sys/fs/nfsclient/nfs_clrpcops.c

   1 /*-
   2  * Copyright (c) 1989, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to Berkeley by
   6  * Rick Macklem at The University of Guelph.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 4. Neither the name of the University nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  *
  32  */
  33
  34 #include <sys/cdefs.h>
  35 __FBSDID("$FreeBSD$");
  36
  37 /*
  38  * Rpc op calls, generally called from the vnode op calls or through the
  39  * buffer cache, for NFS v2, 3 and 4.
  40  * These do not normally make any changes to vnode arguments or use
  41  * structures that might change between the VFS variants. The returned
  42  * arguments are all at the end, after the NFSPROC_T *p one.
  43  */
  44
  45 #ifndef APPLEKEXT
  46 #include "opt_inet6.h"
  47
  48 #include <fs/nfs/nfsport.h>
  49 #include <sys/sysctl.h>
  50
  51 SYSCTL_DECL(_vfs_nfs);
  52
  53 static int      nfsignore_eexist = 0;
  54 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
  55     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
  56
  57 /*
  58  * Global variables
  59  */
  60 extern int nfs_numnfscbd;
  61 extern struct timeval nfsboottime;
  62 extern u_int32_t newnfs_false, newnfs_true;
  63 extern nfstype nfsv34_type[9];
  64 extern int nfsrv_useacl;
  65 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
  66 extern int nfscl_debuglevel;
  67 NFSCLSTATEMUTEX;
  68 int nfstest_outofseq = 0;
  69 int nfscl_assumeposixlocks = 1;
  70 int nfscl_enablecallb = 0;
  71 short nfsv4_cbport = NFSV4_CBPORT;
  72 int nfstest_openallsetattr = 0;
  73 #endif  /* !APPLEKEXT */
  74
  75 #define DIRHDSIZ        (sizeof (struct dirent) - (MAXNAMLEN + 1))
  76
  77 /*
  78  * nfscl_getsameserver() can return one of three values:
  79  * NFSDSP_USETHISSESSION - Use this session for the DS.
  80  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
  81  *     session.
  82  * NFSDSP_NOTFOUND - No matching server was found.
  83  */
  84 enum nfsclds_state {
  85         NFSDSP_USETHISSESSION = 0,
  86         NFSDSP_SEQTHISSESSION = 1,
  87         NFSDSP_NOTFOUND = 2,
  88 };
  89
  90 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
  91     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
  92 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
  93     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
  94 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
  95     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
  96     void *);
  97 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
  98     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
  99     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
 100 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
 101     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
 102     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
 103     int *, void *, int *);
 104 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
 105     struct nfscllockowner *, u_int64_t, u_int64_t,
 106     u_int32_t, struct ucred *, NFSPROC_T *, int);
 107 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
 108     struct acl *, nfsv4stateid_t *, void *);
 109 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
 110     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
 111     struct ucred *, NFSPROC_T *);
 112 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_storage *,
 113     struct nfsclds **, NFSPROC_T *);
 114 static void nfscl_initsessionslots(struct nfsclsession *);
 115 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
 116     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
 117     struct nfsclflayout *, uint64_t, uint64_t, struct ucred *, NFSPROC_T *);
 118 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
 119     struct nfsclds *, uint64_t, int, struct nfsfh *, struct ucred *,
 120     NFSPROC_T *);
 121 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
 122     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
 123     struct nfsfh *, int, struct ucred *, NFSPROC_T *);
 124 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
 125     struct nfsclds *, struct nfsclds **);
 126 #ifdef notyet
 127 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
 128     struct nfsfh *, struct ucred *, NFSPROC_T *, void *);
 129 #endif
 130
 131 /*
 132  * nfs null call from vfs.
 133  */
 134 APPLESTATIC int
 135 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
 136 {
 137         int error;
 138         struct nfsrv_descript nfsd, *nd = &nfsd;
 139
 140         NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
 141         error = nfscl_request(nd, vp, p, cred, NULL);
 142         if (nd->nd_repstat && !error)
 143                 error = nd->nd_repstat;
 144         mbuf_freem(nd->nd_mrep);
 145         return (error);
 146 }
 147
 148 /*
 149  * nfs access rpc op.
 150  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
 151  * modes are changed on the server, accesses might still fail later.
 152  */
 153 APPLESTATIC int
 154 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
 155     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
 156 {
 157         int error;
 158         u_int32_t mode, rmode;
 159
 160         if (acmode & VREAD)
 161                 mode = NFSACCESS_READ;
 162         else
 163                 mode = 0;
 164         if (vnode_vtype(vp) == VDIR) {
 165                 if (acmode & VWRITE)
 166                         mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
 167                                  NFSACCESS_DELETE);
 168                 if (acmode & VEXEC)
 169                         mode |= NFSACCESS_LOOKUP;
 170         } else {
 171                 if (acmode & VWRITE)
 172                         mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
 173                 if (acmode & VEXEC)
 174                         mode |= NFSACCESS_EXECUTE;
 175         }
 176
 177         /*
 178          * Now, just call nfsrpc_accessrpc() to do the actual RPC.
 179          */
 180         error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
 181             NULL);
 182
 183         /*
 184          * The NFS V3 spec does not clarify whether or not
 185          * the returned access bits can be a superset of
 186          * the ones requested, so...
 187          */
 188         if (!error && (rmode & mode) != mode)
 189                 error = EACCES;
 190         return (error);
 191 }
 192
 193 /*
 194  * The actual rpc, separated out for Darwin.
 195  */
 196 APPLESTATIC int
 197 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
 198     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
 199     void *stuff)
 200 {
 201         u_int32_t *tl;
 202         u_int32_t supported, rmode;
 203         int error;
 204         struct nfsrv_descript nfsd, *nd = &nfsd;
 205         nfsattrbit_t attrbits;
 206
 207         *attrflagp = 0;
 208         supported = mode;
 209         NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
 210         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 211         *tl = txdr_unsigned(mode);
 212         if (nd->nd_flag & ND_NFSV4) {
 213                 /*
 214                  * And do a Getattr op.
 215                  */
 216                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 217                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
 218                 NFSGETATTR_ATTRBIT(&attrbits);
 219                 (void) nfsrv_putattrbit(nd, &attrbits);
 220         }
 221         error = nfscl_request(nd, vp, p, cred, stuff);
 222         if (error)
 223                 return (error);
 224         if (nd->nd_flag & ND_NFSV3) {
 225                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
 226                 if (error)
 227                         goto nfsmout;
 228         }
 229         if (!nd->nd_repstat) {
 230                 if (nd->nd_flag & ND_NFSV4) {
 231                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 232                         supported = fxdr_unsigned(u_int32_t, *tl++);
 233                 } else {
 234                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
 235                 }
 236                 rmode = fxdr_unsigned(u_int32_t, *tl);
 237                 if (nd->nd_flag & ND_NFSV4)
 238                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
 239
 240                 /*
 241                  * It's not obvious what should be done about
 242                  * unsupported access modes. For now, be paranoid
 243                  * and clear the unsupported ones.
 244                  */
 245                 rmode &= supported;
 246                 *rmodep = rmode;
 247         } else
 248                 error = nd->nd_repstat;
 249 nfsmout:
 250         mbuf_freem(nd->nd_mrep);
 251         return (error);
 252 }
 253
 254 /*
 255  * nfs open rpc
 256  */
 257 APPLESTATIC int
 258 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
 259 {
 260         struct nfsclopen *op;
 261         struct nfscldeleg *dp;
 262         struct nfsfh *nfhp;
 263         struct nfsnode *np = VTONFS(vp);
 264         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
 265         u_int32_t mode, clidrev;
 266         int ret, newone, error, expireret = 0, retrycnt;
 267
 268         /*
 269          * For NFSv4, Open Ops are only done on Regular Files.
 270          */
 271         if (vnode_vtype(vp) != VREG)
 272                 return (0);
 273         mode = 0;
 274         if (amode & FREAD)
 275                 mode |= NFSV4OPEN_ACCESSREAD;
 276         if (amode & FWRITE)
 277                 mode |= NFSV4OPEN_ACCESSWRITE;
 278         nfhp = np->n_fhp;
 279
 280         retrycnt = 0;
 281 #ifdef notdef
 282 { char name[100]; int namel;
 283 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
 284 bcopy(NFS4NODENAME(np->n_v4), name, namel);
 285 name[namel] = '\0';
 286 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
 287 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
 288 else printf(" fhl=0\n");
 289 }
 290 #endif
 291         do {
 292             dp = NULL;
 293             error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
 294                 cred, p, NULL, &op, &newone, &ret, 1);
 295             if (error) {
 296                 return (error);
 297             }
 298             if (nmp->nm_clp != NULL)
 299                 clidrev = nmp->nm_clp->nfsc_clientidrev;
 300             else
 301                 clidrev = 0;
 302             if (ret == NFSCLOPEN_DOOPEN) {
 303                 if (np->n_v4 != NULL) {
 304                         error = nfsrpc_openrpc(nmp, vp, np->n_v4->n4_data,
 305                            np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
 306                            np->n_fhp->nfh_len, mode, op,
 307                            NFS4NODENAME(np->n_v4), np->n_v4->n4_namelen, &dp,
 308                            0, 0x0, cred, p, 0, 0);
 309                         if (dp != NULL) {
 310 #ifdef APPLE
 311                                 OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
 312 #else
 313                                 NFSLOCKNODE(np);
 314                                 np->n_flag &= ~NDELEGMOD;
 315                                 /*
 316                                  * Invalidate the attribute cache, so that
 317                                  * attributes that pre-date the issue of a
 318                                  * delegation are not cached, since the
 319                                  * cached attributes will remain valid while
 320                                  * the delegation is held.
 321                                  */
 322                                 NFSINVALATTRCACHE(np);
 323                                 NFSUNLOCKNODE(np);
 324 #endif
 325                                 (void) nfscl_deleg(nmp->nm_mountp,
 326                                     op->nfso_own->nfsow_clp,
 327                                     nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
 328                         }
 329                 } else {
 330                         error = EIO;
 331                 }
 332                 newnfs_copyincred(cred, &op->nfso_cred);
 333             } else if (ret == NFSCLOPEN_SETCRED)
 334                 /*
 335                  * This is a new local open on a delegation. It needs
 336                  * to have credentials so that an open can be done
 337                  * against the server during recovery.
 338                  */
 339                 newnfs_copyincred(cred, &op->nfso_cred);
 340
 341             /*
 342              * nfso_opencnt is the count of how many VOP_OPEN()s have
 343              * been done on this Open successfully and a VOP_CLOSE()
 344              * is expected for each of these.
 345              * If error is non-zero, don't increment it, since the Open
 346              * hasn't succeeded yet.
 347              */
 348             if (!error)
 349                 op->nfso_opencnt++;
 350             nfscl_openrelease(op, error, newone);
 351             if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
 352                 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
 353                 error == NFSERR_BADSESSION) {
 354                 (void) nfs_catnap(PZERO, error, "nfs_open");
 355             } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
 356                 && clidrev != 0) {
 357                 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
 358                 retrycnt++;
 359             }
 360         } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
 361             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
 362             error == NFSERR_BADSESSION ||
 363             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
 364              expireret == 0 && clidrev != 0 && retrycnt < 4));
 365         if (error && retrycnt >= 4)
 366                 error = EIO;
 367         return (error);
 368 }
 369
 370 /*
 371  * the actual open rpc
 372  */
 373 APPLESTATIC int
 374 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
 375     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
 376     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
 377     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
 378     int syscred, int recursed)
 379 {
 380         u_int32_t *tl;
 381         struct nfsrv_descript nfsd, *nd = &nfsd;
 382         struct nfscldeleg *dp, *ndp = NULL;
 383         struct nfsvattr nfsva;
 384         u_int32_t rflags, deleg;
 385         nfsattrbit_t attrbits;
 386         int error, ret, acesize, limitby;
 387
 388         dp = *dpp;
 389         *dpp = NULL;
 390         nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL);
 391         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
 392         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 393         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
 394         *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
 395         *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
 396         *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
 397         (void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
 398         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 399         *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
 400         if (reclaim) {
 401                 *tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
 402                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 403                 *tl = txdr_unsigned(delegtype);
 404         } else {
 405                 if (dp != NULL) {
 406                         *tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
 407                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
 408                         if (NFSHASNFSV4N(nmp))
 409                                 *tl++ = 0;
 410                         else
 411                                 *tl++ = dp->nfsdl_stateid.seqid;
 412                         *tl++ = dp->nfsdl_stateid.other[0];
 413                         *tl++ = dp->nfsdl_stateid.other[1];
 414                         *tl = dp->nfsdl_stateid.other[2];
 415                 } else {
 416                         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
 417                 }
 418                 (void) nfsm_strtom(nd, name, namelen);
 419         }
 420         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 421         *tl = txdr_unsigned(NFSV4OP_GETATTR);
 422         NFSZERO_ATTRBIT(&attrbits);
 423         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
 424         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
 425         (void) nfsrv_putattrbit(nd, &attrbits);
 426         if (syscred)
 427                 nd->nd_flag |= ND_USEGSSNAME;
 428         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
 429             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 430         if (error)
 431                 return (error);
 432         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 433         if (!nd->nd_repstat) {
 434                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
 435                     6 * NFSX_UNSIGNED);
 436                 op->nfso_stateid.seqid = *tl++;
 437                 op->nfso_stateid.other[0] = *tl++;
 438                 op->nfso_stateid.other[1] = *tl++;
 439                 op->nfso_stateid.other[2] = *tl;
 440                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
 441                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
 442                 if (error)
 443                         goto nfsmout;
 444                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
 445                 deleg = fxdr_unsigned(u_int32_t, *tl);
 446                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
 447                     deleg == NFSV4OPEN_DELEGATEWRITE) {
 448                         if (!(op->nfso_own->nfsow_clp->nfsc_flags &
 449                               NFSCLFLAGS_FIRSTDELEG))
 450                                 op->nfso_own->nfsow_clp->nfsc_flags |=
 451                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
 452                         MALLOC(ndp, struct nfscldeleg *,
 453                             sizeof (struct nfscldeleg) + newfhlen,
 454                             M_NFSCLDELEG, M_WAITOK);
 455                         LIST_INIT(&ndp->nfsdl_owner);
 456                         LIST_INIT(&ndp->nfsdl_lock);
 457                         ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
 458                         ndp->nfsdl_fhlen = newfhlen;
 459                         NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
 460                         newnfs_copyincred(cred, &ndp->nfsdl_cred);
 461                         nfscl_lockinit(&ndp->nfsdl_rwlock);
 462                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
 463                             NFSX_UNSIGNED);
 464                         ndp->nfsdl_stateid.seqid = *tl++;
 465                         ndp->nfsdl_stateid.other[0] = *tl++;
 466                         ndp->nfsdl_stateid.other[1] = *tl++;
 467                         ndp->nfsdl_stateid.other[2] = *tl++;
 468                         ret = fxdr_unsigned(int, *tl);
 469                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
 470                                 ndp->nfsdl_flags = NFSCLDL_WRITE;
 471                                 /*
 472                                  * Indicates how much the file can grow.
 473                                  */
 474                                 NFSM_DISSECT(tl, u_int32_t *,
 475                                     3 * NFSX_UNSIGNED);
 476                                 limitby = fxdr_unsigned(int, *tl++);
 477                                 switch (limitby) {
 478                                 case NFSV4OPEN_LIMITSIZE:
 479                                         ndp->nfsdl_sizelimit = fxdr_hyper(tl);
 480                                         break;
 481                                 case NFSV4OPEN_LIMITBLOCKS:
 482                                         ndp->nfsdl_sizelimit =
 483                                             fxdr_unsigned(u_int64_t, *tl++);
 484                                         ndp->nfsdl_sizelimit *=
 485                                             fxdr_unsigned(u_int64_t, *tl);
 486                                         break;
 487                                 default:
 488                                         error = NFSERR_BADXDR;
 489                                         goto nfsmout;
 490                                 };
 491                         } else {
 492                                 ndp->nfsdl_flags = NFSCLDL_READ;
 493                         }
 494                         if (ret)
 495                                 ndp->nfsdl_flags |= NFSCLDL_RECALL;
 496                         error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
 497                             &acesize, p);
 498                         if (error)
 499                                 goto nfsmout;
 500                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
 501                         error = NFSERR_BADXDR;
 502                         goto nfsmout;
 503                 }
 504                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 505                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
 506                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
 507                     NULL, NULL, NULL, p, cred);
 508                 if (error)
 509                         goto nfsmout;
 510                 if (ndp != NULL) {
 511                         ndp->nfsdl_change = nfsva.na_filerev;
 512                         ndp->nfsdl_modtime = nfsva.na_mtime;
 513                         ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
 514                 }
 515                 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
 516                     do {
 517                         ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
 518                             cred, p);
 519                         if (ret == NFSERR_DELAY)
 520                             (void) nfs_catnap(PZERO, ret, "nfs_open");
 521                     } while (ret == NFSERR_DELAY);
 522                     error = ret;
 523                 }
 524                 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
 525                     nfscl_assumeposixlocks)
 526                     op->nfso_posixlock = 1;
 527                 else
 528                     op->nfso_posixlock = 0;
 529
 530                 /*
 531                  * If the server is handing out delegations, but we didn't
 532                  * get one because an OpenConfirm was required, try the
 533                  * Open again, to get a delegation. This is a harmless no-op,
 534                  * from a server's point of view.
 535                  */
 536                 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
 537                     (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
 538                     && !error && dp == NULL && ndp == NULL && !recursed) {
 539                     do {
 540                         ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
 541                             newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
 542                             cred, p, syscred, 1);
 543                         if (ret == NFSERR_DELAY)
 544                             (void) nfs_catnap(PZERO, ret, "nfs_open2");
 545                     } while (ret == NFSERR_DELAY);
 546                     if (ret) {
 547                         if (ndp != NULL)
 548                                 FREE((caddr_t)ndp, M_NFSCLDELEG);
 549                         if (ret == NFSERR_STALECLIENTID ||
 550                             ret == NFSERR_STALEDONTRECOVER ||
 551                             ret == NFSERR_BADSESSION)
 552                                 error = ret;
 553                     }
 554                 }
 555         }
 556         if (nd->nd_repstat != 0 && error == 0)
 557                 error = nd->nd_repstat;
 558         if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
 559                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 560 nfsmout:
 561         if (!error)
 562                 *dpp = ndp;
 563         else if (ndp != NULL)
 564                 FREE((caddr_t)ndp, M_NFSCLDELEG);
 565         mbuf_freem(nd->nd_mrep);
 566         return (error);
 567 }
 568
 569 /*
 570  * open downgrade rpc
 571  */
 572 APPLESTATIC int
 573 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
 574     struct ucred *cred, NFSPROC_T *p)
 575 {
 576         u_int32_t *tl;
 577         struct nfsrv_descript nfsd, *nd = &nfsd;
 578         int error;
 579
 580         NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
 581         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
 582         if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
 583                 *tl++ = 0;
 584         else
 585                 *tl++ = op->nfso_stateid.seqid;
 586         *tl++ = op->nfso_stateid.other[0];
 587         *tl++ = op->nfso_stateid.other[1];
 588         *tl++ = op->nfso_stateid.other[2];
 589         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 590         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
 591         *tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
 592         error = nfscl_request(nd, vp, p, cred, NULL);
 593         if (error)
 594                 return (error);
 595         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 596         if (!nd->nd_repstat) {
 597                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 598                 op->nfso_stateid.seqid = *tl++;
 599                 op->nfso_stateid.other[0] = *tl++;
 600                 op->nfso_stateid.other[1] = *tl++;
 601                 op->nfso_stateid.other[2] = *tl;
 602         }
 603         if (nd->nd_repstat && error == 0)
 604                 error = nd->nd_repstat;
 605         if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
 606                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 607 nfsmout:
 608         mbuf_freem(nd->nd_mrep);
 609         return (error);
 610 }
 611
 612 /*
 613  * V4 Close operation.
 614  */
 615 APPLESTATIC int
 616 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
 617 {
 618         struct nfsclclient *clp;
 619         int error;
 620
 621         if (vnode_vtype(vp) != VREG)
 622                 return (0);
 623         if (doclose)
 624                 error = nfscl_doclose(vp, &clp, p);
 625         else
 626                 error = nfscl_getclose(vp, &clp);
 627         if (error)
 628                 return (error);
 629
 630         nfscl_clientrelease(clp);
 631         return (0);
 632 }
 633
 634 /*
 635  * Close the open.
 636  */
 637 APPLESTATIC void
 638 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
 639 {
 640         struct nfsrv_descript nfsd, *nd = &nfsd;
 641         struct nfscllockowner *lp, *nlp;
 642         struct nfscllock *lop, *nlop;
 643         struct ucred *tcred;
 644         u_int64_t off = 0, len = 0;
 645         u_int32_t type = NFSV4LOCKT_READ;
 646         int error, do_unlock, trycnt;
 647
 648         tcred = newnfs_getcred();
 649         newnfs_copycred(&op->nfso_cred, tcred);
 650         /*
 651          * (Theoretically this could be done in the same
 652          *  compound as the close, but having multiple
 653          *  sequenced Ops in the same compound might be
 654          *  too scary for some servers.)
 655          */
 656         if (op->nfso_posixlock) {
 657                 off = 0;
 658                 len = NFS64BITSSET;
 659                 type = NFSV4LOCKT_READ;
 660         }
 661
 662         /*
 663          * Since this function is only called from VOP_INACTIVE(), no
 664          * other thread will be manipulating this Open. As such, the
 665          * lock lists are not being changed by other threads, so it should
 666          * be safe to do this without locking.
 667          */
 668         LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
 669                 do_unlock = 1;
 670                 LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
 671                         if (op->nfso_posixlock == 0) {
 672                                 off = lop->nfslo_first;
 673                                 len = lop->nfslo_end - lop->nfslo_first;
 674                                 if (lop->nfslo_type == F_WRLCK)
 675                                         type = NFSV4LOCKT_WRITE;
 676                                 else
 677                                         type = NFSV4LOCKT_READ;
 678                         }
 679                         if (do_unlock) {
 680                                 trycnt = 0;
 681                                 do {
 682                                         error = nfsrpc_locku(nd, nmp, lp, off,
 683                                             len, type, tcred, p, 0);
 684                                         if ((nd->nd_repstat == NFSERR_GRACE ||
 685                                             nd->nd_repstat == NFSERR_DELAY) &&
 686                                             error == 0)
 687                                                 (void) nfs_catnap(PZERO,
 688                                                     (int)nd->nd_repstat,
 689                                                     "nfs_close");
 690                                 } while ((nd->nd_repstat == NFSERR_GRACE ||
 691                                     nd->nd_repstat == NFSERR_DELAY) &&
 692                                     error == 0 && trycnt++ < 5);
 693                                 if (op->nfso_posixlock)
 694                                         do_unlock = 0;
 695                         }
 696                         nfscl_freelock(lop, 0);
 697                 }
 698                 /*
 699                  * Do a ReleaseLockOwner.
 700                  * The lock owner name nfsl_owner may be used by other opens for
 701                  * other files but the lock_owner4 name that nfsrpc_rellockown()
 702                  * puts on the wire has the file handle for this file appended
 703                  * to it, so it can be done now.
 704                  */
 705                 (void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
 706                     lp->nfsl_open->nfso_fhlen, tcred, p);
 707         }
 708
 709         /*
 710          * There could be other Opens for different files on the same
 711          * OpenOwner, so locking is required.
 712          */
 713         NFSLOCKCLSTATE();
 714         nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
 715         NFSUNLOCKCLSTATE();
 716         do {
 717                 error = nfscl_tryclose(op, tcred, nmp, p);
 718                 if (error == NFSERR_GRACE)
 719                         (void) nfs_catnap(PZERO, error, "nfs_close");
 720         } while (error == NFSERR_GRACE);
 721         NFSLOCKCLSTATE();
 722         nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
 723
 724         LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
 725                 nfscl_freelockowner(lp, 0);
 726         nfscl_freeopen(op, 0);
 727         NFSUNLOCKCLSTATE();
 728         NFSFREECRED(tcred);
 729 }
 730
 731 /*
 732  * The actual Close RPC.
 733  */
 734 APPLESTATIC int
 735 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
 736     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
 737     int syscred)
 738 {
 739         u_int32_t *tl;
 740         int error;
 741
 742         nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
 743             op->nfso_fhlen, NULL, NULL);
 744         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
 745         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 746         if (NFSHASNFSV4N(nmp))
 747                 *tl++ = 0;
 748         else
 749                 *tl++ = op->nfso_stateid.seqid;
 750         *tl++ = op->nfso_stateid.other[0];
 751         *tl++ = op->nfso_stateid.other[1];
 752         *tl = op->nfso_stateid.other[2];
 753         if (syscred)
 754                 nd->nd_flag |= ND_USEGSSNAME;
 755         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
 756             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 757         if (error)
 758                 return (error);
 759         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 760         if (nd->nd_repstat == 0)
 761                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 762         error = nd->nd_repstat;
 763         if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
 764                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 765 nfsmout:
 766         mbuf_freem(nd->nd_mrep);
 767         return (error);
 768 }
 769
 770 /*
 771  * V4 Open Confirm RPC.
