sys/fs/nfsclient/nfs_clrpcops.c

   1 /*-
   2  * Copyright (c) 1989, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to Berkeley by
   6  * Rick Macklem at The University of Guelph.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of the University nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  *
  32  */
  33
  34 #include <sys/cdefs.h>
  35 __FBSDID("$FreeBSD$");
  36
  37 /*
  38  * Rpc op calls, generally called from the vnode op calls or through the
  39  * buffer cache, for NFS v2, 3 and 4.
  40  * These do not normally make any changes to vnode arguments or use
  41  * structures that might change between the VFS variants. The returned
  42  * arguments are all at the end, after the NFSPROC_T *p one.
  43  */
  44
  45 #ifndef APPLEKEXT
  46 #include "opt_inet6.h"
  47
  48 #include <fs/nfs/nfsport.h>
  49 #include <sys/sysctl.h>
  50 #include <sys/taskqueue.h>
  51
  52 SYSCTL_DECL(_vfs_nfs);
  53
  54 static int      nfsignore_eexist = 0;
  55 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
  56     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
  57
  58 /*
  59  * Global variables
  60  */
  61 extern int nfs_numnfscbd;
  62 extern struct timeval nfsboottime;
  63 extern u_int32_t newnfs_false, newnfs_true;
  64 extern nfstype nfsv34_type[9];
  65 extern int nfsrv_useacl;
  66 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
  67 extern int nfscl_debuglevel;
  68 extern int nfs_pnfsiothreads;
  69 NFSCLSTATEMUTEX;
  70 int nfstest_outofseq = 0;
  71 int nfscl_assumeposixlocks = 1;
  72 int nfscl_enablecallb = 0;
  73 short nfsv4_cbport = NFSV4_CBPORT;
  74 int nfstest_openallsetattr = 0;
  75 #endif  /* !APPLEKEXT */
  76
  77 #define DIRHDSIZ        offsetof(struct dirent, d_name)
  78
  79 /*
  80  * nfscl_getsameserver() can return one of three values:
  81  * NFSDSP_USETHISSESSION - Use this session for the DS.
  82  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
  83  *     session.
  84  * NFSDSP_NOTFOUND - No matching server was found.
  85  */
  86 enum nfsclds_state {
  87         NFSDSP_USETHISSESSION = 0,
  88         NFSDSP_SEQTHISSESSION = 1,
  89         NFSDSP_NOTFOUND = 2,
  90 };
  91
  92 /*
  93  * Do a write RPC on a DS data file, using this structure for the arguments,
  94  * so that this function can be executed by a separate kernel process.
  95  */
  96 struct nfsclwritedsdorpc {
  97         int                     done;
  98         int                     inprog;
  99         struct task             tsk;
 100         struct vnode            *vp;
 101         int                     iomode;
 102         int                     must_commit;
 103         nfsv4stateid_t          *stateidp;
 104         struct nfsclds          *dsp;
 105         uint64_t                off;
 106         int                     len;
 107         struct nfsfh            *fhp;
 108         struct mbuf             *m;
 109         int                     vers;
 110         int                     minorvers;
 111         struct ucred            *cred;
 112         NFSPROC_T               *p;
 113         int                     err;
 114 };
 115
 116 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
 117     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
 118 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
 119     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
 120 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
 121     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
 122     void *);
 123 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
 124     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
 125     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
 126 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
 127     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
 128     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
 129     int *, void *, int *);
 130 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
 131     struct nfscllockowner *, u_int64_t, u_int64_t,
 132     u_int32_t, struct ucred *, NFSPROC_T *, int);
 133 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
 134     struct acl *, nfsv4stateid_t *, void *);
 135 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
 136     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
 137     struct ucred *, NFSPROC_T *);
 138 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
 139     struct sockaddr_in6 *, sa_family_t, int, struct nfsclds **, NFSPROC_T *);
 140 static void nfscl_initsessionslots(struct nfsclsession *);
 141 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
 142     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
 143     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
 144     NFSPROC_T *);
 145 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
 146     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
 147     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
 148     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
 149 static struct mbuf *nfsm_copym(struct mbuf *, int, int);
 150 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
 151     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
 152     struct ucred *, NFSPROC_T *);
 153 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
 154     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
 155     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
 156 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
 157     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
 158     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
 159 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
 160     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
 161     struct ucred *, NFSPROC_T *);
 162 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
 163     struct nfsclds *, struct nfsclds **);
 164 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
 165     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
 166     NFSPROC_T *);
 167 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
 168     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
 169 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
 170     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
 171 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
 172     NFSPROC_T *);
 173 static int nfsrv_parselayoutget(struct nfsrv_descript *, nfsv4stateid_t *,
 174     int *, struct nfsclflayouthead *);
 175 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
 176     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
 177     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
 178 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
 179     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
 180     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
 181     struct nfsfh **, int *, int *, void *, int *);
 182 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
 183     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
 184     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
 185     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
 186 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
 187     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
 188     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
 189     struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
 190     int, int, int, int *, struct nfsclflayouthead *, int *);
 191 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
 192     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
 193     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
 194 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
 195     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
 196     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
 197
 198 int nfs_pnfsio(task_fn_t *, void *);
 199
 200 /*
 201  * nfs null call from vfs.
 202  */
 203 APPLESTATIC int
 204 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
 205 {
 206         int error;
 207         struct nfsrv_descript nfsd, *nd = &nfsd;
 208
 209         NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
 210         error = nfscl_request(nd, vp, p, cred, NULL);
 211         if (nd->nd_repstat && !error)
 212                 error = nd->nd_repstat;
 213         mbuf_freem(nd->nd_mrep);
 214         return (error);
 215 }
 216
 217 /*
 218  * nfs access rpc op.
 219  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
 220  * modes are changed on the server, accesses might still fail later.
 221  */
 222 APPLESTATIC int
 223 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
 224     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
 225 {
 226         int error;
 227         u_int32_t mode, rmode;
 228
 229         if (acmode & VREAD)
 230                 mode = NFSACCESS_READ;
 231         else
 232                 mode = 0;
 233         if (vnode_vtype(vp) == VDIR) {
 234                 if (acmode & VWRITE)
 235                         mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
 236                                  NFSACCESS_DELETE);
 237                 if (acmode & VEXEC)
 238                         mode |= NFSACCESS_LOOKUP;
 239         } else {
 240                 if (acmode & VWRITE)
 241                         mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
 242                 if (acmode & VEXEC)
 243                         mode |= NFSACCESS_EXECUTE;
 244         }
 245
 246         /*
 247          * Now, just call nfsrpc_accessrpc() to do the actual RPC.
 248          */
 249         error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
 250             NULL);
 251
 252         /*
 253          * The NFS V3 spec does not clarify whether or not
 254          * the returned access bits can be a superset of
 255          * the ones requested, so...
 256          */
 257         if (!error && (rmode & mode) != mode)
 258                 error = EACCES;
 259         return (error);
 260 }
 261
 262 /*
 263  * The actual rpc, separated out for Darwin.
 264  */
 265 APPLESTATIC int
 266 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
 267     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
 268     void *stuff)
 269 {
 270         u_int32_t *tl;
 271         u_int32_t supported, rmode;
 272         int error;
 273         struct nfsrv_descript nfsd, *nd = &nfsd;
 274         nfsattrbit_t attrbits;
 275
 276         *attrflagp = 0;
 277         supported = mode;
 278         NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
 279         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 280         *tl = txdr_unsigned(mode);
 281         if (nd->nd_flag & ND_NFSV4) {
 282                 /*
 283                  * And do a Getattr op.
 284                  */
 285                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 286                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
 287                 NFSGETATTR_ATTRBIT(&attrbits);
 288                 (void) nfsrv_putattrbit(nd, &attrbits);
 289         }
 290         error = nfscl_request(nd, vp, p, cred, stuff);
 291         if (error)
 292                 return (error);
 293         if (nd->nd_flag & ND_NFSV3) {
 294                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
 295                 if (error)
 296                         goto nfsmout;
 297         }
 298         if (!nd->nd_repstat) {
 299                 if (nd->nd_flag & ND_NFSV4) {
 300                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 301                         supported = fxdr_unsigned(u_int32_t, *tl++);
 302                 } else {
 303                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
 304                 }
 305                 rmode = fxdr_unsigned(u_int32_t, *tl);
 306                 if (nd->nd_flag & ND_NFSV4)
 307                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
 308
 309                 /*
 310                  * It's not obvious what should be done about
 311                  * unsupported access modes. For now, be paranoid
 312                  * and clear the unsupported ones.
 313                  */
 314                 rmode &= supported;
 315                 *rmodep = rmode;
 316         } else
 317                 error = nd->nd_repstat;
 318 nfsmout:
 319         mbuf_freem(nd->nd_mrep);
 320         return (error);
 321 }
 322
 323 /*
 324  * nfs open rpc
 325  */
 326 APPLESTATIC int
 327 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
 328 {
 329         struct nfsclopen *op;
 330         struct nfscldeleg *dp;
 331         struct nfsfh *nfhp;
 332         struct nfsnode *np = VTONFS(vp);
 333         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
 334         u_int32_t mode, clidrev;
 335         int ret, newone, error, expireret = 0, retrycnt;
 336
 337         /*
 338          * For NFSv4, Open Ops are only done on Regular Files.
 339          */
 340         if (vnode_vtype(vp) != VREG)
 341                 return (0);
 342         mode = 0;
 343         if (amode & FREAD)
 344                 mode |= NFSV4OPEN_ACCESSREAD;
 345         if (amode & FWRITE)
 346                 mode |= NFSV4OPEN_ACCESSWRITE;
 347         nfhp = np->n_fhp;
 348
 349         retrycnt = 0;
 350 #ifdef notdef
 351 { char name[100]; int namel;
 352 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
 353 bcopy(NFS4NODENAME(np->n_v4), name, namel);
 354 name[namel] = '\0';
 355 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
 356 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
 357 else printf(" fhl=0\n");
 358 }
 359 #endif
 360         do {
 361             dp = NULL;
 362             error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
 363                 cred, p, NULL, &op, &newone, &ret, 1);
 364             if (error) {
 365                 return (error);
 366             }
 367             if (nmp->nm_clp != NULL)
 368                 clidrev = nmp->nm_clp->nfsc_clientidrev;
 369             else
 370                 clidrev = 0;
 371             if (ret == NFSCLOPEN_DOOPEN) {
 372                 if (np->n_v4 != NULL) {
 373                         /*
 374                          * For the first attempt, try and get a layout, if
 375                          * pNFS is enabled for the mount.
 376                          */
 377                         if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
 378                             nfs_numnfscbd == 0 ||
 379                             (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
 380                                 error = nfsrpc_openrpc(nmp, vp,
 381                                     np->n_v4->n4_data,
 382                                     np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
 383                                     np->n_fhp->nfh_len, mode, op,
 384                                     NFS4NODENAME(np->n_v4),
 385                                     np->n_v4->n4_namelen,
 386                                     &dp, 0, 0x0, cred, p, 0, 0);
 387                         else
 388                                 error = nfsrpc_getopenlayout(nmp, vp,
 389                                     np->n_v4->n4_data,
 390                                     np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
 391                                     np->n_fhp->nfh_len, mode, op,
 392                                     NFS4NODENAME(np->n_v4),
 393                                     np->n_v4->n4_namelen, &dp, cred, p);
 394                         if (dp != NULL) {
 395 #ifdef APPLE
 396                                 OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
 397 #else
 398                                 NFSLOCKNODE(np);
 399                                 np->n_flag &= ~NDELEGMOD;
 400                                 /*
 401                                  * Invalidate the attribute cache, so that
 402                                  * attributes that pre-date the issue of a
 403                                  * delegation are not cached, since the
 404                                  * cached attributes will remain valid while
 405                                  * the delegation is held.
 406                                  */
 407                                 NFSINVALATTRCACHE(np);
 408                                 NFSUNLOCKNODE(np);
 409 #endif
 410                                 (void) nfscl_deleg(nmp->nm_mountp,
 411                                     op->nfso_own->nfsow_clp,
 412                                     nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
 413                         }
 414                 } else {
 415                         error = EIO;
 416                 }
 417                 newnfs_copyincred(cred, &op->nfso_cred);
 418             } else if (ret == NFSCLOPEN_SETCRED)
 419                 /*
 420                  * This is a new local open on a delegation. It needs
 421                  * to have credentials so that an open can be done
 422                  * against the server during recovery.
 423                  */
 424                 newnfs_copyincred(cred, &op->nfso_cred);
 425
 426             /*
 427              * nfso_opencnt is the count of how many VOP_OPEN()s have
 428              * been done on this Open successfully and a VOP_CLOSE()
 429              * is expected for each of these.
 430              * If error is non-zero, don't increment it, since the Open
 431              * hasn't succeeded yet.
 432              */
 433             if (!error)
 434                 op->nfso_opencnt++;
 435             nfscl_openrelease(nmp, op, error, newone);
 436             if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
 437                 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
 438                 error == NFSERR_BADSESSION) {
 439                 (void) nfs_catnap(PZERO, error, "nfs_open");
 440             } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
 441                 && clidrev != 0) {
 442                 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
 443                 retrycnt++;
 444             }
 445         } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
 446             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
 447             error == NFSERR_BADSESSION ||
 448             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
 449              expireret == 0 && clidrev != 0 && retrycnt < 4));
 450         if (error && retrycnt >= 4)
 451                 error = EIO;
 452         return (error);
 453 }
 454
 455 /*
 456  * the actual open rpc
 457  */
 458 APPLESTATIC int
 459 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
 460     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
 461     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
 462     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
 463     int syscred, int recursed)
 464 {
 465         u_int32_t *tl;
 466         struct nfsrv_descript nfsd, *nd = &nfsd;
 467         struct nfscldeleg *dp, *ndp = NULL;
 468         struct nfsvattr nfsva;
 469         u_int32_t rflags, deleg;
 470         nfsattrbit_t attrbits;
 471         int error, ret, acesize, limitby;
 472         struct nfsclsession *tsep;
 473
 474         dp = *dpp;
 475         *dpp = NULL;
 476         nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
 477         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
 478         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 479         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
 480         *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
 481         tsep = nfsmnt_mdssession(nmp);
 482         *tl++ = tsep->nfsess_clientid.lval[0];
 483         *tl = tsep->nfsess_clientid.lval[1];
 484         (void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
 485         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 486         *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
 487         if (reclaim) {
 488                 *tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
 489                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 490                 *tl = txdr_unsigned(delegtype);
 491         } else {
 492                 if (dp != NULL) {
 493                         *tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
 494                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
 495                         if (NFSHASNFSV4N(nmp))
 496                                 *tl++ = 0;
 497                         else
 498                                 *tl++ = dp->nfsdl_stateid.seqid;
 499                         *tl++ = dp->nfsdl_stateid.other[0];
 500                         *tl++ = dp->nfsdl_stateid.other[1];
 501                         *tl = dp->nfsdl_stateid.other[2];
 502                 } else {
 503                         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
 504                 }
 505                 (void) nfsm_strtom(nd, name, namelen);
 506         }
 507         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 508         *tl = txdr_unsigned(NFSV4OP_GETATTR);
 509         NFSZERO_ATTRBIT(&attrbits);
 510         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
 511         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
 512         (void) nfsrv_putattrbit(nd, &attrbits);
 513         if (syscred)
 514                 nd->nd_flag |= ND_USEGSSNAME;
 515         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
 516             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 517         if (error)
 518                 return (error);
 519         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 520         if (!nd->nd_repstat) {
 521                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
 522                     6 * NFSX_UNSIGNED);
 523                 op->nfso_stateid.seqid = *tl++;
 524                 op->nfso_stateid.other[0] = *tl++;
 525                 op->nfso_stateid.other[1] = *tl++;
 526                 op->nfso_stateid.other[2] = *tl;
 527                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
 528                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
 529                 if (error)
 530                         goto nfsmout;
 531                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
 532                 deleg = fxdr_unsigned(u_int32_t, *tl);
 533                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
 534                     deleg == NFSV4OPEN_DELEGATEWRITE) {
 535                         if (!(op->nfso_own->nfsow_clp->nfsc_flags &
 536                               NFSCLFLAGS_FIRSTDELEG))
 537                                 op->nfso_own->nfsow_clp->nfsc_flags |=
 538                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
 539                         MALLOC(ndp, struct nfscldeleg *,
 540                             sizeof (struct nfscldeleg) + newfhlen,
 541                             M_NFSCLDELEG, M_WAITOK);
 542                         LIST_INIT(&ndp->nfsdl_owner);
 543                         LIST_INIT(&ndp->nfsdl_lock);
 544                         ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
 545                         ndp->nfsdl_fhlen = newfhlen;
 546                         NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
 547                         newnfs_copyincred(cred, &ndp->nfsdl_cred);
 548                         nfscl_lockinit(&ndp->nfsdl_rwlock);
 549                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
 550                             NFSX_UNSIGNED);
 551                         ndp->nfsdl_stateid.seqid = *tl++;
 552                         ndp->nfsdl_stateid.other[0] = *tl++;
 553                         ndp->nfsdl_stateid.other[1] = *tl++;
 554                         ndp->nfsdl_stateid.other[2] = *tl++;
 555                         ret = fxdr_unsigned(int, *tl);
 556                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
 557                                 ndp->nfsdl_flags = NFSCLDL_WRITE;
 558                                 /*
 559                                  * Indicates how much the file can grow.
 560                                  */
 561                                 NFSM_DISSECT(tl, u_int32_t *,
 562                                     3 * NFSX_UNSIGNED);
 563                                 limitby = fxdr_unsigned(int, *tl++);
 564                                 switch (limitby) {
 565                                 case NFSV4OPEN_LIMITSIZE:
 566                                         ndp->nfsdl_sizelimit = fxdr_hyper(tl);
 567                                         break;
 568                                 case NFSV4OPEN_LIMITBLOCKS:
 569                                         ndp->nfsdl_sizelimit =
 570                                             fxdr_unsigned(u_int64_t, *tl++);
 571                                         ndp->nfsdl_sizelimit *=
 572                                             fxdr_unsigned(u_int64_t, *tl);
 573                                         break;
 574                                 default:
 575                                         error = NFSERR_BADXDR;
 576                                         goto nfsmout;
 577                                 }
 578                         } else {
 579                                 ndp->nfsdl_flags = NFSCLDL_READ;
 580                         }
 581                         if (ret)
 582                                 ndp->nfsdl_flags |= NFSCLDL_RECALL;
 583                         error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
 584                             &acesize, p);
 585                         if (error)
 586                                 goto nfsmout;
 587                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
 588                         error = NFSERR_BADXDR;
 589                         goto nfsmout;
 590                 }
 591                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 592                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
 593                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
 594                     NULL, NULL, NULL, p, cred);
 595                 if (error)
 596                         goto nfsmout;
 597                 if (ndp != NULL) {
 598                         ndp->nfsdl_change = nfsva.na_filerev;
 599                         ndp->nfsdl_modtime = nfsva.na_mtime;
 600                         ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
 601                 }
 602                 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
 603                     do {
 604                         ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
 605                             cred, p);
 606                         if (ret == NFSERR_DELAY)
 607                             (void) nfs_catnap(PZERO, ret, "nfs_open");
 608                     } while (ret == NFSERR_DELAY);
 609                     error = ret;
 610                 }
 611                 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
 612                     nfscl_assumeposixlocks)
 613                     op->nfso_posixlock = 1;
 614                 else
 615                     op->nfso_posixlock = 0;
 616
 617                 /*
 618                  * If the server is handing out delegations, but we didn't
 619                  * get one because an OpenConfirm was required, try the
 620                  * Open again, to get a delegation. This is a harmless no-op,
 621                  * from a server's point of view.
 622                  */
 623                 if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
 624                     (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
 625                     && !error && dp == NULL && ndp == NULL && !recursed) {
 626                     do {
 627                         ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
 628                             newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
 629                             cred, p, syscred, 1);
 630                         if (ret == NFSERR_DELAY)
 631                             (void) nfs_catnap(PZERO, ret, "nfs_open2");
 632                     } while (ret == NFSERR_DELAY);
 633                     if (ret) {
 634                         if (ndp != NULL) {
 635                                 FREE((caddr_t)ndp, M_NFSCLDELEG);
 636                                 ndp = NULL;
 637                         }
 638                         if (ret == NFSERR_STALECLIENTID ||
 639                             ret == NFSERR_STALEDONTRECOVER ||
 640                             ret == NFSERR_BADSESSION)
 641                                 error = ret;
 642                     }
 643                 }
 644         }
 645         if (nd->nd_repstat != 0 && error == 0)
 646                 error = nd->nd_repstat;
 647         if (error == NFSERR_STALECLIENTID)
 648                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 649 nfsmout:
 650         if (!error)
 651                 *dpp = ndp;
 652         else if (ndp != NULL)
 653                 FREE((caddr_t)ndp, M_NFSCLDELEG);
 654         mbuf_freem(nd->nd_mrep);
 655         return (error);
 656 }
 657
 658 /*
 659  * open downgrade rpc
 660  */
 661 APPLESTATIC int
 662 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
 663     struct ucred *cred, NFSPROC_T *p)
 664 {
 665         u_int32_t *tl;
 666         struct nfsrv_descript nfsd, *nd = &nfsd;
 667         int error;
 668
 669         NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
 670         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
 671         if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
 672                 *tl++ = 0;
 673         else
 674                 *tl++ = op->nfso_stateid.seqid;
 675         *tl++ = op->nfso_stateid.other[0];
 676         *tl++ = op->nfso_stateid.other[1];
 677         *tl++ = op->nfso_stateid.other[2];
 678         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 679         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
 680         *tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
 681         error = nfscl_request(nd, vp, p, cred, NULL);
 682         if (error)
 683                 return (error);
 684         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 685         if (!nd->nd_repstat) {
 686                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 687                 op->nfso_stateid.seqid = *tl++;
 688                 op->nfso_stateid.other[0] = *tl++;
 689                 op->nfso_stateid.other[1] = *tl++;
 690                 op->nfso_stateid.other[2] = *tl;
 691         }
 692         if (nd->nd_repstat && error == 0)
 693                 error = nd->nd_repstat;
 694         if (error == NFSERR_STALESTATEID)
 695                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 696 nfsmout:
 697         mbuf_freem(nd->nd_mrep);
 698         return (error);
 699 }
 700
 701 /*
 702  * V4 Close operation.
 703  */
 704 APPLESTATIC int
 705 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
 706 {
 707         struct nfsclclient *clp;
 708         int error;
 709
 710         if (vnode_vtype(vp) != VREG)
 711                 return (0);
 712         if (doclose)
 713                 error = nfscl_doclose(vp, &clp, p);
 714         else
 715                 error = nfscl_getclose(vp, &clp);
 716         if (error)
 717                 return (error);
 718
 719         nfscl_clientrelease(clp);
 720         return (0);
 721 }
 722
 723 /*
 724  * Close the open.
