sys/fs/nfsclient/nfs_clrpcops.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (c) 1989, 1993
   5  *      The Regents of the University of California.  All rights reserved.
   6  *
   7  * This code is derived from software contributed to Berkeley by
   8  * Rick Macklem at The University of Guelph.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  * 1. Redistributions of source code must retain the above copyright
  14  *    notice, this list of conditions and the following disclaimer.
  15  * 2. Redistributions in binary form must reproduce the above copyright
  16  *    notice, this list of conditions and the following disclaimer in the
  17  *    documentation and/or other materials provided with the distribution.
  18  * 3. Neither the name of the University nor the names of its contributors
  19  *    may be used to endorse or promote products derived from this software
  20  *    without specific prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  */
  35
  36 #include <sys/cdefs.h>
  37 __FBSDID("$FreeBSD$");
  38
  39 /*
  40  * Rpc op calls, generally called from the vnode op calls or through the
  41  * buffer cache, for NFS v2, 3 and 4.
  42  * These do not normally make any changes to vnode arguments or use
  43  * structures that might change between the VFS variants. The returned
  44  * arguments are all at the end, after the NFSPROC_T *p one.
  45  */
  46
  47 #ifndef APPLEKEXT
  48 #include "opt_inet6.h"
  49
  50 #include <fs/nfs/nfsport.h>
  51 #include <sys/sysctl.h>
  52 #include <sys/taskqueue.h>
  53
  54 SYSCTL_DECL(_vfs_nfs);
  55
  56 static int      nfsignore_eexist = 0;
  57 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
  58     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
  59
  60 static int      nfscl_dssameconn = 0;
  61 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
  62     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
  63
  64 /*
  65  * Global variables
  66  */
  67 extern int nfs_numnfscbd;
  68 extern struct timeval nfsboottime;
  69 extern u_int32_t newnfs_false, newnfs_true;
  70 extern nfstype nfsv34_type[9];
  71 extern int nfsrv_useacl;
  72 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
  73 extern int nfscl_debuglevel;
  74 extern int nfs_pnfsiothreads;
  75 NFSCLSTATEMUTEX;
  76 int nfstest_outofseq = 0;
  77 int nfscl_assumeposixlocks = 1;
  78 int nfscl_enablecallb = 0;
  79 short nfsv4_cbport = NFSV4_CBPORT;
  80 int nfstest_openallsetattr = 0;
  81 #endif  /* !APPLEKEXT */
  82
  83 #define DIRHDSIZ        offsetof(struct dirent, d_name)
  84
  85 /*
  86  * nfscl_getsameserver() can return one of three values:
  87  * NFSDSP_USETHISSESSION - Use this session for the DS.
  88  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
  89  *     session.
  90  * NFSDSP_NOTFOUND - No matching server was found.
  91  */
  92 enum nfsclds_state {
  93         NFSDSP_USETHISSESSION = 0,
  94         NFSDSP_SEQTHISSESSION = 1,
  95         NFSDSP_NOTFOUND = 2,
  96 };
  97
  98 /*
  99  * Do a write RPC on a DS data file, using this structure for the arguments,
 100  * so that this function can be executed by a separate kernel process.
 101  */
 102 struct nfsclwritedsdorpc {
 103         int                     done;
 104         int                     inprog;
 105         struct task             tsk;
 106         struct vnode            *vp;
 107         int                     iomode;
 108         int                     must_commit;
 109         nfsv4stateid_t          *stateidp;
 110         struct nfsclds          *dsp;
 111         uint64_t                off;
 112         int                     len;
 113         struct nfsfh            *fhp;
 114         struct mbuf             *m;
 115         int                     vers;
 116         int                     minorvers;
 117         struct ucred            *cred;
 118         NFSPROC_T               *p;
 119         int                     err;
 120 };
 121
 122 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
 123     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
 124 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
 125     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
 126 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
 127     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
 128     void *);
 129 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
 130     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
 131     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
 132 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
 133     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
 134     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
 135     int *, void *, int *);
 136 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
 137     struct nfscllockowner *, u_int64_t, u_int64_t,
 138     u_int32_t, struct ucred *, NFSPROC_T *, int);
 139 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
 140     struct acl *, nfsv4stateid_t *, void *);
 141 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
 142     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
 143     struct ucred *, NFSPROC_T *);
 144 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
 145     struct sockaddr_in6 *, sa_family_t, int, struct nfsclds **, NFSPROC_T *);
 146 static void nfscl_initsessionslots(struct nfsclsession *);
 147 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
 148     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
 149     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
 150     NFSPROC_T *);
 151 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
 152     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
 153     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
 154     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
 155 static struct mbuf *nfsm_copym(struct mbuf *, int, int);
 156 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
 157     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
 158     struct ucred *, NFSPROC_T *);
 159 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
 160     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
 161     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
 162 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
 163     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
 164     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
 165 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
 166     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
 167     struct ucred *, NFSPROC_T *);
 168 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
 169     struct nfsclds *, struct nfsclds **, uint32_t *);
 170 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
 171     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
 172     NFSPROC_T *);
 173 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
 174     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
 175 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
 176     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
 177 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
 178     NFSPROC_T *);
 179 static int nfsrv_parselayoutget(struct nfsrv_descript *, nfsv4stateid_t *,
 180     int *, struct nfsclflayouthead *);
 181 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
 182     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
 183     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
 184 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
 185     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
 186     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
 187     struct nfsfh **, int *, int *, void *, int *);
 188 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
 189     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
 190     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
 191     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
 192 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
 193     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
 194     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
 195     struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
 196     int, int, int, int *, struct nfsclflayouthead *, int *);
 197 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
 198     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
 199     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
 200 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
 201     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
 202     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
 203
 204 int nfs_pnfsio(task_fn_t *, void *);
 205
 206 /*
 207  * nfs null call from vfs.
 208  */
 209 APPLESTATIC int
 210 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
 211 {
 212         int error;
 213         struct nfsrv_descript nfsd, *nd = &nfsd;
 214
 215         NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
 216         error = nfscl_request(nd, vp, p, cred, NULL);
 217         if (nd->nd_repstat && !error)
 218                 error = nd->nd_repstat;
 219         mbuf_freem(nd->nd_mrep);
 220         return (error);
 221 }
 222
 223 /*
 224  * nfs access rpc op.
 225  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
 226  * modes are changed on the server, accesses might still fail later.
 227  */
 228 APPLESTATIC int
 229 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
 230     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
 231 {
 232         int error;
 233         u_int32_t mode, rmode;
 234
 235         if (acmode & VREAD)
 236                 mode = NFSACCESS_READ;
 237         else
 238                 mode = 0;
 239         if (vnode_vtype(vp) == VDIR) {
 240                 if (acmode & VWRITE)
 241                         mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
 242                                  NFSACCESS_DELETE);
 243                 if (acmode & VEXEC)
 244                         mode |= NFSACCESS_LOOKUP;
 245         } else {
 246                 if (acmode & VWRITE)
 247                         mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
 248                 if (acmode & VEXEC)
 249                         mode |= NFSACCESS_EXECUTE;
 250         }
 251
 252         /*
 253          * Now, just call nfsrpc_accessrpc() to do the actual RPC.
 254          */
 255         error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
 256             NULL);
 257
 258         /*
 259          * The NFS V3 spec does not clarify whether or not
 260          * the returned access bits can be a superset of
 261          * the ones requested, so...
 262          */
 263         if (!error && (rmode & mode) != mode)
 264                 error = EACCES;
 265         return (error);
 266 }
 267
 268 /*
 269  * The actual rpc, separated out for Darwin.
 270  */
 271 APPLESTATIC int
 272 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
 273     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
 274     void *stuff)
 275 {
 276         u_int32_t *tl;
 277         u_int32_t supported, rmode;
 278         int error;
 279         struct nfsrv_descript nfsd, *nd = &nfsd;
 280         nfsattrbit_t attrbits;
 281
 282         *attrflagp = 0;
 283         supported = mode;
 284         NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
 285         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 286         *tl = txdr_unsigned(mode);
 287         if (nd->nd_flag & ND_NFSV4) {
 288                 /*
 289                  * And do a Getattr op.
 290                  */
 291                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 292                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
 293                 NFSGETATTR_ATTRBIT(&attrbits);
 294                 (void) nfsrv_putattrbit(nd, &attrbits);
 295         }
 296         error = nfscl_request(nd, vp, p, cred, stuff);
 297         if (error)
 298                 return (error);
 299         if (nd->nd_flag & ND_NFSV3) {
 300                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
 301                 if (error)
 302                         goto nfsmout;
 303         }
 304         if (!nd->nd_repstat) {
 305                 if (nd->nd_flag & ND_NFSV4) {
 306                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 307                         supported = fxdr_unsigned(u_int32_t, *tl++);
 308                 } else {
 309                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
 310                 }
 311                 rmode = fxdr_unsigned(u_int32_t, *tl);
 312                 if (nd->nd_flag & ND_NFSV4)
 313                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
 314
 315                 /*
 316                  * It's not obvious what should be done about
 317                  * unsupported access modes. For now, be paranoid
 318                  * and clear the unsupported ones.
 319                  */
 320                 rmode &= supported;
 321                 *rmodep = rmode;
 322         } else
 323                 error = nd->nd_repstat;
 324 nfsmout:
 325         mbuf_freem(nd->nd_mrep);
 326         return (error);
 327 }
 328
 329 /*
 330  * nfs open rpc
 331  */
 332 APPLESTATIC int
 333 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
 334 {
 335         struct nfsclopen *op;
 336         struct nfscldeleg *dp;
 337         struct nfsfh *nfhp;
 338         struct nfsnode *np = VTONFS(vp);
 339         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
 340         u_int32_t mode, clidrev;
 341         int ret, newone, error, expireret = 0, retrycnt;
 342
 343         /*
 344          * For NFSv4, Open Ops are only done on Regular Files.
 345          */
 346         if (vnode_vtype(vp) != VREG)
 347                 return (0);
 348         mode = 0;
 349         if (amode & FREAD)
 350                 mode |= NFSV4OPEN_ACCESSREAD;
 351         if (amode & FWRITE)
 352                 mode |= NFSV4OPEN_ACCESSWRITE;
 353         nfhp = np->n_fhp;
 354
 355         retrycnt = 0;
 356 #ifdef notdef
 357 { char name[100]; int namel;
 358 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
 359 bcopy(NFS4NODENAME(np->n_v4), name, namel);
 360 name[namel] = '\0';
 361 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
 362 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
 363 else printf(" fhl=0\n");
 364 }
 365 #endif
 366         do {
 367             dp = NULL;
 368             error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
 369                 cred, p, NULL, &op, &newone, &ret, 1);
 370             if (error) {
 371                 return (error);
 372             }
 373             if (nmp->nm_clp != NULL)
 374                 clidrev = nmp->nm_clp->nfsc_clientidrev;
 375             else
 376                 clidrev = 0;
 377             if (ret == NFSCLOPEN_DOOPEN) {
 378                 if (np->n_v4 != NULL) {
 379                         /*
 380                          * For the first attempt, try and get a layout, if
 381                          * pNFS is enabled for the mount.
 382                          */
 383                         if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
 384                             nfs_numnfscbd == 0 ||
 385                             (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
 386                                 error = nfsrpc_openrpc(nmp, vp,
 387                                     np->n_v4->n4_data,
 388                                     np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
 389                                     np->n_fhp->nfh_len, mode, op,
 390                                     NFS4NODENAME(np->n_v4),
 391                                     np->n_v4->n4_namelen,
 392                                     &dp, 0, 0x0, cred, p, 0, 0);
 393                         else
 394                                 error = nfsrpc_getopenlayout(nmp, vp,
 395                                     np->n_v4->n4_data,
 396                                     np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
 397                                     np->n_fhp->nfh_len, mode, op,
 398                                     NFS4NODENAME(np->n_v4),
 399                                     np->n_v4->n4_namelen, &dp, cred, p);
 400                         if (dp != NULL) {
 401 #ifdef APPLE
 402                                 OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
 403 #else
 404                                 NFSLOCKNODE(np);
 405                                 np->n_flag &= ~NDELEGMOD;
 406                                 /*
 407                                  * Invalidate the attribute cache, so that
 408                                  * attributes that pre-date the issue of a
 409                                  * delegation are not cached, since the
 410                                  * cached attributes will remain valid while
 411                                  * the delegation is held.
 412                                  */
 413                                 NFSINVALATTRCACHE(np);
 414                                 NFSUNLOCKNODE(np);
 415 #endif
 416                                 (void) nfscl_deleg(nmp->nm_mountp,
 417                                     op->nfso_own->nfsow_clp,
 418                                     nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
 419                         }
 420                 } else {
 421                         error = EIO;
 422                 }
 423                 newnfs_copyincred(cred, &op->nfso_cred);
 424             } else if (ret == NFSCLOPEN_SETCRED)
 425                 /*
 426                  * This is a new local open on a delegation. It needs
 427                  * to have credentials so that an open can be done
 428                  * against the server during recovery.
 429                  */
 430                 newnfs_copyincred(cred, &op->nfso_cred);
 431
 432             /*
 433              * nfso_opencnt is the count of how many VOP_OPEN()s have
 434              * been done on this Open successfully and a VOP_CLOSE()
 435              * is expected for each of these.
 436              * If error is non-zero, don't increment it, since the Open
 437              * hasn't succeeded yet.
 438              */
 439             if (!error)
 440                 op->nfso_opencnt++;
 441             nfscl_openrelease(nmp, op, error, newone);
 442             if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
 443                 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
 444                 error == NFSERR_BADSESSION) {
 445                 (void) nfs_catnap(PZERO, error, "nfs_open");
 446             } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
 447                 && clidrev != 0) {
 448                 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
 449                 retrycnt++;
 450             }
 451         } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
 452             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
 453             error == NFSERR_BADSESSION ||
 454             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
 455              expireret == 0 && clidrev != 0 && retrycnt < 4));
 456         if (error && retrycnt >= 4)
 457                 error = EIO;
 458         return (error);
 459 }
 460
 461 /*
 462  * the actual open rpc
 463  */
 464 APPLESTATIC int
 465 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
 466     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
 467     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
 468     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
 469     int syscred, int recursed)
 470 {
 471         u_int32_t *tl;
 472         struct nfsrv_descript nfsd, *nd = &nfsd;
 473         struct nfscldeleg *dp, *ndp = NULL;
 474         struct nfsvattr nfsva;
 475         u_int32_t rflags, deleg;
 476         nfsattrbit_t attrbits;
 477         int error, ret, acesize, limitby;
 478         struct nfsclsession *tsep;
 479
 480         dp = *dpp;
 481         *dpp = NULL;
 482         nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
 483         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
 484         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 485         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
 486         *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
 487         tsep = nfsmnt_mdssession(nmp);
 488         *tl++ = tsep->nfsess_clientid.lval[0];
 489         *tl = tsep->nfsess_clientid.lval[1];
 490         (void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
 491         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 492         *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
 493         if (reclaim) {
 494                 *tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
 495                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 496                 *tl = txdr_unsigned(delegtype);
 497         } else {
 498                 if (dp != NULL) {
 499                         *tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
 500                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
 501                         if (NFSHASNFSV4N(nmp))
 502                                 *tl++ = 0;
 503                         else
 504                                 *tl++ = dp->nfsdl_stateid.seqid;
 505                         *tl++ = dp->nfsdl_stateid.other[0];
 506                         *tl++ = dp->nfsdl_stateid.other[1];
 507                         *tl = dp->nfsdl_stateid.other[2];
 508                 } else {
 509                         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
 510                 }
 511                 (void) nfsm_strtom(nd, name, namelen);
 512         }
 513         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 514         *tl = txdr_unsigned(NFSV4OP_GETATTR);
 515         NFSZERO_ATTRBIT(&attrbits);
 516         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
 517         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
 518         (void) nfsrv_putattrbit(nd, &attrbits);
 519         if (syscred)
 520                 nd->nd_flag |= ND_USEGSSNAME;
 521         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
 522             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 523         if (error)
 524                 return (error);
 525         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 526         if (!nd->nd_repstat) {
 527                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
 528                     6 * NFSX_UNSIGNED);
 529                 op->nfso_stateid.seqid = *tl++;
 530                 op->nfso_stateid.other[0] = *tl++;
 531                 op->nfso_stateid.other[1] = *tl++;
 532                 op->nfso_stateid.other[2] = *tl;
 533                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
 534                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
 535                 if (error)
 536                         goto nfsmout;
 537                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
 538                 deleg = fxdr_unsigned(u_int32_t, *tl);
 539                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
 540                     deleg == NFSV4OPEN_DELEGATEWRITE) {
 541                         if (!(op->nfso_own->nfsow_clp->nfsc_flags &
 542                               NFSCLFLAGS_FIRSTDELEG))
 543                                 op->nfso_own->nfsow_clp->nfsc_flags |=
 544                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
 545                         ndp = malloc(
 546                             sizeof (struct nfscldeleg) + newfhlen,
 547                             M_NFSCLDELEG, M_WAITOK);
 548                         LIST_INIT(&ndp->nfsdl_owner);
 549                         LIST_INIT(&ndp->nfsdl_lock);
 550                         ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
 551                         ndp->nfsdl_fhlen = newfhlen;
 552                         NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
 553                         newnfs_copyincred(cred, &ndp->nfsdl_cred);
 554                         nfscl_lockinit(&ndp->nfsdl_rwlock);
 555                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
 556                             NFSX_UNSIGNED);
 557                         ndp->nfsdl_stateid.seqid = *tl++;
 558                         ndp->nfsdl_stateid.other[0] = *tl++;
 559                         ndp->nfsdl_stateid.other[1] = *tl++;
 560                         ndp->nfsdl_stateid.other[2] = *tl++;
 561                         ret = fxdr_unsigned(int, *tl);
 562                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
 563                                 ndp->nfsdl_flags = NFSCLDL_WRITE;
 564                                 /*
 565                                  * Indicates how much the file can grow.
 566                                  */
 567                                 NFSM_DISSECT(tl, u_int32_t *,
 568                                     3 * NFSX_UNSIGNED);
 569                                 limitby = fxdr_unsigned(int, *tl++);
 570                                 switch (limitby) {
 571                                 case NFSV4OPEN_LIMITSIZE:
 572                                         ndp->nfsdl_sizelimit = fxdr_hyper(tl);
 573                                         break;
 574                                 case NFSV4OPEN_LIMITBLOCKS:
 575                                         ndp->nfsdl_sizelimit =
 576                                             fxdr_unsigned(u_int64_t, *tl++);
 577                                         ndp->nfsdl_sizelimit *=
 578                                             fxdr_unsigned(u_int64_t, *tl);
 579                                         break;
 580                                 default:
 581                                         error = NFSERR_BADXDR;
 582                                         goto nfsmout;
 583                                 }
 584                         } else {
 585                                 ndp->nfsdl_flags = NFSCLDL_READ;
 586                         }
 587                         if (ret)
 588                                 ndp->nfsdl_flags |= NFSCLDL_RECALL;
 589                         error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
 590                             &acesize, p);
 591                         if (error)
 592                                 goto nfsmout;
 593                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
 594                         error = NFSERR_BADXDR;
 595                         goto nfsmout;
 596                 }
 597                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 598                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
 599                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
 600                     NULL, NULL, NULL, p, cred);
 601                 if (error)
 602                         goto nfsmout;
 603                 if (ndp != NULL) {
 604                         ndp->nfsdl_change = nfsva.na_filerev;
 605                         ndp->nfsdl_modtime = nfsva.na_mtime;
 606                         ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
 607                 }
 608                 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
 609                     do {
 610                         ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
 611                             cred, p);
 612                         if (ret == NFSERR_DELAY)
 613                             (void) nfs_catnap(PZERO, ret, "nfs_open");
 614                     } while (ret == NFSERR_DELAY);
 615                     error = ret;
 616                 }
 617                 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
 618                     nfscl_assumeposixlocks)
 619                     op->nfso_posixlock = 1;
 620                 else
 621                     op->nfso_posixlock = 0;
 622
 623                 /*
 624                  * If the server is handing out delegations, but we didn't
 625                  * get one because an OpenConfirm was required, try the
 626                  * Open again, to get a delegation. This is a harmless no-op,
 627                  * from a server's point of view.
 628                  */
 629                 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
 630                     (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
 631                     && !error && dp == NULL && ndp == NULL && !recursed) {
 632                     do {
 633                         ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
 634                             newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
 635                             cred, p, syscred, 1);
 636                         if (ret == NFSERR_DELAY)
 637                             (void) nfs_catnap(PZERO, ret, "nfs_open2");
 638                     } while (ret == NFSERR_DELAY);
 639                     if (ret) {
 640                         if (ndp != NULL) {
 641                                 free(ndp, M_NFSCLDELEG);
 642                                 ndp = NULL;
 643                         }
 644                         if (ret == NFSERR_STALECLIENTID ||
 645                             ret == NFSERR_STALEDONTRECOVER ||
 646                             ret == NFSERR_BADSESSION)
 647                                 error = ret;
 648                     }
 649                 }
 650         }
 651         if (nd->nd_repstat != 0 && error == 0)
 652                 error = nd->nd_repstat;
 653         if (error == NFSERR_STALECLIENTID)
 654                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 655 nfsmout:
 656         if (!error)
 657                 *dpp = ndp;
 658         else if (ndp != NULL)
 659                 free(ndp, M_NFSCLDELEG);
 660         mbuf_freem(nd->nd_mrep);
 661         return (error);
 662 }
 663
 664 /*
 665  * open downgrade rpc
 666  */
 667 APPLESTATIC int
 668 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
 669     struct ucred *cred, NFSPROC_T *p)
 670 {
 671         u_int32_t *tl;
 672         struct nfsrv_descript nfsd, *nd = &nfsd;
 673         int error;
 674
 675         NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
 676         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
 677         if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
 678                 *tl++ = 0;
 679         else
 680                 *tl++ = op->nfso_stateid.seqid;
 681         *tl++ = op->nfso_stateid.other[0];
 682         *tl++ = op->nfso_stateid.other[1];
 683         *tl++ = op->nfso_stateid.other[2];
 684         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 685         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
 686         *tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
 687         error = nfscl_request(nd, vp, p, cred, NULL);
 688         if (error)
 689                 return (error);
 690         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 691         if (!nd->nd_repstat) {
 692                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 693                 op->nfso_stateid.seqid = *tl++;
 694                 op->nfso_stateid.other[0] = *tl++;
 695                 op->nfso_stateid.other[1] = *tl++;
 696                 op->nfso_stateid.other[2] = *tl;
 697         }
 698         if (nd->nd_repstat && error == 0)
 699                 error = nd->nd_repstat;
 700         if (error == NFSERR_STALESTATEID)
 701                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 702 nfsmout:
 703         mbuf_freem(nd->nd_mrep);
 704         return (error);
 705 }
 706
 707 /*
 708  * V4 Close operation.
 709  */
 710 APPLESTATIC int
 711 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
 712 {
 713         struct nfsclclient *clp;
 714         int error;
 715
 716         if (vnode_vtype(vp) != VREG)
 717                 return (0);
 718         if (doclose)
 719                 error = nfscl_doclose(vp, &clp, p);
 720         else
 721                 error = nfscl_getclose(vp, &clp);
 722         if (error)
 723                 return (error);
 724
 725         nfscl_clientrelease(clp);
 726         return (0);
 727 }
 728
 729 /*
 730  * Close the open.
 731  */
 732 APPLESTATIC void
 733 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
 734 {
 735         struct nfsrv_descript nfsd, *nd = &nfsd;
 736         struct nfscllockowner *lp, *nlp;
 737         struct nfscllock *lop, *nlop;
 738         struct ucred *tcred;
 739         u_int64_t off = 0, len = 0;
 740         u_int32_t type = NFSV4LOCKT_READ;
 741         int error, do_unlock, trycnt;
 742
 743         tcred = newnfs_getcred();
 744         newnfs_copycred(&op->nfso_cred, tcred);
 745         /*
 746          * (Theoretically this could be done in the same
 747          *  compound as the close, but having multiple
 748          *  sequenced Ops in the same compound might be
 749          *  too scary for some servers.)
 750          */
 751         if (op->nfso_posixlock) {
 752                 off = 0;
 753                 len = NFS64BITSSET;
 754                 type = NFSV4LOCKT_READ;
 755         }
 756
 757         /*
 758          * Since this function is only called from VOP_INACTIVE(), no
 759          * other thread will be manipulating this Open. As such, the
 760          * lock lists are not being changed by other threads, so it should
 761          * be safe to do this without locking.
 762          */
 763         LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
 764                 do_unlock = 1;
 765                 LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
 766                         if (op->nfso_posixlock == 0) {
 767                                 off = lop->nfslo_first;
 768                                 len = lop->nfslo_end - lop->nfslo_first;
 769                                 if (lop->nfslo_type == F_WRLCK)
 770                                         type = NFSV4LOCKT_WRITE;
 771                                 else
 772                                         type = NFSV4LOCKT_READ;
 773                         }
 774                         if (do_unlock) {
 775                                 trycnt = 0;
 776                                 do {
 777                                         error = nfsrpc_locku(nd, nmp, lp, off,
 778                                             len, type, tcred, p, 0);
 779                                         if ((nd->nd_repstat == NFSERR_GRACE ||
 780                                             nd->nd_repstat == NFSERR_DELAY) &&
 781                                             error == 0)
 782                                                 (void) nfs_catnap(PZERO,
 783                                                     (int)nd->nd_repstat,
 784                                                     "nfs_close");
 785                                 } while ((nd->nd_repstat == NFSERR_GRACE ||
 786                                     nd->nd_repstat == NFSERR_DELAY) &&
 787                                     error == 0 && trycnt++ < 5);
 788                                 if (op->nfso_posixlock)
 789                                         do_unlock = 0;
 790                         }
 791                         nfscl_freelock(lop, 0);
 792                 }
 793                 /*
 794                  * Do a ReleaseLockOwner.
 795                  * The lock owner name nfsl_owner may be used by other opens for
 796                  * other files but the lock_owner4 name that nfsrpc_rellockown()
 797                  * puts on the wire has the file handle for this file appended
 798                  * to it, so it can be done now.
 799                  */
 800                 (void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
 801                     lp->nfsl_open->nfso_fhlen, tcred, p);
 802         }
 803
 804         /*
 805          * There could be other Opens for different files on the same
 806          * OpenOwner, so locking is required.
 807          */
 808         NFSLOCKCLSTATE();
 809         nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
 810         NFSUNLOCKCLSTATE();
 811         do {
 812                 error = nfscl_tryclose(op, tcred, nmp, p);
 813                 if (error == NFSERR_GRACE)
 814                         (void) nfs_catnap(PZERO, error, "nfs_close");
 815         } while (error == NFSERR_GRACE);
 816         NFSLOCKCLSTATE();
 817         nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
 818
 819         LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
 820                 nfscl_freelockowner(lp, 0);
 821         nfscl_freeopen(op, 0);
 822         NFSUNLOCKCLSTATE();
 823         NFSFREECRED(tcred);
 824 }
 825
 826 /*
 827  * The actual Close RPC.
 828  */
 829 APPLESTATIC int
 830 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
 831     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
 832     int syscred)
 833 {
 834         u_int32_t *tl;
 835         int error;
 836
 837         nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
 838             op->nfso_fhlen, NULL, NULL, 0, 0);
 839         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
 840         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 841         if (NFSHASNFSV4N(nmp))
 842                 *tl++ = 0;
 843         else
 844                 *tl++ = op->nfso_stateid.seqid;
 845         *tl++ = op->nfso_stateid.other[0];
 846         *tl++ = op->nfso_stateid.other[1];
 847         *tl = op->nfso_stateid.other[2];
 848         if (syscred)
 849                 nd->nd_flag |= ND_USEGSSNAME;
 850         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
 851             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 852         if (error)
 853                 return (error);
 854         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 855         if (nd->nd_repstat == 0)
 856                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 857         error = nd->nd_repstat;
 858         if (error == NFSERR_STALESTATEID)
 859                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 860 nfsmout:
 861         mbuf_freem(nd->nd_mrep);
 862         return (error);
 863 }
 864
 865 /*
 866  * V4 Open Confirm RPC.
 867  */
 868 APPLESTATIC int
 869 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
 870     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
 871 {
 872         u_int32_t *tl;
 873         struct nfsrv_descript nfsd, *nd = &nfsd;
 874         struct nfsmount *nmp;
 875         int error;
 876
 877         nmp = VFSTONFS(vnode_mount(vp));
 878         if (NFSHASNFSV4N(nmp))
 879                 return (0);             /* No confirmation for NFSv4.1. */
 880         nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
 881             0, 0);
 882         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
 883         *tl++ = op->nfso_stateid.seqid;
 884         *tl++ = op->nfso_stateid.other[0];
 885         *tl++ = op->nfso_stateid.other[1];
 886         *tl++ = op->nfso_stateid.other[2];
 887         *tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
 888         error = nfscl_request(nd, vp, p, cred, NULL);
 889         if (error)
 890                 return (error);
 891         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 892         if (!nd->nd_repstat) {
 893                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 894                 op->nfso_stateid.seqid = *tl++;
 895                 op->nfso_stateid.other[0] = *tl++;
 896                 op->nfso_stateid.other[1] = *tl++;
 897                 op->nfso_stateid.other[2] = *tl;
 898         }
 899         error = nd->nd_repstat;
 900         if (error == NFSERR_STALESTATEID)
 901                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 902 nfsmout:
 903         mbuf_freem(nd->nd_mrep);
 904         return (error);
 905 }
 906
 907 /*
 908  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
 909  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
 910  */
 911 APPLESTATIC int
 912 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
 913     struct ucred *cred, NFSPROC_T *p)
 914 {
 915         u_int32_t *tl;
 916         struct nfsrv_descript nfsd;
 917         struct nfsrv_descript *nd = &nfsd;
 918         nfsattrbit_t attrbits;
 919         u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
 920         u_short port;
 921         int error, isinet6 = 0, callblen;
 922         nfsquad_t confirm;
 923         u_int32_t lease;
 924         static u_int32_t rev = 0;
 925         struct nfsclds *dsp;
 926         struct in6_addr a6;
 927         struct nfsclsession *tsep;
 928
 929         if (nfsboottime.tv_sec == 0)
 930                 NFSSETBOOTTIME(nfsboottime);
 931         clp->nfsc_rev = rev++;
 932         if (NFSHASNFSV4N(nmp)) {
 933                 /*
 934                  * Either there was no previous session or the
 935                  * previous session has failed, so...
 936                  * do an ExchangeID followed by the CreateSession.
 937                  */
 938                 error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
 939                     NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
 940                 NFSCL_DEBUG(1, "aft exch=%d\n", error);
 941                 if (error == 0)
 942                         error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
 943                             &nmp->nm_sockreq,
 944                             dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
 945                 if (error == 0) {
 946                         NFSLOCKMNT(nmp);
 947                         /*
 948                          * The old sessions cannot be safely free'd
 949                          * here, since they may still be used by
 950                          * in-progress RPCs.
 951                          */
 952                         tsep = NULL;
 953                         if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
 954                                 tsep = NFSMNT_MDSSESSION(nmp);
 955                         TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
 956                             nfsclds_list);
 957                         /*
 958                          * Wake up RPCs waiting for a slot on the
 959                          * old session. These will then fail with
 960                          * NFSERR_BADSESSION and be retried with the
 961                          * new session by nfsv4_setsequence().
 962                          * Also wakeup() processes waiting for the
 963                          * new session.
 964                          */
 965                         if (tsep != NULL)
 966                                 wakeup(&tsep->nfsess_slots);
 967                         wakeup(&nmp->nm_sess);
 968                         NFSUNLOCKMNT(nmp);
 969                 } else
 970                         nfscl_freenfsclds(dsp);
 971                 NFSCL_DEBUG(1, "aft createsess=%d\n", error);
 972                 if (error == 0 && reclaim == 0) {
 973                         error = nfsrpc_reclaimcomplete(nmp, cred, p);
 974                         NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
 975                         if (error == NFSERR_COMPLETEALREADY ||
 976                             error == NFSERR_NOTSUPP)
 977                                 /* Ignore this error. */
 978                                 error = 0;
 979                 }
 980                 return (error);
 981         }
 982
 983         /*
 984          * Allocate a single session structure for NFSv4.0, because some of
 985          * the fields are used by NFSv4.0 although it doesn't do a session.
