst_t *sip);
  84 
  85 /*
  86  * Couple of simple init/destroy functions for a general waiter
  87  */
  88 void
  89 rfs4_sw_init(rfs4_state_wait_t *swp)
  90 {
  91         mutex_init(swp->sw_cv_lock, NULL, MUTEX_DEFAULT, NULL);
  92         cv_init(swp->sw_cv, NULL, CV_DEFAULT, NULL);
  93         swp->sw_active = FALSE;
  94         swp->sw_wait_count = 0;
  95 }
  96 
  97 void
  98 rfs4_sw_destroy(rfs4_state_wait_t *swp)
  99 {
 100         mutex_destroy(swp->sw_cv_lock);
 101         cv_destroy(swp->sw_cv);
 102 }
 103 
 104 void
 105 rfs4_sw_enter(rfs4_state_wait_t *swp)
 106 {
 107         mutex_enter(swp->sw_cv_lock);
 108         while (swp->sw_active) {
 109                 swp->sw_wait_count++;
 110                 cv_wait(swp->sw_cv, swp->sw_cv_lock);
 111                 swp->sw_wait_count--;
 112         }
 113         ASSERT(swp->sw_active == FALSE);
 114         swp->sw_active = TRUE;
 115         mutex_exit(swp->sw_cv_lock);
 116 }
 117 
 118 void
 119 rfs4_sw_exit(rfs4_state_wait_t *swp)
 120 {
 121         mutex_enter(swp->sw_cv_lock);
 122         ASSERT(swp->sw_active == TRUE);
 123         swp->sw_active = FALSE;
 124         if (swp->sw_wait_count != 0)
 125                 cv_broadcast(swp->sw_cv);
 126         mutex_exit(swp->sw_cv_lock);
 127 }
 128 
 129 static void
 130 deep_lock_copy(LOCK4res *dres, LOCK4res *sres)
 131 {
 132         lock_owner4 *slo = &sres->LOCK4res_u.denied.owner;
 133         lock_owner4 *dlo = &dres->LOCK4res_u.denied.owner;
 134 
 135         if (sres->status == NFS4ERR_DENIED) {
 136                 dlo->owner_val = kmem_alloc(slo->owner_len, KM_SLEEP);
 137                 bcopy(slo->owner_val, dlo->owner_val, slo->owner_len);
 138         }
 139 }
 140 
 141 /*
 142  * CPR callback id -- not related to v4 callbacks
 143  */
 144 static callb_id_t cpr_id = 0;
 145 
 146 static void
 147 deep_lock_free(LOCK4res *res)
 148 {
 149         lock_owner4 *lo = &res->LOCK4res_u.denied.owner;
 150 
 151         if (res->status == NFS4ERR_DENIED)
 152                 kmem_free(lo->owner_val, lo->owner_len);
 153 }
 154 
 155 static void
 156 deep_open_copy(OPEN4res *dres, OPEN4res *sres)
 157 {
 158         nfsace4 *sacep, *dacep;
 159 
 160         if (sres->status != NFS4_OK) {
 161                 return;
 162         }
 163 
 164         dres->attrset = sres->attrset;
 165 
 166         switch (sres->delegation.delegation_type) {
 167         case OPEN_DELEGATE_NONE:
 168                 return;
 169         case OPEN_DELEGATE_READ:
 170                 sacep = &sres->delegation.open_delegation4_u.read.permissions;
 171                 dacep = &dres->delegation.open_delegation4_u.read.permissions;
 172                 break;
 173         case OPEN_DELEGATE_WRITE:
 174                 sacep = &sres->delegation.open_delegation4_u.write.permissions;
 175                 dacep = &dres->delegation.open_delegation4_u.write.permissions;
 176                 break;
 177         }
 178         dacep->who.utf8string_val =
 179             kmem_alloc(sacep->who.utf8string_len, KM_SLEEP);
 180         bcopy(sacep->who.utf8string_val, dacep->who.utf8string_val,
 181             sacep->who.utf8string_len);
 182 }
 183 
 184 static void
 185 deep_open_free(OPEN4res *res)
 186 {
 187         nfsace4 *acep;
 188         if (res->status != NFS4_OK)
 189                 return;
 190 
 191         switch (res->delegation.delegation_type) {
 192         case OPEN_DELEGATE_NONE:
 193                 return;
 194         case OPEN_DELEGATE_READ:
 195                 acep = &res->delegation.open_delegation4_u.read.permissions;
 196                 break;
 197         case OPEN_DELEGATE_WRITE:
 198                 acep = &res->delegation.open_delegation4_u.write.permissions;
 199                 break;
 200         }
 201 
 202         if (acep->who.utf8string_val) {
 203                 kmem_free(acep->who.utf8string_val, acep->who.utf8string_len);
 204                 acep->who.utf8string_val = NULL;
 205         }
 206 }
 207 
 208 void
 209 rfs4_free_reply(nfs_resop4 *rp)
 210 {
 211         switch (rp->resop) {
 212         case OP_LOCK:
 213                 deep_lock_free(&rp->nfs_resop4_u.oplock);
 214                 break;
 215         case OP_OPEN:
 216                 deep_open_free(&rp->nfs_resop4_u.opopen);
 217         default:
 218                 break;
 219         }
 220 }
 221 
 222 void
 223 rfs4_copy_reply(nfs_resop4 *dst, nfs_resop4 *src)
 224 {
 225         *dst = *src;
 226 
 227         /* Handle responses that need deep copy */
 228         switch (src->resop) {
 229         case OP_LOCK:
 230                 deep_lock_copy(&dst->nfs_resop4_u.oplock,
 231                     &src->nfs_resop4_u.oplock);
 232                 break;
 233         case OP_OPEN:
 234                 deep_open_copy(&dst->nfs_resop4_u.opopen,
 235                     &src->nfs_resop4_u.opopen);
 236                 break;
 237         default:
 238                 break;
 239         };
 240 }
 241 
 242 /*
 243  * This is the implementation of the underlying state engine. The
 244  * public interface to this engine is described by
 245  * nfs4_state.h. Callers to the engine should hold no state engine
 246  * locks when they call in to it. If the protocol needs to lock data
 247  * structures it should do so after acquiring all references to them
 248  * first and then follow the following lock order:
 249  *
 250  *      client > openowner > state > lo_state > lockowner > file.
 251  *
 252  * Internally we only allow a thread to hold one hash bucket lock at a
 253  * time and the lock is higher in the lock order (must be acquired
 254  * first) than the data structure that is on that hash list.
 255  *
 256  * If a new reference was acquired by the caller, that reference needs
 257  * to be released after releasing all acquired locks with the
 258  * corresponding rfs4_*_rele routine.
 259  */
 260 
 261 /*
 262  * This code is some what prototypical for now. Its purpose currently is to
 263  * implement the interfaces sufficiently to finish the higher protocol
 264  * elements. This will be replaced by a dynamically resizeable tables
 265  * backed by kmem_cache allocator. However synchronization is handled
 266  * correctly (I hope) and will not change by much.  The mutexes for
 267  * the hash buckets that can be used to create new instances of data
 268  * structures  might be good candidates to evolve into reader writer
 269  * locks. If it has to do a creation, it would be holding the
 270  * mutex across a kmem_alloc with KM_SLEEP specified.
 271  */
 272 
 273 #ifdef DEBUG
 274 #define TABSIZE 17
 275 #else
 276 #define TABSIZE 2047
 277 #endif
 278 
 279 #define ADDRHASH(key) ((unsigned long)(key) >> 3)
 280 
 281 #define MAXTABSZ 1024*1024
 282 
 283 /* The values below are rfs4_lease_time units */
 284 
 285 #ifdef DEBUG
 286 #define CLIENT_CACHE_TIME 1
 287 #define OPENOWNER_CACHE_TIME 1
 288 #define STATE_CACHE_TIME 1
 289 #define LO_STATE_CACHE_TIME 1
 290 #define LOCKOWNER_CACHE_TIME 1
 291 #define FILE_CACHE_TIME 3
 292 #define DELEG_STATE_CACHE_TIME 1
 293 #else
 294 #define CLIENT_CACHE_TIME 10
 295 #define OPENOWNER_CACHE_TIME 5
 296 #define STATE_CACHE_TIME 1
 297 #define LO_STATE_CACHE_TIME 1
 298 #define LOCKOWNER_CACHE_TIME 3
 299 #define FILE_CACHE_TIME 40
 300 #define DELEG_STATE_CACHE_TIME 1
 301 #endif
 302 
 303 /*
 304  * NFSv4 server state databases
 305  *
 306  * Initilized when the module is loaded and used by NFSv4 state tables.
 307  * These kmem_cache databases are global, the tables that make use of these
 308  * are per zone.
 309  */
 310 kmem_cache_t *rfs4_client_mem_cache;
 311 kmem_cache_t *rfs4_clntIP_mem_cache;
 312 kmem_cache_t *rfs4_openown_mem_cache;
 313 kmem_cache_t *rfs4_openstID_mem_cache;
 314 kmem_cache_t *rfs4_lockstID_mem_cache;
 315 kmem_cache_t *rfs4_lockown_mem_cache;
 316 kmem_cache_t *rfs4_file_mem_cache;
 317 kmem_cache_t *rfs4_delegstID_mem_cache;
 318 
 319 /*
 320  * NFSv4 state table functions
 321  */
 322 static bool_t rfs4_client_create(rfs4_entry_t, void *);
 323 static void rfs4_dss_remove_cpleaf(rfs4_client_t *);
 324 static void rfs4_dss_remove_leaf(rfs4_servinst_t *, char *, char *);
 325 static void rfs4_client_destroy(rfs4_entry_t);
 326 static bool_t rfs4_client_expiry(rfs4_entry_t);
 327 static uint32_t clientid_hash(void *);
 328 static bool_t clientid_compare(rfs4_entry_t, void *);
 329 static void *clientid_mkkey(rfs4_entry_t);
 330 static uint32_t nfsclnt_hash(void *);
 331 static bool_t nfsclnt_compare(rfs4_entry_t, void *);
 332 static void *nfsclnt_mkkey(rfs4_entry_t);
 333 static bool_t rfs4_clntip_expiry(rfs4_entry_t);
 334 static void rfs4_clntip_destroy(rfs4_entry_t);
 335 static bool_t rfs4_clntip_create(rfs4_entry_t, void *);
 336 static uint32_t clntip_hash(void *);
 337 static bool_t clntip_compare(rfs4_entry_t, void *);
 338 static void *clntip_mkkey(rfs4_entry_t);
 339 static bool_t rfs4_openowner_create(rfs4_entry_t, void *);
 340 static void rfs4_openowner_destroy(rfs4_entry_t);
 341 static bool_t rfs4_openowner_expiry(rfs4_entry_t);
 342 static uint32_t openowner_hash(void *);
 343 static bool_t openowner_compare(rfs4_entry_t, void *);
 344 static void *openowner_mkkey(rfs4_entry_t);
 345 static bool_t rfs4_state_create(rfs4_entry_t, void *);
 346 static void rfs4_state_destroy(rfs4_entry_t);
 347 static bool_t rfs4_state_expiry(rfs4_entry_t);
 348 static uint32_t state_hash(void *);
 349 static bool_t state_compare(rfs4_entry_t, void *);
 350 static void *state_mkkey(rfs4_entry_t);
 351 static uint32_t state_owner_file_hash(void *);
 352 static bool_t state_owner_file_compare(rfs4_entry_t, void *);
 353 static void *state_owner_file_mkkey(rfs4_entry_t);
 354 static uint32_t state_file_hash(void *);
 355 static bool_t state_file_compare(rfs4_entry_t, void *);
 356 static void *state_file_mkkey(rfs4_entry_t);
 357 static bool_t rfs4_lo_state_create(rfs4_entry_t, void *);
 358 static void rfs4_lo_state_destroy(rfs4_entry_t);
 359 static bool_t rfs4_lo_state_expiry(rfs4_entry_t);
 360 static uint32_t lo_state_hash(void *);
 361 static bool_t lo_state_compare(rfs4_entry_t, void *);
 362 static void *lo_state_mkkey(rfs4_entry_t);
 363 static uint32_t lo_state_lo_hash(void *);
 364 static bool_t lo_state_lo_compare(rfs4_entry_t, void *);
 365 static void *lo_state_lo_mkkey(rfs4_entry_t);
 366 static bool_t rfs4_lockowner_create(rfs4_entry_t, void *);
 367 static void rfs4_lockowner_destroy(rfs4_entry_t);
 368 static bool_t rfs4_lockowner_expiry(rfs4_entry_t);
 369 static uint32_t lockowner_hash(void *);
 370 static bool_t lockowner_compare(rfs4_entry_t, void *);
 371 static void *lockowner_mkkey(rfs4_entry_t);
 372 static uint32_t pid_hash(void *);
 373 static bool_t pid_compare(rfs4_entry_t, void *);
 374 static void *pid_mkkey(rfs4_entry_t);
 375 static bool_t rfs4_file_create(rfs4_entry_t, void *);
 376 static void rfs4_file_destroy(rfs4_entry_t);
 377 static uint32_t file_hash(void *);
 378 static bool_t file_compare(rfs4_entry_t, void *);
 379 static void *file_mkkey(rfs4_entry_t);
 380 static bool_t rfs4_deleg_state_create(rfs4_entry_t, void *);
 381 static void rfs4_deleg_state_destroy(rfs4_entry_t);
 382 static bool_t rfs4_deleg_state_expiry(rfs4_entry_t);
 383 static uint32_t deleg_hash(void *);
 384 static bool_t deleg_compare(rfs4_entry_t, void *);
 385 static void *deleg_mkkey(rfs4_entry_t);
 386 static uint32_t deleg_state_hash(void *);
 387 static bool_t deleg_state_compare(rfs4_entry_t, void *);
 388 static void *deleg_state_mkkey(rfs4_entry_t);
 389 
 390 static void rfs4_state_rele_nounlock(rfs4_state_t *);
 391 
 392 static int rfs4_ss_enabled = 0;
 393 
 394 extern void (*rfs4_client_clrst)(struct nfs4clrst_args *);
 395 
 396 void
 397 rfs4_ss_pnfree(rfs4_ss_pn_t *ss_pn)
 398 {
 399         kmem_free(ss_pn, sizeof (rfs4_ss_pn_t));
 400 }
 401 
 402 static rfs4_ss_pn_t *
 403 rfs4_ss_pnalloc(char *dir, char *leaf)
 404 {
 405         rfs4_ss_pn_t *ss_pn;
 406         int     dir_len, leaf_len;
 407 
 408         /*
 409          * validate we have a resonable path
 410          * (account for the '/' and trailing null)
 411          */
 412         if ((dir_len = strlen(dir)) > MAXPATHLEN ||
 413             (leaf_len = strlen(leaf)) > MAXNAMELEN ||
 414             (dir_len + leaf_len + 2) > MAXPATHLEN) {
 415                 return (NULL);
 416         }
 417 
 418         ss_pn = kmem_alloc(sizeof (rfs4_ss_pn_t), KM_SLEEP);
 419 
 420         (void) snprintf(ss_pn->pn, MAXPATHLEN, "%s/%s", dir, leaf);
 421         /* Handy pointer to just the leaf name */
 422         ss_pn->leaf = ss_pn->pn + dir_len + 1;
 423         return (ss_pn);
 424 }
 425 
 426 
 427 /*
 428  * Move the "leaf" filename from "sdir" directory
 429  * to the "ddir" directory. Return the pathname of
 430  * the destination unless the rename fails in which
 431  * case we need to return the source pathname.
 432  */
 433 static rfs4_ss_pn_t *
 434 rfs4_ss_movestate(char *sdir, char *ddir, char *leaf)
 435 {
 436         rfs4_ss_pn_t *src, *dst;
 437 
 438         if ((src = rfs4_ss_pnalloc(sdir, leaf)) == NULL)
 439                 return (NULL);
 440 
 441         if ((dst = rfs4_ss_pnalloc(ddir, leaf)) == NULL) {
 442                 rfs4_ss_pnfree(src);
 443                 return (NULL);
 444         }
 445 
 446         /*
 447          * If the rename fails we shall return the src
 448          * pathname and free the dst. Otherwise we need
 449          * to free the src and return the dst pathanme.
 450          */
 451         if (vn_rename(src->pn, dst->pn, UIO_SYSSPACE)) {
 452                 rfs4_ss_pnfree(dst);
 453                 return (src);
 454         }
 455         rfs4_ss_pnfree(src);
 456         return (dst);
 457 }
 458 
 459 
 460 static rfs4_oldstate_t *
 461 rfs4_ss_getstate(vnode_t *dvp, rfs4_ss_pn_t *ss_pn)
 462 {
 463         struct uio uio;
 464         struct iovec iov[3];
 465 
 466         rfs4_oldstate_t *cl_ss = NULL;
 467         vnode_t *vp;
 468         vattr_t va;
 469         uint_t id_len;
 470         int err, kill_file, file_vers;
 471 
 472         if (ss_pn == NULL)
 473                 return (NULL);
 474 
 475         /*
 476          * open the state file.
 477          */
 478         if (vn_open(ss_pn->pn, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0) != 0) {
 479                 return (NULL);
 480         }
 481 
 482         if (vp->v_type != VREG) {
 483                 (void) VOP_CLOSE(vp, FREAD, 1, (offset_t)0, CRED(), NULL);
 484                 VN_RELE(vp);
 485                 return (NULL);
 486         }
 487 
 488         err = VOP_ACCESS(vp, VREAD, 0, CRED(), NULL);
 489         if (err) {
 490                 /*
 491                  * We don't have read access? better get the heck out.
 492                  */
 493                 (void) VOP_CLOSE(vp, FREAD, 1, (offset_t)0, CRED(), NULL);
 494                 VN_RELE(vp);
 495                 return (NULL);
 496         }
 497 
 498         (void) VOP_RWLOCK(vp, V_WRITELOCK_FALSE, NULL);
 499         /*
 500          * get the file size to do some basic validation
 501          */
 502         va.va_mask = AT_SIZE;
 503         err = VOP_GETATTR(vp, &va, 0, CRED(), NULL);
 504 
 505         kill_file = (va.va_size == 0 || va.va_size <
 506             (NFS4_VERIFIER_SIZE + sizeof (uint_t)+1));
 507 
 508         if (err || kill_file) {
 509                 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, NULL);
 510                 (void) VOP_CLOSE(vp, FREAD, 1, (offset_t)0, CRED(), NULL);
 511                 VN_RELE(vp);
 512                 if (kill_file) {
 513                         (void) VOP_REMOVE(dvp, ss_pn->leaf, CRED(), NULL, 0);
 514                 }
 515                 return (NULL);
 516         }
 517 
 518         cl_ss = kmem_alloc(sizeof (rfs4_oldstate_t), KM_SLEEP);
 519 
 520         /*
 521          * build iovecs to read in the file_version, verifier and id_len
 522          */
 523         iov[0].iov_base = (caddr_t)&file_vers;
 524         iov[0].iov_len = sizeof (int);
 525         iov[1].iov_base = (caddr_t)&cl_ss->cl_id4.verifier;
 526         iov[1].iov_len = NFS4_VERIFIER_SIZE;
 527         iov[2].iov_base = (caddr_t)&id_len;
 528         iov[2].iov_len = sizeof (uint_t);
 529 
 530         uio.uio_iov = iov;
 531         uio.uio_iovcnt = 3;
 532         uio.uio_segflg = UIO_SYSSPACE;
 533         uio.uio_loffset = 0;
 534         uio.uio_resid = sizeof (int) + NFS4_VERIFIER_SIZE + sizeof (uint_t);
 535 
 536         if (err = VOP_READ(vp, &uio, FREAD, CRED(), NULL)) {
 537                 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, NULL);
 538                 (void) VOP_CLOSE(vp, FREAD, 1, (offset_t)0, CRED(), NULL);
 539                 VN_RELE(vp);
 540                 kmem_free(cl_ss, sizeof (rfs4_oldstate_t));
 541                 return (NULL);
 542         }
 543 
 544         /*
 545          * if the file_version doesn't match or if the
 546          * id_len is zero or the combination of the verifier,
 547          * id_len and id_val is bigger than the file we have
 548          * a problem. If so ditch the file.