 725  */
 726 APPLESTATIC void
 727 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
 728 {
 729         struct nfsrv_descript nfsd, *nd = &nfsd;
 730         struct nfscllockowner *lp, *nlp;
 731         struct nfscllock *lop, *nlop;
 732         struct ucred *tcred;
 733         u_int64_t off = 0, len = 0;
 734         u_int32_t type = NFSV4LOCKT_READ;
 735         int error, do_unlock, trycnt;
 736
 737         tcred = newnfs_getcred();
 738         newnfs_copycred(&op->nfso_cred, tcred);
 739         /*
 740          * (Theoretically this could be done in the same
 741          *  compound as the close, but having multiple
 742          *  sequenced Ops in the same compound might be
 743          *  too scary for some servers.)
 744          */
 745         if (op->nfso_posixlock) {
 746                 off = 0;
 747                 len = NFS64BITSSET;
 748                 type = NFSV4LOCKT_READ;
 749         }
 750
 751         /*
 752          * Since this function is only called from VOP_INACTIVE(), no
 753          * other thread will be manipulating this Open. As such, the
 754          * lock lists are not being changed by other threads, so it should
 755          * be safe to do this without locking.
 756          */
 757         LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
 758                 do_unlock = 1;
 759                 LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
 760                         if (op->nfso_posixlock == 0) {
 761                                 off = lop->nfslo_first;
 762                                 len = lop->nfslo_end - lop->nfslo_first;
 763                                 if (lop->nfslo_type == F_WRLCK)
 764                                         type = NFSV4LOCKT_WRITE;
 765                                 else
 766                                         type = NFSV4LOCKT_READ;
 767                         }
 768                         if (do_unlock) {
 769                                 trycnt = 0;
 770                                 do {
 771                                         error = nfsrpc_locku(nd, nmp, lp, off,
 772                                             len, type, tcred, p, 0);
 773                                         if ((nd->nd_repstat == NFSERR_GRACE ||
 774                                             nd->nd_repstat == NFSERR_DELAY) &&
 775                                             error == 0)
 776                                                 (void) nfs_catnap(PZERO,
 777                                                     (int)nd->nd_repstat,
 778                                                     "nfs_close");
 779                                 } while ((nd->nd_repstat == NFSERR_GRACE ||
 780                                     nd->nd_repstat == NFSERR_DELAY) &&
 781                                     error == 0 && trycnt++ < 5);
 782                                 if (op->nfso_posixlock)
 783                                         do_unlock = 0;
 784                         }
 785                         nfscl_freelock(lop, 0);
 786                 }
 787                 /*
 788                  * Do a ReleaseLockOwner.
 789                  * The lock owner name nfsl_owner may be used by other opens for
 790                  * other files but the lock_owner4 name that nfsrpc_rellockown()
 791                  * puts on the wire has the file handle for this file appended
 792                  * to it, so it can be done now.
 793                  */
 794                 (void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
 795                     lp->nfsl_open->nfso_fhlen, tcred, p);
 796         }
 797
 798         /*
 799          * There could be other Opens for different files on the same
 800          * OpenOwner, so locking is required.
 801          */
 802         NFSLOCKCLSTATE();
 803         nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
 804         NFSUNLOCKCLSTATE();
 805         do {
 806                 error = nfscl_tryclose(op, tcred, nmp, p);
 807                 if (error == NFSERR_GRACE)
 808                         (void) nfs_catnap(PZERO, error, "nfs_close");
 809         } while (error == NFSERR_GRACE);
 810         NFSLOCKCLSTATE();
 811         nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
 812
 813         LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
 814                 nfscl_freelockowner(lp, 0);
 815         nfscl_freeopen(op, 0);
 816         NFSUNLOCKCLSTATE();
 817         NFSFREECRED(tcred);
 818 }
 819
 820 /*
 821  * The actual Close RPC.
 822  */
 823 APPLESTATIC int
 824 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
 825     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
 826     int syscred)
 827 {
 828         u_int32_t *tl;
 829         int error;
 830
 831         nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
 832             op->nfso_fhlen, NULL, NULL, 0, 0);
 833         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
 834         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
 835         if (NFSHASNFSV4N(nmp))
 836                 *tl++ = 0;
 837         else
 838                 *tl++ = op->nfso_stateid.seqid;
 839         *tl++ = op->nfso_stateid.other[0];
 840         *tl++ = op->nfso_stateid.other[1];
 841         *tl = op->nfso_stateid.other[2];
 842         if (syscred)
 843                 nd->nd_flag |= ND_USEGSSNAME;
 844         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
 845             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
 846         if (error)
 847                 return (error);
 848         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 849         if (nd->nd_repstat == 0)
 850                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 851         error = nd->nd_repstat;
 852         if (error == NFSERR_STALESTATEID)
 853                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 854 nfsmout:
 855         mbuf_freem(nd->nd_mrep);
 856         return (error);
 857 }
 858
 859 /*
 860  * V4 Open Confirm RPC.
 861  */
 862 APPLESTATIC int
 863 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
 864     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
 865 {
 866         u_int32_t *tl;
 867         struct nfsrv_descript nfsd, *nd = &nfsd;
 868         struct nfsmount *nmp;
 869         int error;
 870
 871         nmp = VFSTONFS(vnode_mount(vp));
 872         if (NFSHASNFSV4N(nmp))
 873                 return (0);             /* No confirmation for NFSv4.1. */
 874         nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
 875             0, 0);
 876         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
 877         *tl++ = op->nfso_stateid.seqid;
 878         *tl++ = op->nfso_stateid.other[0];
 879         *tl++ = op->nfso_stateid.other[1];
 880         *tl++ = op->nfso_stateid.other[2];
 881         *tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
 882         error = nfscl_request(nd, vp, p, cred, NULL);
 883         if (error)
 884                 return (error);
 885         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
 886         if (!nd->nd_repstat) {
 887                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
 888                 op->nfso_stateid.seqid = *tl++;
 889                 op->nfso_stateid.other[0] = *tl++;
 890                 op->nfso_stateid.other[1] = *tl++;
 891                 op->nfso_stateid.other[2] = *tl;
 892         }
 893         error = nd->nd_repstat;
 894         if (error == NFSERR_STALESTATEID)
 895                 nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
 896 nfsmout:
 897         mbuf_freem(nd->nd_mrep);
 898         return (error);
 899 }
 900
 901 /*
 902  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
 903  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
 904  */
 905 APPLESTATIC int
 906 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
 907     struct ucred *cred, NFSPROC_T *p)
 908 {
 909         u_int32_t *tl;
 910         struct nfsrv_descript nfsd;
 911         struct nfsrv_descript *nd = &nfsd;
 912         nfsattrbit_t attrbits;
 913         u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
 914         u_short port;
 915         int error, isinet6 = 0, callblen;
 916         nfsquad_t confirm;
 917         u_int32_t lease;
 918         static u_int32_t rev = 0;
 919         struct nfsclds *dsp;
 920         struct in6_addr a6;
 921         struct nfsclsession *tsep;
 922
 923         if (nfsboottime.tv_sec == 0)
 924                 NFSSETBOOTTIME(nfsboottime);
 925         clp->nfsc_rev = rev++;
 926         if (NFSHASNFSV4N(nmp)) {
 927                 /*
 928                  * Either there was no previous session or the
 929                  * previous session has failed, so...
 930                  * do an ExchangeID followed by the CreateSession.
 931                  */
 932                 error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
 933                     NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
 934                 NFSCL_DEBUG(1, "aft exch=%d\n", error);
 935                 if (error == 0)
 936                         error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
 937                             &nmp->nm_sockreq,
 938                             dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
 939                 if (error == 0) {
 940                         NFSLOCKMNT(nmp);
 941                         /*
 942                          * The old sessions cannot be safely free'd
 943                          * here, since they may still be used by
 944                          * in-progress RPCs.
 945                          */
 946                         tsep = NULL;
 947                         if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
 948                                 tsep = NFSMNT_MDSSESSION(nmp);
 949                         TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
 950                             nfsclds_list);
 951                         /*
 952                          * Wake up RPCs waiting for a slot on the
 953                          * old session. These will then fail with
 954                          * NFSERR_BADSESSION and be retried with the
 955                          * new session by nfsv4_setsequence().
 956                          * Also wakeup() processes waiting for the
 957                          * new session.
 958                          */
 959                         if (tsep != NULL)
 960                                 wakeup(&tsep->nfsess_slots);
 961                         wakeup(&nmp->nm_sess);
 962                         NFSUNLOCKMNT(nmp);
 963                 } else
 964                         nfscl_freenfsclds(dsp);
 965                 NFSCL_DEBUG(1, "aft createsess=%d\n", error);
 966                 if (error == 0 && reclaim == 0) {
 967                         error = nfsrpc_reclaimcomplete(nmp, cred, p);
 968                         NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
 969                         if (error == NFSERR_COMPLETEALREADY ||
 970                             error == NFSERR_NOTSUPP)
 971                                 /* Ignore this error. */
 972                                 error = 0;
 973                 }
 974                 return (error);
 975         }
 976
 977         /*
 978          * Allocate a single session structure for NFSv4.0, because some of
 979          * the fields are used by NFSv4.0 although it doesn't do a session.
 980          */
 981         dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
 982         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
 983         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
 984         NFSLOCKMNT(nmp);
 985         TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
 986         tsep = NFSMNT_MDSSESSION(nmp);
 987         NFSUNLOCKMNT(nmp);
 988
 989         nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
 990         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 991         *tl++ = txdr_unsigned(nfsboottime.tv_sec);
 992         *tl = txdr_unsigned(clp->nfsc_rev);
 993         (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
 994
 995         /*
 996          * set up the callback address
 997          */
 998         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
 999         *tl = txdr_unsigned(NFS_CALLBCKPROG);
1000         callblen = strlen(nfsv4_callbackaddr);
1001         if (callblen == 0)
1002                 cp = nfscl_getmyip(nmp, &a6, &isinet6);
1003         if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1004             (callblen > 0 || cp != NULL)) {
1005                 port = htons(nfsv4_cbport);
1006                 cp2 = (u_int8_t *)&port;
1007 #ifdef INET6
1008                 if ((callblen > 0 &&
1009                      strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1010                         char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1011
1012                         (void) nfsm_strtom(nd, "tcp6", 4);
1013                         if (callblen == 0) {
1014                                 ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1015                                 ip6add = ip6buf;
1016                         } else {
1017                                 ip6add = nfsv4_callbackaddr;
1018                         }
1019                         snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1020                             ip6add, cp2[0], cp2[1]);
1021                 } else
1022 #endif
1023                 {
1024                         (void) nfsm_strtom(nd, "tcp", 3);
1025                         if (callblen == 0)
1026                                 snprintf(addr, INET6_ADDRSTRLEN + 9,
1027                                     "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1028                                     cp[2], cp[3], cp2[0], cp2[1]);
1029                         else
1030                                 snprintf(addr, INET6_ADDRSTRLEN + 9,
1031                                     "%s.%d.%d", nfsv4_callbackaddr,
1032                                     cp2[0], cp2[1]);
1033                 }
1034                 (void) nfsm_strtom(nd, addr, strlen(addr));
1035         } else {
1036                 (void) nfsm_strtom(nd, "tcp", 3);
1037                 (void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1038         }
1039         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1040         *tl = txdr_unsigned(clp->nfsc_cbident);
1041         nd->nd_flag |= ND_USEGSSNAME;
1042         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1043                 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1044         if (error)
1045                 return (error);
1046         if (nd->nd_repstat == 0) {
1047             NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1048             tsep->nfsess_clientid.lval[0] = *tl++;
1049             tsep->nfsess_clientid.lval[1] = *tl++;
1050             confirm.lval[0] = *tl++;
1051             confirm.lval[1] = *tl;
1052             mbuf_freem(nd->nd_mrep);
1053             nd->nd_mrep = NULL;
1054
1055             /*
1056              * and confirm it.
1057              */
1058             nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1059                 NULL, 0, 0);
1060             NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1061             *tl++ = tsep->nfsess_clientid.lval[0];
1062             *tl++ = tsep->nfsess_clientid.lval[1];
1063             *tl++ = confirm.lval[0];
1064             *tl = confirm.lval[1];
1065             nd->nd_flag |= ND_USEGSSNAME;
1066             error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1067                 cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1068             if (error)
1069                 return (error);
1070             mbuf_freem(nd->nd_mrep);
1071             nd->nd_mrep = NULL;
1072             if (nd->nd_repstat == 0) {
1073                 nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
1074                     nmp->nm_fhsize, NULL, NULL, 0, 0);
1075                 NFSZERO_ATTRBIT(&attrbits);
1076                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1077                 (void) nfsrv_putattrbit(nd, &attrbits);
1078                 nd->nd_flag |= ND_USEGSSNAME;
1079                 error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1080                     cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1081                 if (error)
1082                     return (error);
1083                 if (nd->nd_repstat == 0) {
1084                     error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
1085                         NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1086                     if (error)
1087                         goto nfsmout;
1088                     clp->nfsc_renew = NFSCL_RENEW(lease);
1089                     clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1090                     clp->nfsc_clientidrev++;
1091                     if (clp->nfsc_clientidrev == 0)
1092                         clp->nfsc_clientidrev++;
1093                 }
1094             }
1095         }
1096         error = nd->nd_repstat;
1097 nfsmout:
1098         mbuf_freem(nd->nd_mrep);
1099         return (error);
1100 }
1101
1102 /*
1103  * nfs getattr call.
1104  */
1105 APPLESTATIC int
1106 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1107     struct nfsvattr *nap, void *stuff)
1108 {
1109         struct nfsrv_descript nfsd, *nd = &nfsd;
1110         int error;
1111         nfsattrbit_t attrbits;
1112
1113         NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1114         if (nd->nd_flag & ND_NFSV4) {
1115                 NFSGETATTR_ATTRBIT(&attrbits);
1116                 (void) nfsrv_putattrbit(nd, &attrbits);
1117         }
1118         error = nfscl_request(nd, vp, p, cred, stuff);
1119         if (error)
1120                 return (error);
1121         if (!nd->nd_repstat)
1122                 error = nfsm_loadattr(nd, nap);
1123         else
1124                 error = nd->nd_repstat;
1125         mbuf_freem(nd->nd_mrep);
1126         return (error);
1127 }
1128
1129 /*
1130  * nfs getattr call with non-vnode arguemnts.
1131  */
1132 APPLESTATIC int
1133 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1134     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1135     uint32_t *leasep)
1136 {
1137         struct nfsrv_descript nfsd, *nd = &nfsd;
1138         int error, vers = NFS_VER2;
1139         nfsattrbit_t attrbits;
1140
1141         nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1142         if (nd->nd_flag & ND_NFSV4) {
1143                 vers = NFS_VER4;
1144                 NFSGETATTR_ATTRBIT(&attrbits);
1145                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1146                 (void) nfsrv_putattrbit(nd, &attrbits);
1147         } else if (nd->nd_flag & ND_NFSV3) {
1148                 vers = NFS_VER3;
1149         }
1150         if (syscred)
1151                 nd->nd_flag |= ND_USEGSSNAME;
1152         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1153             NFS_PROG, vers, NULL, 1, xidp, NULL);
1154         if (error)
1155                 return (error);
1156         if (nd->nd_repstat == 0) {
1157                 if ((nd->nd_flag & ND_NFSV4) != 0)
1158                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1159                             NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1160                             NULL, NULL);
1161                 else
1162                         error = nfsm_loadattr(nd, nap);
1163         } else
1164                 error = nd->nd_repstat;
1165         mbuf_freem(nd->nd_mrep);
1166         return (error);
1167 }
1168
1169 /*
1170  * Do an nfs setattr operation.
1171  */
1172 APPLESTATIC int
1173 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1174     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1175     void *stuff)
1176 {
1177         int error, expireret = 0, openerr, retrycnt;
1178         u_int32_t clidrev = 0, mode;
1179         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1180         struct nfsfh *nfhp;
1181         nfsv4stateid_t stateid;
1182         void *lckp;
1183
1184         if (nmp->nm_clp != NULL)
1185                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1186         if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1187                 mode = NFSV4OPEN_ACCESSWRITE;
1188         else
1189                 mode = NFSV4OPEN_ACCESSREAD;
1190         retrycnt = 0;
1191         do {
1192                 lckp = NULL;
1193                 openerr = 1;
1194                 if (NFSHASNFSV4(nmp)) {
1195                         nfhp = VTONFS(vp)->n_fhp;
1196                         error = nfscl_getstateid(vp, nfhp->nfh_fh,
1197                             nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1198                         if (error && vnode_vtype(vp) == VREG &&
1199                             (mode == NFSV4OPEN_ACCESSWRITE ||
1200                              nfstest_openallsetattr)) {
1201                                 /*
1202                                  * No Open stateid, so try and open the file
1203                                  * now.
1204                                  */
1205                                 if (mode == NFSV4OPEN_ACCESSWRITE)
1206                                         openerr = nfsrpc_open(vp, FWRITE, cred,
1207                                             p);
1208                                 else
1209                                         openerr = nfsrpc_open(vp, FREAD, cred,
1210                                             p);
1211                                 if (!openerr)
1212                                         (void) nfscl_getstateid(vp,
1213                                             nfhp->nfh_fh, nfhp->nfh_len,
1214                                             mode, 0, cred, p, &stateid, &lckp);
1215                         }
1216                 }
1217                 if (vap != NULL)
1218                         error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1219                             rnap, attrflagp, stuff);
1220                 else
1221                         error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1222                             stuff);
1223                 if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1224                         NFSLOCKMNT(nmp);
1225                         nmp->nm_state |= NFSSTA_OPENMODE;
1226                         NFSUNLOCKMNT(nmp);
1227                 }
1228                 if (error == NFSERR_STALESTATEID)
1229                         nfscl_initiate_recovery(nmp->nm_clp);
1230                 if (lckp != NULL)
1231                         nfscl_lockderef(lckp);
1232                 if (!openerr)
1233                         (void) nfsrpc_close(vp, 0, p);
1234                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1235                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1236                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1237                         (void) nfs_catnap(PZERO, error, "nfs_setattr");
1238                 } else if ((error == NFSERR_EXPIRED ||
1239                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1240                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1241                 }
1242                 retrycnt++;
1243         } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1244             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1245             error == NFSERR_BADSESSION ||
1246             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1247             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1248              expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1249             (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1250              retrycnt < 4));
1251         if (error && retrycnt >= 4)
1252                 error = EIO;
1253         return (error);
1254 }
1255
1256 static int
1257 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1258     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1259     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1260 {
1261         u_int32_t *tl;
1262         struct nfsrv_descript nfsd, *nd = &nfsd;
1263         int error;
1264         nfsattrbit_t attrbits;
1265
1266         *attrflagp = 0;
1267         NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1268         if (nd->nd_flag & ND_NFSV4)
1269                 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1270         vap->va_type = vnode_vtype(vp);
1271         nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1272         if (nd->nd_flag & ND_NFSV3) {
1273                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1274                 *tl = newnfs_false;
1275         } else if (nd->nd_flag & ND_NFSV4) {
1276                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1277                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1278                 NFSGETATTR_ATTRBIT(&attrbits);
1279                 (void) nfsrv_putattrbit(nd, &attrbits);
1280         }
1281         error = nfscl_request(nd, vp, p, cred, stuff);
1282         if (error)
1283                 return (error);
1284         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1285                 error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1286         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1287                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1288         if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1289                 error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1290         mbuf_freem(nd->nd_mrep);
1291         if (nd->nd_repstat && !error)
1292                 error = nd->nd_repstat;
1293         return (error);
1294 }
1295
1296 /*
1297  * nfs lookup rpc
1298  */
1299 APPLESTATIC int
1300 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1301     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1302     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1303 {
1304         u_int32_t *tl;
1305         struct nfsrv_descript nfsd, *nd = &nfsd;
1306         struct nfsmount *nmp;
1307         struct nfsnode *np;
1308         struct nfsfh *nfhp;
1309         nfsattrbit_t attrbits;
1310         int error = 0, lookupp = 0;
1311
1312         *attrflagp = 0;
1313         *dattrflagp = 0;
1314         if (vnode_vtype(dvp) != VDIR)
1315                 return (ENOTDIR);
1316         nmp = VFSTONFS(vnode_mount(dvp));
1317         if (len > NFS_MAXNAMLEN)
1318                 return (ENAMETOOLONG);
1319         if (NFSHASNFSV4(nmp) && len == 1 &&
1320                 name[0] == '.') {
1321                 /*
1322                  * Just return the current dir's fh.
1323                  */
1324                 np = VTONFS(dvp);
1325                 MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1326                         np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1327                 nfhp->nfh_len = np->n_fhp->nfh_len;
1328                 NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1329                 *nfhpp = nfhp;
1330                 return (0);
1331         }
1332         if (NFSHASNFSV4(nmp) && len == 2 &&
1333                 name[0] == '.' && name[1] == '.') {
1334                 lookupp = 1;
1335                 NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1336         } else {
1337                 NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1338                 (void) nfsm_strtom(nd, name, len);
1339         }
1340         if (nd->nd_flag & ND_NFSV4) {
1341                 NFSGETATTR_ATTRBIT(&attrbits);
1342                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1343                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1344                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1345                 (void) nfsrv_putattrbit(nd, &attrbits);
1346         }
1347         error = nfscl_request(nd, dvp, p, cred, stuff);
1348         if (error)
1349                 return (error);
1350         if (nd->nd_repstat) {
1351                 /*
1352                  * When an NFSv4 Lookupp returns ENOENT, it means that
1353                  * the lookup is at the root of an fs, so return this dir.
1354                  */
1355                 if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1356                     np = VTONFS(dvp);
1357                     MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1358                         np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1359                     nfhp->nfh_len = np->n_fhp->nfh_len;
1360                     NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1361                     *nfhpp = nfhp;
1362                     mbuf_freem(nd->nd_mrep);
1363                     return (0);
1364                 }
1365                 if (nd->nd_flag & ND_NFSV3)
1366                     error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1367                 else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1368                     ND_NFSV4) {
1369                         /* Load the directory attributes. */
1370                         error = nfsm_loadattr(nd, dnap);
1371                         if (error == 0)
1372                                 *dattrflagp = 1;
1373                 }
1374                 goto nfsmout;
1375         }
1376         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1377                 /* Load the directory attributes. */
1378                 error = nfsm_loadattr(nd, dnap);
1379                 if (error != 0)
1380                         goto nfsmout;
1381                 *dattrflagp = 1;
1382                 /* Skip over the Lookup and GetFH operation status values. */
1383                 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1384         }
1385         error = nfsm_getfh(nd, nfhpp);
1386         if (error)
1387                 goto nfsmout;
1388
1389         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1390         if ((nd->nd_flag & ND_NFSV3) && !error)
1391                 error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1392 nfsmout:
1393         mbuf_freem(nd->nd_mrep);
1394         if (!error && nd->nd_repstat)
1395                 error = nd->nd_repstat;
1396         return (error);
1397 }
1398
1399 /*
1400  * Do a readlink rpc.