 986          */
 987         dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
 988         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
 989         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
 990         NFSLOCKMNT(nmp);
 991         TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
 992         tsep = NFSMNT_MDSSESSION(nmp);
 993         NFSUNLOCKMNT(nmp);
 994
 995         nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
 996         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 997         *tl++ = txdr_unsigned(nfsboottime.tv_sec);
 998         *tl = txdr_unsigned(clp->nfsc_rev);
 999         (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1000
1001         /*
1002          * set up the callback address
1003          */
1004         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1005         *tl = txdr_unsigned(NFS_CALLBCKPROG);
1006         callblen = strlen(nfsv4_callbackaddr);
1007         if (callblen == 0)
1008                 cp = nfscl_getmyip(nmp, &a6, &isinet6);
1009         if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1010             (callblen > 0 || cp != NULL)) {
1011                 port = htons(nfsv4_cbport);
1012                 cp2 = (u_int8_t *)&port;
1013 #ifdef INET6
1014                 if ((callblen > 0 &&
1015                      strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1016                         char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1017
1018                         (void) nfsm_strtom(nd, "tcp6", 4);
1019                         if (callblen == 0) {
1020                                 ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1021                                 ip6add = ip6buf;
1022                         } else {
1023                                 ip6add = nfsv4_callbackaddr;
1024                         }
1025                         snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1026                             ip6add, cp2[0], cp2[1]);
1027                 } else
1028 #endif
1029                 {
1030                         (void) nfsm_strtom(nd, "tcp", 3);
1031                         if (callblen == 0)
1032                                 snprintf(addr, INET6_ADDRSTRLEN + 9,
1033                                     "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1034                                     cp[2], cp[3], cp2[0], cp2[1]);
1035                         else
1036                                 snprintf(addr, INET6_ADDRSTRLEN + 9,
1037                                     "%s.%d.%d", nfsv4_callbackaddr,
1038                                     cp2[0], cp2[1]);
1039                 }
1040                 (void) nfsm_strtom(nd, addr, strlen(addr));
1041         } else {
1042                 (void) nfsm_strtom(nd, "tcp", 3);
1043                 (void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1044         }
1045         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1046         *tl = txdr_unsigned(clp->nfsc_cbident);
1047         nd->nd_flag |= ND_USEGSSNAME;
1048         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1049                 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1050         if (error)
1051                 return (error);
1052         if (nd->nd_repstat == 0) {
1053             NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1054             tsep->nfsess_clientid.lval[0] = *tl++;
1055             tsep->nfsess_clientid.lval[1] = *tl++;
1056             confirm.lval[0] = *tl++;
1057             confirm.lval[1] = *tl;
1058             mbuf_freem(nd->nd_mrep);
1059             nd->nd_mrep = NULL;
1060
1061             /*
1062              * and confirm it.
1063              */
1064             nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1065                 NULL, 0, 0);
1066             NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1067             *tl++ = tsep->nfsess_clientid.lval[0];
1068             *tl++ = tsep->nfsess_clientid.lval[1];
1069             *tl++ = confirm.lval[0];
1070             *tl = confirm.lval[1];
1071             nd->nd_flag |= ND_USEGSSNAME;
1072             error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1073                 cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1074             if (error)
1075                 return (error);
1076             mbuf_freem(nd->nd_mrep);
1077             nd->nd_mrep = NULL;
1078             if (nd->nd_repstat == 0) {
1079                 nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
1080                     nmp->nm_fhsize, NULL, NULL, 0, 0);
1081                 NFSZERO_ATTRBIT(&attrbits);
1082                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1083                 (void) nfsrv_putattrbit(nd, &attrbits);
1084                 nd->nd_flag |= ND_USEGSSNAME;
1085                 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1086                     cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1087                 if (error)
1088                     return (error);
1089                 if (nd->nd_repstat == 0) {
1090                     error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
1091                         NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1092                     if (error)
1093                         goto nfsmout;
1094                     clp->nfsc_renew = NFSCL_RENEW(lease);
1095                     clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1096                     clp->nfsc_clientidrev++;
1097                     if (clp->nfsc_clientidrev == 0)
1098                         clp->nfsc_clientidrev++;
1099                 }
1100             }
1101         }
1102         error = nd->nd_repstat;
1103 nfsmout:
1104         mbuf_freem(nd->nd_mrep);
1105         return (error);
1106 }
1107
1108 /*
1109  * nfs getattr call.
1110  */
1111 APPLESTATIC int
1112 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1113     struct nfsvattr *nap, void *stuff)
1114 {
1115         struct nfsrv_descript nfsd, *nd = &nfsd;
1116         int error;
1117         nfsattrbit_t attrbits;
1118
1119         NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1120         if (nd->nd_flag & ND_NFSV4) {
1121                 NFSGETATTR_ATTRBIT(&attrbits);
1122                 (void) nfsrv_putattrbit(nd, &attrbits);
1123         }
1124         error = nfscl_request(nd, vp, p, cred, stuff);
1125         if (error)
1126                 return (error);
1127         if (!nd->nd_repstat)
1128                 error = nfsm_loadattr(nd, nap);
1129         else
1130                 error = nd->nd_repstat;
1131         mbuf_freem(nd->nd_mrep);
1132         return (error);
1133 }
1134
1135 /*
1136  * nfs getattr call with non-vnode arguemnts.
1137  */
1138 APPLESTATIC int
1139 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1140     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1141     uint32_t *leasep)
1142 {
1143         struct nfsrv_descript nfsd, *nd = &nfsd;
1144         int error, vers = NFS_VER2;
1145         nfsattrbit_t attrbits;
1146
1147         nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1148         if (nd->nd_flag & ND_NFSV4) {
1149                 vers = NFS_VER4;
1150                 NFSGETATTR_ATTRBIT(&attrbits);
1151                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1152                 (void) nfsrv_putattrbit(nd, &attrbits);
1153         } else if (nd->nd_flag & ND_NFSV3) {
1154                 vers = NFS_VER3;
1155         }
1156         if (syscred)
1157                 nd->nd_flag |= ND_USEGSSNAME;
1158         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1159             NFS_PROG, vers, NULL, 1, xidp, NULL);
1160         if (error)
1161                 return (error);
1162         if (nd->nd_repstat == 0) {
1163                 if ((nd->nd_flag & ND_NFSV4) != 0)
1164                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1165                             NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1166                             NULL, NULL);
1167                 else
1168                         error = nfsm_loadattr(nd, nap);
1169         } else
1170                 error = nd->nd_repstat;
1171         mbuf_freem(nd->nd_mrep);
1172         return (error);
1173 }
1174
1175 /*
1176  * Do an nfs setattr operation.
1177  */
1178 APPLESTATIC int
1179 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1180     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1181     void *stuff)
1182 {
1183         int error, expireret = 0, openerr, retrycnt;
1184         u_int32_t clidrev = 0, mode;
1185         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1186         struct nfsfh *nfhp;
1187         nfsv4stateid_t stateid;
1188         void *lckp;
1189
1190         if (nmp->nm_clp != NULL)
1191                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1192         if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1193                 mode = NFSV4OPEN_ACCESSWRITE;
1194         else
1195                 mode = NFSV4OPEN_ACCESSREAD;
1196         retrycnt = 0;
1197         do {
1198                 lckp = NULL;
1199                 openerr = 1;
1200                 if (NFSHASNFSV4(nmp)) {
1201                         nfhp = VTONFS(vp)->n_fhp;
1202                         error = nfscl_getstateid(vp, nfhp->nfh_fh,
1203                             nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1204                         if (error && vnode_vtype(vp) == VREG &&
1205                             (mode == NFSV4OPEN_ACCESSWRITE ||
1206                              nfstest_openallsetattr)) {
1207                                 /*
1208                                  * No Open stateid, so try and open the file
1209                                  * now.
1210                                  */
1211                                 if (mode == NFSV4OPEN_ACCESSWRITE)
1212                                         openerr = nfsrpc_open(vp, FWRITE, cred,
1213                                             p);
1214                                 else
1215                                         openerr = nfsrpc_open(vp, FREAD, cred,
1216                                             p);
1217                                 if (!openerr)
1218                                         (void) nfscl_getstateid(vp,
1219                                             nfhp->nfh_fh, nfhp->nfh_len,
1220                                             mode, 0, cred, p, &stateid, &lckp);
1221                         }
1222                 }
1223                 if (vap != NULL)
1224                         error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1225                             rnap, attrflagp, stuff);
1226                 else
1227                         error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1228                             stuff);
1229                 if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1230                         NFSLOCKMNT(nmp);
1231                         nmp->nm_state |= NFSSTA_OPENMODE;
1232                         NFSUNLOCKMNT(nmp);
1233                 }
1234                 if (error == NFSERR_STALESTATEID)
1235                         nfscl_initiate_recovery(nmp->nm_clp);
1236                 if (lckp != NULL)
1237                         nfscl_lockderef(lckp);
1238                 if (!openerr)
1239                         (void) nfsrpc_close(vp, 0, p);
1240                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1241                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1242                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1243                         (void) nfs_catnap(PZERO, error, "nfs_setattr");
1244                 } else if ((error == NFSERR_EXPIRED ||
1245                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1246                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1247                 }
1248                 retrycnt++;
1249         } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1250             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1251             error == NFSERR_BADSESSION ||
1252             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1253             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1254              expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1255             (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1256              retrycnt < 4));
1257         if (error && retrycnt >= 4)
1258                 error = EIO;
1259         return (error);
1260 }
1261
1262 static int
1263 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1264     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1265     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1266 {
1267         u_int32_t *tl;
1268         struct nfsrv_descript nfsd, *nd = &nfsd;
1269         int error;
1270         nfsattrbit_t attrbits;
1271
1272         *attrflagp = 0;
1273         NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1274         if (nd->nd_flag & ND_NFSV4)
1275                 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1276         vap->va_type = vnode_vtype(vp);
1277         nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1278         if (nd->nd_flag & ND_NFSV3) {
1279                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1280                 *tl = newnfs_false;
1281         } else if (nd->nd_flag & ND_NFSV4) {
1282                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1283                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1284                 NFSGETATTR_ATTRBIT(&attrbits);
1285                 (void) nfsrv_putattrbit(nd, &attrbits);
1286         }
1287         error = nfscl_request(nd, vp, p, cred, stuff);
1288         if (error)
1289                 return (error);
1290         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1291                 error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1292         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1293                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1294         if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1295                 error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1296         mbuf_freem(nd->nd_mrep);
1297         if (nd->nd_repstat && !error)
1298                 error = nd->nd_repstat;
1299         return (error);
1300 }
1301
1302 /*
1303  * nfs lookup rpc
1304  */
1305 APPLESTATIC int
1306 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1307     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1308     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1309 {
1310         u_int32_t *tl;
1311         struct nfsrv_descript nfsd, *nd = &nfsd;
1312         struct nfsmount *nmp;
1313         struct nfsnode *np;
1314         struct nfsfh *nfhp;
1315         nfsattrbit_t attrbits;
1316         int error = 0, lookupp = 0;
1317
1318         *attrflagp = 0;
1319         *dattrflagp = 0;
1320         if (vnode_vtype(dvp) != VDIR)
1321                 return (ENOTDIR);
1322         nmp = VFSTONFS(vnode_mount(dvp));
1323         if (len > NFS_MAXNAMLEN)
1324                 return (ENAMETOOLONG);
1325         if (NFSHASNFSV4(nmp) && len == 1 &&
1326                 name[0] == '.') {
1327                 /*
1328                  * Just return the current dir's fh.
1329                  */
1330                 np = VTONFS(dvp);
1331                 nfhp = malloc(sizeof (struct nfsfh) +
1332                         np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1333                 nfhp->nfh_len = np->n_fhp->nfh_len;
1334                 NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1335                 *nfhpp = nfhp;
1336                 return (0);
1337         }
1338         if (NFSHASNFSV4(nmp) && len == 2 &&
1339                 name[0] == '.' && name[1] == '.') {
1340                 lookupp = 1;
1341                 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1342         } else {
1343                 NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1344                 (void) nfsm_strtom(nd, name, len);
1345         }
1346         if (nd->nd_flag & ND_NFSV4) {
1347                 NFSGETATTR_ATTRBIT(&attrbits);
1348                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1349                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1350                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1351                 (void) nfsrv_putattrbit(nd, &attrbits);
1352         }
1353         error = nfscl_request(nd, dvp, p, cred, stuff);
1354         if (error)
1355                 return (error);
1356         if (nd->nd_repstat) {
1357                 /*
1358                  * When an NFSv4 Lookupp returns ENOENT, it means that
1359                  * the lookup is at the root of an fs, so return this dir.
1360                  */
1361                 if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1362                     np = VTONFS(dvp);
1363                     nfhp = malloc(sizeof (struct nfsfh) +
1364                         np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1365                     nfhp->nfh_len = np->n_fhp->nfh_len;
1366                     NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1367                     *nfhpp = nfhp;
1368                     mbuf_freem(nd->nd_mrep);
1369                     return (0);
1370                 }
1371                 if (nd->nd_flag & ND_NFSV3)
1372                     error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1373                 else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1374                     ND_NFSV4) {
1375                         /* Load the directory attributes. */
1376                         error = nfsm_loadattr(nd, dnap);
1377                         if (error == 0)
1378                                 *dattrflagp = 1;
1379                 }
1380                 goto nfsmout;
1381         }
1382         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1383                 /* Load the directory attributes. */
1384                 error = nfsm_loadattr(nd, dnap);
1385                 if (error != 0)
1386                         goto nfsmout;
1387                 *dattrflagp = 1;
1388                 /* Skip over the Lookup and GetFH operation status values. */
1389                 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1390         }
1391         error = nfsm_getfh(nd, nfhpp);
1392         if (error)
1393                 goto nfsmout;
1394
1395         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1396         if ((nd->nd_flag & ND_NFSV3) && !error)
1397                 error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1398 nfsmout:
1399         mbuf_freem(nd->nd_mrep);
1400         if (!error && nd->nd_repstat)
1401                 error = nd->nd_repstat;
1402         return (error);
1403 }
1404
1405 /*
1406  * Do a readlink rpc.
1407  */
1408 APPLESTATIC int
1409 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1410     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1411 {
1412         u_int32_t *tl;
1413         struct nfsrv_descript nfsd, *nd = &nfsd;
1414         struct nfsnode *np = VTONFS(vp);
1415         nfsattrbit_t attrbits;
1416         int error, len, cangetattr = 1;
1417
1418         *attrflagp = 0;
1419         NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1420         if (nd->nd_flag & ND_NFSV4) {
1421                 /*
1422                  * And do a Getattr op.
1423                  */
1424                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1425                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1426                 NFSGETATTR_ATTRBIT(&attrbits);
1427                 (void) nfsrv_putattrbit(nd, &attrbits);
1428         }
1429         error = nfscl_request(nd, vp, p, cred, stuff);
1430         if (error)
1431                 return (error);
1432         if (nd->nd_flag & ND_NFSV3)
1433                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1434         if (!nd->nd_repstat && !error) {
1435                 NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1436                 /*
1437                  * This seems weird to me, but must have been added to
1438                  * FreeBSD for some reason. The only thing I can think of
1439                  * is that there was/is some server that replies with
1440                  * more link data than it should?
1441                  */
1442                 if (len == NFS_MAXPATHLEN) {
1443                         NFSLOCKNODE(np);
1444                         if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1445                                 len = np->n_size;
1446                                 cangetattr = 0;
1447                         }
1448                         NFSUNLOCKNODE(np);
1449                 }
1450                 error = nfsm_mbufuio(nd, uiop, len);
1451                 if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1452                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1453         }
1454         if (nd->nd_repstat && !error)
1455                 error = nd->nd_repstat;
1456 nfsmout:
1457         mbuf_freem(nd->nd_mrep);
1458         return (error);
1459 }
1460
1461 /*
1462  * Read operation.
1463  */
1464 APPLESTATIC int
1465 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1466     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1467 {
1468         int error, expireret = 0, retrycnt;
1469         u_int32_t clidrev = 0;
1470         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1471         struct nfsnode *np = VTONFS(vp);
1472         struct ucred *newcred;
1473         struct nfsfh *nfhp = NULL;
1474         nfsv4stateid_t stateid;
1475         void *lckp;
1476
1477         if (nmp->nm_clp != NULL)
1478                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1479         newcred = cred;
1480         if (NFSHASNFSV4(nmp)) {
1481                 nfhp = np->n_fhp;
1482                 newcred = NFSNEWCRED(cred);
1483         }
1484         retrycnt = 0;
1485         do {
1486                 lckp = NULL;
1487                 if (NFSHASNFSV4(nmp))
1488                         (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1489                             NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1490                             &lckp);
1491                 error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1492                     attrflagp, stuff);
1493                 if (error == NFSERR_OPENMODE) {
1494                         NFSLOCKMNT(nmp);
1495                         nmp->nm_state |= NFSSTA_OPENMODE;
1496                         NFSUNLOCKMNT(nmp);
1497                 }
1498                 if (error == NFSERR_STALESTATEID)
1499                         nfscl_initiate_recovery(nmp->nm_clp);
1500                 if (lckp != NULL)
1501                         nfscl_lockderef(lckp);
1502                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1503                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1504                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1505                         (void) nfs_catnap(PZERO, error, "nfs_read");
1506                 } else if ((error == NFSERR_EXPIRED ||
1507                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1508                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1509                 }
1510                 retrycnt++;
1511         } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1512             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1513             error == NFSERR_BADSESSION ||
1514             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1515             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1516              expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1517             (error == NFSERR_OPENMODE && retrycnt < 4));
1518         if (error && retrycnt >= 4)
1519                 error = EIO;
1520         if (NFSHASNFSV4(nmp))
1521                 NFSFREECRED(newcred);
1522         return (error);
1523 }
1524
1525 /*
1526  * The actual read RPC.
1527  */
1528 static int
1529 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1530     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1531     int *attrflagp, void *stuff)
1532 {
1533         u_int32_t *tl;
1534         int error = 0, len, retlen, tsiz, eof = 0;
1535         struct nfsrv_descript nfsd;
1536         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1537         struct nfsrv_descript *nd = &nfsd;
1538         int rsize;
1539         off_t tmp_off;
1540
1541         *attrflagp = 0;
1542         tsiz = uio_uio_resid(uiop);
1543         tmp_off = uiop->uio_offset + tsiz;
1544         NFSLOCKMNT(nmp);
1545         if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1546                 NFSUNLOCKMNT(nmp);
1547                 return (EFBIG);
1548         }
1549         rsize = nmp->nm_rsize;
1550         NFSUNLOCKMNT(nmp);
1551         nd->nd_mrep = NULL;
1552         while (tsiz > 0) {
1553                 *attrflagp = 0;
1554                 len = (tsiz > rsize) ? rsize : tsiz;
1555                 NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1556                 if (nd->nd_flag & ND_NFSV4)
1557                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1558                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1559                 if (nd->nd_flag & ND_NFSV2) {
1560                         *tl++ = txdr_unsigned(uiop->uio_offset);
1561                         *tl++ = txdr_unsigned(len);
1562                         *tl = 0;
1563                 } else {
1564                         txdr_hyper(uiop->uio_offset, tl);
1565                         *(tl + 2) = txdr_unsigned(len);
1566                 }
1567                 /*
1568                  * Since I can't do a Getattr for NFSv4 for Write, there
1569                  * doesn't seem any point in doing one here, either.
1570                  * (See the comment in nfsrpc_writerpc() for more info.)
1571                  */
1572                 error = nfscl_request(nd, vp, p, cred, stuff);
1573                 if (error)
1574                         return (error);
1575                 if (nd->nd_flag & ND_NFSV3) {
1576                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1577                 } else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1578                         error = nfsm_loadattr(nd, nap);
1579                         if (!error)
1580                                 *attrflagp = 1;
1581                 }
1582                 if (nd->nd_repstat || error) {
1583                         if (!error)
1584                                 error = nd->nd_repstat;
1585                         goto nfsmout;
1586                 }
1587                 if (nd->nd_flag & ND_NFSV3) {
1588                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1589                         eof = fxdr_unsigned(int, *(tl + 1));
1590                 } else if (nd->nd_flag & ND_NFSV4) {
1591                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1592                         eof = fxdr_unsigned(int, *tl);
1593                 }
1594                 NFSM_STRSIZ(retlen, len);
1595                 error = nfsm_mbufuio(nd, uiop, retlen);
1596                 if (error)
1597                         goto nfsmout;
1598                 mbuf_freem(nd->nd_mrep);
1599                 nd->nd_mrep = NULL;
1600                 tsiz -= retlen;
1601                 if (!(nd->nd_flag & ND_NFSV2)) {
1602                         if (eof || retlen == 0)
1603                                 tsiz = 0;
1604                 } else if (retlen < len)
1605                         tsiz = 0;
1606         }
1607         return (0);
1608 nfsmout:
1609         if (nd->nd_mrep != NULL)
1610                 mbuf_freem(nd->nd_mrep);
1611         return (error);
1612 }
1613
1614 /*
1615  * nfs write operation
1616  * When called_from_strategy != 0, it should return EIO for an error that
1617  * indicates recovery is in progress, so that the buffer will be left
1618  * dirty and be written back to the server later. If it loops around,
1619  * the recovery thread could get stuck waiting for the buffer and recovery
1620  * will then deadlock.
1621  */
1622 APPLESTATIC int
1623 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1624     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1625     void *stuff, int called_from_strategy)
1626 {
1627         int error, expireret = 0, retrycnt, nostateid;
1628         u_int32_t clidrev = 0;
1629         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1630         struct nfsnode *np = VTONFS(vp);
1631         struct ucred *newcred;
1632         struct nfsfh *nfhp = NULL;
1633         nfsv4stateid_t stateid;
1634         void *lckp;
1635
1636         *must_commit = 0;
1637         if (nmp->nm_clp != NULL)
1638                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1639         newcred = cred;
1640         if (NFSHASNFSV4(nmp)) {
1641                 newcred = NFSNEWCRED(cred);
1642                 nfhp = np->n_fhp;
1643         }
1644         retrycnt = 0;
1645         do {
1646                 lckp = NULL;
1647                 nostateid = 0;
1648                 if (NFSHASNFSV4(nmp)) {
1649                         (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1650                             NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1651                             &lckp);
1652                         if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1653                             stateid.other[2] == 0) {
1654                                 nostateid = 1;
1655                                 NFSCL_DEBUG(1, "stateid0 in write\n");
1656                         }
1657                 }
1658
1659                 /*
1660                  * If there is no stateid for NFSv4, it means this is an
1661                  * extraneous write after close. Basically a poorly
1662                  * implemented buffer cache. Just don't do the write.
1663                  */
1664                 if (nostateid)
1665                         error = 0;
1666                 else
1667                         error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1668                             newcred, &stateid, p, nap, attrflagp, stuff);
1669                 if (error == NFSERR_STALESTATEID)
1670                         nfscl_initiate_recovery(nmp->nm_clp);
1671                 if (lckp != NULL)
1672                         nfscl_lockderef(lckp);
1673                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1674                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1675                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1676                         (void) nfs_catnap(PZERO, error, "nfs_write");
1677                 } else if ((error == NFSERR_EXPIRED ||
1678                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1679                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1680                 }
1681                 retrycnt++;
1682         } while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1683             ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1684               error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1685             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1686             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1687              expireret == 0 && clidrev != 0 && retrycnt < 4));
1688         if (error != 0 && (retrycnt >= 4 ||
1689             ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1690               error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1691                 error = EIO;
1692         if (NFSHASNFSV4(nmp))
1693                 NFSFREECRED(newcred);
1694         return (error);
1695 }
1696
1697 /*
1698  * The actual write RPC.
1699  */
1700 static int
1701 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1702     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1703     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1704 {
1705         u_int32_t *tl;
1706         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1707         struct nfsnode *np = VTONFS(vp);
1708         int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1709         int wccflag = 0, wsize;
1710         int32_t backup;
1711         struct nfsrv_descript nfsd;
1712         struct nfsrv_descript *nd = &nfsd;
1713         nfsattrbit_t attrbits;
1714         off_t tmp_off;
1715
1716         KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1717         *attrflagp = 0;
1718         tsiz = uio_uio_resid(uiop);
1719         tmp_off = uiop->uio_offset + tsiz;
1720         NFSLOCKMNT(nmp);
1721         if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1722                 NFSUNLOCKMNT(nmp);
1723                 return (EFBIG);
1724         }
1725         wsize = nmp->nm_wsize;
1726         NFSUNLOCKMNT(nmp);
1727         nd->nd_mrep = NULL;     /* NFSv2 sometimes does a write with */
1728         nd->nd_repstat = 0;     /* uio_resid == 0, so the while is not done */
1729         while (tsiz > 0) {
1730                 *attrflagp = 0;
1731                 len = (tsiz > wsize) ? wsize : tsiz;
1732                 NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1733                 if (nd->nd_flag & ND_NFSV4) {
1734                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1735                         NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1736                         txdr_hyper(uiop->uio_offset, tl);
1737                         tl += 2;
1738                         *tl++ = txdr_unsigned(*iomode);
1739                         *tl = txdr_unsigned(len);
1740                 } else if (nd->nd_flag & ND_NFSV3) {
1741                         NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1742                         txdr_hyper(uiop->uio_offset, tl);
1743                         tl += 2;
1744                         *tl++ = txdr_unsigned(len);
1745                         *tl++ = txdr_unsigned(*iomode);
1746                         *tl = txdr_unsigned(len);
1747                 } else {
1748                         u_int32_t x;
1749
1750                         NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1751                         /*
1752                          * Not sure why someone changed this, since the
1753                          * RFC clearly states that "beginoffset" and
1754                          * "totalcount" are ignored, but it wouldn't
1755                          * surprise me if there's a busted server out there.
1756                          */
1757                         /* Set both "begin" and "current" to non-garbage. */
1758                         x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1759                         *tl++ = x;      /* "begin offset" */
1760                         *tl++ = x;      /* "current offset" */
1761                         x = txdr_unsigned(len);
1762                         *tl++ = x;      /* total to this offset */
1763                         *tl = x;        /* size of this write */
1764
1765                 }
1766                 nfsm_uiombuf(nd, uiop, len);
1767                 /*
1768                  * Although it is tempting to do a normal Getattr Op in the
1769                  * NFSv4 compound, the result can be a nearly hung client
1770                  * system if the Getattr asks for Owner and/or OwnerGroup.
1771                  * It occurs when the client can't map either the Owner or
1772                  * Owner_group name in the Getattr reply to a uid/gid. When
1773                  * there is a cache miss, the kernel does an upcall to the
1774                  * nfsuserd. Then, it can try and read the local /etc/passwd
1775                  * or /etc/group file. It can then block in getnewbuf(),
1776                  * waiting for dirty writes to be pushed to the NFS server.
1777                  * The only reason this doesn't result in a complete
1778                  * deadlock, is that the upcall times out and allows
1779                  * the write to complete. However, progress is so slow
1780                  * that it might just as well be deadlocked.
1781                  * As such, we get the rest of the attributes, but not
1782                  * Owner or Owner_group.
1783                  * nb: nfscl_loadattrcache() needs to be told that these
1784                  *     partial attributes from a write rpc are being
1785                  *     passed in, via a argument flag.
1786                  */
1787                 if (nd->nd_flag & ND_NFSV4) {
1788                         NFSWRITEGETATTR_ATTRBIT(&attrbits);
1789                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1790                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
1791                         (void) nfsrv_putattrbit(nd, &attrbits);
1792                 }
1793                 error = nfscl_request(nd, vp, p, cred, stuff);
1794                 if (error)
1795                         return (error);
1796                 if (nd->nd_repstat) {
1797                         /*
1798                          * In case the rpc gets retried, roll
1799                          * the uio fileds changed by nfsm_uiombuf()
1800                          * back.
1801                          */
1802                         uiop->uio_offset -= len;
1803                         uio_uio_resid_add(uiop, len);
1804                         uio_iov_base_add(uiop, -len);
1805                         uio_iov_len_add(uiop, len);
1806                 }
1807                 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1808                         error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1809                             &wccflag, stuff);
1810                         if (error)
1811                                 goto nfsmout;
1812                 }
1813                 if (!nd->nd_repstat) {
1814                         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1815                                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1816                                         + NFSX_VERF);
1817                                 rlen = fxdr_unsigned(int, *tl++);
1818                                 if (rlen == 0) {
1819                                         error = NFSERR_IO;
1820                                         goto nfsmout;
1821                                 } else if (rlen < len) {
1822                                         backup = len - rlen;
1823                                         uio_iov_base_add(uiop, -(backup));
1824                                         uio_iov_len_add(uiop, backup);
1825                                         uiop->uio_offset -= backup;
1826                                         uio_uio_resid_add(uiop, backup);
1827                                         len = rlen;
1828                                 }
1829                                 commit = fxdr_unsigned(int, *tl++);
1830
1831                                 /*
1832                                  * Return the lowest commitment level
1833                                  * obtained by any of the RPCs.
1834                                  */
1835                                 if (committed == NFSWRITE_FILESYNC)
1836                                         committed = commit;
1837                                 else if (committed == NFSWRITE_DATASYNC &&
1838                                         commit == NFSWRITE_UNSTABLE)
1839                                         committed = commit;
1840                                 NFSLOCKMNT(nmp);
1841                                 if (!NFSHASWRITEVERF(nmp)) {
1842                                         NFSBCOPY((caddr_t)tl,
1843                                             (caddr_t)&nmp->nm_verf[0],
1844                                             NFSX_VERF);
1845                                         NFSSETWRITEVERF(nmp);
1846                                 } else if (NFSBCMP(tl, nmp->nm_verf,
1847                                     NFSX_VERF)) {
1848                                         *must_commit = 1;
1849                                         NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1850                                 }
1851                                 NFSUNLOCKMNT(nmp);
1852                         }
1853                         if (nd->nd_flag & ND_NFSV4)
1854                                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1855                         if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1856                                 error = nfsm_loadattr(nd, nap);
1857                                 if (!error)
1858                                         *attrflagp = NFS_LATTR_NOSHRINK;
1859                         }
1860                 } else {
1861                         error = nd->nd_repstat;
1862                 }
1863                 if (error)
1864                         goto nfsmout;
1865                 NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1866                 mbuf_freem(nd->nd_mrep);
1867                 nd->nd_mrep = NULL;
1868                 tsiz -= len;
1869         }
1870 nfsmout:
1871         if (nd->nd_mrep != NULL)
1872                 mbuf_freem(nd->nd_mrep);
1873         *iomode = committed;
1874         if (nd->nd_repstat && !error)
1875                 error = nd->nd_repstat;
1876         return (error);
1877 }
1878
1879 /*
1880  * nfs mknod rpc
1881  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1882  * mode set to specify the file type and the size field for rdev.
1883  */
1884 APPLESTATIC int
1885 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1886     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1887     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1888     int *attrflagp, int *dattrflagp, void *dstuff)
1889 {
1890         u_int32_t *tl;
1891         int error = 0;
1892         struct nfsrv_descript nfsd, *nd = &nfsd;
1893         nfsattrbit_t attrbits;
1894
1895         *nfhpp = NULL;
1896         *attrflagp = 0;
1897         *dattrflagp = 0;
1898         if (namelen > NFS_MAXNAMLEN)
1899                 return (ENAMETOOLONG);
1900         NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1901         if (nd->nd_flag & ND_NFSV4) {
1902                 if (vtyp == VBLK || vtyp == VCHR) {
1903                         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1904                         *tl++ = vtonfsv34_type(vtyp);
1905                         *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1906                         *tl = txdr_unsigned(NFSMINOR(rdev));
1907                 } else {
1908                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1909                         *tl = vtonfsv34_type(vtyp);
1910                 }
1911         }
1912         (void) nfsm_strtom(nd, name, namelen);
1913         if (nd->nd_flag & ND_NFSV3) {
1914                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1915                 *tl = vtonfsv34_type(vtyp);
1916         }
1917         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1918                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
1919         if ((nd->nd_flag & ND_NFSV3) &&
1920             (vtyp == VCHR || vtyp == VBLK)) {
1921                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1922                 *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1923                 *tl = txdr_unsigned(NFSMINOR(rdev));
1924         }
1925         if (nd->nd_flag & ND_NFSV4) {
1926                 NFSGETATTR_ATTRBIT(&attrbits);
1927                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1928                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1929                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1930                 (void) nfsrv_putattrbit(nd, &attrbits);
1931         }
1932         if (nd->nd_flag & ND_NFSV2)
1933                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1934         error = nfscl_request(nd, dvp, p, cred, dstuff);
1935         if (error)
1936                 return (error);
1937         if (nd->nd_flag & ND_NFSV4)
1938                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1939         if (!nd->nd_repstat) {
1940                 if (nd->nd_flag & ND_NFSV4) {
1941                         NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1942                         error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1943                         if (error)
1944                                 goto nfsmout;
1945                 }
1946                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1947                 if (error)
1948                         goto nfsmout;
1949         }
1950         if (nd->nd_flag & ND_NFSV3)
1951                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1952         if (!error && nd->nd_repstat)
1953                 error = nd->nd_repstat;
1954 nfsmout:
1955         mbuf_freem(nd->nd_mrep);
1956         return (error);
1957 }
1958
1959 /*
1960  * nfs file create call
1961  * Mostly just call the approriate routine. (I separated out v4, so that
1962  * error recovery wouldn't be as difficult.)