 772  */
 773 APPLESTATIC int
 774 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
 775     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
 776 {
 777         u_int32_t *tl;
 778         struct nfsrv_descript nfsd, *nd = &nfsd;
 779         struct nfsmount *nmp;
 780         int error;
 781
 782         nmp = VFSTONFS(vnode_mount(vp));
 783         if (NFSHASNFSV4N(nmp))
 784                 return (0);             /* No confirmation for NFSv4.1. */
 785         nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL);
 786         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
 787         *tl++ = op->nfso_stateid.seqid;
 788         *tl++ = op->nfso_stateid.other[0];
 789         *tl++ = op->nfso_stateid.other[1];
 790         *tl++ = op->nfso_stateid.other[2];
 791         *tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
 792         error = nfscl_request(nd, vp, p, cred, NULL);
 793         if (error)
 794                 return (error);
 795         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 796         if (!nd->nd_repstat) {
 797                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 798                 op->nfso_stateid.seqid = *tl++;
 799                 op->nfso_stateid.other[0] = *tl++;
 800                 op->nfso_stateid.other[1] = *tl++;
 801                 op->nfso_stateid.other[2] = *tl;
 802         }
 803         error = nd->nd_repstat;
 804         if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
 805                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 806 nfsmout:
 807         mbuf_freem(nd->nd_mrep);
 808         return (error);
 809 }
 810
 811 /*
 812  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
 813  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
 814  */
 815 APPLESTATIC int
 816 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
 817     struct ucred *cred, NFSPROC_T *p)
 818 {
 819         u_int32_t *tl;
 820         struct nfsrv_descript nfsd;
 821         struct nfsrv_descript *nd = &nfsd;
 822         nfsattrbit_t attrbits;
 823         u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
 824         u_short port;
 825         int error, isinet6 = 0, callblen;
 826         nfsquad_t confirm;
 827         u_int32_t lease;
 828         static u_int32_t rev = 0;
 829         struct nfsclds *dsp, *ndsp, *tdsp;
 830
 831         if (nfsboottime.tv_sec == 0)
 832                 NFSSETBOOTTIME(nfsboottime);
 833         clp->nfsc_rev = rev++;
 834         if (NFSHASNFSV4N(nmp)) {
 835                 error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
 836                     NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
 837                 NFSCL_DEBUG(1, "aft exch=%d\n", error);
 838                 if (error == 0) {
 839                         error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
 840                             &nmp->nm_sockreq,
 841                             dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
 842                         if (error == 0) {
 843                                 NFSLOCKMNT(nmp);
 844                                 TAILQ_FOREACH_SAFE(tdsp, &nmp->nm_sess,
 845                                     nfsclds_list, ndsp)
 846                                         nfscl_freenfsclds(tdsp);
 847                                 TAILQ_INIT(&nmp->nm_sess);
 848                                 TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
 849                                     nfsclds_list);
 850                                 NFSUNLOCKMNT(nmp);
 851                         } else
 852                                 nfscl_freenfsclds(dsp);
 853                         NFSCL_DEBUG(1, "aft createsess=%d\n", error);
 854                 }
 855                 if (error == 0 && reclaim == 0) {
 856                         error = nfsrpc_reclaimcomplete(nmp, cred, p);
 857                         NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
 858                         if (error == NFSERR_COMPLETEALREADY ||
 859                             error == NFSERR_NOTSUPP)
 860                                 /* Ignore this error. */
 861                                 error = 0;
 862                 }
 863                 return (error);
 864         }
 865
 866         /*
 867          * Allocate a single session structure for NFSv4.0, because some of
 868          * the fields are used by NFSv4.0 although it doesn't do a session.
 869          */
 870         dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
 871         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
 872         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
 873         NFSLOCKMNT(nmp);
 874         TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
 875         NFSUNLOCKMNT(nmp);
 876
 877         nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL);
 878         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 879         *tl++ = txdr_unsigned(nfsboottime.tv_sec);
 880         *tl = txdr_unsigned(clp->nfsc_rev);
 881         (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
 882
 883         /*
 884          * set up the callback address
 885          */
 886         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 887         *tl = txdr_unsigned(NFS_CALLBCKPROG);
 888         callblen = strlen(nfsv4_callbackaddr);
 889         if (callblen == 0)
 890                 cp = nfscl_getmyip(nmp, &isinet6);
 891         if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
 892             (callblen > 0 || cp != NULL)) {
 893                 port = htons(nfsv4_cbport);
 894                 cp2 = (u_int8_t *)&port;
 895 #ifdef INET6
 896                 if ((callblen > 0 &&
 897                      strchr(nfsv4_callbackaddr, ':')) || isinet6) {
 898                         char ip6buf[INET6_ADDRSTRLEN], *ip6add;
 899
 900                         (void) nfsm_strtom(nd, "tcp6", 4);
 901                         if (callblen == 0) {
 902                                 ip6_sprintf(ip6buf, (struct in6_addr *)cp);
 903                                 ip6add = ip6buf;
 904                         } else {
 905                                 ip6add = nfsv4_callbackaddr;
 906                         }
 907                         snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
 908                             ip6add, cp2[0], cp2[1]);
 909                 } else
 910 #endif
 911                 {
 912                         (void) nfsm_strtom(nd, "tcp", 3);
 913                         if (callblen == 0)
 914                                 snprintf(addr, INET6_ADDRSTRLEN + 9,
 915                                     "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
 916                                     cp[2], cp[3], cp2[0], cp2[1]);
 917                         else
 918                                 snprintf(addr, INET6_ADDRSTRLEN + 9,
 919                                     "%s.%d.%d", nfsv4_callbackaddr,
 920                                     cp2[0], cp2[1]);
 921                 }
 922                 (void) nfsm_strtom(nd, addr, strlen(addr));
 923         } else {
 924                 (void) nfsm_strtom(nd, "tcp", 3);
 925                 (void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
 926         }
 927         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 928         *tl = txdr_unsigned(clp->nfsc_cbident);
 929         nd->nd_flag |= ND_USEGSSNAME;
 930         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
 931                 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 932         if (error)
 933                 return (error);
 934         if (nd->nd_repstat == 0) {
 935             NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
 936             NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0] = *tl++;
 937             NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1] = *tl++;
 938             confirm.lval[0] = *tl++;
 939             confirm.lval[1] = *tl;
 940             mbuf_freem(nd->nd_mrep);
 941             nd->nd_mrep = NULL;
 942
 943             /*
 944              * and confirm it.
 945              */
 946             nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
 947                 NULL);
 948             NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
 949             *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
 950             *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
 951             *tl++ = confirm.lval[0];
 952             *tl = confirm.lval[1];
 953             nd->nd_flag |= ND_USEGSSNAME;
 954             error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
 955                 cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 956             if (error)
 957                 return (error);
 958             mbuf_freem(nd->nd_mrep);
 959             nd->nd_mrep = NULL;
 960             if (nd->nd_repstat == 0) {
 961                 nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
 962                     nmp->nm_fhsize, NULL, NULL);
 963                 NFSZERO_ATTRBIT(&attrbits);
 964                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
 965                 (void) nfsrv_putattrbit(nd, &attrbits);
 966                 nd->nd_flag |= ND_USEGSSNAME;
 967                 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
 968                     cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 969                 if (error)
 970                     return (error);
 971                 if (nd->nd_repstat == 0) {
 972                     error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
 973                         NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
 974                     if (error)
 975                         goto nfsmout;
 976                     clp->nfsc_renew = NFSCL_RENEW(lease);
 977                     clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
 978                     clp->nfsc_clientidrev++;
 979                     if (clp->nfsc_clientidrev == 0)
 980                         clp->nfsc_clientidrev++;
 981                 }
 982             }
 983         }
 984         error = nd->nd_repstat;
 985 nfsmout:
 986         mbuf_freem(nd->nd_mrep);
 987         return (error);
 988 }
 989
 990 /*
 991  * nfs getattr call.
 992  */
 993 APPLESTATIC int
 994 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
 995     struct nfsvattr *nap, void *stuff)
 996 {
 997         struct nfsrv_descript nfsd, *nd = &nfsd;
 998         int error;
 999         nfsattrbit_t attrbits;
1000
1001         NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1002         if (nd->nd_flag & ND_NFSV4) {
1003                 NFSGETATTR_ATTRBIT(&attrbits);
1004                 (void) nfsrv_putattrbit(nd, &attrbits);
1005         }
1006         error = nfscl_request(nd, vp, p, cred, stuff);
1007         if (error)
1008                 return (error);
1009         if (!nd->nd_repstat)
1010                 error = nfsm_loadattr(nd, nap);
1011         else
1012                 error = nd->nd_repstat;
1013         mbuf_freem(nd->nd_mrep);
1014         return (error);
1015 }
1016
1017 /*
1018  * nfs getattr call with non-vnode arguemnts.
1019  */
1020 APPLESTATIC int
1021 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1022     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1023     uint32_t *leasep)
1024 {
1025         struct nfsrv_descript nfsd, *nd = &nfsd;
1026         int error, vers = NFS_VER2;
1027         nfsattrbit_t attrbits;
1028
1029         nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL);
1030         if (nd->nd_flag & ND_NFSV4) {
1031                 vers = NFS_VER4;
1032                 NFSGETATTR_ATTRBIT(&attrbits);
1033                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1034                 (void) nfsrv_putattrbit(nd, &attrbits);
1035         } else if (nd->nd_flag & ND_NFSV3) {
1036                 vers = NFS_VER3;
1037         }
1038         if (syscred)
1039                 nd->nd_flag |= ND_USEGSSNAME;
1040         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1041             NFS_PROG, vers, NULL, 1, xidp, NULL);
1042         if (error)
1043                 return (error);
1044         if (nd->nd_repstat == 0) {
1045                 if ((nd->nd_flag & ND_NFSV4) != 0)
1046                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1047                             NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1048                             NULL, NULL);
1049                 else
1050                         error = nfsm_loadattr(nd, nap);
1051         } else
1052                 error = nd->nd_repstat;
1053         mbuf_freem(nd->nd_mrep);
1054         return (error);
1055 }
1056
1057 /*
1058  * Do an nfs setattr operation.
1059  */
1060 APPLESTATIC int
1061 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1062     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1063     void *stuff)
1064 {
1065         int error, expireret = 0, openerr, retrycnt;
1066         u_int32_t clidrev = 0, mode;
1067         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1068         struct nfsfh *nfhp;
1069         nfsv4stateid_t stateid;
1070         void *lckp;
1071
1072         if (nmp->nm_clp != NULL)
1073                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1074         if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1075                 mode = NFSV4OPEN_ACCESSWRITE;
1076         else
1077                 mode = NFSV4OPEN_ACCESSREAD;
1078         retrycnt = 0;
1079         do {
1080                 lckp = NULL;
1081                 openerr = 1;
1082                 if (NFSHASNFSV4(nmp)) {
1083                         nfhp = VTONFS(vp)->n_fhp;
1084                         error = nfscl_getstateid(vp, nfhp->nfh_fh,
1085                             nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1086                         if (error && vnode_vtype(vp) == VREG &&
1087                             (mode == NFSV4OPEN_ACCESSWRITE ||
1088                              nfstest_openallsetattr)) {
1089                                 /*
1090                                  * No Open stateid, so try and open the file
1091                                  * now.
1092                                  */
1093                                 if (mode == NFSV4OPEN_ACCESSWRITE)
1094                                         openerr = nfsrpc_open(vp, FWRITE, cred,
1095                                             p);
1096                                 else
1097                                         openerr = nfsrpc_open(vp, FREAD, cred,
1098                                             p);
1099                                 if (!openerr)
1100                                         (void) nfscl_getstateid(vp,
1101                                             nfhp->nfh_fh, nfhp->nfh_len,
1102                                             mode, 0, cred, p, &stateid, &lckp);
1103                         }
1104                 }
1105                 if (vap != NULL)
1106                         error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1107                             rnap, attrflagp, stuff);
1108                 else
1109                         error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1110                             stuff);
1111                 if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1112                         nfscl_initiate_recovery(nmp->nm_clp);
1113                 if (lckp != NULL)
1114                         nfscl_lockderef(lckp);
1115                 if (!openerr)
1116                         (void) nfsrpc_close(vp, 0, p);
1117                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1118                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1119                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1120                         (void) nfs_catnap(PZERO, error, "nfs_setattr");
1121                 } else if ((error == NFSERR_EXPIRED ||
1122                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1123                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1124                 }
1125                 retrycnt++;
1126         } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1127             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1128             error == NFSERR_BADSESSION ||
1129             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1130             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1131              expireret == 0 && clidrev != 0 && retrycnt < 4));
1132         if (error && retrycnt >= 4)
1133                 error = EIO;
1134         return (error);
1135 }
1136
1137 static int
1138 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1139     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1140     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1141 {
1142         u_int32_t *tl;
1143         struct nfsrv_descript nfsd, *nd = &nfsd;
1144         int error;
1145         nfsattrbit_t attrbits;
1146
1147         *attrflagp = 0;
1148         NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1149         if (nd->nd_flag & ND_NFSV4)
1150                 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1151         vap->va_type = vnode_vtype(vp);
1152         nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1153         if (nd->nd_flag & ND_NFSV3) {
1154                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1155                 *tl = newnfs_false;
1156         } else if (nd->nd_flag & ND_NFSV4) {
1157                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1158                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1159                 NFSGETATTR_ATTRBIT(&attrbits);
1160                 (void) nfsrv_putattrbit(nd, &attrbits);
1161         }
1162         error = nfscl_request(nd, vp, p, cred, stuff);
1163         if (error)
1164                 return (error);
1165         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1166                 error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1167         if ((nd->nd_flag & ND_NFSV4) && !error)
1168                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1169         if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1170                 error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1171         mbuf_freem(nd->nd_mrep);
1172         if (nd->nd_repstat && !error)
1173                 error = nd->nd_repstat;
1174         return (error);
1175 }
1176
1177 /*
1178  * nfs lookup rpc
1179  */
1180 APPLESTATIC int
1181 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1182     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1183     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1184 {
1185         u_int32_t *tl;
1186         struct nfsrv_descript nfsd, *nd = &nfsd;
1187         struct nfsmount *nmp;
1188         struct nfsnode *np;
1189         struct nfsfh *nfhp;
1190         nfsattrbit_t attrbits;
1191         int error = 0, lookupp = 0;
1192
1193         *attrflagp = 0;
1194         *dattrflagp = 0;
1195         if (vnode_vtype(dvp) != VDIR)
1196                 return (ENOTDIR);
1197         nmp = VFSTONFS(vnode_mount(dvp));
1198         if (len > NFS_MAXNAMLEN)
1199                 return (ENAMETOOLONG);
1200         if (NFSHASNFSV4(nmp) && len == 1 &&
1201                 name[0] == '.') {
1202                 /*
1203                  * Just return the current dir's fh.
1204                  */
1205                 np = VTONFS(dvp);
1206                 MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1207                         np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1208                 nfhp->nfh_len = np->n_fhp->nfh_len;
1209                 NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1210                 *nfhpp = nfhp;
1211                 return (0);
1212         }
1213         if (NFSHASNFSV4(nmp) && len == 2 &&
1214                 name[0] == '.' && name[1] == '.') {
1215                 lookupp = 1;
1216                 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1217         } else {
1218                 NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1219                 (void) nfsm_strtom(nd, name, len);
1220         }
1221         if (nd->nd_flag & ND_NFSV4) {
1222                 NFSGETATTR_ATTRBIT(&attrbits);
1223                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1224                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1225                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1226                 (void) nfsrv_putattrbit(nd, &attrbits);
1227         }
1228         error = nfscl_request(nd, dvp, p, cred, stuff);
1229         if (error)
1230                 return (error);
1231         if (nd->nd_repstat) {
1232                 /*
1233                  * When an NFSv4 Lookupp returns ENOENT, it means that
1234                  * the lookup is at the root of an fs, so return this dir.
1235                  */
1236                 if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1237                     np = VTONFS(dvp);
1238                     MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1239                         np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1240                     nfhp->nfh_len = np->n_fhp->nfh_len;
1241                     NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1242                     *nfhpp = nfhp;
1243                     mbuf_freem(nd->nd_mrep);
1244                     return (0);
1245                 }
1246                 if (nd->nd_flag & ND_NFSV3)
1247                     error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1248                 else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1249                     ND_NFSV4) {
1250                         /* Load the directory attributes. */
1251                         error = nfsm_loadattr(nd, dnap);
1252                         if (error == 0)
1253                                 *dattrflagp = 1;
1254                 }
1255                 goto nfsmout;
1256         }
1257         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1258                 /* Load the directory attributes. */
1259                 error = nfsm_loadattr(nd, dnap);
1260                 if (error != 0)
1261                         goto nfsmout;
1262                 *dattrflagp = 1;
1263                 /* Skip over the Lookup and GetFH operation status values. */
1264                 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1265         }
1266         error = nfsm_getfh(nd, nfhpp);
1267         if (error)
1268                 goto nfsmout;
1269
1270         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1271         if ((nd->nd_flag & ND_NFSV3) && !error)
1272                 error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1273 nfsmout:
1274         mbuf_freem(nd->nd_mrep);
1275         if (!error && nd->nd_repstat)
1276                 error = nd->nd_repstat;
1277         return (error);
1278 }
1279
1280 /*
1281  * Do a readlink rpc.
1282  */
1283 APPLESTATIC int
1284 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1285     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1286 {
1287         u_int32_t *tl;
1288         struct nfsrv_descript nfsd, *nd = &nfsd;
1289         struct nfsnode *np = VTONFS(vp);
1290         nfsattrbit_t attrbits;
1291         int error, len, cangetattr = 1;
1292
1293         *attrflagp = 0;
1294         NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1295         if (nd->nd_flag & ND_NFSV4) {
1296                 /*
1297                  * And do a Getattr op.
1298                  */
1299                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1300                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1301                 NFSGETATTR_ATTRBIT(&attrbits);
1302                 (void) nfsrv_putattrbit(nd, &attrbits);
1303         }
1304         error = nfscl_request(nd, vp, p, cred, stuff);
1305         if (error)
1306                 return (error);
1307         if (nd->nd_flag & ND_NFSV3)
1308                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1309         if (!nd->nd_repstat && !error) {
1310                 NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1311                 /*
1312                  * This seems weird to me, but must have been added to
1313                  * FreeBSD for some reason. The only thing I can think of
1314                  * is that there was/is some server that replies with
1315                  * more link data than it should?
1316                  */
1317                 if (len == NFS_MAXPATHLEN) {
1318                         NFSLOCKNODE(np);
1319                         if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1320                                 len = np->n_size;
1321                                 cangetattr = 0;
1322                         }
1323                         NFSUNLOCKNODE(np);
1324                 }
1325                 error = nfsm_mbufuio(nd, uiop, len);
1326                 if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1327                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1328         }
1329         if (nd->nd_repstat && !error)
1330                 error = nd->nd_repstat;
1331 nfsmout:
1332         mbuf_freem(nd->nd_mrep);
1333         return (error);
1334 }
1335
1336 /*
1337  * Read operation.
1338  */
1339 APPLESTATIC int
1340 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1341     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1342 {
1343         int error, expireret = 0, retrycnt;
1344         u_int32_t clidrev = 0;
1345         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1346         struct nfsnode *np = VTONFS(vp);
1347         struct ucred *newcred;
1348         struct nfsfh *nfhp = NULL;
1349         nfsv4stateid_t stateid;
1350         void *lckp;
1351
1352         if (nmp->nm_clp != NULL)
1353                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1354         newcred = cred;
1355         if (NFSHASNFSV4(nmp)) {
1356                 nfhp = np->n_fhp;
1357                 newcred = NFSNEWCRED(cred);
1358         }
1359         retrycnt = 0;
1360         do {
1361                 lckp = NULL;
1362                 if (NFSHASNFSV4(nmp))
1363                         (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1364                             NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1365                             &lckp);
1366                 error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1367                     attrflagp, stuff);
1368                 if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1369                         nfscl_initiate_recovery(nmp->nm_clp);
1370                 if (lckp != NULL)
1371                         nfscl_lockderef(lckp);
1372                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1373                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1374                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1375                         (void) nfs_catnap(PZERO, error, "nfs_read");
1376                 } else if ((error == NFSERR_EXPIRED ||
1377                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1378                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1379                 }
1380                 retrycnt++;
1381         } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1382             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1383             error == NFSERR_BADSESSION ||
1384             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1385             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1386              expireret == 0 && clidrev != 0 && retrycnt < 4));
1387         if (error && retrycnt >= 4)
1388                 error = EIO;
1389         if (NFSHASNFSV4(nmp))
1390                 NFSFREECRED(newcred);
1391         return (error);
1392 }
1393
1394 /*
1395  * The actual read RPC.
1396  */
1397 static int
1398 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1399     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1400     int *attrflagp, void *stuff)
1401 {
1402         u_int32_t *tl;
1403         int error = 0, len, retlen, tsiz, eof = 0;
1404         struct nfsrv_descript nfsd;
1405         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1406         struct nfsrv_descript *nd = &nfsd;
1407         int rsize;
1408         off_t tmp_off;
1409
1410         *attrflagp = 0;
1411         tsiz = uio_uio_resid(uiop);
1412         tmp_off = uiop->uio_offset + tsiz;
1413         NFSLOCKMNT(nmp);
1414         if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1415                 NFSUNLOCKMNT(nmp);
1416                 return (EFBIG);
1417         }
1418         rsize = nmp->nm_rsize;
1419         NFSUNLOCKMNT(nmp);
1420         nd->nd_mrep = NULL;
1421         while (tsiz > 0) {
1422                 *attrflagp = 0;
1423                 len = (tsiz > rsize) ? rsize : tsiz;
1424                 NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1425                 if (nd->nd_flag & ND_NFSV4)
1426                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1427                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1428                 if (nd->nd_flag & ND_NFSV2) {
1429                         *tl++ = txdr_unsigned(uiop->uio_offset);
1430                         *tl++ = txdr_unsigned(len);
1431                         *tl = 0;
1432                 } else {
1433                         txdr_hyper(uiop->uio_offset, tl);
1434                         *(tl + 2) = txdr_unsigned(len);
1435                 }
1436                 /*
1437                  * Since I can't do a Getattr for NFSv4 for Write, there
1438                  * doesn't seem any point in doing one here, either.
1439                  * (See the comment in nfsrpc_writerpc() for more info.)
1440                  */
1441                 error = nfscl_request(nd, vp, p, cred, stuff);
1442                 if (error)
1443                         return (error);
1444                 if (nd->nd_flag & ND_NFSV3) {
1445                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1446                 } else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1447                         error = nfsm_loadattr(nd, nap);
1448                         if (!error)
1449                                 *attrflagp = 1;
1450                 }
1451                 if (nd->nd_repstat || error) {
1452                         if (!error)
1453                                 error = nd->nd_repstat;
1454                         goto nfsmout;
1455                 }
1456                 if (nd->nd_flag & ND_NFSV3) {
1457                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1458                         eof = fxdr_unsigned(int, *(tl + 1));
1459                 } else if (nd->nd_flag & ND_NFSV4) {
1460                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1461                         eof = fxdr_unsigned(int, *tl);
1462                 }
1463                 NFSM_STRSIZ(retlen, len);
1464                 error = nfsm_mbufuio(nd, uiop, retlen);
1465                 if (error)
1466                         goto nfsmout;
1467                 mbuf_freem(nd->nd_mrep);
1468                 nd->nd_mrep = NULL;
1469                 tsiz -= retlen;
1470                 if (!(nd->nd_flag & ND_NFSV2)) {
1471                         if (eof || retlen == 0)
1472                                 tsiz = 0;
1473                 } else if (retlen < len)
1474                         tsiz = 0;
1475         }
1476         return (0);
1477 nfsmout:
1478         if (nd->nd_mrep != NULL)
1479                 mbuf_freem(nd->nd_mrep);
1480         return (error);
1481 }
1482
1483 /*
1484  * nfs write operation
1485  * When called_from_strategy != 0, it should return EIO for an error that
1486  * indicates recovery is in progress, so that the buffer will be left
1487  * dirty and be written back to the server later. If it loops around,
1488  * the recovery thread could get stuck waiting for the buffer and recovery
1489  * will then deadlock.
1490  */
1491 APPLESTATIC int
1492 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1493     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1494     void *stuff, int called_from_strategy)
1495 {
1496         int error, expireret = 0, retrycnt, nostateid;
1497         u_int32_t clidrev = 0;
1498         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1499         struct nfsnode *np = VTONFS(vp);
1500         struct ucred *newcred;
1501         struct nfsfh *nfhp = NULL;
1502         nfsv4stateid_t stateid;
1503         void *lckp;
1504
1505         *must_commit = 0;
1506         if (nmp->nm_clp != NULL)
1507                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1508         newcred = cred;
1509         if (NFSHASNFSV4(nmp)) {
1510                 newcred = NFSNEWCRED(cred);
1511                 nfhp = np->n_fhp;
1512         }
1513         retrycnt = 0;
1514         do {
1515                 lckp = NULL;
1516                 nostateid = 0;
1517                 if (NFSHASNFSV4(nmp)) {
1518                         (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1519                             NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1520                             &lckp);
1521                         if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1522                             stateid.other[2] == 0) {
1523                                 nostateid = 1;
1524                                 NFSCL_DEBUG(1, "stateid0 in write\n");
1525                         }
1526                 }
1527
1528                 /*
1529                  * If there is no stateid for NFSv4, it means this is an
1530                  * extraneous write after close. Basically a poorly
1531                  * implemented buffer cache. Just don't do the write.
1532                  */
1533                 if (nostateid)
1534                         error = 0;
1535                 else
1536                         error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1537                             newcred, &stateid, p, nap, attrflagp, stuff);
1538                 if (error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION)
1539                         nfscl_initiate_recovery(nmp->nm_clp);
1540                 if (lckp != NULL)
1541                         nfscl_lockderef(lckp);
1542                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1543                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1544                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1545                         (void) nfs_catnap(PZERO, error, "nfs_write");
1546                 } else if ((error == NFSERR_EXPIRED ||
1547                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1548                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1549                 }
1550                 retrycnt++;
1551         } while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1552             ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1553               error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1554             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1555             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1556              expireret == 0 && clidrev != 0 && retrycnt < 4));
1557         if (error != 0 && (retrycnt >= 4 ||
1558             ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1559               error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1560                 error = EIO;
1561         if (NFSHASNFSV4(nmp))
1562                 NFSFREECRED(newcred);
1563         return (error);
1564 }
1565
1566 /*
1567  * The actual write RPC.