1401  */
1402 APPLESTATIC int
1403 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1404     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1405 {
1406         u_int32_t *tl;
1407         struct nfsrv_descript nfsd, *nd = &nfsd;
1408         struct nfsnode *np = VTONFS(vp);
1409         nfsattrbit_t attrbits;
1410         int error, len, cangetattr = 1;
1411
1412         *attrflagp = 0;
1413         NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1414         if (nd->nd_flag & ND_NFSV4) {
1415                 /*
1416                  * And do a Getattr op.
1417                  */
1418                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1419                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1420                 NFSGETATTR_ATTRBIT(&attrbits);
1421                 (void) nfsrv_putattrbit(nd, &attrbits);
1422         }
1423         error = nfscl_request(nd, vp, p, cred, stuff);
1424         if (error)
1425                 return (error);
1426         if (nd->nd_flag & ND_NFSV3)
1427                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1428         if (!nd->nd_repstat && !error) {
1429                 NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1430                 /*
1431                  * This seems weird to me, but must have been added to
1432                  * FreeBSD for some reason. The only thing I can think of
1433                  * is that there was/is some server that replies with
1434                  * more link data than it should?
1435                  */
1436                 if (len == NFS_MAXPATHLEN) {
1437                         NFSLOCKNODE(np);
1438                         if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1439                                 len = np->n_size;
1440                                 cangetattr = 0;
1441                         }
1442                         NFSUNLOCKNODE(np);
1443                 }
1444                 error = nfsm_mbufuio(nd, uiop, len);
1445                 if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1446                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1447         }
1448         if (nd->nd_repstat && !error)
1449                 error = nd->nd_repstat;
1450 nfsmout:
1451         mbuf_freem(nd->nd_mrep);
1452         return (error);
1453 }
1454
1455 /*
1456  * Read operation.
1457  */
1458 APPLESTATIC int
1459 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1460     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1461 {
1462         int error, expireret = 0, retrycnt;
1463         u_int32_t clidrev = 0;
1464         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1465         struct nfsnode *np = VTONFS(vp);
1466         struct ucred *newcred;
1467         struct nfsfh *nfhp = NULL;
1468         nfsv4stateid_t stateid;
1469         void *lckp;
1470
1471         if (nmp->nm_clp != NULL)
1472                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1473         newcred = cred;
1474         if (NFSHASNFSV4(nmp)) {
1475                 nfhp = np->n_fhp;
1476                 newcred = NFSNEWCRED(cred);
1477         }
1478         retrycnt = 0;
1479         do {
1480                 lckp = NULL;
1481                 if (NFSHASNFSV4(nmp))
1482                         (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1483                             NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1484                             &lckp);
1485                 error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1486                     attrflagp, stuff);
1487                 if (error == NFSERR_OPENMODE) {
1488                         NFSLOCKMNT(nmp);
1489                         nmp->nm_state |= NFSSTA_OPENMODE;
1490                         NFSUNLOCKMNT(nmp);
1491                 }
1492                 if (error == NFSERR_STALESTATEID)
1493                         nfscl_initiate_recovery(nmp->nm_clp);
1494                 if (lckp != NULL)
1495                         nfscl_lockderef(lckp);
1496                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1497                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1498                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1499                         (void) nfs_catnap(PZERO, error, "nfs_read");
1500                 } else if ((error == NFSERR_EXPIRED ||
1501                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1502                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1503                 }
1504                 retrycnt++;
1505         } while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1506             error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1507             error == NFSERR_BADSESSION ||
1508             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1509             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1510              expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1511             (error == NFSERR_OPENMODE && retrycnt < 4));
1512         if (error && retrycnt >= 4)
1513                 error = EIO;
1514         if (NFSHASNFSV4(nmp))
1515                 NFSFREECRED(newcred);
1516         return (error);
1517 }
1518
1519 /*
1520  * The actual read RPC.
1521  */
1522 static int
1523 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1524     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1525     int *attrflagp, void *stuff)
1526 {
1527         u_int32_t *tl;
1528         int error = 0, len, retlen, tsiz, eof = 0;
1529         struct nfsrv_descript nfsd;
1530         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1531         struct nfsrv_descript *nd = &nfsd;
1532         int rsize;
1533         off_t tmp_off;
1534
1535         *attrflagp = 0;
1536         tsiz = uio_uio_resid(uiop);
1537         tmp_off = uiop->uio_offset + tsiz;
1538         NFSLOCKMNT(nmp);
1539         if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1540                 NFSUNLOCKMNT(nmp);
1541                 return (EFBIG);
1542         }
1543         rsize = nmp->nm_rsize;
1544         NFSUNLOCKMNT(nmp);
1545         nd->nd_mrep = NULL;
1546         while (tsiz > 0) {
1547                 *attrflagp = 0;
1548                 len = (tsiz > rsize) ? rsize : tsiz;
1549                 NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1550                 if (nd->nd_flag & ND_NFSV4)
1551                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1552                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1553                 if (nd->nd_flag & ND_NFSV2) {
1554                         *tl++ = txdr_unsigned(uiop->uio_offset);
1555                         *tl++ = txdr_unsigned(len);
1556                         *tl = 0;
1557                 } else {
1558                         txdr_hyper(uiop->uio_offset, tl);
1559                         *(tl + 2) = txdr_unsigned(len);
1560                 }
1561                 /*
1562                  * Since I can't do a Getattr for NFSv4 for Write, there
1563                  * doesn't seem any point in doing one here, either.
1564                  * (See the comment in nfsrpc_writerpc() for more info.)
1565                  */
1566                 error = nfscl_request(nd, vp, p, cred, stuff);
1567                 if (error)
1568                         return (error);
1569                 if (nd->nd_flag & ND_NFSV3) {
1570                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1571                 } else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1572                         error = nfsm_loadattr(nd, nap);
1573                         if (!error)
1574                                 *attrflagp = 1;
1575                 }
1576                 if (nd->nd_repstat || error) {
1577                         if (!error)
1578                                 error = nd->nd_repstat;
1579                         goto nfsmout;
1580                 }
1581                 if (nd->nd_flag & ND_NFSV3) {
1582                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1583                         eof = fxdr_unsigned(int, *(tl + 1));
1584                 } else if (nd->nd_flag & ND_NFSV4) {
1585                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1586                         eof = fxdr_unsigned(int, *tl);
1587                 }
1588                 NFSM_STRSIZ(retlen, len);
1589                 error = nfsm_mbufuio(nd, uiop, retlen);
1590                 if (error)
1591                         goto nfsmout;
1592                 mbuf_freem(nd->nd_mrep);
1593                 nd->nd_mrep = NULL;
1594                 tsiz -= retlen;
1595                 if (!(nd->nd_flag & ND_NFSV2)) {
1596                         if (eof || retlen == 0)
1597                                 tsiz = 0;
1598                 } else if (retlen < len)
1599                         tsiz = 0;
1600         }
1601         return (0);
1602 nfsmout:
1603         if (nd->nd_mrep != NULL)
1604                 mbuf_freem(nd->nd_mrep);
1605         return (error);
1606 }
1607
1608 /*
1609  * nfs write operation
1610  * When called_from_strategy != 0, it should return EIO for an error that
1611  * indicates recovery is in progress, so that the buffer will be left
1612  * dirty and be written back to the server later. If it loops around,
1613  * the recovery thread could get stuck waiting for the buffer and recovery
1614  * will then deadlock.
1615  */
1616 APPLESTATIC int
1617 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1618     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1619     void *stuff, int called_from_strategy)
1620 {
1621         int error, expireret = 0, retrycnt, nostateid;
1622         u_int32_t clidrev = 0;
1623         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1624         struct nfsnode *np = VTONFS(vp);
1625         struct ucred *newcred;
1626         struct nfsfh *nfhp = NULL;
1627         nfsv4stateid_t stateid;
1628         void *lckp;
1629
1630         *must_commit = 0;
1631         if (nmp->nm_clp != NULL)
1632                 clidrev = nmp->nm_clp->nfsc_clientidrev;
1633         newcred = cred;
1634         if (NFSHASNFSV4(nmp)) {
1635                 newcred = NFSNEWCRED(cred);
1636                 nfhp = np->n_fhp;
1637         }
1638         retrycnt = 0;
1639         do {
1640                 lckp = NULL;
1641                 nostateid = 0;
1642                 if (NFSHASNFSV4(nmp)) {
1643                         (void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1644                             NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1645                             &lckp);
1646                         if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1647                             stateid.other[2] == 0) {
1648                                 nostateid = 1;
1649                                 NFSCL_DEBUG(1, "stateid0 in write\n");
1650                         }
1651                 }
1652
1653                 /*
1654                  * If there is no stateid for NFSv4, it means this is an
1655                  * extraneous write after close. Basically a poorly
1656                  * implemented buffer cache. Just don't do the write.
1657                  */
1658                 if (nostateid)
1659                         error = 0;
1660                 else
1661                         error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1662                             newcred, &stateid, p, nap, attrflagp, stuff);
1663                 if (error == NFSERR_STALESTATEID)
1664                         nfscl_initiate_recovery(nmp->nm_clp);
1665                 if (lckp != NULL)
1666                         nfscl_lockderef(lckp);
1667                 if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1668                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1669                     error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1670                         (void) nfs_catnap(PZERO, error, "nfs_write");
1671                 } else if ((error == NFSERR_EXPIRED ||
1672                     error == NFSERR_BADSTATEID) && clidrev != 0) {
1673                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1674                 }
1675                 retrycnt++;
1676         } while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1677             ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1678               error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1679             (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1680             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1681              expireret == 0 && clidrev != 0 && retrycnt < 4));
1682         if (error != 0 && (retrycnt >= 4 ||
1683             ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1684               error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1685                 error = EIO;
1686         if (NFSHASNFSV4(nmp))
1687                 NFSFREECRED(newcred);
1688         return (error);
1689 }
1690
1691 /*
1692  * The actual write RPC.
1693  */
1694 static int
1695 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1696     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1697     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1698 {
1699         u_int32_t *tl;
1700         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1701         struct nfsnode *np = VTONFS(vp);
1702         int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1703         int wccflag = 0, wsize;
1704         int32_t backup;
1705         struct nfsrv_descript nfsd;
1706         struct nfsrv_descript *nd = &nfsd;
1707         nfsattrbit_t attrbits;
1708         off_t tmp_off;
1709
1710         KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1711         *attrflagp = 0;
1712         tsiz = uio_uio_resid(uiop);
1713         tmp_off = uiop->uio_offset + tsiz;
1714         NFSLOCKMNT(nmp);
1715         if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1716                 NFSUNLOCKMNT(nmp);
1717                 return (EFBIG);
1718         }
1719         wsize = nmp->nm_wsize;
1720         NFSUNLOCKMNT(nmp);
1721         nd->nd_mrep = NULL;     /* NFSv2 sometimes does a write with */
1722         nd->nd_repstat = 0;     /* uio_resid == 0, so the while is not done */
1723         while (tsiz > 0) {
1724                 *attrflagp = 0;
1725                 len = (tsiz > wsize) ? wsize : tsiz;
1726                 NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1727                 if (nd->nd_flag & ND_NFSV4) {
1728                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1729                         NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1730                         txdr_hyper(uiop->uio_offset, tl);
1731                         tl += 2;
1732                         *tl++ = txdr_unsigned(*iomode);
1733                         *tl = txdr_unsigned(len);
1734                 } else if (nd->nd_flag & ND_NFSV3) {
1735                         NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1736                         txdr_hyper(uiop->uio_offset, tl);
1737                         tl += 2;
1738                         *tl++ = txdr_unsigned(len);
1739                         *tl++ = txdr_unsigned(*iomode);
1740                         *tl = txdr_unsigned(len);
1741                 } else {
1742                         u_int32_t x;
1743
1744                         NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1745                         /*
1746                          * Not sure why someone changed this, since the
1747                          * RFC clearly states that "beginoffset" and
1748                          * "totalcount" are ignored, but it wouldn't
1749                          * surprise me if there's a busted server out there.
1750                          */
1751                         /* Set both "begin" and "current" to non-garbage. */
1752                         x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1753                         *tl++ = x;      /* "begin offset" */
1754                         *tl++ = x;      /* "current offset" */
1755                         x = txdr_unsigned(len);
1756                         *tl++ = x;      /* total to this offset */
1757                         *tl = x;        /* size of this write */
1758
1759                 }
1760                 nfsm_uiombuf(nd, uiop, len);
1761                 /*
1762                  * Although it is tempting to do a normal Getattr Op in the
1763                  * NFSv4 compound, the result can be a nearly hung client
1764                  * system if the Getattr asks for Owner and/or OwnerGroup.
1765                  * It occurs when the client can't map either the Owner or
1766                  * Owner_group name in the Getattr reply to a uid/gid. When
1767                  * there is a cache miss, the kernel does an upcall to the
1768                  * nfsuserd. Then, it can try and read the local /etc/passwd
1769                  * or /etc/group file. It can then block in getnewbuf(),
1770                  * waiting for dirty writes to be pushed to the NFS server.
1771                  * The only reason this doesn't result in a complete
1772                  * deadlock, is that the upcall times out and allows
1773                  * the write to complete. However, progress is so slow
1774                  * that it might just as well be deadlocked.
1775                  * As such, we get the rest of the attributes, but not
1776                  * Owner or Owner_group.
1777                  * nb: nfscl_loadattrcache() needs to be told that these
1778                  *     partial attributes from a write rpc are being
1779                  *     passed in, via a argument flag.
1780                  */
1781                 if (nd->nd_flag & ND_NFSV4) {
1782                         NFSWRITEGETATTR_ATTRBIT(&attrbits);
1783                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1784                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
1785                         (void) nfsrv_putattrbit(nd, &attrbits);
1786                 }
1787                 error = nfscl_request(nd, vp, p, cred, stuff);
1788                 if (error)
1789                         return (error);
1790                 if (nd->nd_repstat) {
1791                         /*
1792                          * In case the rpc gets retried, roll
1793                          * the uio fileds changed by nfsm_uiombuf()
1794                          * back.
1795                          */
1796                         uiop->uio_offset -= len;
1797                         uio_uio_resid_add(uiop, len);
1798                         uio_iov_base_add(uiop, -len);
1799                         uio_iov_len_add(uiop, len);
1800                 }
1801                 if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1802                         error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1803                             &wccflag, stuff);
1804                         if (error)
1805                                 goto nfsmout;
1806                 }
1807                 if (!nd->nd_repstat) {
1808                         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1809                                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1810                                         + NFSX_VERF);
1811                                 rlen = fxdr_unsigned(int, *tl++);
1812                                 if (rlen == 0) {
1813                                         error = NFSERR_IO;
1814                                         goto nfsmout;
1815                                 } else if (rlen < len) {
1816                                         backup = len - rlen;
1817                                         uio_iov_base_add(uiop, -(backup));
1818                                         uio_iov_len_add(uiop, backup);
1819                                         uiop->uio_offset -= backup;
1820                                         uio_uio_resid_add(uiop, backup);
1821                                         len = rlen;
1822                                 }
1823                                 commit = fxdr_unsigned(int, *tl++);
1824
1825                                 /*
1826                                  * Return the lowest commitment level
1827                                  * obtained by any of the RPCs.
1828                                  */
1829                                 if (committed == NFSWRITE_FILESYNC)
1830                                         committed = commit;
1831                                 else if (committed == NFSWRITE_DATASYNC &&
1832                                         commit == NFSWRITE_UNSTABLE)
1833                                         committed = commit;
1834                                 NFSLOCKMNT(nmp);
1835                                 if (!NFSHASWRITEVERF(nmp)) {
1836                                         NFSBCOPY((caddr_t)tl,
1837                                             (caddr_t)&nmp->nm_verf[0],
1838                                             NFSX_VERF);
1839                                         NFSSETWRITEVERF(nmp);
1840                                 } else if (NFSBCMP(tl, nmp->nm_verf,
1841                                     NFSX_VERF)) {
1842                                         *must_commit = 1;
1843                                         NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1844                                 }
1845                                 NFSUNLOCKMNT(nmp);
1846                         }
1847                         if (nd->nd_flag & ND_NFSV4)
1848                                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1849                         if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1850                                 error = nfsm_loadattr(nd, nap);
1851                                 if (!error)
1852                                         *attrflagp = NFS_LATTR_NOSHRINK;
1853                         }
1854                 } else {
1855                         error = nd->nd_repstat;
1856                 }
1857                 if (error)
1858                         goto nfsmout;
1859                 NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1860                 mbuf_freem(nd->nd_mrep);
1861                 nd->nd_mrep = NULL;
1862                 tsiz -= len;
1863         }
1864 nfsmout:
1865         if (nd->nd_mrep != NULL)
1866                 mbuf_freem(nd->nd_mrep);
1867         *iomode = committed;
1868         if (nd->nd_repstat && !error)
1869                 error = nd->nd_repstat;
1870         return (error);
1871 }
1872
1873 /*
1874  * nfs mknod rpc
1875  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1876  * mode set to specify the file type and the size field for rdev.
1877  */
1878 APPLESTATIC int
1879 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1880     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1881     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1882     int *attrflagp, int *dattrflagp, void *dstuff)
1883 {
1884         u_int32_t *tl;
1885         int error = 0;
1886         struct nfsrv_descript nfsd, *nd = &nfsd;
1887         nfsattrbit_t attrbits;
1888
1889         *nfhpp = NULL;
1890         *attrflagp = 0;
1891         *dattrflagp = 0;
1892         if (namelen > NFS_MAXNAMLEN)
1893                 return (ENAMETOOLONG);
1894         NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1895         if (nd->nd_flag & ND_NFSV4) {
1896                 if (vtyp == VBLK || vtyp == VCHR) {
1897                         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1898                         *tl++ = vtonfsv34_type(vtyp);
1899                         *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1900                         *tl = txdr_unsigned(NFSMINOR(rdev));
1901                 } else {
1902                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1903                         *tl = vtonfsv34_type(vtyp);
1904                 }
1905         }
1906         (void) nfsm_strtom(nd, name, namelen);
1907         if (nd->nd_flag & ND_NFSV3) {
1908                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1909                 *tl = vtonfsv34_type(vtyp);
1910         }
1911         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1912                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
1913         if ((nd->nd_flag & ND_NFSV3) &&
1914             (vtyp == VCHR || vtyp == VBLK)) {
1915                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1916                 *tl++ = txdr_unsigned(NFSMAJOR(rdev));
1917                 *tl = txdr_unsigned(NFSMINOR(rdev));
1918         }
1919         if (nd->nd_flag & ND_NFSV4) {
1920                 NFSGETATTR_ATTRBIT(&attrbits);
1921                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1922                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
1923                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
1924                 (void) nfsrv_putattrbit(nd, &attrbits);
1925         }
1926         if (nd->nd_flag & ND_NFSV2)
1927                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1928         error = nfscl_request(nd, dvp, p, cred, dstuff);
1929         if (error)
1930                 return (error);
1931         if (nd->nd_flag & ND_NFSV4)
1932                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1933         if (!nd->nd_repstat) {
1934                 if (nd->nd_flag & ND_NFSV4) {
1935                         NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1936                         error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1937                         if (error)
1938                                 goto nfsmout;
1939                 }
1940                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1941                 if (error)
1942                         goto nfsmout;
1943         }
1944         if (nd->nd_flag & ND_NFSV3)
1945                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1946         if (!error && nd->nd_repstat)
1947                 error = nd->nd_repstat;
1948 nfsmout:
1949         mbuf_freem(nd->nd_mrep);
1950         return (error);
1951 }
1952
1953 /*
1954  * nfs file create call
1955  * Mostly just call the approriate routine. (I separated out v4, so that
1956  * error recovery wouldn't be as difficult.)
1957  */
1958 APPLESTATIC int
1959 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1960     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1961     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1962     int *attrflagp, int *dattrflagp, void *dstuff)
1963 {
1964         int error = 0, newone, expireret = 0, retrycnt, unlocked;
1965         struct nfsclowner *owp;
1966         struct nfscldeleg *dp;
1967         struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1968         u_int32_t clidrev;
1969
1970         if (NFSHASNFSV4(nmp)) {
1971             retrycnt = 0;
1972             do {
1973                 dp = NULL;
1974                 error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1975                     NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1976                     NULL, 1);
1977                 if (error)
1978                         return (error);
1979                 if (nmp->nm_clp != NULL)
1980                         clidrev = nmp->nm_clp->nfsc_clientidrev;
1981                 else
1982                         clidrev = 0;
1983                 if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
1984                     nfs_numnfscbd == 0 || retrycnt > 0)
1985                         error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
1986                           fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1987                           attrflagp, dattrflagp, dstuff, &unlocked);
1988                 else
1989                         error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
1990                           cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
1991                           attrflagp, dattrflagp, dstuff, &unlocked);
1992                 /*
1993                  * There is no need to invalidate cached attributes here,
1994                  * since new post-delegation issue attributes are always
1995                  * returned by nfsrpc_createv4() and these will update the
1996                  * attribute cache.
1997                  */
1998                 if (dp != NULL)
1999                         (void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2000                             (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2001                 nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2002                 if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2003                     error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2004                     error == NFSERR_BADSESSION) {
2005                         (void) nfs_catnap(PZERO, error, "nfs_open");
2006                 } else if ((error == NFSERR_EXPIRED ||
2007                     error == NFSERR_BADSTATEID) && clidrev != 0) {
2008                         expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2009                         retrycnt++;
2010                 }
2011             } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2012                 error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2013                 error == NFSERR_BADSESSION ||
2014                 ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2015                  expireret == 0 && clidrev != 0 && retrycnt < 4));
2016             if (error && retrycnt >= 4)
2017                     error = EIO;
2018         } else {
2019                 error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2020                     fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2021                     dstuff);
2022         }
2023         return (error);
2024 }
2025
2026 /*
2027  * The create rpc for v2 and 3.