 549          */
 550         kill_file = (file_vers != NFS4_SS_VERSION || id_len == 0 ||
 551             (id_len + NFS4_VERIFIER_SIZE + sizeof (uint_t)) > va.va_size);
 552 
 553         if (err || kill_file) {
 554                 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, NULL);
 555                 (void) VOP_CLOSE(vp, FREAD, 1, (offset_t)0, CRED(), NULL);
 556                 VN_RELE(vp);
 557                 kmem_free(cl_ss, sizeof (rfs4_oldstate_t));
 558                 if (kill_file) {
 559                         (void) VOP_REMOVE(dvp, ss_pn->leaf, CRED(), NULL, 0);
 560                 }
 561                 return (NULL);
 562         }
 563 
 564         /*
 565          * now get the client id value
 566          */
 567         cl_ss->cl_id4.id_val = kmem_alloc(id_len, KM_SLEEP);
 568         iov[0].iov_base = cl_ss->cl_id4.id_val;
 569         iov[0].iov_len = id_len;
 570 
 571         uio.uio_iov = iov;
 572         uio.uio_iovcnt = 1;
 573         uio.uio_segflg = UIO_SYSSPACE;
 574         uio.uio_resid = cl_ss->cl_id4.id_len = id_len;
 575 
 576         if (err = VOP_READ(vp, &uio, FREAD, CRED(), NULL)) {
 577                 VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, NULL);
 578                 (void) VOP_CLOSE(vp, FREAD, 1, (offset_t)0, CRED(), NULL);
 579                 VN_RELE(vp);
 580                 kmem_free(cl_ss->cl_id4.id_val, id_len);
 581                 kmem_free(cl_ss, sizeof (rfs4_oldstate_t));
 582                 return (NULL);
 583         }
 584 
 585         VOP_RWUNLOCK(vp, V_WRITELOCK_FALSE, NULL);
 586         (void) VOP_CLOSE(vp, FREAD, 1, (offset_t)0, CRED(), NULL);
 587         VN_RELE(vp);
 588         return (cl_ss);
 589 }
 590 
 591 #ifdef  nextdp
 592 #undef nextdp
 593 #endif
 594 #define nextdp(dp)      ((struct dirent64 *)((char *)(dp) + (dp)->d_reclen))
 595 
 596 /*
 597  * Add entries from statedir to supplied oldstate list.
 598  * Optionally, move all entries from statedir -> destdir.
 599  */
 600 void
 601 rfs4_ss_oldstate(rfs4_oldstate_t *oldstate, char *statedir, char *destdir)
 602 {
 603         rfs4_ss_pn_t *ss_pn;
 604         rfs4_oldstate_t *cl_ss = NULL;
 605         char    *dirt = NULL;
 606         int     err, dir_eof = 0, size = 0;
 607         vnode_t *dvp;
 608         struct iovec iov;
 609         struct uio uio;
 610         struct dirent64 *dep;
 611         offset_t dirchunk_offset = 0;
 612 
 613         /*
 614          * open the state directory
 615          */
 616         if (vn_open(statedir, UIO_SYSSPACE, FREAD, 0, &dvp, 0, 0))
 617                 return;
 618 
 619         if (dvp->v_type != VDIR || VOP_ACCESS(dvp, VREAD, 0, CRED(), NULL))
 620                 goto out;
 621 
 622         dirt = kmem_alloc(RFS4_SS_DIRSIZE, KM_SLEEP);
 623 
 624         /*
 625          * Get and process the directory entries
 626          */
 627         while (!dir_eof) {
 628                 (void) VOP_RWLOCK(dvp, V_WRITELOCK_FALSE, NULL);
 629                 iov.iov_base = dirt;
 630                 iov.iov_len = RFS4_SS_DIRSIZE;
 631                 uio.uio_iov = &iov;
 632                 uio.uio_iovcnt = 1;
 633                 uio.uio_segflg = UIO_SYSSPACE;
 634                 uio.uio_loffset = dirchunk_offset;
 635                 uio.uio_resid = RFS4_SS_DIRSIZE;
 636 
 637                 err = VOP_READDIR(dvp, &uio, CRED(), &dir_eof, NULL, 0);
 638                 VOP_RWUNLOCK(dvp, V_WRITELOCK_FALSE, NULL);
 639                 if (err)
 640                         goto out;
 641 
 642                 size = RFS4_SS_DIRSIZE - uio.uio_resid;
 643 
 644                 /*
 645                  * Process all the directory entries in this
 646                  * readdir chunk
 647                  */
 648                 for (dep = (struct dirent64 *)dirt; size > 0;
 649                     dep = nextdp(dep)) {
 650 
 651                         size -= dep->d_reclen;
 652                         dirchunk_offset = dep->d_off;
 653 
 654                         /*
 655                          * Skip '.' and '..'
 656                          */
 657                         if (NFS_IS_DOTNAME(dep->d_name))
 658                                 continue;
 659 
 660                         ss_pn = rfs4_ss_pnalloc(statedir, dep->d_name);
 661                         if (ss_pn == NULL)
 662                                 continue;
 663 
 664                         if (cl_ss = rfs4_ss_getstate(dvp, ss_pn)) {
 665                                 if (destdir != NULL) {
 666                                         rfs4_ss_pnfree(ss_pn);
 667                                         cl_ss->ss_pn = rfs4_ss_movestate(
 668                                             statedir, destdir, dep->d_name);
 669                                 } else {
 670                                         cl_ss->ss_pn = ss_pn;
 671                                 }
 672                                 insque(cl_ss, oldstate);
 673                         } else {
 674                                 rfs4_ss_pnfree(ss_pn);
 675                         }
 676                 }
 677         }
 678 
 679 out:
 680         (void) VOP_CLOSE(dvp, FREAD, 1, (offset_t)0, CRED(), NULL);
 681         VN_RELE(dvp);
 682         if (dirt)
 683                 kmem_free((caddr_t)dirt, RFS4_SS_DIRSIZE);
 684 }
 685 
 686 static void
 687 rfs4_ss_init(nfs4_srv_t *nsrv4)
 688 {
 689         int npaths = 1;
 690         char *default_dss_path = NFS4_DSS_VAR_DIR;
 691 
 692         /* read the default stable storage state */
 693         rfs4_dss_readstate(nsrv4, npaths, &default_dss_path);
 694 
 695         rfs4_ss_enabled = 1;
 696 }
 697 
 698 static void
 699 rfs4_ss_fini(nfs4_srv_t *nsrv4)
 700 {
 701         rfs4_servinst_t *sip;
 702 
 703         mutex_enter(&nsrv4->servinst_lock);
 704         sip = nsrv4->nfs4_cur_servinst;
 705         while (sip != NULL) {
 706                 rfs4_dss_clear_oldstate(sip);
 707                 sip = sip->next;
 708         }
 709         mutex_exit(&nsrv4->servinst_lock);
 710 }
 711 
 712 /*
 713  * Remove all oldstate files referenced by this servinst.
 714  */
 715 static void
 716 rfs4_dss_clear_oldstate(rfs4_servinst_t *sip)
 717 {
 718         rfs4_oldstate_t *os_head, *osp;
 719 
 720         rw_enter(&sip->oldstate_lock, RW_WRITER);
 721         os_head = sip->oldstate;
 722 
 723         if (os_head == NULL) {
 724                 rw_exit(&sip->oldstate_lock);
 725                 return;
 726         }
 727 
 728         /* skip dummy entry */
 729         osp = os_head->next;
 730         while (osp != os_head) {
 731                 char *leaf = osp->ss_pn->leaf;
 732                 rfs4_oldstate_t *os_next;
 733 
 734                 rfs4_dss_remove_leaf(sip, NFS4_DSS_OLDSTATE_LEAF, leaf);
 735 
 736                 if (osp->cl_id4.id_val)
 737                         kmem_free(osp->cl_id4.id_val, osp->cl_id4.id_len);
 738                 rfs4_ss_pnfree(osp->ss_pn);
 739 
 740                 os_next = osp->next;
 741                 remque(osp);
 742                 kmem_free(osp, sizeof (rfs4_oldstate_t));
 743                 osp = os_next;
 744         }
 745 
 746         rw_exit(&sip->oldstate_lock);
 747 }
 748 
 749 /*
 750  * Form the state and oldstate paths, and read in the stable storage files.
 751  */
 752 void
 753 rfs4_dss_readstate(nfs4_srv_t *nsrv4, int npaths, char **paths)
 754 {
 755         int i;
 756         char *state, *oldstate;
 757 
 758         state = kmem_alloc(MAXPATHLEN, KM_SLEEP);
 759         oldstate = kmem_alloc(MAXPATHLEN, KM_SLEEP);
 760 
 761         for (i = 0; i < npaths; i++) {
 762                 char *path = paths[i];
 763 
 764                 (void) sprintf(state, "%s/%s", path, NFS4_DSS_STATE_LEAF);
 765                 (void) sprintf(oldstate, "%s/%s", path, NFS4_DSS_OLDSTATE_LEAF);
 766 
 767                 /*
 768                  * Populate the current server instance's oldstate list.
 769                  *
 770                  * 1. Read stable storage data from old state directory,
 771                  *    leaving its contents alone.
 772                  *
 773                  * 2. Read stable storage data from state directory,
 774                  *    and move the latter's contents to old state
 775                  *    directory.
 776                  */
 777                 /* CSTYLED */
 778                 rfs4_ss_oldstate(nsrv4->nfs4_cur_servinst->oldstate, oldstate, NULL);
 779                 /* CSTYLED */
 780                 rfs4_ss_oldstate(nsrv4->nfs4_cur_servinst->oldstate, state, oldstate);
 781         }
 782 
 783         kmem_free(state, MAXPATHLEN);
 784         kmem_free(oldstate, MAXPATHLEN);
 785 }
 786 
 787 
 788 /*
 789  * Check if we are still in grace and if the client can be
 790  * granted permission to perform reclaims.
 791  */
 792 void
 793 rfs4_ss_chkclid(nfs4_srv_t *nsrv4, rfs4_client_t *cp)
 794 {
 795         rfs4_servinst_t *sip;
 796 
 797         /*
 798          * It should be sufficient to check the oldstate data for just
 799          * this client's instance. However, since our per-instance
 800          * client grouping is solely temporal, HA-NFSv4 RG failover
 801          * might result in clients of the same RG being partitioned into
 802          * separate instances.
 803          *
 804          * Until the client grouping is improved, we must check the
 805          * oldstate data for all instances with an active grace period.
 806          *
 807          * This also serves as the mechanism to remove stale oldstate data.
 808          * The first time we check an instance after its grace period has
 809          * expired, the oldstate data should be cleared.
 810          *
 811          * Start at the current instance, and walk the list backwards
 812          * to the first.
 813          */
 814         mutex_enter(&nsrv4->servinst_lock);
 815         for (sip = nsrv4->nfs4_cur_servinst; sip != NULL; sip = sip->prev) {
 816                 rfs4_ss_chkclid_sip(cp, sip);
 817 
 818                 /* if the above check found this client, we're done */
 819                 if (cp->rc_can_reclaim)
 820                         break;
 821         }
 822         mutex_exit(&nsrv4->servinst_lock);
 823 }
 824 
 825 static void
 826 rfs4_ss_chkclid_sip(rfs4_client_t *cp, rfs4_servinst_t *sip)
 827 {
 828         rfs4_oldstate_t *osp, *os_head;
 829 
 830         /* short circuit everything if this server instance has no oldstate */
 831         rw_enter(&sip->oldstate_lock, RW_READER);
 832         os_head = sip->oldstate;
 833         rw_exit(&sip->oldstate_lock);
 834         if (os_head == NULL)
 835                 return;
 836 
 837         /*
 838          * If this server instance is no longer in a grace period then
 839          * the client won't be able to reclaim. No further need for this
 840          * instance's oldstate data, so it can be cleared.
 841          */
 842         if (!rfs4_servinst_in_grace(sip))
 843                 return;
 844 
 845         /* this instance is still in grace; search for the clientid */
 846 
 847         rw_enter(&sip->oldstate_lock, RW_READER);
 848 
 849         os_head = sip->oldstate;
 850         /* skip dummy entry */
 851         osp = os_head->next;
 852         while (osp != os_head) {
 853                 if (osp->cl_id4.id_len == cp->rc_nfs_client.id_len) {
 854                         if (bcmp(osp->cl_id4.id_val, cp->rc_nfs_client.id_val,
 855                             osp->cl_id4.id_len) == 0) {
 856                                 cp->rc_can_reclaim = 1;
 857                                 break;
 858                         }
 859                 }
 860                 osp = osp->next;
 861         }
 862 
 863         rw_exit(&sip->oldstate_lock);
 864 }
 865 
 866 /*
 867  * Place client information into stable storage: 1/3.
 868  * First, generate the leaf filename, from the client's IP address and
 869  * the server-generated short-hand clientid.
 870  */
 871 void
 872 rfs4_ss_clid(nfs4_srv_t *nsrv4, rfs4_client_t *cp)
 873 {
 874         const char *kinet_ntop6(uchar_t *, char *, size_t);
 875         char leaf[MAXNAMELEN], buf[INET6_ADDRSTRLEN];
 876         struct sockaddr *ca;
 877         uchar_t *b;
 878 
 879         if (rfs4_ss_enabled == 0) {
 880                 return;
 881         }
 882 
 883         buf[0] = 0;
 884 
 885         ca = (struct sockaddr *)&cp->rc_addr;
 886 
 887         /*
 888          * Convert the caller's IP address to a dotted string
 889          */
 890         if (ca->sa_family == AF_INET) {
 891                 b = (uchar_t *)&((struct sockaddr_in *)ca)->sin_addr;
 892                 (void) sprintf(buf, "%03d.%03d.%03d.%03d", b[0] & 0xFF,
 893                     b[1] & 0xFF, b[2] & 0xFF, b[3] & 0xFF);
 894         } else if (ca->sa_family == AF_INET6) {
 895                 struct sockaddr_in6 *sin6;
 896 
 897                 sin6 = (struct sockaddr_in6 *)ca;
 898                 (void) kinet_ntop6((uchar_t *)&sin6->sin6_addr,
 899                     buf, INET6_ADDRSTRLEN);
 900         }
 901 
 902         (void) snprintf(leaf, MAXNAMELEN, "%s-%llx", buf,
 903             (longlong_t)cp->rc_clientid);
 904         rfs4_ss_clid_write(nsrv4, cp, leaf);
 905 }
 906 
 907 /*
 908  * Place client information into stable storage: 2/3.
 909  * DSS: distributed stable storage: the file may need to be written to
 910  * multiple directories.
 911  */
 912 static void
 913 rfs4_ss_clid_write(nfs4_srv_t *nsrv4, rfs4_client_t *cp, char *leaf)
 914 {
 915         rfs4_servinst_t *sip;
 916 
 917         /*
 918          * It should be sufficient to write the leaf file to (all) DSS paths
 919          * associated with just this client's instance. However, since our
 920          * per-instance client grouping is solely temporal, HA-NFSv4 RG
 921          * failover might result in us losing DSS data.
 922          *
 923          * Until the client grouping is improved, we must write the DSS data
 924          * to all instances' paths. Start at the current instance, and
 925          * walk the list backwards to the first.
 926          */
 927         mutex_enter(&nsrv4->servinst_lock);
 928         for (sip = nsrv4->nfs4_cur_servinst; sip != NULL; sip = sip->prev) {
 929                 int i, npaths = sip->dss_npaths;
 930 
 931                 /* write the leaf file to all DSS paths */
 932                 for (i = 0; i < npaths; i++) {
 933                         rfs4_dss_path_t *dss_path = sip->dss_paths[i];
 934 
 935                         /* HA-NFSv4 path might have been failed-away from us */
 936                         if (dss_path == NULL)
 937                                 continue;
 938 
 939                         rfs4_ss_clid_write_one(cp, dss_path->path, leaf);
 940                 }
 941         }
 942         mutex_exit(&nsrv4->servinst_lock);
 943 }
 944 
 945 /*
 946  * Place client information into stable storage: 3/3.
 947  * Write the stable storage data to the requested file.
 948  */
 949 static void
 950 rfs4_ss_clid_write_one(rfs4_client_t *cp, char *dss_path, char *leaf)
 951 {
 952         int ioflag;
 953         int file_vers = NFS4_SS_VERSION;
 954         size_t dirlen;
 955         struct uio uio;
 956         struct iovec iov[4];
 957         char *dir;
 958         rfs4_ss_pn_t *ss_pn;
 959         vnode_t *vp;
 960         nfs_client_id4 *cl_id4 = &(cp->rc_nfs_client);
 961 
 962         /* allow 2 extra bytes for '/' & NUL */
 963         dirlen = strlen(dss_path) + strlen(NFS4_DSS_STATE_LEAF) + 2;
 964         dir = kmem_alloc(dirlen, KM_SLEEP);
 965         (void) sprintf(dir, "%s/%s", dss_path, NFS4_DSS_STATE_LEAF);
 966 
 967         ss_pn = rfs4_ss_pnalloc(dir, leaf);
 968         /* rfs4_ss_pnalloc takes its own copy */
 969         kmem_free(dir, dirlen);
 970         if (ss_pn == NULL)
 971                 return;
 972 
 973         if (vn_open(ss_pn->pn, UIO_SYSSPACE, FCREAT|FWRITE, 0600, &vp,
 974             CRCREAT, 0)) {
 975                 rfs4_ss_pnfree(ss_pn);
 976                 return;
 977         }
 978 
 979         /*
 980          * We need to record leaf - i.e. the filename - so that we know
 981          * what to remove, in the future. However, the dir part of cp->ss_pn
 982          * should never be referenced directly, since it's potentially only
 983          * one of several paths with this leaf in it.
 984          */
 985         if (cp->rc_ss_pn != NULL) {
 986                 if (strcmp(cp->rc_ss_pn->leaf, leaf) == 0) {
 987                         /* we've already recorded *this* leaf */
 988                         rfs4_ss_pnfree(ss_pn);
 989                 } else {
 990                         /* replace with this leaf */
 991                         rfs4_ss_pnfree(cp->rc_ss_pn);
 992                         cp->rc_ss_pn = ss_pn;
 993                 }
 994         } else {
 995                 cp->rc_ss_pn = ss_pn;
 996         }
 997 
 998         /*
 999          * Build a scatter list that points to the nfs_client_id4
1000          */
1001         iov[0].iov_base = (caddr_t)&file_vers;
1002         iov[0].iov_len = sizeof (int);
1003         iov[1].iov_base = (caddr_t)&(cl_id4->verifier);
1004         iov[1].iov_len = NFS4_VERIFIER_SIZE;
1005         iov[2].iov_base = (caddr_t)&(cl_id4->id_len);
1006         iov[2].iov_len = sizeof (uint_t);
1007         iov[3].iov_base = (caddr_t)cl_id4->id_val;
1008         iov[3].iov_len = cl_id4->id_len;
1009 
1010         uio.uio_iov = iov;
1011         uio.uio_iovcnt = 4;
1012         uio.uio_loffset = 0;
1013         uio.uio_segflg = UIO_SYSSPACE;
1014         uio.uio_llimit = (rlim64_t)MAXOFFSET_T;
1015         uio.uio_resid = cl_id4->id_len + sizeof (int) +
1016             NFS4_VERIFIER_SIZE + sizeof (uint_t);
1017 
1018         ioflag = uio.uio_fmode = (FWRITE|FSYNC);
1019         uio.uio_extflg = UIO_COPY_DEFAULT;
1020 
1021         (void) VOP_RWLOCK(vp, V_WRITELOCK_TRUE, NULL);
1022         /* write the full client id to the file. */
1023         (void) VOP_WRITE(vp, &uio, ioflag, CRED(), NULL);
1024         VOP_RWUNLOCK(vp, V_WRITELOCK_TRUE, NULL);
1025 
1026         (void) VOP_CLOSE(vp, FWRITE, 1, (offset_t)0, CRED(), NULL);
1027         VN_RELE(vp);
1028 }
1029 
1030 /*
1031  * DSS: distributed stable storage.
1032  * Unpack the list of paths passed by nfsd.
1033  * Use nvlist_alloc(9F) to manage the data.
1034  * The caller is responsible for allocating and freeing the buffer.
1035  */
1036 int
1037 rfs4_dss_setpaths(char *buf, size_t buflen)
1038 {
1039         int error;
1040 
1041         /*
1042          * If this is a "warm start", i.e. we previously had DSS paths,
1043          * preserve the old paths.
1044          */
1045         if (rfs4_dss_paths != NULL) {
1046                 /*
1047                  * Before we lose the ptr, destroy the nvlist and pathnames
1048                  * array from the warm start before this one.
1049                  */
1050                 nvlist_free(rfs4_dss_oldpaths);
1051                 rfs4_dss_oldpaths = rfs4_dss_paths;
1052         }
1053 
1054         /* unpack the buffer into a searchable nvlist */
1055         error = nvlist_unpack(buf, buflen, &rfs4_dss_paths, KM_SLEEP);
1056         if (error)
1057                 return (error);
1058 
1059         /*
1060          * Search the nvlist for the pathnames nvpair (which is the only nvpair
1061          * in the list, and record its location.
1062          */
1063         error = nvlist_lookup_string_array(rfs4_dss_paths, NFS4_DSS_NVPAIR_NAME,
1064             &rfs4_dss_newpaths, &rfs4_dss_numnewpaths);
1065         return (error);
1066 }
1067 
1068 /*
1069  * Ultimately the nfssys() call NFS4_CLR_STATE endsup here
1070  * to find and mark the client for forced expire.