1568  */
1569 static int
1570 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1571     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1572     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1573 {
1574         u_int32_t *tl;
1575         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1576         struct nfsnode *np = VTONFS(vp);
1577         int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1578         int wccflag = 0, wsize;
1579         int32_t backup;
1580         struct nfsrv_descript nfsd;
1581         struct nfsrv_descript *nd = &nfsd;
1582         nfsattrbit_t attrbits;
1583         off_t tmp_off;
1584
1585         KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1586         *attrflagp = 0;
1587         tsiz = uio_uio_resid(uiop);
1588         tmp_off = uiop->uio_offset + tsiz;
1589         NFSLOCKMNT(nmp);
1590         if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1591                 NFSUNLOCKMNT(nmp);
1592                 return (EFBIG);
1593         }
1594         wsize = nmp->nm_wsize;
1595         NFSUNLOCKMNT(nmp);
1596         nd->nd_mrep = NULL;     /* NFSv2 sometimes does a write with */
1597         nd->nd_repstat = 0;     /* uio_resid == 0, so the while is not done */
1598         while (tsiz > 0) {
1599                 *attrflagp = 0;
1600                 len = (tsiz > wsize) ? wsize : tsiz;
1601                 NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1602                 if (nd->nd_flag & ND_NFSV4) {
1603                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1604                         NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1605                         txdr_hyper(uiop->uio_offset, tl);
1606                         tl += 2;
1607                         *tl++ = txdr_unsigned(*iomode);
1608                         *tl = txdr_unsigned(len);
1609                 } else if (nd->nd_flag & ND_NFSV3) {
1610                         NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1611                         txdr_hyper(uiop->uio_offset, tl);
1612                         tl += 2;
1613                         *tl++ = txdr_unsigned(len);
1614                         *tl++ = txdr_unsigned(*iomode);
1615                         *tl = txdr_unsigned(len);
1616                 } else {
1617                         u_int32_t x;
1618
1619                         NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1620                         /*
1621                          * Not sure why someone changed this, since the
1622                          * RFC clearly states that "beginoffset" and
1623                          * "totalcount" are ignored, but it wouldn't
1624                          * surprise me if there's a busted server out there.
1625                          */
1626                         /* Set both "begin" and "current" to non-garbage. */
1627                         x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1628                         *tl++ = x;      /* "begin offset" */
1629                         *tl++ = x;      /* "current offset" */
1630                         x = txdr_unsigned(len);
1631                         *tl++ = x;      /* total to this offset */
1632                         *tl = x;        /* size of this write */
1633
1634                 }
1635                 nfsm_uiombuf(nd, uiop, len);
1636                 /*
1637                  * Although it is tempting to do a normal Getattr Op in the
1638                  * NFSv4 compound, the result can be a nearly hung client
1639                  * system if the Getattr asks for Owner and/or OwnerGroup.
1640                  * It occurs when the client can't map either the Owner or
1641                  * Owner_group name in the Getattr reply to a uid/gid. When
1642                  * there is a cache miss, the kernel does an upcall to the
1643                  * nfsuserd. Then, it can try and read the local /etc/passwd
1644                  * or /etc/group file. It can then block in getnewbuf(),
1645                  * waiting for dirty writes to be pushed to the NFS server.
1646                  * The only reason this doesn't result in a complete
1647                  * deadlock, is that the upcall times out and allows
1648                  * the write to complete. However, progress is so slow
1649                  * that it might just as well be deadlocked.
1650                  * As such, we get the rest of the attributes, but not
1651                  * Owner or Owner_group.
1652                  * nb: nfscl_loadattrcache() needs to be told that these
1653                  *     partial attributes from a write rpc are being
1654                  *     passed in, via a argument flag.
1655                  */
1656                 if (nd->nd_flag & ND_NFSV4) {
1657                         NFSWRITEGETATTR_ATTRBIT(&attrbits);
1658                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1659                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
1660                         (void) nfsrv_putattrbit(nd, &attrbits);
1661                 }
1662                 error = nfscl_request(nd, vp, p, cred, stuff);
1663                 if (error)
1664                         return (error);
1665                 if (nd->nd_repstat) {
1666                         /*
1667                          * In case the rpc gets retried, roll
1668                          * the uio fileds changed by nfsm_uiombuf()
1669                          * back.
1670                          */
1671                         uiop->uio_offset -= len;
1672                         uio_uio_resid_add(uiop, len);
1673                         uio_iov_base_add(uiop, -len);
1674                         uio_iov_len_add(uiop, len);
1675                 }
1676                 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1677                         error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1678                             &wccflag, stuff);
1679                         if (error)
1680                                 goto nfsmout;
1681                 }
1682                 if (!nd->nd_repstat) {
1683                         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1684                                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1685                                         + NFSX_VERF);
1686                                 rlen = fxdr_unsigned(int, *tl++);
1687                                 if (rlen == 0) {
1688                                         error = NFSERR_IO;
1689                                         goto nfsmout;
1690                                 } else if (rlen < len) {
1691                                         backup = len - rlen;
1692                                         uio_iov_base_add(uiop, -(backup));
1693                                         uio_iov_len_add(uiop, backup);
1694                                         uiop->uio_offset -= backup;
1695                                         uio_uio_resid_add(uiop, backup);
1696                                         len = rlen;
1697                                 }
1698                                 commit = fxdr_unsigned(int, *tl++);
1699
1700                                 /*
1701                                  * Return the lowest committment level
1702                                  * obtained by any of the RPCs.
1703                                  */
1704                                 if (committed == NFSWRITE_FILESYNC)
1705                                         committed = commit;
1706                                 else if (committed == NFSWRITE_DATASYNC &&
1707                                         commit == NFSWRITE_UNSTABLE)
1708                                         committed = commit;
1709                                 NFSLOCKMNT(nmp);
1710                                 if (!NFSHASWRITEVERF(nmp)) {
1711                                         NFSBCOPY((caddr_t)tl,
1712                                             (caddr_t)&nmp->nm_verf[0],
1713                                             NFSX_VERF);
1714                                         NFSSETWRITEVERF(nmp);
1715                                 } else if (NFSBCMP(tl, nmp->nm_verf,
1716                                     NFSX_VERF)) {
1717                                         *must_commit = 1;
1718                                         NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1719                                 }
1720                                 NFSUNLOCKMNT(nmp);
1721                         }
1722                         if (nd->nd_flag & ND_NFSV4)
1723                                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1724                         if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1725                                 error = nfsm_loadattr(nd, nap);
1726                                 if (!error)
1727                                         *attrflagp = NFS_LATTR_NOSHRINK;
1728                         }
1729                 } else {
1730                         error = nd->nd_repstat;
1731                 }
1732                 if (error)
1733                         goto nfsmout;
1734                 NFSWRITERPC_SETTIME(wccflag, np, (nd->nd_flag & ND_NFSV4));
1735                 mbuf_freem(nd->nd_mrep);
1736                 nd->nd_mrep = NULL;
1737                 tsiz -= len;
1738         }
1739 nfsmout:
1740         if (nd->nd_mrep != NULL)
1741                 mbuf_freem(nd->nd_mrep);
1742         *iomode = committed;
1743         if (nd->nd_repstat && !error)
1744                 error = nd->nd_repstat;
1745         return (error);
1746 }
1747
1748 /*
1749  * nfs mknod rpc
1750  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1751  * mode set to specify the file type and the size field for rdev.
1752  */
1753 APPLESTATIC int
1754 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1755     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1756     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1757     int *attrflagp, int *dattrflagp, void *dstuff)
1758 {
1759         u_int32_t *tl;
1760         int error = 0;
1761         struct nfsrv_descript nfsd, *nd = &nfsd;
1762         nfsattrbit_t attrbits;
1763
1764         *nfhpp = NULL;
1765         *attrflagp = 0;
1766         *dattrflagp = 0;
1767         if (namelen > NFS_MAXNAMLEN)
1768                 return (ENAMETOOLONG);
1769         NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1770         if (nd->nd_flag & ND_NFSV4) {
1771                 if (vtyp == VBLK || vtyp == VCHR) {
1772                         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1773                         *tl++ = vtonfsv34_type(vtyp);
1774                         *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1775                         *tl = txdr_unsigned(NFSMINOR(rdev));
1776                 } else {
1777                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1778                         *tl = vtonfsv34_type(vtyp);
1779                 }
1780         }
1781         (void) nfsm_strtom(nd, name, namelen);
1782         if (nd->nd_flag & ND_NFSV3) {
1783                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1784                 *tl = vtonfsv34_type(vtyp);
1785         }
1786         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1787                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
1788         if ((nd->nd_flag & ND_NFSV3) &&
1789             (vtyp == VCHR || vtyp == VBLK)) {
1790                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1791                 *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1792                 *tl = txdr_unsigned(NFSMINOR(rdev));
1793         }
1794         if (nd->nd_flag & ND_NFSV4) {
1795                 NFSGETATTR_ATTRBIT(&attrbits);
1796                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1797                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1798                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1799                 (void) nfsrv_putattrbit(nd, &attrbits);
1800         }
1801         if (nd->nd_flag & ND_NFSV2)
1802                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1803         error = nfscl_request(nd, dvp, p, cred, dstuff);
1804         if (error)
1805                 return (error);
1806         if (nd->nd_flag & ND_NFSV4)
1807                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1808         if (!nd->nd_repstat) {
1809                 if (nd->nd_flag & ND_NFSV4) {
1810                         NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1811                         error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1812                         if (error)
1813                                 goto nfsmout;
1814                 }
1815                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1816                 if (error)
1817                         goto nfsmout;
1818         }
1819         if (nd->nd_flag & ND_NFSV3)
1820                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1821         if (!error && nd->nd_repstat)
1822                 error = nd->nd_repstat;
1823 nfsmout:
1824         mbuf_freem(nd->nd_mrep);
1825         return (error);
1826 }
1827
1828 /*
1829  * nfs file create call
1830  * Mostly just call the approriate routine. (I separated out v4, so that
1831  * error recovery wouldn't be as difficult.)
1832  */
1833 APPLESTATIC int
1834 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1835     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1836     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1837     int *attrflagp, int *dattrflagp, void *dstuff)
1838 {
1839         int error = 0, newone, expireret = 0, retrycnt, unlocked;
1840         struct nfsclowner *owp;
1841         struct nfscldeleg *dp;
1842         struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1843         u_int32_t clidrev;
1844
1845         if (NFSHASNFSV4(nmp)) {
1846             retrycnt = 0;
1847             do {
1848                 dp = NULL;
1849                 error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1850                     NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1851                     NULL, 1);
1852                 if (error)
1853                         return (error);
1854                 if (nmp->nm_clp != NULL)
1855                         clidrev = nmp->nm_clp->nfsc_clientidrev;
1856                 else
1857                         clidrev = 0;
1858                 error = nfsrpc_createv4(dvp, name, namelen, vap, cverf, fmode,
1859                   owp, &dp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1860                   dstuff, &unlocked);
1861                 /*
1862                  * There is no need to invalidate cached attributes here,
1863                  * since new post-delegation issue attributes are always
1864                  * returned by nfsrpc_createv4() and these will update the
1865                  * attribute cache.
1866                  */
1867                 if (dp != NULL)
1868                         (void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
1869                             (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
1870                 nfscl_ownerrelease(owp, error, newone, unlocked);
1871                 if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1872                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1873                     error == NFSERR_BADSESSION) {
1874                         (void) nfs_catnap(PZERO, error, "nfs_open");
1875                 } else if ((error == NFSERR_EXPIRED ||
1876                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1877                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1878                         retrycnt++;
1879                 }
1880             } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1881                 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1882                 error == NFSERR_BADSESSION ||
1883                 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1884                  expireret == 0 && clidrev != 0 && retrycnt < 4));
1885             if (error && retrycnt >= 4)
1886                     error = EIO;
1887         } else {
1888                 error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
1889                     fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1890                     dstuff);
1891         }
1892         return (error);
1893 }
1894
1895 /*
1896  * The create rpc for v2 and 3.
1897  */
1898 static int
1899 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1900     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1901     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1902     int *attrflagp, int *dattrflagp, void *dstuff)
1903 {
1904         u_int32_t *tl;
1905         int error = 0;
1906         struct nfsrv_descript nfsd, *nd = &nfsd;
1907
1908         *nfhpp = NULL;
1909         *attrflagp = 0;
1910         *dattrflagp = 0;
1911         if (namelen > NFS_MAXNAMLEN)
1912                 return (ENAMETOOLONG);
1913         NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
1914         (void) nfsm_strtom(nd, name, namelen);
1915         if (nd->nd_flag & ND_NFSV3) {
1916                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1917                 if (fmode & O_EXCL) {
1918                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
1919                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
1920                         *tl++ = cverf.lval[0];
1921                         *tl = cverf.lval[1];
1922                 } else {
1923                         *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
1924                         nfscl_fillsattr(nd, vap, dvp, 0, 0);
1925                 }
1926         } else {
1927                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
1928         }
1929         error = nfscl_request(nd, dvp, p, cred, dstuff);
1930         if (error)
1931                 return (error);
1932         if (nd->nd_repstat == 0) {
1933                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1934                 if (error)
1935                         goto nfsmout;
1936         }
1937         if (nd->nd_flag & ND_NFSV3)
1938                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1939         if (nd->nd_repstat != 0 && error == 0)
1940                 error = nd->nd_repstat;
1941 nfsmout:
1942         mbuf_freem(nd->nd_mrep);
1943         return (error);
1944 }
1945
1946 static int
1947 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1948     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
1949     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
1950     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
1951     int *dattrflagp, void *dstuff, int *unlockedp)
1952 {
1953         u_int32_t *tl;
1954         int error = 0, deleg, newone, ret, acesize, limitby;
1955         struct nfsrv_descript nfsd, *nd = &nfsd;
1956         struct nfsclopen *op;
1957         struct nfscldeleg *dp = NULL;
1958         struct nfsnode *np;
1959         struct nfsfh *nfhp;
1960         nfsattrbit_t attrbits;
1961         nfsv4stateid_t stateid;
1962         u_int32_t rflags;
1963         struct nfsmount *nmp;
1964
1965         nmp = VFSTONFS(dvp->v_mount);
1966         np = VTONFS(dvp);
1967         *unlockedp = 0;
1968         *nfhpp = NULL;
1969         *dpp = NULL;
1970         *attrflagp = 0;
1971         *dattrflagp = 0;
1972         if (namelen > NFS_MAXNAMLEN)
1973                 return (ENAMETOOLONG);
1974         NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
1975         /*
1976          * For V4, this is actually an Open op.
1977          */
1978         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1979         *tl++ = txdr_unsigned(owp->nfsow_seqid);
1980         *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
1981             NFSV4OPEN_ACCESSREAD);
1982         *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1983         *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
1984         *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
1985         (void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
1986         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1987         *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
1988         if (fmode & O_EXCL) {
1989                 if (NFSHASNFSV4N(nmp)) {
1990                         if (NFSHASSESSPERSIST(nmp)) {
1991                                 /* Use GUARDED for persistent sessions. */
1992                                 *tl = txdr_unsigned(NFSCREATE_GUARDED);
1993                                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
1994                         } else {
1995                                 /* Otherwise, use EXCLUSIVE4_1. */
1996                                 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
1997                                 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
1998                                 *tl++ = cverf.lval[0];
1999                                 *tl = cverf.lval[1];
2000                                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2001                         }
2002                 } else {
2003                         /* NFSv4.0 */
2004                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2005                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2006                         *tl++ = cverf.lval[0];
2007                         *tl = cverf.lval[1];
2008                 }
2009         } else {
2010                 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2011                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2012         }
2013         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2014         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2015         (void) nfsm_strtom(nd, name, namelen);
2016         /* Get the new file's handle and attributes. */
2017         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2018         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2019         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2020         NFSGETATTR_ATTRBIT(&attrbits);
2021         (void) nfsrv_putattrbit(nd, &attrbits);
2022         /* Get the directory's post-op attributes. */
2023         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2024         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2025         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2026         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2027         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2028         (void) nfsrv_putattrbit(nd, &attrbits);
2029         error = nfscl_request(nd, dvp, p, cred, dstuff);
2030         if (error)
2031                 return (error);
2032         NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2033         if (nd->nd_repstat == 0) {
2034                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2035                     6 * NFSX_UNSIGNED);
2036                 stateid.seqid = *tl++;
2037                 stateid.other[0] = *tl++;
2038                 stateid.other[1] = *tl++;
2039                 stateid.other[2] = *tl;
2040                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2041                 (void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2042                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2043                 deleg = fxdr_unsigned(int, *tl);
2044                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
2045                     deleg == NFSV4OPEN_DELEGATEWRITE) {
2046                         if (!(owp->nfsow_clp->nfsc_flags &
2047                               NFSCLFLAGS_FIRSTDELEG))
2048                                 owp->nfsow_clp->nfsc_flags |=
2049                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2050                         MALLOC(dp, struct nfscldeleg *,
2051                             sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2052                             M_NFSCLDELEG, M_WAITOK);
2053                         LIST_INIT(&dp->nfsdl_owner);
2054                         LIST_INIT(&dp->nfsdl_lock);
2055                         dp->nfsdl_clp = owp->nfsow_clp;
2056                         newnfs_copyincred(cred, &dp->nfsdl_cred);
2057                         nfscl_lockinit(&dp->nfsdl_rwlock);
2058                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2059                             NFSX_UNSIGNED);
2060                         dp->nfsdl_stateid.seqid = *tl++;
2061                         dp->nfsdl_stateid.other[0] = *tl++;
2062                         dp->nfsdl_stateid.other[1] = *tl++;
2063                         dp->nfsdl_stateid.other[2] = *tl++;
2064                         ret = fxdr_unsigned(int, *tl);
2065                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2066                                 dp->nfsdl_flags = NFSCLDL_WRITE;
2067                                 /*
2068                                  * Indicates how much the file can grow.
2069                                  */
2070                                 NFSM_DISSECT(tl, u_int32_t *,
2071                                     3 * NFSX_UNSIGNED);
2072                                 limitby = fxdr_unsigned(int, *tl++);
2073                                 switch (limitby) {
2074                                 case NFSV4OPEN_LIMITSIZE:
2075                                         dp->nfsdl_sizelimit = fxdr_hyper(tl);
2076                                         break;
2077                                 case NFSV4OPEN_LIMITBLOCKS:
2078                                         dp->nfsdl_sizelimit =
2079                                             fxdr_unsigned(u_int64_t, *tl++);
2080                                         dp->nfsdl_sizelimit *=
2081                                             fxdr_unsigned(u_int64_t, *tl);
2082                                         break;
2083                                 default:
2084                                         error = NFSERR_BADXDR;
2085                                         goto nfsmout;
2086                                 };
2087                         } else {
2088                                 dp->nfsdl_flags = NFSCLDL_READ;
2089                         }
2090                         if (ret)
2091                                 dp->nfsdl_flags |= NFSCLDL_RECALL;
2092                         error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2093                             &acesize, p);
2094                         if (error)
2095                                 goto nfsmout;
2096                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
2097                         error = NFSERR_BADXDR;
2098                         goto nfsmout;
2099                 }
2100                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2101                 if (error)
2102                         goto nfsmout;
2103                 /* Get rid of the PutFH and Getattr status values. */
2104                 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2105                 /* Load the directory attributes. */
2106                 error = nfsm_loadattr(nd, dnap);
2107                 if (error)
2108                         goto nfsmout;
2109                 *dattrflagp = 1;
2110                 if (dp != NULL && *attrflagp) {
2111                         dp->nfsdl_change = nnap->na_filerev;
2112                         dp->nfsdl_modtime = nnap->na_mtime;
2113                         dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2114                 }
2115                 /*
2116                  * We can now complete the Open state.
2117                  */
2118                 nfhp = *nfhpp;
2119                 if (dp != NULL) {
2120                         dp->nfsdl_fhlen = nfhp->nfh_len;
2121                         NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2122                 }
2123                 /*
2124                  * Get an Open structure that will be
2125                  * attached to the OpenOwner, acquired already.
2126                  */
2127                 error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2128                     (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2129                     cred, p, NULL, &op, &newone, NULL, 0);
2130                 if (error)
2131                         goto nfsmout;
2132                 op->nfso_stateid = stateid;
2133                 newnfs_copyincred(cred, &op->nfso_cred);
2134                 if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2135                     do {
2136                         ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2137                             nfhp->nfh_len, op, cred, p);
2138                         if (ret == NFSERR_DELAY)
2139                             (void) nfs_catnap(PZERO, ret, "nfs_create");
2140                     } while (ret == NFSERR_DELAY);
2141                     error = ret;
2142                 }
2143
2144                 /*
2145                  * If the server is handing out delegations, but we didn't
2146                  * get one because an OpenConfirm was required, try the
2147                  * Open again, to get a delegation. This is a harmless no-op,
2148                  * from a server's point of view.
2149                  */
2150                 if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2151                     (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2152                     !error && dp == NULL) {
2153                     do {
2154                         ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2155                             np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2156                             nfhp->nfh_fh, nfhp->nfh_len,
2157                             (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2158                             name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2159                         if (ret == NFSERR_DELAY)
2160                             (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2161                     } while (ret == NFSERR_DELAY);
2162                     if (ret) {
2163                         if (dp != NULL) {
2164                                 FREE((caddr_t)dp, M_NFSCLDELEG);
2165                                 dp = NULL;
2166                         }
2167                         if (ret == NFSERR_STALECLIENTID ||
2168                             ret == NFSERR_STALEDONTRECOVER ||
2169                             ret == NFSERR_BADSESSION)
2170                                 error = ret;
2171                     }
2172                 }
2173                 nfscl_openrelease(op, error, newone);
2174                 *unlockedp = 1;
2175         }
2176         if (nd->nd_repstat != 0 && error == 0)
2177                 error = nd->nd_repstat;
2178         if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
2179                 nfscl_initiate_recovery(owp->nfsow_clp);
2180 nfsmout:
2181         if (!error)
2182                 *dpp = dp;
2183         else if (dp != NULL)
2184                 FREE((caddr_t)dp, M_NFSCLDELEG);
2185         mbuf_freem(nd->nd_mrep);
2186         return (error);
2187 }
2188
2189 /*
2190  * Nfs remove rpc
2191  */
2192 APPLESTATIC int
2193 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2194     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2195     void *dstuff)
2196 {
2197         u_int32_t *tl;
2198         struct nfsrv_descript nfsd, *nd = &nfsd;
2199         struct nfsnode *np;
2200         struct nfsmount *nmp;
2201         nfsv4stateid_t dstateid;
2202         int error, ret = 0, i;
2203
2204         *dattrflagp = 0;
2205         if (namelen > NFS_MAXNAMLEN)
2206                 return (ENAMETOOLONG);
2207         nmp = VFSTONFS(vnode_mount(dvp));
2208 tryagain:
2209         if (NFSHASNFSV4(nmp) && ret == 0) {
2210                 ret = nfscl_removedeleg(vp, p, &dstateid);
2211                 if (ret == 1) {
2212                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2213                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2214                             NFSX_UNSIGNED);
2215                         if (NFSHASNFSV4N(nmp))
2216                                 *tl++ = 0;
2217                         else
2218                                 *tl++ = dstateid.seqid;
2219                         *tl++ = dstateid.other[0];
2220                         *tl++ = dstateid.other[1];
2221                         *tl++ = dstateid.other[2];
2222                         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2223                         np = VTONFS(dvp);
2224                         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2225                             np->n_fhp->nfh_len, 0);
2226                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2227                         *tl = txdr_unsigned(NFSV4OP_REMOVE);
2228                 }
2229         } else {
2230                 ret = 0;
2231         }
2232         if (ret == 0)
2233                 NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2234         (void) nfsm_strtom(nd, name, namelen);
2235         error = nfscl_request(nd, dvp, p, cred, dstuff);
2236         if (error)
2237                 return (error);
2238         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2239                 /* For NFSv4, parse out any Delereturn replies. */
2240                 if (ret > 0 && nd->nd_repstat != 0 &&
2241                     (nd->nd_flag & ND_NOMOREDATA)) {
2242                         /*
2243                          * If the Delegreturn failed, try again without
2244                          * it. The server will Recall, as required.
2245                          */
2246                         mbuf_freem(nd->nd_mrep);
2247                         goto tryagain;
2248                 }
2249                 for (i = 0; i < (ret * 2); i++) {
2250                         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2251                             ND_NFSV4) {
2252                             NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2253                             if (*(tl + 1))
2254                                 nd->nd_flag |= ND_NOMOREDATA;
2255                         }
2256                 }
2257                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2258         }
2259         if (nd->nd_repstat && !error)
2260                 error = nd->nd_repstat;
2261 nfsmout:
2262         mbuf_freem(nd->nd_mrep);
2263         return (error);
2264 }
2265
2266 /*
2267  * Do an nfs rename rpc.