2028  */
2029 static int
2030 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2031     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2032     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2033     int *attrflagp, int *dattrflagp, void *dstuff)
2034 {
2035         u_int32_t *tl;
2036         int error = 0;
2037         struct nfsrv_descript nfsd, *nd = &nfsd;
2038
2039         *nfhpp = NULL;
2040         *attrflagp = 0;
2041         *dattrflagp = 0;
2042         if (namelen > NFS_MAXNAMLEN)
2043                 return (ENAMETOOLONG);
2044         NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2045         (void) nfsm_strtom(nd, name, namelen);
2046         if (nd->nd_flag & ND_NFSV3) {
2047                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2048                 if (fmode & O_EXCL) {
2049                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2050                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2051                         *tl++ = cverf.lval[0];
2052                         *tl = cverf.lval[1];
2053                 } else {
2054                         *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2055                         nfscl_fillsattr(nd, vap, dvp, 0, 0);
2056                 }
2057         } else {
2058                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2059         }
2060         error = nfscl_request(nd, dvp, p, cred, dstuff);
2061         if (error)
2062                 return (error);
2063         if (nd->nd_repstat == 0) {
2064                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2065                 if (error)
2066                         goto nfsmout;
2067         }
2068         if (nd->nd_flag & ND_NFSV3)
2069                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2070         if (nd->nd_repstat != 0 && error == 0)
2071                 error = nd->nd_repstat;
2072 nfsmout:
2073         mbuf_freem(nd->nd_mrep);
2074         return (error);
2075 }
2076
2077 static int
2078 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2079     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2080     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2081     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2082     int *dattrflagp, void *dstuff, int *unlockedp)
2083 {
2084         u_int32_t *tl;
2085         int error = 0, deleg, newone, ret, acesize, limitby;
2086         struct nfsrv_descript nfsd, *nd = &nfsd;
2087         struct nfsclopen *op;
2088         struct nfscldeleg *dp = NULL;
2089         struct nfsnode *np;
2090         struct nfsfh *nfhp;
2091         nfsattrbit_t attrbits;
2092         nfsv4stateid_t stateid;
2093         u_int32_t rflags;
2094         struct nfsmount *nmp;
2095         struct nfsclsession *tsep;
2096
2097         nmp = VFSTONFS(dvp->v_mount);
2098         np = VTONFS(dvp);
2099         *unlockedp = 0;
2100         *nfhpp = NULL;
2101         *dpp = NULL;
2102         *attrflagp = 0;
2103         *dattrflagp = 0;
2104         if (namelen > NFS_MAXNAMLEN)
2105                 return (ENAMETOOLONG);
2106         NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2107         /*
2108          * For V4, this is actually an Open op.
2109          */
2110         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2111         *tl++ = txdr_unsigned(owp->nfsow_seqid);
2112         *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2113             NFSV4OPEN_ACCESSREAD);
2114         *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2115         tsep = nfsmnt_mdssession(nmp);
2116         *tl++ = tsep->nfsess_clientid.lval[0];
2117         *tl = tsep->nfsess_clientid.lval[1];
2118         (void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2119         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2120         *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2121         if (fmode & O_EXCL) {
2122                 if (NFSHASNFSV4N(nmp)) {
2123                         if (NFSHASSESSPERSIST(nmp)) {
2124                                 /* Use GUARDED for persistent sessions. */
2125                                 *tl = txdr_unsigned(NFSCREATE_GUARDED);
2126                                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2127                         } else {
2128                                 /* Otherwise, use EXCLUSIVE4_1. */
2129                                 *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2130                                 NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2131                                 *tl++ = cverf.lval[0];
2132                                 *tl = cverf.lval[1];
2133                                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2134                         }
2135                 } else {
2136                         /* NFSv4.0 */
2137                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2138                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2139                         *tl++ = cverf.lval[0];
2140                         *tl = cverf.lval[1];
2141                 }
2142         } else {
2143                 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2144                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2145         }
2146         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2147         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2148         (void) nfsm_strtom(nd, name, namelen);
2149         /* Get the new file's handle and attributes. */
2150         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2151         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2152         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2153         NFSGETATTR_ATTRBIT(&attrbits);
2154         (void) nfsrv_putattrbit(nd, &attrbits);
2155         /* Get the directory's post-op attributes. */
2156         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2157         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2158         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2159         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2160         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2161         (void) nfsrv_putattrbit(nd, &attrbits);
2162         error = nfscl_request(nd, dvp, p, cred, dstuff);
2163         if (error)
2164                 return (error);
2165         NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2166         if (nd->nd_repstat == 0) {
2167                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2168                     6 * NFSX_UNSIGNED);
2169                 stateid.seqid = *tl++;
2170                 stateid.other[0] = *tl++;
2171                 stateid.other[1] = *tl++;
2172                 stateid.other[2] = *tl;
2173                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2174                 (void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2175                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2176                 deleg = fxdr_unsigned(int, *tl);
2177                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
2178                     deleg == NFSV4OPEN_DELEGATEWRITE) {
2179                         if (!(owp->nfsow_clp->nfsc_flags &
2180                               NFSCLFLAGS_FIRSTDELEG))
2181                                 owp->nfsow_clp->nfsc_flags |=
2182                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2183                         MALLOC(dp, struct nfscldeleg *,
2184                             sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2185                             M_NFSCLDELEG, M_WAITOK);
2186                         LIST_INIT(&dp->nfsdl_owner);
2187                         LIST_INIT(&dp->nfsdl_lock);
2188                         dp->nfsdl_clp = owp->nfsow_clp;
2189                         newnfs_copyincred(cred, &dp->nfsdl_cred);
2190                         nfscl_lockinit(&dp->nfsdl_rwlock);
2191                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2192                             NFSX_UNSIGNED);
2193                         dp->nfsdl_stateid.seqid = *tl++;
2194                         dp->nfsdl_stateid.other[0] = *tl++;
2195                         dp->nfsdl_stateid.other[1] = *tl++;
2196                         dp->nfsdl_stateid.other[2] = *tl++;
2197                         ret = fxdr_unsigned(int, *tl);
2198                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2199                                 dp->nfsdl_flags = NFSCLDL_WRITE;
2200                                 /*
2201                                  * Indicates how much the file can grow.
2202                                  */
2203                                 NFSM_DISSECT(tl, u_int32_t *,
2204                                     3 * NFSX_UNSIGNED);
2205                                 limitby = fxdr_unsigned(int, *tl++);
2206                                 switch (limitby) {
2207                                 case NFSV4OPEN_LIMITSIZE:
2208                                         dp->nfsdl_sizelimit = fxdr_hyper(tl);
2209                                         break;
2210                                 case NFSV4OPEN_LIMITBLOCKS:
2211                                         dp->nfsdl_sizelimit =
2212                                             fxdr_unsigned(u_int64_t, *tl++);
2213                                         dp->nfsdl_sizelimit *=
2214                                             fxdr_unsigned(u_int64_t, *tl);
2215                                         break;
2216                                 default:
2217                                         error = NFSERR_BADXDR;
2218                                         goto nfsmout;
2219                                 }
2220                         } else {
2221                                 dp->nfsdl_flags = NFSCLDL_READ;
2222                         }
2223                         if (ret)
2224                                 dp->nfsdl_flags |= NFSCLDL_RECALL;
2225                         error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2226                             &acesize, p);
2227                         if (error)
2228                                 goto nfsmout;
2229                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
2230                         error = NFSERR_BADXDR;
2231                         goto nfsmout;
2232                 }
2233                 error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2234                 if (error)
2235                         goto nfsmout;
2236                 /* Get rid of the PutFH and Getattr status values. */
2237                 NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2238                 /* Load the directory attributes. */
2239                 error = nfsm_loadattr(nd, dnap);
2240                 if (error)
2241                         goto nfsmout;
2242                 *dattrflagp = 1;
2243                 if (dp != NULL && *attrflagp) {
2244                         dp->nfsdl_change = nnap->na_filerev;
2245                         dp->nfsdl_modtime = nnap->na_mtime;
2246                         dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2247                 }
2248                 /*
2249                  * We can now complete the Open state.
2250                  */
2251                 nfhp = *nfhpp;
2252                 if (dp != NULL) {
2253                         dp->nfsdl_fhlen = nfhp->nfh_len;
2254                         NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2255                 }
2256                 /*
2257                  * Get an Open structure that will be
2258                  * attached to the OpenOwner, acquired already.
2259                  */
2260                 error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2261                     (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2262                     cred, p, NULL, &op, &newone, NULL, 0);
2263                 if (error)
2264                         goto nfsmout;
2265                 op->nfso_stateid = stateid;
2266                 newnfs_copyincred(cred, &op->nfso_cred);
2267                 if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2268                     do {
2269                         ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2270                             nfhp->nfh_len, op, cred, p);
2271                         if (ret == NFSERR_DELAY)
2272                             (void) nfs_catnap(PZERO, ret, "nfs_create");
2273                     } while (ret == NFSERR_DELAY);
2274                     error = ret;
2275                 }
2276
2277                 /*
2278                  * If the server is handing out delegations, but we didn't
2279                  * get one because an OpenConfirm was required, try the
2280                  * Open again, to get a delegation. This is a harmless no-op,
2281                  * from a server's point of view.
2282                  */
2283                 if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2284                     (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2285                     !error && dp == NULL) {
2286                     do {
2287                         ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2288                             np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2289                             nfhp->nfh_fh, nfhp->nfh_len,
2290                             (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2291                             name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2292                         if (ret == NFSERR_DELAY)
2293                             (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2294                     } while (ret == NFSERR_DELAY);
2295                     if (ret) {
2296                         if (dp != NULL) {
2297                                 FREE((caddr_t)dp, M_NFSCLDELEG);
2298                                 dp = NULL;
2299                         }
2300                         if (ret == NFSERR_STALECLIENTID ||
2301                             ret == NFSERR_STALEDONTRECOVER ||
2302                             ret == NFSERR_BADSESSION)
2303                                 error = ret;
2304                     }
2305                 }
2306                 nfscl_openrelease(nmp, op, error, newone);
2307                 *unlockedp = 1;
2308         }
2309         if (nd->nd_repstat != 0 && error == 0)
2310                 error = nd->nd_repstat;
2311         if (error == NFSERR_STALECLIENTID)
2312                 nfscl_initiate_recovery(owp->nfsow_clp);
2313 nfsmout:
2314         if (!error)
2315                 *dpp = dp;
2316         else if (dp != NULL)
2317                 FREE((caddr_t)dp, M_NFSCLDELEG);
2318         mbuf_freem(nd->nd_mrep);
2319         return (error);
2320 }
2321
2322 /*
2323  * Nfs remove rpc
2324  */
2325 APPLESTATIC int
2326 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2327     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2328     void *dstuff)
2329 {
2330         u_int32_t *tl;
2331         struct nfsrv_descript nfsd, *nd = &nfsd;
2332         struct nfsnode *np;
2333         struct nfsmount *nmp;
2334         nfsv4stateid_t dstateid;
2335         int error, ret = 0, i;
2336
2337         *dattrflagp = 0;
2338         if (namelen > NFS_MAXNAMLEN)
2339                 return (ENAMETOOLONG);
2340         nmp = VFSTONFS(vnode_mount(dvp));
2341 tryagain:
2342         if (NFSHASNFSV4(nmp) && ret == 0) {
2343                 ret = nfscl_removedeleg(vp, p, &dstateid);
2344                 if (ret == 1) {
2345                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2346                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2347                             NFSX_UNSIGNED);
2348                         if (NFSHASNFSV4N(nmp))
2349                                 *tl++ = 0;
2350                         else
2351                                 *tl++ = dstateid.seqid;
2352                         *tl++ = dstateid.other[0];
2353                         *tl++ = dstateid.other[1];
2354                         *tl++ = dstateid.other[2];
2355                         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2356                         np = VTONFS(dvp);
2357                         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2358                             np->n_fhp->nfh_len, 0);
2359                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2360                         *tl = txdr_unsigned(NFSV4OP_REMOVE);
2361                 }
2362         } else {
2363                 ret = 0;
2364         }
2365         if (ret == 0)
2366                 NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2367         (void) nfsm_strtom(nd, name, namelen);
2368         error = nfscl_request(nd, dvp, p, cred, dstuff);
2369         if (error)
2370                 return (error);
2371         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2372                 /* For NFSv4, parse out any Delereturn replies. */
2373                 if (ret > 0 && nd->nd_repstat != 0 &&
2374                     (nd->nd_flag & ND_NOMOREDATA)) {
2375                         /*
2376                          * If the Delegreturn failed, try again without
2377                          * it. The server will Recall, as required.
2378                          */
2379                         mbuf_freem(nd->nd_mrep);
2380                         goto tryagain;
2381                 }
2382                 for (i = 0; i < (ret * 2); i++) {
2383                         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2384                             ND_NFSV4) {
2385                             NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2386                             if (*(tl + 1))
2387                                 nd->nd_flag |= ND_NOMOREDATA;
2388                         }
2389                 }
2390                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2391         }
2392         if (nd->nd_repstat && !error)
2393                 error = nd->nd_repstat;
2394 nfsmout:
2395         mbuf_freem(nd->nd_mrep);
2396         return (error);
2397 }
2398
2399 /*
2400  * Do an nfs rename rpc.
2401  */
2402 APPLESTATIC int
2403 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2404     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2405     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2406     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2407 {
2408         u_int32_t *tl;
2409         struct nfsrv_descript nfsd, *nd = &nfsd;
2410         struct nfsmount *nmp;
2411         struct nfsnode *np;
2412         nfsattrbit_t attrbits;
2413         nfsv4stateid_t fdstateid, tdstateid;
2414         int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2415
2416         *fattrflagp = 0;
2417         *tattrflagp = 0;
2418         nmp = VFSTONFS(vnode_mount(fdvp));
2419         if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2420                 return (ENAMETOOLONG);
2421 tryagain:
2422         if (NFSHASNFSV4(nmp) && ret == 0) {
2423                 ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2424                     &tdstateid, &gottd, p);
2425                 if (gotfd && gottd) {
2426                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2427                 } else if (gotfd) {
2428                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2429                 } else if (gottd) {
2430                         NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2431                 }
2432                 if (gotfd) {
2433                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2434                         if (NFSHASNFSV4N(nmp))
2435                                 *tl++ = 0;
2436                         else
2437                                 *tl++ = fdstateid.seqid;
2438                         *tl++ = fdstateid.other[0];
2439                         *tl++ = fdstateid.other[1];
2440                         *tl = fdstateid.other[2];
2441                         if (gottd) {
2442                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2443                                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2444                                 np = VTONFS(tvp);
2445                                 (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2446                                     np->n_fhp->nfh_len, 0);
2447                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2448                                 *tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2449                         }
2450                 }
2451                 if (gottd) {
2452                         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2453                         if (NFSHASNFSV4N(nmp))
2454                                 *tl++ = 0;
2455                         else
2456                                 *tl++ = tdstateid.seqid;
2457                         *tl++ = tdstateid.other[0];
2458                         *tl++ = tdstateid.other[1];
2459                         *tl = tdstateid.other[2];
2460                 }
2461                 if (ret > 0) {
2462                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2463                         *tl = txdr_unsigned(NFSV4OP_PUTFH);
2464                         np = VTONFS(fdvp);
2465                         (void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2466                             np->n_fhp->nfh_len, 0);
2467                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2468                         *tl = txdr_unsigned(NFSV4OP_SAVEFH);
2469                 }
2470         } else {
2471                 ret = 0;
2472         }
2473         if (ret == 0)
2474                 NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2475         if (nd->nd_flag & ND_NFSV4) {
2476                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2477                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2478                 NFSWCCATTR_ATTRBIT(&attrbits);
2479                 (void) nfsrv_putattrbit(nd, &attrbits);
2480                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2481                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2482                 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2483                     VTONFS(tdvp)->n_fhp->nfh_len, 0);
2484                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2485                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2486                 (void) nfsrv_putattrbit(nd, &attrbits);
2487                 nd->nd_flag |= ND_V4WCCATTR;
2488                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2489                 *tl = txdr_unsigned(NFSV4OP_RENAME);
2490         }
2491         (void) nfsm_strtom(nd, fnameptr, fnamelen);
2492         if (!(nd->nd_flag & ND_NFSV4))
2493                 (void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2494                         VTONFS(tdvp)->n_fhp->nfh_len, 0);
2495         (void) nfsm_strtom(nd, tnameptr, tnamelen);
2496         error = nfscl_request(nd, fdvp, p, cred, fstuff);
2497         if (error)
2498                 return (error);
2499         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2500                 /* For NFSv4, parse out any Delereturn replies. */
2501                 if (ret > 0 && nd->nd_repstat != 0 &&
2502                     (nd->nd_flag & ND_NOMOREDATA)) {
2503                         /*
2504                          * If the Delegreturn failed, try again without
2505                          * it. The server will Recall, as required.
2506                          */
2507                         mbuf_freem(nd->nd_mrep);
2508                         goto tryagain;
2509                 }
2510                 for (i = 0; i < (ret * 2); i++) {
2511                         if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2512                             ND_NFSV4) {
2513                             NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2514                             if (*(tl + 1)) {
2515                                 if (i == 0 && ret > 1) {
2516                                     /*
2517                                      * If the Delegreturn failed, try again
2518                                      * without it. The server will Recall, as
2519                                      * required.
2520                                      * If ret > 1, the first iteration of this
2521                                      * loop is the second DelegReturn result.
2522                                      */
2523                                     mbuf_freem(nd->nd_mrep);
2524                                     goto tryagain;
2525                                 } else {
2526                                     nd->nd_flag |= ND_NOMOREDATA;
2527                                 }
2528                             }
2529                         }
2530                 }
2531                 /* Now, the first wcc attribute reply. */
2532                 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2533                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2534                         if (*(tl + 1))
2535                                 nd->nd_flag |= ND_NOMOREDATA;
2536                 }
2537                 error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2538                     fstuff);
2539                 /* and the second wcc attribute reply. */
2540                 if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2541                     !error) {
2542                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2543                         if (*(tl + 1))
2544                                 nd->nd_flag |= ND_NOMOREDATA;
2545                 }
2546                 if (!error)
2547                         error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2548                             NULL, tstuff);
2549         }
2550         if (nd->nd_repstat && !error)
2551                 error = nd->nd_repstat;
2552 nfsmout:
2553         mbuf_freem(nd->nd_mrep);
2554         return (error);
2555 }
2556
2557 /*
2558  * nfs hard link create rpc
2559  */
2560 APPLESTATIC int
2561 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2562     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2563     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2564 {
2565         u_int32_t *tl;
2566         struct nfsrv_descript nfsd, *nd = &nfsd;
2567         nfsattrbit_t attrbits;
2568         int error = 0;
2569
2570         *attrflagp = 0;
2571         *dattrflagp = 0;
2572         if (namelen > NFS_MAXNAMLEN)
2573                 return (ENAMETOOLONG);
2574         NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2575         if (nd->nd_flag & ND_NFSV4) {
2576                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2577                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2578         }
2579         (void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2580                 VTONFS(dvp)->n_fhp->nfh_len, 0);
2581         if (nd->nd_flag & ND_NFSV4) {
2582                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2583                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2584                 NFSWCCATTR_ATTRBIT(&attrbits);
2585                 (void) nfsrv_putattrbit(nd, &attrbits);
2586                 nd->nd_flag |= ND_V4WCCATTR;
2587                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2588                 *tl = txdr_unsigned(NFSV4OP_LINK);
2589         }
2590         (void) nfsm_strtom(nd, name, namelen);
2591         error = nfscl_request(nd, vp, p, cred, dstuff);
2592         if (error)
2593                 return (error);
2594         if (nd->nd_flag & ND_NFSV3) {
2595                 error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2596                 if (!error)
2597                         error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2598                             NULL, dstuff);
2599         } else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2600                 /*
2601                  * First, parse out the PutFH and Getattr result.
2602                  */
2603                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2604                 if (!(*(tl + 1)))
2605                         NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2606                 if (*(tl + 1))
2607                         nd->nd_flag |= ND_NOMOREDATA;
2608                 /*
2609                  * Get the pre-op attributes.
2610                  */
2611                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2612         }
2613         if (nd->nd_repstat && !error)
2614                 error = nd->nd_repstat;
2615 nfsmout:
2616         mbuf_freem(nd->nd_mrep);
2617         return (error);
2618 }
2619
2620 /*
2621  * nfs symbolic link create rpc
2622  */
2623 APPLESTATIC int
2624 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, char *target,
2625     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2626     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2627     int *dattrflagp, void *dstuff)
2628 {
2629         u_int32_t *tl;
2630         struct nfsrv_descript nfsd, *nd = &nfsd;
2631         struct nfsmount *nmp;
2632         int slen, error = 0;
2633
2634         *nfhpp = NULL;
2635         *attrflagp = 0;
2636         *dattrflagp = 0;
2637         nmp = VFSTONFS(vnode_mount(dvp));
2638         slen = strlen(target);
2639         if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2640                 return (ENAMETOOLONG);
2641         NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2642         if (nd->nd_flag & ND_NFSV4) {
2643                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2644                 *tl = txdr_unsigned(NFLNK);
2645                 (void) nfsm_strtom(nd, target, slen);
2646         }
2647         (void) nfsm_strtom(nd, name, namelen);
2648         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2649                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
2650         if (!(nd->nd_flag & ND_NFSV4))
2651                 (void) nfsm_strtom(nd, target, slen);
2652         if (nd->nd_flag & ND_NFSV2)
2653                 nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2654         error = nfscl_request(nd, dvp, p, cred, dstuff);
2655         if (error)
2656                 return (error);
2657         if (nd->nd_flag & ND_NFSV4)
2658                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2659         if ((nd->nd_flag & ND_NFSV3) && !error) {
2660                 if (!nd->nd_repstat)
2661                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2662                 if (!error)
2663                         error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2664                             NULL, dstuff);
2665         }
2666         if (nd->nd_repstat && !error)
2667                 error = nd->nd_repstat;
2668         mbuf_freem(nd->nd_mrep);
2669         /*
2670          * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2671          * Only do this if vfs.nfs.ignore_eexist is set.
2672          * Never do this for NFSv4.1 or later minor versions, since sessions
2673          * should guarantee "exactly once" RPC semantics.