1071  */
1072 static void
1073 rfs4_client_scrub(rfs4_entry_t ent, void *arg)
1074 {
1075         rfs4_client_t *cp = (rfs4_client_t *)ent;
1076         struct nfs4clrst_args *clr = arg;
1077         struct sockaddr_in6 *ent_sin6;
1078         struct in6_addr  clr_in6;
1079         struct sockaddr_in  *ent_sin;
1080         struct in_addr   clr_in;
1081 
1082         if (clr->addr_type != cp->rc_addr.ss_family) {
1083                 return;
1084         }
1085 
1086         switch (clr->addr_type) {
1087 
1088         case AF_INET6:
1089                 /* copyin the address from user space */
1090                 if (copyin(clr->ap, &clr_in6, sizeof (clr_in6))) {
1091                         break;
1092                 }
1093 
1094                 ent_sin6 = (struct sockaddr_in6 *)&cp->rc_addr;
1095 
1096                 /*
1097                  * now compare, and if equivalent mark entry
1098                  * for forced expiration
1099                  */
1100                 if (IN6_ARE_ADDR_EQUAL(&ent_sin6->sin6_addr, &clr_in6)) {
1101                         cp->rc_forced_expire = 1;
1102                 }
1103                 break;
1104 
1105         case AF_INET:
1106                 /* copyin the address from user space */
1107                 if (copyin(clr->ap, &clr_in, sizeof (clr_in))) {
1108                         break;
1109                 }
1110 
1111                 ent_sin = (struct sockaddr_in *)&cp->rc_addr;
1112 
1113                 /*
1114                  * now compare, and if equivalent mark entry
1115                  * for forced expiration
1116                  */
1117                 if (ent_sin->sin_addr.s_addr == clr_in.s_addr) {
1118                         cp->rc_forced_expire = 1;
1119                 }
1120                 break;
1121 
1122         default:
1123                 /* force this assert to fail */
1124                 ASSERT(clr->addr_type != clr->addr_type);
1125         }
1126 }
1127 
1128 /*
1129  * This is called from nfssys() in order to clear server state
1130  * for the specified client IP Address.
1131  */
1132 void
1133 rfs4_clear_client_state(struct nfs4clrst_args *clr)
1134 {
1135         nfs4_srv_t *nsrv4;
1136         nsrv4 = nfs4_get_srv();
1137         (void) rfs4_dbe_walk(nsrv4->rfs4_client_tab, rfs4_client_scrub, clr);
1138 }
1139 
1140 /*
1141  * Used to initialize the NFSv4 server's state or database.  All of
1142  * the tables are created and timers are set.
1143  */
1144 void
1145 rfs4_state_g_init()
1146 {
1147         extern boolean_t rfs4_cpr_callb(void *, int);
1148         /*
1149          * Add a CPR callback so that we can update client
1150          * access times to extend the lease after a suspend
1151          * and resume (using the same class as rpcmod/connmgr)
1152          */
1153         cpr_id = callb_add(rfs4_cpr_callb, 0, CB_CL_CPR_RPC, "rfs4");
1154 
1155         /*
1156          * NFSv4 server state databases
1157          *
1158          * Initialized when the module is loaded and used by NFSv4 state
1159          * tables.  These kmem_cache free pools are used globally, the NFSv4
1160          * state tables which make use of these kmem_cache free pools are per
1161          * zone.
1162          *
1163          * initialize the global kmem_cache free pools which will be used by
1164          * the NFSv4 state tables.
1165          */
1166         /* CSTYLED */
1167         rfs4_client_mem_cache = nfs4_init_mem_cache("Client_entry_cache", 2, sizeof (rfs4_client_t), 0);
1168         /* CSTYLED */
1169         rfs4_clntIP_mem_cache = nfs4_init_mem_cache("ClntIP_entry_cache", 1, sizeof (rfs4_clntip_t), 1);
1170         /* CSTYLED */
1171         rfs4_openown_mem_cache = nfs4_init_mem_cache("OpenOwner_entry_cache", 1, sizeof (rfs4_openowner_t), 2);
1172         /* CSTYLED */
1173         rfs4_openstID_mem_cache = nfs4_init_mem_cache("OpenStateID_entry_cache", 3, sizeof (rfs4_state_t), 3);
1174         /* CSTYLED */
1175         rfs4_lockstID_mem_cache = nfs4_init_mem_cache("LockStateID_entry_cache", 3, sizeof (rfs4_lo_state_t), 4);
1176         /* CSTYLED */
1177         rfs4_lockown_mem_cache = nfs4_init_mem_cache("Lockowner_entry_cache", 2, sizeof (rfs4_lockowner_t), 5);
1178         /* CSTYLED */
1179         rfs4_file_mem_cache = nfs4_init_mem_cache("File_entry_cache", 1, sizeof (rfs4_file_t), 6);
1180         /* CSTYLED */
1181         rfs4_delegstID_mem_cache = nfs4_init_mem_cache("DelegStateID_entry_cache", 2, sizeof (rfs4_deleg_state_t), 7);
1182 
1183         rfs4_client_clrst = rfs4_clear_client_state;
1184 }
1185 
1186 
1187 /*
1188  * Used at server shutdown to cleanup all of the NFSv4 server's structures
1189  * and other state.
1190  */
1191 void
1192 rfs4_state_g_fini()
1193 {
1194         int i;
1195         /*
1196          * Cleanup the CPR callback.
1197          */
1198         if (cpr_id)
1199                 (void) callb_delete(cpr_id);
1200 
1201         rfs4_client_clrst = NULL;
1202 
1203         /* free the NFSv4 state databases */
1204         for (i = 0; i < RFS4_DB_MEM_CACHE_NUM; i++) {
1205                 kmem_cache_destroy(rfs4_db_mem_cache_table[i].r_db_mem_cache);
1206                 rfs4_db_mem_cache_table[i].r_db_mem_cache = NULL;
1207         }
1208 
1209         rfs4_client_mem_cache = NULL;
1210         rfs4_clntIP_mem_cache = NULL;
1211         rfs4_openown_mem_cache = NULL;
1212         rfs4_openstID_mem_cache = NULL;
1213         rfs4_lockstID_mem_cache = NULL;
1214         rfs4_lockown_mem_cache = NULL;
1215         rfs4_file_mem_cache = NULL;
1216         rfs4_delegstID_mem_cache = NULL;
1217 
1218         /* DSS: distributed stable storage */
1219         nvlist_free(rfs4_dss_oldpaths);
1220         nvlist_free(rfs4_dss_paths);
1221         rfs4_dss_paths = rfs4_dss_oldpaths = NULL;
1222 }
1223 
1224 /*
1225  * Used to initialize the per zone NFSv4 server's state
1226  */
1227 void
1228 rfs4_state_zone_init(nfs4_srv_t *nsrv4)
1229 {
1230         time_t start_time;
1231         int start_grace;
1232         char *dss_path = NFS4_DSS_VAR_DIR;
1233 
1234         /* DSS: distributed stable storage: initialise served paths list */
1235         nsrv4->dss_pathlist = NULL;
1236 
1237         /*
1238          * Set the boot time.  If the server
1239          * has been restarted quickly and has had the opportunity to
1240          * service clients, then the start_time needs to be bumped
1241          * regardless.  A small window but it exists...
1242          */
1243         start_time = gethrestime_sec();
1244         if (nsrv4->rfs4_start_time < start_time)
1245                 nsrv4->rfs4_start_time = start_time;
1246         else
1247                 nsrv4->rfs4_start_time++;
1248 
1249         /*
1250          * Create the first server instance, or a new one if the server has
1251          * been restarted; see above comments on rfs4_start_time. Don't
1252          * start its grace period; that will be done later, to maximise the
1253          * clients' recovery window.
1254          */
1255         start_grace = 0;
1256         if (curzone == global_zone && rfs4_dss_numnewpaths > 0) {
1257                 int i;
1258                 char **dss_allpaths = NULL;
1259                 dss_allpaths = kmem_alloc(sizeof (char *) * (rfs4_dss_numnewpaths + 1), KM_SLEEP);
1260                 /*
1261                  * Add the default path into the list of paths for saving
1262                  * state informantion.
1263                  */
1264                 dss_allpaths[0] = dss_path;
1265                 for ( i = 0; i < rfs4_dss_numnewpaths; i++) {
1266                         dss_allpaths[i + 1] = rfs4_dss_newpaths[i];
1267                 }
1268                 rfs4_servinst_create(nsrv4, start_grace, (rfs4_dss_numnewpaths + 1), dss_allpaths);
1269                 kmem_free(dss_allpaths, (sizeof (char *) * (rfs4_dss_numnewpaths + 1)));
1270         } else {
1271                 rfs4_servinst_create(nsrv4, start_grace, 1, &dss_path);
1272         }
1273 
1274         /* reset the "first NFSv4 request" status */
1275         nsrv4->seen_first_compound = 0;
1276 
1277         mutex_enter(&nsrv4->state_lock);
1278 
1279         /*
1280          * If the server state database has already been initialized,
1281          * skip it
1282          */
1283         if (nsrv4->nfs4_server_state != NULL) {
1284                 mutex_exit(&nsrv4->state_lock);
1285                 return;
1286         }
1287 
1288         rw_init(&nsrv4->rfs4_findclient_lock, NULL, RW_DEFAULT, NULL);
1289 
1290         /* set the various cache timers for table creation */
1291         if (nsrv4->rfs4_client_cache_time == 0)
1292                 nsrv4->rfs4_client_cache_time = CLIENT_CACHE_TIME;
1293         if (nsrv4->rfs4_openowner_cache_time == 0)
1294                 nsrv4->rfs4_openowner_cache_time = OPENOWNER_CACHE_TIME;
1295         if (nsrv4->rfs4_state_cache_time == 0)
1296                 nsrv4->rfs4_state_cache_time = STATE_CACHE_TIME;
1297         if (nsrv4->rfs4_lo_state_cache_time == 0)
1298                 nsrv4->rfs4_lo_state_cache_time = LO_STATE_CACHE_TIME;
1299         if (nsrv4->rfs4_lockowner_cache_time == 0)
1300                 nsrv4->rfs4_lockowner_cache_time = LOCKOWNER_CACHE_TIME;
1301         if (nsrv4->rfs4_file_cache_time == 0)
1302                 nsrv4->rfs4_file_cache_time = FILE_CACHE_TIME;
1303         if (nsrv4->rfs4_deleg_state_cache_time == 0)
1304                 nsrv4->rfs4_deleg_state_cache_time = DELEG_STATE_CACHE_TIME;
1305 
1306         /* Create the overall database to hold all server state */
1307         nsrv4->nfs4_server_state = rfs4_database_create(rfs4_database_debug);
1308 
1309         /* Now create the individual tables */
1310         nsrv4->rfs4_client_cache_time *= rfs4_lease_time;
1311         nsrv4->rfs4_client_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1312             "Client",
1313             nsrv4->rfs4_client_cache_time,
1314             2,
1315             rfs4_client_create,
1316             rfs4_client_destroy,
1317             rfs4_client_expiry,
1318             sizeof (rfs4_client_t),
1319             TABSIZE,
1320             MAXTABSZ/8, 100);
1321         nsrv4->rfs4_nfsclnt_idx = rfs4_index_create(nsrv4->rfs4_client_tab,
1322             "nfs_client_id4", nfsclnt_hash,
1323             nfsclnt_compare, nfsclnt_mkkey,
1324             TRUE);
1325         nsrv4->rfs4_clientid_idx = rfs4_index_create(nsrv4->rfs4_client_tab,
1326             "client_id", clientid_hash,
1327             clientid_compare, clientid_mkkey,
1328             FALSE);
1329 
1330         nsrv4->rfs4_clntip_cache_time = 86400 * 365; /* about a year */
1331         nsrv4->rfs4_clntip_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1332             "ClntIP",
1333             nsrv4->rfs4_clntip_cache_time,
1334             1,
1335             rfs4_clntip_create,
1336             rfs4_clntip_destroy,
1337             rfs4_clntip_expiry,
1338             sizeof (rfs4_clntip_t),
1339             TABSIZE,
1340             MAXTABSZ, 100);
1341         nsrv4->rfs4_clntip_idx = rfs4_index_create(nsrv4->rfs4_clntip_tab,
1342             "client_ip", clntip_hash,
1343             clntip_compare, clntip_mkkey,
1344             TRUE);
1345 
1346         nsrv4->rfs4_openowner_cache_time *= rfs4_lease_time;
1347         nsrv4->rfs4_openowner_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1348             "OpenOwner",
1349             nsrv4->rfs4_openowner_cache_time,
1350             1,
1351             rfs4_openowner_create,
1352             rfs4_openowner_destroy,
1353             rfs4_openowner_expiry,
1354             sizeof (rfs4_openowner_t),
1355             TABSIZE,
1356             MAXTABSZ, 100);
1357         nsrv4->rfs4_openowner_idx = rfs4_index_create(nsrv4->rfs4_openowner_tab,
1358             "open_owner4", openowner_hash,
1359             openowner_compare,
1360             openowner_mkkey, TRUE);
1361 
1362         nsrv4->rfs4_state_cache_time *= rfs4_lease_time;
1363         nsrv4->rfs4_state_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1364             "OpenStateID",
1365             nsrv4->rfs4_state_cache_time,
1366             3,
1367             rfs4_state_create,
1368             rfs4_state_destroy,
1369             rfs4_state_expiry,
1370             sizeof (rfs4_state_t),
1371             TABSIZE,
1372             MAXTABSZ, 100);
1373 
1374         /* CSTYLED */
1375         nsrv4->rfs4_state_owner_file_idx = rfs4_index_create(nsrv4->rfs4_state_tab,
1376             "Openowner-File",
1377             state_owner_file_hash,
1378             state_owner_file_compare,
1379             state_owner_file_mkkey, TRUE);
1380 
1381         nsrv4->rfs4_state_idx = rfs4_index_create(nsrv4->rfs4_state_tab,
1382             "State-id", state_hash,
1383             state_compare, state_mkkey, FALSE);
1384 
1385         nsrv4->rfs4_state_file_idx = rfs4_index_create(nsrv4->rfs4_state_tab,
1386             "File", state_file_hash,
1387             state_file_compare, state_file_mkkey,
1388             FALSE);
1389 
1390         nsrv4->rfs4_lo_state_cache_time *= rfs4_lease_time;
1391         nsrv4->rfs4_lo_state_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1392             "LockStateID",
1393             nsrv4->rfs4_lo_state_cache_time,
1394             2,
1395             rfs4_lo_state_create,
1396             rfs4_lo_state_destroy,
1397             rfs4_lo_state_expiry,
1398             sizeof (rfs4_lo_state_t),
1399             TABSIZE,
1400             MAXTABSZ, 100);
1401 
1402         /* CSTYLED */
1403         nsrv4->rfs4_lo_state_owner_idx = rfs4_index_create(nsrv4->rfs4_lo_state_tab,
1404             "lockownerxstate",
1405             lo_state_lo_hash,
1406             lo_state_lo_compare,
1407             lo_state_lo_mkkey, TRUE);
1408 
1409         nsrv4->rfs4_lo_state_idx = rfs4_index_create(nsrv4->rfs4_lo_state_tab,
1410             "State-id",
1411             lo_state_hash, lo_state_compare,
1412             lo_state_mkkey, FALSE);
1413 
1414         nsrv4->rfs4_lockowner_cache_time *= rfs4_lease_time;
1415 
1416         nsrv4->rfs4_lockowner_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1417             "Lockowner",
1418             nsrv4->rfs4_lockowner_cache_time,
1419             2,
1420             rfs4_lockowner_create,
1421             rfs4_lockowner_destroy,
1422             rfs4_lockowner_expiry,
1423             sizeof (rfs4_lockowner_t),
1424             TABSIZE,
1425             MAXTABSZ, 100);
1426 
1427         nsrv4->rfs4_lockowner_idx = rfs4_index_create(nsrv4->rfs4_lockowner_tab,
1428             "lock_owner4", lockowner_hash,
1429             lockowner_compare,
1430             lockowner_mkkey, TRUE);
1431 
1432         /* CSTYLED */
1433         nsrv4->rfs4_lockowner_pid_idx = rfs4_index_create(nsrv4->rfs4_lockowner_tab,
1434             "pid", pid_hash,
1435             pid_compare, pid_mkkey,
1436             FALSE);
1437 
1438         nsrv4->rfs4_file_cache_time *= rfs4_lease_time;
1439         nsrv4->rfs4_file_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1440             "File",
1441             nsrv4->rfs4_file_cache_time,
1442             1,
1443             rfs4_file_create,
1444             rfs4_file_destroy,
1445             NULL,
1446             sizeof (rfs4_file_t),
1447             TABSIZE,
1448             MAXTABSZ, -1);
1449 
1450         nsrv4->rfs4_file_idx = rfs4_index_create(nsrv4->rfs4_file_tab,
1451             "Filehandle", file_hash,
1452             file_compare, file_mkkey, TRUE);
1453 
1454         nsrv4->rfs4_deleg_state_cache_time *= rfs4_lease_time;
1455         /* CSTYLED */
1456         nsrv4->rfs4_deleg_state_tab = rfs4_table_create(nsrv4->nfs4_server_state,
1457             "DelegStateID",
1458             nsrv4->rfs4_deleg_state_cache_time,
1459             2,
1460             rfs4_deleg_state_create,
1461             rfs4_deleg_state_destroy,
1462             rfs4_deleg_state_expiry,
1463             sizeof (rfs4_deleg_state_t),
1464             TABSIZE,
1465             MAXTABSZ, 100);
1466         nsrv4->rfs4_deleg_idx = rfs4_index_create(nsrv4->rfs4_deleg_state_tab,
1467             "DelegByFileClient",
1468             deleg_hash,
1469             deleg_compare,
1470             deleg_mkkey, TRUE);
1471 
1472         /* CSTYLED */
1473         nsrv4->rfs4_deleg_state_idx = rfs4_index_create(nsrv4->rfs4_deleg_state_tab,
1474             "DelegState",
1475             deleg_state_hash,
1476             deleg_state_compare,
1477             deleg_state_mkkey, FALSE);
1478 
1479         mutex_exit(&nsrv4->state_lock);
1480 
1481         /*
1482          * Init the stable storage.
1483          */
1484         rfs4_ss_init(nsrv4);
1485 }
1486 
1487 /*
1488  * Used at server shutdown to cleanup all of NFSv4 server's zone structures
1489  * and state.
1490  */
1491 void
1492 rfs4_state_zone_fini()
1493 {
1494         rfs4_database_t *dbp;
1495         nfs4_srv_t *nsrv4;
1496         nsrv4 = nfs4_get_srv();
1497 
1498         rfs4_set_deleg_policy(nsrv4, SRV_NEVER_DELEGATE);
1499 
1500         /*
1501          * Clean up any dangling stable storage structures BEFORE calling
1502          * rfs4_servinst_destroy_all() so there are no dangling structures
1503          * (i.e. the srvinsts are all cleared of danglers BEFORE they get
1504          * freed).
1505          */
1506         rfs4_ss_fini(nsrv4);
1507 
1508         mutex_enter(&nsrv4->state_lock);
1509 
1510         if (nsrv4->nfs4_server_state == NULL) {
1511                 mutex_exit(&nsrv4->state_lock);
1512                 return;
1513         }
1514 
1515         /* destroy server instances and current instance ptr */
1516         rfs4_servinst_destroy_all(nsrv4);
1517 
1518         /* reset the "first NFSv4 request" status */
1519         nsrv4->seen_first_compound = 0;
1520 
1521         dbp = nsrv4->nfs4_server_state;
1522         nsrv4->nfs4_server_state = NULL;
1523 
1524         rw_destroy(&nsrv4->rfs4_findclient_lock);
1525 
1526         /* First stop all of the reaper threads in the database */
1527         rfs4_database_shutdown(dbp);
1528 
1529         /*
1530          * WARNING: There may be consumers of the rfs4 database still
1531          * active as we destroy these.  IF that's the case, consider putting
1532          * some of their _zone_fini()-like functions into the zsd key as
1533          * ~~SHUTDOWN~~ functions instead of ~~DESTROY~~ functions.  We can
1534          * maintain some ordering guarantees better that way.