1963  */
1964 APPLESTATIC int
1965 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1966     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1967     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1968     int *attrflagp, int *dattrflagp, void *dstuff)
1969 {
1970         int error = 0, newone, expireret = 0, retrycnt, unlocked;
1971         struct nfsclowner *owp;
1972         struct nfscldeleg *dp;
1973         struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1974         u_int32_t clidrev;
1975
1976         if (NFSHASNFSV4(nmp)) {
1977             retrycnt = 0;
1978             do {
1979                 dp = NULL;
1980                 error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1981                     NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1982                     NULL, 1);
1983                 if (error)
1984                         return (error);
1985                 if (nmp->nm_clp != NULL)
1986                         clidrev = nmp->nm_clp->nfsc_clientidrev;
1987                 else
1988                         clidrev = 0;
1989                 if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
1990                     nfs_numnfscbd == 0 || retrycnt > 0)
1991                         error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
1992                           fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1993                           attrflagp, dattrflagp, dstuff, &unlocked);
1994                 else
1995                         error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
1996                           cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1997                           attrflagp, dattrflagp, dstuff, &unlocked);
1998                 /*
1999                  * There is no need to invalidate cached attributes here,
2000                  * since new post-delegation issue attributes are always
2001                  * returned by nfsrpc_createv4() and these will update the
2002                  * attribute cache.
2003                  */
2004                 if (dp != NULL)
2005                         (void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2006                             (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2007                 nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2008                 if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2009                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2010                     error == NFSERR_BADSESSION) {
2011                         (void) nfs_catnap(PZERO, error, "nfs_open");
2012                 } else if ((error == NFSERR_EXPIRED ||
2013                     error == NFSERR_BADSTATEID) && clidrev != 0) {
2014                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2015                         retrycnt++;
2016                 }
2017             } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2018                 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2019                 error == NFSERR_BADSESSION ||
2020                 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2021                  expireret == 0 && clidrev != 0 && retrycnt < 4));
2022             if (error && retrycnt >= 4)
2023                     error = EIO;
2024         } else {
2025                 error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2026                     fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2027                     dstuff);
2028         }
2029         return (error);
2030 }
2031
2032 /*
2033  * The create rpc for v2 and 3.
2034  */
2035 static int
2036 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2037     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2038     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2039     int *attrflagp, int *dattrflagp, void *dstuff)
2040 {
2041         u_int32_t *tl;
2042         int error = 0;
2043         struct nfsrv_descript nfsd, *nd = &nfsd;
2044
2045         *nfhpp = NULL;
2046         *attrflagp = 0;
2047         *dattrflagp = 0;
2048         if (namelen > NFS_MAXNAMLEN)
2049                 return (ENAMETOOLONG);
2050         NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2051         (void) nfsm_strtom(nd, name, namelen);
2052         if (nd->nd_flag & ND_NFSV3) {
2053                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2054                 if (fmode & O_EXCL) {
2055                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2056                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2057                         *tl++ = cverf.lval[0];
2058                         *tl = cverf.lval[1];
2059                 } else {
2060                         *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2061                         nfscl_fillsattr(nd, vap, dvp, 0, 0);
2062                 }
2063         } else {
2064                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2065         }
2066         error = nfscl_request(nd, dvp, p, cred, dstuff);
2067         if (error)
2068                 return (error);
2069         if (nd->nd_repstat == 0) {
2070                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2071                 if (error)
2072                         goto nfsmout;
2073         }
2074         if (nd->nd_flag & ND_NFSV3)
2075                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2076         if (nd->nd_repstat != 0 && error == 0)
2077                 error = nd->nd_repstat;
2078 nfsmout:
2079         mbuf_freem(nd->nd_mrep);
2080         return (error);
2081 }
2082
2083 static int
2084 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2085     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2086     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2087     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2088     int *dattrflagp, void *dstuff, int *unlockedp)
2089 {
2090         u_int32_t *tl;
2091         int error = 0, deleg, newone, ret, acesize, limitby;
2092         struct nfsrv_descript nfsd, *nd = &nfsd;
2093         struct nfsclopen *op;
2094         struct nfscldeleg *dp = NULL;
2095         struct nfsnode *np;
2096         struct nfsfh *nfhp;
2097         nfsattrbit_t attrbits;
2098         nfsv4stateid_t stateid;
2099         u_int32_t rflags;
2100         struct nfsmount *nmp;
2101         struct nfsclsession *tsep;
2102
2103         nmp = VFSTONFS(dvp->v_mount);
2104         np = VTONFS(dvp);
2105         *unlockedp = 0;
2106         *nfhpp = NULL;
2107         *dpp = NULL;
2108         *attrflagp = 0;
2109         *dattrflagp = 0;
2110         if (namelen > NFS_MAXNAMLEN)
2111                 return (ENAMETOOLONG);
2112         NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2113         /*
2114          * For V4, this is actually an Open op.
2115          */
2116         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2117         *tl++ = txdr_unsigned(owp->nfsow_seqid);
2118         *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2119             NFSV4OPEN_ACCESSREAD);
2120         *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2121         tsep = nfsmnt_mdssession(nmp);
2122         *tl++ = tsep->nfsess_clientid.lval[0];
2123         *tl = tsep->nfsess_clientid.lval[1];
2124         (void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2125         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2126         *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2127         if (fmode & O_EXCL) {
2128                 if (NFSHASNFSV4N(nmp)) {
2129                         if (NFSHASSESSPERSIST(nmp)) {
2130                                 /* Use GUARDED for persistent sessions. */
2131                                 *tl = txdr_unsigned(NFSCREATE_GUARDED);
2132                                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2133                         } else {
2134                                 /* Otherwise, use EXCLUSIVE4_1. */
2135                                 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2136                                 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2137                                 *tl++ = cverf.lval[0];
2138                                 *tl = cverf.lval[1];
2139                                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2140                         }
2141                 } else {
2142                         /* NFSv4.0 */
2143                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2144                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2145                         *tl++ = cverf.lval[0];
2146                         *tl = cverf.lval[1];
2147                 }
2148         } else {
2149                 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2150                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2151         }
2152         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2153         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2154         (void) nfsm_strtom(nd, name, namelen);
2155         /* Get the new file's handle and attributes. */
2156         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2157         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2158         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2159         NFSGETATTR_ATTRBIT(&attrbits);
2160         (void) nfsrv_putattrbit(nd, &attrbits);
2161         /* Get the directory's post-op attributes. */
2162         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2163         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2164         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2165         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2166         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2167         (void) nfsrv_putattrbit(nd, &attrbits);
2168         error = nfscl_request(nd, dvp, p, cred, dstuff);
2169         if (error)
2170                 return (error);
2171         NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2172         if (nd->nd_repstat == 0) {
2173                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2174                     6 * NFSX_UNSIGNED);
2175                 stateid.seqid = *tl++;
2176                 stateid.other[0] = *tl++;
2177                 stateid.other[1] = *tl++;
2178                 stateid.other[2] = *tl;
2179                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2180                 (void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2181                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2182                 deleg = fxdr_unsigned(int, *tl);
2183                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
2184                     deleg == NFSV4OPEN_DELEGATEWRITE) {
2185                         if (!(owp->nfsow_clp->nfsc_flags &
2186                               NFSCLFLAGS_FIRSTDELEG))
2187                                 owp->nfsow_clp->nfsc_flags |=
2188                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2189                         dp = malloc(
2190                             sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2191                             M_NFSCLDELEG, M_WAITOK);
2192                         LIST_INIT(&dp->nfsdl_owner);
2193                         LIST_INIT(&dp->nfsdl_lock);
2194                         dp->nfsdl_clp = owp->nfsow_clp;
2195                         newnfs_copyincred(cred, &dp->nfsdl_cred);
2196                         nfscl_lockinit(&dp->nfsdl_rwlock);
2197                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2198                             NFSX_UNSIGNED);
2199                         dp->nfsdl_stateid.seqid = *tl++;
2200                         dp->nfsdl_stateid.other[0] = *tl++;
2201                         dp->nfsdl_stateid.other[1] = *tl++;
2202                         dp->nfsdl_stateid.other[2] = *tl++;
2203                         ret = fxdr_unsigned(int, *tl);
2204                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2205                                 dp->nfsdl_flags = NFSCLDL_WRITE;
2206                                 /*
2207                                  * Indicates how much the file can grow.
2208                                  */
2209                                 NFSM_DISSECT(tl, u_int32_t *,
2210                                     3 * NFSX_UNSIGNED);
2211                                 limitby = fxdr_unsigned(int, *tl++);
2212                                 switch (limitby) {
2213                                 case NFSV4OPEN_LIMITSIZE:
2214                                         dp->nfsdl_sizelimit = fxdr_hyper(tl);
2215                                         break;
2216                                 case NFSV4OPEN_LIMITBLOCKS:
2217                                         dp->nfsdl_sizelimit =
2218                                             fxdr_unsigned(u_int64_t, *tl++);
2219                                         dp->nfsdl_sizelimit *=
2220                                             fxdr_unsigned(u_int64_t, *tl);
2221                                         break;
2222                                 default:
2223                                         error = NFSERR_BADXDR;
2224                                         goto nfsmout;
2225                                 }
2226                         } else {
2227                                 dp->nfsdl_flags = NFSCLDL_READ;
2228                         }
2229                         if (ret)
2230                                 dp->nfsdl_flags |= NFSCLDL_RECALL;
2231                         error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2232                             &acesize, p);
2233                         if (error)
2234                                 goto nfsmout;
2235                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
2236                         error = NFSERR_BADXDR;
2237                         goto nfsmout;
2238                 }
2239                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2240                 if (error)
2241                         goto nfsmout;
2242                 /* Get rid of the PutFH and Getattr status values. */
2243                 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2244                 /* Load the directory attributes. */
2245                 error = nfsm_loadattr(nd, dnap);
2246                 if (error)
2247                         goto nfsmout;
2248                 *dattrflagp = 1;
2249                 if (dp != NULL && *attrflagp) {
2250                         dp->nfsdl_change = nnap->na_filerev;
2251                         dp->nfsdl_modtime = nnap->na_mtime;
2252                         dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2253                 }
2254                 /*
2255                  * We can now complete the Open state.
2256                  */
2257                 nfhp = *nfhpp;
2258                 if (dp != NULL) {
2259                         dp->nfsdl_fhlen = nfhp->nfh_len;
2260                         NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2261                 }
2262                 /*
2263                  * Get an Open structure that will be
2264                  * attached to the OpenOwner, acquired already.
2265                  */
2266                 error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2267                     (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2268                     cred, p, NULL, &op, &newone, NULL, 0);
2269                 if (error)
2270                         goto nfsmout;
2271                 op->nfso_stateid = stateid;
2272                 newnfs_copyincred(cred, &op->nfso_cred);
2273                 if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2274                     do {
2275                         ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2276                             nfhp->nfh_len, op, cred, p);
2277                         if (ret == NFSERR_DELAY)
2278                             (void) nfs_catnap(PZERO, ret, "nfs_create");
2279                     } while (ret == NFSERR_DELAY);
2280                     error = ret;
2281                 }
2282
2283                 /*
2284                  * If the server is handing out delegations, but we didn't
2285                  * get one because an OpenConfirm was required, try the
2286                  * Open again, to get a delegation. This is a harmless no-op,
2287                  * from a server's point of view.
2288                  */
2289                 if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2290                     (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2291                     !error && dp == NULL) {
2292                     do {
2293                         ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2294                             np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2295                             nfhp->nfh_fh, nfhp->nfh_len,
2296                             (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2297                             name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2298                         if (ret == NFSERR_DELAY)
2299                             (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2300                     } while (ret == NFSERR_DELAY);
2301                     if (ret) {
2302                         if (dp != NULL) {
2303                                 free(dp, M_NFSCLDELEG);
2304                                 dp = NULL;
2305                         }
2306                         if (ret == NFSERR_STALECLIENTID ||
2307                             ret == NFSERR_STALEDONTRECOVER ||
2308                             ret == NFSERR_BADSESSION)
2309                                 error = ret;
2310                     }
2311                 }
2312                 nfscl_openrelease(nmp, op, error, newone);
2313                 *unlockedp = 1;
2314         }
2315         if (nd->nd_repstat != 0 && error == 0)
2316                 error = nd->nd_repstat;
2317         if (error == NFSERR_STALECLIENTID)
2318                 nfscl_initiate_recovery(owp->nfsow_clp);
2319 nfsmout:
2320         if (!error)
2321                 *dpp = dp;
2322         else if (dp != NULL)
2323                 free(dp, M_NFSCLDELEG);
2324         mbuf_freem(nd->nd_mrep);
2325         return (error);
2326 }
2327
2328 /*
2329  * Nfs remove rpc
2330  */
2331 APPLESTATIC int
2332 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2333     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2334     void *dstuff)
2335 {
2336         u_int32_t *tl;
2337         struct nfsrv_descript nfsd, *nd = &nfsd;
2338         struct nfsnode *np;
2339         struct nfsmount *nmp;
2340         nfsv4stateid_t dstateid;
2341         int error, ret = 0, i;
2342
2343         *dattrflagp = 0;
2344         if (namelen > NFS_MAXNAMLEN)
2345                 return (ENAMETOOLONG);
2346         nmp = VFSTONFS(vnode_mount(dvp));
2347 tryagain:
2348         if (NFSHASNFSV4(nmp) && ret == 0) {
2349                 ret = nfscl_removedeleg(vp, p, &dstateid);
2350                 if (ret == 1) {
2351                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2352                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2353                             NFSX_UNSIGNED);
2354                         if (NFSHASNFSV4N(nmp))
2355                                 *tl++ = 0;
2356                         else
2357                                 *tl++ = dstateid.seqid;
2358                         *tl++ = dstateid.other[0];
2359                         *tl++ = dstateid.other[1];
2360                         *tl++ = dstateid.other[2];
2361                         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2362                         np = VTONFS(dvp);
2363                         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2364                             np->n_fhp->nfh_len, 0);
2365                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2366                         *tl = txdr_unsigned(NFSV4OP_REMOVE);
2367                 }
2368         } else {
2369                 ret = 0;
2370         }
2371         if (ret == 0)
2372                 NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2373         (void) nfsm_strtom(nd, name, namelen);
2374         error = nfscl_request(nd, dvp, p, cred, dstuff);
2375         if (error)
2376                 return (error);
2377         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2378                 /* For NFSv4, parse out any Delereturn replies. */
2379                 if (ret > 0 && nd->nd_repstat != 0 &&
2380                     (nd->nd_flag & ND_NOMOREDATA)) {
2381                         /*
2382                          * If the Delegreturn failed, try again without
2383                          * it. The server will Recall, as required.
2384                          */
2385                         mbuf_freem(nd->nd_mrep);
2386                         goto tryagain;
2387                 }
2388                 for (i = 0; i < (ret * 2); i++) {
2389                         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2390                             ND_NFSV4) {
2391                             NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2392                             if (*(tl + 1))
2393                                 nd->nd_flag |= ND_NOMOREDATA;
2394                         }
2395                 }
2396                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2397         }
2398         if (nd->nd_repstat && !error)
2399                 error = nd->nd_repstat;
2400 nfsmout:
2401         mbuf_freem(nd->nd_mrep);
2402         return (error);
2403 }
2404
2405 /*
2406  * Do an nfs rename rpc.
2407  */
2408 APPLESTATIC int
2409 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2410     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2411     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2412     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2413 {
2414         u_int32_t *tl;
2415         struct nfsrv_descript nfsd, *nd = &nfsd;
2416         struct nfsmount *nmp;
2417         struct nfsnode *np;
2418         nfsattrbit_t attrbits;
2419         nfsv4stateid_t fdstateid, tdstateid;
2420         int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2421
2422         *fattrflagp = 0;
2423         *tattrflagp = 0;
2424         nmp = VFSTONFS(vnode_mount(fdvp));
2425         if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2426                 return (ENAMETOOLONG);
2427 tryagain:
2428         if (NFSHASNFSV4(nmp) && ret == 0) {
2429                 ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2430                     &tdstateid, &gottd, p);
2431                 if (gotfd && gottd) {
2432                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2433                 } else if (gotfd) {
2434                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2435                 } else if (gottd) {
2436                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2437                 }
2438                 if (gotfd) {
2439                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2440                         if (NFSHASNFSV4N(nmp))
2441                                 *tl++ = 0;
2442                         else
2443                                 *tl++ = fdstateid.seqid;
2444                         *tl++ = fdstateid.other[0];
2445                         *tl++ = fdstateid.other[1];
2446                         *tl = fdstateid.other[2];
2447                         if (gottd) {
2448                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2449                                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2450                                 np = VTONFS(tvp);
2451                                 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2452                                     np->n_fhp->nfh_len, 0);
2453                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2454                                 *tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2455                         }
2456                 }
2457                 if (gottd) {
2458                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2459                         if (NFSHASNFSV4N(nmp))
2460                                 *tl++ = 0;
2461                         else
2462                                 *tl++ = tdstateid.seqid;
2463                         *tl++ = tdstateid.other[0];
2464                         *tl++ = tdstateid.other[1];
2465                         *tl = tdstateid.other[2];
2466                 }
2467                 if (ret > 0) {
2468                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2469                         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2470                         np = VTONFS(fdvp);
2471                         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2472                             np->n_fhp->nfh_len, 0);
2473                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2474                         *tl = txdr_unsigned(NFSV4OP_SAVEFH);
2475                 }
2476         } else {
2477                 ret = 0;
2478         }
2479         if (ret == 0)
2480                 NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2481         if (nd->nd_flag & ND_NFSV4) {
2482                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2483                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2484                 NFSWCCATTR_ATTRBIT(&attrbits);
2485                 (void) nfsrv_putattrbit(nd, &attrbits);
2486                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2487                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2488                 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2489                     VTONFS(tdvp)->n_fhp->nfh_len, 0);
2490                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2491                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2492                 (void) nfsrv_putattrbit(nd, &attrbits);
2493                 nd->nd_flag |= ND_V4WCCATTR;
2494                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2495                 *tl = txdr_unsigned(NFSV4OP_RENAME);
2496         }
2497         (void) nfsm_strtom(nd, fnameptr, fnamelen);
2498         if (!(nd->nd_flag & ND_NFSV4))
2499                 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2500                         VTONFS(tdvp)->n_fhp->nfh_len, 0);
2501         (void) nfsm_strtom(nd, tnameptr, tnamelen);
2502         error = nfscl_request(nd, fdvp, p, cred, fstuff);
2503         if (error)
2504                 return (error);
2505         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2506                 /* For NFSv4, parse out any Delereturn replies. */
2507                 if (ret > 0 && nd->nd_repstat != 0 &&
2508                     (nd->nd_flag & ND_NOMOREDATA)) {
2509                         /*
2510                          * If the Delegreturn failed, try again without
2511                          * it. The server will Recall, as required.
2512                          */
2513                         mbuf_freem(nd->nd_mrep);
2514                         goto tryagain;
2515                 }
2516                 for (i = 0; i < (ret * 2); i++) {
2517                         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2518                             ND_NFSV4) {
2519                             NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2520                             if (*(tl + 1)) {
2521                                 if (i == 0 && ret > 1) {
2522                                     /*
2523                                      * If the Delegreturn failed, try again
2524                                      * without it. The server will Recall, as
2525                                      * required.
2526                                      * If ret > 1, the first iteration of this
2527                                      * loop is the second DelegReturn result.
2528                                      */
2529                                     mbuf_freem(nd->nd_mrep);
2530                                     goto tryagain;
2531                                 } else {
2532                                     nd->nd_flag |= ND_NOMOREDATA;
2533                                 }
2534                             }
2535                         }
2536                 }
2537                 /* Now, the first wcc attribute reply. */
2538                 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2539                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2540                         if (*(tl + 1))
2541                                 nd->nd_flag |= ND_NOMOREDATA;
2542                 }
2543                 error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2544                     fstuff);
2545                 /* and the second wcc attribute reply. */
2546                 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2547                     !error) {
2548                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2549                         if (*(tl + 1))
2550                                 nd->nd_flag |= ND_NOMOREDATA;
2551                 }
2552                 if (!error)
2553                         error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2554                             NULL, tstuff);
2555         }
2556         if (nd->nd_repstat && !error)
2557                 error = nd->nd_repstat;
2558 nfsmout:
2559         mbuf_freem(nd->nd_mrep);
2560         return (error);
2561 }
2562
2563 /*
2564  * nfs hard link create rpc
2565  */
2566 APPLESTATIC int
2567 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2568     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2569     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2570 {
2571         u_int32_t *tl;
2572         struct nfsrv_descript nfsd, *nd = &nfsd;
2573         nfsattrbit_t attrbits;
2574         int error = 0;
2575
2576         *attrflagp = 0;
2577         *dattrflagp = 0;
2578         if (namelen > NFS_MAXNAMLEN)
2579                 return (ENAMETOOLONG);
2580         NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2581         if (nd->nd_flag & ND_NFSV4) {
2582                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2583                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2584         }
2585         (void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2586                 VTONFS(dvp)->n_fhp->nfh_len, 0);
2587         if (nd->nd_flag & ND_NFSV4) {
2588                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2589                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2590                 NFSWCCATTR_ATTRBIT(&attrbits);
2591                 (void) nfsrv_putattrbit(nd, &attrbits);
2592                 nd->nd_flag |= ND_V4WCCATTR;
2593                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2594                 *tl = txdr_unsigned(NFSV4OP_LINK);
2595         }
2596         (void) nfsm_strtom(nd, name, namelen);
2597         error = nfscl_request(nd, vp, p, cred, dstuff);
2598         if (error)
2599                 return (error);
2600         if (nd->nd_flag & ND_NFSV3) {
2601                 error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2602                 if (!error)
2603                         error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2604                             NULL, dstuff);
2605         } else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2606                 /*
2607                  * First, parse out the PutFH and Getattr result.
2608                  */
2609                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2610                 if (!(*(tl + 1)))
2611                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2612                 if (*(tl + 1))
2613                         nd->nd_flag |= ND_NOMOREDATA;
2614                 /*
2615                  * Get the pre-op attributes.
2616                  */
2617                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2618         }
2619         if (nd->nd_repstat && !error)
2620                 error = nd->nd_repstat;
2621 nfsmout:
2622         mbuf_freem(nd->nd_mrep);
2623         return (error);
2624 }
2625
2626 /*
2627  * nfs symbolic link create rpc
2628  */
2629 APPLESTATIC int
2630 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, char *target,
2631     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2632     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2633     int *dattrflagp, void *dstuff)
2634 {
2635         u_int32_t *tl;
2636         struct nfsrv_descript nfsd, *nd = &nfsd;
2637         struct nfsmount *nmp;
2638         int slen, error = 0;
2639
2640         *nfhpp = NULL;
2641         *attrflagp = 0;
2642         *dattrflagp = 0;
2643         nmp = VFSTONFS(vnode_mount(dvp));
2644         slen = strlen(target);
2645         if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2646                 return (ENAMETOOLONG);
2647         NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2648         if (nd->nd_flag & ND_NFSV4) {
2649                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2650                 *tl = txdr_unsigned(NFLNK);
2651                 (void) nfsm_strtom(nd, target, slen);
2652         }
2653         (void) nfsm_strtom(nd, name, namelen);
2654         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2655                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2656         if (!(nd->nd_flag & ND_NFSV4))
2657                 (void) nfsm_strtom(nd, target, slen);
2658         if (nd->nd_flag & ND_NFSV2)
2659                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2660         error = nfscl_request(nd, dvp, p, cred, dstuff);
2661         if (error)
2662                 return (error);
2663         if (nd->nd_flag & ND_NFSV4)
2664                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2665         if ((nd->nd_flag & ND_NFSV3) && !error) {
2666                 if (!nd->nd_repstat)
2667                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2668                 if (!error)
2669                         error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2670                             NULL, dstuff);
2671         }
2672         if (nd->nd_repstat && !error)
2673                 error = nd->nd_repstat;
2674         mbuf_freem(nd->nd_mrep);
2675         /*
2676          * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2677          * Only do this if vfs.nfs.ignore_eexist is set.
2678          * Never do this for NFSv4.1 or later minor versions, since sessions
2679          * should guarantee "exactly once" RPC semantics.
2680          */
2681         if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2682             nmp->nm_minorvers == 0))
2683                 error = 0;
2684         return (error);
2685 }
2686
2687 /*
2688  * nfs make dir rpc
2689  */
2690 APPLESTATIC int
2691 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2692     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2693     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2694     int *dattrflagp, void *dstuff)
2695 {
2696         u_int32_t *tl;
2697         struct nfsrv_descript nfsd, *nd = &nfsd;
2698         nfsattrbit_t attrbits;
2699         int error = 0;
2700         struct nfsfh *fhp;
2701         struct nfsmount *nmp;
2702
2703         *nfhpp = NULL;
2704         *attrflagp = 0;
2705         *dattrflagp = 0;
2706         nmp = VFSTONFS(vnode_mount(dvp));
2707         fhp = VTONFS(dvp)->n_fhp;
2708         if (namelen > NFS_MAXNAMLEN)
2709                 return (ENAMETOOLONG);
2710         NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2711         if (nd->nd_flag & ND_NFSV4) {
2712                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2713                 *tl = txdr_unsigned(NFDIR);
2714         }
2715         (void) nfsm_strtom(nd, name, namelen);
2716         nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2717         if (nd->nd_flag & ND_NFSV4) {
2718                 NFSGETATTR_ATTRBIT(&attrbits);
2719                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2720                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2721                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2722                 (void) nfsrv_putattrbit(nd, &attrbits);
2723                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2724                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2725                 (void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2726                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2727                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2728                 (void) nfsrv_putattrbit(nd, &attrbits);
2729         }
2730         error = nfscl_request(nd, dvp, p, cred, dstuff);
2731         if (error)
2732                 return (error);
2733         if (nd->nd_flag & ND_NFSV4)
2734                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2735         if (!nd->nd_repstat && !error) {
2736                 if (nd->nd_flag & ND_NFSV4) {
2737                         NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2738                         error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2739                 }
2740                 if (!error)
2741                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2742                 if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2743                         /* Get rid of the PutFH and Getattr status values. */
2744                         NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2745                         /* Load the directory attributes. */
2746                         error = nfsm_loadattr(nd, dnap);
2747                         if (error == 0)
2748                                 *dattrflagp = 1;
2749                 }
2750         }
2751         if ((nd->nd_flag & ND_NFSV3) && !error)
2752                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2753         if (nd->nd_repstat && !error)
2754                 error = nd->nd_repstat;
2755 nfsmout:
2756         mbuf_freem(nd->nd_mrep);
2757         /*
2758          * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2759          * Only do this if vfs.nfs.ignore_eexist is set.
2760          * Never do this for NFSv4.1 or later minor versions, since sessions
2761          * should guarantee "exactly once" RPC semantics.
2762          */
2763         if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2764             nmp->nm_minorvers == 0))
2765                 error = 0;
2766         return (error);
2767 }
2768
2769 /*
2770  * nfs remove directory call
2771  */
2772 APPLESTATIC int
2773 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2774     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2775 {
2776         struct nfsrv_descript nfsd, *nd = &nfsd;
2777         int error = 0;
2778
2779         *dattrflagp = 0;
2780         if (namelen > NFS_MAXNAMLEN)
2781                 return (ENAMETOOLONG);
2782         NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2783         (void) nfsm_strtom(nd, name, namelen);
2784         error = nfscl_request(nd, dvp, p, cred, dstuff);
2785         if (error)
2786                 return (error);
2787         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2788                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2789         if (nd->nd_repstat && !error)
2790                 error = nd->nd_repstat;
2791         mbuf_freem(nd->nd_mrep);
2792         /*
2793          * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2794          */
2795         if (error == ENOENT)
2796                 error = 0;
2797         return (error);
2798 }
2799
2800 /*
2801  * Readdir rpc.
2802  * Always returns with either uio_resid unchanged, if you are at the
2803  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2804  * filled in.
2805  * I felt this would allow caching of directory blocks more easily
2806  * than returning a pertially filled block.
2807  * Directory offset cookies:
2808  * Oh my, what to do with them...
2809  * I can think of three ways to deal with them:
2810  * 1 - have the layer above these RPCs maintain a map between logical
2811  *     directory byte offsets and the NFS directory offset cookies
2812  * 2 - pass the opaque directory offset cookies up into userland
2813  *     and let the libc functions deal with them, via the system call
2814  * 3 - return them to userland in the "struct dirent", so future versions
2815  *     of libc can use them and do whatever is necessary to make things work
2816  *     above these rpc calls, in the meantime
2817  * For now, I do #3 by "hiding" the directory offset cookies after the
2818  * d_name field in struct dirent. This is space inside d_reclen that
2819  * will be ignored by anything that doesn't know about them.
2820  * The directory offset cookies are filled in as the last 8 bytes of
2821  * each directory entry, after d_name. Someday, the userland libc
2822  * functions may be able to use these. In the meantime, it satisfies
2823  * OpenBSD's requirements for cookies being returned.
2824  * If expects the directory offset cookie for the read to be in uio_offset
2825  * and returns the one for the next entry after this directory block in
2826  * there, as well.
2827  */
2828 APPLESTATIC int
2829 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2830     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2831     int *eofp, void *stuff)
2832 {
2833         int len, left;
2834         struct dirent *dp = NULL;
2835         u_int32_t *tl;
2836         nfsquad_t cookie, ncookie;
2837         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2838         struct nfsnode *dnp = VTONFS(vp);
2839         struct nfsvattr nfsva;
2840         struct nfsrv_descript nfsd, *nd = &nfsd;
2841         int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2842         int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2843         u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2844         char *cp;
2845         nfsattrbit_t attrbits, dattrbits;
2846         u_int32_t rderr, *tl2 = NULL;
2847         size_t tresid;
2848
2849         KASSERT(uiop->uio_iovcnt == 1 &&
2850             (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2851             ("nfs readdirrpc bad uio"));
2852         ncookie.lval[0] = ncookie.lval[1] = 0;
2853         /*
2854          * There is no point in reading a lot more than uio_resid, however
2855          * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2856          * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2857          * will never make readsize > nm_readdirsize.
2858          */
2859         readsize = nmp->nm_readdirsize;
2860         if (readsize > uio_uio_resid(uiop))
2861                 readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2862
2863         *attrflagp = 0;
2864         if (eofp)
2865                 *eofp = 0;
2866         tresid = uio_uio_resid(uiop);
2867         cookie.lval[0] = cookiep->nfsuquad[0];
2868         cookie.lval[1] = cookiep->nfsuquad[1];
2869         nd->nd_mrep = NULL;
2870
2871         /*
2872          * For NFSv4, first create the "." and ".." entries.
2873          */
2874         if (NFSHASNFSV4(nmp)) {
2875                 reqsize = 6 * NFSX_UNSIGNED;
2876                 NFSGETATTR_ATTRBIT(&dattrbits);
2877                 NFSZERO_ATTRBIT(&attrbits);
2878                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2879                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2880                 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2881                     NFSATTRBIT_MOUNTEDONFILEID)) {
2882                         NFSSETBIT_ATTRBIT(&attrbits,
2883                             NFSATTRBIT_MOUNTEDONFILEID);
2884                         gotmnton = 1;
2885                 } else {
2886                         /*
2887                          * Must fake it. Use the fileno, except when the
2888                          * fsid is != to that of the directory. For that
2889                          * case, generate a fake fileno that is not the same.
2890                          */
2891                         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2892                         gotmnton = 0;
2893                 }
2894
2895                 /*
2896                  * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2897                  */
2898                 if (uiop->uio_offset == 0) {
2899                         NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2900                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2901                         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2902                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2903                         (void) nfsrv_putattrbit(nd, &attrbits);
2904                         error = nfscl_request(nd, vp, p, cred, stuff);
2905                         if (error)
2906                             return (error);
2907                         dotfileid = 0;  /* Fake out the compiler. */
2908                         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2909                             error = nfsm_loadattr(nd, &nfsva);
2910                             if (error != 0)
2911                                 goto nfsmout;
2912                             dotfileid = nfsva.na_fileid;
2913                         }
2914                         if (nd->nd_repstat == 0) {
2915                             NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2916                             len = fxdr_unsigned(int, *(tl + 4));
2917                             if (len > 0 && len <= NFSX_V4FHMAX)
2918                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2919                             else
2920                                 error = EPERM;
2921                             if (!error) {
2922                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2923                                 nfsva.na_mntonfileno = UINT64_MAX;
2924                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2925                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2926                                     NULL, NULL, NULL, p, cred);
2927                                 if (error) {
2928                                     dotdotfileid = dotfileid;
2929                                 } else if (gotmnton) {
2930                                     if (nfsva.na_mntonfileno != UINT64_MAX)
2931                                         dotdotfileid = nfsva.na_mntonfileno;
2932                                     else
2933                                         dotdotfileid = nfsva.na_fileid;
2934                                 } else if (nfsva.na_filesid[0] ==
2935                                     dnp->n_vattr.na_filesid[0] &&
2936                                     nfsva.na_filesid[1] ==
2937                                     dnp->n_vattr.na_filesid[1]) {
2938                                     dotdotfileid = nfsva.na_fileid;
2939                                 } else {
2940                                     do {
2941                                         fakefileno--;
2942                                     } while (fakefileno ==
2943                                         nfsva.na_fileid);
2944                                     dotdotfileid = fakefileno;
2945                                 }
2946                             }
2947                         } else if (nd->nd_repstat == NFSERR_NOENT) {
2948                             /*
2949                              * Lookupp returns NFSERR_NOENT when we are
2950                              * at the root, so just use the current dir.