2268  */
2269 APPLESTATIC int
2270 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2271     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2272     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2273     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2274 {
2275         u_int32_t *tl;
2276         struct nfsrv_descript nfsd, *nd = &nfsd;
2277         struct nfsmount *nmp;
2278         struct nfsnode *np;
2279         nfsattrbit_t attrbits;
2280         nfsv4stateid_t fdstateid, tdstateid;
2281         int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2282
2283         *fattrflagp = 0;
2284         *tattrflagp = 0;
2285         nmp = VFSTONFS(vnode_mount(fdvp));
2286         if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2287                 return (ENAMETOOLONG);
2288 tryagain:
2289         if (NFSHASNFSV4(nmp) && ret == 0) {
2290                 ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2291                     &tdstateid, &gottd, p);
2292                 if (gotfd && gottd) {
2293                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2294                 } else if (gotfd) {
2295                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2296                 } else if (gottd) {
2297                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2298                 }
2299                 if (gotfd) {
2300                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2301                         if (NFSHASNFSV4N(nmp))
2302                                 *tl++ = 0;
2303                         else
2304                                 *tl++ = fdstateid.seqid;
2305                         *tl++ = fdstateid.other[0];
2306                         *tl++ = fdstateid.other[1];
2307                         *tl = fdstateid.other[2];
2308                         if (gottd) {
2309                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2310                                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2311                                 np = VTONFS(tvp);
2312                                 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2313                                     np->n_fhp->nfh_len, 0);
2314                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2315                                 *tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2316                         }
2317                 }
2318                 if (gottd) {
2319                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2320                         if (NFSHASNFSV4N(nmp))
2321                                 *tl++ = 0;
2322                         else
2323                                 *tl++ = tdstateid.seqid;
2324                         *tl++ = tdstateid.other[0];
2325                         *tl++ = tdstateid.other[1];
2326                         *tl = tdstateid.other[2];
2327                 }
2328                 if (ret > 0) {
2329                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2330                         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2331                         np = VTONFS(fdvp);
2332                         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2333                             np->n_fhp->nfh_len, 0);
2334                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2335                         *tl = txdr_unsigned(NFSV4OP_SAVEFH);
2336                 }
2337         } else {
2338                 ret = 0;
2339         }
2340         if (ret == 0)
2341                 NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2342         if (nd->nd_flag & ND_NFSV4) {
2343                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2344                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2345                 NFSWCCATTR_ATTRBIT(&attrbits);
2346                 (void) nfsrv_putattrbit(nd, &attrbits);
2347                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2348                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2349                 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2350                     VTONFS(tdvp)->n_fhp->nfh_len, 0);
2351                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2352                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2353                 (void) nfsrv_putattrbit(nd, &attrbits);
2354                 nd->nd_flag |= ND_V4WCCATTR;
2355                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2356                 *tl = txdr_unsigned(NFSV4OP_RENAME);
2357         }
2358         (void) nfsm_strtom(nd, fnameptr, fnamelen);
2359         if (!(nd->nd_flag & ND_NFSV4))
2360                 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2361                         VTONFS(tdvp)->n_fhp->nfh_len, 0);
2362         (void) nfsm_strtom(nd, tnameptr, tnamelen);
2363         error = nfscl_request(nd, fdvp, p, cred, fstuff);
2364         if (error)
2365                 return (error);
2366         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2367                 /* For NFSv4, parse out any Delereturn replies. */
2368                 if (ret > 0 && nd->nd_repstat != 0 &&
2369                     (nd->nd_flag & ND_NOMOREDATA)) {
2370                         /*
2371                          * If the Delegreturn failed, try again without
2372                          * it. The server will Recall, as required.
2373                          */
2374                         mbuf_freem(nd->nd_mrep);
2375                         goto tryagain;
2376                 }
2377                 for (i = 0; i < (ret * 2); i++) {
2378                         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2379                             ND_NFSV4) {
2380                             NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2381                             if (*(tl + 1)) {
2382                                 if (i == 0 && ret > 1) {
2383                                     /*
2384                                      * If the Delegreturn failed, try again
2385                                      * without it. The server will Recall, as
2386                                      * required.
2387                                      * If ret > 1, the first iteration of this
2388                                      * loop is the second DelegReturn result.
2389                                      */
2390                                     mbuf_freem(nd->nd_mrep);
2391                                     goto tryagain;
2392                                 } else {
2393                                     nd->nd_flag |= ND_NOMOREDATA;
2394                                 }
2395                             }
2396                         }
2397                 }
2398                 /* Now, the first wcc attribute reply. */
2399                 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2400                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2401                         if (*(tl + 1))
2402                                 nd->nd_flag |= ND_NOMOREDATA;
2403                 }
2404                 error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2405                     fstuff);
2406                 /* and the second wcc attribute reply. */
2407                 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2408                     !error) {
2409                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2410                         if (*(tl + 1))
2411                                 nd->nd_flag |= ND_NOMOREDATA;
2412                 }
2413                 if (!error)
2414                         error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2415                             NULL, tstuff);
2416         }
2417         if (nd->nd_repstat && !error)
2418                 error = nd->nd_repstat;
2419 nfsmout:
2420         mbuf_freem(nd->nd_mrep);
2421         return (error);
2422 }
2423
2424 /*
2425  * nfs hard link create rpc
2426  */
2427 APPLESTATIC int
2428 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2429     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2430     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2431 {
2432         u_int32_t *tl;
2433         struct nfsrv_descript nfsd, *nd = &nfsd;
2434         nfsattrbit_t attrbits;
2435         int error = 0;
2436
2437         *attrflagp = 0;
2438         *dattrflagp = 0;
2439         if (namelen > NFS_MAXNAMLEN)
2440                 return (ENAMETOOLONG);
2441         NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2442         if (nd->nd_flag & ND_NFSV4) {
2443                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2444                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2445         }
2446         (void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2447                 VTONFS(dvp)->n_fhp->nfh_len, 0);
2448         if (nd->nd_flag & ND_NFSV4) {
2449                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2450                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2451                 NFSWCCATTR_ATTRBIT(&attrbits);
2452                 (void) nfsrv_putattrbit(nd, &attrbits);
2453                 nd->nd_flag |= ND_V4WCCATTR;
2454                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2455                 *tl = txdr_unsigned(NFSV4OP_LINK);
2456         }
2457         (void) nfsm_strtom(nd, name, namelen);
2458         error = nfscl_request(nd, vp, p, cred, dstuff);
2459         if (error)
2460                 return (error);
2461         if (nd->nd_flag & ND_NFSV3) {
2462                 error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2463                 if (!error)
2464                         error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2465                             NULL, dstuff);
2466         } else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2467                 /*
2468                  * First, parse out the PutFH and Getattr result.
2469                  */
2470                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2471                 if (!(*(tl + 1)))
2472                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2473                 if (*(tl + 1))
2474                         nd->nd_flag |= ND_NOMOREDATA;
2475                 /*
2476                  * Get the pre-op attributes.
2477                  */
2478                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2479         }
2480         if (nd->nd_repstat && !error)
2481                 error = nd->nd_repstat;
2482 nfsmout:
2483         mbuf_freem(nd->nd_mrep);
2484         return (error);
2485 }
2486
2487 /*
2488  * nfs symbolic link create rpc
2489  */
2490 APPLESTATIC int
2491 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, char *target,
2492     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2493     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2494     int *dattrflagp, void *dstuff)
2495 {
2496         u_int32_t *tl;
2497         struct nfsrv_descript nfsd, *nd = &nfsd;
2498         struct nfsmount *nmp;
2499         int slen, error = 0;
2500
2501         *nfhpp = NULL;
2502         *attrflagp = 0;
2503         *dattrflagp = 0;
2504         nmp = VFSTONFS(vnode_mount(dvp));
2505         slen = strlen(target);
2506         if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2507                 return (ENAMETOOLONG);
2508         NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2509         if (nd->nd_flag & ND_NFSV4) {
2510                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2511                 *tl = txdr_unsigned(NFLNK);
2512                 (void) nfsm_strtom(nd, target, slen);
2513         }
2514         (void) nfsm_strtom(nd, name, namelen);
2515         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2516                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2517         if (!(nd->nd_flag & ND_NFSV4))
2518                 (void) nfsm_strtom(nd, target, slen);
2519         if (nd->nd_flag & ND_NFSV2)
2520                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2521         error = nfscl_request(nd, dvp, p, cred, dstuff);
2522         if (error)
2523                 return (error);
2524         if (nd->nd_flag & ND_NFSV4)
2525                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2526         if ((nd->nd_flag & ND_NFSV3) && !error) {
2527                 if (!nd->nd_repstat)
2528                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2529                 if (!error)
2530                         error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2531                             NULL, dstuff);
2532         }
2533         if (nd->nd_repstat && !error)
2534                 error = nd->nd_repstat;
2535         mbuf_freem(nd->nd_mrep);
2536         /*
2537          * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2538          * Only do this if vfs.nfs.ignore_eexist is set.
2539          * Never do this for NFSv4.1 or later minor versions, since sessions
2540          * should guarantee "exactly once" RPC semantics.
2541          */
2542         if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2543             nmp->nm_minorvers == 0))
2544                 error = 0;
2545         return (error);
2546 }
2547
2548 /*
2549  * nfs make dir rpc
2550  */
2551 APPLESTATIC int
2552 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2553     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2554     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2555     int *dattrflagp, void *dstuff)
2556 {
2557         u_int32_t *tl;
2558         struct nfsrv_descript nfsd, *nd = &nfsd;
2559         nfsattrbit_t attrbits;
2560         int error = 0;
2561         struct nfsfh *fhp;
2562         struct nfsmount *nmp;
2563
2564         *nfhpp = NULL;
2565         *attrflagp = 0;
2566         *dattrflagp = 0;
2567         nmp = VFSTONFS(vnode_mount(dvp));
2568         fhp = VTONFS(dvp)->n_fhp;
2569         if (namelen > NFS_MAXNAMLEN)
2570                 return (ENAMETOOLONG);
2571         NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2572         if (nd->nd_flag & ND_NFSV4) {
2573                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2574                 *tl = txdr_unsigned(NFDIR);
2575         }
2576         (void) nfsm_strtom(nd, name, namelen);
2577         nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2578         if (nd->nd_flag & ND_NFSV4) {
2579                 NFSGETATTR_ATTRBIT(&attrbits);
2580                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2581                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2582                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2583                 (void) nfsrv_putattrbit(nd, &attrbits);
2584                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2585                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2586                 (void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2587                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2588                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2589                 (void) nfsrv_putattrbit(nd, &attrbits);
2590         }
2591         error = nfscl_request(nd, dvp, p, cred, dstuff);
2592         if (error)
2593                 return (error);
2594         if (nd->nd_flag & ND_NFSV4)
2595                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2596         if (!nd->nd_repstat && !error) {
2597                 if (nd->nd_flag & ND_NFSV4) {
2598                         NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2599                         error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2600                 }
2601                 if (!error)
2602                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2603                 if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2604                         /* Get rid of the PutFH and Getattr status values. */
2605                         NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2606                         /* Load the directory attributes. */
2607                         error = nfsm_loadattr(nd, dnap);
2608                         if (error == 0)
2609                                 *dattrflagp = 1;
2610                 }
2611         }
2612         if ((nd->nd_flag & ND_NFSV3) && !error)
2613                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2614         if (nd->nd_repstat && !error)
2615                 error = nd->nd_repstat;
2616 nfsmout:
2617         mbuf_freem(nd->nd_mrep);
2618         /*
2619          * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2620          * Only do this if vfs.nfs.ignore_eexist is set.
2621          * Never do this for NFSv4.1 or later minor versions, since sessions
2622          * should guarantee "exactly once" RPC semantics.
2623          */
2624         if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2625             nmp->nm_minorvers == 0))
2626                 error = 0;
2627         return (error);
2628 }
2629
2630 /*
2631  * nfs remove directory call
2632  */
2633 APPLESTATIC int
2634 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2635     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2636 {
2637         struct nfsrv_descript nfsd, *nd = &nfsd;
2638         int error = 0;
2639
2640         *dattrflagp = 0;
2641         if (namelen > NFS_MAXNAMLEN)
2642                 return (ENAMETOOLONG);
2643         NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2644         (void) nfsm_strtom(nd, name, namelen);
2645         error = nfscl_request(nd, dvp, p, cred, dstuff);
2646         if (error)
2647                 return (error);
2648         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2649                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2650         if (nd->nd_repstat && !error)
2651                 error = nd->nd_repstat;
2652         mbuf_freem(nd->nd_mrep);
2653         /*
2654          * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2655          */
2656         if (error == ENOENT)
2657                 error = 0;
2658         return (error);
2659 }
2660
2661 /*
2662  * Readdir rpc.
2663  * Always returns with either uio_resid unchanged, if you are at the
2664  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2665  * filled in.
2666  * I felt this would allow caching of directory blocks more easily
2667  * than returning a pertially filled block.
2668  * Directory offset cookies:
2669  * Oh my, what to do with them...
2670  * I can think of three ways to deal with them:
2671  * 1 - have the layer above these RPCs maintain a map between logical
2672  *     directory byte offsets and the NFS directory offset cookies
2673  * 2 - pass the opaque directory offset cookies up into userland
2674  *     and let the libc functions deal with them, via the system call
2675  * 3 - return them to userland in the "struct dirent", so future versions
2676  *     of libc can use them and do whatever is necessary to amke things work
2677  *     above these rpc calls, in the meantime
2678  * For now, I do #3 by "hiding" the directory offset cookies after the
2679  * d_name field in struct dirent. This is space inside d_reclen that
2680  * will be ignored by anything that doesn't know about them.
2681  * The directory offset cookies are filled in as the last 8 bytes of
2682  * each directory entry, after d_name. Someday, the userland libc
2683  * functions may be able to use these. In the meantime, it satisfies
2684  * OpenBSD's requirements for cookies being returned.
2685  * If expects the directory offset cookie for the read to be in uio_offset
2686  * and returns the one for the next entry after this directory block in
2687  * there, as well.
2688  */
2689 APPLESTATIC int
2690 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2691     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2692     int *eofp, void *stuff)
2693 {
2694         int len, left;
2695         struct dirent *dp = NULL;
2696         u_int32_t *tl;
2697         nfsquad_t cookie, ncookie;
2698         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2699         struct nfsnode *dnp = VTONFS(vp);
2700         struct nfsvattr nfsva;
2701         struct nfsrv_descript nfsd, *nd = &nfsd;
2702         int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2703         int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2704         long dotfileid, dotdotfileid = 0;
2705         u_int32_t fakefileno = 0xffffffff, rderr;
2706         char *cp;
2707         nfsattrbit_t attrbits, dattrbits;
2708         u_int32_t *tl2 = NULL;
2709         size_t tresid;
2710
2711         KASSERT(uiop->uio_iovcnt == 1 &&
2712             (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2713             ("nfs readdirrpc bad uio"));
2714
2715         /*
2716          * There is no point in reading a lot more than uio_resid, however
2717          * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2718          * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2719          * will never make readsize > nm_readdirsize.
2720          */
2721         readsize = nmp->nm_readdirsize;
2722         if (readsize > uio_uio_resid(uiop))
2723                 readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2724
2725         *attrflagp = 0;
2726         if (eofp)
2727                 *eofp = 0;
2728         tresid = uio_uio_resid(uiop);
2729         cookie.lval[0] = cookiep->nfsuquad[0];
2730         cookie.lval[1] = cookiep->nfsuquad[1];
2731         nd->nd_mrep = NULL;
2732
2733         /*
2734          * For NFSv4, first create the "." and ".." entries.
2735          */
2736         if (NFSHASNFSV4(nmp)) {
2737                 reqsize = 6 * NFSX_UNSIGNED;
2738                 NFSGETATTR_ATTRBIT(&dattrbits);
2739                 NFSZERO_ATTRBIT(&attrbits);
2740                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2741                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2742                 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2743                     NFSATTRBIT_MOUNTEDONFILEID)) {
2744                         NFSSETBIT_ATTRBIT(&attrbits,
2745                             NFSATTRBIT_MOUNTEDONFILEID);
2746                         gotmnton = 1;
2747                 } else {
2748                         /*
2749                          * Must fake it. Use the fileno, except when the
2750                          * fsid is != to that of the directory. For that
2751                          * case, generate a fake fileno that is not the same.
2752                          */
2753                         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2754                         gotmnton = 0;
2755                 }
2756
2757                 /*
2758                  * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2759                  */
2760                 if (uiop->uio_offset == 0) {
2761                         NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2762                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2763                         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2764                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2765                         (void) nfsrv_putattrbit(nd, &attrbits);
2766                         error = nfscl_request(nd, vp, p, cred, stuff);
2767                         if (error)
2768                             return (error);
2769                         dotfileid = 0;  /* Fake out the compiler. */
2770                         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2771                             error = nfsm_loadattr(nd, &nfsva);
2772                             if (error != 0)
2773                                 goto nfsmout;
2774                             dotfileid = nfsva.na_fileid;
2775                         }
2776                         if (nd->nd_repstat == 0) {
2777                             NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2778                             len = fxdr_unsigned(int, *(tl + 4));
2779                             if (len > 0 && len <= NFSX_V4FHMAX)
2780                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2781                             else
2782                                 error = EPERM;
2783                             if (!error) {
2784                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2785                                 nfsva.na_mntonfileno = 0xffffffff;
2786                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2787                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2788                                     NULL, NULL, NULL, p, cred);
2789                                 if (error) {
2790                                     dotdotfileid = dotfileid;
2791                                 } else if (gotmnton) {
2792                                     if (nfsva.na_mntonfileno != 0xffffffff)
2793                                         dotdotfileid = nfsva.na_mntonfileno;
2794                                     else
2795                                         dotdotfileid = nfsva.na_fileid;
2796                                 } else if (nfsva.na_filesid[0] ==
2797                                     dnp->n_vattr.na_filesid[0] &&
2798                                     nfsva.na_filesid[1] ==
2799                                     dnp->n_vattr.na_filesid[1]) {
2800                                     dotdotfileid = nfsva.na_fileid;
2801                                 } else {
2802                                     do {
2803                                         fakefileno--;
2804                                     } while (fakefileno ==
2805                                         nfsva.na_fileid);
2806                                     dotdotfileid = fakefileno;
2807                                 }
2808                             }
2809                         } else if (nd->nd_repstat == NFSERR_NOENT) {
2810                             /*
2811                              * Lookupp returns NFSERR_NOENT when we are
2812                              * at the root, so just use the current dir.
2813                              */
2814                             nd->nd_repstat = 0;
2815                             dotdotfileid = dotfileid;
2816                         } else {
2817                             error = nd->nd_repstat;
2818                         }
2819                         mbuf_freem(nd->nd_mrep);
2820                         if (error)
2821                             return (error);
2822                         nd->nd_mrep = NULL;
2823                         dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2824                         dp->d_type = DT_DIR;
2825                         dp->d_fileno = dotfileid;
2826                         dp->d_namlen = 1;
2827                         dp->d_name[0] = '.';
2828                         dp->d_name[1] = '\0';
2829                         dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2830                         /*
2831                          * Just make these offset cookie 0.
2832                          */
2833                         tl = (u_int32_t *)&dp->d_name[4];
2834                         *tl++ = 0;
2835                         *tl = 0;
2836                         blksiz += dp->d_reclen;
2837                         uio_uio_resid_add(uiop, -(dp->d_reclen));
2838                         uiop->uio_offset += dp->d_reclen;
2839                         uio_iov_base_add(uiop, dp->d_reclen);
2840                         uio_iov_len_add(uiop, -(dp->d_reclen));
2841                         dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2842                         dp->d_type = DT_DIR;
2843                         dp->d_fileno = dotdotfileid;
2844                         dp->d_namlen = 2;
2845                         dp->d_name[0] = '.';
2846                         dp->d_name[1] = '.';
2847                         dp->d_name[2] = '\0';
2848                         dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2849                         /*
2850                          * Just make these offset cookie 0.
2851                          */
2852                         tl = (u_int32_t *)&dp->d_name[4];
2853                         *tl++ = 0;
2854                         *tl = 0;
2855                         blksiz += dp->d_reclen;
2856                         uio_uio_resid_add(uiop, -(dp->d_reclen));
2857                         uiop->uio_offset += dp->d_reclen;
2858                         uio_iov_base_add(uiop, dp->d_reclen);
2859                         uio_iov_len_add(uiop, -(dp->d_reclen));
2860                 }
2861                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
2862         } else {
2863                 reqsize = 5 * NFSX_UNSIGNED;
2864         }
2865
2866
2867         /*
2868          * Loop around doing readdir rpc's of size readsize.
2869          * The stopping criteria is EOF or buffer full.
2870          */
2871         while (more_dirs && bigenough) {
2872                 *attrflagp = 0;
2873                 NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
2874                 if (nd->nd_flag & ND_NFSV2) {
2875                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2876                         *tl++ = cookie.lval[1];
2877                         *tl = txdr_unsigned(readsize);
2878                 } else {
2879                         NFSM_BUILD(tl, u_int32_t *, reqsize);
2880                         *tl++ = cookie.lval[0];
2881                         *tl++ = cookie.lval[1];
2882                         if (cookie.qval == 0) {
2883                                 *tl++ = 0;
2884                                 *tl++ = 0;
2885                         } else {
2886                                 NFSLOCKNODE(dnp);
2887                                 *tl++ = dnp->n_cookieverf.nfsuquad[0];
2888                                 *tl++ = dnp->n_cookieverf.nfsuquad[1];
2889                                 NFSUNLOCKNODE(dnp);
2890                         }
2891                         if (nd->nd_flag & ND_NFSV4) {
2892                                 *tl++ = txdr_unsigned(readsize);
2893                                 *tl = txdr_unsigned(readsize);
2894                                 (void) nfsrv_putattrbit(nd, &attrbits);
2895                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2896                                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2897                                 (void) nfsrv_putattrbit(nd, &dattrbits);
2898                         } else {
2899                                 *tl = txdr_unsigned(readsize);
2900                         }
2901                 }
2902                 error = nfscl_request(nd, vp, p, cred, stuff);
2903                 if (error)
2904                         return (error);
2905                 if (!(nd->nd_flag & ND_NFSV2)) {
2906                         if (nd->nd_flag & ND_NFSV3)
2907                                 error = nfscl_postop_attr(nd, nap, attrflagp,
2908                                     stuff);
2909                         if (!nd->nd_repstat && !error) {
2910                                 NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
2911                                 NFSLOCKNODE(dnp);
2912                                 dnp->n_cookieverf.nfsuquad[0] = *tl++;
2913                                 dnp->n_cookieverf.nfsuquad[1] = *tl;
2914                                 NFSUNLOCKNODE(dnp);
2915                         }
2916                 }
2917                 if (nd->nd_repstat || error) {
2918                         if (!error)
2919                                 error = nd->nd_repstat;
2920                         goto nfsmout;
2921                 }
2922                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2923                 more_dirs = fxdr_unsigned(int, *tl);
2924                 if (!more_dirs)
2925                         tryformoredirs = 0;
2926
2927                 /* loop thru the dir entries, doctoring them to 4bsd form */
2928                 while (more_dirs && bigenough) {
2929                         if (nd->nd_flag & ND_NFSV4) {
2930                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2931                                 ncookie.lval[0] = *tl++;
2932                                 ncookie.lval[1] = *tl++;
2933                                 len = fxdr_unsigned(int, *tl);
2934                         } else if (nd->nd_flag & ND_NFSV3) {
2935                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2936                                 nfsva.na_fileid = fxdr_hyper(tl);
2937                                 tl += 2;
2938                                 len = fxdr_unsigned(int, *tl);
2939                         } else {
2940                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2941                                 nfsva.na_fileid =
2942                                     fxdr_unsigned(long, *tl++);
2943                                 len = fxdr_unsigned(int, *tl);
2944                         }
2945                         if (len <= 0 || len > NFS_MAXNAMLEN) {
2946                                 error = EBADRPC;
2947                                 goto nfsmout;
2948                         }
2949                         tlen = NFSM_RNDUP(len);
2950                         if (tlen == len)
2951                                 tlen += 4;  /* To ensure null termination */
2952                         left = DIRBLKSIZ - blksiz;
2953                         if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > left) {
2954                                 dp->d_reclen += left;
2955                                 uio_iov_base_add(uiop, left);
2956                                 uio_iov_len_add(uiop, -(left));
2957                                 uio_uio_resid_add(uiop, -(left));
2958                                 uiop->uio_offset += left;
2959                                 blksiz = 0;
2960                         }
2961                         if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
2962                                 bigenough = 0;
2963                         if (bigenough) {
2964                                 dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2965                                 dp->d_namlen = len;
2966                                 dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
2967                                 dp->d_type = DT_UNKNOWN;
2968                                 blksiz += dp->d_reclen;
2969                                 if (blksiz == DIRBLKSIZ)
2970                                         blksiz = 0;
2971                                 uio_uio_resid_add(uiop, -(DIRHDSIZ));
2972                                 uiop->uio_offset += DIRHDSIZ;
2973                                 uio_iov_base_add(uiop, DIRHDSIZ);
2974                                 uio_iov_len_add(uiop, -(DIRHDSIZ));
2975                                 error = nfsm_mbufuio(nd, uiop, len);
2976                                 if (error)
2977                                         goto nfsmout;
2978                                 cp = CAST_DOWN(caddr_t, uio_iov_base(uiop));
2979                                 tlen -= len;
2980                                 *cp = '\0';     /* null terminate */
2981                                 cp += tlen;     /* points to cookie storage */
2982                                 tl2 = (u_int32_t *)cp;
2983                                 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
2984                                 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
2985                                 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
2986                                 uiop->uio_offset += (tlen + NFSX_HYPER);
2987                         } else {
2988                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2989                                 if (error)
2990                                         goto nfsmout;
2991                         }
2992                         if (nd->nd_flag & ND_NFSV4) {
2993                                 rderr = 0;
2994                                 nfsva.na_mntonfileno = 0xffffffff;
2995                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2996                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2997                                     NULL, NULL, &rderr, p, cred);
2998                                 if (error)
2999                                         goto nfsmout;
3000                                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3001                         } else if (nd->nd_flag & ND_NFSV3) {
3002                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3003                                 ncookie.lval[0] = *tl++;
3004                                 ncookie.lval[1] = *tl++;
3005                         } else {
3006                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3007                                 ncookie.lval[0] = 0;
3008                                 ncookie.lval[1] = *tl++;
3009                         }
3010                         if (bigenough) {
3011                             if (nd->nd_flag & ND_NFSV4) {
3012                                 if (rderr) {
3013                                     dp->d_fileno = 0;
3014                                 } else {
3015                                     if (gotmnton) {
3016                                         if (nfsva.na_mntonfileno != 0xffffffff)
3017                                             dp->d_fileno = nfsva.na_mntonfileno;
3018                                         else
3019                                             dp->d_fileno = nfsva.na_fileid;
3020                                     } else if (nfsva.na_filesid[0] ==
3021                                         dnp->n_vattr.na_filesid[0] &&
3022                                         nfsva.na_filesid[1] ==
3023                                         dnp->n_vattr.na_filesid[1]) {
3024                                         dp->d_fileno = nfsva.na_fileid;
3025                                     } else {
3026                                         do {
3027                                             fakefileno--;
3028                                         } while (fakefileno ==
3029                                             nfsva.na_fileid);
3030                                         dp->d_fileno = fakefileno;
3031                                     }
3032                                     dp->d_type = vtonfs_dtype(nfsva.na_type);
3033                                 }
3034                             } else {
3035                                 dp->d_fileno = nfsva.na_fileid;
3036                             }
3037                             *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3038                                 ncookie.lval[0];
3039                             *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3040                                 ncookie.lval[1];
3041                         }
3042                         more_dirs = fxdr_unsigned(int, *tl);
3043                 }
3044                 /*
3045                  * If at end of rpc data, get the eof boolean
3046                  */
3047                 if (!more_dirs) {
3048                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3049                         eof = fxdr_unsigned(int, *tl);
3050                         if (tryformoredirs)
3051                                 more_dirs = !eof;
3052                         if (nd->nd_flag & ND_NFSV4) {
3053                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3054                                     stuff);
3055                                 if (error)
3056                                         goto nfsmout;
3057                         }
3058                 }
3059                 mbuf_freem(nd->nd_mrep);
3060                 nd->nd_mrep = NULL;
3061         }
3062         /*
3063          * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3064          * by increasing d_reclen for the last record.