2674          */
2675         if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2676             nmp->nm_minorvers == 0))
2677                 error = 0;
2678         return (error);
2679 }
2680
2681 /*
2682  * nfs make dir rpc
2683  */
2684 APPLESTATIC int
2685 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2686     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2687     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2688     int *dattrflagp, void *dstuff)
2689 {
2690         u_int32_t *tl;
2691         struct nfsrv_descript nfsd, *nd = &nfsd;
2692         nfsattrbit_t attrbits;
2693         int error = 0;
2694         struct nfsfh *fhp;
2695         struct nfsmount *nmp;
2696
2697         *nfhpp = NULL;
2698         *attrflagp = 0;
2699         *dattrflagp = 0;
2700         nmp = VFSTONFS(vnode_mount(dvp));
2701         fhp = VTONFS(dvp)->n_fhp;
2702         if (namelen > NFS_MAXNAMLEN)
2703                 return (ENAMETOOLONG);
2704         NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2705         if (nd->nd_flag & ND_NFSV4) {
2706                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2707                 *tl = txdr_unsigned(NFDIR);
2708         }
2709         (void) nfsm_strtom(nd, name, namelen);
2710         nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2711         if (nd->nd_flag & ND_NFSV4) {
2712                 NFSGETATTR_ATTRBIT(&attrbits);
2713                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2714                 *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2715                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2716                 (void) nfsrv_putattrbit(nd, &attrbits);
2717                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2718                 *tl = txdr_unsigned(NFSV4OP_PUTFH);
2719                 (void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2720                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2721                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
2722                 (void) nfsrv_putattrbit(nd, &attrbits);
2723         }
2724         error = nfscl_request(nd, dvp, p, cred, dstuff);
2725         if (error)
2726                 return (error);
2727         if (nd->nd_flag & ND_NFSV4)
2728                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2729         if (!nd->nd_repstat && !error) {
2730                 if (nd->nd_flag & ND_NFSV4) {
2731                         NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2732                         error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2733                 }
2734                 if (!error)
2735                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2736                 if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2737                         /* Get rid of the PutFH and Getattr status values. */
2738                         NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2739                         /* Load the directory attributes. */
2740                         error = nfsm_loadattr(nd, dnap);
2741                         if (error == 0)
2742                                 *dattrflagp = 1;
2743                 }
2744         }
2745         if ((nd->nd_flag & ND_NFSV3) && !error)
2746                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2747         if (nd->nd_repstat && !error)
2748                 error = nd->nd_repstat;
2749 nfsmout:
2750         mbuf_freem(nd->nd_mrep);
2751         /*
2752          * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2753          * Only do this if vfs.nfs.ignore_eexist is set.
2754          * Never do this for NFSv4.1 or later minor versions, since sessions
2755          * should guarantee "exactly once" RPC semantics.
2756          */
2757         if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2758             nmp->nm_minorvers == 0))
2759                 error = 0;
2760         return (error);
2761 }
2762
2763 /*
2764  * nfs remove directory call
2765  */
2766 APPLESTATIC int
2767 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2768     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2769 {
2770         struct nfsrv_descript nfsd, *nd = &nfsd;
2771         int error = 0;
2772
2773         *dattrflagp = 0;
2774         if (namelen > NFS_MAXNAMLEN)
2775                 return (ENAMETOOLONG);
2776         NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2777         (void) nfsm_strtom(nd, name, namelen);
2778         error = nfscl_request(nd, dvp, p, cred, dstuff);
2779         if (error)
2780                 return (error);
2781         if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2782                 error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2783         if (nd->nd_repstat && !error)
2784                 error = nd->nd_repstat;
2785         mbuf_freem(nd->nd_mrep);
2786         /*
2787          * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2788          */
2789         if (error == ENOENT)
2790                 error = 0;
2791         return (error);
2792 }
2793
2794 /*
2795  * Readdir rpc.
2796  * Always returns with either uio_resid unchanged, if you are at the
2797  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2798  * filled in.
2799  * I felt this would allow caching of directory blocks more easily
2800  * than returning a pertially filled block.
2801  * Directory offset cookies:
2802  * Oh my, what to do with them...
2803  * I can think of three ways to deal with them:
2804  * 1 - have the layer above these RPCs maintain a map between logical
2805  *     directory byte offsets and the NFS directory offset cookies
2806  * 2 - pass the opaque directory offset cookies up into userland
2807  *     and let the libc functions deal with them, via the system call
2808  * 3 - return them to userland in the "struct dirent", so future versions
2809  *     of libc can use them and do whatever is necessary to make things work
2810  *     above these rpc calls, in the meantime
2811  * For now, I do #3 by "hiding" the directory offset cookies after the
2812  * d_name field in struct dirent. This is space inside d_reclen that
2813  * will be ignored by anything that doesn't know about them.
2814  * The directory offset cookies are filled in as the last 8 bytes of
2815  * each directory entry, after d_name. Someday, the userland libc
2816  * functions may be able to use these. In the meantime, it satisfies
2817  * OpenBSD's requirements for cookies being returned.
2818  * If expects the directory offset cookie for the read to be in uio_offset
2819  * and returns the one for the next entry after this directory block in
2820  * there, as well.
2821  */
2822 APPLESTATIC int
2823 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2824     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2825     int *eofp, void *stuff)
2826 {
2827         int len, left;
2828         struct dirent *dp = NULL;
2829         u_int32_t *tl;
2830         nfsquad_t cookie, ncookie;
2831         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2832         struct nfsnode *dnp = VTONFS(vp);
2833         struct nfsvattr nfsva;
2834         struct nfsrv_descript nfsd, *nd = &nfsd;
2835         int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2836         int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2837         u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2838         char *cp;
2839         nfsattrbit_t attrbits, dattrbits;
2840         u_int32_t rderr, *tl2 = NULL;
2841         size_t tresid;
2842
2843         KASSERT(uiop->uio_iovcnt == 1 &&
2844             (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2845             ("nfs readdirrpc bad uio"));
2846
2847         /*
2848          * There is no point in reading a lot more than uio_resid, however
2849          * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2850          * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2851          * will never make readsize > nm_readdirsize.
2852          */
2853         readsize = nmp->nm_readdirsize;
2854         if (readsize > uio_uio_resid(uiop))
2855                 readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2856
2857         *attrflagp = 0;
2858         if (eofp)
2859                 *eofp = 0;
2860         tresid = uio_uio_resid(uiop);
2861         cookie.lval[0] = cookiep->nfsuquad[0];
2862         cookie.lval[1] = cookiep->nfsuquad[1];
2863         nd->nd_mrep = NULL;
2864
2865         /*
2866          * For NFSv4, first create the "." and ".." entries.
2867          */
2868         if (NFSHASNFSV4(nmp)) {
2869                 reqsize = 6 * NFSX_UNSIGNED;
2870                 NFSGETATTR_ATTRBIT(&dattrbits);
2871                 NFSZERO_ATTRBIT(&attrbits);
2872                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2873                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2874                 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2875                     NFSATTRBIT_MOUNTEDONFILEID)) {
2876                         NFSSETBIT_ATTRBIT(&attrbits,
2877                             NFSATTRBIT_MOUNTEDONFILEID);
2878                         gotmnton = 1;
2879                 } else {
2880                         /*
2881                          * Must fake it. Use the fileno, except when the
2882                          * fsid is != to that of the directory. For that
2883                          * case, generate a fake fileno that is not the same.
2884                          */
2885                         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2886                         gotmnton = 0;
2887                 }
2888
2889                 /*
2890                  * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2891                  */
2892                 if (uiop->uio_offset == 0) {
2893                         NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2894                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2895                         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
2896                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
2897                         (void) nfsrv_putattrbit(nd, &attrbits);
2898                         error = nfscl_request(nd, vp, p, cred, stuff);
2899                         if (error)
2900                             return (error);
2901                         dotfileid = 0;  /* Fake out the compiler. */
2902                         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2903                             error = nfsm_loadattr(nd, &nfsva);
2904                             if (error != 0)
2905                                 goto nfsmout;
2906                             dotfileid = nfsva.na_fileid;
2907                         }
2908                         if (nd->nd_repstat == 0) {
2909                             NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2910                             len = fxdr_unsigned(int, *(tl + 4));
2911                             if (len > 0 && len <= NFSX_V4FHMAX)
2912                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2913                             else
2914                                 error = EPERM;
2915                             if (!error) {
2916                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2917                                 nfsva.na_mntonfileno = UINT64_MAX;
2918                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2919                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2920                                     NULL, NULL, NULL, p, cred);
2921                                 if (error) {
2922                                     dotdotfileid = dotfileid;
2923                                 } else if (gotmnton) {
2924                                     if (nfsva.na_mntonfileno != UINT64_MAX)
2925                                         dotdotfileid = nfsva.na_mntonfileno;
2926                                     else
2927                                         dotdotfileid = nfsva.na_fileid;
2928                                 } else if (nfsva.na_filesid[0] ==
2929                                     dnp->n_vattr.na_filesid[0] &&
2930                                     nfsva.na_filesid[1] ==
2931                                     dnp->n_vattr.na_filesid[1]) {
2932                                     dotdotfileid = nfsva.na_fileid;
2933                                 } else {
2934                                     do {
2935                                         fakefileno--;
2936                                     } while (fakefileno ==
2937                                         nfsva.na_fileid);
2938                                     dotdotfileid = fakefileno;
2939                                 }
2940                             }
2941                         } else if (nd->nd_repstat == NFSERR_NOENT) {
2942                             /*
2943                              * Lookupp returns NFSERR_NOENT when we are
2944                              * at the root, so just use the current dir.
2945                              */
2946                             nd->nd_repstat = 0;
2947                             dotdotfileid = dotfileid;
2948                         } else {
2949                             error = nd->nd_repstat;
2950                         }
2951                         mbuf_freem(nd->nd_mrep);
2952                         if (error)
2953                             return (error);
2954                         nd->nd_mrep = NULL;
2955                         dp = (struct dirent *)uio_iov_base(uiop);
2956                         dp->d_off = 0;
2957                         dp->d_type = DT_DIR;
2958                         dp->d_fileno = dotfileid;
2959                         dp->d_namlen = 1;
2960                         *((uint64_t *)dp->d_name) = 0;  /* Zero pad it. */
2961                         dp->d_name[0] = '.';
2962                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2963                         /*
2964                          * Just make these offset cookie 0.
2965                          */
2966                         tl = (u_int32_t *)&dp->d_name[8];
2967                         *tl++ = 0;
2968                         *tl = 0;
2969                         blksiz += dp->d_reclen;
2970                         uio_uio_resid_add(uiop, -(dp->d_reclen));
2971                         uiop->uio_offset += dp->d_reclen;
2972                         uio_iov_base_add(uiop, dp->d_reclen);
2973                         uio_iov_len_add(uiop, -(dp->d_reclen));
2974                         dp = (struct dirent *)uio_iov_base(uiop);
2975                         dp->d_off = 0;
2976                         dp->d_type = DT_DIR;
2977                         dp->d_fileno = dotdotfileid;
2978                         dp->d_namlen = 2;
2979                         *((uint64_t *)dp->d_name) = 0;
2980                         dp->d_name[0] = '.';
2981                         dp->d_name[1] = '.';
2982                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
2983                         /*
2984                          * Just make these offset cookie 0.
2985                          */
2986                         tl = (u_int32_t *)&dp->d_name[8];
2987                         *tl++ = 0;
2988                         *tl = 0;
2989                         blksiz += dp->d_reclen;
2990                         uio_uio_resid_add(uiop, -(dp->d_reclen));
2991                         uiop->uio_offset += dp->d_reclen;
2992                         uio_iov_base_add(uiop, dp->d_reclen);
2993                         uio_iov_len_add(uiop, -(dp->d_reclen));
2994                 }
2995                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
2996         } else {
2997                 reqsize = 5 * NFSX_UNSIGNED;
2998         }
2999
3000
3001         /*
3002          * Loop around doing readdir rpc's of size readsize.
3003          * The stopping criteria is EOF or buffer full.
3004          */
3005         while (more_dirs && bigenough) {
3006                 *attrflagp = 0;
3007                 NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3008                 if (nd->nd_flag & ND_NFSV2) {
3009                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3010                         *tl++ = cookie.lval[1];
3011                         *tl = txdr_unsigned(readsize);
3012                 } else {
3013                         NFSM_BUILD(tl, u_int32_t *, reqsize);
3014                         *tl++ = cookie.lval[0];
3015                         *tl++ = cookie.lval[1];
3016                         if (cookie.qval == 0) {
3017                                 *tl++ = 0;
3018                                 *tl++ = 0;
3019                         } else {
3020                                 NFSLOCKNODE(dnp);
3021                                 *tl++ = dnp->n_cookieverf.nfsuquad[0];
3022                                 *tl++ = dnp->n_cookieverf.nfsuquad[1];
3023                                 NFSUNLOCKNODE(dnp);
3024                         }
3025                         if (nd->nd_flag & ND_NFSV4) {
3026                                 *tl++ = txdr_unsigned(readsize);
3027                                 *tl = txdr_unsigned(readsize);
3028                                 (void) nfsrv_putattrbit(nd, &attrbits);
3029                                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3030                                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3031                                 (void) nfsrv_putattrbit(nd, &dattrbits);
3032                         } else {
3033                                 *tl = txdr_unsigned(readsize);
3034                         }
3035                 }
3036                 error = nfscl_request(nd, vp, p, cred, stuff);
3037                 if (error)
3038                         return (error);
3039                 if (!(nd->nd_flag & ND_NFSV2)) {
3040                         if (nd->nd_flag & ND_NFSV3)
3041                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3042                                     stuff);
3043                         if (!nd->nd_repstat && !error) {
3044                                 NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3045                                 NFSLOCKNODE(dnp);
3046                                 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3047                                 dnp->n_cookieverf.nfsuquad[1] = *tl;
3048                                 NFSUNLOCKNODE(dnp);
3049                         }
3050                 }
3051                 if (nd->nd_repstat || error) {
3052                         if (!error)
3053                                 error = nd->nd_repstat;
3054                         goto nfsmout;
3055                 }
3056                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3057                 more_dirs = fxdr_unsigned(int, *tl);
3058                 if (!more_dirs)
3059                         tryformoredirs = 0;
3060
3061                 /* loop through the dir entries, doctoring them to 4bsd form */
3062                 while (more_dirs && bigenough) {
3063                         if (nd->nd_flag & ND_NFSV4) {
3064                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3065                                 ncookie.lval[0] = *tl++;
3066                                 ncookie.lval[1] = *tl++;
3067                                 len = fxdr_unsigned(int, *tl);
3068                         } else if (nd->nd_flag & ND_NFSV3) {
3069                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3070                                 nfsva.na_fileid = fxdr_hyper(tl);
3071                                 tl += 2;
3072                                 len = fxdr_unsigned(int, *tl);
3073                         } else {
3074                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3075                                 nfsva.na_fileid = fxdr_unsigned(uint64_t,
3076                                     *tl++);
3077                                 len = fxdr_unsigned(int, *tl);
3078                         }
3079                         if (len <= 0 || len > NFS_MAXNAMLEN) {
3080                                 error = EBADRPC;
3081                                 goto nfsmout;
3082                         }
3083                         tlen = roundup2(len, 8);
3084                         if (tlen == len)
3085                                 tlen += 8;  /* To ensure null termination. */
3086                         left = DIRBLKSIZ - blksiz;
3087                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3088                                 dp->d_reclen += left;
3089                                 uio_iov_base_add(uiop, left);
3090                                 uio_iov_len_add(uiop, -(left));
3091                                 uio_uio_resid_add(uiop, -(left));
3092                                 uiop->uio_offset += left;
3093                                 blksiz = 0;
3094                         }
3095                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3096                             uio_uio_resid(uiop))
3097                                 bigenough = 0;
3098                         if (bigenough) {
3099                                 dp = (struct dirent *)uio_iov_base(uiop);
3100                                 dp->d_off = 0;
3101                                 dp->d_namlen = len;
3102                                 dp->d_reclen = _GENERIC_DIRLEN(len) +
3103                                     NFSX_HYPER;
3104                                 dp->d_type = DT_UNKNOWN;
3105                                 blksiz += dp->d_reclen;
3106                                 if (blksiz == DIRBLKSIZ)
3107                                         blksiz = 0;
3108                                 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3109                                 uiop->uio_offset += DIRHDSIZ;
3110                                 uio_iov_base_add(uiop, DIRHDSIZ);
3111                                 uio_iov_len_add(uiop, -(DIRHDSIZ));
3112                                 error = nfsm_mbufuio(nd, uiop, len);
3113                                 if (error)
3114                                         goto nfsmout;
3115                                 cp = uio_iov_base(uiop);
3116                                 tlen -= len;
3117                                 *cp = '\0';     /* null terminate */
3118                                 cp += tlen;     /* points to cookie storage */
3119                                 tl2 = (u_int32_t *)cp;
3120                                 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3121                                 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3122                                 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3123                                 uiop->uio_offset += (tlen + NFSX_HYPER);
3124                         } else {
3125                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3126                                 if (error)
3127                                         goto nfsmout;
3128                         }
3129                         if (nd->nd_flag & ND_NFSV4) {
3130                                 rderr = 0;
3131                                 nfsva.na_mntonfileno = UINT64_MAX;
3132                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3133                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3134                                     NULL, NULL, &rderr, p, cred);
3135                                 if (error)
3136                                         goto nfsmout;
3137                                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3138                         } else if (nd->nd_flag & ND_NFSV3) {
3139                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3140                                 ncookie.lval[0] = *tl++;
3141                                 ncookie.lval[1] = *tl++;
3142                         } else {
3143                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3144                                 ncookie.lval[0] = 0;
3145                                 ncookie.lval[1] = *tl++;
3146                         }
3147                         if (bigenough) {
3148                             if (nd->nd_flag & ND_NFSV4) {
3149                                 if (rderr) {
3150                                     dp->d_fileno = 0;
3151                                 } else {
3152                                     if (gotmnton) {
3153                                         if (nfsva.na_mntonfileno != UINT64_MAX)
3154                                             dp->d_fileno = nfsva.na_mntonfileno;
3155                                         else
3156                                             dp->d_fileno = nfsva.na_fileid;
3157                                     } else if (nfsva.na_filesid[0] ==
3158                                         dnp->n_vattr.na_filesid[0] &&
3159                                         nfsva.na_filesid[1] ==
3160                                         dnp->n_vattr.na_filesid[1]) {
3161                                         dp->d_fileno = nfsva.na_fileid;
3162                                     } else {
3163                                         do {
3164                                             fakefileno--;
3165                                         } while (fakefileno ==
3166                                             nfsva.na_fileid);
3167                                         dp->d_fileno = fakefileno;
3168                                     }
3169                                     dp->d_type = vtonfs_dtype(nfsva.na_type);
3170                                 }
3171                             } else {
3172                                 dp->d_fileno = nfsva.na_fileid;
3173                             }
3174                             *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3175                                 ncookie.lval[0];
3176                             *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3177                                 ncookie.lval[1];
3178                         }
3179                         more_dirs = fxdr_unsigned(int, *tl);
3180                 }
3181                 /*
3182                  * If at end of rpc data, get the eof boolean
3183                  */
3184                 if (!more_dirs) {
3185                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3186                         eof = fxdr_unsigned(int, *tl);
3187                         if (tryformoredirs)
3188                                 more_dirs = !eof;
3189                         if (nd->nd_flag & ND_NFSV4) {
3190                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3191                                     stuff);
3192                                 if (error)
3193                                         goto nfsmout;
3194                         }
3195                 }
3196                 mbuf_freem(nd->nd_mrep);
3197                 nd->nd_mrep = NULL;
3198         }
3199         /*
3200          * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3201          * by increasing d_reclen for the last record.
3202          */
3203         if (blksiz > 0) {
3204                 left = DIRBLKSIZ - blksiz;
3205                 dp->d_reclen += left;
3206                 uio_iov_base_add(uiop, left);
3207                 uio_iov_len_add(uiop, -(left));
3208                 uio_uio_resid_add(uiop, -(left));
3209                 uiop->uio_offset += left;
3210         }
3211
3212         /*
3213          * If returning no data, assume end of file.
3214          * If not bigenough, return not end of file, since you aren't
3215          *    returning all the data
3216          * Otherwise, return the eof flag from the server.
3217          */
3218         if (eofp) {
3219                 if (tresid == ((size_t)(uio_uio_resid(uiop))))
3220                         *eofp = 1;
3221                 else if (!bigenough)
3222                         *eofp = 0;
3223                 else
3224                         *eofp = eof;
3225         }
3226
3227         /*
3228          * Add extra empty records to any remaining DIRBLKSIZ chunks.
3229          */
3230         while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3231                 dp = (struct dirent *)uio_iov_base(uiop);
3232                 dp->d_type = DT_UNKNOWN;
3233                 dp->d_fileno = 0;
3234                 dp->d_namlen = 0;
3235                 dp->d_name[0] = '\0';
3236                 tl = (u_int32_t *)&dp->d_name[4];
3237                 *tl++ = cookie.lval[0];
3238                 *tl = cookie.lval[1];
3239                 dp->d_reclen = DIRBLKSIZ;
3240                 uio_iov_base_add(uiop, DIRBLKSIZ);
3241                 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3242                 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3243                 uiop->uio_offset += DIRBLKSIZ;
3244         }
3245
3246 nfsmout:
3247         if (nd->nd_mrep != NULL)
3248                 mbuf_freem(nd->nd_mrep);
3249         return (error);
3250 }
3251
3252 #ifndef APPLE
3253 /*
3254  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3255  * (Also used for NFS V4 when mount flag set.)
3256  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3257  */
3258 APPLESTATIC int
3259 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3260     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3261     int *eofp, void *stuff)
3262 {
3263         int len, left;
3264         struct dirent *dp = NULL;
3265         u_int32_t *tl;
3266         vnode_t newvp = NULLVP;
3267         struct nfsrv_descript nfsd, *nd = &nfsd;
3268         struct nameidata nami, *ndp = &nami;
3269         struct componentname *cnp = &ndp->ni_cnd;
3270         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3271         struct nfsnode *dnp = VTONFS(vp), *np;
3272         struct nfsvattr nfsva;
3273         struct nfsfh *nfhp;
3274         nfsquad_t cookie, ncookie;
3275         int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3276         int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3277         int isdotdot = 0, unlocknewvp = 0;
3278         u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3279         u_int64_t fileno = 0;
3280         char *cp;
3281         nfsattrbit_t attrbits, dattrbits;
3282         size_t tresid;
3283         u_int32_t *tl2 = NULL, rderr;
3284         struct timespec dctime;
3285
3286         KASSERT(uiop->uio_iovcnt == 1 &&
3287             (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3288             ("nfs readdirplusrpc bad uio"));
3289         timespecclear(&dctime);
3290         *attrflagp = 0;
3291         if (eofp != NULL)
3292                 *eofp = 0;
3293         ndp->ni_dvp = vp;
3294         nd->nd_mrep = NULL;
3295         cookie.lval[0] = cookiep->nfsuquad[0];
3296         cookie.lval[1] = cookiep->nfsuquad[1];
3297         tresid = uio_uio_resid(uiop);
3298
3299         /*
3300          * For NFSv4, first create the "." and ".." entries.