1535          */
1536         /* Now destroy/release the database tables */
1537         rfs4_database_destroy(dbp);
1538 
1539         /* Reset the cache timers for next time */
1540         nsrv4->rfs4_client_cache_time = 0;
1541         nsrv4->rfs4_openowner_cache_time = 0;
1542         nsrv4->rfs4_state_cache_time = 0;
1543         nsrv4->rfs4_lo_state_cache_time = 0;
1544         nsrv4->rfs4_lockowner_cache_time = 0;
1545         nsrv4->rfs4_file_cache_time = 0;
1546         nsrv4->rfs4_deleg_state_cache_time = 0;
1547 
1548         mutex_exit(&nsrv4->state_lock);
1549 }
1550 
1551 typedef union {
1552         struct {
1553                 uint32_t start_time;
1554                 uint32_t c_id;
1555         } impl_id;
1556         clientid4 id4;
1557 } cid;
1558 
1559 static int foreign_stateid(stateid_t *id);
1560 static int foreign_clientid(cid *cidp);
1561 static void embed_nodeid(cid *cidp);
1562 
1563 typedef union {
1564         struct {
1565                 uint32_t c_id;
1566                 uint32_t gen_num;
1567         } cv_impl;
1568         verifier4       confirm_verf;
1569 } scid_confirm_verf;
1570 
1571 static uint32_t
1572 clientid_hash(void *key)
1573 {
1574         cid *idp = key;
1575 
1576         return (idp->impl_id.c_id);
1577 }
1578 
1579 static bool_t
1580 clientid_compare(rfs4_entry_t entry, void *key)
1581 {
1582         rfs4_client_t *cp = (rfs4_client_t *)entry;
1583         clientid4 *idp = key;
1584 
1585         return (*idp == cp->rc_clientid);
1586 }
1587 
1588 static void *
1589 clientid_mkkey(rfs4_entry_t entry)
1590 {
1591         rfs4_client_t *cp = (rfs4_client_t *)entry;
1592 
1593         return (&cp->rc_clientid);
1594 }
1595 
1596 static uint32_t
1597 nfsclnt_hash(void *key)
1598 {
1599         nfs_client_id4 *client = key;
1600         int i;
1601         uint32_t hash = 0;
1602 
1603         for (i = 0; i < client->id_len; i++) {
1604                 hash <<= 1;
1605                 hash += (uint_t)client->id_val[i];
1606         }
1607         return (hash);
1608 }
1609 
1610 
1611 static bool_t
1612 nfsclnt_compare(rfs4_entry_t entry, void *key)
1613 {
1614         rfs4_client_t *cp = (rfs4_client_t *)entry;
1615         nfs_client_id4 *nfs_client = key;
1616 
1617         if (cp->rc_nfs_client.id_len != nfs_client->id_len)
1618                 return (FALSE);
1619 
1620         return (bcmp(cp->rc_nfs_client.id_val, nfs_client->id_val,
1621             nfs_client->id_len) == 0);
1622 }
1623 
1624 static void *
1625 nfsclnt_mkkey(rfs4_entry_t entry)
1626 {
1627         rfs4_client_t *cp = (rfs4_client_t *)entry;
1628 
1629         return (&cp->rc_nfs_client);
1630 }
1631 
1632 static bool_t
1633 rfs4_client_expiry(rfs4_entry_t u_entry)
1634 {
1635         rfs4_client_t *cp = (rfs4_client_t *)u_entry;
1636         bool_t cp_expired;
1637 
1638         if (rfs4_dbe_is_invalid(cp->rc_dbe)) {
1639                 cp->rc_ss_remove = 1;
1640                 return (TRUE);
1641         }
1642         /*
1643          * If the sysadmin has used clear_locks for this
1644          * entry then forced_expire will be set and we
1645          * want this entry to be reaped. Or the entry
1646          * has exceeded its lease period.
1647          */
1648         cp_expired = (cp->rc_forced_expire ||
1649             (gethrestime_sec() - cp->rc_last_access
1650             > rfs4_lease_time));
1651 
1652         if (!cp->rc_ss_remove && cp_expired)
1653                 cp->rc_ss_remove = 1;
1654         return (cp_expired);
1655 }
1656 
1657 /*
1658  * Remove the leaf file from all distributed stable storage paths.
1659  */
1660 static void
1661 rfs4_dss_remove_cpleaf(rfs4_client_t *cp)
1662 {
1663         nfs4_srv_t *nsrv4;
1664         rfs4_servinst_t *sip;
1665         char *leaf = cp->rc_ss_pn->leaf;
1666 
1667         /*
1668          * since the state files are written to all DSS
1669          * paths we must remove this leaf file instance
1670          * from all server instances.
1671          */
1672 
1673         nsrv4 = nfs4_get_srv();
1674         mutex_enter(&nsrv4->servinst_lock);
1675         for (sip = nsrv4->nfs4_cur_servinst; sip != NULL; sip = sip->prev) {
1676                 /* remove the leaf file associated with this server instance */
1677                 rfs4_dss_remove_leaf(sip, NFS4_DSS_STATE_LEAF, leaf);
1678         }
1679         mutex_exit(&nsrv4->servinst_lock);
1680 }
1681 
1682 static void
1683 rfs4_dss_remove_leaf(rfs4_servinst_t *sip, char *dir_leaf, char *leaf)
1684 {
1685         int i, npaths = sip->dss_npaths;
1686 
1687         for (i = 0; i < npaths; i++) {
1688                 rfs4_dss_path_t *dss_path = sip->dss_paths[i];
1689                 char *path, *dir;
1690                 size_t pathlen;
1691 
1692                 /* the HA-NFSv4 path might have been failed-over away from us */
1693                 if (dss_path == NULL)
1694                         continue;
1695 
1696                 dir = dss_path->path;
1697 
1698                 /* allow 3 extra bytes for two '/' & a NUL */
1699                 pathlen = strlen(dir) + strlen(dir_leaf) + strlen(leaf) + 3;
1700                 path = kmem_alloc(pathlen, KM_SLEEP);
1701                 (void) sprintf(path, "%s/%s/%s", dir, dir_leaf, leaf);
1702 
1703                 (void) vn_remove(path, UIO_SYSSPACE, RMFILE);
1704 
1705                 kmem_free(path, pathlen);
1706         }
1707 }
1708 
1709 static void
1710 rfs4_client_destroy(rfs4_entry_t u_entry)
1711 {
1712         rfs4_client_t *cp = (rfs4_client_t *)u_entry;
1713 
1714         mutex_destroy(cp->rc_cbinfo.cb_lock);
1715         cv_destroy(cp->rc_cbinfo.cb_cv);
1716         cv_destroy(cp->rc_cbinfo.cb_cv_nullcaller);
1717         list_destroy(&cp->rc_openownerlist);
1718 
1719         /* free callback info */
1720         rfs4_cbinfo_free(&cp->rc_cbinfo);
1721 
1722         if (cp->rc_cp_confirmed)
1723                 rfs4_client_rele(cp->rc_cp_confirmed);
1724 
1725         if (cp->rc_ss_pn) {
1726                 /* check if the stable storage files need to be removed */
1727                 if (cp->rc_ss_remove)
1728                         rfs4_dss_remove_cpleaf(cp);
1729                 rfs4_ss_pnfree(cp->rc_ss_pn);
1730         }
1731 
1732         /* Free the client supplied client id */
1733         kmem_free(cp->rc_nfs_client.id_val, cp->rc_nfs_client.id_len);
1734 
1735         if (cp->rc_sysidt != LM_NOSYSID)
1736                 lm_free_sysidt(cp->rc_sysidt);
1737 }
1738 
1739 static bool_t
1740 rfs4_client_create(rfs4_entry_t u_entry, void *arg)
1741 {
1742         rfs4_client_t *cp = (rfs4_client_t *)u_entry;
1743         nfs_client_id4 *client = (nfs_client_id4 *)arg;
1744         struct sockaddr *ca;
1745         cid *cidp;
1746         scid_confirm_verf *scvp;
1747         nfs4_srv_t *nsrv4;
1748 
1749         nsrv4 = nfs4_get_srv();
1750 
1751         /* Get a clientid to give to the client */
1752         cidp = (cid *)&cp->rc_clientid;
1753         cidp->impl_id.start_time = nsrv4->rfs4_start_time;
1754         cidp->impl_id.c_id = (uint32_t)rfs4_dbe_getid(cp->rc_dbe);
1755 
1756         /* If we are booted as a cluster node, embed our nodeid */
1757         if (cluster_bootflags & CLUSTER_BOOTED)
1758                 embed_nodeid(cidp);
1759 
1760         /* Allocate and copy client's client id value */
1761         cp->rc_nfs_client.id_val = kmem_alloc(client->id_len, KM_SLEEP);
1762         cp->rc_nfs_client.id_len = client->id_len;
1763         bcopy(client->id_val, cp->rc_nfs_client.id_val, client->id_len);
1764         cp->rc_nfs_client.verifier = client->verifier;
1765 
1766         /* Copy client's IP address */
1767         ca = client->cl_addr;
1768         if (ca->sa_family == AF_INET)
1769                 bcopy(ca, &cp->rc_addr, sizeof (struct sockaddr_in));
1770         else if (ca->sa_family == AF_INET6)
1771                 bcopy(ca, &cp->rc_addr, sizeof (struct sockaddr_in6));
1772         cp->rc_nfs_client.cl_addr = (struct sockaddr *)&cp->rc_addr;
1773 
1774         /* Init the value for the SETCLIENTID_CONFIRM verifier */
1775         scvp = (scid_confirm_verf *)&cp->rc_confirm_verf;
1776         scvp->cv_impl.c_id = cidp->impl_id.c_id;
1777         scvp->cv_impl.gen_num = 0;
1778 
1779         /* An F_UNLKSYS has been done for this client */
1780         cp->rc_unlksys_completed = FALSE;
1781 
1782         /* We need the client to ack us */
1783         cp->rc_need_confirm = TRUE;
1784         cp->rc_cp_confirmed = NULL;
1785 
1786         /* TRUE all the time until the callback path actually fails */
1787         cp->rc_cbinfo.cb_notified_of_cb_path_down = TRUE;
1788 
1789         /* Initialize the access time to now */
1790         cp->rc_last_access = gethrestime_sec();
1791 
1792         cp->rc_cr_set = NULL;
1793 
1794         cp->rc_sysidt = LM_NOSYSID;
1795 
1796         list_create(&cp->rc_openownerlist, sizeof (rfs4_openowner_t),
1797             offsetof(rfs4_openowner_t, ro_node));
1798 
1799         /* set up the callback control structure */
1800         cp->rc_cbinfo.cb_state = CB_UNINIT;
1801         mutex_init(cp->rc_cbinfo.cb_lock, NULL, MUTEX_DEFAULT, NULL);
1802         cv_init(cp->rc_cbinfo.cb_cv, NULL, CV_DEFAULT, NULL);
1803         cv_init(cp->rc_cbinfo.cb_cv_nullcaller, NULL, CV_DEFAULT, NULL);
1804 
1805         /*
1806          * Associate the client_t with the current server instance.
1807          * The hold is solely to satisfy the calling requirement of
1808          * rfs4_servinst_assign(). In this case it's not strictly necessary.
1809          */
1810         rfs4_dbe_hold(cp->rc_dbe);
1811         rfs4_servinst_assign(nsrv4, cp, nsrv4->nfs4_cur_servinst);
1812         rfs4_dbe_rele(cp->rc_dbe);
1813 
1814         return (TRUE);
1815 }
1816 
1817 /*
1818  * Caller wants to generate/update the setclientid_confirm verifier
1819  * associated with a client.  This is done during the SETCLIENTID
1820  * processing.
1821  */
1822 void
1823 rfs4_client_scv_next(rfs4_client_t *cp)
1824 {
1825         scid_confirm_verf *scvp;
1826 
1827         /* Init the value for the SETCLIENTID_CONFIRM verifier */
1828         scvp = (scid_confirm_verf *)&cp->rc_confirm_verf;
1829         scvp->cv_impl.gen_num++;
1830 }
1831 
1832 void
1833 rfs4_client_rele(rfs4_client_t *cp)
1834 {
1835         rfs4_dbe_rele(cp->rc_dbe);
1836 }
1837 
1838 rfs4_client_t *
1839 rfs4_findclient(nfs_client_id4 *client, bool_t *create, rfs4_client_t *oldcp)
1840 {
1841         rfs4_client_t *cp;
1842         nfs4_srv_t *nsrv4;
1843         nsrv4 = nfs4_get_srv();
1844 
1845 
1846         if (oldcp) {
1847                 rw_enter(&nsrv4->rfs4_findclient_lock, RW_WRITER);
1848                 rfs4_dbe_hide(oldcp->rc_dbe);
1849         } else {
1850                 rw_enter(&nsrv4->rfs4_findclient_lock, RW_READER);
1851         }
1852 
1853         cp = (rfs4_client_t *)rfs4_dbsearch(nsrv4->rfs4_nfsclnt_idx, client,
1854             create, (void *)client, RFS4_DBS_VALID);
1855 
1856         if (oldcp)
1857                 rfs4_dbe_unhide(oldcp->rc_dbe);
1858 
1859         rw_exit(&nsrv4->rfs4_findclient_lock);
1860 
1861         return (cp);
1862 }
1863 
1864 rfs4_client_t *
1865 rfs4_findclient_by_id(clientid4 clientid, bool_t find_unconfirmed)
1866 {
1867         rfs4_client_t *cp;
1868         bool_t create = FALSE;
1869         cid *cidp = (cid *)&clientid;
1870         nfs4_srv_t *nsrv4 = nfs4_get_srv();
1871 
1872         /* If we're a cluster and the nodeid isn't right, short-circuit */
1873         if (cluster_bootflags & CLUSTER_BOOTED && foreign_clientid(cidp))
1874                 return (NULL);
1875 
1876         rw_enter(&nsrv4->rfs4_findclient_lock, RW_READER);
1877 
1878         cp = (rfs4_client_t *)rfs4_dbsearch(nsrv4->rfs4_clientid_idx, &clientid,
1879             &create, NULL, RFS4_DBS_VALID);
1880 
1881         rw_exit(&nsrv4->rfs4_findclient_lock);
1882 
1883         if (cp && cp->rc_need_confirm && find_unconfirmed == FALSE) {
1884                 rfs4_client_rele(cp);
1885                 return (NULL);
1886         } else {
1887                 return (cp);
1888         }
1889 }
1890 
1891 static uint32_t
1892 clntip_hash(void *key)
1893 {
1894         struct sockaddr *addr = key;
1895         int i, len = 0;
1896         uint32_t hash = 0;
1897         char *ptr;
1898 
1899         if (addr->sa_family == AF_INET) {
1900                 struct sockaddr_in *a = (struct sockaddr_in *)addr;
1901                 len = sizeof (struct in_addr);
1902                 ptr = (char *)&a->sin_addr;
1903         } else if (addr->sa_family == AF_INET6) {
1904                 struct sockaddr_in6 *a = (struct sockaddr_in6 *)addr;
1905                 len = sizeof (struct in6_addr);
1906                 ptr = (char *)&a->sin6_addr;
1907         } else
1908                 return (0);
1909 
1910         for (i = 0; i < len; i++) {
1911                 hash <<= 1;
1912                 hash += (uint_t)ptr[i];
1913         }
1914         return (hash);
1915 }
1916 
1917 static bool_t
1918 clntip_compare(rfs4_entry_t entry, void *key)
1919 {
1920         rfs4_clntip_t *cp = (rfs4_clntip_t *)entry;
1921         struct sockaddr *addr = key;
1922         int len = 0;
1923         char *p1, *p2;
1924 
1925         if (addr->sa_family == AF_INET) {
1926                 struct sockaddr_in *a1 = (struct sockaddr_in *)&cp->ri_addr;
1927                 struct sockaddr_in *a2 = (struct sockaddr_in *)addr;
1928                 len = sizeof (struct in_addr);
1929                 p1 = (char *)&a1->sin_addr;
1930                 p2 = (char *)&a2->sin_addr;
1931         } else if (addr->sa_family == AF_INET6) {
1932                 struct sockaddr_in6 *a1 = (struct sockaddr_in6 *)&cp->ri_addr;
1933                 struct sockaddr_in6 *a2 = (struct sockaddr_in6 *)addr;
1934                 len = sizeof (struct in6_addr);
1935                 p1 = (char *)&a1->sin6_addr;
1936                 p2 = (char *)&a2->sin6_addr;
1937         } else
1938                 return (0);
1939 
1940         return (bcmp(p1, p2, len) == 0);
1941 }
1942 
1943 static void *
1944 clntip_mkkey(rfs4_entry_t entry)
1945 {
1946         rfs4_clntip_t *cp = (rfs4_clntip_t *)entry;
1947 
1948         return (&cp->ri_addr);
1949 }
1950 
1951 static bool_t
1952 rfs4_clntip_expiry(rfs4_entry_t u_entry)
1953 {
1954         rfs4_clntip_t *cp = (rfs4_clntip_t *)u_entry;
1955 
1956         if (rfs4_dbe_is_invalid(cp->ri_dbe))
1957                 return (TRUE);
1958         return (FALSE);
1959 }
1960 
1961 /* ARGSUSED */
1962 static void
1963 rfs4_clntip_destroy(rfs4_entry_t u_entry)
1964 {
1965 }
1966 
1967 static bool_t
1968 rfs4_clntip_create(rfs4_entry_t u_entry, void *arg)
1969 {
1970         rfs4_clntip_t *cp = (rfs4_clntip_t *)u_entry;
1971         struct sockaddr *ca = (struct sockaddr *)arg;
1972 
1973         /* Copy client's IP address */
1974         if (ca->sa_family == AF_INET)
1975                 bcopy(ca, &cp->ri_addr, sizeof (struct sockaddr_in));
1976         else if (ca->sa_family == AF_INET6)
1977                 bcopy(ca, &cp->ri_addr, sizeof (struct sockaddr_in6));
1978         else
1979                 return (FALSE);
1980         cp->ri_no_referrals = 1;
1981 
1982         return (TRUE);
1983 }
1984 
1985 rfs4_clntip_t *
1986 rfs4_find_clntip(struct sockaddr *addr, bool_t *create)
1987 {
1988         rfs4_clntip_t *cp;
1989         nfs4_srv_t *nsrv4;
1990 
1991         nsrv4 = nfs4_get_srv();
1992 
1993         rw_enter(&nsrv4->rfs4_findclient_lock, RW_READER);
1994 
1995         cp = (rfs4_clntip_t *)rfs4_dbsearch(nsrv4->rfs4_clntip_idx, addr,
1996             create, addr, RFS4_DBS_VALID);
1997 
1998         rw_exit(&nsrv4->rfs4_findclient_lock);
1999 
2000         return (cp);
2001 }
2002 
2003 void
2004 rfs4_invalidate_clntip(struct sockaddr *addr)
2005 {
2006         rfs4_clntip_t *cp;
2007         bool_t create = FALSE;
2008         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2009 
2010         rw_enter(&nsrv4->rfs4_findclient_lock, RW_READER);
2011 
2012         cp = (rfs4_clntip_t *)rfs4_dbsearch(nsrv4->rfs4_clntip_idx, addr,
2013             &create, NULL, RFS4_DBS_VALID);
2014         if (cp == NULL) {
2015                 rw_exit(&nsrv4->rfs4_findclient_lock);
2016                 return;
2017         }
2018         rfs4_dbe_invalidate(cp->ri_dbe);
2019         rfs4_dbe_rele(cp->ri_dbe);
2020 
2021         rw_exit(&nsrv4->rfs4_findclient_lock);
2022 }
2023 
2024 bool_t
2025 rfs4_lease_expired(rfs4_client_t *cp)
2026 {
2027         bool_t rc;
2028 
2029         rfs4_dbe_lock(cp->rc_dbe);
2030 
2031         /*
2032          * If the admin has executed clear_locks for this
2033          * client id, force expire will be set, so no need
2034          * to calculate anything because it's "outa here".
2035          */
2036         if (cp->rc_forced_expire) {
2037                 rc = TRUE;
2038         } else {
2039                 rc = (gethrestime_sec() - cp->rc_last_access > rfs4_lease_time);
2040         }
2041 
2042         /*
2043          * If the lease has expired we will also want
2044          * to remove any stable storage state data. So
2045          * mark the client id accordingly.