3065          */
3066         if (blksiz > 0) {
3067                 left = DIRBLKSIZ - blksiz;
3068                 dp->d_reclen += left;
3069                 uio_iov_base_add(uiop, left);
3070                 uio_iov_len_add(uiop, -(left));
3071                 uio_uio_resid_add(uiop, -(left));
3072                 uiop->uio_offset += left;
3073         }
3074
3075         /*
3076          * If returning no data, assume end of file.
3077          * If not bigenough, return not end of file, since you aren't
3078          *    returning all the data
3079          * Otherwise, return the eof flag from the server.
3080          */
3081         if (eofp) {
3082                 if (tresid == ((size_t)(uio_uio_resid(uiop))))
3083                         *eofp = 1;
3084                 else if (!bigenough)
3085                         *eofp = 0;
3086                 else
3087                         *eofp = eof;
3088         }
3089
3090         /*
3091          * Add extra empty records to any remaining DIRBLKSIZ chunks.
3092          */
3093         while (uio_uio_resid(uiop) > 0 && ((size_t)(uio_uio_resid(uiop))) != tresid) {
3094                 dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
3095                 dp->d_type = DT_UNKNOWN;
3096                 dp->d_fileno = 0;
3097                 dp->d_namlen = 0;
3098                 dp->d_name[0] = '\0';
3099                 tl = (u_int32_t *)&dp->d_name[4];
3100                 *tl++ = cookie.lval[0];
3101                 *tl = cookie.lval[1];
3102                 dp->d_reclen = DIRBLKSIZ;
3103                 uio_iov_base_add(uiop, DIRBLKSIZ);
3104                 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3105                 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3106                 uiop->uio_offset += DIRBLKSIZ;
3107         }
3108
3109 nfsmout:
3110         if (nd->nd_mrep != NULL)
3111                 mbuf_freem(nd->nd_mrep);
3112         return (error);
3113 }
3114
3115 #ifndef APPLE
3116 /*
3117  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3118  * (Also used for NFS V4 when mount flag set.)
3119  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3120  */
3121 APPLESTATIC int
3122 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3123     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3124     int *eofp, void *stuff)
3125 {
3126         int len, left;
3127         struct dirent *dp = NULL;
3128         u_int32_t *tl;
3129         vnode_t newvp = NULLVP;
3130         struct nfsrv_descript nfsd, *nd = &nfsd;
3131         struct nameidata nami, *ndp = &nami;
3132         struct componentname *cnp = &ndp->ni_cnd;
3133         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3134         struct nfsnode *dnp = VTONFS(vp), *np;
3135         struct nfsvattr nfsva;
3136         struct nfsfh *nfhp;
3137         nfsquad_t cookie, ncookie;
3138         int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3139         int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3140         int isdotdot = 0, unlocknewvp = 0;
3141         long dotfileid, dotdotfileid = 0, fileno = 0;
3142         char *cp;
3143         nfsattrbit_t attrbits, dattrbits;
3144         size_t tresid;
3145         u_int32_t *tl2 = NULL, fakefileno = 0xffffffff, rderr;
3146         struct timespec dctime;
3147
3148         KASSERT(uiop->uio_iovcnt == 1 &&
3149             (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3150             ("nfs readdirplusrpc bad uio"));
3151         timespecclear(&dctime);
3152         *attrflagp = 0;
3153         if (eofp != NULL)
3154                 *eofp = 0;
3155         ndp->ni_dvp = vp;
3156         nd->nd_mrep = NULL;
3157         cookie.lval[0] = cookiep->nfsuquad[0];
3158         cookie.lval[1] = cookiep->nfsuquad[1];
3159         tresid = uio_uio_resid(uiop);
3160
3161         /*
3162          * For NFSv4, first create the "." and ".." entries.
3163          */
3164         if (NFSHASNFSV4(nmp)) {
3165                 NFSGETATTR_ATTRBIT(&dattrbits);
3166                 NFSZERO_ATTRBIT(&attrbits);
3167                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3168                 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3169                     NFSATTRBIT_MOUNTEDONFILEID)) {
3170                         NFSSETBIT_ATTRBIT(&attrbits,
3171                             NFSATTRBIT_MOUNTEDONFILEID);
3172                         gotmnton = 1;
3173                 } else {
3174                         /*
3175                          * Must fake it. Use the fileno, except when the
3176                          * fsid is != to that of the directory. For that
3177                          * case, generate a fake fileno that is not the same.
3178                          */
3179                         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3180                         gotmnton = 0;
3181                 }
3182
3183                 /*
3184                  * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3185                  */
3186                 if (uiop->uio_offset == 0) {
3187                         NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3188                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3189                         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
3190                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
3191                         (void) nfsrv_putattrbit(nd, &attrbits);
3192                         error = nfscl_request(nd, vp, p, cred, stuff);
3193                         if (error)
3194                             return (error);
3195                         dotfileid = 0;  /* Fake out the compiler. */
3196                         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3197                             error = nfsm_loadattr(nd, &nfsva);
3198                             if (error != 0)
3199                                 goto nfsmout;
3200                             dctime = nfsva.na_ctime;
3201                             dotfileid = nfsva.na_fileid;
3202                         }
3203                         if (nd->nd_repstat == 0) {
3204                             NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3205                             len = fxdr_unsigned(int, *(tl + 4));
3206                             if (len > 0 && len <= NFSX_V4FHMAX)
3207                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3208                             else
3209                                 error = EPERM;
3210                             if (!error) {
3211                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3212                                 nfsva.na_mntonfileno = 0xffffffff;
3213                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3214                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3215                                     NULL, NULL, NULL, p, cred);
3216                                 if (error) {
3217                                     dotdotfileid = dotfileid;
3218                                 } else if (gotmnton) {
3219                                     if (nfsva.na_mntonfileno != 0xffffffff)
3220                                         dotdotfileid = nfsva.na_mntonfileno;
3221                                     else
3222                                         dotdotfileid = nfsva.na_fileid;
3223                                 } else if (nfsva.na_filesid[0] ==
3224                                     dnp->n_vattr.na_filesid[0] &&
3225                                     nfsva.na_filesid[1] ==
3226                                     dnp->n_vattr.na_filesid[1]) {
3227                                     dotdotfileid = nfsva.na_fileid;
3228                                 } else {
3229                                     do {
3230                                         fakefileno--;
3231                                     } while (fakefileno ==
3232                                         nfsva.na_fileid);
3233                                     dotdotfileid = fakefileno;
3234                                 }
3235                             }
3236                         } else if (nd->nd_repstat == NFSERR_NOENT) {
3237                             /*
3238                              * Lookupp returns NFSERR_NOENT when we are
3239                              * at the root, so just use the current dir.
3240                              */
3241                             nd->nd_repstat = 0;
3242                             dotdotfileid = dotfileid;
3243                         } else {
3244                             error = nd->nd_repstat;
3245                         }
3246                         mbuf_freem(nd->nd_mrep);
3247                         if (error)
3248                             return (error);
3249                         nd->nd_mrep = NULL;
3250                         dp = (struct dirent *)uio_iov_base(uiop);
3251                         dp->d_type = DT_DIR;
3252                         dp->d_fileno = dotfileid;
3253                         dp->d_namlen = 1;
3254                         dp->d_name[0] = '.';
3255                         dp->d_name[1] = '\0';
3256                         dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3257                         /*
3258                          * Just make these offset cookie 0.
3259                          */
3260                         tl = (u_int32_t *)&dp->d_name[4];
3261                         *tl++ = 0;
3262                         *tl = 0;
3263                         blksiz += dp->d_reclen;
3264                         uio_uio_resid_add(uiop, -(dp->d_reclen));
3265                         uiop->uio_offset += dp->d_reclen;
3266                         uio_iov_base_add(uiop, dp->d_reclen);
3267                         uio_iov_len_add(uiop, -(dp->d_reclen));
3268                         dp = (struct dirent *)uio_iov_base(uiop);
3269                         dp->d_type = DT_DIR;
3270                         dp->d_fileno = dotdotfileid;
3271                         dp->d_namlen = 2;
3272                         dp->d_name[0] = '.';
3273                         dp->d_name[1] = '.';
3274                         dp->d_name[2] = '\0';
3275                         dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3276                         /*
3277                          * Just make these offset cookie 0.
3278                          */
3279                         tl = (u_int32_t *)&dp->d_name[4];
3280                         *tl++ = 0;
3281                         *tl = 0;
3282                         blksiz += dp->d_reclen;
3283                         uio_uio_resid_add(uiop, -(dp->d_reclen));
3284                         uiop->uio_offset += dp->d_reclen;
3285                         uio_iov_base_add(uiop, dp->d_reclen);
3286                         uio_iov_len_add(uiop, -(dp->d_reclen));
3287                 }
3288                 NFSREADDIRPLUS_ATTRBIT(&attrbits);
3289                 if (gotmnton)
3290                         NFSSETBIT_ATTRBIT(&attrbits,
3291                             NFSATTRBIT_MOUNTEDONFILEID);
3292         }
3293
3294         /*
3295          * Loop around doing readdir rpc's of size nm_readdirsize.
3296          * The stopping criteria is EOF or buffer full.
3297          */
3298         while (more_dirs && bigenough) {
3299                 *attrflagp = 0;
3300                 NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3301                 NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3302                 *tl++ = cookie.lval[0];
3303                 *tl++ = cookie.lval[1];
3304                 if (cookie.qval == 0) {
3305                         *tl++ = 0;
3306                         *tl++ = 0;
3307                 } else {
3308                         NFSLOCKNODE(dnp);
3309                         *tl++ = dnp->n_cookieverf.nfsuquad[0];
3310                         *tl++ = dnp->n_cookieverf.nfsuquad[1];
3311                         NFSUNLOCKNODE(dnp);
3312                 }
3313                 *tl++ = txdr_unsigned(nmp->nm_readdirsize);
3314                 *tl = txdr_unsigned(nmp->nm_readdirsize);
3315                 if (nd->nd_flag & ND_NFSV4) {
3316                         (void) nfsrv_putattrbit(nd, &attrbits);
3317                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3318                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
3319                         (void) nfsrv_putattrbit(nd, &dattrbits);
3320                 }
3321                 error = nfscl_request(nd, vp, p, cred, stuff);
3322                 if (error)
3323                         return (error);
3324                 if (nd->nd_flag & ND_NFSV3)
3325                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3326                 if (nd->nd_repstat || error) {
3327                         if (!error)
3328                                 error = nd->nd_repstat;
3329                         goto nfsmout;
3330                 }
3331                 if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3332                         dctime = nap->na_ctime;
3333                 NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3334                 NFSLOCKNODE(dnp);
3335                 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3336                 dnp->n_cookieverf.nfsuquad[1] = *tl++;
3337                 NFSUNLOCKNODE(dnp);
3338                 more_dirs = fxdr_unsigned(int, *tl);
3339                 if (!more_dirs)
3340                         tryformoredirs = 0;
3341
3342                 /* loop thru the dir entries, doctoring them to 4bsd form */
3343                 while (more_dirs && bigenough) {
3344                         NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3345                         if (nd->nd_flag & ND_NFSV4) {
3346                                 ncookie.lval[0] = *tl++;
3347                                 ncookie.lval[1] = *tl++;
3348                         } else {
3349                                 fileno = fxdr_unsigned(long, *++tl);
3350                                 tl++;
3351                         }
3352                         len = fxdr_unsigned(int, *tl);
3353                         if (len <= 0 || len > NFS_MAXNAMLEN) {
3354                                 error = EBADRPC;
3355                                 goto nfsmout;
3356                         }
3357                         tlen = NFSM_RNDUP(len);
3358                         if (tlen == len)
3359                                 tlen += 4;  /* To ensure null termination */
3360                         left = DIRBLKSIZ - blksiz;
3361                         if ((tlen + DIRHDSIZ + NFSX_HYPER) > left) {
3362                                 dp->d_reclen += left;
3363                                 uio_iov_base_add(uiop, left);
3364                                 uio_iov_len_add(uiop, -(left));
3365                                 uio_uio_resid_add(uiop, -(left));
3366                                 uiop->uio_offset += left;
3367                                 blksiz = 0;
3368                         }
3369                         if ((tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
3370                                 bigenough = 0;
3371                         if (bigenough) {
3372                                 dp = (struct dirent *)uio_iov_base(uiop);
3373                                 dp->d_namlen = len;
3374                                 dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
3375                                 dp->d_type = DT_UNKNOWN;
3376                                 blksiz += dp->d_reclen;
3377                                 if (blksiz == DIRBLKSIZ)
3378                                         blksiz = 0;
3379                                 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3380                                 uiop->uio_offset += DIRHDSIZ;
3381                                 uio_iov_base_add(uiop, DIRHDSIZ);
3382                                 uio_iov_len_add(uiop, -(DIRHDSIZ));
3383                                 cnp->cn_nameptr = uio_iov_base(uiop);
3384                                 cnp->cn_namelen = len;
3385                                 NFSCNHASHZERO(cnp);
3386                                 error = nfsm_mbufuio(nd, uiop, len);
3387                                 if (error)
3388                                         goto nfsmout;
3389                                 cp = uio_iov_base(uiop);
3390                                 tlen -= len;
3391                                 *cp = '\0';
3392                                 cp += tlen;     /* points to cookie storage */
3393                                 tl2 = (u_int32_t *)cp;
3394                                 if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3395                                     cnp->cn_nameptr[1] == '.')
3396                                         isdotdot = 1;
3397                                 else
3398                                         isdotdot = 0;
3399                                 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3400                                 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3401                                 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3402                                 uiop->uio_offset += (tlen + NFSX_HYPER);
3403                         } else {
3404                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3405                                 if (error)
3406                                         goto nfsmout;
3407                         }
3408                         nfhp = NULL;
3409                         if (nd->nd_flag & ND_NFSV3) {
3410                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3411                                 ncookie.lval[0] = *tl++;
3412                                 ncookie.lval[1] = *tl++;
3413                                 attrflag = fxdr_unsigned(int, *tl);
3414                                 if (attrflag) {
3415                                   error = nfsm_loadattr(nd, &nfsva);
3416                                   if (error)
3417                                         goto nfsmout;
3418                                 }
3419                                 NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3420                                 if (*tl) {
3421                                         error = nfsm_getfh(nd, &nfhp);
3422                                         if (error)
3423                                             goto nfsmout;
3424                                 }
3425                                 if (!attrflag && nfhp != NULL) {
3426                                         FREE((caddr_t)nfhp, M_NFSFH);
3427                                         nfhp = NULL;
3428                                 }
3429                         } else {
3430                                 rderr = 0;
3431                                 nfsva.na_mntonfileno = 0xffffffff;
3432                                 error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3433                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3434                                     NULL, NULL, &rderr, p, cred);
3435                                 if (error)
3436                                         goto nfsmout;
3437                         }
3438
3439                         if (bigenough) {
3440                             if (nd->nd_flag & ND_NFSV4) {
3441                                 if (rderr) {
3442                                     dp->d_fileno = 0;
3443                                 } else if (gotmnton) {
3444                                     if (nfsva.na_mntonfileno != 0xffffffff)
3445                                         dp->d_fileno = nfsva.na_mntonfileno;
3446                                     else
3447                                         dp->d_fileno = nfsva.na_fileid;
3448                                 } else if (nfsva.na_filesid[0] ==
3449                                     dnp->n_vattr.na_filesid[0] &&
3450                                     nfsva.na_filesid[1] ==
3451                                     dnp->n_vattr.na_filesid[1]) {
3452                                     dp->d_fileno = nfsva.na_fileid;
3453                                 } else {
3454                                     do {
3455                                         fakefileno--;
3456                                     } while (fakefileno ==
3457                                         nfsva.na_fileid);
3458                                     dp->d_fileno = fakefileno;
3459                                 }
3460                             } else {
3461                                 dp->d_fileno = fileno;
3462                             }
3463                             *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3464                                 ncookie.lval[0];
3465                             *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3466                                 ncookie.lval[1];
3467
3468                             if (nfhp != NULL) {
3469                                 if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3470                                     dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3471                                     VREF(vp);
3472                                     newvp = vp;
3473                                     unlocknewvp = 0;
3474                                     FREE((caddr_t)nfhp, M_NFSFH);
3475                                     np = dnp;
3476                                 } else if (isdotdot != 0) {
3477                                     /*
3478                                      * Skip doing a nfscl_nget() call for "..".
3479                                      * There's a race between acquiring the nfs
3480                                      * node here and lookups that look for the
3481                                      * directory being read (in the parent).
3482                                      * It would try to get a lock on ".." here,
3483                                      * owning the lock on the directory being
3484                                      * read. Lookup will hold the lock on ".."
3485                                      * and try to acquire the lock on the
3486                                      * directory being read.
3487                                      * If the directory is unlocked/relocked,
3488                                      * then there is a LOR with the buflock
3489                                      * vp is relocked.
3490                                      */
3491                                     free(nfhp, M_NFSFH);
3492                                 } else {
3493                                     error = nfscl_nget(vnode_mount(vp), vp,
3494                                       nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3495                                     if (!error) {
3496                                         newvp = NFSTOV(np);
3497                                         unlocknewvp = 1;
3498                                     }
3499                                 }
3500                                 nfhp = NULL;
3501                                 if (newvp != NULLVP) {
3502                                     error = nfscl_loadattrcache(&newvp,
3503                                         &nfsva, NULL, NULL, 0, 0);
3504                                     if (error) {
3505                                         if (unlocknewvp)
3506                                             vput(newvp);
3507                                         else
3508                                             vrele(newvp);
3509                                         goto nfsmout;
3510                                     }
3511                                     dp->d_type =
3512                                         vtonfs_dtype(np->n_vattr.na_type);
3513                                     ndp->ni_vp = newvp;
3514                                     NFSCNHASH(cnp, HASHINIT);
3515                                     if (cnp->cn_namelen <= NCHNAMLEN &&
3516                                         (newvp->v_type != VDIR ||
3517                                          dctime.tv_sec != 0)) {
3518                                         cache_enter_time(ndp->ni_dvp,
3519                                             ndp->ni_vp, cnp,
3520                                             &nfsva.na_ctime,
3521                                             newvp->v_type != VDIR ? NULL :
3522                                             &dctime);
3523                                     }
3524                                     if (unlocknewvp)
3525                                         vput(newvp);
3526                                     else
3527                                         vrele(newvp);
3528                                     newvp = NULLVP;
3529                                 }
3530                             }
3531                         } else if (nfhp != NULL) {
3532                             FREE((caddr_t)nfhp, M_NFSFH);
3533                         }
3534                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3535                         more_dirs = fxdr_unsigned(int, *tl);
3536                 }
3537                 /*
3538                  * If at end of rpc data, get the eof boolean
3539                  */
3540                 if (!more_dirs) {
3541                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3542                         eof = fxdr_unsigned(int, *tl);
3543                         if (tryformoredirs)
3544                                 more_dirs = !eof;
3545                         if (nd->nd_flag & ND_NFSV4) {
3546                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3547                                     stuff);
3548                                 if (error)
3549                                         goto nfsmout;
3550                         }
3551                 }
3552                 mbuf_freem(nd->nd_mrep);
3553                 nd->nd_mrep = NULL;
3554         }
3555         /*
3556          * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3557          * by increasing d_reclen for the last record.
3558          */
3559         if (blksiz > 0) {
3560                 left = DIRBLKSIZ - blksiz;
3561                 dp->d_reclen += left;
3562                 uio_iov_base_add(uiop, left);
3563                 uio_iov_len_add(uiop, -(left));
3564                 uio_uio_resid_add(uiop, -(left));
3565                 uiop->uio_offset += left;
3566         }
3567
3568         /*
3569          * If returning no data, assume end of file.
3570          * If not bigenough, return not end of file, since you aren't
3571          *    returning all the data
3572          * Otherwise, return the eof flag from the server.
3573          */
3574         if (eofp != NULL) {
3575                 if (tresid == uio_uio_resid(uiop))
3576                         *eofp = 1;
3577                 else if (!bigenough)
3578                         *eofp = 0;
3579                 else
3580                         *eofp = eof;
3581         }
3582
3583         /*
3584          * Add extra empty records to any remaining DIRBLKSIZ chunks.
3585          */
3586         while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3587                 dp = (struct dirent *)uio_iov_base(uiop);
3588                 dp->d_type = DT_UNKNOWN;
3589                 dp->d_fileno = 0;
3590                 dp->d_namlen = 0;
3591                 dp->d_name[0] = '\0';
3592                 tl = (u_int32_t *)&dp->d_name[4];
3593                 *tl++ = cookie.lval[0];
3594                 *tl = cookie.lval[1];
3595                 dp->d_reclen = DIRBLKSIZ;
3596                 uio_iov_base_add(uiop, DIRBLKSIZ);
3597                 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3598                 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3599                 uiop->uio_offset += DIRBLKSIZ;
3600         }
3601
3602 nfsmout:
3603         if (nd->nd_mrep != NULL)
3604                 mbuf_freem(nd->nd_mrep);
3605         return (error);
3606 }
3607 #endif  /* !APPLE */
3608
3609 /*
3610  * Nfs commit rpc
3611  */
3612 APPLESTATIC int
3613 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3614     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3615 {
3616         u_int32_t *tl;
3617         struct nfsrv_descript nfsd, *nd = &nfsd;
3618         nfsattrbit_t attrbits;
3619         int error;
3620         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3621
3622         *attrflagp = 0;
3623         NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3624         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3625         txdr_hyper(offset, tl);
3626         tl += 2;
3627         *tl = txdr_unsigned(cnt);
3628         if (nd->nd_flag & ND_NFSV4) {
3629                 /*
3630                  * And do a Getattr op.
3631                  */
3632                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3633                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3634                 NFSGETATTR_ATTRBIT(&attrbits);
3635                 (void) nfsrv_putattrbit(nd, &attrbits);
3636         }
3637         error = nfscl_request(nd, vp, p, cred, stuff);
3638         if (error)
3639                 return (error);
3640         error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3641         if (!error && !nd->nd_repstat) {
3642                 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3643                 NFSLOCKMNT(nmp);
3644                 if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3645                         NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3646                         nd->nd_repstat = NFSERR_STALEWRITEVERF;
3647                 }
3648                 NFSUNLOCKMNT(nmp);
3649                 if (nd->nd_flag & ND_NFSV4)
3650                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3651         }
3652 nfsmout:
3653         if (!error && nd->nd_repstat)
3654                 error = nd->nd_repstat;
3655         mbuf_freem(nd->nd_mrep);
3656         return (error);
3657 }
3658
3659 /*
3660  * NFS byte range lock rpc.
3661  * (Mostly just calls one of the three lower level RPC routines.)
3662  */
3663 APPLESTATIC int
3664 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3665     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3666 {
3667         struct nfscllockowner *lp;
3668         struct nfsclclient *clp;
3669         struct nfsfh *nfhp;
3670         struct nfsrv_descript nfsd, *nd = &nfsd;
3671         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3672         u_int64_t off, len;
3673         off_t start, end;
3674         u_int32_t clidrev = 0;
3675         int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3676         int callcnt, dorpc;
3677
3678         /*
3679          * Convert the flock structure into a start and end and do POSIX
3680          * bounds checking.
3681          */
3682         switch (fl->l_whence) {
3683         case SEEK_SET:
3684         case SEEK_CUR:
3685                 /*
3686                  * Caller is responsible for adding any necessary offset
3687                  * when SEEK_CUR is used.
3688                  */
3689                 start = fl->l_start;
3690                 off = fl->l_start;
3691                 break;
3692         case SEEK_END:
3693                 start = size + fl->l_start;
3694                 off = size + fl->l_start;
3695                 break;
3696         default:
3697                 return (EINVAL);
3698         };
3699         if (start < 0)
3700                 return (EINVAL);
3701         if (fl->l_len != 0) {
3702                 end = start + fl->l_len - 1;
3703                 if (end < start)
3704                         return (EINVAL);
3705         }
3706
3707         len = fl->l_len;
3708         if (len == 0)
3709                 len = NFS64BITSSET;
3710         retrycnt = 0;
3711         do {
3712             nd->nd_repstat = 0;
3713             if (op == F_GETLK) {
3714                 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3715                 if (error)
3716                         return (error);
3717                 error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3718                 if (!error) {
3719                         clidrev = clp->nfsc_clientidrev;
3720                         error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3721                             p, id, flags);
3722                 } else if (error == -1) {
3723                         error = 0;
3724                 }
3725                 nfscl_clientrelease(clp);
3726             } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3727                 /*
3728                  * We must loop around for all lockowner cases.