2951                              */
2952                             nd->nd_repstat = 0;
2953                             dotdotfileid = dotfileid;
2954                         } else {
2955                             error = nd->nd_repstat;
2956                         }
2957                         mbuf_freem(nd->nd_mrep);
2958                         if (error)
2959                             return (error);
2960                         nd->nd_mrep = NULL;
2961                         dp = (struct dirent *)uio_iov_base(uiop);
2962                         dp->d_off = 0;
2963                         dp->d_type = DT_DIR;
2964                         dp->d_fileno = dotfileid;
2965                         dp->d_namlen = 1;
2966                         *((uint64_t *)dp->d_name) = 0;  /* Zero pad it. */
2967                         dp->d_name[0] = '.';
2968                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2969                         /*
2970                          * Just make these offset cookie 0.
2971                          */
2972                         tl = (u_int32_t *)&dp->d_name[8];
2973                         *tl++ = 0;
2974                         *tl = 0;
2975                         blksiz += dp->d_reclen;
2976                         uio_uio_resid_add(uiop, -(dp->d_reclen));
2977                         uiop->uio_offset += dp->d_reclen;
2978                         uio_iov_base_add(uiop, dp->d_reclen);
2979                         uio_iov_len_add(uiop, -(dp->d_reclen));
2980                         dp = (struct dirent *)uio_iov_base(uiop);
2981                         dp->d_off = 0;
2982                         dp->d_type = DT_DIR;
2983                         dp->d_fileno = dotdotfileid;
2984                         dp->d_namlen = 2;
2985                         *((uint64_t *)dp->d_name) = 0;
2986                         dp->d_name[0] = '.';
2987                         dp->d_name[1] = '.';
2988                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2989                         /*
2990                          * Just make these offset cookie 0.
2991                          */
2992                         tl = (u_int32_t *)&dp->d_name[8];
2993                         *tl++ = 0;
2994                         *tl = 0;
2995                         blksiz += dp->d_reclen;
2996                         uio_uio_resid_add(uiop, -(dp->d_reclen));
2997                         uiop->uio_offset += dp->d_reclen;
2998                         uio_iov_base_add(uiop, dp->d_reclen);
2999                         uio_iov_len_add(uiop, -(dp->d_reclen));
3000                 }
3001                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3002         } else {
3003                 reqsize = 5 * NFSX_UNSIGNED;
3004         }
3005
3006
3007         /*
3008          * Loop around doing readdir rpc's of size readsize.
3009          * The stopping criteria is EOF or buffer full.
3010          */
3011         while (more_dirs && bigenough) {
3012                 *attrflagp = 0;
3013                 NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3014                 if (nd->nd_flag & ND_NFSV2) {
3015                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3016                         *tl++ = cookie.lval[1];
3017                         *tl = txdr_unsigned(readsize);
3018                 } else {
3019                         NFSM_BUILD(tl, u_int32_t *, reqsize);
3020                         *tl++ = cookie.lval[0];
3021                         *tl++ = cookie.lval[1];
3022                         if (cookie.qval == 0) {
3023                                 *tl++ = 0;
3024                                 *tl++ = 0;
3025                         } else {
3026                                 NFSLOCKNODE(dnp);
3027                                 *tl++ = dnp->n_cookieverf.nfsuquad[0];
3028                                 *tl++ = dnp->n_cookieverf.nfsuquad[1];
3029                                 NFSUNLOCKNODE(dnp);
3030                         }
3031                         if (nd->nd_flag & ND_NFSV4) {
3032                                 *tl++ = txdr_unsigned(readsize);
3033                                 *tl = txdr_unsigned(readsize);
3034                                 (void) nfsrv_putattrbit(nd, &attrbits);
3035                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3036                                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3037                                 (void) nfsrv_putattrbit(nd, &dattrbits);
3038                         } else {
3039                                 *tl = txdr_unsigned(readsize);
3040                         }
3041                 }
3042                 error = nfscl_request(nd, vp, p, cred, stuff);
3043                 if (error)
3044                         return (error);
3045                 if (!(nd->nd_flag & ND_NFSV2)) {
3046                         if (nd->nd_flag & ND_NFSV3)
3047                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3048                                     stuff);
3049                         if (!nd->nd_repstat && !error) {
3050                                 NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3051                                 NFSLOCKNODE(dnp);
3052                                 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3053                                 dnp->n_cookieverf.nfsuquad[1] = *tl;
3054                                 NFSUNLOCKNODE(dnp);
3055                         }
3056                 }
3057                 if (nd->nd_repstat || error) {
3058                         if (!error)
3059                                 error = nd->nd_repstat;
3060                         goto nfsmout;
3061                 }
3062                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3063                 more_dirs = fxdr_unsigned(int, *tl);
3064                 if (!more_dirs)
3065                         tryformoredirs = 0;
3066
3067                 /* loop through the dir entries, doctoring them to 4bsd form */
3068                 while (more_dirs && bigenough) {
3069                         if (nd->nd_flag & ND_NFSV4) {
3070                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3071                                 ncookie.lval[0] = *tl++;
3072                                 ncookie.lval[1] = *tl++;
3073                                 len = fxdr_unsigned(int, *tl);
3074                         } else if (nd->nd_flag & ND_NFSV3) {
3075                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3076                                 nfsva.na_fileid = fxdr_hyper(tl);
3077                                 tl += 2;
3078                                 len = fxdr_unsigned(int, *tl);
3079                         } else {
3080                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3081                                 nfsva.na_fileid = fxdr_unsigned(uint64_t,
3082                                     *tl++);
3083                                 len = fxdr_unsigned(int, *tl);
3084                         }
3085                         if (len <= 0 || len > NFS_MAXNAMLEN) {
3086                                 error = EBADRPC;
3087                                 goto nfsmout;
3088                         }
3089                         tlen = roundup2(len, 8);
3090                         if (tlen == len)
3091                                 tlen += 8;  /* To ensure null termination. */
3092                         left = DIRBLKSIZ - blksiz;
3093                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3094                                 dp->d_reclen += left;
3095                                 uio_iov_base_add(uiop, left);
3096                                 uio_iov_len_add(uiop, -(left));
3097                                 uio_uio_resid_add(uiop, -(left));
3098                                 uiop->uio_offset += left;
3099                                 blksiz = 0;
3100                         }
3101                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3102                             uio_uio_resid(uiop))
3103                                 bigenough = 0;
3104                         if (bigenough) {
3105                                 dp = (struct dirent *)uio_iov_base(uiop);
3106                                 dp->d_off = 0;
3107                                 dp->d_namlen = len;
3108                                 dp->d_reclen = _GENERIC_DIRLEN(len) +
3109                                     NFSX_HYPER;
3110                                 dp->d_type = DT_UNKNOWN;
3111                                 blksiz += dp->d_reclen;
3112                                 if (blksiz == DIRBLKSIZ)
3113                                         blksiz = 0;
3114                                 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3115                                 uiop->uio_offset += DIRHDSIZ;
3116                                 uio_iov_base_add(uiop, DIRHDSIZ);
3117                                 uio_iov_len_add(uiop, -(DIRHDSIZ));
3118                                 error = nfsm_mbufuio(nd, uiop, len);
3119                                 if (error)
3120                                         goto nfsmout;
3121                                 cp = uio_iov_base(uiop);
3122                                 tlen -= len;
3123                                 *cp = '\0';     /* null terminate */
3124                                 cp += tlen;     /* points to cookie storage */
3125                                 tl2 = (u_int32_t *)cp;
3126                                 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3127                                 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3128                                 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3129                                 uiop->uio_offset += (tlen + NFSX_HYPER);
3130                         } else {
3131                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3132                                 if (error)
3133                                         goto nfsmout;
3134                         }
3135                         if (nd->nd_flag & ND_NFSV4) {
3136                                 rderr = 0;
3137                                 nfsva.na_mntonfileno = UINT64_MAX;
3138                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3139                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3140                                     NULL, NULL, &rderr, p, cred);
3141                                 if (error)
3142                                         goto nfsmout;
3143                                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3144                         } else if (nd->nd_flag & ND_NFSV3) {
3145                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3146                                 ncookie.lval[0] = *tl++;
3147                                 ncookie.lval[1] = *tl++;
3148                         } else {
3149                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3150                                 ncookie.lval[0] = 0;
3151                                 ncookie.lval[1] = *tl++;
3152                         }
3153                         if (bigenough) {
3154                             if (nd->nd_flag & ND_NFSV4) {
3155                                 if (rderr) {
3156                                     dp->d_fileno = 0;
3157                                 } else {
3158                                     if (gotmnton) {
3159                                         if (nfsva.na_mntonfileno != UINT64_MAX)
3160                                             dp->d_fileno = nfsva.na_mntonfileno;
3161                                         else
3162                                             dp->d_fileno = nfsva.na_fileid;
3163                                     } else if (nfsva.na_filesid[0] ==
3164                                         dnp->n_vattr.na_filesid[0] &&
3165                                         nfsva.na_filesid[1] ==
3166                                         dnp->n_vattr.na_filesid[1]) {
3167                                         dp->d_fileno = nfsva.na_fileid;
3168                                     } else {
3169                                         do {
3170                                             fakefileno--;
3171                                         } while (fakefileno ==
3172                                             nfsva.na_fileid);
3173                                         dp->d_fileno = fakefileno;
3174                                     }
3175                                     dp->d_type = vtonfs_dtype(nfsva.na_type);
3176                                 }
3177                             } else {
3178                                 dp->d_fileno = nfsva.na_fileid;
3179                             }
3180                             *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3181                                 ncookie.lval[0];
3182                             *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3183                                 ncookie.lval[1];
3184                         }
3185                         more_dirs = fxdr_unsigned(int, *tl);
3186                 }
3187                 /*
3188                  * If at end of rpc data, get the eof boolean
3189                  */
3190                 if (!more_dirs) {
3191                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3192                         eof = fxdr_unsigned(int, *tl);
3193                         if (tryformoredirs)
3194                                 more_dirs = !eof;
3195                         if (nd->nd_flag & ND_NFSV4) {
3196                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3197                                     stuff);
3198                                 if (error)
3199                                         goto nfsmout;
3200                         }
3201                 }
3202                 mbuf_freem(nd->nd_mrep);
3203                 nd->nd_mrep = NULL;
3204         }
3205         /*
3206          * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3207          * by increasing d_reclen for the last record.
3208          */
3209         if (blksiz > 0) {
3210                 left = DIRBLKSIZ - blksiz;
3211                 dp->d_reclen += left;
3212                 uio_iov_base_add(uiop, left);
3213                 uio_iov_len_add(uiop, -(left));
3214                 uio_uio_resid_add(uiop, -(left));
3215                 uiop->uio_offset += left;
3216         }
3217
3218         /*
3219          * If returning no data, assume end of file.
3220          * If not bigenough, return not end of file, since you aren't
3221          *    returning all the data
3222          * Otherwise, return the eof flag from the server.
3223          */
3224         if (eofp) {
3225                 if (tresid == ((size_t)(uio_uio_resid(uiop))))
3226                         *eofp = 1;
3227                 else if (!bigenough)
3228                         *eofp = 0;
3229                 else
3230                         *eofp = eof;
3231         }
3232
3233         /*
3234          * Add extra empty records to any remaining DIRBLKSIZ chunks.
3235          */
3236         while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3237                 dp = (struct dirent *)uio_iov_base(uiop);
3238                 dp->d_type = DT_UNKNOWN;
3239                 dp->d_fileno = 0;
3240                 dp->d_namlen = 0;
3241                 dp->d_name[0] = '\0';
3242                 tl = (u_int32_t *)&dp->d_name[4];
3243                 *tl++ = cookie.lval[0];
3244                 *tl = cookie.lval[1];
3245                 dp->d_reclen = DIRBLKSIZ;
3246                 uio_iov_base_add(uiop, DIRBLKSIZ);
3247                 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3248                 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3249                 uiop->uio_offset += DIRBLKSIZ;
3250         }
3251
3252 nfsmout:
3253         if (nd->nd_mrep != NULL)
3254                 mbuf_freem(nd->nd_mrep);
3255         return (error);
3256 }
3257
3258 #ifndef APPLE
3259 /*
3260  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3261  * (Also used for NFS V4 when mount flag set.)
3262  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3263  */
3264 APPLESTATIC int
3265 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3266     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3267     int *eofp, void *stuff)
3268 {
3269         int len, left;
3270         struct dirent *dp = NULL;
3271         u_int32_t *tl;
3272         vnode_t newvp = NULLVP;
3273         struct nfsrv_descript nfsd, *nd = &nfsd;
3274         struct nameidata nami, *ndp = &nami;
3275         struct componentname *cnp = &ndp->ni_cnd;
3276         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3277         struct nfsnode *dnp = VTONFS(vp), *np;
3278         struct nfsvattr nfsva;
3279         struct nfsfh *nfhp;
3280         nfsquad_t cookie, ncookie;
3281         int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3282         int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3283         int isdotdot = 0, unlocknewvp = 0;
3284         u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3285         u_int64_t fileno = 0;
3286         char *cp;
3287         nfsattrbit_t attrbits, dattrbits;
3288         size_t tresid;
3289         u_int32_t *tl2 = NULL, rderr;
3290         struct timespec dctime;
3291
3292         KASSERT(uiop->uio_iovcnt == 1 &&
3293             (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3294             ("nfs readdirplusrpc bad uio"));
3295         ncookie.lval[0] = ncookie.lval[1] = 0;
3296         timespecclear(&dctime);
3297         *attrflagp = 0;
3298         if (eofp != NULL)
3299                 *eofp = 0;
3300         ndp->ni_dvp = vp;
3301         nd->nd_mrep = NULL;
3302         cookie.lval[0] = cookiep->nfsuquad[0];
3303         cookie.lval[1] = cookiep->nfsuquad[1];
3304         tresid = uio_uio_resid(uiop);
3305
3306         /*
3307          * For NFSv4, first create the "." and ".." entries.
3308          */
3309         if (NFSHASNFSV4(nmp)) {
3310                 NFSGETATTR_ATTRBIT(&dattrbits);
3311                 NFSZERO_ATTRBIT(&attrbits);
3312                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3313                 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3314                     NFSATTRBIT_MOUNTEDONFILEID)) {
3315                         NFSSETBIT_ATTRBIT(&attrbits,
3316                             NFSATTRBIT_MOUNTEDONFILEID);
3317                         gotmnton = 1;
3318                 } else {
3319                         /*
3320                          * Must fake it. Use the fileno, except when the
3321                          * fsid is != to that of the directory. For that
3322                          * case, generate a fake fileno that is not the same.
3323                          */
3324                         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3325                         gotmnton = 0;
3326                 }
3327
3328                 /*
3329                  * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3330                  */
3331                 if (uiop->uio_offset == 0) {
3332                         NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3333                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3334                         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
3335                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
3336                         (void) nfsrv_putattrbit(nd, &attrbits);
3337                         error = nfscl_request(nd, vp, p, cred, stuff);
3338                         if (error)
3339                             return (error);
3340                         dotfileid = 0;  /* Fake out the compiler. */
3341                         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3342                             error = nfsm_loadattr(nd, &nfsva);
3343                             if (error != 0)
3344                                 goto nfsmout;
3345                             dctime = nfsva.na_ctime;
3346                             dotfileid = nfsva.na_fileid;
3347                         }
3348                         if (nd->nd_repstat == 0) {
3349                             NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3350                             len = fxdr_unsigned(int, *(tl + 4));
3351                             if (len > 0 && len <= NFSX_V4FHMAX)
3352                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3353                             else
3354                                 error = EPERM;
3355                             if (!error) {
3356                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3357                                 nfsva.na_mntonfileno = UINT64_MAX;
3358                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3359                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3360                                     NULL, NULL, NULL, p, cred);
3361                                 if (error) {
3362                                     dotdotfileid = dotfileid;
3363                                 } else if (gotmnton) {
3364                                     if (nfsva.na_mntonfileno != UINT64_MAX)
3365                                         dotdotfileid = nfsva.na_mntonfileno;
3366                                     else
3367                                         dotdotfileid = nfsva.na_fileid;
3368                                 } else if (nfsva.na_filesid[0] ==
3369                                     dnp->n_vattr.na_filesid[0] &&
3370                                     nfsva.na_filesid[1] ==
3371                                     dnp->n_vattr.na_filesid[1]) {
3372                                     dotdotfileid = nfsva.na_fileid;
3373                                 } else {
3374                                     do {
3375                                         fakefileno--;
3376                                     } while (fakefileno ==
3377                                         nfsva.na_fileid);
3378                                     dotdotfileid = fakefileno;
3379                                 }
3380                             }
3381                         } else if (nd->nd_repstat == NFSERR_NOENT) {
3382                             /*
3383                              * Lookupp returns NFSERR_NOENT when we are
3384                              * at the root, so just use the current dir.
3385                              */
3386                             nd->nd_repstat = 0;
3387                             dotdotfileid = dotfileid;
3388                         } else {
3389                             error = nd->nd_repstat;
3390                         }
3391                         mbuf_freem(nd->nd_mrep);
3392                         if (error)
3393                             return (error);
3394                         nd->nd_mrep = NULL;
3395                         dp = (struct dirent *)uio_iov_base(uiop);
3396                         dp->d_off = 0;
3397                         dp->d_type = DT_DIR;
3398                         dp->d_fileno = dotfileid;
3399                         dp->d_namlen = 1;
3400                         *((uint64_t *)dp->d_name) = 0;  /* Zero pad it. */
3401                         dp->d_name[0] = '.';
3402                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3403                         /*
3404                          * Just make these offset cookie 0.
3405                          */
3406                         tl = (u_int32_t *)&dp->d_name[8];
3407                         *tl++ = 0;
3408                         *tl = 0;
3409                         blksiz += dp->d_reclen;
3410                         uio_uio_resid_add(uiop, -(dp->d_reclen));
3411                         uiop->uio_offset += dp->d_reclen;
3412                         uio_iov_base_add(uiop, dp->d_reclen);
3413                         uio_iov_len_add(uiop, -(dp->d_reclen));
3414                         dp = (struct dirent *)uio_iov_base(uiop);
3415                         dp->d_off = 0;
3416                         dp->d_type = DT_DIR;
3417                         dp->d_fileno = dotdotfileid;
3418                         dp->d_namlen = 2;
3419                         *((uint64_t *)dp->d_name) = 0;
3420                         dp->d_name[0] = '.';
3421                         dp->d_name[1] = '.';
3422                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3423                         /*
3424                          * Just make these offset cookie 0.
3425                          */
3426                         tl = (u_int32_t *)&dp->d_name[8];
3427                         *tl++ = 0;
3428                         *tl = 0;
3429                         blksiz += dp->d_reclen;
3430                         uio_uio_resid_add(uiop, -(dp->d_reclen));
3431                         uiop->uio_offset += dp->d_reclen;
3432                         uio_iov_base_add(uiop, dp->d_reclen);
3433                         uio_iov_len_add(uiop, -(dp->d_reclen));
3434                 }
3435                 NFSREADDIRPLUS_ATTRBIT(&attrbits);
3436                 if (gotmnton)
3437                         NFSSETBIT_ATTRBIT(&attrbits,
3438                             NFSATTRBIT_MOUNTEDONFILEID);
3439         }
3440
3441         /*
3442          * Loop around doing readdir rpc's of size nm_readdirsize.
3443          * The stopping criteria is EOF or buffer full.
3444          */
3445         while (more_dirs && bigenough) {
3446                 *attrflagp = 0;
3447                 NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3448                 NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3449                 *tl++ = cookie.lval[0];
3450                 *tl++ = cookie.lval[1];
3451                 if (cookie.qval == 0) {
3452                         *tl++ = 0;
3453                         *tl++ = 0;
3454                 } else {
3455                         NFSLOCKNODE(dnp);
3456                         *tl++ = dnp->n_cookieverf.nfsuquad[0];
3457                         *tl++ = dnp->n_cookieverf.nfsuquad[1];
3458                         NFSUNLOCKNODE(dnp);
3459                 }
3460                 *tl++ = txdr_unsigned(nmp->nm_readdirsize);
3461                 *tl = txdr_unsigned(nmp->nm_readdirsize);
3462                 if (nd->nd_flag & ND_NFSV4) {
3463                         (void) nfsrv_putattrbit(nd, &attrbits);
3464                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3465                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
3466                         (void) nfsrv_putattrbit(nd, &dattrbits);
3467                 }
3468                 error = nfscl_request(nd, vp, p, cred, stuff);
3469                 if (error)
3470                         return (error);
3471                 if (nd->nd_flag & ND_NFSV3)
3472                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3473                 if (nd->nd_repstat || error) {
3474                         if (!error)
3475                                 error = nd->nd_repstat;
3476                         goto nfsmout;
3477                 }
3478                 if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3479                         dctime = nap->na_ctime;
3480                 NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3481                 NFSLOCKNODE(dnp);
3482                 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3483                 dnp->n_cookieverf.nfsuquad[1] = *tl++;
3484                 NFSUNLOCKNODE(dnp);
3485                 more_dirs = fxdr_unsigned(int, *tl);
3486                 if (!more_dirs)
3487                         tryformoredirs = 0;
3488
3489                 /* loop through the dir entries, doctoring them to 4bsd form */
3490                 while (more_dirs && bigenough) {
3491                         NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3492                         if (nd->nd_flag & ND_NFSV4) {
3493                                 ncookie.lval[0] = *tl++;
3494                                 ncookie.lval[1] = *tl++;
3495                         } else {
3496                                 fileno = fxdr_hyper(tl);
3497                                 tl += 2;
3498                         }
3499                         len = fxdr_unsigned(int, *tl);
3500                         if (len <= 0 || len > NFS_MAXNAMLEN) {
3501                                 error = EBADRPC;
3502                                 goto nfsmout;
3503                         }
3504                         tlen = roundup2(len, 8);
3505                         if (tlen == len)
3506                                 tlen += 8;  /* To ensure null termination. */
3507                         left = DIRBLKSIZ - blksiz;
3508                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3509                                 dp->d_reclen += left;
3510                                 uio_iov_base_add(uiop, left);
3511                                 uio_iov_len_add(uiop, -(left));
3512                                 uio_uio_resid_add(uiop, -(left));
3513                                 uiop->uio_offset += left;
3514                                 blksiz = 0;
3515                         }
3516                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3517                             uio_uio_resid(uiop))
3518                                 bigenough = 0;
3519                         if (bigenough) {
3520                                 dp = (struct dirent *)uio_iov_base(uiop);
3521                                 dp->d_off = 0;
3522                                 dp->d_namlen = len;
3523                                 dp->d_reclen = _GENERIC_DIRLEN(len) +
3524                                     NFSX_HYPER;
3525                                 dp->d_type = DT_UNKNOWN;
3526                                 blksiz += dp->d_reclen;
3527                                 if (blksiz == DIRBLKSIZ)
3528                                         blksiz = 0;
3529                                 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3530                                 uiop->uio_offset += DIRHDSIZ;
3531                                 uio_iov_base_add(uiop, DIRHDSIZ);
3532                                 uio_iov_len_add(uiop, -(DIRHDSIZ));
3533                                 cnp->cn_nameptr = uio_iov_base(uiop);
3534                                 cnp->cn_namelen = len;
3535                                 NFSCNHASHZERO(cnp);
3536                                 error = nfsm_mbufuio(nd, uiop, len);
3537                                 if (error)
3538                                         goto nfsmout;
3539                                 cp = uio_iov_base(uiop);
3540                                 tlen -= len;
3541                                 *cp = '\0';
3542                                 cp += tlen;     /* points to cookie storage */
3543                                 tl2 = (u_int32_t *)cp;
3544                                 if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3545                                     cnp->cn_nameptr[1] == '.')
3546                                         isdotdot = 1;
3547                                 else
3548                                         isdotdot = 0;
3549                                 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3550                                 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3551                                 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3552                                 uiop->uio_offset += (tlen + NFSX_HYPER);
3553                         } else {
3554                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3555                                 if (error)
3556                                         goto nfsmout;
3557                         }
3558                         nfhp = NULL;
3559                         if (nd->nd_flag & ND_NFSV3) {
3560                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3561                                 ncookie.lval[0] = *tl++;
3562                                 ncookie.lval[1] = *tl++;
3563                                 attrflag = fxdr_unsigned(int, *tl);
3564                                 if (attrflag) {
3565                                   error = nfsm_loadattr(nd, &nfsva);
3566                                   if (error)
3567                                         goto nfsmout;
3568                                 }
3569                                 NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3570                                 if (*tl) {
3571                                         error = nfsm_getfh(nd, &nfhp);
3572                                         if (error)
3573                                             goto nfsmout;
3574                                 }
3575                                 if (!attrflag && nfhp != NULL) {
3576                                         free(nfhp, M_NFSFH);
3577                                         nfhp = NULL;
3578                                 }
3579                         } else {
3580                                 rderr = 0;
3581                                 nfsva.na_mntonfileno = 0xffffffff;
3582                                 error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3583                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3584                                     NULL, NULL, &rderr, p, cred);
3585                                 if (error)
3586                                         goto nfsmout;
3587                         }
3588
3589                         if (bigenough) {
3590                             if (nd->nd_flag & ND_NFSV4) {
3591                                 if (rderr) {
3592                                     dp->d_fileno = 0;
3593                                 } else if (gotmnton) {
3594                                     if (nfsva.na_mntonfileno != 0xffffffff)
3595                                         dp->d_fileno = nfsva.na_mntonfileno;
3596                                     else
3597                                         dp->d_fileno = nfsva.na_fileid;
3598                                 } else if (nfsva.na_filesid[0] ==
3599                                     dnp->n_vattr.na_filesid[0] &&
3600                                     nfsva.na_filesid[1] ==
3601                                     dnp->n_vattr.na_filesid[1]) {
3602                                     dp->d_fileno = nfsva.na_fileid;
3603                                 } else {
3604                                     do {
3605                                         fakefileno--;
3606                                     } while (fakefileno ==
3607                                         nfsva.na_fileid);
3608                                     dp->d_fileno = fakefileno;
3609                                 }
3610                             } else {
3611                                 dp->d_fileno = fileno;
3612                             }
3613                             *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3614                                 ncookie.lval[0];
3615                             *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3616                                 ncookie.lval[1];
3617
3618                             if (nfhp != NULL) {
3619                                 if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3620                                     dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3621                                     VREF(vp);
3622                                     newvp = vp;
3623                                     unlocknewvp = 0;
3624                                     free(nfhp, M_NFSFH);
3625                                     np = dnp;
3626                                 } else if (isdotdot != 0) {
3627                                     /*
3628                                      * Skip doing a nfscl_nget() call for "..".
3629                                      * There's a race between acquiring the nfs
3630                                      * node here and lookups that look for the
3631                                      * directory being read (in the parent).
3632                                      * It would try to get a lock on ".." here,
3633                                      * owning the lock on the directory being
3634                                      * read. Lookup will hold the lock on ".."
3635                                      * and try to acquire the lock on the
3636                                      * directory being read.
3637                                      * If the directory is unlocked/relocked,
3638                                      * then there is a LOR with the buflock
3639                                      * vp is relocked.
3640                                      */
3641                                     free(nfhp, M_NFSFH);
3642                                 } else {
3643                                     error = nfscl_nget(vnode_mount(vp), vp,
3644                                       nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3645                                     if (!error) {
3646                                         newvp = NFSTOV(np);
3647                                         unlocknewvp = 1;
3648                                     }
3649                                 }
3650                                 nfhp = NULL;
3651                                 if (newvp != NULLVP) {
3652                                     error = nfscl_loadattrcache(&newvp,
3653                                         &nfsva, NULL, NULL, 0, 0);
3654                                     if (error) {
3655                                         if (unlocknewvp)
3656                                             vput(newvp);
3657                                         else
3658                                             vrele(newvp);
3659                                         goto nfsmout;
3660                                     }
3661                                     dp->d_type =
3662                                         vtonfs_dtype(np->n_vattr.na_type);
3663                                     ndp->ni_vp = newvp;
3664                                     NFSCNHASH(cnp, HASHINIT);
3665                                     if (cnp->cn_namelen <= NCHNAMLEN &&
3666                                         (newvp->v_type != VDIR ||
3667                                          dctime.tv_sec != 0)) {
3668                                         cache_enter_time(ndp->ni_dvp,
3669                                             ndp->ni_vp, cnp,
3670                                             &nfsva.na_ctime,
3671                                             newvp->v_type != VDIR ? NULL :
3672                                             &dctime);
3673                                     }
3674                                     if (unlocknewvp)
3675                                         vput(newvp);
3676                                     else
3677                                         vrele(newvp);
3678                                     newvp = NULLVP;
3679                                 }
3680                             }
3681                         } else if (nfhp != NULL) {
3682                             free(nfhp, M_NFSFH);
3683                         }
3684                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3685                         more_dirs = fxdr_unsigned(int, *tl);
3686                 }
3687                 /*
3688                  * If at end of rpc data, get the eof boolean
3689                  */
3690                 if (!more_dirs) {
3691                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3692                         eof = fxdr_unsigned(int, *tl);
3693                         if (tryformoredirs)
3694                                 more_dirs = !eof;
3695                         if (nd->nd_flag & ND_NFSV4) {
3696                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3697                                     stuff);
3698                                 if (error)
3699                                         goto nfsmout;
3700                         }
3701                 }
3702                 mbuf_freem(nd->nd_mrep);
3703                 nd->nd_mrep = NULL;
3704         }
3705         /*
3706          * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3707          * by increasing d_reclen for the last record.
3708          */
3709         if (blksiz > 0) {
3710                 left = DIRBLKSIZ - blksiz;
3711                 dp->d_reclen += left;
3712                 uio_iov_base_add(uiop, left);
3713                 uio_iov_len_add(uiop, -(left));
3714                 uio_uio_resid_add(uiop, -(left));
3715                 uiop->uio_offset += left;
3716         }
3717
3718         /*
3719          * If returning no data, assume end of file.
3720          * If not bigenough, return not end of file, since you aren't
3721          *    returning all the data
3722          * Otherwise, return the eof flag from the server.
3723          */
3724         if (eofp != NULL) {
3725                 if (tresid == uio_uio_resid(uiop))
3726                         *eofp = 1;
3727                 else if (!bigenough)
3728                         *eofp = 0;
3729                 else
3730                         *eofp = eof;
3731         }
3732
3733         /*
3734          * Add extra empty records to any remaining DIRBLKSIZ chunks.
3735          */
3736         while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3737                 dp = (struct dirent *)uio_iov_base(uiop);
3738                 dp->d_type = DT_UNKNOWN;
3739                 dp->d_fileno = 0;
3740                 dp->d_namlen = 0;
3741                 dp->d_name[0] = '\0';
3742                 tl = (u_int32_t *)&dp->d_name[4];
3743                 *tl++ = cookie.lval[0];
3744                 *tl = cookie.lval[1];
3745                 dp->d_reclen = DIRBLKSIZ;
3746                 uio_iov_base_add(uiop, DIRBLKSIZ);
3747                 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3748                 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3749                 uiop->uio_offset += DIRBLKSIZ;
3750         }
3751
3752 nfsmout:
3753         if (nd->nd_mrep != NULL)
3754                 mbuf_freem(nd->nd_mrep);
3755         return (error);
3756 }
3757 #endif  /* !APPLE */
3758
3759 /*
3760  * Nfs commit rpc
3761  */
3762 APPLESTATIC int
3763 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3764     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3765 {
3766         u_int32_t *tl;
3767         struct nfsrv_descript nfsd, *nd = &nfsd;
3768         nfsattrbit_t attrbits;
3769         int error;
3770         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3771
3772         *attrflagp = 0;
3773         NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3774         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3775         txdr_hyper(offset, tl);
3776         tl += 2;
3777         *tl = txdr_unsigned(cnt);
3778         if (nd->nd_flag & ND_NFSV4) {
3779                 /*
3780                  * And do a Getattr op.
3781                  */
3782                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3783                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3784                 NFSGETATTR_ATTRBIT(&attrbits);
3785                 (void) nfsrv_putattrbit(nd, &attrbits);
3786         }
3787         error = nfscl_request(nd, vp, p, cred, stuff);
3788         if (error)
3789                 return (error);
3790         error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3791         if (!error && !nd->nd_repstat) {
3792                 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3793                 NFSLOCKMNT(nmp);
3794                 if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3795                         NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3796                         nd->nd_repstat = NFSERR_STALEWRITEVERF;
3797                 }
3798                 NFSUNLOCKMNT(nmp);
3799                 if (nd->nd_flag & ND_NFSV4)
3800                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3801         }
3802 nfsmout:
3803         if (!error && nd->nd_repstat)
3804                 error = nd->nd_repstat;
3805         mbuf_freem(nd->nd_mrep);
3806         return (error);
3807 }
3808
3809 /*
3810  * NFS byte range lock rpc.