2046          */
2047         if (!cp->rc_ss_remove)
2048                 cp->rc_ss_remove = (rc == TRUE);
2049 
2050         rfs4_dbe_unlock(cp->rc_dbe);
2051 
2052         return (rc);
2053 }
2054 
2055 void
2056 rfs4_update_lease(rfs4_client_t *cp)
2057 {
2058         rfs4_dbe_lock(cp->rc_dbe);
2059         if (!cp->rc_forced_expire)
2060                 cp->rc_last_access = gethrestime_sec();
2061         rfs4_dbe_unlock(cp->rc_dbe);
2062 }
2063 
2064 
2065 static bool_t
2066 EQOPENOWNER(open_owner4 *a, open_owner4 *b)
2067 {
2068         bool_t rc;
2069 
2070         if (a->clientid != b->clientid)
2071                 return (FALSE);
2072 
2073         if (a->owner_len != b->owner_len)
2074                 return (FALSE);
2075 
2076         rc = (bcmp(a->owner_val, b->owner_val, a->owner_len) == 0);
2077 
2078         return (rc);
2079 }
2080 
2081 static uint_t
2082 openowner_hash(void *key)
2083 {
2084         int i;
2085         open_owner4 *openowner = key;
2086         uint_t hash = 0;
2087 
2088         for (i = 0; i < openowner->owner_len; i++) {
2089                 hash <<= 4;
2090                 hash += (uint_t)openowner->owner_val[i];
2091         }
2092         hash += (uint_t)openowner->clientid;
2093         hash |= (openowner->clientid >> 32);
2094 
2095         return (hash);
2096 }
2097 
2098 static bool_t
2099 openowner_compare(rfs4_entry_t u_entry, void *key)
2100 {
2101         rfs4_openowner_t *oo = (rfs4_openowner_t *)u_entry;
2102         open_owner4 *arg = key;
2103 
2104         return (EQOPENOWNER(&oo->ro_owner, arg));
2105 }
2106 
2107 void *
2108 openowner_mkkey(rfs4_entry_t u_entry)
2109 {
2110         rfs4_openowner_t *oo = (rfs4_openowner_t *)u_entry;
2111 
2112         return (&oo->ro_owner);
2113 }
2114 
2115 /* ARGSUSED */
2116 static bool_t
2117 rfs4_openowner_expiry(rfs4_entry_t u_entry)
2118 {
2119         /* openstateid held us and did all needed delay */
2120         return (TRUE);
2121 }
2122 
2123 static void
2124 rfs4_openowner_destroy(rfs4_entry_t u_entry)
2125 {
2126         rfs4_openowner_t *oo = (rfs4_openowner_t *)u_entry;
2127 
2128         /* Remove open owner from client's lists of open owners */
2129         rfs4_dbe_lock(oo->ro_client->rc_dbe);
2130         list_remove(&oo->ro_client->rc_openownerlist, oo);
2131         rfs4_dbe_unlock(oo->ro_client->rc_dbe);
2132 
2133         /* One less reference to the client */
2134         rfs4_client_rele(oo->ro_client);
2135         oo->ro_client = NULL;
2136 
2137         /* Free the last reply for this lock owner */
2138         rfs4_free_reply(&oo->ro_reply);
2139 
2140         if (oo->ro_reply_fh.nfs_fh4_val) {
2141                 kmem_free(oo->ro_reply_fh.nfs_fh4_val,
2142                     oo->ro_reply_fh.nfs_fh4_len);
2143                 oo->ro_reply_fh.nfs_fh4_val = NULL;
2144                 oo->ro_reply_fh.nfs_fh4_len = 0;
2145         }
2146 
2147         rfs4_sw_destroy(&oo->ro_sw);
2148         list_destroy(&oo->ro_statelist);
2149 
2150         /* Free the lock owner id */
2151         kmem_free(oo->ro_owner.owner_val, oo->ro_owner.owner_len);
2152 }
2153 
2154 void
2155 rfs4_openowner_rele(rfs4_openowner_t *oo)
2156 {
2157         rfs4_dbe_rele(oo->ro_dbe);
2158 }
2159 
2160 static bool_t
2161 rfs4_openowner_create(rfs4_entry_t u_entry, void *arg)
2162 {
2163         rfs4_openowner_t *oo = (rfs4_openowner_t *)u_entry;
2164         rfs4_openowner_t *argp = (rfs4_openowner_t *)arg;
2165         open_owner4 *openowner = &argp->ro_owner;
2166         seqid4 seqid = argp->ro_open_seqid;
2167         rfs4_client_t *cp;
2168         bool_t create = FALSE;
2169         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2170 
2171         rw_enter(&nsrv4->rfs4_findclient_lock, RW_READER);
2172 
2173         cp = (rfs4_client_t *)rfs4_dbsearch(nsrv4->rfs4_clientid_idx,
2174             &openowner->clientid,
2175             &create, NULL, RFS4_DBS_VALID);
2176 
2177         rw_exit(&nsrv4->rfs4_findclient_lock);
2178 
2179         if (cp == NULL)
2180                 return (FALSE);
2181 
2182         oo->ro_reply_fh.nfs_fh4_len = 0;
2183         oo->ro_reply_fh.nfs_fh4_val = NULL;
2184 
2185         oo->ro_owner.clientid = openowner->clientid;
2186         oo->ro_owner.owner_val =
2187             kmem_alloc(openowner->owner_len, KM_SLEEP);
2188 
2189         bcopy(openowner->owner_val,
2190             oo->ro_owner.owner_val, openowner->owner_len);
2191 
2192         oo->ro_owner.owner_len = openowner->owner_len;
2193 
2194         oo->ro_need_confirm = TRUE;
2195 
2196         rfs4_sw_init(&oo->ro_sw);
2197 
2198         oo->ro_open_seqid = seqid;
2199         bzero(&oo->ro_reply, sizeof (nfs_resop4));
2200         oo->ro_client = cp;
2201         oo->ro_cr_set = NULL;
2202 
2203         list_create(&oo->ro_statelist, sizeof (rfs4_state_t),
2204             offsetof(rfs4_state_t, rs_node));
2205 
2206         /* Insert openowner into client's open owner list */
2207         rfs4_dbe_lock(cp->rc_dbe);
2208         list_insert_tail(&cp->rc_openownerlist, oo);
2209         rfs4_dbe_unlock(cp->rc_dbe);
2210 
2211         return (TRUE);
2212 }
2213 
2214 rfs4_openowner_t *
2215 rfs4_findopenowner(open_owner4 *openowner, bool_t *create, seqid4 seqid)
2216 {
2217         rfs4_openowner_t *oo;
2218         rfs4_openowner_t arg;
2219         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2220 
2221         arg.ro_owner = *openowner;
2222         arg.ro_open_seqid = seqid;
2223         /* CSTYLED */
2224         oo = (rfs4_openowner_t *)rfs4_dbsearch(nsrv4->rfs4_openowner_idx, openowner,
2225             create, &arg, RFS4_DBS_VALID);
2226 
2227         return (oo);
2228 }
2229 
2230 void
2231 rfs4_update_open_sequence(rfs4_openowner_t *oo)
2232 {
2233 
2234         rfs4_dbe_lock(oo->ro_dbe);
2235 
2236         oo->ro_open_seqid++;
2237 
2238         rfs4_dbe_unlock(oo->ro_dbe);
2239 }
2240 
2241 void
2242 rfs4_update_open_resp(rfs4_openowner_t *oo, nfs_resop4 *resp, nfs_fh4 *fh)
2243 {
2244 
2245         rfs4_dbe_lock(oo->ro_dbe);
2246 
2247         rfs4_free_reply(&oo->ro_reply);
2248 
2249         rfs4_copy_reply(&oo->ro_reply, resp);
2250 
2251         /* Save the filehandle if provided and free if not used */
2252         if (resp->nfs_resop4_u.opopen.status == NFS4_OK &&
2253             fh && fh->nfs_fh4_len) {
2254                 if (oo->ro_reply_fh.nfs_fh4_val == NULL)
2255                         oo->ro_reply_fh.nfs_fh4_val =
2256                             kmem_alloc(fh->nfs_fh4_len, KM_SLEEP);
2257                 nfs_fh4_copy(fh, &oo->ro_reply_fh);
2258         } else {
2259                 if (oo->ro_reply_fh.nfs_fh4_val) {
2260                         kmem_free(oo->ro_reply_fh.nfs_fh4_val,
2261                             oo->ro_reply_fh.nfs_fh4_len);
2262                         oo->ro_reply_fh.nfs_fh4_val = NULL;
2263                         oo->ro_reply_fh.nfs_fh4_len = 0;
2264                 }
2265         }
2266 
2267         rfs4_dbe_unlock(oo->ro_dbe);
2268 }
2269 
2270 static bool_t
2271 lockowner_compare(rfs4_entry_t u_entry, void *key)
2272 {
2273         rfs4_lockowner_t *lo = (rfs4_lockowner_t *)u_entry;
2274         lock_owner4 *b = (lock_owner4 *)key;
2275 
2276         if (lo->rl_owner.clientid != b->clientid)
2277                 return (FALSE);
2278 
2279         if (lo->rl_owner.owner_len != b->owner_len)
2280                 return (FALSE);
2281 
2282         return (bcmp(lo->rl_owner.owner_val, b->owner_val,
2283             lo->rl_owner.owner_len) == 0);
2284 }
2285 
2286 void *
2287 lockowner_mkkey(rfs4_entry_t u_entry)
2288 {
2289         rfs4_lockowner_t *lo = (rfs4_lockowner_t *)u_entry;
2290 
2291         return (&lo->rl_owner);
2292 }
2293 
2294 static uint32_t
2295 lockowner_hash(void *key)
2296 {
2297         int i;
2298         lock_owner4 *lockowner = key;
2299         uint_t hash = 0;
2300 
2301         for (i = 0; i < lockowner->owner_len; i++) {
2302                 hash <<= 4;
2303                 hash += (uint_t)lockowner->owner_val[i];
2304         }
2305         hash += (uint_t)lockowner->clientid;
2306         hash |= (lockowner->clientid >> 32);
2307 
2308         return (hash);
2309 }
2310 
2311 static uint32_t
2312 pid_hash(void *key)
2313 {
2314         return ((uint32_t)(uintptr_t)key);
2315 }
2316 
2317 static void *
2318 pid_mkkey(rfs4_entry_t u_entry)
2319 {
2320         rfs4_lockowner_t *lo = (rfs4_lockowner_t *)u_entry;
2321 
2322         return ((void *)(uintptr_t)lo->rl_pid);
2323 }
2324 
2325 static bool_t
2326 pid_compare(rfs4_entry_t u_entry, void *key)
2327 {
2328         rfs4_lockowner_t *lo = (rfs4_lockowner_t *)u_entry;
2329 
2330         return (lo->rl_pid == (pid_t)(uintptr_t)key);
2331 }
2332 
2333 static void
2334 rfs4_lockowner_destroy(rfs4_entry_t u_entry)
2335 {
2336         rfs4_lockowner_t *lo = (rfs4_lockowner_t *)u_entry;
2337 
2338         /* Free the lock owner id */
2339         kmem_free(lo->rl_owner.owner_val, lo->rl_owner.owner_len);
2340         rfs4_client_rele(lo->rl_client);
2341 }
2342 
2343 void
2344 rfs4_lockowner_rele(rfs4_lockowner_t *lo)
2345 {
2346         rfs4_dbe_rele(lo->rl_dbe);
2347 }
2348 
2349 /* ARGSUSED */
2350 static bool_t
2351 rfs4_lockowner_expiry(rfs4_entry_t u_entry)
2352 {
2353         /*
2354          * Since expiry is called with no other references on
2355          * this struct, go ahead and have it removed.
2356          */
2357         return (TRUE);
2358 }
2359 
2360 static bool_t
2361 rfs4_lockowner_create(rfs4_entry_t u_entry, void *arg)
2362 {
2363         rfs4_lockowner_t *lo = (rfs4_lockowner_t *)u_entry;
2364         lock_owner4 *lockowner = (lock_owner4 *)arg;
2365         rfs4_client_t *cp;
2366         bool_t create = FALSE;
2367         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2368 
2369         rw_enter(&nsrv4->rfs4_findclient_lock, RW_READER);
2370 
2371         cp = (rfs4_client_t *)rfs4_dbsearch(nsrv4->rfs4_clientid_idx,
2372             &lockowner->clientid,
2373             &create, NULL, RFS4_DBS_VALID);
2374 
2375         rw_exit(&nsrv4->rfs4_findclient_lock);
2376 
2377         if (cp == NULL)
2378                 return (FALSE);
2379 
2380         /* Reference client */
2381         lo->rl_client = cp;
2382         lo->rl_owner.clientid = lockowner->clientid;
2383         lo->rl_owner.owner_val = kmem_alloc(lockowner->owner_len, KM_SLEEP);
2384         bcopy(lockowner->owner_val, lo->rl_owner.owner_val,
2385             lockowner->owner_len);
2386         lo->rl_owner.owner_len = lockowner->owner_len;
2387         lo->rl_pid = rfs4_dbe_getid(lo->rl_dbe);
2388 
2389         return (TRUE);
2390 }
2391 
2392 rfs4_lockowner_t *
2393 rfs4_findlockowner(lock_owner4 *lockowner, bool_t *create)
2394 {
2395         rfs4_lockowner_t *lo;
2396         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2397 
2398         /* CSTYLED */
2399         lo = (rfs4_lockowner_t *)rfs4_dbsearch(nsrv4->rfs4_lockowner_idx, lockowner,
2400             create, lockowner, RFS4_DBS_VALID);
2401 
2402         return (lo);
2403 }
2404 
2405 rfs4_lockowner_t *
2406 rfs4_findlockowner_by_pid(pid_t pid)
2407 {
2408         rfs4_lockowner_t *lo;
2409         bool_t create = FALSE;
2410         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2411 
2412         lo = (rfs4_lockowner_t *)rfs4_dbsearch(nsrv4->rfs4_lockowner_pid_idx,
2413             (void *)(uintptr_t)pid, &create, NULL, RFS4_DBS_VALID);
2414 
2415         return (lo);
2416 }
2417 
2418 
2419 static uint32_t
2420 file_hash(void *key)
2421 {
2422         return (ADDRHASH(key));
2423 }
2424 
2425 static void *
2426 file_mkkey(rfs4_entry_t u_entry)
2427 {
2428         rfs4_file_t *fp = (rfs4_file_t *)u_entry;
2429 
2430         return (fp->rf_vp);
2431 }
2432 
2433 static bool_t
2434 file_compare(rfs4_entry_t u_entry, void *key)
2435 {
2436         rfs4_file_t *fp = (rfs4_file_t *)u_entry;
2437 
2438         return (fp->rf_vp == (vnode_t *)key);
2439 }
2440 
2441 static void
2442 rfs4_file_destroy(rfs4_entry_t u_entry)
2443 {
2444         rfs4_file_t *fp = (rfs4_file_t *)u_entry;
2445 
2446         list_destroy(&fp->rf_delegstatelist);
2447 
2448         if (fp->rf_filehandle.nfs_fh4_val)
2449                 kmem_free(fp->rf_filehandle.nfs_fh4_val,
2450                     fp->rf_filehandle.nfs_fh4_len);
2451         cv_destroy(fp->rf_dinfo.rd_recall_cv);
2452         if (fp->rf_vp) {
2453                 vnode_t *vp = fp->rf_vp;
2454 
2455                 mutex_enter(&vp->v_vsd_lock);
2456                 (void) vsd_set(vp, nfs4_srv_vkey, NULL);
2457                 mutex_exit(&vp->v_vsd_lock);
2458                 VN_RELE(vp);
2459                 fp->rf_vp = NULL;
2460         }
2461         rw_destroy(&fp->rf_file_rwlock);
2462 }
2463 
2464 /*
2465  * Used to unlock the underlying dbe struct only
2466  */
2467 void
2468 rfs4_file_rele(rfs4_file_t *fp)
2469 {
2470         rfs4_dbe_rele(fp->rf_dbe);
2471 }
2472 
2473 typedef struct {
2474     vnode_t *vp;
2475     nfs_fh4 *fh;
2476 } rfs4_fcreate_arg;
2477 
2478 static bool_t
2479 rfs4_file_create(rfs4_entry_t u_entry, void *arg)
2480 {
2481         rfs4_file_t *fp = (rfs4_file_t *)u_entry;
2482         rfs4_fcreate_arg *ap = (rfs4_fcreate_arg *)arg;
2483         vnode_t *vp = ap->vp;
2484         nfs_fh4 *fh = ap->fh;
2485 
2486         VN_HOLD(vp);
2487 
2488         fp->rf_filehandle.nfs_fh4_len = 0;
2489         fp->rf_filehandle.nfs_fh4_val = NULL;
2490         ASSERT(fh && fh->nfs_fh4_len);
2491         if (fh && fh->nfs_fh4_len) {
2492                 fp->rf_filehandle.nfs_fh4_val =
2493                     kmem_alloc(fh->nfs_fh4_len, KM_SLEEP);
2494                 nfs_fh4_copy(fh, &fp->rf_filehandle);
2495         }
2496         fp->rf_vp = vp;
2497 
2498         list_create(&fp->rf_delegstatelist, sizeof (rfs4_deleg_state_t),
2499             offsetof(rfs4_deleg_state_t, rds_node));
2500 
2501         fp->rf_share_deny = fp->rf_share_access = fp->rf_access_read = 0;
2502         fp->rf_access_write = fp->rf_deny_read = fp->rf_deny_write = 0;
2503 
2504         mutex_init(fp->rf_dinfo.rd_recall_lock, NULL, MUTEX_DEFAULT, NULL);
2505         cv_init(fp->rf_dinfo.rd_recall_cv, NULL, CV_DEFAULT, NULL);
2506 
2507         fp->rf_dinfo.rd_dtype = OPEN_DELEGATE_NONE;
2508 
2509         rw_init(&fp->rf_file_rwlock, NULL, RW_DEFAULT, NULL);
2510 
2511         mutex_enter(&vp->v_vsd_lock);
2512         VERIFY(vsd_set(vp, nfs4_srv_vkey, (void *)fp) == 0);
2513         mutex_exit(&vp->v_vsd_lock);
2514 
2515         return (TRUE);
2516 }
2517 
2518 rfs4_file_t *
2519 rfs4_findfile(vnode_t *vp, nfs_fh4 *fh, bool_t *create)
2520 {
2521         rfs4_file_t *fp;
2522         rfs4_fcreate_arg arg;
2523         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2524 
2525         arg.vp = vp;
2526         arg.fh = fh;
2527 
2528         if (*create == TRUE)
2529                 /* CSTYLED */
2530                 fp = (rfs4_file_t *)rfs4_dbsearch(nsrv4->rfs4_file_idx, vp, create,
2531                     &arg, RFS4_DBS_VALID);
2532         else {
2533                 mutex_enter(&vp->v_vsd_lock);
2534                 fp = (rfs4_file_t *)vsd_get(vp, nfs4_srv_vkey);
2535                 if (fp) {
2536                         rfs4_dbe_lock(fp->rf_dbe);
2537                         if (rfs4_dbe_is_invalid(fp->rf_dbe) ||
2538                             (rfs4_dbe_refcnt(fp->rf_dbe) == 0)) {
2539                                 rfs4_dbe_unlock(fp->rf_dbe);
2540                                 fp = NULL;
2541                         } else {
2542                                 rfs4_dbe_hold(fp->rf_dbe);
2543                                 rfs4_dbe_unlock(fp->rf_dbe);
2544                         }
2545                 }
2546                 mutex_exit(&vp->v_vsd_lock);
2547         }
2548         return (fp);
2549 }
2550 
2551 /*
2552  * Find a file in the db and once it is located, take the rw lock.
2553  * Need to check the vnode pointer and if it does not exist (it was
2554  * removed between the db location and check) redo the find.  This
2555  * assumes that a file struct that has a NULL vnode pointer is marked
2556  * at 'invalid' and will not be found in the db the second time
2557  * around.