3729                  */
3730                 callcnt = 0;
3731                 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3732                 if (error)
3733                         return (error);
3734                 do {
3735                     error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3736                         clp, id, flags, &lp, &dorpc);
3737                     /*
3738                      * If it returns a NULL lp, we're done.
3739                      */
3740                     if (lp == NULL) {
3741                         if (callcnt == 0)
3742                             nfscl_clientrelease(clp);
3743                         else
3744                             nfscl_releasealllocks(clp, vp, p, id, flags);
3745                         return (error);
3746                     }
3747                     if (nmp->nm_clp != NULL)
3748                         clidrev = nmp->nm_clp->nfsc_clientidrev;
3749                     else
3750                         clidrev = 0;
3751                     /*
3752                      * If the server doesn't support Posix lock semantics,
3753                      * only allow locks on the entire file, since it won't
3754                      * handle overlapping byte ranges.
3755                      * There might still be a problem when a lock
3756                      * upgrade/downgrade (read<->write) occurs, since the
3757                      * server "might" expect an unlock first?
3758                      */
3759                     if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3760                         (off == 0 && len == NFS64BITSSET))) {
3761                         /*
3762                          * Since the lock records will go away, we must
3763                          * wait for grace and delay here.
3764                          */
3765                         do {
3766                             error = nfsrpc_locku(nd, nmp, lp, off, len,
3767                                 NFSV4LOCKT_READ, cred, p, 0);
3768                             if ((nd->nd_repstat == NFSERR_GRACE ||
3769                                  nd->nd_repstat == NFSERR_DELAY) &&
3770                                 error == 0)
3771                                 (void) nfs_catnap(PZERO, (int)nd->nd_repstat,
3772                                     "nfs_advlock");
3773                         } while ((nd->nd_repstat == NFSERR_GRACE ||
3774                             nd->nd_repstat == NFSERR_DELAY) && error == 0);
3775                     }
3776                     callcnt++;
3777                 } while (error == 0 && nd->nd_repstat == 0);
3778                 nfscl_releasealllocks(clp, vp, p, id, flags);
3779             } else if (op == F_SETLK) {
3780                 error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
3781                     NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
3782                 if (error || donelocally) {
3783                         return (error);
3784                 }
3785                 if (nmp->nm_clp != NULL)
3786                         clidrev = nmp->nm_clp->nfsc_clientidrev;
3787                 else
3788                         clidrev = 0;
3789                 nfhp = VTONFS(vp)->n_fhp;
3790                 if (!lp->nfsl_open->nfso_posixlock &&
3791                     (off != 0 || len != NFS64BITSSET)) {
3792                         error = EINVAL;
3793                 } else {
3794                         error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
3795                             nfhp->nfh_len, lp, newone, reclaim, off,
3796                             len, fl->l_type, cred, p, 0);
3797                 }
3798                 if (!error)
3799                         error = nd->nd_repstat;
3800                 nfscl_lockrelease(lp, error, newone);
3801             } else {
3802                 error = EINVAL;
3803             }
3804             if (!error)
3805                 error = nd->nd_repstat;
3806             if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
3807                 error == NFSERR_STALEDONTRECOVER ||
3808                 error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3809                 error == NFSERR_BADSESSION) {
3810                 (void) nfs_catnap(PZERO, error, "nfs_advlock");
3811             } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
3812                 && clidrev != 0) {
3813                 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
3814                 retrycnt++;
3815             }
3816         } while (error == NFSERR_GRACE ||
3817             error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3818             error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
3819             error == NFSERR_BADSESSION ||
3820             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
3821              expireret == 0 && clidrev != 0 && retrycnt < 4));
3822         if (error && retrycnt >= 4)
3823                 error = EIO;
3824         return (error);
3825 }
3826
3827 /*
3828  * The lower level routine for the LockT case.
3829  */
3830 APPLESTATIC int
3831 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
3832     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
3833     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3834 {
3835         u_int32_t *tl;
3836         int error, type, size;
3837         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3838         struct nfsnode *np;
3839         struct nfsmount *nmp;
3840
3841         nmp = VFSTONFS(vp->v_mount);
3842         NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
3843         NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3844         if (fl->l_type == F_RDLCK)
3845                 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3846         else
3847                 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3848         txdr_hyper(off, tl);
3849         tl += 2;
3850         txdr_hyper(len, tl);
3851         tl += 2;
3852         *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
3853         *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
3854         nfscl_filllockowner(id, own, flags);
3855         np = VTONFS(vp);
3856         NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
3857             np->n_fhp->nfh_len);
3858         (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
3859         error = nfscl_request(nd, vp, p, cred, NULL);
3860         if (error)
3861                 return (error);
3862         if (nd->nd_repstat == 0) {
3863                 fl->l_type = F_UNLCK;
3864         } else if (nd->nd_repstat == NFSERR_DENIED) {
3865                 nd->nd_repstat = 0;
3866                 fl->l_whence = SEEK_SET;
3867                 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
3868                 fl->l_start = fxdr_hyper(tl);
3869                 tl += 2;
3870                 len = fxdr_hyper(tl);
3871                 tl += 2;
3872                 if (len == NFS64BITSSET)
3873                         fl->l_len = 0;
3874                 else
3875                         fl->l_len = len;
3876                 type = fxdr_unsigned(int, *tl++);
3877                 if (type == NFSV4LOCKT_WRITE)
3878                         fl->l_type = F_WRLCK;
3879                 else
3880                         fl->l_type = F_RDLCK;
3881                 /*
3882                  * XXX For now, I have no idea what to do with the
3883                  * conflicting lock_owner, so I'll just set the pid == 0
3884                  * and skip over the lock_owner.
3885                  */
3886                 fl->l_pid = (pid_t)0;
3887                 tl += 2;
3888                 size = fxdr_unsigned(int, *tl);
3889                 if (size < 0 || size > NFSV4_OPAQUELIMIT)
3890                         error = EBADRPC;
3891                 if (!error)
3892                         error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
3893         } else if (nd->nd_repstat == NFSERR_STALECLIENTID ||
3894             nd->nd_repstat == NFSERR_BADSESSION)
3895                 nfscl_initiate_recovery(clp);
3896 nfsmout:
3897         mbuf_freem(nd->nd_mrep);
3898         return (error);
3899 }
3900
3901 /*
3902  * Lower level function that performs the LockU RPC.
3903  */
3904 static int
3905 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
3906     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
3907     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
3908 {
3909         u_int32_t *tl;
3910         int error;
3911
3912         nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
3913             lp->nfsl_open->nfso_fhlen, NULL, NULL);
3914         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
3915         *tl++ = txdr_unsigned(type);
3916         *tl = txdr_unsigned(lp->nfsl_seqid);
3917         if (nfstest_outofseq &&
3918             (arc4random() % nfstest_outofseq) == 0)
3919                 *tl = txdr_unsigned(lp->nfsl_seqid + 1);
3920         tl++;
3921         if (NFSHASNFSV4N(nmp))
3922                 *tl++ = 0;
3923         else
3924                 *tl++ = lp->nfsl_stateid.seqid;
3925         *tl++ = lp->nfsl_stateid.other[0];
3926         *tl++ = lp->nfsl_stateid.other[1];
3927         *tl++ = lp->nfsl_stateid.other[2];
3928         txdr_hyper(off, tl);
3929         tl += 2;
3930         txdr_hyper(len, tl);
3931         if (syscred)
3932                 nd->nd_flag |= ND_USEGSSNAME;
3933         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
3934             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
3935         NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
3936         if (error)
3937                 return (error);
3938         if (nd->nd_repstat == 0) {
3939                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
3940                 lp->nfsl_stateid.seqid = *tl++;
3941                 lp->nfsl_stateid.other[0] = *tl++;
3942                 lp->nfsl_stateid.other[1] = *tl++;
3943                 lp->nfsl_stateid.other[2] = *tl;
3944         } else if (nd->nd_repstat == NFSERR_STALESTATEID ||
3945             nd->nd_repstat == NFSERR_BADSESSION)
3946                 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
3947 nfsmout:
3948         mbuf_freem(nd->nd_mrep);
3949         return (error);
3950 }
3951
3952 /*
3953  * The actual Lock RPC.
3954  */
3955 APPLESTATIC int
3956 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
3957     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
3958     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
3959     NFSPROC_T *p, int syscred)
3960 {
3961         u_int32_t *tl;
3962         int error, size;
3963         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3964
3965         nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL);
3966         NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3967         if (type == F_RDLCK)
3968                 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3969         else
3970                 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3971         *tl++ = txdr_unsigned(reclaim);
3972         txdr_hyper(off, tl);
3973         tl += 2;
3974         txdr_hyper(len, tl);
3975         tl += 2;
3976         if (newone) {
3977             *tl = newnfs_true;
3978             NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
3979                 2 * NFSX_UNSIGNED + NFSX_HYPER);
3980             *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
3981             if (NFSHASNFSV4N(nmp))
3982                 *tl++ = 0;
3983             else
3984                 *tl++ = lp->nfsl_open->nfso_stateid.seqid;
3985             *tl++ = lp->nfsl_open->nfso_stateid.other[0];
3986             *tl++ = lp->nfsl_open->nfso_stateid.other[1];
3987             *tl++ = lp->nfsl_open->nfso_stateid.other[2];
3988             *tl++ = txdr_unsigned(lp->nfsl_seqid);
3989             *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
3990             *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
3991             NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
3992             NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
3993             (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
3994         } else {
3995             *tl = newnfs_false;
3996             NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
3997             if (NFSHASNFSV4N(nmp))
3998                 *tl++ = 0;
3999             else
4000                 *tl++ = lp->nfsl_stateid.seqid;
4001             *tl++ = lp->nfsl_stateid.other[0];
4002             *tl++ = lp->nfsl_stateid.other[1];
4003             *tl++ = lp->nfsl_stateid.other[2];
4004             *tl = txdr_unsigned(lp->nfsl_seqid);
4005             if (nfstest_outofseq &&
4006                 (arc4random() % nfstest_outofseq) == 0)
4007                     *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4008         }
4009         if (syscred)
4010                 nd->nd_flag |= ND_USEGSSNAME;
4011         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4012             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4013         if (error)
4014                 return (error);
4015         if (newone)
4016             NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4017         NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4018         if (nd->nd_repstat == 0) {
4019                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4020                 lp->nfsl_stateid.seqid = *tl++;
4021                 lp->nfsl_stateid.other[0] = *tl++;
4022                 lp->nfsl_stateid.other[1] = *tl++;
4023                 lp->nfsl_stateid.other[2] = *tl;
4024         } else if (nd->nd_repstat == NFSERR_DENIED) {
4025                 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4026                 size = fxdr_unsigned(int, *(tl + 7));
4027                 if (size < 0 || size > NFSV4_OPAQUELIMIT)
4028                         error = EBADRPC;
4029                 if (!error)
4030                         error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4031         } else if (nd->nd_repstat == NFSERR_STALESTATEID ||
4032             nd->nd_repstat == NFSERR_BADSESSION)
4033                 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4034 nfsmout:
4035         mbuf_freem(nd->nd_mrep);
4036         return (error);
4037 }
4038
4039 /*
4040  * nfs statfs rpc
4041  * (always called with the vp for the mount point)
4042  */
4043 APPLESTATIC int
4044 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4045     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4046     void *stuff)
4047 {
4048         u_int32_t *tl = NULL;
4049         struct nfsrv_descript nfsd, *nd = &nfsd;
4050         struct nfsmount *nmp;
4051         nfsattrbit_t attrbits;
4052         int error;
4053
4054         *attrflagp = 0;
4055         nmp = VFSTONFS(vnode_mount(vp));
4056         if (NFSHASNFSV4(nmp)) {
4057                 /*
4058                  * For V4, you actually do a getattr.
4059                  */
4060                 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4061                 NFSSTATFS_GETATTRBIT(&attrbits);
4062                 (void) nfsrv_putattrbit(nd, &attrbits);
4063                 nd->nd_flag |= ND_USEGSSNAME;
4064                 error = nfscl_request(nd, vp, p, cred, stuff);
4065                 if (error)
4066                         return (error);
4067                 if (nd->nd_repstat == 0) {
4068                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4069                             NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4070                             cred);
4071                         if (!error) {
4072                                 nmp->nm_fsid[0] = nap->na_filesid[0];
4073                                 nmp->nm_fsid[1] = nap->na_filesid[1];
4074                                 NFSSETHASSETFSID(nmp);
4075                                 *attrflagp = 1;
4076                         }
4077                 } else {
4078                         error = nd->nd_repstat;
4079                 }
4080                 if (error)
4081                         goto nfsmout;
4082         } else {
4083                 NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4084                 error = nfscl_request(nd, vp, p, cred, stuff);
4085                 if (error)
4086                         return (error);
4087                 if (nd->nd_flag & ND_NFSV3) {
4088                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4089                         if (error)
4090                                 goto nfsmout;
4091                 }
4092                 if (nd->nd_repstat) {
4093                         error = nd->nd_repstat;
4094                         goto nfsmout;
4095                 }
4096                 NFSM_DISSECT(tl, u_int32_t *,
4097                     NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4098         }
4099         if (NFSHASNFSV3(nmp)) {
4100                 sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4101                 sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4102                 sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4103                 sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4104                 sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4105                 sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4106                 sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4107         } else if (NFSHASNFSV4(nmp) == 0) {
4108                 sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4109                 sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4110                 sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4111                 sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4112                 sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4113         }
4114 nfsmout:
4115         mbuf_freem(nd->nd_mrep);
4116         return (error);
4117 }
4118
4119 /*
4120  * nfs pathconf rpc
4121  */
4122 APPLESTATIC int
4123 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4124     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4125     void *stuff)
4126 {
4127         struct nfsrv_descript nfsd, *nd = &nfsd;
4128         struct nfsmount *nmp;
4129         u_int32_t *tl;
4130         nfsattrbit_t attrbits;
4131         int error;
4132
4133         *attrflagp = 0;
4134         nmp = VFSTONFS(vnode_mount(vp));
4135         if (NFSHASNFSV4(nmp)) {
4136                 /*
4137                  * For V4, you actually do a getattr.
4138                  */
4139                 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4140                 NFSPATHCONF_GETATTRBIT(&attrbits);
4141                 (void) nfsrv_putattrbit(nd, &attrbits);
4142                 nd->nd_flag |= ND_USEGSSNAME;
4143                 error = nfscl_request(nd, vp, p, cred, stuff);
4144                 if (error)
4145                         return (error);
4146                 if (nd->nd_repstat == 0) {
4147                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4148                             pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4149                             cred);
4150                         if (!error)
4151                                 *attrflagp = 1;
4152                 } else {
4153                         error = nd->nd_repstat;
4154                 }
4155         } else {
4156                 NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4157                 error = nfscl_request(nd, vp, p, cred, stuff);
4158                 if (error)
4159                         return (error);
4160                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4161                 if (nd->nd_repstat && !error)
4162                         error = nd->nd_repstat;
4163                 if (!error) {
4164                         NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4165                         pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4166                         pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4167                         pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4168                         pc->pc_chownrestricted =
4169                             fxdr_unsigned(u_int32_t, *tl++);
4170                         pc->pc_caseinsensitive =
4171                             fxdr_unsigned(u_int32_t, *tl++);
4172                         pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4173                 }
4174         }
4175 nfsmout:
4176         mbuf_freem(nd->nd_mrep);
4177         return (error);
4178 }
4179
4180 /*
4181  * nfs version 3 fsinfo rpc call
4182  */
4183 APPLESTATIC int
4184 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4185     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4186 {
4187         u_int32_t *tl;
4188         struct nfsrv_descript nfsd, *nd = &nfsd;
4189         int error;
4190
4191         *attrflagp = 0;
4192         NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4193         error = nfscl_request(nd, vp, p, cred, stuff);
4194         if (error)
4195                 return (error);
4196         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4197         if (nd->nd_repstat && !error)
4198                 error = nd->nd_repstat;
4199         if (!error) {
4200                 NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4201                 fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4202                 fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4203                 fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4204                 fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4205                 fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4206                 fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4207                 fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4208                 fsp->fs_maxfilesize = fxdr_hyper(tl);
4209                 tl += 2;
4210                 fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4211                 tl += 2;
4212                 fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4213         }
4214 nfsmout:
4215         mbuf_freem(nd->nd_mrep);
4216         return (error);
4217 }
4218
4219 /*
4220  * This function performs the Renew RPC.
4221  */
4222 APPLESTATIC int
4223 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4224     NFSPROC_T *p)
4225 {
4226         u_int32_t *tl;
4227         struct nfsrv_descript nfsd;
4228         struct nfsrv_descript *nd = &nfsd;
4229         struct nfsmount *nmp;
4230         int error;
4231         struct nfssockreq *nrp;
4232
4233         nmp = clp->nfsc_nmp;
4234         if (nmp == NULL)
4235                 return (0);
4236         nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4237             &dsp->nfsclds_sess);
4238         if (!NFSHASNFSV4N(nmp)) {
4239                 /* NFSv4.1 just uses a Sequence Op and not a Renew. */
4240                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4241                 *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4242                 *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4243         }
4244         nrp = dsp->nfsclds_sockp;
4245         if (nrp == NULL)
4246                 /* If NULL, use the MDS socket. */
4247                 nrp = &nmp->nm_sockreq;
4248         nd->nd_flag |= ND_USEGSSNAME;
4249         error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4250             NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4251         if (error)
4252                 return (error);
4253         error = nd->nd_repstat;
4254         mbuf_freem(nd->nd_mrep);
4255         return (error);
4256 }
4257
4258 /*
4259  * This function performs the Releaselockowner RPC.
4260  */
4261 APPLESTATIC int
4262 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4263     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4264 {
4265         struct nfsrv_descript nfsd, *nd = &nfsd;
4266         u_int32_t *tl;
4267         int error;
4268         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4269
4270         if (NFSHASNFSV4N(nmp)) {
4271                 /* For NFSv4.1, do a FreeStateID. */
4272                 nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4273                     NULL);
4274                 nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4275         } else {
4276                 nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4277                     NULL);
4278                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4279                 *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4280                 *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4281                 NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4282                 NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4283                 (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4284         }
4285         nd->nd_flag |= ND_USEGSSNAME;
4286         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4287             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4288         if (error)
4289                 return (error);
4290         error = nd->nd_repstat;
4291         mbuf_freem(nd->nd_mrep);
4292         return (error);
4293 }
4294
4295 /*
4296  * This function performs the Compound to get the mount pt FH.
4297  */
4298 APPLESTATIC int
4299 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4300     NFSPROC_T *p)
4301 {
4302         u_int32_t *tl;
4303         struct nfsrv_descript nfsd;
4304         struct nfsrv_descript *nd = &nfsd;
4305         u_char *cp, *cp2;
4306         int error, cnt, len, setnil;
4307         u_int32_t *opcntp;
4308
4309         nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL);
4310         cp = dirpath;
4311         cnt = 0;
4312         do {
4313                 setnil = 0;
4314                 while (*cp == '/')
4315                         cp++;
4316                 cp2 = cp;
4317                 while (*cp2 != '\0' && *cp2 != '/')
4318                         cp2++;
4319                 if (*cp2 == '/') {
4320                         setnil = 1;
4321                         *cp2 = '\0';
4322                 }
4323                 if (cp2 != cp) {
4324                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4325                         *tl = txdr_unsigned(NFSV4OP_LOOKUP);
4326                         nfsm_strtom(nd, cp, strlen(cp));
4327                         cnt++;
4328                 }
4329                 if (setnil)
4330                         *cp2++ = '/';
4331                 cp = cp2;
4332         } while (*cp != '\0');
4333         if (NFSHASNFSV4N(nmp))
4334                 /* Has a Sequence Op done by nfscl_reqstart(). */
4335                 *opcntp = txdr_unsigned(3 + cnt);
4336         else
4337                 *opcntp = txdr_unsigned(2 + cnt);
4338         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4339         *tl = txdr_unsigned(NFSV4OP_GETFH);
4340         nd->nd_flag |= ND_USEGSSNAME;
4341         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4342                 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4343         if (error)
4344                 return (error);
4345         if (nd->nd_repstat == 0) {
4346                 NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4347                 tl += (2 + 2 * cnt);
4348                 if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4349                         len > NFSX_FHMAX) {
4350                         nd->nd_repstat = NFSERR_BADXDR;
4351                 } else {
4352                         nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4353                         if (nd->nd_repstat == 0)
4354                                 nmp->nm_fhsize = len;
4355                 }
4356         }
4357         error = nd->nd_repstat;
4358 nfsmout:
4359         mbuf_freem(nd->nd_mrep);
4360         return (error);
4361 }
4362
4363 /*
4364  * This function performs the Delegreturn RPC.
4365  */
4366 APPLESTATIC int
4367 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4368     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4369 {
4370         u_int32_t *tl;
4371         struct nfsrv_descript nfsd;
4372         struct nfsrv_descript *nd = &nfsd;
4373         int error;
4374
4375         nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4376             dp->nfsdl_fhlen, NULL, NULL);
4377         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4378         if (NFSHASNFSV4N(nmp))
4379                 *tl++ = 0;
4380         else
4381                 *tl++ = dp->nfsdl_stateid.seqid;
4382         *tl++ = dp->nfsdl_stateid.other[0];
4383         *tl++ = dp->nfsdl_stateid.other[1];
4384         *tl = dp->nfsdl_stateid.other[2];
4385         if (syscred)
4386                 nd->nd_flag |= ND_USEGSSNAME;
4387         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4388             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4389         if (error)
4390                 return (error);
4391         error = nd->nd_repstat;
4392         mbuf_freem(nd->nd_mrep);
4393         return (error);
4394 }
4395
4396 /*
4397  * nfs getacl call.
4398  */
4399 APPLESTATIC int
4400 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4401     struct acl *aclp, void *stuff)
4402 {
4403         struct nfsrv_descript nfsd, *nd = &nfsd;
4404         int error;
4405         nfsattrbit_t attrbits;
4406         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4407
4408         if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4409                 return (EOPNOTSUPP);
4410         NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4411         NFSZERO_ATTRBIT(&attrbits);
4412         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4413         (void) nfsrv_putattrbit(nd, &attrbits);
4414         error = nfscl_request(nd, vp, p, cred, stuff);
4415         if (error)
4416                 return (error);
4417         if (!nd->nd_repstat)
4418                 error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4419                     NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4420         else
4421                 error = nd->nd_repstat;
4422         mbuf_freem(nd->nd_mrep);
4423         return (error);
4424 }
4425
4426 /*
4427  * nfs setacl call.
4428  */
4429 APPLESTATIC int
4430 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4431     struct acl *aclp, void *stuff)
4432 {
4433         int error;
4434         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4435
4436         if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4437                 return (EOPNOTSUPP);
4438         error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4439         return (error);
4440 }
4441
4442 /*
4443  * nfs setacl call.
4444  */
4445 static int
4446 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4447     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4448 {
4449         struct nfsrv_descript nfsd, *nd = &nfsd;
4450         int error;
4451         nfsattrbit_t attrbits;
4452         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4453
4454         if (!NFSHASNFSV4(nmp))
4455                 return (EOPNOTSUPP);
4456         NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4457         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4458         NFSZERO_ATTRBIT(&attrbits);
4459         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4460         (void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4461             &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0);
4462         error = nfscl_request(nd, vp, p, cred, stuff);
4463         if (error)
4464                 return (error);
4465         /* Don't care about the pre/postop attributes */
4466         mbuf_freem(nd->nd_mrep);
4467         return (nd->nd_repstat);
4468 }
4469
4470 /*
4471  * Do the NFSv4.1 Exchange ID.
4472  */
4473 int
4474 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4475     struct nfssockreq *nrp, uint32_t exchflags, struct nfsclds **dspp,
4476     struct ucred *cred, NFSPROC_T *p)
4477 {
4478         uint32_t *tl, v41flags;
4479         struct nfsrv_descript nfsd;
4480         struct nfsrv_descript *nd = &nfsd;
4481         struct nfsclds *dsp;
4482         struct timespec verstime;
4483         int error, len;
4484
4485         *dspp = NULL;
4486         nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL);
4487         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4488         *tl++ = txdr_unsigned(nfsboottime.tv_sec);      /* Client owner */
4489         *tl = txdr_unsigned(clp->nfsc_rev);
4490         (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4491
4492         NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4493         *tl++ = txdr_unsigned(exchflags);
4494         *tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4495
4496         /* Set the implementation id4 */
4497         *tl = txdr_unsigned(1);
4498         (void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4499         (void) nfsm_strtom(nd, version, strlen(version));
4500         NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4501         verstime.tv_sec = 1293840000;           /* Jan 1, 2011 */
4502         verstime.tv_nsec = 0;
4503         txdr_nfsv4time(&verstime, tl);
4504         nd->nd_flag |= ND_USEGSSNAME;
4505         error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4506             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4507         NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4508             (int)nd->nd_repstat);
4509         if (error != 0)
4510                 return (error);
4511         if (nd->nd_repstat == 0) {
4512                 NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4513                 len = fxdr_unsigned(int, *(tl + 7));
4514                 if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4515                         error = NFSERR_BADXDR;
4516                         goto nfsmout;
4517                 }
4518                 dsp = malloc(sizeof(struct nfsclds) + len, M_NFSCLDS,
4519                     M_WAITOK | M_ZERO);
4520                 dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4521                 dsp->nfsclds_servownlen = len;
4522                 dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4523                 dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4524                 dsp->nfsclds_sess.nfsess_sequenceid =
4525                     fxdr_unsigned(uint32_t, *tl++);
4526                 v41flags = fxdr_unsigned(uint32_t, *tl);
4527                 if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4528                     NFSHASPNFSOPT(nmp)) {
4529                         NFSCL_DEBUG(1, "set PNFS\n");
4530                         NFSLOCKMNT(nmp);
4531                         nmp->nm_state |= NFSSTA_PNFS;
4532                         NFSUNLOCKMNT(nmp);
4533                         dsp->nfsclds_flags |= NFSCLDS_MDS;
4534                 }
4535                 if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4536                         dsp->nfsclds_flags |= NFSCLDS_DS;
4537                 if (len > 0)
4538                         nd->nd_repstat = nfsrv_mtostr(nd,
4539                             dsp->nfsclds_serverown, len);
4540                 if (nd->nd_repstat == 0) {
4541                         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4542                         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4543                             NULL, MTX_DEF);
4544                         nfscl_initsessionslots(&dsp->nfsclds_sess);
4545                         *dspp = dsp;
4546                 } else
4547                         free(dsp, M_NFSCLDS);
4548         }
4549         error = nd->nd_repstat;
4550 nfsmout:
4551         mbuf_freem(nd->nd_mrep);
4552         return (error);
4553 }
4554
4555 /*
4556  * Do the NFSv4.1 Create Session.