3811  * (Mostly just calls one of the three lower level RPC routines.)
3812  */
3813 APPLESTATIC int
3814 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3815     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3816 {
3817         struct nfscllockowner *lp;
3818         struct nfsclclient *clp;
3819         struct nfsfh *nfhp;
3820         struct nfsrv_descript nfsd, *nd = &nfsd;
3821         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3822         u_int64_t off, len;
3823         off_t start, end;
3824         u_int32_t clidrev = 0;
3825         int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3826         int callcnt, dorpc;
3827
3828         /*
3829          * Convert the flock structure into a start and end and do POSIX
3830          * bounds checking.
3831          */
3832         switch (fl->l_whence) {
3833         case SEEK_SET:
3834         case SEEK_CUR:
3835                 /*
3836                  * Caller is responsible for adding any necessary offset
3837                  * when SEEK_CUR is used.
3838                  */
3839                 start = fl->l_start;
3840                 off = fl->l_start;
3841                 break;
3842         case SEEK_END:
3843                 start = size + fl->l_start;
3844                 off = size + fl->l_start;
3845                 break;
3846         default:
3847                 return (EINVAL);
3848         }
3849         if (start < 0)
3850                 return (EINVAL);
3851         if (fl->l_len != 0) {
3852                 end = start + fl->l_len - 1;
3853                 if (end < start)
3854                         return (EINVAL);
3855         }
3856
3857         len = fl->l_len;
3858         if (len == 0)
3859                 len = NFS64BITSSET;
3860         retrycnt = 0;
3861         do {
3862             nd->nd_repstat = 0;
3863             if (op == F_GETLK) {
3864                 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3865                 if (error)
3866                         return (error);
3867                 error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3868                 if (!error) {
3869                         clidrev = clp->nfsc_clientidrev;
3870                         error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3871                             p, id, flags);
3872                 } else if (error == -1) {
3873                         error = 0;
3874                 }
3875                 nfscl_clientrelease(clp);
3876             } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3877                 /*
3878                  * We must loop around for all lockowner cases.
3879                  */
3880                 callcnt = 0;
3881                 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3882                 if (error)
3883                         return (error);
3884                 do {
3885                     error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3886                         clp, id, flags, &lp, &dorpc);
3887                     /*
3888                      * If it returns a NULL lp, we're done.
3889                      */
3890                     if (lp == NULL) {
3891                         if (callcnt == 0)
3892                             nfscl_clientrelease(clp);
3893                         else
3894                             nfscl_releasealllocks(clp, vp, p, id, flags);
3895                         return (error);
3896                     }
3897                     if (nmp->nm_clp != NULL)
3898                         clidrev = nmp->nm_clp->nfsc_clientidrev;
3899                     else
3900                         clidrev = 0;
3901                     /*
3902                      * If the server doesn't support Posix lock semantics,
3903                      * only allow locks on the entire file, since it won't
3904                      * handle overlapping byte ranges.
3905                      * There might still be a problem when a lock
3906                      * upgrade/downgrade (read<->write) occurs, since the
3907                      * server "might" expect an unlock first?
3908                      */
3909                     if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3910                         (off == 0 && len == NFS64BITSSET))) {
3911                         /*
3912                          * Since the lock records will go away, we must
3913                          * wait for grace and delay here.
3914                          */
3915                         do {
3916                             error = nfsrpc_locku(nd, nmp, lp, off, len,
3917                                 NFSV4LOCKT_READ, cred, p, 0);
3918                             if ((nd->nd_repstat == NFSERR_GRACE ||
3919                                  nd->nd_repstat == NFSERR_DELAY) &&
3920                                 error == 0)
3921                                 (void) nfs_catnap(PZERO, (int)nd->nd_repstat,
3922                                     "nfs_advlock");
3923                         } while ((nd->nd_repstat == NFSERR_GRACE ||
3924                             nd->nd_repstat == NFSERR_DELAY) && error == 0);
3925                     }
3926                     callcnt++;
3927                 } while (error == 0 && nd->nd_repstat == 0);
3928                 nfscl_releasealllocks(clp, vp, p, id, flags);
3929             } else if (op == F_SETLK) {
3930                 error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
3931                     NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
3932                 if (error || donelocally) {
3933                         return (error);
3934                 }
3935                 if (nmp->nm_clp != NULL)
3936                         clidrev = nmp->nm_clp->nfsc_clientidrev;
3937                 else
3938                         clidrev = 0;
3939                 nfhp = VTONFS(vp)->n_fhp;
3940                 if (!lp->nfsl_open->nfso_posixlock &&
3941                     (off != 0 || len != NFS64BITSSET)) {
3942                         error = EINVAL;
3943                 } else {
3944                         error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
3945                             nfhp->nfh_len, lp, newone, reclaim, off,
3946                             len, fl->l_type, cred, p, 0);
3947                 }
3948                 if (!error)
3949                         error = nd->nd_repstat;
3950                 nfscl_lockrelease(lp, error, newone);
3951             } else {
3952                 error = EINVAL;
3953             }
3954             if (!error)
3955                 error = nd->nd_repstat;
3956             if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
3957                 error == NFSERR_STALEDONTRECOVER ||
3958                 error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3959                 error == NFSERR_BADSESSION) {
3960                 (void) nfs_catnap(PZERO, error, "nfs_advlock");
3961             } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
3962                 && clidrev != 0) {
3963                 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
3964                 retrycnt++;
3965             }
3966         } while (error == NFSERR_GRACE ||
3967             error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3968             error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
3969             error == NFSERR_BADSESSION ||
3970             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
3971              expireret == 0 && clidrev != 0 && retrycnt < 4));
3972         if (error && retrycnt >= 4)
3973                 error = EIO;
3974         return (error);
3975 }
3976
3977 /*
3978  * The lower level routine for the LockT case.
3979  */
3980 APPLESTATIC int
3981 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
3982     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
3983     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3984 {
3985         u_int32_t *tl;
3986         int error, type, size;
3987         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3988         struct nfsnode *np;
3989         struct nfsmount *nmp;
3990         struct nfsclsession *tsep;
3991
3992         nmp = VFSTONFS(vp->v_mount);
3993         NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
3994         NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3995         if (fl->l_type == F_RDLCK)
3996                 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3997         else
3998                 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3999         txdr_hyper(off, tl);
4000         tl += 2;
4001         txdr_hyper(len, tl);
4002         tl += 2;
4003         tsep = nfsmnt_mdssession(nmp);
4004         *tl++ = tsep->nfsess_clientid.lval[0];
4005         *tl = tsep->nfsess_clientid.lval[1];
4006         nfscl_filllockowner(id, own, flags);
4007         np = VTONFS(vp);
4008         NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4009             np->n_fhp->nfh_len);
4010         (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4011         error = nfscl_request(nd, vp, p, cred, NULL);
4012         if (error)
4013                 return (error);
4014         if (nd->nd_repstat == 0) {
4015                 fl->l_type = F_UNLCK;
4016         } else if (nd->nd_repstat == NFSERR_DENIED) {
4017                 nd->nd_repstat = 0;
4018                 fl->l_whence = SEEK_SET;
4019                 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4020                 fl->l_start = fxdr_hyper(tl);
4021                 tl += 2;
4022                 len = fxdr_hyper(tl);
4023                 tl += 2;
4024                 if (len == NFS64BITSSET)
4025                         fl->l_len = 0;
4026                 else
4027                         fl->l_len = len;
4028                 type = fxdr_unsigned(int, *tl++);
4029                 if (type == NFSV4LOCKT_WRITE)
4030                         fl->l_type = F_WRLCK;
4031                 else
4032                         fl->l_type = F_RDLCK;
4033                 /*
4034                  * XXX For now, I have no idea what to do with the
4035                  * conflicting lock_owner, so I'll just set the pid == 0
4036                  * and skip over the lock_owner.
4037                  */
4038                 fl->l_pid = (pid_t)0;
4039                 tl += 2;
4040                 size = fxdr_unsigned(int, *tl);
4041                 if (size < 0 || size > NFSV4_OPAQUELIMIT)
4042                         error = EBADRPC;
4043                 if (!error)
4044                         error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4045         } else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4046                 nfscl_initiate_recovery(clp);
4047 nfsmout:
4048         mbuf_freem(nd->nd_mrep);
4049         return (error);
4050 }
4051
4052 /*
4053  * Lower level function that performs the LockU RPC.
4054  */
4055 static int
4056 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4057     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4058     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4059 {
4060         u_int32_t *tl;
4061         int error;
4062
4063         nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4064             lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4065         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4066         *tl++ = txdr_unsigned(type);
4067         *tl = txdr_unsigned(lp->nfsl_seqid);
4068         if (nfstest_outofseq &&
4069             (arc4random() % nfstest_outofseq) == 0)
4070                 *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4071         tl++;
4072         if (NFSHASNFSV4N(nmp))
4073                 *tl++ = 0;
4074         else
4075                 *tl++ = lp->nfsl_stateid.seqid;
4076         *tl++ = lp->nfsl_stateid.other[0];
4077         *tl++ = lp->nfsl_stateid.other[1];
4078         *tl++ = lp->nfsl_stateid.other[2];
4079         txdr_hyper(off, tl);
4080         tl += 2;
4081         txdr_hyper(len, tl);
4082         if (syscred)
4083                 nd->nd_flag |= ND_USEGSSNAME;
4084         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4085             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4086         NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4087         if (error)
4088                 return (error);
4089         if (nd->nd_repstat == 0) {
4090                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4091                 lp->nfsl_stateid.seqid = *tl++;
4092                 lp->nfsl_stateid.other[0] = *tl++;
4093                 lp->nfsl_stateid.other[1] = *tl++;
4094                 lp->nfsl_stateid.other[2] = *tl;
4095         } else if (nd->nd_repstat == NFSERR_STALESTATEID)
4096                 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4097 nfsmout:
4098         mbuf_freem(nd->nd_mrep);
4099         return (error);
4100 }
4101
4102 /*
4103  * The actual Lock RPC.
4104  */
4105 APPLESTATIC int
4106 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4107     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4108     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4109     NFSPROC_T *p, int syscred)
4110 {
4111         u_int32_t *tl;
4112         int error, size;
4113         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4114         struct nfsclsession *tsep;
4115
4116         nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4117         NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4118         if (type == F_RDLCK)
4119                 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4120         else
4121                 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4122         *tl++ = txdr_unsigned(reclaim);
4123         txdr_hyper(off, tl);
4124         tl += 2;
4125         txdr_hyper(len, tl);
4126         tl += 2;
4127         if (newone) {
4128             *tl = newnfs_true;
4129             NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4130                 2 * NFSX_UNSIGNED + NFSX_HYPER);
4131             *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4132             if (NFSHASNFSV4N(nmp))
4133                 *tl++ = 0;
4134             else
4135                 *tl++ = lp->nfsl_open->nfso_stateid.seqid;
4136             *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4137             *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4138             *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4139             *tl++ = txdr_unsigned(lp->nfsl_seqid);
4140             tsep = nfsmnt_mdssession(nmp);
4141             *tl++ = tsep->nfsess_clientid.lval[0];
4142             *tl = tsep->nfsess_clientid.lval[1];
4143             NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4144             NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4145             (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4146         } else {
4147             *tl = newnfs_false;
4148             NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4149             if (NFSHASNFSV4N(nmp))
4150                 *tl++ = 0;
4151             else
4152                 *tl++ = lp->nfsl_stateid.seqid;
4153             *tl++ = lp->nfsl_stateid.other[0];
4154             *tl++ = lp->nfsl_stateid.other[1];
4155             *tl++ = lp->nfsl_stateid.other[2];
4156             *tl = txdr_unsigned(lp->nfsl_seqid);
4157             if (nfstest_outofseq &&
4158                 (arc4random() % nfstest_outofseq) == 0)
4159                     *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4160         }
4161         if (syscred)
4162                 nd->nd_flag |= ND_USEGSSNAME;
4163         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4164             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4165         if (error)
4166                 return (error);
4167         if (newone)
4168             NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4169         NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4170         if (nd->nd_repstat == 0) {
4171                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4172                 lp->nfsl_stateid.seqid = *tl++;
4173                 lp->nfsl_stateid.other[0] = *tl++;
4174                 lp->nfsl_stateid.other[1] = *tl++;
4175                 lp->nfsl_stateid.other[2] = *tl;
4176         } else if (nd->nd_repstat == NFSERR_DENIED) {
4177                 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4178                 size = fxdr_unsigned(int, *(tl + 7));
4179                 if (size < 0 || size > NFSV4_OPAQUELIMIT)
4180                         error = EBADRPC;
4181                 if (!error)
4182                         error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4183         } else if (nd->nd_repstat == NFSERR_STALESTATEID)
4184                 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4185 nfsmout:
4186         mbuf_freem(nd->nd_mrep);
4187         return (error);
4188 }
4189
4190 /*
4191  * nfs statfs rpc
4192  * (always called with the vp for the mount point)
4193  */
4194 APPLESTATIC int
4195 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4196     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4197     void *stuff)
4198 {
4199         u_int32_t *tl = NULL;
4200         struct nfsrv_descript nfsd, *nd = &nfsd;
4201         struct nfsmount *nmp;
4202         nfsattrbit_t attrbits;
4203         int error;
4204
4205         *attrflagp = 0;
4206         nmp = VFSTONFS(vnode_mount(vp));
4207         if (NFSHASNFSV4(nmp)) {
4208                 /*
4209                  * For V4, you actually do a getattr.
4210                  */
4211                 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4212                 NFSSTATFS_GETATTRBIT(&attrbits);
4213                 (void) nfsrv_putattrbit(nd, &attrbits);
4214                 nd->nd_flag |= ND_USEGSSNAME;
4215                 error = nfscl_request(nd, vp, p, cred, stuff);
4216                 if (error)
4217                         return (error);
4218                 if (nd->nd_repstat == 0) {
4219                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4220                             NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4221                             cred);
4222                         if (!error) {
4223                                 nmp->nm_fsid[0] = nap->na_filesid[0];
4224                                 nmp->nm_fsid[1] = nap->na_filesid[1];
4225                                 NFSSETHASSETFSID(nmp);
4226                                 *attrflagp = 1;
4227                         }
4228                 } else {
4229                         error = nd->nd_repstat;
4230                 }
4231                 if (error)
4232                         goto nfsmout;
4233         } else {
4234                 NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4235                 error = nfscl_request(nd, vp, p, cred, stuff);
4236                 if (error)
4237                         return (error);
4238                 if (nd->nd_flag & ND_NFSV3) {
4239                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4240                         if (error)
4241                                 goto nfsmout;
4242                 }
4243                 if (nd->nd_repstat) {
4244                         error = nd->nd_repstat;
4245                         goto nfsmout;
4246                 }
4247                 NFSM_DISSECT(tl, u_int32_t *,
4248                     NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4249         }
4250         if (NFSHASNFSV3(nmp)) {
4251                 sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4252                 sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4253                 sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4254                 sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4255                 sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4256                 sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4257                 sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4258         } else if (NFSHASNFSV4(nmp) == 0) {
4259                 sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4260                 sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4261                 sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4262                 sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4263                 sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4264         }
4265 nfsmout:
4266         mbuf_freem(nd->nd_mrep);
4267         return (error);
4268 }
4269
4270 /*
4271  * nfs pathconf rpc
4272  */
4273 APPLESTATIC int
4274 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4275     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4276     void *stuff)
4277 {
4278         struct nfsrv_descript nfsd, *nd = &nfsd;
4279         struct nfsmount *nmp;
4280         u_int32_t *tl;
4281         nfsattrbit_t attrbits;
4282         int error;
4283
4284         *attrflagp = 0;
4285         nmp = VFSTONFS(vnode_mount(vp));
4286         if (NFSHASNFSV4(nmp)) {
4287                 /*
4288                  * For V4, you actually do a getattr.
4289                  */
4290                 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4291                 NFSPATHCONF_GETATTRBIT(&attrbits);
4292                 (void) nfsrv_putattrbit(nd, &attrbits);
4293                 nd->nd_flag |= ND_USEGSSNAME;
4294                 error = nfscl_request(nd, vp, p, cred, stuff);
4295                 if (error)
4296                         return (error);
4297                 if (nd->nd_repstat == 0) {
4298                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4299                             pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4300                             cred);
4301                         if (!error)
4302                                 *attrflagp = 1;
4303                 } else {
4304                         error = nd->nd_repstat;
4305                 }
4306         } else {
4307                 NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4308                 error = nfscl_request(nd, vp, p, cred, stuff);
4309                 if (error)
4310                         return (error);
4311                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4312                 if (nd->nd_repstat && !error)
4313                         error = nd->nd_repstat;
4314                 if (!error) {
4315                         NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4316                         pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4317                         pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4318                         pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4319                         pc->pc_chownrestricted =
4320                             fxdr_unsigned(u_int32_t, *tl++);
4321                         pc->pc_caseinsensitive =
4322                             fxdr_unsigned(u_int32_t, *tl++);
4323                         pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4324                 }
4325         }
4326 nfsmout:
4327         mbuf_freem(nd->nd_mrep);
4328         return (error);
4329 }
4330
4331 /*
4332  * nfs version 3 fsinfo rpc call
4333  */
4334 APPLESTATIC int
4335 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4336     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4337 {
4338         u_int32_t *tl;
4339         struct nfsrv_descript nfsd, *nd = &nfsd;
4340         int error;
4341
4342         *attrflagp = 0;
4343         NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4344         error = nfscl_request(nd, vp, p, cred, stuff);
4345         if (error)
4346                 return (error);
4347         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4348         if (nd->nd_repstat && !error)
4349                 error = nd->nd_repstat;
4350         if (!error) {
4351                 NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4352                 fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4353                 fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4354                 fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4355                 fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4356                 fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4357                 fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4358                 fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4359                 fsp->fs_maxfilesize = fxdr_hyper(tl);
4360                 tl += 2;
4361                 fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4362                 tl += 2;
4363                 fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4364         }
4365 nfsmout:
4366         mbuf_freem(nd->nd_mrep);
4367         return (error);
4368 }
4369
4370 /*
4371  * This function performs the Renew RPC.
4372  */
4373 APPLESTATIC int
4374 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4375     NFSPROC_T *p)
4376 {
4377         u_int32_t *tl;
4378         struct nfsrv_descript nfsd;
4379         struct nfsrv_descript *nd = &nfsd;
4380         struct nfsmount *nmp;
4381         int error;
4382         struct nfssockreq *nrp;
4383         struct nfsclsession *tsep;
4384
4385         nmp = clp->nfsc_nmp;
4386         if (nmp == NULL)
4387                 return (0);
4388         if (dsp == NULL)
4389                 nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4390                     0);
4391         else
4392                 nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4393                     &dsp->nfsclds_sess, 0, 0);
4394         if (!NFSHASNFSV4N(nmp)) {
4395                 /* NFSv4.1 just uses a Sequence Op and not a Renew. */
4396                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4397                 tsep = nfsmnt_mdssession(nmp);
4398                 *tl++ = tsep->nfsess_clientid.lval[0];
4399                 *tl = tsep->nfsess_clientid.lval[1];
4400         }
4401         nrp = NULL;
4402         if (dsp != NULL)
4403                 nrp = dsp->nfsclds_sockp;
4404         if (nrp == NULL)
4405                 /* If NULL, use the MDS socket. */
4406                 nrp = &nmp->nm_sockreq;
4407         nd->nd_flag |= ND_USEGSSNAME;
4408         if (dsp == NULL)
4409                 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4410                     NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4411         else
4412                 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4413                     NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4414         if (error)
4415                 return (error);
4416         error = nd->nd_repstat;
4417         mbuf_freem(nd->nd_mrep);
4418         return (error);
4419 }
4420
4421 /*
4422  * This function performs the Releaselockowner RPC.
4423  */
4424 APPLESTATIC int
4425 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4426     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4427 {
4428         struct nfsrv_descript nfsd, *nd = &nfsd;
4429         u_int32_t *tl;
4430         int error;
4431         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4432         struct nfsclsession *tsep;
4433
4434         if (NFSHASNFSV4N(nmp)) {
4435                 /* For NFSv4.1, do a FreeStateID. */
4436                 nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4437                     NULL, 0, 0);
4438                 nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4439         } else {
4440                 nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4441                     NULL, 0, 0);
4442                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4443                 tsep = nfsmnt_mdssession(nmp);
4444                 *tl++ = tsep->nfsess_clientid.lval[0];
4445                 *tl = tsep->nfsess_clientid.lval[1];
4446                 NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4447                 NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4448                 (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4449         }
4450         nd->nd_flag |= ND_USEGSSNAME;
4451         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4452             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4453         if (error)
4454                 return (error);
4455         error = nd->nd_repstat;
4456         mbuf_freem(nd->nd_mrep);
4457         return (error);
4458 }
4459
4460 /*
4461  * This function performs the Compound to get the mount pt FH.
4462  */
4463 APPLESTATIC int
4464 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4465     NFSPROC_T *p)
4466 {
4467         u_int32_t *tl;
4468         struct nfsrv_descript nfsd;
4469         struct nfsrv_descript *nd = &nfsd;
4470         u_char *cp, *cp2;
4471         int error, cnt, len, setnil;
4472         u_int32_t *opcntp;
4473
4474         nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4475             0);
4476         cp = dirpath;
4477         cnt = 0;
4478         do {
4479                 setnil = 0;
4480                 while (*cp == '/')
4481                         cp++;
4482                 cp2 = cp;
4483                 while (*cp2 != '\0' && *cp2 != '/')
4484                         cp2++;
4485                 if (*cp2 == '/') {
4486                         setnil = 1;
4487                         *cp2 = '\0';
4488                 }
4489                 if (cp2 != cp) {
4490                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4491                         *tl = txdr_unsigned(NFSV4OP_LOOKUP);
4492                         nfsm_strtom(nd, cp, strlen(cp));
4493                         cnt++;
4494                 }
4495                 if (setnil)
4496                         *cp2++ = '/';
4497                 cp = cp2;
4498         } while (*cp != '\0');
4499         if (NFSHASNFSV4N(nmp))
4500                 /* Has a Sequence Op done by nfscl_reqstart(). */
4501                 *opcntp = txdr_unsigned(3 + cnt);
4502         else
4503                 *opcntp = txdr_unsigned(2 + cnt);
4504         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4505         *tl = txdr_unsigned(NFSV4OP_GETFH);
4506         nd->nd_flag |= ND_USEGSSNAME;
4507         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4508                 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4509         if (error)
4510                 return (error);
4511         if (nd->nd_repstat == 0) {
4512                 NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4513                 tl += (2 + 2 * cnt);
4514                 if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4515                         len > NFSX_FHMAX) {
4516                         nd->nd_repstat = NFSERR_BADXDR;
4517                 } else {
4518                         nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4519                         if (nd->nd_repstat == 0)
4520                                 nmp->nm_fhsize = len;
4521                 }
4522         }
4523         error = nd->nd_repstat;
4524 nfsmout:
4525         mbuf_freem(nd->nd_mrep);
4526         return (error);
4527 }
4528
4529 /*
4530  * This function performs the Delegreturn RPC.
4531  */
4532 APPLESTATIC int
4533 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4534     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4535 {
4536         u_int32_t *tl;
4537         struct nfsrv_descript nfsd;
4538         struct nfsrv_descript *nd = &nfsd;
4539         int error;
4540
4541         nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4542             dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4543         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4544         if (NFSHASNFSV4N(nmp))
4545                 *tl++ = 0;
4546         else
4547                 *tl++ = dp->nfsdl_stateid.seqid;
4548         *tl++ = dp->nfsdl_stateid.other[0];
4549         *tl++ = dp->nfsdl_stateid.other[1];
4550         *tl = dp->nfsdl_stateid.other[2];
4551         if (syscred)
4552                 nd->nd_flag |= ND_USEGSSNAME;
4553         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4554             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4555         if (error)
4556                 return (error);
4557         error = nd->nd_repstat;
4558         mbuf_freem(nd->nd_mrep);
4559         return (error);
4560 }
4561
4562 /*
4563  * nfs getacl call.
4564  */
4565 APPLESTATIC int
4566 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4567     struct acl *aclp, void *stuff)
4568 {
4569         struct nfsrv_descript nfsd, *nd = &nfsd;
4570         int error;
4571         nfsattrbit_t attrbits;
4572         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4573
4574         if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4575                 return (EOPNOTSUPP);
4576         NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4577         NFSZERO_ATTRBIT(&attrbits);
4578         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4579         (void) nfsrv_putattrbit(nd, &attrbits);
4580         error = nfscl_request(nd, vp, p, cred, stuff);
4581         if (error)
4582                 return (error);
4583         if (!nd->nd_repstat)
4584                 error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4585                     NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4586         else
4587                 error = nd->nd_repstat;
4588         mbuf_freem(nd->nd_mrep);
4589         return (error);
4590 }
4591
4592 /*
4593  * nfs setacl call.
4594  */
4595 APPLESTATIC int
4596 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4597     struct acl *aclp, void *stuff)
4598 {
4599         int error;
4600         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4601
4602         if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4603                 return (EOPNOTSUPP);
4604         error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4605         return (error);
4606 }
4607
4608 /*
4609  * nfs setacl call.
4610  */
4611 static int
4612 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4613     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4614 {
4615         struct nfsrv_descript nfsd, *nd = &nfsd;
4616         int error;
4617         nfsattrbit_t attrbits;
4618         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4619
4620         if (!NFSHASNFSV4(nmp))
4621                 return (EOPNOTSUPP);
4622         NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4623         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4624         NFSZERO_ATTRBIT(&attrbits);
4625         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4626         (void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4627             &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4628         error = nfscl_request(nd, vp, p, cred, stuff);
4629         if (error)
4630                 return (error);
4631         /* Don't care about the pre/postop attributes */
4632         mbuf_freem(nd->nd_mrep);
4633         return (nd->nd_repstat);
4634 }
4635
4636 /*
4637  * Do the NFSv4.1 Exchange ID.
4638  */
4639 int
4640 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4641     struct nfssockreq *nrp, uint32_t exchflags, struct nfsclds **dspp,
4642     struct ucred *cred, NFSPROC_T *p)
4643 {
4644         uint32_t *tl, v41flags;
4645         struct nfsrv_descript nfsd;
4646         struct nfsrv_descript *nd = &nfsd;
4647         struct nfsclds *dsp;
4648         struct timespec verstime;
4649         int error, len;
4650
4651         *dspp = NULL;
4652         nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL, 0, 0);
4653         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4654         *tl++ = txdr_unsigned(nfsboottime.tv_sec);      /* Client owner */
4655         *tl = txdr_unsigned(clp->nfsc_rev);
4656         (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4657
4658         NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4659         *tl++ = txdr_unsigned(exchflags);
4660         *tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4661
4662         /* Set the implementation id4 */
4663         *tl = txdr_unsigned(1);
4664         (void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4665         (void) nfsm_strtom(nd, version, strlen(version));
4666         NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4667         verstime.tv_sec = 1293840000;           /* Jan 1, 2011 */
4668         verstime.tv_nsec = 0;
4669         txdr_nfsv4time(&verstime, tl);
4670         nd->nd_flag |= ND_USEGSSNAME;
4671         error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4672             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4673         NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4674             (int)nd->nd_repstat);
4675         if (error != 0)
4676                 return (error);
4677         if (nd->nd_repstat == 0) {
4678                 NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4679                 len = fxdr_unsigned(int, *(tl + 7));
4680                 if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4681                         error = NFSERR_BADXDR;
4682                         goto nfsmout;
4683                 }
4684                 dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4685                     M_WAITOK | M_ZERO);
4686                 dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4687                 dsp->nfsclds_servownlen = len;
4688                 dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4689                 dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4690                 dsp->nfsclds_sess.nfsess_sequenceid =
4691                     fxdr_unsigned(uint32_t, *tl++);
4692                 v41flags = fxdr_unsigned(uint32_t, *tl);
4693                 if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4694                     NFSHASPNFSOPT(nmp)) {
4695                         NFSCL_DEBUG(1, "set PNFS\n");
4696                         NFSLOCKMNT(nmp);
4697                         nmp->nm_state |= NFSSTA_PNFS;
4698                         NFSUNLOCKMNT(nmp);
4699                         dsp->nfsclds_flags |= NFSCLDS_MDS;
4700                 }
4701                 if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4702                         dsp->nfsclds_flags |= NFSCLDS_DS;
4703                 if (len > 0)
4704                         nd->nd_repstat = nfsrv_mtostr(nd,
4705                             dsp->nfsclds_serverown, len);
4706                 if (nd->nd_repstat == 0) {
4707                         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4708                         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4709                             NULL, MTX_DEF);
4710                         nfscl_initsessionslots(&dsp->nfsclds_sess);
4711                         *dspp = dsp;
4712                 } else
4713                         free(dsp, M_NFSCLDS);
4714         }
4715         error = nd->nd_repstat;
4716 nfsmout:
4717         mbuf_freem(nd->nd_mrep);
4718         return (error);
4719 }
4720
4721 /*
4722  * Do the NFSv4.1 Create Session.
4723  */
4724 int
4725 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4726     struct nfssockreq *nrp, uint32_t sequenceid, int mds, struct ucred *cred,
4727     NFSPROC_T *p)
4728 {
4729         uint32_t crflags, maxval, *tl;
4730         struct nfsrv_descript nfsd;
4731         struct nfsrv_descript *nd = &nfsd;
4732         int error, irdcnt;
4733
4734         /* Make sure nm_rsize, nm_wsize is set. */
4735         if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4736                 nmp->nm_rsize = NFS_MAXBSIZE;
4737         if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4738                 nmp->nm_wsize = NFS_MAXBSIZE;
4739         nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL, 0,
4740             0);
4741         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4742         *tl++ = sep->nfsess_clientid.lval[0];
4743         *tl++ = sep->nfsess_clientid.lval[1];
4744         *tl++ = txdr_unsigned(sequenceid);
4745         crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4746         if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4747                 crflags |= NFSV4CRSESS_CONNBACKCHAN;
4748         *tl = txdr_unsigned(crflags);
4749
4750         /* Fill in fore channel attributes. */
4751         NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4752         *tl++ = 0;                              /* Header pad size */
4753         *tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);/* Max request size */
4754         *tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);/* Max reply size */
4755         *tl++ = txdr_unsigned(4096);            /* Max response size cached */
4756         *tl++ = txdr_unsigned(20);              /* Max operations */
4757         *tl++ = txdr_unsigned(64);              /* Max slots */
4758         *tl = 0;                                /* No rdma ird */
4759
4760         /* Fill in back channel attributes. */
4761         NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4762         *tl++ = 0;                              /* Header pad size */
4763         *tl++ = txdr_unsigned(10000);           /* Max request size */
4764         *tl++ = txdr_unsigned(10000);           /* Max response size */
4765         *tl++ = txdr_unsigned(4096);            /* Max response size cached */
4766         *tl++ = txdr_unsigned(4);               /* Max operations */
4767         *tl++ = txdr_unsigned(NFSV4_CBSLOTS);   /* Max slots */
4768         *tl = 0;                                /* No rdma ird */
4769
4770         NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4771         *tl++ = txdr_unsigned(NFS_CALLBCKPROG); /* Call back prog # */
4772
4773         /* Allow AUTH_SYS callbacks as uid, gid == 0. */
4774         *tl++ = txdr_unsigned(1);               /* Auth_sys only */
4775         *tl++ = txdr_unsigned(AUTH_SYS);        /* AUTH_SYS type */
4776         *tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4777         *tl++ = 0;                              /* Null machine name */
4778         *tl++ = 0;                              /* Uid == 0 */
4779         *tl++ = 0;                              /* Gid == 0 */
4780         *tl = 0;                                /* No additional gids */
4781         nd->nd_flag |= ND_USEGSSNAME;
4782         error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4783             NFS_VER4, NULL, 1, NULL, NULL);
4784         if (error != 0)
4785                 return (error);
4786         if (nd->nd_repstat == 0) {
4787                 NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4788                     2 * NFSX_UNSIGNED);
4789                 bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4790                 tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4791                 sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4792                 crflags = fxdr_unsigned(uint32_t, *tl);
4793                 if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4794                         NFSLOCKMNT(nmp);
4795                         nmp->nm_state |= NFSSTA_SESSPERSIST;
4796                         NFSUNLOCKMNT(nmp);
4797                 }
4798
4799                 /* Get the fore channel slot count. */
4800                 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4801                 tl++;                   /* Skip the header pad size. */
4802
4803                 /* Make sure nm_wsize is small enough. */
4804                 maxval = fxdr_unsigned(uint32_t, *tl++);
4805                 while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4806                         if (nmp->nm_wsize > 8096)
4807                                 nmp->nm_wsize /= 2;
4808                         else
4809                                 break;
4810                 }
4811
4812                 /* Make sure nm_rsize is small enough. */
4813                 maxval = fxdr_unsigned(uint32_t, *tl++);
4814                 while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4815                         if (nmp->nm_rsize > 8096)
4816                                 nmp->nm_rsize /= 2;
4817                         else
4818                                 break;
4819                 }
4820
4821                 sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4822                 tl++;
4823                 sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4824                 NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4825                 irdcnt = fxdr_unsigned(int, *tl);
4826                 if (irdcnt > 0)
4827                         NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4828
4829                 /* and the back channel slot count. */
4830                 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4831                 tl += 5;
4832                 sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4833                 NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4834         }
4835         error = nd->nd_repstat;
4836 nfsmout:
4837         mbuf_freem(nd->nd_mrep);
4838         return (error);
4839 }
4840
4841 /*
4842  * Do the NFSv4.1 Destroy Session.