2558  */
2559 rfs4_file_t *
2560 rfs4_findfile_withlock(vnode_t *vp, nfs_fh4 *fh, bool_t *create)
2561 {
2562         rfs4_file_t *fp;
2563         rfs4_fcreate_arg arg;
2564         bool_t screate = *create;
2565         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2566 
2567         if (screate == FALSE) {
2568                 mutex_enter(&vp->v_vsd_lock);
2569                 fp = (rfs4_file_t *)vsd_get(vp, nfs4_srv_vkey);
2570                 if (fp) {
2571                         rfs4_dbe_lock(fp->rf_dbe);
2572                         if (rfs4_dbe_is_invalid(fp->rf_dbe) ||
2573                             (rfs4_dbe_refcnt(fp->rf_dbe) == 0)) {
2574                                 rfs4_dbe_unlock(fp->rf_dbe);
2575                                 mutex_exit(&vp->v_vsd_lock);
2576                                 fp = NULL;
2577                         } else {
2578                                 rfs4_dbe_hold(fp->rf_dbe);
2579                                 rfs4_dbe_unlock(fp->rf_dbe);
2580                                 mutex_exit(&vp->v_vsd_lock);
2581                                 rw_enter(&fp->rf_file_rwlock, RW_WRITER);
2582                                 if (fp->rf_vp == NULL) {
2583                                         rw_exit(&fp->rf_file_rwlock);
2584                                         rfs4_file_rele(fp);
2585                                         fp = NULL;
2586                                 }
2587                         }
2588                 } else {
2589                         mutex_exit(&vp->v_vsd_lock);
2590                 }
2591         } else {
2592 retry:
2593                 arg.vp = vp;
2594                 arg.fh = fh;
2595 
2596                 fp = (rfs4_file_t *)rfs4_dbsearch(nsrv4->rfs4_file_idx, vp,
2597                     create, &arg, RFS4_DBS_VALID);
2598                 if (fp != NULL) {
2599                         rw_enter(&fp->rf_file_rwlock, RW_WRITER);
2600                         if (fp->rf_vp == NULL) {
2601                                 rw_exit(&fp->rf_file_rwlock);
2602                                 rfs4_file_rele(fp);
2603                                 *create = screate;
2604                                 goto retry;
2605                         }
2606                 }
2607         }
2608 
2609         return (fp);
2610 }
2611 
2612 static uint32_t
2613 lo_state_hash(void *key)
2614 {
2615         stateid_t *id = key;
2616 
2617         return (id->bits.ident+id->bits.pid);
2618 }
2619 
2620 static bool_t
2621 lo_state_compare(rfs4_entry_t u_entry, void *key)
2622 {
2623         rfs4_lo_state_t *lsp = (rfs4_lo_state_t *)u_entry;
2624         stateid_t *id = key;
2625         bool_t rc;
2626 
2627         rc = (lsp->rls_lockid.bits.boottime == id->bits.boottime &&
2628             lsp->rls_lockid.bits.type == id->bits.type &&
2629             lsp->rls_lockid.bits.ident == id->bits.ident &&
2630             lsp->rls_lockid.bits.pid == id->bits.pid);
2631 
2632         return (rc);
2633 }
2634 
2635 static void *
2636 lo_state_mkkey(rfs4_entry_t u_entry)
2637 {
2638         rfs4_lo_state_t *lsp = (rfs4_lo_state_t *)u_entry;
2639 
2640         return (&lsp->rls_lockid);
2641 }
2642 
2643 static bool_t
2644 rfs4_lo_state_expiry(rfs4_entry_t u_entry)
2645 {
2646         rfs4_lo_state_t *lsp = (rfs4_lo_state_t *)u_entry;
2647 
2648         if (rfs4_dbe_is_invalid(lsp->rls_dbe))
2649                 return (TRUE);
2650         if (lsp->rls_state->rs_closed)
2651                 return (TRUE);
2652         return ((gethrestime_sec() -
2653             lsp->rls_state->rs_owner->ro_client->rc_last_access
2654             > rfs4_lease_time));
2655 }
2656 
2657 static void
2658 rfs4_lo_state_destroy(rfs4_entry_t u_entry)
2659 {
2660         rfs4_lo_state_t *lsp = (rfs4_lo_state_t *)u_entry;
2661 
2662         rfs4_dbe_lock(lsp->rls_state->rs_dbe);
2663         list_remove(&lsp->rls_state->rs_lostatelist, lsp);
2664         rfs4_dbe_unlock(lsp->rls_state->rs_dbe);
2665 
2666         rfs4_sw_destroy(&lsp->rls_sw);
2667 
2668         /* Make sure to release the file locks */
2669         if (lsp->rls_locks_cleaned == FALSE) {
2670                 lsp->rls_locks_cleaned = TRUE;
2671                 if (lsp->rls_locker->rl_client->rc_sysidt != LM_NOSYSID) {
2672                         /* Is the PxFS kernel module loaded? */
2673                         if (lm_remove_file_locks != NULL) {
2674                                 int new_sysid;
2675 
2676                                 /* Encode the cluster nodeid in new sysid */
2677                                 new_sysid =
2678                                     lsp->rls_locker->rl_client->rc_sysidt;
2679                                 lm_set_nlmid_flk(&new_sysid);
2680 
2681                                 /*
2682                                  * This PxFS routine removes file locks for a
2683                                  * client over all nodes of a cluster.
2684                                  */
2685                                 DTRACE_PROBE1(nfss_i_clust_rm_lck,
2686                                     int, new_sysid);
2687                                 (*lm_remove_file_locks)(new_sysid);
2688                         } else {
2689                                 (void) cleanlocks(
2690                                     lsp->rls_state->rs_finfo->rf_vp,
2691                                     lsp->rls_locker->rl_pid,
2692                                     lsp->rls_locker->rl_client->rc_sysidt);
2693                         }
2694                 }
2695         }
2696 
2697         /* Free the last reply for this state */
2698         rfs4_free_reply(&lsp->rls_reply);
2699 
2700         rfs4_lockowner_rele(lsp->rls_locker);
2701         lsp->rls_locker = NULL;
2702 
2703         rfs4_state_rele_nounlock(lsp->rls_state);
2704         lsp->rls_state = NULL;
2705 }
2706 
2707 static bool_t
2708 rfs4_lo_state_create(rfs4_entry_t u_entry, void *arg)
2709 {
2710         rfs4_lo_state_t *lsp = (rfs4_lo_state_t *)u_entry;
2711         rfs4_lo_state_t *argp = (rfs4_lo_state_t *)arg;
2712         rfs4_lockowner_t *lo = argp->rls_locker;
2713         rfs4_state_t *sp = argp->rls_state;
2714 
2715         lsp->rls_state = sp;
2716 
2717         lsp->rls_lockid = sp->rs_stateid;
2718         lsp->rls_lockid.bits.type = LOCKID;
2719         lsp->rls_lockid.bits.chgseq = 0;
2720         lsp->rls_lockid.bits.pid = lo->rl_pid;
2721 
2722         lsp->rls_locks_cleaned = FALSE;
2723         lsp->rls_lock_completed = FALSE;
2724 
2725         rfs4_sw_init(&lsp->rls_sw);
2726 
2727         /* Attached the supplied lock owner */
2728         rfs4_dbe_hold(lo->rl_dbe);
2729         lsp->rls_locker = lo;
2730 
2731         rfs4_dbe_lock(sp->rs_dbe);
2732         list_insert_tail(&sp->rs_lostatelist, lsp);
2733         rfs4_dbe_hold(sp->rs_dbe);
2734         rfs4_dbe_unlock(sp->rs_dbe);
2735 
2736         return (TRUE);
2737 }
2738 
2739 void
2740 rfs4_lo_state_rele(rfs4_lo_state_t *lsp, bool_t unlock_fp)
2741 {
2742         if (unlock_fp == TRUE)
2743                 rw_exit(&lsp->rls_state->rs_finfo->rf_file_rwlock);
2744         rfs4_dbe_rele(lsp->rls_dbe);
2745 }
2746 
2747 static rfs4_lo_state_t *
2748 rfs4_findlo_state(stateid_t *id, bool_t lock_fp)
2749 {
2750         rfs4_lo_state_t *lsp;
2751         bool_t create = FALSE;
2752         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2753 
2754         lsp = (rfs4_lo_state_t *)rfs4_dbsearch(nsrv4->rfs4_lo_state_idx, id,
2755             &create, NULL, RFS4_DBS_VALID);
2756         if (lock_fp == TRUE && lsp != NULL)
2757                 rw_enter(&lsp->rls_state->rs_finfo->rf_file_rwlock, RW_READER);
2758 
2759         return (lsp);
2760 }
2761 
2762 
2763 static uint32_t
2764 lo_state_lo_hash(void *key)
2765 {
2766         rfs4_lo_state_t *lsp = key;
2767 
2768         return (ADDRHASH(lsp->rls_locker) ^ ADDRHASH(lsp->rls_state));
2769 }
2770 
2771 static bool_t
2772 lo_state_lo_compare(rfs4_entry_t u_entry, void *key)
2773 {
2774         rfs4_lo_state_t *lsp = (rfs4_lo_state_t *)u_entry;
2775         rfs4_lo_state_t *keyp = key;
2776 
2777         return (keyp->rls_locker == lsp->rls_locker &&
2778             keyp->rls_state == lsp->rls_state);
2779 }
2780 
2781 static void *
2782 lo_state_lo_mkkey(rfs4_entry_t u_entry)
2783 {
2784         return (u_entry);
2785 }
2786 
2787 rfs4_lo_state_t *
2788 rfs4_findlo_state_by_owner(rfs4_lockowner_t *lo, rfs4_state_t *sp,
2789     bool_t *create)
2790 {
2791         rfs4_lo_state_t *lsp;
2792         rfs4_lo_state_t arg;
2793         nfs4_srv_t *nsrv4 = nfs4_get_srv();
2794 
2795         arg.rls_locker = lo;
2796         arg.rls_state = sp;
2797 
2798         lsp = (rfs4_lo_state_t *)rfs4_dbsearch(nsrv4->rfs4_lo_state_owner_idx,
2799             &arg, create, &arg, RFS4_DBS_VALID);
2800 
2801         return (lsp);
2802 }
2803 
2804 static stateid_t
2805 get_stateid(id_t eid)
2806 {
2807         stateid_t id;
2808         nfs4_srv_t *nsrv4;
2809 
2810         nsrv4 = nfs4_get_srv();
2811 
2812         id.bits.boottime = nsrv4->rfs4_start_time;
2813         id.bits.ident = eid;
2814         id.bits.chgseq = 0;
2815         id.bits.type = 0;
2816         id.bits.pid = 0;
2817 
2818         /*
2819          * If we are booted as a cluster node, embed our nodeid.
2820          * We've already done sanity checks in rfs4_client_create() so no
2821          * need to repeat them here.
2822          */
2823         id.bits.clnodeid = (cluster_bootflags & CLUSTER_BOOTED) ?
2824             clconf_get_nodeid() : 0;
2825 
2826         return (id);
2827 }
2828 
2829 /*
2830  * For use only when booted as a cluster node.
2831  * Returns TRUE if the embedded nodeid indicates that this stateid was
2832  * generated on another node.
2833  */
2834 static int
2835 foreign_stateid(stateid_t *id)
2836 {
2837         ASSERT(cluster_bootflags & CLUSTER_BOOTED);
2838         return (id->bits.clnodeid != (uint32_t)clconf_get_nodeid());
2839 }
2840 
2841 /*
2842  * For use only when booted as a cluster node.
2843  * Returns TRUE if the embedded nodeid indicates that this clientid was
2844  * generated on another node.
2845  */
2846 static int
2847 foreign_clientid(cid *cidp)
2848 {
2849         ASSERT(cluster_bootflags & CLUSTER_BOOTED);
2850         return (cidp->impl_id.c_id >> CLUSTER_NODEID_SHIFT !=
2851             (uint32_t)clconf_get_nodeid());
2852 }
2853 
2854 /*
2855  * For use only when booted as a cluster node.
2856  * Embed our cluster nodeid into the clientid.
2857  */
2858 static void
2859 embed_nodeid(cid *cidp)
2860 {
2861         int clnodeid;
2862         /*
2863          * Currently, our state tables are small enough that their
2864          * ids will leave enough bits free for the nodeid. If the
2865          * tables become larger, we mustn't overwrite the id.
2866          * Equally, we only have room for so many bits of nodeid, so
2867          * must check that too.
2868          */
2869         ASSERT(cluster_bootflags & CLUSTER_BOOTED);
2870         ASSERT(cidp->impl_id.c_id >> CLUSTER_NODEID_SHIFT == 0);
2871         clnodeid = clconf_get_nodeid();
2872         ASSERT(clnodeid <= CLUSTER_MAX_NODEID);
2873         ASSERT(clnodeid != NODEID_UNKNOWN);
2874         cidp->impl_id.c_id |= (clnodeid << CLUSTER_NODEID_SHIFT);
2875 }
2876 
2877 static uint32_t
2878 state_hash(void *key)
2879 {
2880         stateid_t *ip = (stateid_t *)key;
2881 
2882         return (ip->bits.ident);
2883 }
2884 
2885 static bool_t
2886 state_compare(rfs4_entry_t u_entry, void *key)
2887 {
2888         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
2889         stateid_t *id = (stateid_t *)key;
2890         bool_t rc;
2891 
2892         rc = (sp->rs_stateid.bits.boottime == id->bits.boottime &&
2893             sp->rs_stateid.bits.ident == id->bits.ident);
2894 
2895         return (rc);
2896 }
2897 
2898 static void *
2899 state_mkkey(rfs4_entry_t u_entry)
2900 {
2901         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
2902 
2903         return (&sp->rs_stateid);
2904 }
2905 
2906 static void
2907 rfs4_state_destroy(rfs4_entry_t u_entry)
2908 {
2909         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
2910 
2911         /* remove from openowner list */
2912         rfs4_dbe_lock(sp->rs_owner->ro_dbe);
2913         list_remove(&sp->rs_owner->ro_statelist, sp);
2914         rfs4_dbe_unlock(sp->rs_owner->ro_dbe);
2915 
2916         list_destroy(&sp->rs_lostatelist);
2917 
2918         /* release any share locks for this stateid if it's still open */
2919         if (!sp->rs_closed) {
2920                 rfs4_dbe_lock(sp->rs_dbe);
2921                 (void) rfs4_unshare(sp);
2922                 rfs4_dbe_unlock(sp->rs_dbe);
2923         }
2924 
2925         /* Were done with the file */
2926         rfs4_file_rele(sp->rs_finfo);
2927         sp->rs_finfo = NULL;
2928 
2929         /* And now with the openowner */
2930         rfs4_openowner_rele(sp->rs_owner);
2931         sp->rs_owner = NULL;
2932 }
2933 
2934 static void
2935 rfs4_state_rele_nounlock(rfs4_state_t *sp)
2936 {
2937         rfs4_dbe_rele(sp->rs_dbe);
2938 }
2939 
2940 void
2941 rfs4_state_rele(rfs4_state_t *sp)
2942 {
2943         rw_exit(&sp->rs_finfo->rf_file_rwlock);
2944         rfs4_dbe_rele(sp->rs_dbe);
2945 }
2946 
2947 static uint32_t
2948 deleg_hash(void *key)
2949 {
2950         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)key;
2951 
2952         return (ADDRHASH(dsp->rds_client) ^ ADDRHASH(dsp->rds_finfo));
2953 }
2954 
2955 static bool_t
2956 deleg_compare(rfs4_entry_t u_entry, void *key)
2957 {
2958         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)u_entry;
2959         rfs4_deleg_state_t *kdsp = (rfs4_deleg_state_t *)key;
2960 
2961         return (dsp->rds_client == kdsp->rds_client &&
2962             dsp->rds_finfo == kdsp->rds_finfo);
2963 }
2964 
2965 static void *
2966 deleg_mkkey(rfs4_entry_t u_entry)
2967 {
2968         return (u_entry);
2969 }
2970 
2971 static uint32_t
2972 deleg_state_hash(void *key)
2973 {
2974         stateid_t *ip = (stateid_t *)key;
2975 
2976         return (ip->bits.ident);
2977 }
2978 
2979 static bool_t
2980 deleg_state_compare(rfs4_entry_t u_entry, void *key)
2981 {
2982         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)u_entry;
2983         stateid_t *id = (stateid_t *)key;
2984         bool_t rc;
2985 
2986         if (id->bits.type != DELEGID)
2987                 return (FALSE);
2988 
2989         rc = (dsp->rds_delegid.bits.boottime == id->bits.boottime &&
2990             dsp->rds_delegid.bits.ident == id->bits.ident);
2991 
2992         return (rc);
2993 }
2994 
2995 static void *
2996 deleg_state_mkkey(rfs4_entry_t u_entry)
2997 {
2998         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)u_entry;
2999 
3000         return (&dsp->rds_delegid);
3001 }
3002 
3003 static bool_t
3004 rfs4_deleg_state_expiry(rfs4_entry_t u_entry)
3005 {
3006         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)u_entry;
3007 
3008         if (rfs4_dbe_is_invalid(dsp->rds_dbe))
3009                 return (TRUE);
3010 
3011         if (dsp->rds_dtype == OPEN_DELEGATE_NONE)
3012                 return (TRUE);
3013 
3014         if ((gethrestime_sec() - dsp->rds_client->rc_last_access
3015             > rfs4_lease_time)) {
3016                 rfs4_dbe_invalidate(dsp->rds_dbe);
3017                 return (TRUE);
3018         }
3019 
3020         return (FALSE);
3021 }
3022 
3023 static bool_t
3024 rfs4_deleg_state_create(rfs4_entry_t u_entry, void *argp)
3025 {
3026         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)u_entry;
3027         rfs4_file_t *fp = ((rfs4_deleg_state_t *)argp)->rds_finfo;
3028         rfs4_client_t *cp = ((rfs4_deleg_state_t *)argp)->rds_client;
3029 
3030         rfs4_dbe_hold(fp->rf_dbe);
3031         rfs4_dbe_hold(cp->rc_dbe);
3032 
3033         dsp->rds_delegid = get_stateid(rfs4_dbe_getid(dsp->rds_dbe));
3034         dsp->rds_delegid.bits.type = DELEGID;
3035         dsp->rds_finfo = fp;
3036         dsp->rds_client = cp;
3037         dsp->rds_dtype = OPEN_DELEGATE_NONE;
3038 
3039         dsp->rds_time_granted = gethrestime_sec();   /* observability */
3040         dsp->rds_time_revoked = 0;
3041 
3042         list_link_init(&dsp->rds_node);
3043 
3044         return (TRUE);
3045 }
3046 
3047 static void
3048 rfs4_deleg_state_destroy(rfs4_entry_t u_entry)
3049 {
3050         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)u_entry;
3051 
3052         /* return delegation if necessary */
3053         rfs4_return_deleg(dsp, FALSE);
3054 
3055         /* Were done with the file */
3056         rfs4_file_rele(dsp->rds_finfo);
3057         dsp->rds_finfo = NULL;
3058 
3059         /* And now with the openowner */
3060         rfs4_client_rele(dsp->rds_client);
3061         dsp->rds_client = NULL;
3062 }
3063 
3064 rfs4_deleg_state_t *
3065 rfs4_finddeleg(rfs4_state_t *sp, bool_t *create)
3066 {
3067         rfs4_deleg_state_t ds, *dsp;
3068         nfs4_srv_t *nsrv4 = nfs4_get_srv();
3069 
3070         ds.rds_client = sp->rs_owner->ro_client;
3071         ds.rds_finfo = sp->rs_finfo;
3072 
3073         dsp = (rfs4_deleg_state_t *)rfs4_dbsearch(nsrv4->rfs4_deleg_idx, &ds,
3074             create, &ds, RFS4_DBS_VALID);
3075 
3076         return (dsp);
3077 }
3078 
3079 rfs4_deleg_state_t *
3080 rfs4_finddelegstate(stateid_t *id)
3081 {
3082         rfs4_deleg_state_t *dsp;
3083         bool_t create = FALSE;
3084         nfs4_srv_t *nsrv4 = nfs4_get_srv();
3085 
3086         dsp = (rfs4_deleg_state_t *)rfs4_dbsearch(nsrv4->rfs4_deleg_state_idx,
3087             id, &create, NULL, RFS4_DBS_VALID);
3088 
3089         return (dsp);
3090 }
3091 
3092 void
3093 rfs4_deleg_state_rele(rfs4_deleg_state_t *dsp)
3094 {
3095         rfs4_dbe_rele(dsp->rds_dbe);
3096 }
3097 
3098 void
3099 rfs4_update_lock_sequence(rfs4_lo_state_t *lsp)
3100 {
3101 
3102         rfs4_dbe_lock(lsp->rls_dbe);
3103 
3104         /*
3105          * If we are skipping sequence id checking, this means that
3106          * this is the first lock request and therefore the sequence
3107          * id does not need to be updated.  This only happens on the
3108          * first lock request for a lockowner
3109          */
3110         if (!lsp->rls_skip_seqid_check)
3111                 lsp->rls_seqid++;
3112 
3113         rfs4_dbe_unlock(lsp->rls_dbe);
3114 }
3115 
3116 void
3117 rfs4_update_lock_resp(rfs4_lo_state_t *lsp, nfs_resop4 *resp)
3118 {
3119 
3120         rfs4_dbe_lock(lsp->rls_dbe);
3121 
3122         rfs4_free_reply(&lsp->rls_reply);
3123 
3124         rfs4_copy_reply(&lsp->rls_reply, resp);
3125 
3126         rfs4_dbe_unlock(lsp->rls_dbe);
3127 }
3128 
3129 void
3130 rfs4_free_opens(rfs4_openowner_t *oo, bool_t invalidate,
3131     bool_t close_of_client)
3132 {
3133         rfs4_state_t *sp;
3134 
3135         rfs4_dbe_lock(oo->ro_dbe);
3136 
3137         for (sp = list_head(&oo->ro_statelist); sp != NULL;
3138             sp = list_next(&oo->ro_statelist, sp)) {
3139                 rfs4_state_close(sp, FALSE, close_of_client, CRED());
3140                 if (invalidate == TRUE)
3141                         rfs4_dbe_invalidate(sp->rs_dbe);
3142         }
3143 
3144         rfs4_dbe_invalidate(oo->ro_dbe);
3145         rfs4_dbe_unlock(oo->ro_dbe);
3146 }
3147 
3148 static uint32_t
3149 state_owner_file_hash(void *key)
3150 {
3151         rfs4_state_t *sp = key;
3152 
3153         return (ADDRHASH(sp->rs_owner) ^ ADDRHASH(sp->rs_finfo));
3154 }
3155 
3156 static bool_t
3157 state_owner_file_compare(rfs4_entry_t u_entry, void *key)
3158 {
3159         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
3160         rfs4_state_t *arg = key;
3161 
3162         if (sp->rs_closed == TRUE)
3163                 return (FALSE);
3164 
3165         return (arg->rs_owner == sp->rs_owner && arg->rs_finfo == sp->rs_finfo);
3166 }
3167 
3168 static void *
3169 state_owner_file_mkkey(rfs4_entry_t u_entry)
3170 {
3171         return (u_entry);
3172 }
3173 
3174 static uint32_t
3175 state_file_hash(void *key)
3176 {
3177         return (ADDRHASH(key));
3178 }
3179 
3180 static bool_t
3181 state_file_compare(rfs4_entry_t u_entry, void *key)
3182 {
3183         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
3184         rfs4_file_t *fp = key;
3185 
3186         if (sp->rs_closed == TRUE)
3187                 return (FALSE);
3188 
3189         return (fp == sp->rs_finfo);
3190 }
3191 
3192 static void *
3193 state_file_mkkey(rfs4_entry_t u_entry)
3194 {
3195         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
3196 
3197         return (sp->rs_finfo);
3198 }
3199 
3200 rfs4_state_t *
3201 rfs4_findstate_by_owner_file(rfs4_openowner_t *oo, rfs4_file_t *fp,
3202     bool_t *create)
3203 {
3204         rfs4_state_t *sp;
3205         rfs4_state_t key;
3206         nfs4_srv_t *nsrv4 = nfs4_get_srv();
3207 
3208         key.rs_owner = oo;
3209         key.rs_finfo = fp;
3210 
3211         sp = (rfs4_state_t *)rfs4_dbsearch(nsrv4->rfs4_state_owner_file_idx,
3212             &key, create, &key, RFS4_DBS_VALID);
3213 
3214         return (sp);
3215 }
3216 
3217 /* This returns ANY state struct that refers to this file */
3218 static rfs4_state_t *
3219 rfs4_findstate_by_file(rfs4_file_t *fp)
3220 {
3221         bool_t create = FALSE;
3222         nfs4_srv_t *nsrv4 = nfs4_get_srv();
3223 
3224         return ((rfs4_state_t *)rfs4_dbsearch(nsrv4->rfs4_state_file_idx, fp,
3225             &create, fp, RFS4_DBS_VALID));
3226 }
3227 
3228 static bool_t
3229 rfs4_state_expiry(rfs4_entry_t u_entry)
3230 {
3231         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
3232 
3233         if (rfs4_dbe_is_invalid(sp->rs_dbe))
3234                 return (TRUE);
3235 
3236         if (sp->rs_closed == TRUE &&
3237             ((gethrestime_sec() - rfs4_dbe_get_timerele(sp->rs_dbe))
3238             > rfs4_lease_time))
3239                 return (TRUE);
3240 
3241         return ((gethrestime_sec() - sp->rs_owner->ro_client->rc_last_access
3242             > rfs4_lease_time));
3243 }
3244 
3245 static bool_t
3246 rfs4_state_create(rfs4_entry_t u_entry, void *argp)
3247 {
3248         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
3249         rfs4_file_t *fp = ((rfs4_state_t *)argp)->rs_finfo;
3250         rfs4_openowner_t *oo = ((rfs4_state_t *)argp)->rs_owner;
3251 
3252         rfs4_dbe_hold(fp->rf_dbe);
3253         rfs4_dbe_hold(oo->ro_dbe);
3254         sp->rs_stateid = get_stateid(rfs4_dbe_getid(sp->rs_dbe));
3255         sp->rs_stateid.bits.type = OPENID;
3256         sp->rs_owner = oo;
3257         sp->rs_finfo = fp;
3258 
3259         list_create(&sp->rs_lostatelist, sizeof (rfs4_lo_state_t),
3260             offsetof(rfs4_lo_state_t, rls_node));
3261 
3262         /* Insert state on per open owner's list */
3263         rfs4_dbe_lock(oo->ro_dbe);
3264         list_insert_tail(&oo->ro_statelist, sp);
3265         rfs4_dbe_unlock(oo->ro_dbe);
3266 
3267         return (TRUE);
3268 }
3269 
3270 static rfs4_state_t *
3271 rfs4_findstate(stateid_t *id, rfs4_dbsearch_type_t find_invalid, bool_t lock_fp)
3272 {
3273         rfs4_state_t *sp;
3274         bool_t create = FALSE;
3275         nfs4_srv_t *nsrv4 = nfs4_get_srv();
3276 
3277         sp = (rfs4_state_t *)rfs4_dbsearch(nsrv4->rfs4_state_idx, id,
3278             &create, NULL, find_invalid);
3279         if (lock_fp == TRUE && sp != NULL)
3280                 rw_enter(&sp->rs_finfo->rf_file_rwlock, RW_READER);
3281 
3282         return (sp);
3283 }
3284 
3285 void
3286 rfs4_state_close(rfs4_state_t *sp, bool_t lock_held, bool_t close_of_client,
3287     cred_t *cr)
3288 {
3289         /* Remove the associated lo_state owners */
3290         if (!lock_held)
3291                 rfs4_dbe_lock(sp->rs_dbe);
3292 
3293         /*
3294          * If refcnt == 0, the dbe is about to be destroyed.