4557  */
4558 int
4559 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4560     struct nfssockreq *nrp, uint32_t sequenceid, int mds, struct ucred *cred,
4561     NFSPROC_T *p)
4562 {
4563         uint32_t crflags, *tl;
4564         struct nfsrv_descript nfsd;
4565         struct nfsrv_descript *nd = &nfsd;
4566         int error, irdcnt;
4567
4568         nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL);
4569         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4570         *tl++ = sep->nfsess_clientid.lval[0];
4571         *tl++ = sep->nfsess_clientid.lval[1];
4572         *tl++ = txdr_unsigned(sequenceid);
4573         crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4574         if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0)
4575                 crflags |= NFSV4CRSESS_CONNBACKCHAN;
4576         *tl = txdr_unsigned(crflags);
4577
4578         /* Fill in fore channel attributes. */
4579         NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4580         *tl++ = 0;                              /* Header pad size */
4581         *tl++ = txdr_unsigned(100000);          /* Max request size */
4582         *tl++ = txdr_unsigned(100000);          /* Max response size */
4583         *tl++ = txdr_unsigned(4096);            /* Max response size cached */
4584         *tl++ = txdr_unsigned(20);              /* Max operations */
4585         *tl++ = txdr_unsigned(64);              /* Max slots */
4586         *tl = 0;                                /* No rdma ird */
4587
4588         /* Fill in back channel attributes. */
4589         NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4590         *tl++ = 0;                              /* Header pad size */
4591         *tl++ = txdr_unsigned(10000);           /* Max request size */
4592         *tl++ = txdr_unsigned(10000);           /* Max response size */
4593         *tl++ = txdr_unsigned(4096);            /* Max response size cached */
4594         *tl++ = txdr_unsigned(4);               /* Max operations */
4595         *tl++ = txdr_unsigned(NFSV4_CBSLOTS);   /* Max slots */
4596         *tl = 0;                                /* No rdma ird */
4597
4598         NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4599         *tl++ = txdr_unsigned(NFS_CALLBCKPROG); /* Call back prog # */
4600
4601         /* Allow AUTH_SYS callbacks as uid, gid == 0. */
4602         *tl++ = txdr_unsigned(1);               /* Auth_sys only */
4603         *tl++ = txdr_unsigned(AUTH_SYS);        /* AUTH_SYS type */
4604         *tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4605         *tl++ = 0;                              /* Null machine name */
4606         *tl++ = 0;                              /* Uid == 0 */
4607         *tl++ = 0;                              /* Gid == 0 */
4608         *tl = 0;                                /* No additional gids */
4609         nd->nd_flag |= ND_USEGSSNAME;
4610         error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4611             NFS_VER4, NULL, 1, NULL, NULL);
4612         if (error != 0)
4613                 return (error);
4614         if (nd->nd_repstat == 0) {
4615                 NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4616                     2 * NFSX_UNSIGNED);
4617                 bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4618                 tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4619                 sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4620                 crflags = fxdr_unsigned(uint32_t, *tl);
4621                 if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4622                         NFSLOCKMNT(nmp);
4623                         nmp->nm_state |= NFSSTA_SESSPERSIST;
4624                         NFSUNLOCKMNT(nmp);
4625                 }
4626
4627                 /* Get the fore channel slot count. */
4628                 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4629                 tl += 3;                /* Skip the other counts. */
4630                 sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4631                 tl++;
4632                 sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4633                 NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4634                 irdcnt = fxdr_unsigned(int, *tl);
4635                 if (irdcnt > 0)
4636                         NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4637
4638                 /* and the back channel slot count. */
4639                 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4640                 tl += 5;
4641                 sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4642                 NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4643         }
4644         error = nd->nd_repstat;
4645 nfsmout:
4646         mbuf_freem(nd->nd_mrep);
4647         return (error);
4648 }
4649
4650 /*
4651  * Do the NFSv4.1 Destroy Session.
4652  */
4653 int
4654 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4655     struct ucred *cred, NFSPROC_T *p)
4656 {
4657         uint32_t *tl;
4658         struct nfsrv_descript nfsd;
4659         struct nfsrv_descript *nd = &nfsd;
4660         int error;
4661
4662         nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL);
4663         NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4664         bcopy(NFSMNT_MDSSESSION(nmp)->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4665         nd->nd_flag |= ND_USEGSSNAME;
4666         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4667             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4668         if (error != 0)
4669                 return (error);
4670         error = nd->nd_repstat;
4671         mbuf_freem(nd->nd_mrep);
4672         return (error);
4673 }
4674
4675 /*
4676  * Do the NFSv4.1 Destroy Client.
4677  */
4678 int
4679 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4680     struct ucred *cred, NFSPROC_T *p)
4681 {
4682         uint32_t *tl;
4683         struct nfsrv_descript nfsd;
4684         struct nfsrv_descript *nd = &nfsd;
4685         int error;
4686
4687         nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL);
4688         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4689         *tl++ = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[0];
4690         *tl = NFSMNT_MDSSESSION(nmp)->nfsess_clientid.lval[1];
4691         nd->nd_flag |= ND_USEGSSNAME;
4692         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4693             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4694         if (error != 0)
4695                 return (error);
4696         error = nd->nd_repstat;
4697         mbuf_freem(nd->nd_mrep);
4698         return (error);
4699 }
4700
4701 /*
4702  * Do the NFSv4.1 LayoutGet.
4703  */
4704 int
4705 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
4706     uint64_t offset, uint64_t len, uint64_t minlen, int layoutlen,
4707     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp,
4708     struct ucred *cred, NFSPROC_T *p, void *stuff)
4709 {
4710         uint32_t *tl;
4711         struct nfsrv_descript nfsd, *nd = &nfsd;
4712         struct nfsfh *nfhp;
4713         struct nfsclflayout *flp, *prevflp, *tflp;
4714         int cnt, error, gotiomode, fhcnt, nfhlen, i, j;
4715         uint8_t *cp;
4716         uint64_t retlen;
4717
4718         flp = NULL;
4719         gotiomode = -1;
4720         nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL);
4721         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
4722             NFSX_STATEID);
4723         *tl++ = newnfs_false;           /* Don't signal availability. */
4724         *tl++ = txdr_unsigned(NFSLAYOUT_NFSV4_1_FILES);
4725         *tl++ = txdr_unsigned(iomode);
4726         txdr_hyper(offset, tl);
4727         tl += 2;
4728         txdr_hyper(len, tl);
4729         tl += 2;
4730         txdr_hyper(minlen, tl);
4731         tl += 2;
4732         *tl++ = txdr_unsigned(stateidp->seqid);
4733         NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
4734         *tl++ = stateidp->other[0];
4735         *tl++ = stateidp->other[1];
4736         *tl++ = stateidp->other[2];
4737         *tl = txdr_unsigned(layoutlen);
4738         nd->nd_flag |= ND_USEGSSNAME;
4739         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4740             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4741         if (error != 0)
4742                 return (error);
4743         if (nd->nd_repstat == 0) {
4744                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
4745                 if (*tl++ != 0)
4746                         *retonclosep = 1;
4747                 else
4748                         *retonclosep = 0;
4749                 stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
4750                 NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
4751                     (int)stateidp->seqid);
4752                 stateidp->other[0] = *tl++;
4753                 stateidp->other[1] = *tl++;
4754                 stateidp->other[2] = *tl++;
4755                 cnt = fxdr_unsigned(int, *tl);
4756                 NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
4757                 if (cnt <= 0 || cnt > 10000) {
4758                         /* Don't accept more than 10000 layouts in reply. */
4759                         error = NFSERR_BADXDR;
4760                         goto nfsmout;
4761                 }
4762                 for (i = 0; i < cnt; i++) {
4763                         /* Dissect all the way to the file handle cnt. */
4764                         NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_HYPER +
4765                             6 * NFSX_UNSIGNED + NFSX_V4DEVICEID);
4766                         fhcnt = fxdr_unsigned(int, *(tl + 11 +
4767                             NFSX_V4DEVICEID / NFSX_UNSIGNED));
4768                         NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
4769                         if (fhcnt < 0 || fhcnt > 100) {
4770                                 /* Don't accept more than 100 file handles. */
4771                                 error = NFSERR_BADXDR;
4772                                 goto nfsmout;
4773                         }
4774                         if (fhcnt > 1)
4775                                 flp = malloc(sizeof(*flp) + (fhcnt - 1) *
4776                                     sizeof(struct nfsfh *),
4777                                     M_NFSFLAYOUT, M_WAITOK);
4778                         else
4779                                 flp = malloc(sizeof(*flp),
4780                                     M_NFSFLAYOUT, M_WAITOK);
4781                         flp->nfsfl_flags = 0;
4782                         flp->nfsfl_fhcnt = 0;
4783                         flp->nfsfl_devp = NULL;
4784                         flp->nfsfl_off = fxdr_hyper(tl); tl += 2;
4785                         retlen = fxdr_hyper(tl); tl += 2;
4786                         if (flp->nfsfl_off + retlen < flp->nfsfl_off)
4787                                 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
4788                         else
4789                                 flp->nfsfl_end = flp->nfsfl_off + retlen;
4790                         flp->nfsfl_iomode = fxdr_unsigned(int, *tl++);
4791                         if (gotiomode == -1)
4792                                 gotiomode = flp->nfsfl_iomode;
4793                         NFSCL_DEBUG(4, "layg reqiom=%d retiom=%d\n", iomode,
4794                             (int)flp->nfsfl_iomode);
4795                         if (fxdr_unsigned(int, *tl++) !=
4796                             NFSLAYOUT_NFSV4_1_FILES) {
4797                                 printf("NFSv4.1: got non-files layout\n");
4798                                 error = NFSERR_BADXDR;
4799                                 goto nfsmout;
4800                         }
4801                         NFSBCOPY(++tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
4802                         tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4803                         flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
4804                         NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
4805                         flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
4806                         flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
4807                         if (fxdr_unsigned(int, *tl) != fhcnt) {
4808                                 printf("EEK! bad fhcnt\n");
4809                                 error = NFSERR_BADXDR;
4810                                 goto nfsmout;
4811                         }
4812                         for (j = 0; j < fhcnt; j++) {
4813                                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4814                                 nfhlen = fxdr_unsigned(int, *tl);
4815                                 if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
4816                                         error = NFSERR_BADXDR;
4817                                         goto nfsmout;
4818                                 }
4819                                 nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
4820                                     M_NFSFH, M_WAITOK);
4821                                 flp->nfsfl_fh[j] = nfhp;
4822                                 flp->nfsfl_fhcnt++;
4823                                 nfhp->nfh_len = nfhlen;
4824                                 NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
4825                                 NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
4826                         }
4827                         if (flp->nfsfl_iomode == gotiomode) {
4828                                 /* Keep the list in increasing offset order. */
4829                                 tflp = LIST_FIRST(flhp);
4830                                 prevflp = NULL;
4831                                 while (tflp != NULL &&
4832                                     tflp->nfsfl_off < flp->nfsfl_off) {
4833                                         prevflp = tflp;
4834                                         tflp = LIST_NEXT(tflp, nfsfl_list);
4835                                 }
4836                                 if (prevflp == NULL)
4837                                         LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
4838                                 else
4839                                         LIST_INSERT_AFTER(prevflp, flp,
4840                                             nfsfl_list);
4841                         } else {
4842                                 printf("nfscl_layoutget(): got wrong iomode\n");
4843                                 nfscl_freeflayout(flp);
4844                         }
4845                         flp = NULL;
4846                 }
4847         }
4848         if (nd->nd_repstat != 0 && error == 0)
4849                 error = nd->nd_repstat;
4850 nfsmout:
4851         if (error != 0 && flp != NULL)
4852                 nfscl_freeflayout(flp);
4853         mbuf_freem(nd->nd_mrep);
4854         return (error);
4855 }
4856
4857 /*
4858  * Do the NFSv4.1 Get Device Info.
4859  */
4860 int
4861 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
4862     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
4863     NFSPROC_T *p)
4864 {
4865         uint32_t cnt, *tl;
4866         struct nfsrv_descript nfsd;
4867         struct nfsrv_descript *nd = &nfsd;
4868         struct sockaddr_storage ss;
4869         struct nfsclds *dsp = NULL, **dspp;
4870         struct nfscldevinfo *ndi;
4871         int addrcnt, bitcnt, error, i, isudp, j, pos, safilled, stripecnt;
4872         uint8_t stripeindex;
4873
4874         *ndip = NULL;
4875         ndi = NULL;
4876         nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL);
4877         NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
4878         NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
4879         tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4880         *tl++ = txdr_unsigned(layouttype);
4881         *tl++ = txdr_unsigned(100000);
4882         if (notifybitsp != NULL && *notifybitsp != 0) {
4883                 *tl = txdr_unsigned(1);         /* One word of bits. */
4884                 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
4885                 *tl = txdr_unsigned(*notifybitsp);
4886         } else
4887                 *tl = txdr_unsigned(0);
4888         nd->nd_flag |= ND_USEGSSNAME;
4889         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4890             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4891         if (error != 0)
4892                 return (error);
4893         if (nd->nd_repstat == 0) {
4894                 NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4895                 if (layouttype != fxdr_unsigned(int, *tl++))
4896                         printf("EEK! devinfo layout type not same!\n");
4897                 stripecnt = fxdr_unsigned(int, *++tl);
4898                 NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
4899                 if (stripecnt < 1 || stripecnt > 4096) {
4900                         printf("NFS devinfo stripecnt %d: out of range\n",
4901                             stripecnt);
4902                         error = NFSERR_BADXDR;
4903                         goto nfsmout;
4904                 }
4905                 NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) * NFSX_UNSIGNED);
4906                 addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
4907                 NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
4908                 if (addrcnt < 1 || addrcnt > 128) {
4909                         printf("NFS devinfo addrcnt %d: out of range\n",
4910                             addrcnt);
4911                         error = NFSERR_BADXDR;
4912                         goto nfsmout;
4913                 }
4914
4915                 /*
4916                  * Now we know how many stripe indices and addresses, so
4917                  * we can allocate the structure the correct size.
4918                  */
4919                 i = (stripecnt * sizeof(uint8_t)) / sizeof(struct nfsclds *)
4920                     + 1;
4921                 NFSCL_DEBUG(4, "stripeindices=%d\n", i);
4922                 ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
4923                     sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK | M_ZERO);
4924                 NFSBCOPY(deviceid, ndi->nfsdi_deviceid, NFSX_V4DEVICEID);
4925                 ndi->nfsdi_refcnt = 0;
4926                 ndi->nfsdi_stripecnt = stripecnt;
4927                 ndi->nfsdi_addrcnt = addrcnt;
4928                 /* Fill in the stripe indices. */
4929                 for (i = 0; i < stripecnt; i++) {
4930                         stripeindex = fxdr_unsigned(uint8_t, *tl++);
4931                         NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
4932                         if (stripeindex >= addrcnt) {
4933                                 printf("NFS devinfo stripeindex %d: too big\n",
4934                                     (int)stripeindex);
4935                                 error = NFSERR_BADXDR;
4936                                 goto nfsmout;
4937                         }
4938                         nfsfldi_setstripeindex(ndi, i, stripeindex);
4939                 }
4940
4941                 /* Now, dissect the server address(es). */
4942                 safilled = 0;
4943                 for (i = 0; i < addrcnt; i++) {
4944                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4945                         cnt = fxdr_unsigned(uint32_t, *tl);
4946                         if (cnt == 0) {
4947                                 printf("NFS devinfo 0 len addrlist\n");
4948                                 error = NFSERR_BADXDR;
4949                                 goto nfsmout;
4950                         }
4951                         dspp = nfsfldi_addr(ndi, i);
4952                         pos = arc4random() % cnt;       /* Choose one. */
4953                         safilled = 0;
4954                         for (j = 0; j < cnt; j++) {
4955                                 error = nfsv4_getipaddr(nd, &ss, &isudp);
4956                                 if (error != 0 && error != EPERM) {
4957                                         error = NFSERR_BADXDR;
4958                                         goto nfsmout;
4959                                 }
4960                                 if (error == 0 && isudp == 0) {
4961                                         /*
4962                                          * The algorithm is:
4963                                          * - use "pos" entry if it is of the
4964                                          *   same af_family or none of them
4965                                          *   is of the same af_family
4966                                          * else
4967                                          * - use the first one of the same
4968                                          *   af_family.
4969                                          */
4970                                         if ((safilled == 0 && ss.ss_family ==
4971                                              nmp->nm_nam->sa_family) ||
4972                                             (j == pos &&
4973                                              (safilled == 0 || ss.ss_family ==
4974                                               nmp->nm_nam->sa_family)) ||
4975                                             (safilled == 1 && ss.ss_family ==
4976                                              nmp->nm_nam->sa_family)) {
4977                                                 error = nfsrpc_fillsa(nmp, &ss,
4978                                                     &dsp, p);
4979                                                 if (error == 0) {
4980                                                         *dspp = dsp;
4981                                                         if (ss.ss_family ==
4982                                                          nmp->nm_nam->sa_family)
4983                                                                 safilled = 2;
4984                                                         else
4985                                                                 safilled = 1;
4986                                                 }
4987                                         }
4988                                 }
4989                         }
4990                         if (safilled == 0)
4991                                 break;
4992                 }
4993
4994                 /* And the notify bits. */
4995                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4996                 if (safilled != 0) {
4997                         bitcnt = fxdr_unsigned(int, *tl);
4998                         if (bitcnt > 0) {
4999                                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5000                                 if (notifybitsp != NULL)
5001                                         *notifybitsp =
5002                                             fxdr_unsigned(uint32_t, *tl);
5003                         }
5004                         *ndip = ndi;
5005                 } else
5006                         error = EPERM;
5007         }
5008         if (nd->nd_repstat != 0)
5009                 error = nd->nd_repstat;
5010 nfsmout:
5011         if (error != 0 && ndi != NULL)
5012                 nfscl_freedevinfo(ndi);
5013         mbuf_freem(nd->nd_mrep);
5014         return (error);
5015 }
5016
5017 /*
5018  * Do the NFSv4.1 LayoutCommit.
5019  */
5020 int
5021 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5022     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5023     int layouttype, int layoutupdatecnt, uint8_t *layp, struct ucred *cred,
5024     NFSPROC_T *p, void *stuff)
5025 {
5026         uint32_t *tl;
5027         struct nfsrv_descript nfsd, *nd = &nfsd;
5028         int error, outcnt, i;
5029         uint8_t *cp;
5030
5031         nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL);
5032         NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5033             NFSX_STATEID);
5034         txdr_hyper(off, tl);
5035         tl += 2;
5036         txdr_hyper(len, tl);
5037         tl += 2;
5038         if (reclaim != 0)
5039                 *tl++ = newnfs_true;
5040         else
5041                 *tl++ = newnfs_false;
5042         *tl++ = txdr_unsigned(stateidp->seqid);
5043         *tl++ = stateidp->other[0];
5044         *tl++ = stateidp->other[1];
5045         *tl++ = stateidp->other[2];
5046         *tl++ = newnfs_true;
5047         if (lastbyte < off)
5048                 lastbyte = off;
5049         else if (lastbyte >= (off + len))
5050                 lastbyte = off + len - 1;
5051         txdr_hyper(lastbyte, tl);
5052         tl += 2;
5053         *tl++ = newnfs_false;
5054         *tl++ = txdr_unsigned(layouttype);
5055         *tl = txdr_unsigned(layoutupdatecnt);
5056         if (layoutupdatecnt > 0) {
5057                 KASSERT(layouttype != NFSLAYOUT_NFSV4_1_FILES,
5058                     ("Must be nil for Files Layout"));
5059                 outcnt = NFSM_RNDUP(layoutupdatecnt);
5060                 NFSM_BUILD(cp, uint8_t *, outcnt);
5061                 NFSBCOPY(layp, cp, layoutupdatecnt);
5062                 cp += layoutupdatecnt;
5063                 for (i = 0; i < (outcnt - layoutupdatecnt); i++)
5064                         *cp++ = 0x0;
5065         }
5066         nd->nd_flag |= ND_USEGSSNAME;
5067         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5068             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5069         if (error != 0)
5070                 return (error);
5071         error = nd->nd_repstat;
5072         mbuf_freem(nd->nd_mrep);
5073         return (error);
5074 }
5075
5076 /*
5077  * Do the NFSv4.1 LayoutReturn.
5078  */
5079 int
5080 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5081     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5082     uint64_t len, nfsv4stateid_t *stateidp, int layoutcnt, uint32_t *layp,
5083     struct ucred *cred, NFSPROC_T *p, void *stuff)
5084 {
5085         uint32_t *tl;
5086         struct nfsrv_descript nfsd, *nd = &nfsd;
5087         int error, outcnt, i;
5088         uint8_t *cp;
5089
5090         nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL);
5091         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5092         if (reclaim != 0)
5093                 *tl++ = newnfs_true;
5094         else
5095                 *tl++ = newnfs_false;
5096         *tl++ = txdr_unsigned(layouttype);
5097         *tl++ = txdr_unsigned(iomode);
5098         *tl = txdr_unsigned(layoutreturn);
5099         if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5100                 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5101                     NFSX_UNSIGNED);
5102                 txdr_hyper(offset, tl);
5103                 tl += 2;
5104                 txdr_hyper(len, tl);
5105                 tl += 2;
5106                 NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5107                 *tl++ = txdr_unsigned(stateidp->seqid);
5108                 *tl++ = stateidp->other[0];
5109                 *tl++ = stateidp->other[1];
5110                 *tl++ = stateidp->other[2];
5111                 *tl = txdr_unsigned(layoutcnt);
5112                 if (layoutcnt > 0) {
5113                         outcnt = NFSM_RNDUP(layoutcnt);
5114                         NFSM_BUILD(cp, uint8_t *, outcnt);
5115                         NFSBCOPY(layp, cp, layoutcnt);
5116                         cp += layoutcnt;
5117                         for (i = 0; i < (outcnt - layoutcnt); i++)
5118                                 *cp++ = 0x0;
5119                 }
5120         }
5121         nd->nd_flag |= ND_USEGSSNAME;
5122         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5123             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5124         if (error != 0)
5125                 return (error);
5126         if (nd->nd_repstat == 0) {
5127                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5128                 if (*tl != 0) {
5129                         NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5130                         stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5131                         stateidp->other[0] = *tl++;
5132                         stateidp->other[1] = *tl++;
5133                         stateidp->other[2] = *tl;
5134                 }
5135         } else
5136                 error = nd->nd_repstat;
5137 nfsmout:
5138         mbuf_freem(nd->nd_mrep);
5139         return (error);
5140 }
5141
5142 /*
5143  * Acquire a layout and devinfo, if possible. The caller must have acquired
5144  * a reference count on the nfsclclient structure before calling this.
5145  * Return the layout in lypp with a reference count on it, if successful.
5146  */
5147 static int
5148 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5149     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5150     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5151 {
5152         struct nfscllayout *lyp;
5153         struct nfsclflayout *flp, *tflp;
5154         struct nfscldevinfo *dip;
5155         struct nfsclflayouthead flh;
5156         int error = 0, islocked, layoutlen, recalled, retonclose;
5157         nfsv4stateid_t stateid;
5158
5159         *lypp = NULL;
5160         /*
5161          * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5162          * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5163          * flp == NULL.
5164          */
5165         lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5166             off, &flp, &recalled);
5167         islocked = 0;
5168         if (lyp == NULL || flp == NULL) {
5169                 if (recalled != 0)
5170                         return (EIO);
5171                 LIST_INIT(&flh);
5172                 layoutlen = NFSMNT_MDSSESSION(nmp)->nfsess_maxcache -
5173                     (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5174                 if (lyp == NULL) {
5175                         stateid.seqid = 0;
5176                         stateid.other[0] = stateidp->other[0];
5177                         stateid.other[1] = stateidp->other[1];
5178                         stateid.other[2] = stateidp->other[2];
5179                         error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5180                             nfhp->nfh_len, iomode, (uint64_t)0, INT64_MAX,
5181                             (uint64_t)0, layoutlen, &stateid, &retonclose,
5182                             &flh, cred, p, NULL);
5183                 } else {
5184                         islocked = 1;
5185                         stateid.seqid = lyp->nfsly_stateid.seqid;
5186                         stateid.other[0] = lyp->nfsly_stateid.other[0];
5187                         stateid.other[1] = lyp->nfsly_stateid.other[1];
5188                         stateid.other[2] = lyp->nfsly_stateid.other[2];
5189                         error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5190                             nfhp->nfh_len, iomode, off, INT64_MAX,
5191                             (uint64_t)0, layoutlen, &stateid, &retonclose,
5192                             &flh, cred, p, NULL);
5193                 }
5194                 if (error == 0)
5195                         LIST_FOREACH(tflp, &flh, nfsfl_list) {
5196                                 error = nfscl_adddevinfo(nmp, NULL, tflp);
5197                                 if (error != 0) {
5198                                         error = nfsrpc_getdeviceinfo(nmp,
5199                                             tflp->nfsfl_dev,
5200                                             NFSLAYOUT_NFSV4_1_FILES,
5201                                             notifybitsp, &dip, cred, p);
5202                                         if (error != 0)
5203                                                 break;
5204                                         error = nfscl_adddevinfo(nmp, dip,
5205                                             tflp);
5206                                         if (error != 0)
5207                                                 printf(
5208                                                     "getlayout: cannot add\n");
5209                                 }
5210                         }
5211                 if (error == 0) {
5212                         /*
5213                          * nfscl_layout() always returns with the nfsly_lock
5214                          * set to a refcnt (shared lock).