4843  */
4844 int
4845 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4846     struct ucred *cred, NFSPROC_T *p)
4847 {
4848         uint32_t *tl;
4849         struct nfsrv_descript nfsd;
4850         struct nfsrv_descript *nd = &nfsd;
4851         int error;
4852         struct nfsclsession *tsep;
4853
4854         nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
4855             0);
4856         NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4857         tsep = nfsmnt_mdssession(nmp);
4858         bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4859         nd->nd_flag |= ND_USEGSSNAME;
4860         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4861             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4862         if (error != 0)
4863                 return (error);
4864         error = nd->nd_repstat;
4865         mbuf_freem(nd->nd_mrep);
4866         return (error);
4867 }
4868
4869 /*
4870  * Do the NFSv4.1 Destroy Client.
4871  */
4872 int
4873 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4874     struct ucred *cred, NFSPROC_T *p)
4875 {
4876         uint32_t *tl;
4877         struct nfsrv_descript nfsd;
4878         struct nfsrv_descript *nd = &nfsd;
4879         int error;
4880         struct nfsclsession *tsep;
4881
4882         nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
4883             0);
4884         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4885         tsep = nfsmnt_mdssession(nmp);
4886         *tl++ = tsep->nfsess_clientid.lval[0];
4887         *tl = tsep->nfsess_clientid.lval[1];
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         error = nd->nd_repstat;
4894         mbuf_freem(nd->nd_mrep);
4895         return (error);
4896 }
4897
4898 /*
4899  * Do the NFSv4.1 LayoutGet.
4900  */
4901 static int
4902 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
4903     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
4904     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
4905     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
4906     void *stuff)
4907 {
4908         struct nfsrv_descript nfsd, *nd = &nfsd;
4909         int error;
4910
4911         nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
4912             0);
4913         nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
4914             layouttype, layoutlen, 0);
4915         nd->nd_flag |= ND_USEGSSNAME;
4916         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4917             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4918         NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
4919         if (error != 0)
4920                 return (error);
4921         if (nd->nd_repstat == 0)
4922                 error = nfsrv_parselayoutget(nd, stateidp, retonclosep, flhp);
4923         if (error == 0 && nd->nd_repstat != 0)
4924                 error = nd->nd_repstat;
4925         mbuf_freem(nd->nd_mrep);
4926         return (error);
4927 }
4928
4929 /*
4930  * Do the NFSv4.1 Get Device Info.
4931  */
4932 int
4933 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
4934     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
4935     NFSPROC_T *p)
4936 {
4937         uint32_t cnt, *tl, vers, minorvers;
4938         struct nfsrv_descript nfsd;
4939         struct nfsrv_descript *nd = &nfsd;
4940         struct sockaddr_in sin, ssin;
4941         struct sockaddr_in6 sin6, ssin6;
4942         struct nfsclds *dsp = NULL, **dspp, **gotdspp;
4943         struct nfscldevinfo *ndi;
4944         int addrcnt = 0, bitcnt, error, gotvers, i, isudp, j, stripecnt;
4945         uint8_t stripeindex;
4946         sa_family_t af, safilled;
4947
4948         *ndip = NULL;
4949         ndi = NULL;
4950         gotdspp = NULL;
4951         nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
4952             0);
4953         NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
4954         NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
4955         tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4956         *tl++ = txdr_unsigned(layouttype);
4957         *tl++ = txdr_unsigned(100000);
4958         if (notifybitsp != NULL && *notifybitsp != 0) {
4959                 *tl = txdr_unsigned(1);         /* One word of bits. */
4960                 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
4961                 *tl = txdr_unsigned(*notifybitsp);
4962         } else
4963                 *tl = txdr_unsigned(0);
4964         nd->nd_flag |= ND_USEGSSNAME;
4965         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4966             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4967         if (error != 0)
4968                 return (error);
4969         if (nd->nd_repstat == 0) {
4970                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4971                 if (layouttype != fxdr_unsigned(int, *tl))
4972                         printf("EEK! devinfo layout type not same!\n");
4973                 if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
4974                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4975                         stripecnt = fxdr_unsigned(int, *tl);
4976                         NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
4977                         if (stripecnt < 1 || stripecnt > 4096) {
4978                                 printf("pNFS File layout devinfo stripecnt %d:"
4979                                     " out of range\n", stripecnt);
4980                                 error = NFSERR_BADXDR;
4981                                 goto nfsmout;
4982                         }
4983                         NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
4984                             NFSX_UNSIGNED);
4985                         addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
4986                         NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
4987                         if (addrcnt < 1 || addrcnt > 128) {
4988                                 printf("NFS devinfo addrcnt %d: out of range\n",
4989                                     addrcnt);
4990                                 error = NFSERR_BADXDR;
4991                                 goto nfsmout;
4992                         }
4993
4994                         /*
4995                          * Now we know how many stripe indices and addresses, so
4996                          * we can allocate the structure the correct size.
4997                          */
4998                         i = (stripecnt * sizeof(uint8_t)) /
4999                             sizeof(struct nfsclds *) + 1;
5000                         NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5001                         ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5002                             sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5003                             M_ZERO);
5004                         NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5005                             NFSX_V4DEVICEID);
5006                         ndi->nfsdi_refcnt = 0;
5007                         ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5008                         ndi->nfsdi_stripecnt = stripecnt;
5009                         ndi->nfsdi_addrcnt = addrcnt;
5010                         /* Fill in the stripe indices. */
5011                         for (i = 0; i < stripecnt; i++) {
5012                                 stripeindex = fxdr_unsigned(uint8_t, *tl++);
5013                                 NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5014                                 if (stripeindex >= addrcnt) {
5015                                         printf("pNFS File Layout devinfo"
5016                                             " stripeindex %d: too big\n",
5017                                             (int)stripeindex);
5018                                         error = NFSERR_BADXDR;
5019                                         goto nfsmout;
5020                                 }
5021                                 nfsfldi_setstripeindex(ndi, i, stripeindex);
5022                         }
5023                 } else if (layouttype == NFSLAYOUT_FLEXFILE) {
5024                         /* For Flex File, we only get one address list. */
5025                         ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5026                             M_NFSDEVINFO, M_WAITOK | M_ZERO);
5027                         NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5028                             NFSX_V4DEVICEID);
5029                         ndi->nfsdi_refcnt = 0;
5030                         ndi->nfsdi_flags = NFSDI_FLEXFILE;
5031                         addrcnt = ndi->nfsdi_addrcnt = 1;
5032                 }
5033
5034                 /* Now, dissect the server address(es). */
5035                 safilled = AF_UNSPEC;
5036                 for (i = 0; i < addrcnt; i++) {
5037                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5038                         cnt = fxdr_unsigned(uint32_t, *tl);
5039                         if (cnt == 0) {
5040                                 printf("NFS devinfo 0 len addrlist\n");
5041                                 error = NFSERR_BADXDR;
5042                                 goto nfsmout;
5043                         }
5044                         dspp = nfsfldi_addr(ndi, i);
5045                         safilled = AF_UNSPEC;
5046                         for (j = 0; j < cnt; j++) {
5047                                 error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5048                                     &isudp);
5049                                 if (error != 0 && error != EPERM) {
5050                                         error = NFSERR_BADXDR;
5051                                         goto nfsmout;
5052                                 }
5053                                 if (error == 0 && isudp == 0) {
5054                                         /*
5055                                          * The priority is:
5056                                          * - Same address family.
5057                                          * Save the address and dspp, so that
5058                                          * the connection can be done after
5059                                          * parsing is complete.
5060                                          */
5061                                         if (safilled == AF_UNSPEC ||
5062                                             (af == nmp->nm_nam->sa_family &&
5063                                              safilled != nmp->nm_nam->sa_family)
5064                                            ) {
5065                                                 if (af == AF_INET)
5066                                                         ssin = sin;
5067                                                 else
5068                                                         ssin6 = sin6;
5069                                                 safilled = af;
5070                                                 gotdspp = dspp;
5071                                         }
5072                                 }
5073                         }
5074                 }
5075
5076                 gotvers = NFS_VER4;     /* Always NFSv4 for File Layout. */
5077                 /* For Flex File, we will take one of the versions to use. */
5078                 if (layouttype == NFSLAYOUT_FLEXFILE) {
5079                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5080                         j = fxdr_unsigned(int, *tl);
5081                         if (j < 1 || j > NFSDEV_MAXVERS) {
5082                                 printf("pNFS: too many versions\n");
5083                                 error = NFSERR_BADXDR;
5084                                 goto nfsmout;
5085                         }
5086                         gotvers = 0;
5087                         for (i = 0; i < j; i++) {
5088                                 NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5089                                 vers = fxdr_unsigned(uint32_t, *tl++);
5090                                 minorvers = fxdr_unsigned(uint32_t, *tl++);
5091                                 if ((vers == NFS_VER4 && minorvers ==
5092                                     NFSV41_MINORVERSION) || (vers == NFS_VER3 &&
5093                                     gotvers == 0)) {
5094                                         gotvers = vers;
5095                                         /* We'll take this one. */
5096                                         ndi->nfsdi_versindex = i;
5097                                         ndi->nfsdi_vers = vers;
5098                                         ndi->nfsdi_minorvers = minorvers;
5099                                         ndi->nfsdi_rsize = fxdr_unsigned(
5100                                             uint32_t, *tl++);
5101                                         ndi->nfsdi_wsize = fxdr_unsigned(
5102                                             uint32_t, *tl++);
5103                                         if (*tl == newnfs_true)
5104                                                 ndi->nfsdi_flags |=
5105                                                     NFSDI_TIGHTCOUPLED;
5106                                         else
5107                                                 ndi->nfsdi_flags &=
5108                                                     ~NFSDI_TIGHTCOUPLED;
5109                                 }
5110                         }
5111                         if (gotvers == 0) {
5112                                 printf("pNFS: no NFSv3 or NFSv4.1\n");
5113                                 error = NFSERR_BADXDR;
5114                                 goto nfsmout;
5115                         }
5116                 }
5117
5118                 /* And the notify bits. */
5119                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5120                 bitcnt = fxdr_unsigned(int, *tl);
5121                 if (bitcnt > 0) {
5122                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5123                         if (notifybitsp != NULL)
5124                                 *notifybitsp =
5125                                     fxdr_unsigned(uint32_t, *tl);
5126                 }
5127                 if (safilled != AF_UNSPEC) {
5128                         KASSERT(ndi != NULL, ("ndi is NULL"));
5129                         *ndip = ndi;
5130                 } else
5131                         error = EPERM;
5132                 if (error == 0) {
5133                         /*
5134                          * Now we can do a TCP connection for the correct
5135                          * NFS version and IP address.
5136                          */
5137                         error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5138                             gotvers, &dsp, p);
5139                 }
5140                 if (error == 0) {
5141                         KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5142                         *gotdspp = dsp;
5143                 }
5144         }
5145         if (nd->nd_repstat != 0 && error == 0)
5146                 error = nd->nd_repstat;
5147 nfsmout:
5148         if (error != 0 && ndi != NULL)
5149                 nfscl_freedevinfo(ndi);
5150         mbuf_freem(nd->nd_mrep);
5151         return (error);
5152 }
5153
5154 /*
5155  * Do the NFSv4.1 LayoutCommit.
5156  */
5157 int
5158 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5159     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5160     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5161 {
5162         uint32_t *tl;
5163         struct nfsrv_descript nfsd, *nd = &nfsd;
5164         int error;
5165
5166         nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5167             0, 0);
5168         NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5169             NFSX_STATEID);
5170         txdr_hyper(off, tl);
5171         tl += 2;
5172         txdr_hyper(len, tl);
5173         tl += 2;
5174         if (reclaim != 0)
5175                 *tl++ = newnfs_true;
5176         else
5177                 *tl++ = newnfs_false;
5178         *tl++ = txdr_unsigned(stateidp->seqid);
5179         *tl++ = stateidp->other[0];
5180         *tl++ = stateidp->other[1];
5181         *tl++ = stateidp->other[2];
5182         *tl++ = newnfs_true;
5183         if (lastbyte < off)
5184                 lastbyte = off;
5185         else if (lastbyte >= (off + len))
5186                 lastbyte = off + len - 1;
5187         txdr_hyper(lastbyte, tl);
5188         tl += 2;
5189         *tl++ = newnfs_false;
5190         *tl++ = txdr_unsigned(layouttype);
5191         /* All supported layouts are 0 length. */
5192         *tl = txdr_unsigned(0);
5193         nd->nd_flag |= ND_USEGSSNAME;
5194         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5195             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5196         if (error != 0)
5197                 return (error);
5198         error = nd->nd_repstat;
5199         mbuf_freem(nd->nd_mrep);
5200         return (error);
5201 }
5202
5203 /*
5204  * Do the NFSv4.1 LayoutReturn.
5205  */
5206 int
5207 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5208     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5209     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5210     uint32_t stat, uint32_t op, char *devid)
5211 {
5212         uint32_t *tl;
5213         struct nfsrv_descript nfsd, *nd = &nfsd;
5214         uint64_t tu64;
5215         int error;
5216
5217         nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5218             0, 0);
5219         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5220         if (reclaim != 0)
5221                 *tl++ = newnfs_true;
5222         else
5223                 *tl++ = newnfs_false;
5224         *tl++ = txdr_unsigned(layouttype);
5225         *tl++ = txdr_unsigned(iomode);
5226         *tl = txdr_unsigned(layoutreturn);
5227         if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5228                 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5229                     NFSX_UNSIGNED);
5230                 txdr_hyper(offset, tl);
5231                 tl += 2;
5232                 txdr_hyper(len, tl);
5233                 tl += 2;
5234                 NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5235                 *tl++ = txdr_unsigned(stateidp->seqid);
5236                 *tl++ = stateidp->other[0];
5237                 *tl++ = stateidp->other[1];
5238                 *tl++ = stateidp->other[2];
5239                 if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5240                         *tl = txdr_unsigned(0);
5241                 else if (layouttype == NFSLAYOUT_FLEXFILE) {
5242                         if (stat != 0) {
5243                                 *tl = txdr_unsigned(2 * NFSX_HYPER +
5244                                     NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5245                                     NFSX_UNSIGNED);
5246                                 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5247                                     NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5248                                     NFSX_UNSIGNED);
5249                                 *tl++ = txdr_unsigned(1);       /* One error. */
5250                                 tu64 = 0;                       /* Offset. */
5251                                 txdr_hyper(tu64, tl); tl += 2;
5252                                 tu64 = UINT64_MAX;              /* Length. */
5253                                 txdr_hyper(tu64, tl); tl += 2;
5254                                 NFSBCOPY(stateidp, tl, NFSX_STATEID);
5255                                 tl += (NFSX_STATEID / NFSX_UNSIGNED);
5256                                 *tl++ = txdr_unsigned(1);       /* One error. */
5257                                 NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5258                                 tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5259                                 *tl++ = txdr_unsigned(stat);
5260                                 *tl++ = txdr_unsigned(op);
5261                         } else {
5262                                 *tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5263                                 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5264                                 /* No ioerrs. */
5265                                 *tl++ = 0;
5266                         }
5267                         *tl = 0;        /* No stats yet. */
5268                 }
5269         }
5270         nd->nd_flag |= ND_USEGSSNAME;
5271         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5272             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5273         if (error != 0)
5274                 return (error);
5275         if (nd->nd_repstat == 0) {
5276                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5277                 if (*tl != 0) {
5278                         NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5279                         stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5280                         stateidp->other[0] = *tl++;
5281                         stateidp->other[1] = *tl++;
5282                         stateidp->other[2] = *tl;
5283                 }
5284         } else
5285                 error = nd->nd_repstat;
5286 nfsmout:
5287         mbuf_freem(nd->nd_mrep);
5288         return (error);
5289 }
5290
5291 /*
5292  * Acquire a layout and devinfo, if possible. The caller must have acquired
5293  * a reference count on the nfsclclient structure before calling this.
5294  * Return the layout in lypp with a reference count on it, if successful.
5295  */
5296 static int
5297 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5298     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5299     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5300 {
5301         struct nfscllayout *lyp;
5302         struct nfsclflayout *flp;
5303         struct nfsclflayouthead flh;
5304         int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5305         nfsv4stateid_t stateid;
5306         struct nfsclsession *tsep;
5307
5308         *lypp = NULL;
5309         if (NFSHASFLEXFILE(nmp))
5310                 layouttype = NFSLAYOUT_FLEXFILE;
5311         else
5312                 layouttype = NFSLAYOUT_NFSV4_1_FILES;
5313         /*
5314          * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5315          * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5316          * flp == NULL.
5317          */
5318         lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5319             off, &flp, &recalled);
5320         islocked = 0;
5321         if (lyp == NULL || flp == NULL) {
5322                 if (recalled != 0)
5323                         return (EIO);
5324                 LIST_INIT(&flh);
5325                 tsep = nfsmnt_mdssession(nmp);
5326                 layoutlen = tsep->nfsess_maxcache -
5327                     (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5328                 if (lyp == NULL) {
5329                         stateid.seqid = 0;
5330                         stateid.other[0] = stateidp->other[0];
5331                         stateid.other[1] = stateidp->other[1];
5332                         stateid.other[2] = stateidp->other[2];
5333                         error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5334                             nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5335                             (uint64_t)0, layouttype, layoutlen, &stateid,
5336                             &retonclose, &flh, cred, p, NULL);
5337                 } else {
5338                         islocked = 1;
5339                         stateid.seqid = lyp->nfsly_stateid.seqid;
5340                         stateid.other[0] = lyp->nfsly_stateid.other[0];
5341                         stateid.other[1] = lyp->nfsly_stateid.other[1];
5342                         stateid.other[2] = lyp->nfsly_stateid.other[2];
5343                         error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5344                             nfhp->nfh_len, iomode, off, UINT64_MAX,
5345                             (uint64_t)0, layouttype, layoutlen, &stateid,
5346                             &retonclose, &flh, cred, p, NULL);
5347                 }
5348                 error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5349                     nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5350                     &flh, layouttype, error, NULL, cred, p);
5351                 if (error == 0)
5352                         *lypp = lyp;
5353                 else if (islocked != 0)
5354                         nfscl_rellayout(lyp, 1);
5355         } else
5356                 *lypp = lyp;
5357         return (error);
5358 }
5359
5360 /*
5361  * Do a TCP connection plus exchange id and create session.
5362  * If successful, a "struct nfsclds" is linked into the list for the
5363  * mount point and a pointer to it is returned.
5364  */
5365 static int
5366 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5367     struct sockaddr_in6 *sin6, sa_family_t af, int vers, struct nfsclds **dspp,
5368     NFSPROC_T *p)
5369 {
5370         struct sockaddr_in *msad, *sad;
5371         struct sockaddr_in6 *msad6, *sad6;
5372         struct nfsclclient *clp;
5373         struct nfssockreq *nrp;
5374         struct nfsclds *dsp, *tdsp;
5375         int error;
5376         enum nfsclds_state retv;
5377         uint32_t sequenceid;
5378
5379         KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5380             ("nfsrpc_fillsa: NULL nr_cred"));
5381         NFSLOCKCLSTATE();
5382         clp = nmp->nm_clp;
5383         NFSUNLOCKCLSTATE();
5384         if (clp == NULL)
5385                 return (EPERM);
5386         if (af == AF_INET) {
5387                 NFSLOCKMNT(nmp);
5388                 /*
5389                  * Check to see if we already have a session for this
5390                  * address that is usable for a DS.
5391                  * Note that the MDS's address is in a different place
5392                  * than the sessions already acquired for DS's.
5393                  */
5394                 msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5395                 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5396                 while (tdsp != NULL) {
5397                         if (msad != NULL && msad->sin_family == AF_INET &&
5398                             sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5399                             sin->sin_port == msad->sin_port &&
5400                             (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5401                             tdsp->nfsclds_sess.nfsess_defunct == 0) {
5402                                 *dspp = tdsp;
5403                                 NFSUNLOCKMNT(nmp);
5404                                 NFSCL_DEBUG(4, "fnd same addr\n");
5405                                 return (0);
5406                         }
5407                         tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5408                         if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5409                                 msad = (struct sockaddr_in *)
5410                                     tdsp->nfsclds_sockp->nr_nam;
5411                         else
5412                                 msad = NULL;
5413                 }
5414                 NFSUNLOCKMNT(nmp);
5415
5416                 /* No IP address match, so look for new/trunked one. */
5417                 sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5418                 sad->sin_len = sizeof(*sad);
5419                 sad->sin_family = AF_INET;
5420                 sad->sin_port = sin->sin_port;
5421                 sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5422                 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5423                 nrp->nr_nam = (struct sockaddr *)sad;
5424         } else if (af == AF_INET6) {
5425                 NFSLOCKMNT(nmp);
5426                 /*
5427                  * Check to see if we already have a session for this
5428                  * address that is usable for a DS.
5429                  * Note that the MDS's address is in a different place
5430                  * than the sessions already acquired for DS's.
5431                  */
5432                 msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5433                 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5434                 while (tdsp != NULL) {
5435                         if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5436                             IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5437                             &msad6->sin6_addr) &&
5438                             sin6->sin6_port == msad6->sin6_port &&
5439                             (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5440                             tdsp->nfsclds_sess.nfsess_defunct == 0) {
5441                                 *dspp = tdsp;
5442                                 NFSUNLOCKMNT(nmp);
5443                                 return (0);
5444                         }
5445                         tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5446                         if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5447                                 msad6 = (struct sockaddr_in6 *)
5448                                     tdsp->nfsclds_sockp->nr_nam;
5449                         else
5450                                 msad6 = NULL;
5451                 }
5452                 NFSUNLOCKMNT(nmp);
5453
5454                 /* No IP address match, so look for new/trunked one. */
5455                 sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5456                 sad6->sin6_len = sizeof(*sad6);
5457                 sad6->sin6_family = AF_INET6;
5458                 sad6->sin6_port = sin6->sin6_port;
5459                 NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5460                     sizeof(struct in6_addr));
5461                 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5462                 nrp->nr_nam = (struct sockaddr *)sad6;
5463         } else
5464                 return (EPERM);
5465
5466         nrp->nr_sotype = SOCK_STREAM;
5467         mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5468         nrp->nr_prog = NFS_PROG;
5469         nrp->nr_vers = vers;
5470
5471         /*
5472          * Use the credentials that were used for the mount, which are
5473          * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5474          * Ref. counting the credentials with crhold() is probably not
5475          * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5476          * unmount, but I did it anyhow.
5477          */
5478         nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5479         error = newnfs_connect(nmp, nrp, NULL, p, 0);
5480         NFSCL_DEBUG(3, "DS connect=%d\n", error);
5481
5482         dsp = NULL;
5483         /* Now, do the exchangeid and create session. */
5484         if (error == 0) {
5485                 if (vers == NFS_VER4) {
5486                         error = nfsrpc_exchangeid(nmp, clp, nrp,
5487                             NFSV4EXCH_USEPNFSDS, &dsp, nrp->nr_cred, p);
5488                         NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5489                         if (error != 0)
5490                                 newnfs_disconnect(nrp);
5491                 } else {
5492                         dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5493                             M_WAITOK | M_ZERO);
5494                         dsp->nfsclds_flags |= NFSCLDS_DS;
5495                         dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5496                         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5497                         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5498                             NULL, MTX_DEF);
5499                 }
5500         }
5501         if (error == 0) {
5502                 dsp->nfsclds_sockp = nrp;
5503                 if (vers == NFS_VER4) {
5504                         NFSLOCKMNT(nmp);
5505                         retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5506                             &sequenceid);
5507                         NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5508                         if (retv == NFSDSP_USETHISSESSION &&
5509                             nfscl_dssameconn != 0) {
5510                                 NFSLOCKDS(tdsp);
5511                                 tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5512                                 NFSUNLOCKDS(tdsp);
5513                                 NFSUNLOCKMNT(nmp);
5514                                 /*
5515                                  * If there is already a session for this
5516                                  * server, use it.
5517                                  */
5518                                 (void)newnfs_disconnect(nrp);
5519                                 nfscl_freenfsclds(dsp);
5520                                 *dspp = tdsp;
5521                                 return (0);
5522                         }
5523                         if (retv == NFSDSP_NOTFOUND)
5524                                 sequenceid =
5525                                     dsp->nfsclds_sess.nfsess_sequenceid;
5526                         NFSUNLOCKMNT(nmp);
5527                         error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5528                             nrp, sequenceid, 0, nrp->nr_cred, p);
5529                         NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5530                 }
5531         } else {
5532                 NFSFREECRED(nrp->nr_cred);
5533                 NFSFREEMUTEX(&nrp->nr_mtx);
5534                 free(nrp->nr_nam, M_SONAME);
5535                 free(nrp, M_NFSSOCKREQ);
5536         }
5537         if (error == 0) {
5538                 NFSCL_DEBUG(3, "add DS session\n");
5539                 /*
5540                  * Put it at the end of the list. That way the list
5541                  * is ordered by when the entry was added. This matters
5542                  * since the one done first is the one that should be
5543                  * used for sequencid'ing any subsequent create sessions.
5544                  */
5545                 NFSLOCKMNT(nmp);
5546                 TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5547                 NFSUNLOCKMNT(nmp);
5548                 *dspp = dsp;
5549         } else if (dsp != NULL) {
5550                 newnfs_disconnect(nrp);
5551                 nfscl_freenfsclds(dsp);
5552         }
5553         return (error);
5554 }
5555
5556 /*
5557  * Do the NFSv4.1 Reclaim Complete.
5558  */
5559 int
5560 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5561 {
5562         uint32_t *tl;
5563         struct nfsrv_descript nfsd;
5564         struct nfsrv_descript *nd = &nfsd;
5565         int error;
5566
5567         nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5568             0);
5569         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5570         *tl = newnfs_false;
5571         nd->nd_flag |= ND_USEGSSNAME;
5572         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5573             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5574         if (error != 0)
5575                 return (error);
5576         error = nd->nd_repstat;
5577         mbuf_freem(nd->nd_mrep);
5578         return (error);
5579 }
5580
5581 /*
5582  * Initialize the slot tables for a session.
5583  */
5584 static void
5585 nfscl_initsessionslots(struct nfsclsession *sep)
5586 {
5587         int i;
5588
5589         for (i = 0; i < NFSV4_CBSLOTS; i++) {
5590                 if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5591                         m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5592                 NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5593         }
5594         for (i = 0; i < 64; i++)
5595                 sep->nfsess_slotseq[i] = 0;
5596         sep->nfsess_slots = 0;
5597 }
5598
5599 /*
5600  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5601  */
5602 int
5603 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5604     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5605 {
5606         struct nfsnode *np = VTONFS(vp);
5607         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5608         struct nfscllayout *layp;
5609         struct nfscldevinfo *dip;
5610         struct nfsclflayout *rflp;
5611         struct mbuf *m;
5612         struct nfsclwritedsdorpc *drpc, *tdrpc;
5613         nfsv4stateid_t stateid;
5614         struct ucred *newcred;
5615         uint64_t lastbyte, len, off, oresid, xfer;
5616         int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5617         void *lckp;
5618         uint8_t *dev;
5619         void *iovbase;
5620         size_t iovlen;
5621         off_t offs;
5622         ssize_t resid;
5623
5624         if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5625             (np->n_flag & NNOLAYOUT) != 0)
5626                 return (EIO);
5627         /* Now, get a reference cnt on the clientid for this mount. */
5628         if (nfscl_getref(nmp) == 0)
5629                 return (EIO);
5630
5631         /* Find an appropriate stateid. */
5632         newcred = NFSNEWCRED(cred);
5633         error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5634             rwaccess, 1, newcred, p, &stateid, &lckp);
5635         if (error != 0) {
5636                 NFSFREECRED(newcred);
5637                 nfscl_relref(nmp);
5638                 return (error);
5639         }
5640         /* Search for a layout for this file. */
5641         off = uiop->uio_offset;
5642         layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5643             np->n_fhp->nfh_len, off, &rflp, &recalled);
5644         if (layp == NULL || rflp == NULL) {
5645                 if (recalled != 0) {
5646                         NFSFREECRED(newcred);
5647                         nfscl_relref(nmp);
5648                         return (EIO);
5649                 }
5650                 if (layp != NULL) {
5651                         nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5652                         layp = NULL;
5653                 }
5654                 /* Try and get a Layout, if it is supported. */
5655                 if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5656                     (np->n_flag & NWRITEOPENED) != 0)
5657                         iolaymode = NFSLAYOUTIOMODE_RW;
5658                 else
5659                         iolaymode = NFSLAYOUTIOMODE_READ;
5660                 error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5661                     NULL, &stateid, off, &layp, newcred, p);
5662                 if (error != 0) {
5663                         NFSLOCKNODE(np);
5664                         np->n_flag |= NNOLAYOUT;
5665                         NFSUNLOCKNODE(np);
5666                         if (lckp != NULL)
5667                                 nfscl_lockderef(lckp);
5668                         NFSFREECRED(newcred);
5669                         if (layp != NULL)
5670                                 nfscl_rellayout(layp, 0);
5671                         nfscl_relref(nmp);
5672                         return (error);
5673                 }
5674         }
5675
5676         /*
5677          * Loop around finding a layout that works for the first part of
5678          * this I/O operation, and then call the function that actually
5679          * does the RPC.
5680          */
5681         eof = 0;
5682         len = (uint64_t)uiop->uio_resid;
5683         while (len > 0 && error == 0 && eof == 0) {
5684                 off = uiop->uio_offset;
5685                 error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5686                 if (error == 0) {
5687                         oresid = xfer = (uint64_t)uiop->uio_resid;
5688                         if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5689                                 xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5690                         /*
5691                          * For Flex File layout with mirrored DSs, select one
5692                          * of them at random for reads. For writes and commits,
5693                          * do all mirrors.
5694                          */
5695                         m = NULL;
5696                         drpc = NULL;
5697                         firstmirror = 0;
5698                         mirrorcnt = 1;
5699                         if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5700                             (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5701                                 if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5702                                         firstmirror = arc4random() % mirrorcnt;
5703                                         mirrorcnt = firstmirror + 1;
5704                                 } else {
5705                                         if (docommit == 0) {
5706                                                 /*
5707                                                  * Save values, so uiop can be
5708                                                  * rolled back upon a write
5709                                                  * error.
5710                                                  */
5711                                                 offs = uiop->uio_offset;
5712                                                 resid = uiop->uio_resid;
5713                                                 iovbase =
5714                                                     uiop->uio_iov->iov_base;
5715                                                 iovlen = uiop->uio_iov->iov_len;
5716                                                 m = nfsm_uiombuflist(uiop, len,
5717                                                     NULL, NULL);
5718                                         }
5719                                         tdrpc = drpc = malloc(sizeof(*drpc) *
5720                                             (mirrorcnt - 1), M_TEMP, M_WAITOK |
5721                                             M_ZERO);
5722                                 }
5723                         }
5724                         for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5725                                 if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0)
5726                                         dev = rflp->nfsfl_ffm[i].dev;
5727                                 else
5728                                         dev = rflp->nfsfl_dev;
5729                                 dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5730                                     rflp->nfsfl_devp);
5731                                 if (dip != NULL) {
5732                                         if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5733                                             != 0)
5734                                                 error = nfscl_dofflayoutio(vp,
5735                                                     uiop, iomode, must_commit,
5736                                                     &eof, &stateid, rwaccess,
5737                                                     dip, layp, rflp, off, xfer,
5738                                                     i, docommit, m, tdrpc,
5739                                                     newcred, p);
5740                                         else
5741                                                 error = nfscl_doflayoutio(vp,
5742                                                     uiop, iomode, must_commit,
5743                                                     &eof, &stateid, rwaccess,
5744                                                     dip, layp, rflp, off, xfer,
5745                                                     docommit, newcred, p);
5746                                         nfscl_reldevinfo(dip);
5747                                 } else
5748                                         error = EIO;
5749                                 tdrpc++;
5750                         }
5751                         if (m != NULL)
5752                                 m_freem(m);
5753                         tdrpc = drpc;
5754                         timo = hz / 50;         /* Wait for 20msec. */
5755                         if (timo < 1)
5756                                 timo = 1;
5757                         for (i = firstmirror; i < mirrorcnt - 1 &&
5758                             tdrpc != NULL; i++, tdrpc++) {
5759                                 /*
5760                                  * For the unused drpc entries, both inprog and
5761                                  * err == 0, so this loop won't break.