3301          */
3302         if (NFSHASNFSV4(nmp)) {
3303                 NFSGETATTR_ATTRBIT(&dattrbits);
3304                 NFSZERO_ATTRBIT(&attrbits);
3305                 NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3306                 if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3307                     NFSATTRBIT_MOUNTEDONFILEID)) {
3308                         NFSSETBIT_ATTRBIT(&attrbits,
3309                             NFSATTRBIT_MOUNTEDONFILEID);
3310                         gotmnton = 1;
3311                 } else {
3312                         /*
3313                          * Must fake it. Use the fileno, except when the
3314                          * fsid is != to that of the directory. For that
3315                          * case, generate a fake fileno that is not the same.
3316                          */
3317                         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3318                         gotmnton = 0;
3319                 }
3320
3321                 /*
3322                  * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3323                  */
3324                 if (uiop->uio_offset == 0) {
3325                         NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3326                         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3327                         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
3328                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
3329                         (void) nfsrv_putattrbit(nd, &attrbits);
3330                         error = nfscl_request(nd, vp, p, cred, stuff);
3331                         if (error)
3332                             return (error);
3333                         dotfileid = 0;  /* Fake out the compiler. */
3334                         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3335                             error = nfsm_loadattr(nd, &nfsva);
3336                             if (error != 0)
3337                                 goto nfsmout;
3338                             dctime = nfsva.na_ctime;
3339                             dotfileid = nfsva.na_fileid;
3340                         }
3341                         if (nd->nd_repstat == 0) {
3342                             NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3343                             len = fxdr_unsigned(int, *(tl + 4));
3344                             if (len > 0 && len <= NFSX_V4FHMAX)
3345                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3346                             else
3347                                 error = EPERM;
3348                             if (!error) {
3349                                 NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3350                                 nfsva.na_mntonfileno = UINT64_MAX;
3351                                 error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3352                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3353                                     NULL, NULL, NULL, p, cred);
3354                                 if (error) {
3355                                     dotdotfileid = dotfileid;
3356                                 } else if (gotmnton) {
3357                                     if (nfsva.na_mntonfileno != UINT64_MAX)
3358                                         dotdotfileid = nfsva.na_mntonfileno;
3359                                     else
3360                                         dotdotfileid = nfsva.na_fileid;
3361                                 } else if (nfsva.na_filesid[0] ==
3362                                     dnp->n_vattr.na_filesid[0] &&
3363                                     nfsva.na_filesid[1] ==
3364                                     dnp->n_vattr.na_filesid[1]) {
3365                                     dotdotfileid = nfsva.na_fileid;
3366                                 } else {
3367                                     do {
3368                                         fakefileno--;
3369                                     } while (fakefileno ==
3370                                         nfsva.na_fileid);
3371                                     dotdotfileid = fakefileno;
3372                                 }
3373                             }
3374                         } else if (nd->nd_repstat == NFSERR_NOENT) {
3375                             /*
3376                              * Lookupp returns NFSERR_NOENT when we are
3377                              * at the root, so just use the current dir.
3378                              */
3379                             nd->nd_repstat = 0;
3380                             dotdotfileid = dotfileid;
3381                         } else {
3382                             error = nd->nd_repstat;
3383                         }
3384                         mbuf_freem(nd->nd_mrep);
3385                         if (error)
3386                             return (error);
3387                         nd->nd_mrep = NULL;
3388                         dp = (struct dirent *)uio_iov_base(uiop);
3389                         dp->d_off = 0;
3390                         dp->d_type = DT_DIR;
3391                         dp->d_fileno = dotfileid;
3392                         dp->d_namlen = 1;
3393                         *((uint64_t *)dp->d_name) = 0;  /* Zero pad it. */
3394                         dp->d_name[0] = '.';
3395                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3396                         /*
3397                          * Just make these offset cookie 0.
3398                          */
3399                         tl = (u_int32_t *)&dp->d_name[8];
3400                         *tl++ = 0;
3401                         *tl = 0;
3402                         blksiz += dp->d_reclen;
3403                         uio_uio_resid_add(uiop, -(dp->d_reclen));
3404                         uiop->uio_offset += dp->d_reclen;
3405                         uio_iov_base_add(uiop, dp->d_reclen);
3406                         uio_iov_len_add(uiop, -(dp->d_reclen));
3407                         dp = (struct dirent *)uio_iov_base(uiop);
3408                         dp->d_off = 0;
3409                         dp->d_type = DT_DIR;
3410                         dp->d_fileno = dotdotfileid;
3411                         dp->d_namlen = 2;
3412                         *((uint64_t *)dp->d_name) = 0;
3413                         dp->d_name[0] = '.';
3414                         dp->d_name[1] = '.';
3415                         dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3416                         /*
3417                          * Just make these offset cookie 0.
3418                          */
3419                         tl = (u_int32_t *)&dp->d_name[8];
3420                         *tl++ = 0;
3421                         *tl = 0;
3422                         blksiz += dp->d_reclen;
3423                         uio_uio_resid_add(uiop, -(dp->d_reclen));
3424                         uiop->uio_offset += dp->d_reclen;
3425                         uio_iov_base_add(uiop, dp->d_reclen);
3426                         uio_iov_len_add(uiop, -(dp->d_reclen));
3427                 }
3428                 NFSREADDIRPLUS_ATTRBIT(&attrbits);
3429                 if (gotmnton)
3430                         NFSSETBIT_ATTRBIT(&attrbits,
3431                             NFSATTRBIT_MOUNTEDONFILEID);
3432         }
3433
3434         /*
3435          * Loop around doing readdir rpc's of size nm_readdirsize.
3436          * The stopping criteria is EOF or buffer full.
3437          */
3438         while (more_dirs && bigenough) {
3439                 *attrflagp = 0;
3440                 NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3441                 NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3442                 *tl++ = cookie.lval[0];
3443                 *tl++ = cookie.lval[1];
3444                 if (cookie.qval == 0) {
3445                         *tl++ = 0;
3446                         *tl++ = 0;
3447                 } else {
3448                         NFSLOCKNODE(dnp);
3449                         *tl++ = dnp->n_cookieverf.nfsuquad[0];
3450                         *tl++ = dnp->n_cookieverf.nfsuquad[1];
3451                         NFSUNLOCKNODE(dnp);
3452                 }
3453                 *tl++ = txdr_unsigned(nmp->nm_readdirsize);
3454                 *tl = txdr_unsigned(nmp->nm_readdirsize);
3455                 if (nd->nd_flag & ND_NFSV4) {
3456                         (void) nfsrv_putattrbit(nd, &attrbits);
3457                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3458                         *tl = txdr_unsigned(NFSV4OP_GETATTR);
3459                         (void) nfsrv_putattrbit(nd, &dattrbits);
3460                 }
3461                 error = nfscl_request(nd, vp, p, cred, stuff);
3462                 if (error)
3463                         return (error);
3464                 if (nd->nd_flag & ND_NFSV3)
3465                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3466                 if (nd->nd_repstat || error) {
3467                         if (!error)
3468                                 error = nd->nd_repstat;
3469                         goto nfsmout;
3470                 }
3471                 if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3472                         dctime = nap->na_ctime;
3473                 NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3474                 NFSLOCKNODE(dnp);
3475                 dnp->n_cookieverf.nfsuquad[0] = *tl++;
3476                 dnp->n_cookieverf.nfsuquad[1] = *tl++;
3477                 NFSUNLOCKNODE(dnp);
3478                 more_dirs = fxdr_unsigned(int, *tl);
3479                 if (!more_dirs)
3480                         tryformoredirs = 0;
3481
3482                 /* loop through the dir entries, doctoring them to 4bsd form */
3483                 while (more_dirs && bigenough) {
3484                         NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3485                         if (nd->nd_flag & ND_NFSV4) {
3486                                 ncookie.lval[0] = *tl++;
3487                                 ncookie.lval[1] = *tl++;
3488                         } else {
3489                                 fileno = fxdr_hyper(tl);
3490                                 tl += 2;
3491                         }
3492                         len = fxdr_unsigned(int, *tl);
3493                         if (len <= 0 || len > NFS_MAXNAMLEN) {
3494                                 error = EBADRPC;
3495                                 goto nfsmout;
3496                         }
3497                         tlen = roundup2(len, 8);
3498                         if (tlen == len)
3499                                 tlen += 8;  /* To ensure null termination. */
3500                         left = DIRBLKSIZ - blksiz;
3501                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3502                                 dp->d_reclen += left;
3503                                 uio_iov_base_add(uiop, left);
3504                                 uio_iov_len_add(uiop, -(left));
3505                                 uio_uio_resid_add(uiop, -(left));
3506                                 uiop->uio_offset += left;
3507                                 blksiz = 0;
3508                         }
3509                         if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3510                             uio_uio_resid(uiop))
3511                                 bigenough = 0;
3512                         if (bigenough) {
3513                                 dp = (struct dirent *)uio_iov_base(uiop);
3514                                 dp->d_off = 0;
3515                                 dp->d_namlen = len;
3516                                 dp->d_reclen = _GENERIC_DIRLEN(len) +
3517                                     NFSX_HYPER;
3518                                 dp->d_type = DT_UNKNOWN;
3519                                 blksiz += dp->d_reclen;
3520                                 if (blksiz == DIRBLKSIZ)
3521                                         blksiz = 0;
3522                                 uio_uio_resid_add(uiop, -(DIRHDSIZ));
3523                                 uiop->uio_offset += DIRHDSIZ;
3524                                 uio_iov_base_add(uiop, DIRHDSIZ);
3525                                 uio_iov_len_add(uiop, -(DIRHDSIZ));
3526                                 cnp->cn_nameptr = uio_iov_base(uiop);
3527                                 cnp->cn_namelen = len;
3528                                 NFSCNHASHZERO(cnp);
3529                                 error = nfsm_mbufuio(nd, uiop, len);
3530                                 if (error)
3531                                         goto nfsmout;
3532                                 cp = uio_iov_base(uiop);
3533                                 tlen -= len;
3534                                 *cp = '\0';
3535                                 cp += tlen;     /* points to cookie storage */
3536                                 tl2 = (u_int32_t *)cp;
3537                                 if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3538                                     cnp->cn_nameptr[1] == '.')
3539                                         isdotdot = 1;
3540                                 else
3541                                         isdotdot = 0;
3542                                 uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3543                                 uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3544                                 uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3545                                 uiop->uio_offset += (tlen + NFSX_HYPER);
3546                         } else {
3547                                 error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3548                                 if (error)
3549                                         goto nfsmout;
3550                         }
3551                         nfhp = NULL;
3552                         if (nd->nd_flag & ND_NFSV3) {
3553                                 NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3554                                 ncookie.lval[0] = *tl++;
3555                                 ncookie.lval[1] = *tl++;
3556                                 attrflag = fxdr_unsigned(int, *tl);
3557                                 if (attrflag) {
3558                                   error = nfsm_loadattr(nd, &nfsva);
3559                                   if (error)
3560                                         goto nfsmout;
3561                                 }
3562                                 NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3563                                 if (*tl) {
3564                                         error = nfsm_getfh(nd, &nfhp);
3565                                         if (error)
3566                                             goto nfsmout;
3567                                 }
3568                                 if (!attrflag && nfhp != NULL) {
3569                                         FREE((caddr_t)nfhp, M_NFSFH);
3570                                         nfhp = NULL;
3571                                 }
3572                         } else {
3573                                 rderr = 0;
3574                                 nfsva.na_mntonfileno = 0xffffffff;
3575                                 error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3576                                     NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3577                                     NULL, NULL, &rderr, p, cred);
3578                                 if (error)
3579                                         goto nfsmout;
3580                         }
3581
3582                         if (bigenough) {
3583                             if (nd->nd_flag & ND_NFSV4) {
3584                                 if (rderr) {
3585                                     dp->d_fileno = 0;
3586                                 } else if (gotmnton) {
3587                                     if (nfsva.na_mntonfileno != 0xffffffff)
3588                                         dp->d_fileno = nfsva.na_mntonfileno;
3589                                     else
3590                                         dp->d_fileno = nfsva.na_fileid;
3591                                 } else if (nfsva.na_filesid[0] ==
3592                                     dnp->n_vattr.na_filesid[0] &&
3593                                     nfsva.na_filesid[1] ==
3594                                     dnp->n_vattr.na_filesid[1]) {
3595                                     dp->d_fileno = nfsva.na_fileid;
3596                                 } else {
3597                                     do {
3598                                         fakefileno--;
3599                                     } while (fakefileno ==
3600                                         nfsva.na_fileid);
3601                                     dp->d_fileno = fakefileno;
3602                                 }
3603                             } else {
3604                                 dp->d_fileno = fileno;
3605                             }
3606                             *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3607                                 ncookie.lval[0];
3608                             *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3609                                 ncookie.lval[1];
3610
3611                             if (nfhp != NULL) {
3612                                 if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3613                                     dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3614                                     VREF(vp);
3615                                     newvp = vp;
3616                                     unlocknewvp = 0;
3617                                     FREE((caddr_t)nfhp, M_NFSFH);
3618                                     np = dnp;
3619                                 } else if (isdotdot != 0) {
3620                                     /*
3621                                      * Skip doing a nfscl_nget() call for "..".
3622                                      * There's a race between acquiring the nfs
3623                                      * node here and lookups that look for the
3624                                      * directory being read (in the parent).
3625                                      * It would try to get a lock on ".." here,
3626                                      * owning the lock on the directory being
3627                                      * read. Lookup will hold the lock on ".."
3628                                      * and try to acquire the lock on the
3629                                      * directory being read.
3630                                      * If the directory is unlocked/relocked,
3631                                      * then there is a LOR with the buflock
3632                                      * vp is relocked.
3633                                      */
3634                                     free(nfhp, M_NFSFH);
3635                                 } else {
3636                                     error = nfscl_nget(vnode_mount(vp), vp,
3637                                       nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3638                                     if (!error) {
3639                                         newvp = NFSTOV(np);
3640                                         unlocknewvp = 1;
3641                                     }
3642                                 }
3643                                 nfhp = NULL;
3644                                 if (newvp != NULLVP) {
3645                                     error = nfscl_loadattrcache(&newvp,
3646                                         &nfsva, NULL, NULL, 0, 0);
3647                                     if (error) {
3648                                         if (unlocknewvp)
3649                                             vput(newvp);
3650                                         else
3651                                             vrele(newvp);
3652                                         goto nfsmout;
3653                                     }
3654                                     dp->d_type =
3655                                         vtonfs_dtype(np->n_vattr.na_type);
3656                                     ndp->ni_vp = newvp;
3657                                     NFSCNHASH(cnp, HASHINIT);
3658                                     if (cnp->cn_namelen <= NCHNAMLEN &&
3659                                         (newvp->v_type != VDIR ||
3660                                          dctime.tv_sec != 0)) {
3661                                         cache_enter_time(ndp->ni_dvp,
3662                                             ndp->ni_vp, cnp,
3663                                             &nfsva.na_ctime,
3664                                             newvp->v_type != VDIR ? NULL :
3665                                             &dctime);
3666                                     }
3667                                     if (unlocknewvp)
3668                                         vput(newvp);
3669                                     else
3670                                         vrele(newvp);
3671                                     newvp = NULLVP;
3672                                 }
3673                             }
3674                         } else if (nfhp != NULL) {
3675                             FREE((caddr_t)nfhp, M_NFSFH);
3676                         }
3677                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3678                         more_dirs = fxdr_unsigned(int, *tl);
3679                 }
3680                 /*
3681                  * If at end of rpc data, get the eof boolean
3682                  */
3683                 if (!more_dirs) {
3684                         NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3685                         eof = fxdr_unsigned(int, *tl);
3686                         if (tryformoredirs)
3687                                 more_dirs = !eof;
3688                         if (nd->nd_flag & ND_NFSV4) {
3689                                 error = nfscl_postop_attr(nd, nap, attrflagp,
3690                                     stuff);
3691                                 if (error)
3692                                         goto nfsmout;
3693                         }
3694                 }
3695                 mbuf_freem(nd->nd_mrep);
3696                 nd->nd_mrep = NULL;
3697         }
3698         /*
3699          * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3700          * by increasing d_reclen for the last record.
3701          */
3702         if (blksiz > 0) {
3703                 left = DIRBLKSIZ - blksiz;
3704                 dp->d_reclen += left;
3705                 uio_iov_base_add(uiop, left);
3706                 uio_iov_len_add(uiop, -(left));
3707                 uio_uio_resid_add(uiop, -(left));
3708                 uiop->uio_offset += left;
3709         }
3710
3711         /*
3712          * If returning no data, assume end of file.
3713          * If not bigenough, return not end of file, since you aren't
3714          *    returning all the data
3715          * Otherwise, return the eof flag from the server.
3716          */
3717         if (eofp != NULL) {
3718                 if (tresid == uio_uio_resid(uiop))
3719                         *eofp = 1;
3720                 else if (!bigenough)
3721                         *eofp = 0;
3722                 else
3723                         *eofp = eof;
3724         }
3725
3726         /*
3727          * Add extra empty records to any remaining DIRBLKSIZ chunks.
3728          */
3729         while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3730                 dp = (struct dirent *)uio_iov_base(uiop);
3731                 dp->d_type = DT_UNKNOWN;
3732                 dp->d_fileno = 0;
3733                 dp->d_namlen = 0;
3734                 dp->d_name[0] = '\0';
3735                 tl = (u_int32_t *)&dp->d_name[4];
3736                 *tl++ = cookie.lval[0];
3737                 *tl = cookie.lval[1];
3738                 dp->d_reclen = DIRBLKSIZ;
3739                 uio_iov_base_add(uiop, DIRBLKSIZ);
3740                 uio_iov_len_add(uiop, -(DIRBLKSIZ));
3741                 uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3742                 uiop->uio_offset += DIRBLKSIZ;
3743         }
3744
3745 nfsmout:
3746         if (nd->nd_mrep != NULL)
3747                 mbuf_freem(nd->nd_mrep);
3748         return (error);
3749 }
3750 #endif  /* !APPLE */
3751
3752 /*
3753  * Nfs commit rpc
3754  */
3755 APPLESTATIC int
3756 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3757     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3758 {
3759         u_int32_t *tl;
3760         struct nfsrv_descript nfsd, *nd = &nfsd;
3761         nfsattrbit_t attrbits;
3762         int error;
3763         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3764
3765         *attrflagp = 0;
3766         NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3767         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3768         txdr_hyper(offset, tl);
3769         tl += 2;
3770         *tl = txdr_unsigned(cnt);
3771         if (nd->nd_flag & ND_NFSV4) {
3772                 /*
3773                  * And do a Getattr op.
3774                  */
3775                 NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3776                 *tl = txdr_unsigned(NFSV4OP_GETATTR);
3777                 NFSGETATTR_ATTRBIT(&attrbits);
3778                 (void) nfsrv_putattrbit(nd, &attrbits);
3779         }
3780         error = nfscl_request(nd, vp, p, cred, stuff);
3781         if (error)
3782                 return (error);
3783         error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3784         if (!error && !nd->nd_repstat) {
3785                 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3786                 NFSLOCKMNT(nmp);
3787                 if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3788                         NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3789                         nd->nd_repstat = NFSERR_STALEWRITEVERF;
3790                 }
3791                 NFSUNLOCKMNT(nmp);
3792                 if (nd->nd_flag & ND_NFSV4)
3793                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3794         }
3795 nfsmout:
3796         if (!error && nd->nd_repstat)
3797                 error = nd->nd_repstat;
3798         mbuf_freem(nd->nd_mrep);
3799         return (error);
3800 }
3801
3802 /*
3803  * NFS byte range lock rpc.
3804  * (Mostly just calls one of the three lower level RPC routines.)
3805  */
3806 APPLESTATIC int
3807 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3808     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3809 {
3810         struct nfscllockowner *lp;
3811         struct nfsclclient *clp;
3812         struct nfsfh *nfhp;
3813         struct nfsrv_descript nfsd, *nd = &nfsd;
3814         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3815         u_int64_t off, len;
3816         off_t start, end;
3817         u_int32_t clidrev = 0;
3818         int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3819         int callcnt, dorpc;
3820
3821         /*
3822          * Convert the flock structure into a start and end and do POSIX
3823          * bounds checking.
3824          */
3825         switch (fl->l_whence) {
3826         case SEEK_SET:
3827         case SEEK_CUR:
3828                 /*
3829                  * Caller is responsible for adding any necessary offset
3830                  * when SEEK_CUR is used.
3831                  */
3832                 start = fl->l_start;
3833                 off = fl->l_start;
3834                 break;
3835         case SEEK_END:
3836                 start = size + fl->l_start;
3837                 off = size + fl->l_start;
3838                 break;
3839         default:
3840                 return (EINVAL);
3841         }
3842         if (start < 0)
3843                 return (EINVAL);
3844         if (fl->l_len != 0) {
3845                 end = start + fl->l_len - 1;
3846                 if (end < start)
3847                         return (EINVAL);
3848         }
3849
3850         len = fl->l_len;
3851         if (len == 0)
3852                 len = NFS64BITSSET;
3853         retrycnt = 0;
3854         do {
3855             nd->nd_repstat = 0;
3856             if (op == F_GETLK) {
3857                 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3858                 if (error)
3859                         return (error);
3860                 error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3861                 if (!error) {
3862                         clidrev = clp->nfsc_clientidrev;
3863                         error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3864                             p, id, flags);
3865                 } else if (error == -1) {
3866                         error = 0;
3867                 }
3868                 nfscl_clientrelease(clp);
3869             } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3870                 /*
3871                  * We must loop around for all lockowner cases.
3872                  */
3873                 callcnt = 0;
3874                 error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3875                 if (error)
3876                         return (error);
3877                 do {
3878                     error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3879                         clp, id, flags, &lp, &dorpc);
3880                     /*
3881                      * If it returns a NULL lp, we're done.
3882                      */
3883                     if (lp == NULL) {
3884                         if (callcnt == 0)
3885                             nfscl_clientrelease(clp);
3886                         else
3887                             nfscl_releasealllocks(clp, vp, p, id, flags);
3888                         return (error);
3889                     }
3890                     if (nmp->nm_clp != NULL)
3891                         clidrev = nmp->nm_clp->nfsc_clientidrev;
3892                     else
3893                         clidrev = 0;
3894                     /*
3895                      * If the server doesn't support Posix lock semantics,
3896                      * only allow locks on the entire file, since it won't
3897                      * handle overlapping byte ranges.
3898                      * There might still be a problem when a lock
3899                      * upgrade/downgrade (read<->write) occurs, since the
3900                      * server "might" expect an unlock first?
3901                      */
3902                     if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3903                         (off == 0 && len == NFS64BITSSET))) {
3904                         /*
3905                          * Since the lock records will go away, we must
3906                          * wait for grace and delay here.