3295          * lock state will be released by the reaper thread.
3296          */
3297 
3298         if (rfs4_dbe_refcnt(sp->rs_dbe) > 0) {
3299                 if (sp->rs_closed == FALSE) {
3300                         rfs4_release_share_lock_state(sp, cr, close_of_client);
3301                         sp->rs_closed = TRUE;
3302                 }
3303         }
3304 
3305         if (!lock_held)
3306                 rfs4_dbe_unlock(sp->rs_dbe);
3307 }
3308 
3309 /*
3310  * Remove all state associated with the given client.
3311  */
3312 void
3313 rfs4_client_state_remove(rfs4_client_t *cp)
3314 {
3315         rfs4_openowner_t *oo;
3316 
3317         rfs4_dbe_lock(cp->rc_dbe);
3318 
3319         for (oo = list_head(&cp->rc_openownerlist); oo != NULL;
3320             oo = list_next(&cp->rc_openownerlist, oo)) {
3321                 rfs4_free_opens(oo, TRUE, TRUE);
3322         }
3323 
3324         rfs4_dbe_unlock(cp->rc_dbe);
3325 }
3326 
3327 void
3328 rfs4_client_close(rfs4_client_t *cp)
3329 {
3330         /* Mark client as going away. */
3331         rfs4_dbe_lock(cp->rc_dbe);
3332         rfs4_dbe_invalidate(cp->rc_dbe);
3333         rfs4_dbe_unlock(cp->rc_dbe);
3334 
3335         rfs4_client_state_remove(cp);
3336 
3337         /* Release the client */
3338         rfs4_client_rele(cp);
3339 }
3340 
3341 nfsstat4
3342 rfs4_check_clientid(clientid4 *cp, int setclid_confirm)
3343 {
3344         cid *cidp = (cid *) cp;
3345         nfs4_srv_t *nsrv4;
3346 
3347         nsrv4 = nfs4_get_srv();
3348 
3349         /*
3350          * If we are booted as a cluster node, check the embedded nodeid.
3351          * If it indicates that this clientid was generated on another node,
3352          * inform the client accordingly.
3353          */
3354         if (cluster_bootflags & CLUSTER_BOOTED && foreign_clientid(cidp))
3355                 return (NFS4ERR_STALE_CLIENTID);
3356 
3357         /*
3358          * If the server start time matches the time provided
3359          * by the client (via the clientid) and this is NOT a
3360          * setclientid_confirm then return EXPIRED.
3361          */
3362         if (!setclid_confirm &&
3363             cidp->impl_id.start_time == nsrv4->rfs4_start_time)
3364                 return (NFS4ERR_EXPIRED);
3365 
3366         return (NFS4ERR_STALE_CLIENTID);
3367 }
3368 
3369 /*
3370  * This is used when a stateid has not been found amongst the
3371  * current server's state.  Check the stateid to see if it
3372  * was from this server instantiation or not.
3373  */
3374 static nfsstat4
3375 what_stateid_error(stateid_t *id, stateid_type_t type)
3376 {
3377         nfs4_srv_t *nsrv4;
3378 
3379         nsrv4 = nfs4_get_srv();
3380 
3381         /* If we are booted as a cluster node, was stateid locally generated? */
3382         if ((cluster_bootflags & CLUSTER_BOOTED) && foreign_stateid(id))
3383                 return (NFS4ERR_STALE_STATEID);
3384 
3385         /* If types don't match then no use checking further */
3386         if (type != id->bits.type)
3387                 return (NFS4ERR_BAD_STATEID);
3388 
3389         /* From a different server instantiation, return STALE */
3390         if (id->bits.boottime != nsrv4->rfs4_start_time)
3391                 return (NFS4ERR_STALE_STATEID);
3392 
3393         /*
3394          * From this server but the state is most likely beyond lease
3395          * timeout: return NFS4ERR_EXPIRED.  However, there is the
3396          * case of a delegation stateid.  For delegations, there is a
3397          * case where the state can be removed without the client's
3398          * knowledge/consent: revocation.  In the case of delegation
3399          * revocation, the delegation state will be removed and will
3400          * not be found.  If the client does something like a
3401          * DELEGRETURN or even a READ/WRITE with a delegatoin stateid
3402          * that has been revoked, the server should return BAD_STATEID
3403          * instead of the more common EXPIRED error.
3404          */
3405         if (id->bits.boottime == nsrv4->rfs4_start_time) {
3406                 if (type == DELEGID)
3407                         return (NFS4ERR_BAD_STATEID);
3408                 else
3409                         return (NFS4ERR_EXPIRED);
3410         }
3411 
3412         return (NFS4ERR_BAD_STATEID);
3413 }
3414 
3415 /*
3416  * Used later on to find the various state structs.  When called from
3417  * rfs4_check_stateid()->rfs4_get_all_state(), no file struct lock is
3418  * taken (it is not needed) and helps on the read/write path with
3419  * respect to performance.
3420  */
3421 static nfsstat4
3422 rfs4_get_state_lockit(stateid4 *stateid, rfs4_state_t **spp,
3423     rfs4_dbsearch_type_t find_invalid, bool_t lock_fp)
3424 {
3425         stateid_t *id = (stateid_t *)stateid;
3426         rfs4_state_t *sp;
3427 
3428         *spp = NULL;
3429 
3430         /* If we are booted as a cluster node, was stateid locally generated? */
3431         if ((cluster_bootflags & CLUSTER_BOOTED) && foreign_stateid(id))
3432                 return (NFS4ERR_STALE_STATEID);
3433 
3434         sp = rfs4_findstate(id, find_invalid, lock_fp);
3435         if (sp == NULL) {
3436                 return (what_stateid_error(id, OPENID));
3437         }
3438 
3439         if (rfs4_lease_expired(sp->rs_owner->ro_client)) {
3440                 if (lock_fp == TRUE)
3441                         rfs4_state_rele(sp);
3442                 else
3443                         rfs4_state_rele_nounlock(sp);
3444                 return (NFS4ERR_EXPIRED);
3445         }
3446 
3447         *spp = sp;
3448 
3449         return (NFS4_OK);
3450 }
3451 
3452 nfsstat4
3453 rfs4_get_state(stateid4 *stateid, rfs4_state_t **spp,
3454     rfs4_dbsearch_type_t find_invalid)
3455 {
3456         return (rfs4_get_state_lockit(stateid, spp, find_invalid, TRUE));
3457 }
3458 
3459 int
3460 rfs4_check_stateid_seqid(rfs4_state_t *sp, stateid4 *stateid)
3461 {
3462         stateid_t *id = (stateid_t *)stateid;
3463 
3464         if (rfs4_lease_expired(sp->rs_owner->ro_client))
3465                 return (NFS4_CHECK_STATEID_EXPIRED);
3466 
3467         /* Stateid is some time in the future - that's bad */
3468         if (sp->rs_stateid.bits.chgseq < id->bits.chgseq)
3469                 return (NFS4_CHECK_STATEID_BAD);
3470 
3471         if (sp->rs_stateid.bits.chgseq == id->bits.chgseq + 1)
3472                 return (NFS4_CHECK_STATEID_REPLAY);
3473 
3474         /* Stateid is some time in the past - that's old */
3475         if (sp->rs_stateid.bits.chgseq > id->bits.chgseq)
3476                 return (NFS4_CHECK_STATEID_OLD);
3477 
3478         /* Caller needs to know about confirmation before closure */
3479         if (sp->rs_owner->ro_need_confirm)
3480                 return (NFS4_CHECK_STATEID_UNCONFIRMED);
3481 
3482         if (sp->rs_closed == TRUE)
3483                 return (NFS4_CHECK_STATEID_CLOSED);
3484 
3485         return (NFS4_CHECK_STATEID_OKAY);
3486 }
3487 
3488 int
3489 rfs4_check_lo_stateid_seqid(rfs4_lo_state_t *lsp, stateid4 *stateid)
3490 {
3491         stateid_t *id = (stateid_t *)stateid;
3492 
3493         if (rfs4_lease_expired(lsp->rls_state->rs_owner->ro_client))
3494                 return (NFS4_CHECK_STATEID_EXPIRED);
3495 
3496         /* Stateid is some time in the future - that's bad */
3497         if (lsp->rls_lockid.bits.chgseq < id->bits.chgseq)
3498                 return (NFS4_CHECK_STATEID_BAD);
3499 
3500         if (lsp->rls_lockid.bits.chgseq == id->bits.chgseq + 1)
3501                 return (NFS4_CHECK_STATEID_REPLAY);
3502 
3503         /* Stateid is some time in the past - that's old */
3504         if (lsp->rls_lockid.bits.chgseq > id->bits.chgseq)
3505                 return (NFS4_CHECK_STATEID_OLD);
3506 
3507         if (lsp->rls_state->rs_closed == TRUE)
3508                 return (NFS4_CHECK_STATEID_CLOSED);
3509 
3510         return (NFS4_CHECK_STATEID_OKAY);
3511 }
3512 
3513 nfsstat4
3514 rfs4_get_deleg_state(stateid4 *stateid, rfs4_deleg_state_t **dspp)
3515 {
3516         stateid_t *id = (stateid_t *)stateid;
3517         rfs4_deleg_state_t *dsp;
3518 
3519         *dspp = NULL;
3520 
3521         /* If we are booted as a cluster node, was stateid locally generated? */
3522         if ((cluster_bootflags & CLUSTER_BOOTED) && foreign_stateid(id))
3523                 return (NFS4ERR_STALE_STATEID);
3524 
3525         dsp = rfs4_finddelegstate(id);
3526         if (dsp == NULL) {
3527                 return (what_stateid_error(id, DELEGID));
3528         }
3529 
3530         if (rfs4_lease_expired(dsp->rds_client)) {
3531                 rfs4_deleg_state_rele(dsp);
3532                 return (NFS4ERR_EXPIRED);
3533         }
3534 
3535         *dspp = dsp;
3536 
3537         return (NFS4_OK);
3538 }
3539 
3540 nfsstat4
3541 rfs4_get_lo_state(stateid4 *stateid, rfs4_lo_state_t **lspp, bool_t lock_fp)
3542 {
3543         stateid_t *id = (stateid_t *)stateid;
3544         rfs4_lo_state_t *lsp;
3545 
3546         *lspp = NULL;
3547 
3548         /* If we are booted as a cluster node, was stateid locally generated? */
3549         if ((cluster_bootflags & CLUSTER_BOOTED) && foreign_stateid(id))
3550                 return (NFS4ERR_STALE_STATEID);
3551 
3552         lsp = rfs4_findlo_state(id, lock_fp);
3553         if (lsp == NULL) {
3554                 return (what_stateid_error(id, LOCKID));
3555         }
3556 
3557         if (rfs4_lease_expired(lsp->rls_state->rs_owner->ro_client)) {
3558                 rfs4_lo_state_rele(lsp, lock_fp);
3559                 return (NFS4ERR_EXPIRED);
3560         }
3561 
3562         *lspp = lsp;
3563 
3564         return (NFS4_OK);
3565 }
3566 
3567 static nfsstat4
3568 rfs4_get_all_state(stateid4 *sid, rfs4_state_t **spp,
3569     rfs4_deleg_state_t **dspp, rfs4_lo_state_t **lspp)
3570 {
3571         rfs4_state_t *sp = NULL;
3572         rfs4_deleg_state_t *dsp = NULL;
3573         rfs4_lo_state_t *lsp = NULL;
3574         stateid_t *id;
3575         nfsstat4 status;
3576 
3577         *spp = NULL; *dspp = NULL; *lspp = NULL;
3578 
3579         id = (stateid_t *)sid;
3580         switch (id->bits.type) {
3581         case OPENID:
3582                 status = rfs4_get_state_lockit(sid, &sp, FALSE, FALSE);
3583                 break;
3584         case DELEGID:
3585                 status = rfs4_get_deleg_state(sid, &dsp);
3586                 break;
3587         case LOCKID:
3588                 status = rfs4_get_lo_state(sid, &lsp, FALSE);
3589                 if (status == NFS4_OK) {
3590                         sp = lsp->rls_state;
3591                         rfs4_dbe_hold(sp->rs_dbe);
3592                 }
3593                 break;
3594         default:
3595                 status = NFS4ERR_BAD_STATEID;
3596         }
3597 
3598         if (status == NFS4_OK) {
3599                 *spp = sp;
3600                 *dspp = dsp;
3601                 *lspp = lsp;
3602         }
3603 
3604         return (status);
3605 }
3606 
3607 /*
3608  * Given the I/O mode (FREAD or FWRITE), this checks whether the
3609  * rfs4_state_t struct has access to do this operation and if so
3610  * return NFS4_OK; otherwise the proper NFSv4 error is returned.
3611  */
3612 nfsstat4
3613 rfs4_state_has_access(rfs4_state_t *sp, int mode, vnode_t *vp)
3614 {
3615         nfsstat4 stat = NFS4_OK;
3616         rfs4_file_t *fp;
3617         bool_t create = FALSE;
3618 
3619         rfs4_dbe_lock(sp->rs_dbe);
3620         if (mode == FWRITE) {
3621                 if (!(sp->rs_share_access & OPEN4_SHARE_ACCESS_WRITE)) {
3622                         stat = NFS4ERR_OPENMODE;
3623                 }
3624         } else if (mode == FREAD) {
3625                 if (!(sp->rs_share_access & OPEN4_SHARE_ACCESS_READ)) {
3626                         /*
3627                          * If we have OPENed the file with DENYing access
3628                          * to both READ and WRITE then no one else could
3629                          * have OPENed the file, hence no conflicting READ
3630                          * deny.  This check is merely an optimization.
3631                          */
3632                         if (sp->rs_share_deny == OPEN4_SHARE_DENY_BOTH)
3633                                 goto out;
3634 
3635                         /* Check against file struct's DENY mode */
3636                         fp = rfs4_findfile(vp, NULL, &create);
3637                         if (fp != NULL) {
3638                                 int deny_read = 0;
3639                                 rfs4_dbe_lock(fp->rf_dbe);
3640                                 /*
3641                                  * Check if any other open owner has the file
3642                                  * OPENed with deny READ.
3643                                  */
3644                                 if (sp->rs_share_deny & OPEN4_SHARE_DENY_READ)
3645                                         deny_read = 1;
3646                                 ASSERT(fp->rf_deny_read >= deny_read);
3647                                 if (fp->rf_deny_read > deny_read)
3648                                         stat = NFS4ERR_OPENMODE;
3649                                 rfs4_dbe_unlock(fp->rf_dbe);
3650                                 rfs4_file_rele(fp);
3651                         }
3652                 }
3653         } else {
3654                 /* Illegal I/O mode */
3655                 stat = NFS4ERR_INVAL;
3656         }
3657 out:
3658         rfs4_dbe_unlock(sp->rs_dbe);
3659         return (stat);
3660 }
3661 
3662 /*
3663  * Given the I/O mode (FREAD or FWRITE), the vnode, the stateid and whether
3664  * the file is being truncated, return NFS4_OK if allowed or appropriate
3665  * V4 error if not. Note NFS4ERR_DELAY will be returned and a recall on
3666  * the associated file will be done if the I/O is not consistent with any
3667  * delegation in effect on the file. Should be holding VOP_RWLOCK, either
3668  * as reader or writer as appropriate. rfs4_op_open will acquire the
3669  * VOP_RWLOCK as writer when setting up delegation. If the stateid is bad
3670  * this routine will return NFS4ERR_BAD_STATEID. In addition, through the
3671  * deleg parameter, we will return whether a write delegation is held by
3672  * the client associated with this stateid.