5762                                  */
5763                                 while (tdrpc->inprog != 0 && tdrpc->done == 0)
5764                                         tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5765                                             timo);
5766                                 if (error == 0 && tdrpc->err != 0)
5767                                         error = tdrpc->err;
5768                         }
5769                         free(drpc, M_TEMP);
5770                         if (error == 0) {
5771                                 if (mirrorcnt > 1 && rwaccess ==
5772                                     NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5773                                         NFSLOCKCLSTATE();
5774                                         layp->nfsly_flags |= NFSLY_WRITTEN;
5775                                         NFSUNLOCKCLSTATE();
5776                                 }
5777                                 lastbyte = off + xfer - 1;
5778                                 NFSLOCKCLSTATE();
5779                                 if (lastbyte > layp->nfsly_lastbyte)
5780                                         layp->nfsly_lastbyte = lastbyte;
5781                                 NFSUNLOCKCLSTATE();
5782                         } else if (error == NFSERR_OPENMODE &&
5783                             rwaccess == NFSV4OPEN_ACCESSREAD) {
5784                                 NFSLOCKMNT(nmp);
5785                                 nmp->nm_state |= NFSSTA_OPENMODE;
5786                                 NFSUNLOCKMNT(nmp);
5787                         } else
5788                                 error = EIO;
5789                         if (error == 0)
5790                                 len -= (oresid - (uint64_t)uiop->uio_resid);
5791                         else if (mirrorcnt > 1 && rwaccess ==
5792                             NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5793                                 /*
5794                                  * In case the rpc gets retried, roll the
5795                                  * uio fields changed by nfsm_uiombuflist()
5796                                  * back.
5797                                  */
5798                                 uiop->uio_offset = offs;
5799                                 uiop->uio_resid = resid;
5800                                 uiop->uio_iov->iov_base = iovbase;
5801                                 uiop->uio_iov->iov_len = iovlen;
5802                         }
5803                 }
5804         }
5805         if (lckp != NULL)
5806                 nfscl_lockderef(lckp);
5807         NFSFREECRED(newcred);
5808         nfscl_rellayout(layp, 0);
5809         nfscl_relref(nmp);
5810         return (error);
5811 }
5812
5813 /*
5814  * Make a copy of the mbuf chain and add an mbuf for null padding, as required.
5815  */
5816 static struct mbuf *
5817 nfsm_copym(struct mbuf *m, int off, int xfer)
5818 {
5819         struct mbuf *m2, *m3, *m4;
5820         uint32_t *tl;
5821         int rem;
5822
5823         m2 = m_copym(m, off, xfer, M_WAITOK);
5824         rem = NFSM_RNDUP(xfer) - xfer;
5825         if (rem > 0) {
5826                 /*
5827                  * The zero padding to a multiple of 4 bytes is required by
5828                  * the XDR. So that the mbufs copied by reference aren't
5829                  * modified, add an mbuf with the zero'd bytes to the list.
5830                  * rem will be a maximum of 3, so one zero'd uint32_t is
5831                  * sufficient.
5832                  */
5833                 m3 = m2;
5834                 while (m3->m_next != NULL)
5835                         m3 = m3->m_next;
5836                 NFSMGET(m4);
5837                 tl = NFSMTOD(m4, uint32_t *);
5838                 *tl = 0;
5839                 mbuf_setlen(m4, rem);
5840                 mbuf_setnext(m3, m4);
5841         }
5842         return (m2);
5843 }
5844
5845 /*
5846  * Find a file layout that will handle the first bytes of the requested
5847  * range and return the information from it needed to the I/O operation.
5848  */
5849 int
5850 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5851     struct nfsclflayout **retflpp)
5852 {
5853         struct nfsclflayout *flp, *nflp, *rflp;
5854         uint32_t rw;
5855
5856         rflp = NULL;
5857         rw = rwaccess;
5858         /* For reading, do the Read list first and then the Write list. */
5859         do {
5860                 if (rw == NFSV4OPEN_ACCESSREAD)
5861                         flp = LIST_FIRST(&lyp->nfsly_flayread);
5862                 else
5863                         flp = LIST_FIRST(&lyp->nfsly_flayrw);
5864                 while (flp != NULL) {
5865                         nflp = LIST_NEXT(flp, nfsfl_list);
5866                         if (flp->nfsfl_off > off)
5867                                 break;
5868                         if (flp->nfsfl_end > off &&
5869                             (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5870                                 rflp = flp;
5871                         flp = nflp;
5872                 }
5873                 if (rw == NFSV4OPEN_ACCESSREAD)
5874                         rw = NFSV4OPEN_ACCESSWRITE;
5875                 else
5876                         rw = 0;
5877         } while (rw != 0);
5878         if (rflp != NULL) {
5879                 /* This one covers the most bytes starting at off. */
5880                 *retflpp = rflp;
5881                 return (0);
5882         }
5883         return (EIO);
5884 }
5885
5886 /*
5887  * Do I/O using an NFSv4.1 file layout.
5888  */
5889 static int
5890 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5891     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5892     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5893     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
5894 {
5895         uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
5896         int commit_thru_mds, error, stripe_index, stripe_pos;
5897         struct nfsnode *np;
5898         struct nfsfh *fhp;
5899         struct nfsclds **dspp;
5900
5901         np = VTONFS(vp);
5902         rel_off = off - flp->nfsfl_patoff;
5903         stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
5904         stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
5905             dp->nfsdi_stripecnt;
5906         transfer = stripe_unit_size - (rel_off % stripe_unit_size);
5907         error = 0;
5908
5909         /* Loop around, doing I/O for each stripe unit. */
5910         while (len > 0 && error == 0) {
5911                 stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
5912                 dspp = nfsfldi_addr(dp, stripe_index);
5913                 if (len > transfer && docommit == 0)
5914                         xfer = transfer;
5915                 else
5916                         xfer = len;
5917                 if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
5918                         /* Dense layout. */
5919                         if (stripe_pos >= flp->nfsfl_fhcnt)
5920                                 return (EIO);
5921                         fhp = flp->nfsfl_fh[stripe_pos];
5922                         io_off = (rel_off / (stripe_unit_size *
5923                             dp->nfsdi_stripecnt)) * stripe_unit_size +
5924                             rel_off % stripe_unit_size;
5925                 } else {
5926                         /* Sparse layout. */
5927                         if (flp->nfsfl_fhcnt > 1) {
5928                                 if (stripe_index >= flp->nfsfl_fhcnt)
5929                                         return (EIO);
5930                                 fhp = flp->nfsfl_fh[stripe_index];
5931                         } else if (flp->nfsfl_fhcnt == 1)
5932                                 fhp = flp->nfsfl_fh[0];
5933                         else
5934                                 fhp = np->n_fhp;
5935                         io_off = off;
5936                 }
5937                 if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
5938                         commit_thru_mds = 1;
5939                         if (docommit != 0)
5940                                 error = EIO;
5941                 } else {
5942                         commit_thru_mds = 0;
5943                         mtx_lock(&np->n_mtx);
5944                         np->n_flag |= NDSCOMMIT;
5945                         mtx_unlock(&np->n_mtx);
5946                 }
5947                 if (docommit != 0) {
5948                         if (error == 0)
5949                                 error = nfsrpc_commitds(vp, io_off, xfer,
5950                                     *dspp, fhp, 0, 0, cred, p);
5951                         if (error == 0) {
5952                                 /*
5953                                  * Set both eof and uio_resid = 0 to end any
5954                                  * loops.
5955                                  */
5956                                 *eofp = 1;
5957                                 uiop->uio_resid = 0;
5958                         } else {
5959                                 mtx_lock(&np->n_mtx);
5960                                 np->n_flag &= ~NDSCOMMIT;
5961                                 mtx_unlock(&np->n_mtx);
5962                         }
5963                 } else if (rwflag == NFSV4OPEN_ACCESSREAD)
5964                         error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
5965                             io_off, xfer, fhp, 0, 0, 0, cred, p);
5966                 else {
5967                         error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
5968                             stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
5969                             0, 0, 0, cred, p);
5970                         if (error == 0) {
5971                                 NFSLOCKCLSTATE();
5972                                 lyp->nfsly_flags |= NFSLY_WRITTEN;
5973                                 NFSUNLOCKCLSTATE();
5974                         }
5975                 }
5976                 if (error == 0) {
5977                         transfer = stripe_unit_size;
5978                         stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
5979                         len -= xfer;
5980                         off += xfer;
5981                 }
5982         }
5983         return (error);
5984 }
5985
5986 /*
5987  * Do I/O using an NFSv4.1 flex file layout.
5988  */
5989 static int
5990 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5991     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5992     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5993     uint64_t len, int mirror, int docommit, struct mbuf *mp,
5994     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
5995 {
5996         uint64_t transfer, xfer;
5997         int error, rel_off;
5998         struct nfsnode *np;
5999         struct nfsfh *fhp;
6000         struct nfsclds **dspp;
6001         struct ucred *tcred;
6002         struct mbuf *m;
6003
6004         np = VTONFS(vp);
6005         error = 0;
6006         rel_off = 0;
6007         NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6008             (uintmax_t)len);
6009         /* Loop around, doing I/O for each stripe unit. */
6010         while (len > 0 && error == 0) {
6011                 dspp = nfsfldi_addr(dp, 0);
6012                 fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6013                 stateidp = &flp->nfsfl_ffm[mirror].st;
6014                 NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6015                     mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6016                 if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6017                         tcred = NFSNEWCRED(cred);
6018                         tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6019                         tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6020                         tcred->cr_ngroups = 1;
6021                 } else
6022                         tcred = cred;
6023                 if (rwflag == NFSV4OPEN_ACCESSREAD)
6024                         transfer = dp->nfsdi_rsize;
6025                 else
6026                         transfer = dp->nfsdi_wsize;
6027                 mtx_lock(&np->n_mtx);
6028                 np->n_flag |= NDSCOMMIT;
6029                 mtx_unlock(&np->n_mtx);
6030                 if (len > transfer && docommit == 0)
6031                         xfer = transfer;
6032                 else
6033                         xfer = len;
6034                 if (docommit != 0) {
6035                         if (error == 0) {
6036                                 /*
6037                                  * Do last mirrored DS commit with this thread.
6038                                  */
6039                                 if (mirror < flp->nfsfl_mirrorcnt - 1)
6040                                         error = nfsio_commitds(vp, off, xfer,
6041                                             *dspp, fhp, dp->nfsdi_vers,
6042                                             dp->nfsdi_minorvers, drpc, tcred,
6043                                             p);
6044                                 else
6045                                         error = nfsrpc_commitds(vp, off, xfer,
6046                                             *dspp, fhp, dp->nfsdi_vers,
6047                                             dp->nfsdi_minorvers, tcred, p);
6048                                 NFSCL_DEBUG(4, "commitds=%d\n", error);
6049                                 if (error != 0 && error != EACCES) {
6050                                         NFSCL_DEBUG(4,
6051                                             "DS layreterr for commit\n");
6052                                         nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6053                                             lyp, *dspp);
6054                                 }
6055                         }
6056                         NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6057                         if (error == 0) {
6058                                 /*
6059                                  * Set both eof and uio_resid = 0 to end any
6060                                  * loops.
6061                                  */
6062                                 *eofp = 1;
6063                                 uiop->uio_resid = 0;
6064                         } else {
6065                                 mtx_lock(&np->n_mtx);
6066                                 np->n_flag &= ~NDSCOMMIT;
6067                                 mtx_unlock(&np->n_mtx);
6068                         }
6069                 } else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6070                         error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6071                             off, xfer, fhp, 1, dp->nfsdi_vers,
6072                             dp->nfsdi_minorvers, tcred, p);
6073                         NFSCL_DEBUG(4, "readds=%d\n", error);
6074                         if (error != 0 && error != EACCES) {
6075                                 NFSCL_DEBUG(4, "DS layreterr for read\n");
6076                                 nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6077                                     *dspp);
6078                         }
6079                 } else {
6080                         if (flp->nfsfl_mirrorcnt == 1) {
6081                                 error = nfsrpc_writeds(vp, uiop, iomode,
6082                                     must_commit, stateidp, *dspp, off, xfer,
6083                                     fhp, 0, 1, dp->nfsdi_vers,
6084                                     dp->nfsdi_minorvers, tcred, p);
6085                                 if (error == 0) {
6086                                         NFSLOCKCLSTATE();
6087                                         lyp->nfsly_flags |= NFSLY_WRITTEN;
6088                                         NFSUNLOCKCLSTATE();
6089                                 }
6090                         } else {
6091                                 m = nfsm_copym(mp, rel_off, xfer);
6092                                 NFSCL_DEBUG(4, "mcopy reloff=%d xfer=%jd\n",
6093                                     rel_off, (uintmax_t)xfer);
6094                                 /*
6095                                  * Do last write to a mirrored DS with this
6096                                  * thread.
6097                                  */
6098                                 if (mirror < flp->nfsfl_mirrorcnt - 1)
6099                                         error = nfsio_writedsmir(vp, iomode,
6100                                             must_commit, stateidp, *dspp, off,
6101                                             xfer, fhp, m, dp->nfsdi_vers,
6102                                             dp->nfsdi_minorvers, drpc, tcred,
6103                                             p);
6104                                 else
6105                                         error = nfsrpc_writedsmir(vp, iomode,
6106                                             must_commit, stateidp, *dspp, off,
6107                                             xfer, fhp, m, dp->nfsdi_vers,
6108                                             dp->nfsdi_minorvers, tcred, p);
6109                                 NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6110                                 if (error != 0 && error != EACCES) {
6111                                         NFSCL_DEBUG(4,
6112                                             "DS layreterr for write\n");
6113                                         nfscl_dserr(NFSV4OP_WRITE, error, dp,
6114                                             lyp, *dspp);
6115                                 }
6116                         }
6117                 }
6118                 NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6119                 if (error == 0) {
6120                         len -= xfer;
6121                         off += xfer;
6122                         rel_off += xfer;
6123                 }
6124                 if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6125                         NFSFREECRED(tcred);
6126         }
6127         NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6128         return (error);
6129 }
6130
6131 /*
6132  * The actual read RPC done to a DS.
6133  */
6134 static int
6135 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6136     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6137     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6138 {
6139         uint32_t *tl;
6140         int attrflag, error, retlen;
6141         struct nfsrv_descript nfsd;
6142         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6143         struct nfsrv_descript *nd = &nfsd;
6144         struct nfssockreq *nrp;
6145         struct nfsvattr na;
6146
6147         nd->nd_mrep = NULL;
6148         if (vers == 0 || vers == NFS_VER4) {
6149                 nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6150                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6151                 vers = NFS_VER4;
6152                 NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6153                 if (flex != 0)
6154                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6155                 else
6156                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6157         } else {
6158                 nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6159                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6160                 NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6161         }
6162         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6163         txdr_hyper(io_off, tl);
6164         *(tl + 2) = txdr_unsigned(len);
6165         nrp = dsp->nfsclds_sockp;
6166         NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6167         if (nrp == NULL)
6168                 /* If NULL, use the MDS socket. */
6169                 nrp = &nmp->nm_sockreq;
6170         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6171             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6172         NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6173             error);
6174         if (error != 0)
6175                 return (error);
6176         if (vers == NFS_VER3) {
6177                 error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6178                 NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6179                 if (error != 0)
6180                         goto nfsmout;
6181         }
6182         if (nd->nd_repstat != 0) {
6183                 error = nd->nd_repstat;
6184                 goto nfsmout;
6185         }
6186         if (vers == NFS_VER3) {
6187                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6188                 *eofp = fxdr_unsigned(int, *(tl + 1));
6189         } else {
6190                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6191                 *eofp = fxdr_unsigned(int, *tl);
6192         }
6193         NFSM_STRSIZ(retlen, len);
6194         NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6195         error = nfsm_mbufuio(nd, uiop, retlen);
6196 nfsmout:
6197         if (nd->nd_mrep != NULL)
6198                 mbuf_freem(nd->nd_mrep);
6199         return (error);
6200 }
6201
6202 /*
6203  * The actual write RPC done to a DS.
6204  */
6205 static int
6206 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6207     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6208     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6209     struct ucred *cred, NFSPROC_T *p)
6210 {
6211         uint32_t *tl;
6212         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6213         int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6214         int32_t backup;
6215         struct nfsrv_descript nfsd;
6216         struct nfsrv_descript *nd = &nfsd;
6217         struct nfssockreq *nrp;
6218         struct nfsvattr na;
6219
6220         KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6221         nd->nd_mrep = NULL;
6222         if (vers == 0 || vers == NFS_VER4) {
6223                 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6224                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6225                 NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6226                 vers = NFS_VER4;
6227                 if (flex != 0)
6228                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6229                 else
6230                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6231                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6232         } else {
6233                 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6234                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6235                 NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6236                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6237         }
6238         txdr_hyper(io_off, tl);
6239         tl += 2;
6240         if (vers == NFS_VER3)
6241                 *tl++ = txdr_unsigned(len);
6242         *tl++ = txdr_unsigned(*iomode);
6243         *tl = txdr_unsigned(len);
6244         nfsm_uiombuf(nd, uiop, len);
6245         nrp = dsp->nfsclds_sockp;
6246         if (nrp == NULL)
6247                 /* If NULL, use the MDS socket. */
6248                 nrp = &nmp->nm_sockreq;
6249         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6250             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6251         NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6252             nd->nd_repstat);
6253         if (error != 0)
6254                 return (error);
6255         if (nd->nd_repstat != 0) {
6256                 /*
6257                  * In case the rpc gets retried, roll
6258                  * the uio fileds changed by nfsm_uiombuf()
6259                  * back.
6260                  */
6261                 uiop->uio_offset -= len;
6262                 uio_uio_resid_add(uiop, len);
6263                 uio_iov_base_add(uiop, -len);
6264                 uio_iov_len_add(uiop, len);
6265                 error = nd->nd_repstat;
6266         } else {
6267                 if (vers == NFS_VER3) {
6268                         error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6269                             NULL);
6270                         NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6271                         if (error != 0)
6272                                 goto nfsmout;
6273                 }
6274                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6275                 rlen = fxdr_unsigned(int, *tl++);
6276                 NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6277                 if (rlen == 0) {
6278                         error = NFSERR_IO;
6279                         goto nfsmout;
6280                 } else if (rlen < len) {
6281                         backup = len - rlen;
6282                         uio_iov_base_add(uiop, -(backup));
6283                         uio_iov_len_add(uiop, backup);
6284                         uiop->uio_offset -= backup;
6285                         uio_uio_resid_add(uiop, backup);
6286                         len = rlen;
6287                 }
6288                 commit = fxdr_unsigned(int, *tl++);
6289
6290                 /*
6291                  * Return the lowest commitment level
6292                  * obtained by any of the RPCs.
6293                  */
6294                 if (committed == NFSWRITE_FILESYNC)
6295                         committed = commit;
6296                 else if (committed == NFSWRITE_DATASYNC &&
6297                     commit == NFSWRITE_UNSTABLE)
6298                         committed = commit;
6299                 if (commit_thru_mds != 0) {
6300                         NFSLOCKMNT(nmp);
6301                         if (!NFSHASWRITEVERF(nmp)) {
6302                                 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6303                                 NFSSETWRITEVERF(nmp);
6304                         } else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6305                                 *must_commit = 1;
6306                                 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6307                         }
6308                         NFSUNLOCKMNT(nmp);
6309                 } else {
6310                         NFSLOCKDS(dsp);
6311                         if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6312                                 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6313                                 dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6314                         } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6315                                 *must_commit = 1;
6316                                 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6317                         }
6318                         NFSUNLOCKDS(dsp);
6319                 }
6320         }
6321 nfsmout:
6322         if (nd->nd_mrep != NULL)
6323                 mbuf_freem(nd->nd_mrep);
6324         *iomode = committed;
6325         if (nd->nd_repstat != 0 && error == 0)
6326                 error = nd->nd_repstat;
6327         return (error);
6328 }
6329
6330 /*
6331  * The actual write RPC done to a DS.
6332  * This variant is called from a separate kernel process for mirrors.
6333  * Any short write is considered an IO error.
6334  */
6335 static int
6336 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6337     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6338     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6339     struct ucred *cred, NFSPROC_T *p)
6340 {
6341         uint32_t *tl;
6342         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6343         int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6344         struct nfsrv_descript nfsd;
6345         struct nfsrv_descript *nd = &nfsd;
6346         struct nfssockreq *nrp;
6347         struct nfsvattr na;
6348
6349         nd->nd_mrep = NULL;
6350         if (vers == 0 || vers == NFS_VER4) {
6351                 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6352                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6353                 vers = NFS_VER4;
6354                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6355                     minorvers);
6356                 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6357                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6358         } else {
6359                 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6360                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6361                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6362                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6363         }
6364         txdr_hyper(io_off, tl);
6365         tl += 2;
6366         if (vers == NFS_VER3)
6367                 *tl++ = txdr_unsigned(len);
6368         *tl++ = txdr_unsigned(*iomode);
6369         *tl = txdr_unsigned(len);
6370         if (len > 0) {
6371                 /* Put data in mbuf chain. */
6372                 nd->nd_mb->m_next = m;
6373                 /* Set nd_mb and nd_bpos to end of data. */
6374                 while (m->m_next != NULL)
6375                         m = m->m_next;
6376                 nd->nd_mb = m;
6377                 nd->nd_bpos = mtod(m, char *) + m->m_len;
6378                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: lastmb len=%d\n", m->m_len);
6379         }
6380         nrp = dsp->nfsclds_sockp;
6381         if (nrp == NULL)
6382                 /* If NULL, use the MDS socket. */
6383                 nrp = &nmp->nm_sockreq;
6384         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6385             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6386         NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6387             nd->nd_repstat);
6388         if (error != 0)
6389                 return (error);
6390         if (nd->nd_repstat != 0)
6391                 error = nd->nd_repstat;
6392         else {
6393                 if (vers == NFS_VER3) {
6394                         error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6395                             NULL);
6396                         NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6397                             error);
6398                         if (error != 0)
6399                                 goto nfsmout;
6400                 }
6401                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6402                 rlen = fxdr_unsigned(int, *tl++);
6403                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6404                     rlen);
6405                 if (rlen != len) {
6406                         error = NFSERR_IO;
6407                         NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6408                             len, rlen);
6409                         goto nfsmout;
6410                 }
6411                 commit = fxdr_unsigned(int, *tl++);
6412
6413                 /*
6414                  * Return the lowest commitment level
6415                  * obtained by any of the RPCs.
6416                  */
6417                 if (committed == NFSWRITE_FILESYNC)
6418                         committed = commit;
6419                 else if (committed == NFSWRITE_DATASYNC &&
6420                     commit == NFSWRITE_UNSTABLE)
6421                         committed = commit;
6422                 NFSLOCKDS(dsp);
6423                 if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6424                         NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6425                         dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6426                 } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6427                         *must_commit = 1;
6428                         NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6429                 }
6430                 NFSUNLOCKDS(dsp);
6431         }
6432 nfsmout:
6433         if (nd->nd_mrep != NULL)
6434                 mbuf_freem(nd->nd_mrep);
6435         *iomode = committed;
6436         if (nd->nd_repstat != 0 && error == 0)
6437                 error = nd->nd_repstat;
6438         return (error);
6439 }
6440
6441 /*
6442  * Start up the thread that will execute nfsrpc_writedsmir().
6443  */
6444 static void
6445 start_writedsmir(void *arg, int pending)
6446 {
6447         struct nfsclwritedsdorpc *drpc;
6448
6449         drpc = (struct nfsclwritedsdorpc *)arg;
6450         drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6451             &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6452             drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6453             drpc->p);
6454         drpc->done = 1;
6455         NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6456 }
6457
6458 /*
6459  * Set up the write DS mirror call for the pNFS I/O thread.
6460  */
6461 static int
6462 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6463     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6464     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6465     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6466 {
6467         int error, ret;
6468
6469         error = 0;
6470         drpc->done = 0;
6471         drpc->vp = vp;
6472         drpc->iomode = *iomode;
6473         drpc->must_commit = *must_commit;
6474         drpc->stateidp = stateidp;
6475         drpc->dsp = dsp;
6476         drpc->off = off;
6477         drpc->len = len;
6478         drpc->fhp = fhp;
6479         drpc->m = m;
6480         drpc->vers = vers;
6481         drpc->minorvers = minorvers;
6482         drpc->cred = cred;
6483         drpc->p = p;
6484         drpc->inprog = 0;
6485         ret = EIO;
6486         if (nfs_pnfsiothreads != 0) {
6487                 ret = nfs_pnfsio(start_writedsmir, drpc);
6488                 NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6489         }
6490         if (ret != 0)
6491                 error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6492                     dsp, off, len, fhp, m, vers, minorvers, cred, p);
6493         NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6494         return (error);
6495 }
6496
6497 /*
6498  * Free up the nfsclds structure.
6499  */
6500 void
6501 nfscl_freenfsclds(struct nfsclds *dsp)
6502 {
6503         int i;
6504
6505         if (dsp == NULL)
6506                 return;
6507         if (dsp->nfsclds_sockp != NULL) {
6508                 NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6509                 NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6510                 free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6511                 free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6512         }
6513         NFSFREEMUTEX(&dsp->nfsclds_mtx);
6514         NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6515         for (i = 0; i < NFSV4_CBSLOTS; i++) {
6516                 if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6517                         m_freem(
6518                             dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6519         }
6520         free(dsp, M_NFSCLDS);
6521 }
6522
6523 static enum nfsclds_state
6524 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6525     struct nfsclds **retdspp, uint32_t *sequencep)
6526 {
6527         struct nfsclds *dsp;
6528         int fndseq;
6529
6530         /*
6531          * Search the list of nfsclds structures for one with the same
6532          * server.
6533          */
6534         fndseq = 0;
6535         TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6536                 if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6537                     dsp->nfsclds_servownlen != 0 &&
6538                     !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6539                     dsp->nfsclds_servownlen) &&
6540                     dsp->nfsclds_sess.nfsess_defunct == 0) {
6541                         NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6542                             TAILQ_FIRST(&nmp->nm_sess), dsp,
6543                             dsp->nfsclds_flags);
6544                         if (fndseq == 0) {
6545                                 /* Get sequenceid# from first entry. */
6546                                 *sequencep =
6547                                     dsp->nfsclds_sess.nfsess_sequenceid;
6548                                 fndseq = 1;
6549                         }
6550                         /* Server major id matches. */
6551                         if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6552                                 *retdspp = dsp;
6553                                 return (NFSDSP_USETHISSESSION);
6554                         }
6555
6556                 }
6557         }
6558         if (fndseq != 0)
6559                 return (NFSDSP_SEQTHISSESSION);
6560         return (NFSDSP_NOTFOUND);
6561 }
6562
6563 /*
6564  * NFS commit rpc to a NFSv4.1 DS.
6565  */
6566 static int
6567 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6568     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6569     NFSPROC_T *p)
6570 {
6571         uint32_t *tl;
6572         struct nfsrv_descript nfsd, *nd = &nfsd;
6573         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6574         struct nfssockreq *nrp;
6575         struct nfsvattr na;
6576         int attrflag, error;
6577
6578         nd->nd_mrep = NULL;
6579         if (vers == 0 || vers == NFS_VER4) {
6580                 nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6581                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6582                 vers = NFS_VER4;
6583         } else
6584                 nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6585                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6586         NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6587             minorvers);
6588         NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6589         txdr_hyper(offset, tl);
6590         tl += 2;
6591         *tl = txdr_unsigned(cnt);
6592         nrp = dsp->nfsclds_sockp;
6593         if (nrp == NULL)
6594                 /* If NULL, use the MDS socket. */
6595                 nrp = &nmp->nm_sockreq;
6596         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6597             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6598         NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6599             nd->nd_repstat);
6600         if (error != 0)
6601                 return (error);
6602         if (nd->nd_repstat == 0) {
6603                 if (vers == NFS_VER3) {
6604                         error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6605                             NULL);
6606                         NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6607                         if (error != 0)
6608                                 goto nfsmout;
6609                 }
6610                 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6611                 NFSLOCKDS(dsp);
6612                 if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6613                         NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6614                         error = NFSERR_STALEWRITEVERF;
6615                 }
6616                 NFSUNLOCKDS(dsp);
6617         }
6618 nfsmout:
6619         if (error == 0 && nd->nd_repstat != 0)
6620                 error = nd->nd_repstat;
6621         mbuf_freem(nd->nd_mrep);
6622         return (error);
6623 }
6624
6625 /*
6626  * Start up the thread that will execute nfsrpc_commitds().
6627  */
6628 static void
6629 start_commitds(void *arg, int pending)
6630 {
6631         struct nfsclwritedsdorpc *drpc;
6632
6633         drpc = (struct nfsclwritedsdorpc *)arg;
6634         drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6635             drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6636             drpc->p);
6637         drpc->done = 1;
6638         NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6639 }
6640
6641 /*
6642  * Set up the commit DS mirror call for the pNFS I/O thread.
6643  */
6644 static int
6645 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6646     struct nfsfh *fhp, int vers, int minorvers,
6647     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6648 {
6649         int error, ret;
6650
6651         error = 0;
6652         drpc->done = 0;
6653         drpc->vp = vp;
6654         drpc->off = offset;
6655         drpc->len = cnt;
6656         drpc->dsp = dsp;
6657         drpc->fhp = fhp;
6658         drpc->vers = vers;
6659         drpc->minorvers = minorvers;
6660         drpc->cred = cred;
6661         drpc->p = p;
6662         drpc->inprog = 0;
6663         ret = EIO;
6664         if (nfs_pnfsiothreads != 0) {
6665                 ret = nfs_pnfsio(start_commitds, drpc);
6666                 NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6667         }
6668         if (ret != 0)
6669                 error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6670                     minorvers, cred, p);
6671         NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6672         return (error);
6673 }
6674
6675 /*
6676  * Set up the XDR arguments for the LayoutGet operation.
6677  */
6678 static void
6679 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
6680     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
6681     int layoutlen, int usecurstateid)
6682 {
6683         uint32_t *tl;
6684
6685         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
6686             NFSX_STATEID);
6687         *tl++ = newnfs_false;           /* Don't signal availability. */
6688         *tl++ = txdr_unsigned(layouttype);
6689         *tl++ = txdr_unsigned(iomode);
6690         txdr_hyper(offset, tl);
6691         tl += 2;
6692         txdr_hyper(len, tl);
6693         tl += 2;
6694         txdr_hyper(minlen, tl);
6695         tl += 2;
6696         if (usecurstateid != 0) {
6697                 /* Special stateid for Current stateid. */
6698                 *tl++ = txdr_unsigned(1);
6699                 *tl++ = 0;
6700                 *tl++ = 0;
6701                 *tl++ = 0;
6702         } else {
6703                 *tl++ = txdr_unsigned(stateidp->seqid);
6704                 NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
6705                 *tl++ = stateidp->other[0];
6706                 *tl++ = stateidp->other[1];
6707                 *tl++ = stateidp->other[2];
6708         }
6709         *tl = txdr_unsigned(layoutlen);
6710 }
6711
6712 /*
6713  * Parse the reply for a successful LayoutGet operation.