3907                          */
3908                         do {
3909                             error = nfsrpc_locku(nd, nmp, lp, off, len,
3910                                 NFSV4LOCKT_READ, cred, p, 0);
3911                             if ((nd->nd_repstat == NFSERR_GRACE ||
3912                                  nd->nd_repstat == NFSERR_DELAY) &&
3913                                 error == 0)
3914                                 (void) nfs_catnap(PZERO, (int)nd->nd_repstat,
3915                                     "nfs_advlock");
3916                         } while ((nd->nd_repstat == NFSERR_GRACE ||
3917                             nd->nd_repstat == NFSERR_DELAY) && error == 0);
3918                     }
3919                     callcnt++;
3920                 } while (error == 0 && nd->nd_repstat == 0);
3921                 nfscl_releasealllocks(clp, vp, p, id, flags);
3922             } else if (op == F_SETLK) {
3923                 error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
3924                     NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
3925                 if (error || donelocally) {
3926                         return (error);
3927                 }
3928                 if (nmp->nm_clp != NULL)
3929                         clidrev = nmp->nm_clp->nfsc_clientidrev;
3930                 else
3931                         clidrev = 0;
3932                 nfhp = VTONFS(vp)->n_fhp;
3933                 if (!lp->nfsl_open->nfso_posixlock &&
3934                     (off != 0 || len != NFS64BITSSET)) {
3935                         error = EINVAL;
3936                 } else {
3937                         error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
3938                             nfhp->nfh_len, lp, newone, reclaim, off,
3939                             len, fl->l_type, cred, p, 0);
3940                 }
3941                 if (!error)
3942                         error = nd->nd_repstat;
3943                 nfscl_lockrelease(lp, error, newone);
3944             } else {
3945                 error = EINVAL;
3946             }
3947             if (!error)
3948                 error = nd->nd_repstat;
3949             if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
3950                 error == NFSERR_STALEDONTRECOVER ||
3951                 error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3952                 error == NFSERR_BADSESSION) {
3953                 (void) nfs_catnap(PZERO, error, "nfs_advlock");
3954             } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
3955                 && clidrev != 0) {
3956                 expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
3957                 retrycnt++;
3958             }
3959         } while (error == NFSERR_GRACE ||
3960             error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3961             error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
3962             error == NFSERR_BADSESSION ||
3963             ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
3964              expireret == 0 && clidrev != 0 && retrycnt < 4));
3965         if (error && retrycnt >= 4)
3966                 error = EIO;
3967         return (error);
3968 }
3969
3970 /*
3971  * The lower level routine for the LockT case.
3972  */
3973 APPLESTATIC int
3974 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
3975     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
3976     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3977 {
3978         u_int32_t *tl;
3979         int error, type, size;
3980         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3981         struct nfsnode *np;
3982         struct nfsmount *nmp;
3983         struct nfsclsession *tsep;
3984
3985         nmp = VFSTONFS(vp->v_mount);
3986         NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
3987         NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3988         if (fl->l_type == F_RDLCK)
3989                 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3990         else
3991                 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3992         txdr_hyper(off, tl);
3993         tl += 2;
3994         txdr_hyper(len, tl);
3995         tl += 2;
3996         tsep = nfsmnt_mdssession(nmp);
3997         *tl++ = tsep->nfsess_clientid.lval[0];
3998         *tl = tsep->nfsess_clientid.lval[1];
3999         nfscl_filllockowner(id, own, flags);
4000         np = VTONFS(vp);
4001         NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4002             np->n_fhp->nfh_len);
4003         (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4004         error = nfscl_request(nd, vp, p, cred, NULL);
4005         if (error)
4006                 return (error);
4007         if (nd->nd_repstat == 0) {
4008                 fl->l_type = F_UNLCK;
4009         } else if (nd->nd_repstat == NFSERR_DENIED) {
4010                 nd->nd_repstat = 0;
4011                 fl->l_whence = SEEK_SET;
4012                 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4013                 fl->l_start = fxdr_hyper(tl);
4014                 tl += 2;
4015                 len = fxdr_hyper(tl);
4016                 tl += 2;
4017                 if (len == NFS64BITSSET)
4018                         fl->l_len = 0;
4019                 else
4020                         fl->l_len = len;
4021                 type = fxdr_unsigned(int, *tl++);
4022                 if (type == NFSV4LOCKT_WRITE)
4023                         fl->l_type = F_WRLCK;
4024                 else
4025                         fl->l_type = F_RDLCK;
4026                 /*
4027                  * XXX For now, I have no idea what to do with the
4028                  * conflicting lock_owner, so I'll just set the pid == 0
4029                  * and skip over the lock_owner.
4030                  */
4031                 fl->l_pid = (pid_t)0;
4032                 tl += 2;
4033                 size = fxdr_unsigned(int, *tl);
4034                 if (size < 0 || size > NFSV4_OPAQUELIMIT)
4035                         error = EBADRPC;
4036                 if (!error)
4037                         error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4038         } else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4039                 nfscl_initiate_recovery(clp);
4040 nfsmout:
4041         mbuf_freem(nd->nd_mrep);
4042         return (error);
4043 }
4044
4045 /*
4046  * Lower level function that performs the LockU RPC.
4047  */
4048 static int
4049 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4050     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4051     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4052 {
4053         u_int32_t *tl;
4054         int error;
4055
4056         nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4057             lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4058         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4059         *tl++ = txdr_unsigned(type);
4060         *tl = txdr_unsigned(lp->nfsl_seqid);
4061         if (nfstest_outofseq &&
4062             (arc4random() % nfstest_outofseq) == 0)
4063                 *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4064         tl++;
4065         if (NFSHASNFSV4N(nmp))
4066                 *tl++ = 0;
4067         else
4068                 *tl++ = lp->nfsl_stateid.seqid;
4069         *tl++ = lp->nfsl_stateid.other[0];
4070         *tl++ = lp->nfsl_stateid.other[1];
4071         *tl++ = lp->nfsl_stateid.other[2];
4072         txdr_hyper(off, tl);
4073         tl += 2;
4074         txdr_hyper(len, tl);
4075         if (syscred)
4076                 nd->nd_flag |= ND_USEGSSNAME;
4077         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4078             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4079         NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4080         if (error)
4081                 return (error);
4082         if (nd->nd_repstat == 0) {
4083                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4084                 lp->nfsl_stateid.seqid = *tl++;
4085                 lp->nfsl_stateid.other[0] = *tl++;
4086                 lp->nfsl_stateid.other[1] = *tl++;
4087                 lp->nfsl_stateid.other[2] = *tl;
4088         } else if (nd->nd_repstat == NFSERR_STALESTATEID)
4089                 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4090 nfsmout:
4091         mbuf_freem(nd->nd_mrep);
4092         return (error);
4093 }
4094
4095 /*
4096  * The actual Lock RPC.
4097  */
4098 APPLESTATIC int
4099 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4100     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4101     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4102     NFSPROC_T *p, int syscred)
4103 {
4104         u_int32_t *tl;
4105         int error, size;
4106         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4107         struct nfsclsession *tsep;
4108
4109         nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4110         NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4111         if (type == F_RDLCK)
4112                 *tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4113         else
4114                 *tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4115         *tl++ = txdr_unsigned(reclaim);
4116         txdr_hyper(off, tl);
4117         tl += 2;
4118         txdr_hyper(len, tl);
4119         tl += 2;
4120         if (newone) {
4121             *tl = newnfs_true;
4122             NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4123                 2 * NFSX_UNSIGNED + NFSX_HYPER);
4124             *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4125             if (NFSHASNFSV4N(nmp))
4126                 *tl++ = 0;
4127             else
4128                 *tl++ = lp->nfsl_open->nfso_stateid.seqid;
4129             *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4130             *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4131             *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4132             *tl++ = txdr_unsigned(lp->nfsl_seqid);
4133             tsep = nfsmnt_mdssession(nmp);
4134             *tl++ = tsep->nfsess_clientid.lval[0];
4135             *tl = tsep->nfsess_clientid.lval[1];
4136             NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4137             NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4138             (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4139         } else {
4140             *tl = newnfs_false;
4141             NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4142             if (NFSHASNFSV4N(nmp))
4143                 *tl++ = 0;
4144             else
4145                 *tl++ = lp->nfsl_stateid.seqid;
4146             *tl++ = lp->nfsl_stateid.other[0];
4147             *tl++ = lp->nfsl_stateid.other[1];
4148             *tl++ = lp->nfsl_stateid.other[2];
4149             *tl = txdr_unsigned(lp->nfsl_seqid);
4150             if (nfstest_outofseq &&
4151                 (arc4random() % nfstest_outofseq) == 0)
4152                     *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4153         }
4154         if (syscred)
4155                 nd->nd_flag |= ND_USEGSSNAME;
4156         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4157             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4158         if (error)
4159                 return (error);
4160         if (newone)
4161             NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4162         NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4163         if (nd->nd_repstat == 0) {
4164                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4165                 lp->nfsl_stateid.seqid = *tl++;
4166                 lp->nfsl_stateid.other[0] = *tl++;
4167                 lp->nfsl_stateid.other[1] = *tl++;
4168                 lp->nfsl_stateid.other[2] = *tl;
4169         } else if (nd->nd_repstat == NFSERR_DENIED) {
4170                 NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4171                 size = fxdr_unsigned(int, *(tl + 7));
4172                 if (size < 0 || size > NFSV4_OPAQUELIMIT)
4173                         error = EBADRPC;
4174                 if (!error)
4175                         error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4176         } else if (nd->nd_repstat == NFSERR_STALESTATEID)
4177                 nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4178 nfsmout:
4179         mbuf_freem(nd->nd_mrep);
4180         return (error);
4181 }
4182
4183 /*
4184  * nfs statfs rpc
4185  * (always called with the vp for the mount point)
4186  */
4187 APPLESTATIC int
4188 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4189     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4190     void *stuff)
4191 {
4192         u_int32_t *tl = NULL;
4193         struct nfsrv_descript nfsd, *nd = &nfsd;
4194         struct nfsmount *nmp;
4195         nfsattrbit_t attrbits;
4196         int error;
4197
4198         *attrflagp = 0;
4199         nmp = VFSTONFS(vnode_mount(vp));
4200         if (NFSHASNFSV4(nmp)) {
4201                 /*
4202                  * For V4, you actually do a getattr.
4203                  */
4204                 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4205                 NFSSTATFS_GETATTRBIT(&attrbits);
4206                 (void) nfsrv_putattrbit(nd, &attrbits);
4207                 nd->nd_flag |= ND_USEGSSNAME;
4208                 error = nfscl_request(nd, vp, p, cred, stuff);
4209                 if (error)
4210                         return (error);
4211                 if (nd->nd_repstat == 0) {
4212                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4213                             NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4214                             cred);
4215                         if (!error) {
4216                                 nmp->nm_fsid[0] = nap->na_filesid[0];
4217                                 nmp->nm_fsid[1] = nap->na_filesid[1];
4218                                 NFSSETHASSETFSID(nmp);
4219                                 *attrflagp = 1;
4220                         }
4221                 } else {
4222                         error = nd->nd_repstat;
4223                 }
4224                 if (error)
4225                         goto nfsmout;
4226         } else {
4227                 NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4228                 error = nfscl_request(nd, vp, p, cred, stuff);
4229                 if (error)
4230                         return (error);
4231                 if (nd->nd_flag & ND_NFSV3) {
4232                         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4233                         if (error)
4234                                 goto nfsmout;
4235                 }
4236                 if (nd->nd_repstat) {
4237                         error = nd->nd_repstat;
4238                         goto nfsmout;
4239                 }
4240                 NFSM_DISSECT(tl, u_int32_t *,
4241                     NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4242         }
4243         if (NFSHASNFSV3(nmp)) {
4244                 sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4245                 sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4246                 sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4247                 sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4248                 sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4249                 sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4250                 sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4251         } else if (NFSHASNFSV4(nmp) == 0) {
4252                 sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4253                 sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4254                 sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4255                 sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4256                 sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4257         }
4258 nfsmout:
4259         mbuf_freem(nd->nd_mrep);
4260         return (error);
4261 }
4262
4263 /*
4264  * nfs pathconf rpc
4265  */
4266 APPLESTATIC int
4267 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4268     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4269     void *stuff)
4270 {
4271         struct nfsrv_descript nfsd, *nd = &nfsd;
4272         struct nfsmount *nmp;
4273         u_int32_t *tl;
4274         nfsattrbit_t attrbits;
4275         int error;
4276
4277         *attrflagp = 0;
4278         nmp = VFSTONFS(vnode_mount(vp));
4279         if (NFSHASNFSV4(nmp)) {
4280                 /*
4281                  * For V4, you actually do a getattr.
4282                  */
4283                 NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4284                 NFSPATHCONF_GETATTRBIT(&attrbits);
4285                 (void) nfsrv_putattrbit(nd, &attrbits);
4286                 nd->nd_flag |= ND_USEGSSNAME;
4287                 error = nfscl_request(nd, vp, p, cred, stuff);
4288                 if (error)
4289                         return (error);
4290                 if (nd->nd_repstat == 0) {
4291                         error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4292                             pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4293                             cred);
4294                         if (!error)
4295                                 *attrflagp = 1;
4296                 } else {
4297                         error = nd->nd_repstat;
4298                 }
4299         } else {
4300                 NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4301                 error = nfscl_request(nd, vp, p, cred, stuff);
4302                 if (error)
4303                         return (error);
4304                 error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4305                 if (nd->nd_repstat && !error)
4306                         error = nd->nd_repstat;
4307                 if (!error) {
4308                         NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4309                         pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4310                         pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4311                         pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4312                         pc->pc_chownrestricted =
4313                             fxdr_unsigned(u_int32_t, *tl++);
4314                         pc->pc_caseinsensitive =
4315                             fxdr_unsigned(u_int32_t, *tl++);
4316                         pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4317                 }
4318         }
4319 nfsmout:
4320         mbuf_freem(nd->nd_mrep);
4321         return (error);
4322 }
4323
4324 /*
4325  * nfs version 3 fsinfo rpc call
4326  */
4327 APPLESTATIC int
4328 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4329     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4330 {
4331         u_int32_t *tl;
4332         struct nfsrv_descript nfsd, *nd = &nfsd;
4333         int error;
4334
4335         *attrflagp = 0;
4336         NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4337         error = nfscl_request(nd, vp, p, cred, stuff);
4338         if (error)
4339                 return (error);
4340         error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4341         if (nd->nd_repstat && !error)
4342                 error = nd->nd_repstat;
4343         if (!error) {
4344                 NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4345                 fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4346                 fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4347                 fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4348                 fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4349                 fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4350                 fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4351                 fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4352                 fsp->fs_maxfilesize = fxdr_hyper(tl);
4353                 tl += 2;
4354                 fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4355                 tl += 2;
4356                 fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4357         }
4358 nfsmout:
4359         mbuf_freem(nd->nd_mrep);
4360         return (error);
4361 }
4362
4363 /*
4364  * This function performs the Renew RPC.
4365  */
4366 APPLESTATIC int
4367 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4368     NFSPROC_T *p)
4369 {
4370         u_int32_t *tl;
4371         struct nfsrv_descript nfsd;
4372         struct nfsrv_descript *nd = &nfsd;
4373         struct nfsmount *nmp;
4374         int error;
4375         struct nfssockreq *nrp;
4376         struct nfsclsession *tsep;
4377
4378         nmp = clp->nfsc_nmp;
4379         if (nmp == NULL)
4380                 return (0);
4381         if (dsp == NULL)
4382                 nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4383                     0);
4384         else
4385                 nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4386                     &dsp->nfsclds_sess, 0, 0);
4387         if (!NFSHASNFSV4N(nmp)) {
4388                 /* NFSv4.1 just uses a Sequence Op and not a Renew. */
4389                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4390                 tsep = nfsmnt_mdssession(nmp);
4391                 *tl++ = tsep->nfsess_clientid.lval[0];
4392                 *tl = tsep->nfsess_clientid.lval[1];
4393         }
4394         nrp = NULL;
4395         if (dsp != NULL)
4396                 nrp = dsp->nfsclds_sockp;
4397         if (nrp == NULL)
4398                 /* If NULL, use the MDS socket. */
4399                 nrp = &nmp->nm_sockreq;
4400         nd->nd_flag |= ND_USEGSSNAME;
4401         if (dsp == NULL)
4402                 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4403                     NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4404         else
4405                 error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4406                     NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4407         if (error)
4408                 return (error);
4409         error = nd->nd_repstat;
4410         mbuf_freem(nd->nd_mrep);
4411         return (error);
4412 }
4413
4414 /*
4415  * This function performs the Releaselockowner RPC.
4416  */
4417 APPLESTATIC int
4418 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4419     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4420 {
4421         struct nfsrv_descript nfsd, *nd = &nfsd;
4422         u_int32_t *tl;
4423         int error;
4424         uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4425         struct nfsclsession *tsep;
4426
4427         if (NFSHASNFSV4N(nmp)) {
4428                 /* For NFSv4.1, do a FreeStateID. */
4429                 nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4430                     NULL, 0, 0);
4431                 nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4432         } else {
4433                 nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4434                     NULL, 0, 0);
4435                 NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4436                 tsep = nfsmnt_mdssession(nmp);
4437                 *tl++ = tsep->nfsess_clientid.lval[0];
4438                 *tl = tsep->nfsess_clientid.lval[1];
4439                 NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4440                 NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4441                 (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4442         }
4443         nd->nd_flag |= ND_USEGSSNAME;
4444         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4445             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4446         if (error)
4447                 return (error);
4448         error = nd->nd_repstat;
4449         mbuf_freem(nd->nd_mrep);
4450         return (error);
4451 }
4452
4453 /*
4454  * This function performs the Compound to get the mount pt FH.
4455  */
4456 APPLESTATIC int
4457 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4458     NFSPROC_T *p)
4459 {
4460         u_int32_t *tl;
4461         struct nfsrv_descript nfsd;
4462         struct nfsrv_descript *nd = &nfsd;
4463         u_char *cp, *cp2;
4464         int error, cnt, len, setnil;
4465         u_int32_t *opcntp;
4466
4467         nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4468             0);
4469         cp = dirpath;
4470         cnt = 0;
4471         do {
4472                 setnil = 0;
4473                 while (*cp == '/')
4474                         cp++;
4475                 cp2 = cp;
4476                 while (*cp2 != '\0' && *cp2 != '/')
4477                         cp2++;
4478                 if (*cp2 == '/') {
4479                         setnil = 1;
4480                         *cp2 = '\0';
4481                 }
4482                 if (cp2 != cp) {
4483                         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4484                         *tl = txdr_unsigned(NFSV4OP_LOOKUP);
4485                         nfsm_strtom(nd, cp, strlen(cp));
4486                         cnt++;
4487                 }
4488                 if (setnil)
4489                         *cp2++ = '/';
4490                 cp = cp2;
4491         } while (*cp != '\0');
4492         if (NFSHASNFSV4N(nmp))
4493                 /* Has a Sequence Op done by nfscl_reqstart(). */
4494                 *opcntp = txdr_unsigned(3 + cnt);
4495         else
4496                 *opcntp = txdr_unsigned(2 + cnt);
4497         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4498         *tl = txdr_unsigned(NFSV4OP_GETFH);
4499         nd->nd_flag |= ND_USEGSSNAME;
4500         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4501                 NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4502         if (error)
4503                 return (error);
4504         if (nd->nd_repstat == 0) {
4505                 NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4506                 tl += (2 + 2 * cnt);
4507                 if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4508                         len > NFSX_FHMAX) {
4509                         nd->nd_repstat = NFSERR_BADXDR;
4510                 } else {
4511                         nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4512                         if (nd->nd_repstat == 0)
4513                                 nmp->nm_fhsize = len;
4514                 }
4515         }
4516         error = nd->nd_repstat;
4517 nfsmout:
4518         mbuf_freem(nd->nd_mrep);
4519         return (error);
4520 }
4521
4522 /*
4523  * This function performs the Delegreturn RPC.
4524  */
4525 APPLESTATIC int
4526 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4527     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4528 {
4529         u_int32_t *tl;
4530         struct nfsrv_descript nfsd;
4531         struct nfsrv_descript *nd = &nfsd;
4532         int error;
4533
4534         nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4535             dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4536         NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4537         if (NFSHASNFSV4N(nmp))
4538                 *tl++ = 0;
4539         else
4540                 *tl++ = dp->nfsdl_stateid.seqid;
4541         *tl++ = dp->nfsdl_stateid.other[0];
4542         *tl++ = dp->nfsdl_stateid.other[1];
4543         *tl = dp->nfsdl_stateid.other[2];
4544         if (syscred)
4545                 nd->nd_flag |= ND_USEGSSNAME;
4546         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4547             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4548         if (error)
4549                 return (error);
4550         error = nd->nd_repstat;
4551         mbuf_freem(nd->nd_mrep);
4552         return (error);
4553 }
4554
4555 /*
4556  * nfs getacl call.
4557  */
4558 APPLESTATIC int
4559 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4560     struct acl *aclp, void *stuff)
4561 {
4562         struct nfsrv_descript nfsd, *nd = &nfsd;
4563         int error;
4564         nfsattrbit_t attrbits;
4565         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4566
4567         if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4568                 return (EOPNOTSUPP);
4569         NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4570         NFSZERO_ATTRBIT(&attrbits);
4571         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4572         (void) nfsrv_putattrbit(nd, &attrbits);
4573         error = nfscl_request(nd, vp, p, cred, stuff);
4574         if (error)
4575                 return (error);
4576         if (!nd->nd_repstat)
4577                 error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4578                     NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4579         else
4580                 error = nd->nd_repstat;
4581         mbuf_freem(nd->nd_mrep);
4582         return (error);
4583 }
4584
4585 /*
4586  * nfs setacl call.
4587  */
4588 APPLESTATIC int
4589 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4590     struct acl *aclp, void *stuff)
4591 {
4592         int error;
4593         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4594
4595         if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4596                 return (EOPNOTSUPP);
4597         error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4598         return (error);
4599 }
4600
4601 /*
4602  * nfs setacl call.
4603  */
4604 static int
4605 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4606     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4607 {
4608         struct nfsrv_descript nfsd, *nd = &nfsd;
4609         int error;
4610         nfsattrbit_t attrbits;
4611         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4612
4613         if (!NFSHASNFSV4(nmp))
4614                 return (EOPNOTSUPP);
4615         NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4616         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4617         NFSZERO_ATTRBIT(&attrbits);
4618         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4619         (void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4620             &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0);
4621         error = nfscl_request(nd, vp, p, cred, stuff);
4622         if (error)
4623                 return (error);
4624         /* Don't care about the pre/postop attributes */
4625         mbuf_freem(nd->nd_mrep);
4626         return (nd->nd_repstat);
4627 }
4628
4629 /*
4630  * Do the NFSv4.1 Exchange ID.