3673  * If the server instance associated with the relevant client is in its
3674  * grace period, return NFS4ERR_GRACE.
3675  */
3676 
3677 nfsstat4
3678 rfs4_check_stateid(int mode, vnode_t *vp,
3679     stateid4 *stateid, bool_t trunc, bool_t *deleg,
3680     bool_t do_access, caller_context_t *ct)
3681 {
3682         rfs4_file_t *fp;
3683         bool_t create = FALSE;
3684         rfs4_state_t *sp;
3685         rfs4_deleg_state_t *dsp;
3686         rfs4_lo_state_t *lsp;
3687         stateid_t *id = (stateid_t *)stateid;
3688         nfsstat4 stat = NFS4_OK;
3689 
3690         if (ct != NULL) {
3691                 ct->cc_sysid = 0;
3692                 ct->cc_pid = 0;
3693                 ct->cc_caller_id = nfs4_srv_caller_id;
3694                 ct->cc_flags = CC_DONTBLOCK;
3695         }
3696 
3697         if (ISSPECIAL(stateid)) {
3698                 fp = rfs4_findfile(vp, NULL, &create);
3699                 if (fp == NULL)
3700                         return (NFS4_OK);
3701                 if (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_NONE) {
3702                         rfs4_file_rele(fp);
3703                         return (NFS4_OK);
3704                 }
3705                 if (mode == FWRITE ||
3706                     fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_WRITE) {
3707                         rfs4_recall_deleg(fp, trunc, NULL);
3708                         rfs4_file_rele(fp);
3709                         return (NFS4ERR_DELAY);
3710                 }
3711                 rfs4_file_rele(fp);
3712                 return (NFS4_OK);
3713         } else {
3714                 stat = rfs4_get_all_state(stateid, &sp, &dsp, &lsp);
3715                 if (stat != NFS4_OK)
3716                         return (stat);
3717                 if (lsp != NULL) {
3718                         /* Is associated server instance in its grace period? */
3719                         if (rfs4_clnt_in_grace(lsp->rls_locker->rl_client)) {
3720                                 rfs4_lo_state_rele(lsp, FALSE);
3721                                 if (sp != NULL)
3722                                         rfs4_state_rele_nounlock(sp);
3723                                 return (NFS4ERR_GRACE);
3724                         }
3725                         if (id->bits.type == LOCKID) {
3726                                 /* Seqid in the future? - that's bad */
3727                                 if (lsp->rls_lockid.bits.chgseq <
3728                                     id->bits.chgseq) {
3729                                         rfs4_lo_state_rele(lsp, FALSE);
3730                                         if (sp != NULL)
3731                                                 rfs4_state_rele_nounlock(sp);
3732                                         return (NFS4ERR_BAD_STATEID);
3733                                 }
3734                                 /* Seqid in the past? - that's old */
3735                                 if (lsp->rls_lockid.bits.chgseq >
3736                                     id->bits.chgseq) {
3737                                         rfs4_lo_state_rele(lsp, FALSE);
3738                                         if (sp != NULL)
3739                                                 rfs4_state_rele_nounlock(sp);
3740                                         return (NFS4ERR_OLD_STATEID);
3741                                 }
3742                                 /* Ensure specified filehandle matches */
3743                                 if (lsp->rls_state->rs_finfo->rf_vp != vp) {
3744                                         rfs4_lo_state_rele(lsp, FALSE);
3745                                         if (sp != NULL)
3746                                                 rfs4_state_rele_nounlock(sp);
3747                                         return (NFS4ERR_BAD_STATEID);
3748                                 }
3749                         }
3750                         if (ct != NULL) {
3751                                 ct->cc_sysid =
3752                                     lsp->rls_locker->rl_client->rc_sysidt;
3753                                 ct->cc_pid = lsp->rls_locker->rl_pid;
3754                         }
3755                         rfs4_lo_state_rele(lsp, FALSE);
3756                 }
3757 
3758                 /* Stateid provided was an "open" stateid */
3759                 if (sp != NULL) {
3760                         /* Is associated server instance in its grace period? */
3761                         if (rfs4_clnt_in_grace(sp->rs_owner->ro_client)) {
3762                                 rfs4_state_rele_nounlock(sp);
3763                                 return (NFS4ERR_GRACE);
3764                         }
3765                         if (id->bits.type == OPENID) {
3766                                 /* Seqid in the future? - that's bad */
3767                                 if (sp->rs_stateid.bits.chgseq <
3768                                     id->bits.chgseq) {
3769                                         rfs4_state_rele_nounlock(sp);
3770                                         return (NFS4ERR_BAD_STATEID);
3771                                 }
3772                                 /* Seqid in the past - that's old */
3773                                 if (sp->rs_stateid.bits.chgseq >
3774                                     id->bits.chgseq) {
3775                                         rfs4_state_rele_nounlock(sp);
3776                                         return (NFS4ERR_OLD_STATEID);
3777                                 }
3778                         }
3779                         /* Ensure specified filehandle matches */
3780                         if (sp->rs_finfo->rf_vp != vp) {
3781                                 rfs4_state_rele_nounlock(sp);
3782                                 return (NFS4ERR_BAD_STATEID);
3783                         }
3784 
3785                         if (sp->rs_owner->ro_need_confirm) {
3786                                 rfs4_state_rele_nounlock(sp);
3787                                 return (NFS4ERR_BAD_STATEID);
3788                         }
3789 
3790                         if (sp->rs_closed == TRUE) {
3791                                 rfs4_state_rele_nounlock(sp);
3792                                 return (NFS4ERR_OLD_STATEID);
3793                         }
3794 
3795                         if (do_access)
3796                                 stat = rfs4_state_has_access(sp, mode, vp);
3797                         else
3798                                 stat = NFS4_OK;
3799 
3800                         /*
3801                          * Return whether this state has write
3802                          * delegation if desired
3803                          */
3804                         if (deleg && (sp->rs_finfo->rf_dinfo.rd_dtype ==
3805                             OPEN_DELEGATE_WRITE))
3806                                 *deleg = TRUE;
3807 
3808                         /*
3809                          * We got a valid stateid, so we update the
3810                          * lease on the client. Ideally we would like
3811                          * to do this after the calling op succeeds,
3812                          * but for now this will be good
3813                          * enough. Callers of this routine are
3814                          * currently insulated from the state stuff.
3815                          */
3816                         rfs4_update_lease(sp->rs_owner->ro_client);
3817 
3818                         /*
3819                          * If a delegation is present on this file and
3820                          * this is a WRITE, then update the lastwrite
3821                          * time to indicate that activity is present.
3822                          */
3823                         if (sp->rs_finfo->rf_dinfo.rd_dtype ==
3824                             OPEN_DELEGATE_WRITE &&
3825                             mode == FWRITE) {
3826                                 sp->rs_finfo->rf_dinfo.rd_time_lastwrite =
3827                                     gethrestime_sec();
3828                         }
3829 
3830                         rfs4_state_rele_nounlock(sp);
3831 
3832                         return (stat);
3833                 }
3834 
3835                 if (dsp != NULL) {
3836                         /* Is associated server instance in its grace period? */
3837                         if (rfs4_clnt_in_grace(dsp->rds_client)) {
3838                                 rfs4_deleg_state_rele(dsp);
3839                                 return (NFS4ERR_GRACE);
3840                         }
3841                         if (dsp->rds_delegid.bits.chgseq != id->bits.chgseq) {
3842                                 rfs4_deleg_state_rele(dsp);
3843                                 return (NFS4ERR_BAD_STATEID);
3844                         }
3845 
3846                         /* Ensure specified filehandle matches */
3847                         if (dsp->rds_finfo->rf_vp != vp) {
3848                                 rfs4_deleg_state_rele(dsp);
3849                                 return (NFS4ERR_BAD_STATEID);
3850                         }
3851                         /*
3852                          * Return whether this state has write
3853                          * delegation if desired
3854                          */
3855                         if (deleg && (dsp->rds_finfo->rf_dinfo.rd_dtype ==
3856                             OPEN_DELEGATE_WRITE))
3857                                 *deleg = TRUE;
3858 
3859                         rfs4_update_lease(dsp->rds_client);
3860 
3861                         /*
3862                          * If a delegation is present on this file and
3863                          * this is a WRITE, then update the lastwrite
3864                          * time to indicate that activity is present.
3865                          */
3866                         if (dsp->rds_finfo->rf_dinfo.rd_dtype ==
3867                             OPEN_DELEGATE_WRITE && mode == FWRITE) {
3868                                 dsp->rds_finfo->rf_dinfo.rd_time_lastwrite =
3869                                     gethrestime_sec();
3870                         }
3871 
3872                         /*
3873                          * XXX - what happens if this is a WRITE and the
3874                          * delegation type of for READ.
3875                          */
3876                         rfs4_deleg_state_rele(dsp);
3877 
3878                         return (stat);
3879                 }
3880                 /*
3881                  * If we got this far, something bad happened
3882                  */
3883                 return (NFS4ERR_BAD_STATEID);
3884         }
3885 }
3886 
3887 
3888 /*
3889  * This is a special function in that for the file struct provided the
3890  * server wants to remove/close all current state associated with the
3891  * file.  The prime use of this would be with OP_REMOVE to force the
3892  * release of state and particularly of file locks.
3893  *
3894  * There is an assumption that there is no delegations outstanding on
3895  * this file at this point.  The caller should have waited for those
3896  * to be returned or revoked.
3897  */
3898 void
3899 rfs4_close_all_state(rfs4_file_t *fp)
3900 {
3901         rfs4_state_t *sp;
3902 
3903         rfs4_dbe_lock(fp->rf_dbe);
3904 
3905 #ifdef DEBUG
3906         /* only applies when server is handing out delegations */
3907         if (nfs4_get_deleg_policy() != SRV_NEVER_DELEGATE)
3908                 ASSERT(fp->rf_dinfo.rd_hold_grant > 0);
3909 #endif
3910 
3911         /* No delegations for this file */
3912         ASSERT(list_is_empty(&fp->rf_delegstatelist));
3913 
3914         /* Make sure that it can not be found */
3915         rfs4_dbe_invalidate(fp->rf_dbe);
3916 
3917         if (fp->rf_vp == NULL) {
3918                 rfs4_dbe_unlock(fp->rf_dbe);
3919                 return;
3920         }
3921         rfs4_dbe_unlock(fp->rf_dbe);
3922 
3923         /*
3924          * Hold as writer to prevent other server threads from
3925          * processing requests related to the file while all state is
3926          * being removed.
3927          */
3928         rw_enter(&fp->rf_file_rwlock, RW_WRITER);
3929 
3930         /* Remove ALL state from the file */
3931         while (sp = rfs4_findstate_by_file(fp)) {
3932                 rfs4_state_close(sp, FALSE, FALSE, CRED());
3933                 rfs4_state_rele_nounlock(sp);
3934         }
3935 
3936         /*
3937          * This is only safe since there are no further references to
3938          * the file.
3939          */
3940         rfs4_dbe_lock(fp->rf_dbe);
3941         if (fp->rf_vp) {
3942                 vnode_t *vp = fp->rf_vp;
3943 
3944                 mutex_enter(&vp->v_vsd_lock);
3945                 (void) vsd_set(vp, nfs4_srv_vkey, NULL);
3946                 mutex_exit(&vp->v_vsd_lock);
3947                 VN_RELE(vp);
3948                 fp->rf_vp = NULL;
3949         }
3950         rfs4_dbe_unlock(fp->rf_dbe);
3951 
3952         /* Finally let other references to proceed */
3953         rw_exit(&fp->rf_file_rwlock);
3954 }
3955 
3956 /*
3957  * This function is used as a target for the rfs4_dbe_walk() call
3958  * below.  The purpose of this function is to see if the
3959  * lockowner_state refers to a file that resides within the exportinfo
3960  * export.  If so, then remove the lock_owner state (file locks and
3961  * share "locks") for this object since the intent is the server is
3962  * unexporting the specified directory.  Be sure to invalidate the
3963  * object after the state has been released
3964  */
3965 static void
3966 rfs4_lo_state_walk_callout(rfs4_entry_t u_entry, void *e)
3967 {
3968         rfs4_lo_state_t *lsp = (rfs4_lo_state_t *)u_entry;
3969         struct exportinfo *exi = (struct exportinfo *)e;
3970         nfs_fh4_fmt_t   fhfmt4, *exi_fhp, *finfo_fhp;
3971         fhandle_t *efhp;
3972 
3973         efhp = (fhandle_t *)&exi->exi_fh;
3974         exi_fhp = (nfs_fh4_fmt_t *)&fhfmt4;
3975 
3976         FH_TO_FMT4(efhp, exi_fhp);
3977 
3978         finfo_fhp = (nfs_fh4_fmt_t *)lsp->rls_state->rs_finfo->
3979             rf_filehandle.nfs_fh4_val;
3980 
3981         if (EQFSID(&finfo_fhp->fh4_fsid, &exi_fhp->fh4_fsid) &&
3982             bcmp(&finfo_fhp->fh4_xdata, &exi_fhp->fh4_xdata,
3983             exi_fhp->fh4_xlen) == 0) {
3984                 rfs4_state_close(lsp->rls_state, FALSE, FALSE, CRED());
3985                 rfs4_dbe_invalidate(lsp->rls_dbe);
3986                 rfs4_dbe_invalidate(lsp->rls_state->rs_dbe);
3987         }
3988 }
3989 
3990 /*
3991  * This function is used as a target for the rfs4_dbe_walk() call
3992  * below.  The purpose of this function is to see if the state refers
3993  * to a file that resides within the exportinfo export.  If so, then
3994  * remove the open state for this object since the intent is the
3995  * server is unexporting the specified directory.  The main result for
3996  * this type of entry is to invalidate it such it will not be found in
3997  * the future.
3998  */
3999 static void
4000 rfs4_state_walk_callout(rfs4_entry_t u_entry, void *e)
4001 {
4002         rfs4_state_t *sp = (rfs4_state_t *)u_entry;
4003         struct exportinfo *exi = (struct exportinfo *)e;
4004         nfs_fh4_fmt_t   fhfmt4, *exi_fhp, *finfo_fhp;
4005         fhandle_t *efhp;
4006 
4007         efhp = (fhandle_t *)&exi->exi_fh;
4008         exi_fhp = (nfs_fh4_fmt_t *)&fhfmt4;
4009 
4010         FH_TO_FMT4(efhp, exi_fhp);
4011 
4012         finfo_fhp =
4013             (nfs_fh4_fmt_t *)sp->rs_finfo->rf_filehandle.nfs_fh4_val;
4014 
4015         if (EQFSID(&finfo_fhp->fh4_fsid, &exi_fhp->fh4_fsid) &&
4016             bcmp(&finfo_fhp->fh4_xdata, &exi_fhp->fh4_xdata,
4017             exi_fhp->fh4_xlen) == 0) {
4018                 rfs4_state_close(sp, TRUE, FALSE, CRED());
4019                 rfs4_dbe_invalidate(sp->rs_dbe);
4020         }
4021 }
4022 
4023 /*
4024  * This function is used as a target for the rfs4_dbe_walk() call
4025  * below.  The purpose of this function is to see if the state refers
4026  * to a file that resides within the exportinfo export.  If so, then
4027  * remove the deleg state for this object since the intent is the
4028  * server is unexporting the specified directory.  The main result for
4029  * this type of entry is to invalidate it such it will not be found in
4030  * the future.
4031  */
4032 static void
4033 rfs4_deleg_state_walk_callout(rfs4_entry_t u_entry, void *e)
4034 {
4035         rfs4_deleg_state_t *dsp = (rfs4_deleg_state_t *)u_entry;
4036         struct exportinfo *exi = (struct exportinfo *)e;
4037         nfs_fh4_fmt_t   fhfmt4, *exi_fhp, *finfo_fhp;
4038         fhandle_t *efhp;
4039 
4040         efhp = (fhandle_t *)&exi->exi_fh;
4041         exi_fhp = (nfs_fh4_fmt_t *)&fhfmt4;
4042 
4043         FH_TO_FMT4(efhp, exi_fhp);
4044 
4045         finfo_fhp =
4046             (nfs_fh4_fmt_t *)dsp->rds_finfo->rf_filehandle.nfs_fh4_val;
4047 
4048         if (EQFSID(&finfo_fhp->fh4_fsid, &exi_fhp->fh4_fsid) &&
4049             bcmp(&finfo_fhp->fh4_xdata, &exi_fhp->fh4_xdata,
4050             exi_fhp->fh4_xlen) == 0) {
4051                 rfs4_dbe_invalidate(dsp->rds_dbe);
4052         }
4053 }
4054 
4055 /*
4056  * This function is used as a target for the rfs4_dbe_walk() call
4057  * below.  The purpose of this function is to see if the state refers
4058  * to a file that resides within the exportinfo export.  If so, then
4059  * release vnode hold for this object since the intent is the server
4060  * is unexporting the specified directory.  Invalidation will prevent
4061  * this struct from being found in the future.
4062  */
4063 static void
4064 rfs4_file_walk_callout(rfs4_entry_t u_entry, void *e)
4065 {
4066         rfs4_file_t *fp = (rfs4_file_t *)u_entry;
4067         struct exportinfo *exi = (struct exportinfo *)e;
4068         nfs_fh4_fmt_t   fhfmt4, *exi_fhp, *finfo_fhp;
4069         fhandle_t *efhp;
4070 
4071         efhp = (fhandle_t *)&exi->exi_fh;
4072         exi_fhp = (nfs_fh4_fmt_t *)&fhfmt4;
4073 
4074         FH_TO_FMT4(efhp, exi_fhp);
4075 
4076         finfo_fhp = (nfs_fh4_fmt_t *)fp->rf_filehandle.nfs_fh4_val;
4077 
4078         if (EQFSID(&finfo_fhp->fh4_fsid, &exi_fhp->fh4_fsid) &&
4079             bcmp(&finfo_fhp->fh4_xdata, &exi_fhp->fh4_xdata,
4080             exi_fhp->fh4_xlen) == 0) {
4081                 if (fp->rf_vp) {
4082                         vnode_t *vp = fp->rf_vp;
4083 
4084                         /*
4085                          * don't leak monitors and remove the reference
4086                          * put on the vnode when the delegation was granted.
4087                          */
4088                         if (fp->rf_dinfo.rd_dtype == OPEN_DELEGATE_READ) {
4089                                 (void) fem_uninstall(vp, deleg_rdops,
4090                                     (void *)fp);
4091                                 vn_open_downgrade(vp, FREAD);
4092                         } else if (fp->rf_dinfo.rd_dtype ==
4093                             OPEN_DELEGATE_WRITE) {
4094                                 (void) fem_uninstall(vp, deleg_wrops,
4095                                     (void *)fp);
4096                                 vn_open_downgrade(vp, FREAD|FWRITE);
4097                         }
4098                         mutex_enter(&vp->v_vsd_lock);
4099                         (void) vsd_set(vp, nfs4_srv_vkey, NULL);
4100                         mutex_exit(&vp->v_vsd_lock);
4101                         VN_RELE(vp);
4102                         fp->rf_vp = NULL;
4103                 }
4104                 rfs4_dbe_invalidate(fp->rf_dbe);
4105         }
4106 }
4107 
4108 /*
4109  * Given a directory that is being unexported, cleanup/release all
4110  * state in the server that refers to objects residing underneath this
4111  * particular export.  The ordering of the release is important.
4112  * Lock_owner, then state and then file.
4113  *
4114  * NFS zones note: nfs_export.c:unexport() calls this from a
4115  * thread in the global zone for NGZ data structures, so we
4116  * CANNOT use zone_getspecific anywhere in this code path.
4117  */
4118 void
4119 rfs4_clean_state_exi(nfs_export_t *ne, struct exportinfo *exi)
4120 {
4121         nfs_globals_t *ng;
4122         nfs4_srv_t *nsrv4;
4123 
4124         ng = ne->ne_globals;
4125         ASSERT(ng->nfs_zoneid == exi->exi_zoneid);
4126         nsrv4 = ng->nfs4_srv;
4127 
4128         mutex_enter(&nsrv4->state_lock);
4129 
4130         if (nsrv4->nfs4_server_state == NULL) {
4131                 mutex_exit(&nsrv4->state_lock);
4132                 return;
4133         }
4134 
4135         rfs4_dbe_walk(nsrv4->rfs4_lo_state_tab,
4136             rfs4_lo_state_walk_callout, exi);
4137         rfs4_dbe_walk(nsrv4->rfs4_state_tab, rfs4_state_walk_callout, exi);
4138         rfs4_dbe_walk(nsrv4->rfs4_deleg_state_tab,
4139             rfs4_deleg_state_walk_callout, exi);
4140         rfs4_dbe_walk(nsrv4->rfs4_file_tab, rfs4_file_walk_callout, exi);
4141 
4142         mutex_exit(&nsrv4->state_lock);
4143 }