5215                          */
5216                         error = nfscl_layout(nmp, vp, nfhp->nfh_fh,
5217                             nfhp->nfh_len, &stateid, retonclose, &flh, &lyp,
5218                             cred, p);
5219                         if (error == 0)
5220                                 *lypp = lyp;
5221                 } else if (islocked != 0)
5222                         nfsv4_unlock(&lyp->nfsly_lock, 0);
5223         } else
5224                 *lypp = lyp;
5225         return (error);
5226 }
5227
5228 /*
5229  * Do a TCP connection plus exchange id and create session.
5230  * If successful, a "struct nfsclds" is linked into the list for the
5231  * mount point and a pointer to it is returned.
5232  */
5233 static int
5234 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_storage *ssp,
5235     struct nfsclds **dspp, NFSPROC_T *p)
5236 {
5237         struct sockaddr_in *msad, *sad, *ssd;
5238         struct sockaddr_in6 *msad6, *sad6, *ssd6;
5239         struct nfsclclient *clp;
5240         struct nfssockreq *nrp;
5241         struct nfsclds *dsp, *tdsp;
5242         int error;
5243         enum nfsclds_state retv;
5244         uint32_t sequenceid;
5245
5246         KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5247             ("nfsrpc_fillsa: NULL nr_cred"));
5248         NFSLOCKCLSTATE();
5249         clp = nmp->nm_clp;
5250         NFSUNLOCKCLSTATE();
5251         if (clp == NULL)
5252                 return (EPERM);
5253         if (ssp->ss_family == AF_INET) {
5254                 ssd = (struct sockaddr_in *)ssp;
5255                 NFSLOCKMNT(nmp);
5256
5257                 /*
5258                  * Check to see if we already have a session for this
5259                  * address that is usable for a DS.
5260                  * Note that the MDS's address is in a different place
5261                  * than the sessions already acquired for DS's.
5262                  */
5263                 msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5264                 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5265                 while (tdsp != NULL) {
5266                         if (msad != NULL && msad->sin_family == AF_INET &&
5267                             ssd->sin_addr.s_addr == msad->sin_addr.s_addr &&
5268                             ssd->sin_port == msad->sin_port &&
5269                             (tdsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5270                                 *dspp = tdsp;
5271                                 NFSUNLOCKMNT(nmp);
5272                                 NFSCL_DEBUG(4, "fnd same addr\n");
5273                                 return (0);
5274                         }
5275                         tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5276                         if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5277                                 msad = (struct sockaddr_in *)
5278                                     tdsp->nfsclds_sockp->nr_nam;
5279                         else
5280                                 msad = NULL;
5281                 }
5282                 NFSUNLOCKMNT(nmp);
5283
5284                 /* No IP address match, so look for new/trunked one. */
5285                 sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5286                 sad->sin_len = sizeof(*sad);
5287                 sad->sin_family = AF_INET;
5288                 sad->sin_port = ssd->sin_port;
5289                 sad->sin_addr.s_addr = ssd->sin_addr.s_addr;
5290                 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5291                 nrp->nr_nam = (struct sockaddr *)sad;
5292         } else if (ssp->ss_family == AF_INET6) {
5293                 ssd6 = (struct sockaddr_in6 *)ssp;
5294                 NFSLOCKMNT(nmp);
5295
5296                 /*
5297                  * Check to see if we already have a session for this
5298                  * address that is usable for a DS.
5299                  * Note that the MDS's address is in a different place
5300                  * than the sessions already acquired for DS's.
5301                  */
5302                 msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5303                 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5304                 while (tdsp != NULL) {
5305                         if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5306                             IN6_ARE_ADDR_EQUAL(&ssd6->sin6_addr,
5307                             &msad6->sin6_addr) &&
5308                             ssd6->sin6_port == msad6->sin6_port &&
5309                             (tdsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5310                                 *dspp = tdsp;
5311                                 NFSUNLOCKMNT(nmp);
5312                                 return (0);
5313                         }
5314                         tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5315                         if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5316                                 msad6 = (struct sockaddr_in6 *)
5317                                     tdsp->nfsclds_sockp->nr_nam;
5318                         else
5319                                 msad6 = NULL;
5320                 }
5321                 NFSUNLOCKMNT(nmp);
5322
5323                 /* No IP address match, so look for new/trunked one. */
5324                 sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5325                 sad6->sin6_len = sizeof(*sad6);
5326                 sad6->sin6_family = AF_INET6;
5327                 sad6->sin6_port = ssd6->sin6_port;
5328                 NFSBCOPY(&ssd6->sin6_addr, &sad6->sin6_addr,
5329                     sizeof(struct in6_addr));
5330                 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5331                 nrp->nr_nam = (struct sockaddr *)sad6;
5332         } else
5333                 return (EPERM);
5334
5335         nrp->nr_sotype = SOCK_STREAM;
5336         mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5337         nrp->nr_prog = NFS_PROG;
5338         nrp->nr_vers = NFS_VER4;
5339
5340         /*
5341          * Use the credentials that were used for the mount, which are
5342          * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5343          * Ref. counting the credentials with crhold() is probably not
5344          * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5345          * unmount, but I did it anyhow.
5346          */
5347         nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5348         error = newnfs_connect(nmp, nrp, NULL, p, 0);
5349         NFSCL_DEBUG(3, "DS connect=%d\n", error);
5350
5351         /* Now, do the exchangeid and create session. */
5352         if (error == 0)
5353                 error = nfsrpc_exchangeid(nmp, clp, nrp, NFSV4EXCH_USEPNFSDS,
5354                     &dsp, nrp->nr_cred, p);
5355         NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5356         if (error == 0) {
5357                 dsp->nfsclds_sockp = nrp;
5358                 NFSLOCKMNT(nmp);
5359                 retv = nfscl_getsameserver(nmp, dsp, &tdsp);
5360                 NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5361                 if (retv == NFSDSP_USETHISSESSION) {
5362                         NFSUNLOCKMNT(nmp);
5363                         /*
5364                          * If there is already a session for this server,
5365                          * use it.
5366                          */
5367                         (void)newnfs_disconnect(nrp);
5368                         nfscl_freenfsclds(dsp);
5369                         *dspp = tdsp;
5370                         return (0);
5371                 }
5372                 if (retv == NFSDSP_SEQTHISSESSION)
5373                         sequenceid = tdsp->nfsclds_sess.nfsess_sequenceid;
5374                 else
5375                         sequenceid = dsp->nfsclds_sess.nfsess_sequenceid;
5376                 NFSUNLOCKMNT(nmp);
5377                 error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5378                     nrp, sequenceid, 0, nrp->nr_cred, p);
5379                 NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5380         } else {
5381                 NFSFREECRED(nrp->nr_cred);
5382                 NFSFREEMUTEX(&nrp->nr_mtx);
5383                 free(nrp->nr_nam, M_SONAME);
5384                 free(nrp, M_NFSSOCKREQ);
5385         }
5386         if (error == 0) {
5387                 NFSCL_DEBUG(3, "add DS session\n");
5388                 /*
5389                  * Put it at the end of the list. That way the list
5390                  * is ordered by when the entry was added. This matters
5391                  * since the one done first is the one that should be
5392                  * used for sequencid'ing any subsequent create sessions.
5393                  */
5394                 NFSLOCKMNT(nmp);
5395                 TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5396                 NFSUNLOCKMNT(nmp);
5397                 *dspp = dsp;
5398         } else if (dsp != NULL)
5399                 nfscl_freenfsclds(dsp);
5400         return (error);
5401 }
5402
5403 /*
5404  * Do the NFSv4.1 Reclaim Complete.
5405  */
5406 int
5407 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5408 {
5409         uint32_t *tl;
5410         struct nfsrv_descript nfsd;
5411         struct nfsrv_descript *nd = &nfsd;
5412         int error;
5413
5414         nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL);
5415         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5416         *tl = newnfs_false;
5417         nd->nd_flag |= ND_USEGSSNAME;
5418         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5419             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5420         if (error != 0)
5421                 return (error);
5422         error = nd->nd_repstat;
5423         mbuf_freem(nd->nd_mrep);
5424         return (error);
5425 }
5426
5427 /*
5428  * Initialize the slot tables for a session.
5429  */
5430 static void
5431 nfscl_initsessionslots(struct nfsclsession *sep)
5432 {
5433         int i;
5434
5435         for (i = 0; i < NFSV4_CBSLOTS; i++) {
5436                 if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5437                         m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5438                 NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5439         }
5440         for (i = 0; i < 64; i++)
5441                 sep->nfsess_slotseq[i] = 0;
5442         sep->nfsess_slots = 0;
5443 }
5444
5445 /*
5446  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5447  */
5448 int
5449 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5450     uint32_t rwaccess, struct ucred *cred, NFSPROC_T *p)
5451 {
5452         struct nfsnode *np = VTONFS(vp);
5453         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5454         struct nfscllayout *layp;
5455         struct nfscldevinfo *dip;
5456         struct nfsclflayout *rflp;
5457         nfsv4stateid_t stateid;
5458         struct ucred *newcred;
5459         uint64_t lastbyte, len, off, oresid, xfer;
5460         int eof, error, iolaymode, recalled;
5461         void *lckp;
5462
5463         if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5464             (np->n_flag & NNOLAYOUT) != 0)
5465                 return (EIO);
5466         /* Now, get a reference cnt on the clientid for this mount. */
5467         if (nfscl_getref(nmp) == 0)
5468                 return (EIO);
5469
5470         /* Find an appropriate stateid. */
5471         newcred = NFSNEWCRED(cred);
5472         error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5473             rwaccess, 1, newcred, p, &stateid, &lckp);
5474         if (error != 0) {
5475                 NFSFREECRED(newcred);
5476                 nfscl_relref(nmp);
5477                 return (error);
5478         }
5479         /* Search for a layout for this file. */
5480         off = uiop->uio_offset;
5481         layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5482             np->n_fhp->nfh_len, off, &rflp, &recalled);
5483         if (layp == NULL || rflp == NULL) {
5484                 if (recalled != 0) {
5485                         NFSFREECRED(newcred);
5486                         nfscl_relref(nmp);
5487                         return (EIO);
5488                 }
5489                 if (layp != NULL) {
5490                         nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5491                         layp = NULL;
5492                 }
5493                 /* Try and get a Layout, if it is supported. */
5494                 if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5495                     (np->n_flag & NWRITEOPENED) != 0)
5496                         iolaymode = NFSLAYOUTIOMODE_RW;
5497                 else
5498                         iolaymode = NFSLAYOUTIOMODE_READ;
5499                 error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5500                     NULL, &stateid, off, &layp, newcred, p);
5501                 if (error != 0) {
5502                         NFSLOCKNODE(np);
5503                         np->n_flag |= NNOLAYOUT;
5504                         NFSUNLOCKNODE(np);
5505                         if (lckp != NULL)
5506                                 nfscl_lockderef(lckp);
5507                         NFSFREECRED(newcred);
5508                         if (layp != NULL)
5509                                 nfscl_rellayout(layp, 0);
5510                         nfscl_relref(nmp);
5511                         return (error);
5512                 }
5513         }
5514
5515         /*
5516          * Loop around finding a layout that works for the first part of
5517          * this I/O operation, and then call the function that actually
5518          * does the RPC.
5519          */
5520         eof = 0;
5521         len = (uint64_t)uiop->uio_resid;
5522         while (len > 0 && error == 0 && eof == 0) {
5523                 off = uiop->uio_offset;
5524                 error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5525                 if (error == 0) {
5526                         oresid = xfer = (uint64_t)uiop->uio_resid;
5527                         if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5528                                 xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5529                         dip = nfscl_getdevinfo(nmp->nm_clp, rflp->nfsfl_dev,
5530                             rflp->nfsfl_devp);
5531                         if (dip != NULL) {
5532                                 error = nfscl_doflayoutio(vp, uiop, iomode,
5533                                     must_commit, &eof, &stateid, rwaccess, dip,
5534                                     layp, rflp, off, xfer, newcred, p);
5535                                 nfscl_reldevinfo(dip);
5536                                 lastbyte = off + xfer - 1;
5537                                 if (error == 0) {
5538                                         NFSLOCKCLSTATE();
5539                                         if (lastbyte > layp->nfsly_lastbyte)
5540                                                 layp->nfsly_lastbyte = lastbyte;
5541                                         NFSUNLOCKCLSTATE();
5542                                 }
5543                         } else
5544                                 error = EIO;
5545                         if (error == 0)
5546                                 len -= (oresid - (uint64_t)uiop->uio_resid);
5547                 }
5548         }
5549         if (lckp != NULL)
5550                 nfscl_lockderef(lckp);
5551         NFSFREECRED(newcred);
5552         nfscl_rellayout(layp, 0);
5553         nfscl_relref(nmp);
5554         return (error);
5555 }
5556
5557 /*
5558  * Find a file layout that will handle the first bytes of the requested
5559  * range and return the information from it needed to to the I/O operation.
5560  */
5561 int
5562 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5563     struct nfsclflayout **retflpp)
5564 {
5565         struct nfsclflayout *flp, *nflp, *rflp;
5566         uint32_t rw;
5567
5568         rflp = NULL;
5569         rw = rwaccess;
5570         /* For reading, do the Read list first and then the Write list. */
5571         do {
5572                 if (rw == NFSV4OPEN_ACCESSREAD)
5573                         flp = LIST_FIRST(&lyp->nfsly_flayread);
5574                 else
5575                         flp = LIST_FIRST(&lyp->nfsly_flayrw);
5576                 while (flp != NULL) {
5577                         nflp = LIST_NEXT(flp, nfsfl_list);
5578                         if (flp->nfsfl_off > off)
5579                                 break;
5580                         if (flp->nfsfl_end > off &&
5581                             (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5582                                 rflp = flp;
5583                         flp = nflp;
5584                 }
5585                 if (rw == NFSV4OPEN_ACCESSREAD)
5586                         rw = NFSV4OPEN_ACCESSWRITE;
5587                 else
5588                         rw = 0;
5589         } while (rw != 0);
5590         if (rflp != NULL) {
5591                 /* This one covers the most bytes starting at off. */
5592                 *retflpp = rflp;
5593                 return (0);
5594         }
5595         return (EIO);
5596 }
5597
5598 /*
5599  * Do I/O using an NFSv4.1 file layout.
5600  */
5601 static int
5602 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5603     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5604     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5605     uint64_t len, struct ucred *cred, NFSPROC_T *p)
5606 {
5607         uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
5608         int commit_thru_mds, error = 0, stripe_index, stripe_pos;
5609         struct nfsnode *np;
5610         struct nfsfh *fhp;
5611         struct nfsclds **dspp;
5612
5613         np = VTONFS(vp);
5614         rel_off = off - flp->nfsfl_patoff;
5615         stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
5616         stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
5617             dp->nfsdi_stripecnt;
5618         transfer = stripe_unit_size - (rel_off % stripe_unit_size);
5619
5620         /* Loop around, doing I/O for each stripe unit. */
5621         while (len > 0 && error == 0) {
5622                 stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
5623                 dspp = nfsfldi_addr(dp, stripe_index);
5624                 if (len > transfer)
5625                         xfer = transfer;
5626                 else
5627                         xfer = len;
5628                 if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
5629                         /* Dense layout. */
5630                         if (stripe_pos >= flp->nfsfl_fhcnt)
5631                                 return (EIO);
5632                         fhp = flp->nfsfl_fh[stripe_pos];
5633                         io_off = (rel_off / (stripe_unit_size *
5634                             dp->nfsdi_stripecnt)) * stripe_unit_size +
5635                             rel_off % stripe_unit_size;
5636                 } else {
5637                         /* Sparse layout. */
5638                         if (flp->nfsfl_fhcnt > 1) {
5639                                 if (stripe_index >= flp->nfsfl_fhcnt)
5640                                         return (EIO);
5641                                 fhp = flp->nfsfl_fh[stripe_index];
5642                         } else if (flp->nfsfl_fhcnt == 1)
5643                                 fhp = flp->nfsfl_fh[0];
5644                         else
5645                                 fhp = np->n_fhp;
5646                         io_off = off;
5647                 }
5648                 if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0)
5649                         commit_thru_mds = 1;
5650                 else
5651                         commit_thru_mds = 0;
5652                 if (rwflag == FREAD)
5653                         error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
5654                             io_off, xfer, fhp, cred, p);
5655                 else {
5656                         error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
5657                             stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
5658                             cred, p);
5659                         if (error == 0) {
5660                                 NFSLOCKCLSTATE();
5661                                 lyp->nfsly_flags |= NFSLY_WRITTEN;
5662                                 NFSUNLOCKCLSTATE();
5663                         }
5664                 }
5665                 if (error == 0) {
5666                         transfer = stripe_unit_size;
5667                         stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
5668                         len -= xfer;
5669                         off += xfer;
5670                 }
5671         }
5672         return (error);
5673 }
5674
5675 /*
5676  * The actual read RPC done to a DS.
5677  */
5678 static int
5679 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
5680     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp,
5681     struct ucred *cred, NFSPROC_T *p)
5682 {
5683         uint32_t *tl;
5684         int error, retlen;
5685         struct nfsrv_descript nfsd;
5686         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5687         struct nfsrv_descript *nd = &nfsd;
5688         struct nfssockreq *nrp;
5689
5690         nd->nd_mrep = NULL;
5691         nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5692             NULL, &dsp->nfsclds_sess);
5693         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5694         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
5695         txdr_hyper(io_off, tl);
5696         *(tl + 2) = txdr_unsigned(len);
5697         nrp = dsp->nfsclds_sockp;
5698         if (nrp == NULL)
5699                 /* If NULL, use the MDS socket. */
5700                 nrp = &nmp->nm_sockreq;
5701         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5702             NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5703         if (error != 0)
5704                 return (error);
5705         if (nd->nd_repstat != 0) {
5706                 error = nd->nd_repstat;
5707                 goto nfsmout;
5708         }
5709         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5710         *eofp = fxdr_unsigned(int, *tl);
5711         NFSM_STRSIZ(retlen, len);
5712         error = nfsm_mbufuio(nd, uiop, retlen);
5713 nfsmout:
5714         if (nd->nd_mrep != NULL)
5715                 mbuf_freem(nd->nd_mrep);
5716         return (error);
5717 }
5718
5719 /*
5720  * The actual write RPC done to a DS.
5721  */
5722 static int
5723 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5724     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
5725     struct nfsfh *fhp, int commit_thru_mds, struct ucred *cred, NFSPROC_T *p)
5726 {
5727         uint32_t *tl;
5728         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5729         int error, rlen, commit, committed = NFSWRITE_FILESYNC;
5730         int32_t backup;
5731         struct nfsrv_descript nfsd;
5732         struct nfsrv_descript *nd = &nfsd;
5733         struct nfssockreq *nrp;
5734
5735         KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
5736         nd->nd_mrep = NULL;
5737         nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5738             NULL, &dsp->nfsclds_sess);
5739         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5740         NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
5741         txdr_hyper(io_off, tl);
5742         tl += 2;
5743         *tl++ = txdr_unsigned(*iomode);
5744         *tl = txdr_unsigned(len);
5745         nfsm_uiombuf(nd, uiop, len);
5746         nrp = dsp->nfsclds_sockp;
5747         if (nrp == NULL)
5748                 /* If NULL, use the MDS socket. */
5749                 nrp = &nmp->nm_sockreq;
5750         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5751             NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5752         if (error != 0)
5753                 return (error);
5754         if (nd->nd_repstat != 0) {
5755                 /*
5756                  * In case the rpc gets retried, roll
5757                  * the uio fileds changed by nfsm_uiombuf()
5758                  * back.
5759                  */
5760                 uiop->uio_offset -= len;
5761                 uio_uio_resid_add(uiop, len);
5762                 uio_iov_base_add(uiop, -len);
5763                 uio_iov_len_add(uiop, len);
5764                 error = nd->nd_repstat;
5765         } else {
5766                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
5767                 rlen = fxdr_unsigned(int, *tl++);
5768                 if (rlen == 0) {
5769                         error = NFSERR_IO;
5770                         goto nfsmout;
5771                 } else if (rlen < len) {
5772                         backup = len - rlen;
5773                         uio_iov_base_add(uiop, -(backup));
5774                         uio_iov_len_add(uiop, backup);
5775                         uiop->uio_offset -= backup;
5776                         uio_uio_resid_add(uiop, backup);
5777                         len = rlen;
5778                 }
5779                 commit = fxdr_unsigned(int, *tl++);
5780
5781                 /*
5782                  * Return the lowest committment level
5783                  * obtained by any of the RPCs.
5784                  */
5785                 if (committed == NFSWRITE_FILESYNC)
5786                         committed = commit;
5787                 else if (committed == NFSWRITE_DATASYNC &&
5788                     commit == NFSWRITE_UNSTABLE)
5789                         committed = commit;
5790                 if (commit_thru_mds != 0) {
5791                         NFSLOCKMNT(nmp);
5792                         if (!NFSHASWRITEVERF(nmp)) {
5793                                 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5794                                 NFSSETWRITEVERF(nmp);
5795                         } else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
5796                                 *must_commit = 1;
5797                                 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5798                         }
5799                         NFSUNLOCKMNT(nmp);
5800                 } else {
5801                         NFSLOCKDS(dsp);
5802                         if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
5803                                 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5804                                 dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
5805                         } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5806                                 *must_commit = 1;
5807                                 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5808                         }
5809                         NFSUNLOCKDS(dsp);
5810                 }
5811         }
5812 nfsmout:
5813         if (nd->nd_mrep != NULL)
5814                 mbuf_freem(nd->nd_mrep);
5815         *iomode = committed;
5816         if (nd->nd_repstat != 0 && error == 0)
5817                 error = nd->nd_repstat;
5818         return (error);
5819 }
5820
5821 /*
5822  * Free up the nfsclds structure.
5823  */
5824 void
5825 nfscl_freenfsclds(struct nfsclds *dsp)
5826 {
5827         int i;
5828
5829         if (dsp == NULL)
5830                 return;
5831         if (dsp->nfsclds_sockp != NULL) {
5832                 NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
5833                 NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
5834                 free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
5835                 free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
5836         }
5837         NFSFREEMUTEX(&dsp->nfsclds_mtx);
5838         NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
5839         for (i = 0; i < NFSV4_CBSLOTS; i++) {
5840                 if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
5841                         m_freem(
5842                             dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
5843         }
5844         free(dsp, M_NFSCLDS);
5845 }
5846
5847 static enum nfsclds_state
5848 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
5849     struct nfsclds **retdspp)
5850 {
5851         struct nfsclds *dsp, *cur_dsp;
5852
5853         /*
5854          * Search the list of nfsclds structures for one with the same
5855          * server.
5856          */
5857         cur_dsp = NULL;
5858         TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
5859                 if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
5860                     dsp->nfsclds_servownlen != 0 &&
5861                     !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
5862                     dsp->nfsclds_servownlen)) {
5863                         NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
5864                             TAILQ_FIRST(&nmp->nm_sess), dsp,
5865                             dsp->nfsclds_flags);
5866                         /* Server major id matches. */
5867                         if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5868                                 *retdspp = dsp;
5869                                 return (NFSDSP_USETHISSESSION);
5870                         }
5871
5872                         /*
5873                          * Note the first match, so it can be used for
5874                          * sequence'ing new sessions.
5875                          */
5876                         if (cur_dsp == NULL)
5877                                 cur_dsp = dsp;
5878                 }
5879         }
5880         if (cur_dsp != NULL) {
5881                 *retdspp = cur_dsp;
5882                 return (NFSDSP_SEQTHISSESSION);
5883         }
5884         return (NFSDSP_NOTFOUND);
5885 }
5886
5887 #ifdef notyet
5888 /*
5889  * NFS commit rpc to a DS.
5890  */
5891 static int
5892 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
5893     struct nfsfh *fhp, struct ucred *cred, NFSPROC_T *p, void *stuff)
5894 {
5895         uint32_t *tl;
5896         struct nfsrv_descript nfsd, *nd = &nfsd;
5897         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5898         struct nfssockreq *nrp;
5899         int error;
5900
5901         nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5902             NULL, &dsp->nfsclds_sess);
5903         NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
5904         txdr_hyper(offset, tl);
5905         tl += 2;
5906         *tl = txdr_unsigned(cnt);
5907         nrp = dsp->nfsclds_sockp;
5908         if (nrp == NULL)
5909                 /* If NULL, use the MDS socket. */
5910                 nrp = &nmp->nm_sockreq;
5911         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5912             NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5913         if (error)
5914                 return (error);
5915         if (nd->nd_repstat == 0) {
5916                 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
5917                 NFSLOCKDS(dsp);
5918                 if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5919                         NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5920                         error = NFSERR_STALEWRITEVERF;
5921                 }
5922                 NFSUNLOCKDS(dsp);
5923         }
5924 nfsmout:
5925         if (error == 0 && nd->nd_repstat != 0)
5926                 error = nd->nd_repstat;
5927         mbuf_freem(nd->nd_mrep);
5928         return (error);
5929 }
5930 #endif
5931