6714  */
6715 static int
6716 nfsrv_parselayoutget(struct nfsrv_descript *nd, nfsv4stateid_t *stateidp,
6717     int *retonclosep, struct nfsclflayouthead *flhp)
6718 {
6719         uint32_t *tl;
6720         struct nfsclflayout *flp, *prevflp, *tflp;
6721         int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
6722         int m, mirrorcnt;
6723         uint64_t retlen, off;
6724         struct nfsfh *nfhp;
6725         uint8_t *cp;
6726         uid_t user;
6727         gid_t grp;
6728
6729         NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
6730         error = 0;
6731         flp = NULL;
6732         gotiomode = -1;
6733         NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
6734         if (*tl++ != 0)
6735                 *retonclosep = 1;
6736         else
6737                 *retonclosep = 0;
6738         stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
6739         NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
6740             (int)stateidp->seqid);
6741         stateidp->other[0] = *tl++;
6742         stateidp->other[1] = *tl++;
6743         stateidp->other[2] = *tl++;
6744         cnt = fxdr_unsigned(int, *tl);
6745         NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
6746         if (cnt <= 0 || cnt > 10000) {
6747                 /* Don't accept more than 10000 layouts in reply. */
6748                 error = NFSERR_BADXDR;
6749                 goto nfsmout;
6750         }
6751         for (i = 0; i < cnt; i++) {
6752                 /* Dissect to the layout type. */
6753                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
6754                     3 * NFSX_UNSIGNED);
6755                 off = fxdr_hyper(tl); tl += 2;
6756                 retlen = fxdr_hyper(tl); tl += 2;
6757                 iomode = fxdr_unsigned(int, *tl++);
6758                 laytype = fxdr_unsigned(int, *tl);
6759                 NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
6760                     (uintmax_t)off, (uintmax_t)retlen, iomode);
6761                 /* Ignore length of layout body for now. */
6762                 if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
6763                         /* Parse the File layout up to fhcnt. */
6764                         NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
6765                             NFSX_HYPER + NFSX_V4DEVICEID);
6766                         fhcnt = fxdr_unsigned(int, *(tl + 4 +
6767                             NFSX_V4DEVICEID / NFSX_UNSIGNED));
6768                         NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6769                         if (fhcnt < 0 || fhcnt > 100) {
6770                                 /* Don't accept more than 100 file handles. */
6771                                 error = NFSERR_BADXDR;
6772                                 goto nfsmout;
6773                         }
6774                         if (fhcnt > 0)
6775                                 flp = malloc(sizeof(*flp) + fhcnt *
6776                                     sizeof(struct nfsfh *), M_NFSFLAYOUT,
6777                                     M_WAITOK);
6778                         else
6779                                 flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
6780                                     M_WAITOK);
6781                         flp->nfsfl_flags = NFSFL_FILE;
6782                         flp->nfsfl_fhcnt = 0;
6783                         flp->nfsfl_devp = NULL;
6784                         flp->nfsfl_off = off;
6785                         if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6786                                 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6787                         else
6788                                 flp->nfsfl_end = flp->nfsfl_off + retlen;
6789                         flp->nfsfl_iomode = iomode;
6790                         if (gotiomode == -1)
6791                                 gotiomode = flp->nfsfl_iomode;
6792                         /* Ignore layout body length for now. */
6793                         NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
6794                         tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
6795                         flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
6796                         NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
6797                         flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
6798                         flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
6799                         NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
6800                             flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
6801                         for (j = 0; j < fhcnt; j++) {
6802                                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6803                                 nfhlen = fxdr_unsigned(int, *tl);
6804                                 if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
6805                                         error = NFSERR_BADXDR;
6806                                         goto nfsmout;
6807                                 }
6808                                 nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
6809                                     M_NFSFH, M_WAITOK);
6810                                 flp->nfsfl_fh[j] = nfhp;
6811                                 flp->nfsfl_fhcnt++;
6812                                 nfhp->nfh_len = nfhlen;
6813                                 NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
6814                                 NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
6815                         }
6816                 } else if (laytype == NFSLAYOUT_FLEXFILE) {
6817                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
6818                             NFSX_HYPER);
6819                         mirrorcnt = fxdr_unsigned(int, *(tl + 2));
6820                         NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
6821                         if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
6822                                 error = NFSERR_BADXDR;
6823                                 goto nfsmout;
6824                         }
6825                         flp = malloc(sizeof(*flp) + mirrorcnt *
6826                             sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
6827                         flp->nfsfl_flags = NFSFL_FLEXFILE;
6828                         flp->nfsfl_mirrorcnt = mirrorcnt;
6829                         flp->nfsfl_devp = NULL;
6830                         flp->nfsfl_off = off;
6831                         if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6832                                 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6833                         else
6834                                 flp->nfsfl_end = flp->nfsfl_off + retlen;
6835                         flp->nfsfl_iomode = iomode;
6836                         if (gotiomode == -1)
6837                                 gotiomode = flp->nfsfl_iomode;
6838                         flp->nfsfl_stripeunit = fxdr_hyper(tl);
6839                         NFSCL_DEBUG(4, "stripeunit=%ju\n",
6840                             (uintmax_t)flp->nfsfl_stripeunit);
6841                         for (j = 0; j < mirrorcnt; j++) {
6842                                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6843                                 k = fxdr_unsigned(int, *tl);
6844                                 if (k < 1 || k > 128) {
6845                                         error = NFSERR_BADXDR;
6846                                         goto nfsmout;
6847                                 }
6848                                 NFSCL_DEBUG(4, "servercnt=%d\n", k);
6849                                 for (l = 0; l < k; l++) {
6850                                         NFSM_DISSECT(tl, uint32_t *,
6851                                             NFSX_V4DEVICEID + NFSX_STATEID +
6852                                             2 * NFSX_UNSIGNED);
6853                                         if (l == 0) {
6854                                                 /* Just use the first server. */
6855                                                 NFSBCOPY(tl,
6856                                                     flp->nfsfl_ffm[j].dev,
6857                                                     NFSX_V4DEVICEID);
6858                                                 tl += (NFSX_V4DEVICEID /
6859                                                     NFSX_UNSIGNED);
6860                                                 tl++;
6861                                                 flp->nfsfl_ffm[j].st.seqid =
6862                                                     *tl++;
6863                                                 flp->nfsfl_ffm[j].st.other[0] =
6864                                                     *tl++;
6865                                                 flp->nfsfl_ffm[j].st.other[1] =
6866                                                     *tl++;
6867                                                 flp->nfsfl_ffm[j].st.other[2] =
6868                                                     *tl++;
6869                                                 NFSCL_DEBUG(4, "st.seqid=%u "
6870                                                  "st.o0=0x%x st.o1=0x%x "
6871                                                  "st.o2=0x%x\n",
6872                                                  flp->nfsfl_ffm[j].st.seqid,
6873                                                  flp->nfsfl_ffm[j].st.other[0],
6874                                                  flp->nfsfl_ffm[j].st.other[1],
6875                                                  flp->nfsfl_ffm[j].st.other[2]);
6876                                         } else
6877                                                 tl += ((NFSX_V4DEVICEID +
6878                                                     NFSX_STATEID +
6879                                                     NFSX_UNSIGNED) /
6880                                                     NFSX_UNSIGNED);
6881                                         fhcnt = fxdr_unsigned(int, *tl);
6882                                         NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6883                                         if (fhcnt < 1 ||
6884                                             fhcnt > NFSDEV_MAXVERS) {
6885                                                 error = NFSERR_BADXDR;
6886                                                 goto nfsmout;
6887                                         }
6888                                         for (m = 0; m < fhcnt; m++) {
6889                                                 NFSM_DISSECT(tl, uint32_t *,
6890                                                     NFSX_UNSIGNED);
6891                                                 nfhlen = fxdr_unsigned(int,
6892                                                     *tl);
6893                                                 NFSCL_DEBUG(4, "nfhlen=%d\n",
6894                                                     nfhlen);
6895                                                 if (nfhlen <= 0 || nfhlen >
6896                                                     NFSX_V4FHMAX) {
6897                                                         error = NFSERR_BADXDR;
6898                                                         goto nfsmout;
6899                                                 }
6900                                                 NFSM_DISSECT(cp, uint8_t *,
6901                                                     NFSM_RNDUP(nfhlen));
6902                                                 if (l == 0) {
6903                                                         flp->nfsfl_ffm[j].fhcnt
6904                                                             = fhcnt;
6905                                                         nfhp = malloc(
6906                                                             sizeof(*nfhp) +
6907                                                             nfhlen - 1, M_NFSFH,
6908                                                             M_WAITOK);
6909                                                         flp->nfsfl_ffm[j].fh[m]
6910                                                             = nfhp;
6911                                                         nfhp->nfh_len = nfhlen;
6912                                                         NFSBCOPY(cp,
6913                                                             nfhp->nfh_fh,
6914                                                             nfhlen);
6915                                                         NFSCL_DEBUG(4,
6916                                                             "got fh\n");
6917                                                 }
6918                                         }
6919                                         /* Now, get the ffsd_user/ffds_group. */
6920                                         error = nfsrv_parseug(nd, 0, &user,
6921                                             &grp, curthread);
6922                                         NFSCL_DEBUG(4, "after parseu=%d\n",
6923                                             error);
6924                                         if (error == 0)
6925                                                 error = nfsrv_parseug(nd, 1,
6926                                                     &user, &grp, curthread);
6927                                         NFSCL_DEBUG(4, "aft parseg=%d\n",
6928                                             grp);
6929                                         if (error != 0)
6930                                                 goto nfsmout;
6931                                         NFSCL_DEBUG(4, "user=%d group=%d\n",
6932                                             user, grp);
6933                                         if (l == 0) {
6934                                                 flp->nfsfl_ffm[j].user = user;
6935                                                 flp->nfsfl_ffm[j].group = grp;
6936                                                 NFSCL_DEBUG(4,
6937                                                     "usr=%d grp=%d\n", user,
6938                                                     grp);
6939                                         }
6940                                 }
6941                         }
6942                         NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6943                         flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
6944                         flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
6945                         NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
6946                             flp->nfsfl_fflags, flp->nfsfl_statshint);
6947                 } else {
6948                         error = NFSERR_BADXDR;
6949                         goto nfsmout;
6950                 }
6951                 if (flp->nfsfl_iomode == gotiomode) {
6952                         /* Keep the list in increasing offset order. */
6953                         tflp = LIST_FIRST(flhp);
6954                         prevflp = NULL;
6955                         while (tflp != NULL &&
6956                             tflp->nfsfl_off < flp->nfsfl_off) {
6957                                 prevflp = tflp;
6958                                 tflp = LIST_NEXT(tflp, nfsfl_list);
6959                         }
6960                         if (prevflp == NULL)
6961                                 LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
6962                         else
6963                                 LIST_INSERT_AFTER(prevflp, flp,
6964                                     nfsfl_list);
6965                         NFSCL_DEBUG(4, "flp inserted\n");
6966                 } else {
6967                         printf("nfscl_layoutget(): got wrong iomode\n");
6968                         nfscl_freeflayout(flp);
6969                 }
6970                 flp = NULL;
6971         }
6972 nfsmout:
6973         NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
6974         if (error != 0 && flp != NULL)
6975                 nfscl_freeflayout(flp);
6976         return (error);
6977 }
6978
6979 /*
6980  * Parse a user/group digit string.
6981  */
6982 static int
6983 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
6984     NFSPROC_T *p)
6985 {
6986         uint32_t *tl;
6987         char *cp, *str, str0[NFSV4_SMALLSTR + 1];
6988         uint32_t len = 0;
6989         int error = 0;
6990
6991         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6992         len = fxdr_unsigned(uint32_t, *tl);
6993         str = NULL;
6994         if (len > NFSV4_OPAQUELIMIT) {
6995                 error = NFSERR_BADXDR;
6996                 goto nfsmout;
6997         }
6998         NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
6999         if (len == 0) {
7000                 if (dogrp != 0)
7001                         *gidp = GID_NOGROUP;
7002                 else
7003                         *uidp = UID_NOBODY;
7004                 return (0);
7005         }
7006         if (len > NFSV4_SMALLSTR)
7007                 str = malloc(len + 1, M_TEMP, M_WAITOK);
7008         else
7009                 str = str0;
7010         NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7011         NFSBCOPY(cp, str, len);
7012         str[len] = '\0';
7013         NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7014         if (dogrp != 0)
7015                 error = nfsv4_strtogid(nd, str, len, gidp, p);
7016         else
7017                 error = nfsv4_strtouid(nd, str, len, uidp, p);
7018 nfsmout:
7019         if (len > NFSV4_SMALLSTR)
7020                 free(str, M_TEMP);
7021         NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7022         return (error);
7023 }
7024
7025 /*
7026  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7027  * so that it does both an Open and a Layoutget.
7028  */
7029 static int
7030 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7031     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7032     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7033     struct ucred *cred, NFSPROC_T *p)
7034 {
7035         struct nfscllayout *lyp;
7036         struct nfsclflayout *flp;
7037         struct nfsclflayouthead flh;
7038         int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7039         int layouttype, laystat;
7040         nfsv4stateid_t stateid;
7041         struct nfsclsession *tsep;
7042
7043         error = 0;
7044         if (NFSHASFLEXFILE(nmp))
7045                 layouttype = NFSLAYOUT_FLEXFILE;
7046         else
7047                 layouttype = NFSLAYOUT_NFSV4_1_FILES;
7048         /*
7049          * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7050          * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7051          * flp == NULL.
7052          */
7053         lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7054             &recalled);
7055         NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7056         if (lyp == NULL)
7057                 islocked = 0;
7058         else if (flp != NULL)
7059                 islocked = 1;
7060         else
7061                 islocked = 2;
7062         if ((lyp == NULL || flp == NULL) && recalled == 0) {
7063                 LIST_INIT(&flh);
7064                 tsep = nfsmnt_mdssession(nmp);
7065                 layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7066                     3 * NFSX_UNSIGNED);
7067                 if (lyp == NULL)
7068                         usecurstateid = 1;
7069                 else {
7070                         usecurstateid = 0;
7071                         stateid.seqid = lyp->nfsly_stateid.seqid;
7072                         stateid.other[0] = lyp->nfsly_stateid.other[0];
7073                         stateid.other[1] = lyp->nfsly_stateid.other[1];
7074                         stateid.other[2] = lyp->nfsly_stateid.other[2];
7075                 }
7076                 error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7077                     newfhp, newfhlen, mode, op, name, namelen,
7078                     dpp, &stateid, usecurstateid, layouttype, layoutlen,
7079                     &retonclose, &flh, &laystat, cred, p);
7080                 NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7081                     laystat, error);
7082                 laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7083                     &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7084                     &islocked, cred, p);
7085         } else
7086                 error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7087                     mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7088         if (islocked == 2)
7089                 nfscl_rellayout(lyp, 1);
7090         else if (islocked == 1)
7091                 nfscl_rellayout(lyp, 0);
7092         return (error);
7093 }
7094
7095 /*
7096  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7097  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7098  * handled by nfsrpc_openrpc().
7099  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7100  * can be NULL.
7101  */
7102 static int
7103 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7104     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7105     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7106     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7107     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7108     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7109 {
7110         uint32_t *tl;
7111         struct nfsrv_descript nfsd, *nd = &nfsd;
7112         struct nfscldeleg *ndp = NULL;
7113         struct nfsvattr nfsva;
7114         struct nfsclsession *tsep;
7115         uint32_t rflags, deleg;
7116         nfsattrbit_t attrbits;
7117         int error, ret, acesize, limitby, iomode;
7118
7119         *dpp = NULL;
7120         *laystatp = ENXIO;
7121         nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7122             0, 0);
7123         NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7124         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7125         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7126         *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7127         tsep = nfsmnt_mdssession(nmp);
7128         *tl++ = tsep->nfsess_clientid.lval[0];
7129         *tl = tsep->nfsess_clientid.lval[1];
7130         nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7131         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7132         *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7133         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7134         nfsm_strtom(nd, name, namelen);
7135         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7136         *tl = txdr_unsigned(NFSV4OP_GETATTR);
7137         NFSZERO_ATTRBIT(&attrbits);
7138         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7139         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7140         nfsrv_putattrbit(nd, &attrbits);
7141         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7142         *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7143         if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7144                 iomode = NFSLAYOUTIOMODE_RW;
7145         else
7146                 iomode = NFSLAYOUTIOMODE_READ;
7147         nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7148             layouttype, layoutlen, usecurstateid);
7149         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7150             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7151         if (error != 0)
7152                 return (error);
7153         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7154         if (nd->nd_repstat != 0)
7155                 *laystatp = nd->nd_repstat;
7156         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7157                 /* ND_NOMOREDATA will be set if the Open operation failed. */
7158                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7159                     6 * NFSX_UNSIGNED);
7160                 op->nfso_stateid.seqid = *tl++;
7161                 op->nfso_stateid.other[0] = *tl++;
7162                 op->nfso_stateid.other[1] = *tl++;
7163                 op->nfso_stateid.other[2] = *tl;
7164                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7165                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7166                 if (error != 0)
7167                         goto nfsmout;
7168                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7169                 deleg = fxdr_unsigned(u_int32_t, *tl);
7170                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7171                     deleg == NFSV4OPEN_DELEGATEWRITE) {
7172                         if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7173                               NFSCLFLAGS_FIRSTDELEG))
7174                                 op->nfso_own->nfsow_clp->nfsc_flags |=
7175                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7176                         ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7177                             M_NFSCLDELEG, M_WAITOK);
7178                         LIST_INIT(&ndp->nfsdl_owner);
7179                         LIST_INIT(&ndp->nfsdl_lock);
7180                         ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7181                         ndp->nfsdl_fhlen = newfhlen;
7182                         NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7183                         newnfs_copyincred(cred, &ndp->nfsdl_cred);
7184                         nfscl_lockinit(&ndp->nfsdl_rwlock);
7185                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7186                             NFSX_UNSIGNED);
7187                         ndp->nfsdl_stateid.seqid = *tl++;
7188                         ndp->nfsdl_stateid.other[0] = *tl++;
7189                         ndp->nfsdl_stateid.other[1] = *tl++;
7190                         ndp->nfsdl_stateid.other[2] = *tl++;
7191                         ret = fxdr_unsigned(int, *tl);
7192                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7193                                 ndp->nfsdl_flags = NFSCLDL_WRITE;
7194                                 /*
7195                                  * Indicates how much the file can grow.
7196                                  */
7197                                 NFSM_DISSECT(tl, u_int32_t *,
7198                                     3 * NFSX_UNSIGNED);
7199                                 limitby = fxdr_unsigned(int, *tl++);
7200                                 switch (limitby) {
7201                                 case NFSV4OPEN_LIMITSIZE:
7202                                         ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7203                                         break;
7204                                 case NFSV4OPEN_LIMITBLOCKS:
7205                                         ndp->nfsdl_sizelimit =
7206                                             fxdr_unsigned(u_int64_t, *tl++);
7207                                         ndp->nfsdl_sizelimit *=
7208                                             fxdr_unsigned(u_int64_t, *tl);
7209                                         break;
7210                                 default:
7211                                         error = NFSERR_BADXDR;
7212                                         goto nfsmout;
7213                                 };
7214                         } else
7215                                 ndp->nfsdl_flags = NFSCLDL_READ;
7216                         if (ret != 0)
7217                                 ndp->nfsdl_flags |= NFSCLDL_RECALL;
7218                         error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7219                             &acesize, p);
7220                         if (error != 0)
7221                                 goto nfsmout;
7222                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7223                         error = NFSERR_BADXDR;
7224                         goto nfsmout;
7225                 }
7226                 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7227                     nfscl_assumeposixlocks)
7228                         op->nfso_posixlock = 1;
7229                 else
7230                         op->nfso_posixlock = 0;
7231                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7232                 /* If the 2nd element == NFS_OK, the Getattr succeeded. */
7233                 if (*++tl == 0) {
7234                         error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7235                             NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7236                             NULL, NULL, NULL, p, cred);
7237                         if (error != 0)
7238                                 goto nfsmout;
7239                         if (ndp != NULL) {
7240                                 ndp->nfsdl_change = nfsva.na_filerev;
7241                                 ndp->nfsdl_modtime = nfsva.na_mtime;
7242                                 ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7243                                 *dpp = ndp;
7244                                 ndp = NULL;
7245                         }
7246                         /*
7247                          * At this point, the Open has succeeded, so set
7248                          * nd_repstat = NFS_OK.  If the Layoutget failed,
7249                          * this function just won't return a layout.
7250                          */
7251                         if (nd->nd_repstat == 0) {
7252                                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7253                                 *laystatp = fxdr_unsigned(int, *++tl);
7254                                 if (*laystatp == 0) {
7255                                         error = nfsrv_parselayoutget(nd,
7256                                             stateidp, retonclosep, flhp);
7257                                         if (error != 0)
7258                                                 *laystatp = error;
7259                                 }
7260                         } else
7261                                 nd->nd_repstat = 0;     /* Return 0 for Open. */
7262                 }
7263         }
7264         if (nd->nd_repstat != 0 && error == 0)
7265                 error = nd->nd_repstat;
7266 nfsmout:
7267         free(ndp, M_NFSCLDELEG);
7268         mbuf_freem(nd->nd_mrep);
7269         return (error);
7270 }
7271
7272 /*
7273  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7274  * Used only for mounts with pNFS enabled.
7275  */
7276 static int
7277 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7278     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7279     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7280     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7281     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7282     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7283     struct nfsclflayouthead *flhp, int *laystatp)
7284 {
7285         uint32_t *tl;
7286         int error = 0, deleg, newone, ret, acesize, limitby;
7287         struct nfsrv_descript nfsd, *nd = &nfsd;
7288         struct nfsclopen *op;
7289         struct nfscldeleg *dp = NULL;
7290         struct nfsnode *np;
7291         struct nfsfh *nfhp;
7292         struct nfsclsession *tsep;
7293         nfsattrbit_t attrbits;
7294         nfsv4stateid_t stateid;
7295         struct nfsmount *nmp;
7296
7297         nmp = VFSTONFS(dvp->v_mount);
7298         np = VTONFS(dvp);
7299         *laystatp = ENXIO;
7300         *unlockedp = 0;
7301         *nfhpp = NULL;
7302         *dpp = NULL;
7303         *attrflagp = 0;
7304         *dattrflagp = 0;
7305         if (namelen > NFS_MAXNAMLEN)
7306                 return (ENAMETOOLONG);
7307         NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7308         /*
7309          * For V4, this is actually an Open op.
7310          */
7311         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7312         *tl++ = txdr_unsigned(owp->nfsow_seqid);
7313         *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7314             NFSV4OPEN_ACCESSREAD);
7315         *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7316         tsep = nfsmnt_mdssession(nmp);
7317         *tl++ = tsep->nfsess_clientid.lval[0];
7318         *tl = tsep->nfsess_clientid.lval[1];
7319         nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7320         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7321         *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7322         if ((fmode & O_EXCL) != 0) {
7323                 if (NFSHASSESSPERSIST(nmp)) {
7324                         /* Use GUARDED for persistent sessions. */
7325                         *tl = txdr_unsigned(NFSCREATE_GUARDED);
7326                         nfscl_fillsattr(nd, vap, dvp, 0, 0);
7327                 } else {
7328                         /* Otherwise, use EXCLUSIVE4_1. */
7329                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7330                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7331                         *tl++ = cverf.lval[0];
7332                         *tl = cverf.lval[1];
7333                         nfscl_fillsattr(nd, vap, dvp, 0, 0);
7334                 }
7335         } else {
7336                 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7337                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7338         }
7339         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7340         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7341         nfsm_strtom(nd, name, namelen);
7342         /* Get the new file's handle and attributes, plus save the FH. */
7343         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7344         *tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7345         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
7346         *tl = txdr_unsigned(NFSV4OP_GETATTR);
7347         NFSGETATTR_ATTRBIT(&attrbits);
7348         nfsrv_putattrbit(nd, &attrbits);
7349         /* Get the directory's post-op attributes. */
7350         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7351         *tl = txdr_unsigned(NFSV4OP_PUTFH);
7352         nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7353         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7354         *tl = txdr_unsigned(NFSV4OP_GETATTR);
7355         nfsrv_putattrbit(nd, &attrbits);
7356         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7357         *tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7358         *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7359         nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7360             layouttype, layoutlen, usecurstateid);
7361         error = nfscl_request(nd, dvp, p, cred, dstuff);
7362         if (error != 0)
7363                 return (error);
7364         NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7365             error);
7366         if (nd->nd_repstat != 0)
7367                 *laystatp = nd->nd_repstat;
7368         NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7369         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7370                 NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7371                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7372                     6 * NFSX_UNSIGNED);
7373                 stateid.seqid = *tl++;
7374                 stateid.other[0] = *tl++;
7375                 stateid.other[1] = *tl++;
7376                 stateid.other[2] = *tl;
7377                 nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7378                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7379                 deleg = fxdr_unsigned(int, *tl);
7380                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7381                     deleg == NFSV4OPEN_DELEGATEWRITE) {
7382                         if (!(owp->nfsow_clp->nfsc_flags &
7383                               NFSCLFLAGS_FIRSTDELEG))
7384                                 owp->nfsow_clp->nfsc_flags |=
7385                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7386                         dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7387                             M_NFSCLDELEG, M_WAITOK);
7388                         LIST_INIT(&dp->nfsdl_owner);
7389                         LIST_INIT(&dp->nfsdl_lock);
7390                         dp->nfsdl_clp = owp->nfsow_clp;
7391                         newnfs_copyincred(cred, &dp->nfsdl_cred);
7392                         nfscl_lockinit(&dp->nfsdl_rwlock);
7393                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7394                             NFSX_UNSIGNED);
7395                         dp->nfsdl_stateid.seqid = *tl++;
7396                         dp->nfsdl_stateid.other[0] = *tl++;
7397                         dp->nfsdl_stateid.other[1] = *tl++;
7398                         dp->nfsdl_stateid.other[2] = *tl++;
7399                         ret = fxdr_unsigned(int, *tl);
7400                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7401                                 dp->nfsdl_flags = NFSCLDL_WRITE;
7402                                 /*
7403                                  * Indicates how much the file can grow.
7404                                  */
7405                                 NFSM_DISSECT(tl, u_int32_t *,
7406                                     3 * NFSX_UNSIGNED);
7407                                 limitby = fxdr_unsigned(int, *tl++);
7408                                 switch (limitby) {
7409                                 case NFSV4OPEN_LIMITSIZE:
7410                                         dp->nfsdl_sizelimit = fxdr_hyper(tl);
7411                                         break;
7412                                 case NFSV4OPEN_LIMITBLOCKS:
7413                                         dp->nfsdl_sizelimit =
7414                                             fxdr_unsigned(u_int64_t, *tl++);
7415                                         dp->nfsdl_sizelimit *=
7416                                             fxdr_unsigned(u_int64_t, *tl);
7417                                         break;
7418                                 default:
7419                                         error = NFSERR_BADXDR;
7420                                         goto nfsmout;
7421                                 };
7422                         } else {
7423                                 dp->nfsdl_flags = NFSCLDL_READ;
7424                         }
7425                         if (ret != 0)
7426                                 dp->nfsdl_flags |= NFSCLDL_RECALL;
7427                         error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7428                             &acesize, p);
7429                         if (error != 0)
7430                                 goto nfsmout;
7431                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7432                         error = NFSERR_BADXDR;
7433                         goto nfsmout;
7434                 }
7435
7436                 /* Now, we should have the status for the SaveFH. */
7437                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7438                 if (*++tl == 0) {
7439                         NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7440                         /*
7441                          * Now, process the GetFH and Getattr for the newly
7442                          * created file. nfscl_mtofh() will set
7443                          * ND_NOMOREDATA if these weren't successful.
7444                          */
7445                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7446                         NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7447                         if (error != 0)
7448                                 goto nfsmout;
7449                 } else
7450                         nd->nd_flag |= ND_NOMOREDATA;
7451                 /* Now we have the PutFH and Getattr for the directory. */
7452                 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7453                         NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7454                         if (*++tl != 0)
7455                                 nd->nd_flag |= ND_NOMOREDATA;
7456                         else {
7457                                 NFSM_DISSECT(tl, uint32_t *, 2 *
7458                                     NFSX_UNSIGNED);
7459                                 if (*++tl != 0)
7460                                         nd->nd_flag |= ND_NOMOREDATA;
7461                         }
7462                 }
7463                 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7464                         /* Load the directory attributes. */
7465                         error = nfsm_loadattr(nd, dnap);
7466                         NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7467                         if (error != 0)
7468                                 goto nfsmout;
7469                         *dattrflagp = 1;
7470                         if (dp != NULL && *attrflagp != 0) {
7471                                 dp->nfsdl_change = nnap->na_filerev;
7472                                 dp->nfsdl_modtime = nnap->na_mtime;
7473                                 dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7474                         }
7475                         /*
7476                          * We can now complete the Open state.
7477                          */
7478                         nfhp = *nfhpp;
7479                         if (dp != NULL) {
7480                                 dp->nfsdl_fhlen = nfhp->nfh_len;
7481                                 NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7482                                     nfhp->nfh_len);
7483                         }
7484                         /*
7485                          * Get an Open structure that will be
7486                          * attached to the OpenOwner, acquired already.
7487                          */
7488                         error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7489                             (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7490                             cred, p, NULL, &op, &newone, NULL, 0);
7491                         if (error != 0)
7492                                 goto nfsmout;
7493                         op->nfso_stateid = stateid;
7494                         newnfs_copyincred(cred, &op->nfso_cred);
7495
7496                         nfscl_openrelease(nmp, op, error, newone);
7497                         *unlockedp = 1;
7498
7499                         /* Now, handle the RestoreFH and LayoutGet. */
7500                         if (nd->nd_repstat == 0) {
7501                                 NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7502                                 *laystatp = fxdr_unsigned(int, *(tl + 3));
7503                                 if (*laystatp == 0) {
7504                                         error = nfsrv_parselayoutget(nd,
7505                                             stateidp, retonclosep, flhp);
7506                                         if (error != 0)
7507                                                 *laystatp = error;
7508                                 }
7509                                 NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7510                                     error);
7511                         } else
7512                                 nd->nd_repstat = 0;
7513                 }
7514         }
7515         if (nd->nd_repstat != 0 && error == 0)
7516                 error = nd->nd_repstat;
7517         if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7518                 nfscl_initiate_recovery(owp->nfsow_clp);
7519 nfsmout:
7520         NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7521         if (error == 0)
7522                 *dpp = dp;
7523         else
7524                 free(dp, M_NFSCLDELEG);
7525         mbuf_freem(nd->nd_mrep);
7526         return (error);
7527 }
7528
7529 /*
7530  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7531  */
7532 static int
7533 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7534     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7535     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7536     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7537     int *dattrflagp, void *dstuff, int *unlockedp)
7538 {
7539         struct nfscllayout *lyp;
7540         struct nfsclflayouthead flh;
7541         struct nfsfh *nfhp;
7542         struct nfsclsession *tsep;
7543         struct nfsmount *nmp;
7544         nfsv4stateid_t stateid;
7545         int error, layoutlen, layouttype, retonclose, laystat;
7546
7547         error = 0;
7548         nmp = VFSTONFS(dvp->v_mount);
7549         if (NFSHASFLEXFILE(nmp))
7550                 layouttype = NFSLAYOUT_FLEXFILE;
7551         else
7552                 layouttype = NFSLAYOUT_NFSV4_1_FILES;
7553         LIST_INIT(&flh);
7554         tsep = nfsmnt_mdssession(nmp);
7555         layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7556         error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7557             owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7558             dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7559             &flh, &laystat);
7560         NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7561             laystat, error);
7562         lyp = NULL;
7563         if (laystat == 0) {
7564                 nfhp = *nfhpp;
7565                 laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
7566                     nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
7567                     layouttype, laystat, NULL, cred, p);
7568         } else
7569                 laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
7570                     retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
7571                     cred, p);
7572         if (laystat == 0)
7573                 nfscl_rellayout(lyp, 0);
7574         return (error);
7575 }
7576
7577 /*
7578  * Process the results of a layoutget() operation.
7579  */
7580 static int
7581 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
7582     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
7583     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
7584     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
7585 {
7586         struct nfsclflayout *tflp;
7587         struct nfscldevinfo *dip;
7588         uint8_t *dev;
7589
7590         if (laystat == NFSERR_UNKNLAYOUTTYPE) {
7591                 NFSLOCKMNT(nmp);
7592                 if (!NFSHASFLEXFILE(nmp)) {
7593                         /* Switch to using Flex File Layout. */
7594                         nmp->nm_state |= NFSSTA_FLEXFILE;
7595                 } else if (layouttype == NFSLAYOUT_FLEXFILE) {
7596                         /* Disable pNFS. */
7597                         NFSCL_DEBUG(1, "disable PNFS\n");
7598                         nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
7599                 }
7600                 NFSUNLOCKMNT(nmp);
7601         }
7602         if (laystat == 0) {
7603                 NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
7604                 LIST_FOREACH(tflp, flhp, nfsfl_list) {
7605                         laystat = nfscl_adddevinfo(nmp, NULL, tflp);
7606                         NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
7607                         if (laystat != 0) {
7608                                 if (layouttype == NFSLAYOUT_FLEXFILE)
7609                                         dev = tflp->nfsfl_ffm[0].dev;
7610                                 else
7611                                         dev = tflp->nfsfl_dev;
7612                                 laystat = nfsrpc_getdeviceinfo(nmp, dev,
7613                                     layouttype, notifybit, &dip, cred, p);
7614                                 NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
7615                                     laystat);
7616                                 if (laystat != 0)
7617                                         break;
7618                                 laystat = nfscl_adddevinfo(nmp, dip, tflp);
7619                                 if (laystat != 0)
7620                                         printf("getlayout: cannot add\n");
7621                         }
7622                 }
7623         }
7624         if (laystat == 0) {
7625                 /*
7626                  * nfscl_layout() always returns with the nfsly_lock
7627                  * set to a refcnt (shared lock).
7628                  * Passing in dvp is sufficient, since it is only used to
7629                  * get the fsid for the file system.
7630                  */
7631                 laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
7632                     layouttype, retonclose, flhp, lypp, cred, p);
7633                 NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
7634                     laystat);
7635                 if (laystat == 0 && islockedp != NULL)
7636                         *islockedp = 1;
7637         }
7638         return (laystat);
7639 }
7640