4631  */
4632 int
4633 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4634     struct nfssockreq *nrp, uint32_t exchflags, struct nfsclds **dspp,
4635     struct ucred *cred, NFSPROC_T *p)
4636 {
4637         uint32_t *tl, v41flags;
4638         struct nfsrv_descript nfsd;
4639         struct nfsrv_descript *nd = &nfsd;
4640         struct nfsclds *dsp;
4641         struct timespec verstime;
4642         int error, len;
4643
4644         *dspp = NULL;
4645         nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL, 0, 0);
4646         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4647         *tl++ = txdr_unsigned(nfsboottime.tv_sec);      /* Client owner */
4648         *tl = txdr_unsigned(clp->nfsc_rev);
4649         (void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4650
4651         NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4652         *tl++ = txdr_unsigned(exchflags);
4653         *tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4654
4655         /* Set the implementation id4 */
4656         *tl = txdr_unsigned(1);
4657         (void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4658         (void) nfsm_strtom(nd, version, strlen(version));
4659         NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4660         verstime.tv_sec = 1293840000;           /* Jan 1, 2011 */
4661         verstime.tv_nsec = 0;
4662         txdr_nfsv4time(&verstime, tl);
4663         nd->nd_flag |= ND_USEGSSNAME;
4664         error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4665             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4666         NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4667             (int)nd->nd_repstat);
4668         if (error != 0)
4669                 return (error);
4670         if (nd->nd_repstat == 0) {
4671                 NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4672                 len = fxdr_unsigned(int, *(tl + 7));
4673                 if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4674                         error = NFSERR_BADXDR;
4675                         goto nfsmout;
4676                 }
4677                 dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4678                     M_WAITOK | M_ZERO);
4679                 dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4680                 dsp->nfsclds_servownlen = len;
4681                 dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4682                 dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4683                 dsp->nfsclds_sess.nfsess_sequenceid =
4684                     fxdr_unsigned(uint32_t, *tl++);
4685                 v41flags = fxdr_unsigned(uint32_t, *tl);
4686                 if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4687                     NFSHASPNFSOPT(nmp)) {
4688                         NFSCL_DEBUG(1, "set PNFS\n");
4689                         NFSLOCKMNT(nmp);
4690                         nmp->nm_state |= NFSSTA_PNFS;
4691                         NFSUNLOCKMNT(nmp);
4692                         dsp->nfsclds_flags |= NFSCLDS_MDS;
4693                 }
4694                 if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4695                         dsp->nfsclds_flags |= NFSCLDS_DS;
4696                 if (len > 0)
4697                         nd->nd_repstat = nfsrv_mtostr(nd,
4698                             dsp->nfsclds_serverown, len);
4699                 if (nd->nd_repstat == 0) {
4700                         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4701                         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4702                             NULL, MTX_DEF);
4703                         nfscl_initsessionslots(&dsp->nfsclds_sess);
4704                         *dspp = dsp;
4705                 } else
4706                         free(dsp, M_NFSCLDS);
4707         }
4708         error = nd->nd_repstat;
4709 nfsmout:
4710         mbuf_freem(nd->nd_mrep);
4711         return (error);
4712 }
4713
4714 /*
4715  * Do the NFSv4.1 Create Session.
4716  */
4717 int
4718 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4719     struct nfssockreq *nrp, uint32_t sequenceid, int mds, struct ucred *cred,
4720     NFSPROC_T *p)
4721 {
4722         uint32_t crflags, maxval, *tl;
4723         struct nfsrv_descript nfsd;
4724         struct nfsrv_descript *nd = &nfsd;
4725         int error, irdcnt;
4726
4727         /* Make sure nm_rsize, nm_wsize is set. */
4728         if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4729                 nmp->nm_rsize = NFS_MAXBSIZE;
4730         if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4731                 nmp->nm_wsize = NFS_MAXBSIZE;
4732         nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL, 0,
4733             0);
4734         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4735         *tl++ = sep->nfsess_clientid.lval[0];
4736         *tl++ = sep->nfsess_clientid.lval[1];
4737         *tl++ = txdr_unsigned(sequenceid);
4738         crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4739         if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4740                 crflags |= NFSV4CRSESS_CONNBACKCHAN;
4741         *tl = txdr_unsigned(crflags);
4742
4743         /* Fill in fore channel attributes. */
4744         NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4745         *tl++ = 0;                              /* Header pad size */
4746         *tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);/* Max request size */
4747         *tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);/* Max reply size */
4748         *tl++ = txdr_unsigned(4096);            /* Max response size cached */
4749         *tl++ = txdr_unsigned(20);              /* Max operations */
4750         *tl++ = txdr_unsigned(64);              /* Max slots */
4751         *tl = 0;                                /* No rdma ird */
4752
4753         /* Fill in back channel attributes. */
4754         NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4755         *tl++ = 0;                              /* Header pad size */
4756         *tl++ = txdr_unsigned(10000);           /* Max request size */
4757         *tl++ = txdr_unsigned(10000);           /* Max response size */
4758         *tl++ = txdr_unsigned(4096);            /* Max response size cached */
4759         *tl++ = txdr_unsigned(4);               /* Max operations */
4760         *tl++ = txdr_unsigned(NFSV4_CBSLOTS);   /* Max slots */
4761         *tl = 0;                                /* No rdma ird */
4762
4763         NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4764         *tl++ = txdr_unsigned(NFS_CALLBCKPROG); /* Call back prog # */
4765
4766         /* Allow AUTH_SYS callbacks as uid, gid == 0. */
4767         *tl++ = txdr_unsigned(1);               /* Auth_sys only */
4768         *tl++ = txdr_unsigned(AUTH_SYS);        /* AUTH_SYS type */
4769         *tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4770         *tl++ = 0;                              /* Null machine name */
4771         *tl++ = 0;                              /* Uid == 0 */
4772         *tl++ = 0;                              /* Gid == 0 */
4773         *tl = 0;                                /* No additional gids */
4774         nd->nd_flag |= ND_USEGSSNAME;
4775         error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4776             NFS_VER4, NULL, 1, NULL, NULL);
4777         if (error != 0)
4778                 return (error);
4779         if (nd->nd_repstat == 0) {
4780                 NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4781                     2 * NFSX_UNSIGNED);
4782                 bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4783                 tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4784                 sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4785                 crflags = fxdr_unsigned(uint32_t, *tl);
4786                 if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4787                         NFSLOCKMNT(nmp);
4788                         nmp->nm_state |= NFSSTA_SESSPERSIST;
4789                         NFSUNLOCKMNT(nmp);
4790                 }
4791
4792                 /* Get the fore channel slot count. */
4793                 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4794                 tl++;                   /* Skip the header pad size. */
4795
4796                 /* Make sure nm_wsize is small enough. */
4797                 maxval = fxdr_unsigned(uint32_t, *tl++);
4798                 while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4799                         if (nmp->nm_wsize > 8096)
4800                                 nmp->nm_wsize /= 2;
4801                         else
4802                                 break;
4803                 }
4804
4805                 /* Make sure nm_rsize is small enough. */
4806                 maxval = fxdr_unsigned(uint32_t, *tl++);
4807                 while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4808                         if (nmp->nm_rsize > 8096)
4809                                 nmp->nm_rsize /= 2;
4810                         else
4811                                 break;
4812                 }
4813
4814                 sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4815                 tl++;
4816                 sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4817                 NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4818                 irdcnt = fxdr_unsigned(int, *tl);
4819                 if (irdcnt > 0)
4820                         NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4821
4822                 /* and the back channel slot count. */
4823                 NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4824                 tl += 5;
4825                 sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4826                 NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4827         }
4828         error = nd->nd_repstat;
4829 nfsmout:
4830         mbuf_freem(nd->nd_mrep);
4831         return (error);
4832 }
4833
4834 /*
4835  * Do the NFSv4.1 Destroy Session.
4836  */
4837 int
4838 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4839     struct ucred *cred, NFSPROC_T *p)
4840 {
4841         uint32_t *tl;
4842         struct nfsrv_descript nfsd;
4843         struct nfsrv_descript *nd = &nfsd;
4844         int error;
4845         struct nfsclsession *tsep;
4846
4847         nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
4848             0);
4849         NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4850         tsep = nfsmnt_mdssession(nmp);
4851         bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4852         nd->nd_flag |= ND_USEGSSNAME;
4853         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4854             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4855         if (error != 0)
4856                 return (error);
4857         error = nd->nd_repstat;
4858         mbuf_freem(nd->nd_mrep);
4859         return (error);
4860 }
4861
4862 /*
4863  * Do the NFSv4.1 Destroy Client.
4864  */
4865 int
4866 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4867     struct ucred *cred, NFSPROC_T *p)
4868 {
4869         uint32_t *tl;
4870         struct nfsrv_descript nfsd;
4871         struct nfsrv_descript *nd = &nfsd;
4872         int error;
4873         struct nfsclsession *tsep;
4874
4875         nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
4876             0);
4877         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4878         tsep = nfsmnt_mdssession(nmp);
4879         *tl++ = tsep->nfsess_clientid.lval[0];
4880         *tl = tsep->nfsess_clientid.lval[1];
4881         nd->nd_flag |= ND_USEGSSNAME;
4882         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4883             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4884         if (error != 0)
4885                 return (error);
4886         error = nd->nd_repstat;
4887         mbuf_freem(nd->nd_mrep);
4888         return (error);
4889 }
4890
4891 /*
4892  * Do the NFSv4.1 LayoutGet.
4893  */
4894 static int
4895 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
4896     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
4897     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
4898     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
4899     void *stuff)
4900 {
4901         struct nfsrv_descript nfsd, *nd = &nfsd;
4902         int error;
4903
4904         nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
4905             0);
4906         nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
4907             layouttype, layoutlen, 0);
4908         nd->nd_flag |= ND_USEGSSNAME;
4909         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4910             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4911         NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
4912         if (error != 0)
4913                 return (error);
4914         if (nd->nd_repstat == 0)
4915                 error = nfsrv_parselayoutget(nd, stateidp, retonclosep, flhp);
4916         if (error == 0 && nd->nd_repstat != 0)
4917                 error = nd->nd_repstat;
4918         mbuf_freem(nd->nd_mrep);
4919         return (error);
4920 }
4921
4922 /*
4923  * Do the NFSv4.1 Get Device Info.
4924  */
4925 int
4926 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
4927     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
4928     NFSPROC_T *p)
4929 {
4930         uint32_t cnt, *tl, vers, minorvers;
4931         struct nfsrv_descript nfsd;
4932         struct nfsrv_descript *nd = &nfsd;
4933         struct sockaddr_in sin, ssin;
4934         struct sockaddr_in6 sin6, ssin6;
4935         struct nfsclds *dsp = NULL, **dspp, **gotdspp;
4936         struct nfscldevinfo *ndi;
4937         int addrcnt = 0, bitcnt, error, gotvers, i, isudp, j, stripecnt;
4938         uint8_t stripeindex;
4939         sa_family_t af, safilled;
4940
4941         *ndip = NULL;
4942         ndi = NULL;
4943         gotdspp = NULL;
4944         nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
4945             0);
4946         NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
4947         NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
4948         tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4949         *tl++ = txdr_unsigned(layouttype);
4950         *tl++ = txdr_unsigned(100000);
4951         if (notifybitsp != NULL && *notifybitsp != 0) {
4952                 *tl = txdr_unsigned(1);         /* One word of bits. */
4953                 NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
4954                 *tl = txdr_unsigned(*notifybitsp);
4955         } else
4956                 *tl = txdr_unsigned(0);
4957         nd->nd_flag |= ND_USEGSSNAME;
4958         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4959             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4960         if (error != 0)
4961                 return (error);
4962         if (nd->nd_repstat == 0) {
4963                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4964                 if (layouttype != fxdr_unsigned(int, *tl))
4965                         printf("EEK! devinfo layout type not same!\n");
4966                 if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
4967                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4968                         stripecnt = fxdr_unsigned(int, *tl);
4969                         NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
4970                         if (stripecnt < 1 || stripecnt > 4096) {
4971                                 printf("pNFS File layout devinfo stripecnt %d:"
4972                                     " out of range\n", stripecnt);
4973                                 error = NFSERR_BADXDR;
4974                                 goto nfsmout;
4975                         }
4976                         NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
4977                             NFSX_UNSIGNED);
4978                         addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
4979                         NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
4980                         if (addrcnt < 1 || addrcnt > 128) {
4981                                 printf("NFS devinfo addrcnt %d: out of range\n",
4982                                     addrcnt);
4983                                 error = NFSERR_BADXDR;
4984                                 goto nfsmout;
4985                         }
4986
4987                         /*
4988                          * Now we know how many stripe indices and addresses, so
4989                          * we can allocate the structure the correct size.
4990                          */
4991                         i = (stripecnt * sizeof(uint8_t)) /
4992                             sizeof(struct nfsclds *) + 1;
4993                         NFSCL_DEBUG(4, "stripeindices=%d\n", i);
4994                         ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
4995                             sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
4996                             M_ZERO);
4997                         NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
4998                             NFSX_V4DEVICEID);
4999                         ndi->nfsdi_refcnt = 0;
5000                         ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5001                         ndi->nfsdi_stripecnt = stripecnt;
5002                         ndi->nfsdi_addrcnt = addrcnt;
5003                         /* Fill in the stripe indices. */
5004                         for (i = 0; i < stripecnt; i++) {
5005                                 stripeindex = fxdr_unsigned(uint8_t, *tl++);
5006                                 NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5007                                 if (stripeindex >= addrcnt) {
5008                                         printf("pNFS File Layout devinfo"
5009                                             " stripeindex %d: too big\n",
5010                                             (int)stripeindex);
5011                                         error = NFSERR_BADXDR;
5012                                         goto nfsmout;
5013                                 }
5014                                 nfsfldi_setstripeindex(ndi, i, stripeindex);
5015                         }
5016                 } else if (layouttype == NFSLAYOUT_FLEXFILE) {
5017                         /* For Flex File, we only get one address list. */
5018                         ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5019                             M_NFSDEVINFO, M_WAITOK | M_ZERO);
5020                         NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5021                             NFSX_V4DEVICEID);
5022                         ndi->nfsdi_refcnt = 0;
5023                         ndi->nfsdi_flags = NFSDI_FLEXFILE;
5024                         addrcnt = ndi->nfsdi_addrcnt = 1;
5025                 }
5026
5027                 /* Now, dissect the server address(es). */
5028                 safilled = AF_UNSPEC;
5029                 for (i = 0; i < addrcnt; i++) {
5030                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5031                         cnt = fxdr_unsigned(uint32_t, *tl);
5032                         if (cnt == 0) {
5033                                 printf("NFS devinfo 0 len addrlist\n");
5034                                 error = NFSERR_BADXDR;
5035                                 goto nfsmout;
5036                         }
5037                         dspp = nfsfldi_addr(ndi, i);
5038                         safilled = AF_UNSPEC;
5039                         for (j = 0; j < cnt; j++) {
5040                                 error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5041                                     &isudp);
5042                                 if (error != 0 && error != EPERM) {
5043                                         error = NFSERR_BADXDR;
5044                                         goto nfsmout;
5045                                 }
5046                                 if (error == 0 && isudp == 0) {
5047                                         /*
5048                                          * The priority is:
5049                                          * - Same address family.
5050                                          * Save the address and dspp, so that
5051                                          * the connection can be done after
5052                                          * parsing is complete.
5053                                          */
5054                                         if (safilled == AF_UNSPEC ||
5055                                             (af == nmp->nm_nam->sa_family &&
5056                                              safilled != nmp->nm_nam->sa_family)
5057                                            ) {
5058                                                 if (af == AF_INET)
5059                                                         ssin = sin;
5060                                                 else
5061                                                         ssin6 = sin6;
5062                                                 safilled = af;
5063                                                 gotdspp = dspp;
5064                                         }
5065                                 }
5066                         }
5067                 }
5068
5069                 gotvers = NFS_VER4;     /* Always NFSv4 for File Layout. */
5070                 /* For Flex File, we will take one of the versions to use. */
5071                 if (layouttype == NFSLAYOUT_FLEXFILE) {
5072                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5073                         j = fxdr_unsigned(int, *tl);
5074                         if (j < 1 || j > NFSDEV_MAXVERS) {
5075                                 printf("pNFS: too many versions\n");
5076                                 error = NFSERR_BADXDR;
5077                                 goto nfsmout;
5078                         }
5079                         gotvers = 0;
5080                         for (i = 0; i < j; i++) {
5081                                 NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5082                                 vers = fxdr_unsigned(uint32_t, *tl++);
5083                                 minorvers = fxdr_unsigned(uint32_t, *tl++);
5084                                 if ((vers == NFS_VER4 && minorvers ==
5085                                     NFSV41_MINORVERSION) || (vers == NFS_VER3 &&
5086                                     gotvers == 0)) {
5087                                         gotvers = vers;
5088                                         /* We'll take this one. */
5089                                         ndi->nfsdi_versindex = i;
5090                                         ndi->nfsdi_vers = vers;
5091                                         ndi->nfsdi_minorvers = minorvers;
5092                                         ndi->nfsdi_rsize = fxdr_unsigned(
5093                                             uint32_t, *tl++);
5094                                         ndi->nfsdi_wsize = fxdr_unsigned(
5095                                             uint32_t, *tl++);
5096                                         if (*tl == newnfs_true)
5097                                                 ndi->nfsdi_flags |=
5098                                                     NFSDI_TIGHTCOUPLED;
5099                                         else
5100                                                 ndi->nfsdi_flags &=
5101                                                     ~NFSDI_TIGHTCOUPLED;
5102                                 }
5103                         }
5104                         if (gotvers == 0) {
5105                                 printf("pNFS: no NFSv3 or NFSv4.1\n");
5106                                 error = NFSERR_BADXDR;
5107                                 goto nfsmout;
5108                         }
5109                 }
5110
5111                 /* And the notify bits. */
5112                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5113                 bitcnt = fxdr_unsigned(int, *tl);
5114                 if (bitcnt > 0) {
5115                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5116                         if (notifybitsp != NULL)
5117                                 *notifybitsp =
5118                                     fxdr_unsigned(uint32_t, *tl);
5119                 }
5120                 if (safilled != AF_UNSPEC) {
5121                         KASSERT(ndi != NULL, ("ndi is NULL"));
5122                         *ndip = ndi;
5123                 } else
5124                         error = EPERM;
5125                 if (error == 0) {
5126                         /*
5127                          * Now we can do a TCP connection for the correct
5128                          * NFS version and IP address.
5129                          */
5130                         error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5131                             gotvers, &dsp, p);
5132                 }
5133                 if (error == 0) {
5134                         KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5135                         *gotdspp = dsp;
5136                 }
5137         }
5138         if (nd->nd_repstat != 0 && error == 0)
5139                 error = nd->nd_repstat;
5140 nfsmout:
5141         if (error != 0 && ndi != NULL)
5142                 nfscl_freedevinfo(ndi);
5143         mbuf_freem(nd->nd_mrep);
5144         return (error);
5145 }
5146
5147 /*
5148  * Do the NFSv4.1 LayoutCommit.
5149  */
5150 int
5151 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5152     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5153     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5154 {
5155         uint32_t *tl;
5156         struct nfsrv_descript nfsd, *nd = &nfsd;
5157         int error;
5158
5159         nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5160             0, 0);
5161         NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5162             NFSX_STATEID);
5163         txdr_hyper(off, tl);
5164         tl += 2;
5165         txdr_hyper(len, tl);
5166         tl += 2;
5167         if (reclaim != 0)
5168                 *tl++ = newnfs_true;
5169         else
5170                 *tl++ = newnfs_false;
5171         *tl++ = txdr_unsigned(stateidp->seqid);
5172         *tl++ = stateidp->other[0];
5173         *tl++ = stateidp->other[1];
5174         *tl++ = stateidp->other[2];
5175         *tl++ = newnfs_true;
5176         if (lastbyte < off)
5177                 lastbyte = off;
5178         else if (lastbyte >= (off + len))
5179                 lastbyte = off + len - 1;
5180         txdr_hyper(lastbyte, tl);
5181         tl += 2;
5182         *tl++ = newnfs_false;
5183         *tl++ = txdr_unsigned(layouttype);
5184         /* All supported layouts are 0 length. */
5185         *tl = txdr_unsigned(0);
5186         nd->nd_flag |= ND_USEGSSNAME;
5187         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5188             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5189         if (error != 0)
5190                 return (error);
5191         error = nd->nd_repstat;
5192         mbuf_freem(nd->nd_mrep);
5193         return (error);
5194 }
5195
5196 /*
5197  * Do the NFSv4.1 LayoutReturn.
5198  */
5199 int
5200 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5201     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5202     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5203     void *stuff)
5204 {
5205         uint32_t *tl;
5206         struct nfsrv_descript nfsd, *nd = &nfsd;
5207         int error;
5208
5209         nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5210             0, 0);
5211         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5212         if (reclaim != 0)
5213                 *tl++ = newnfs_true;
5214         else
5215                 *tl++ = newnfs_false;
5216         *tl++ = txdr_unsigned(layouttype);
5217         *tl++ = txdr_unsigned(iomode);
5218         *tl = txdr_unsigned(layoutreturn);
5219         if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5220                 NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5221                     NFSX_UNSIGNED);
5222                 txdr_hyper(offset, tl);
5223                 tl += 2;
5224                 txdr_hyper(len, tl);
5225                 tl += 2;
5226                 NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5227                 *tl++ = txdr_unsigned(stateidp->seqid);
5228                 *tl++ = stateidp->other[0];
5229                 *tl++ = stateidp->other[1];
5230                 *tl++ = stateidp->other[2];
5231                 if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5232                         *tl = txdr_unsigned(0);
5233                 else if (layouttype == NFSLAYOUT_FLEXFILE) {
5234                         *tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5235                         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5236                         /* No ioerrs or stats yet. */
5237                         *tl++ = 0;
5238                         *tl = 0;
5239                 }
5240         }
5241         nd->nd_flag |= ND_USEGSSNAME;
5242         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5243             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5244         if (error != 0)
5245                 return (error);
5246         if (nd->nd_repstat == 0) {
5247                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5248                 if (*tl != 0) {
5249                         NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5250                         stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5251                         stateidp->other[0] = *tl++;
5252                         stateidp->other[1] = *tl++;
5253                         stateidp->other[2] = *tl;
5254                 }
5255         } else
5256                 error = nd->nd_repstat;
5257 nfsmout:
5258         mbuf_freem(nd->nd_mrep);
5259         return (error);
5260 }
5261
5262 /*
5263  * Acquire a layout and devinfo, if possible. The caller must have acquired
5264  * a reference count on the nfsclclient structure before calling this.
5265  * Return the layout in lypp with a reference count on it, if successful.
5266  */
5267 static int
5268 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5269     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5270     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5271 {
5272         struct nfscllayout *lyp;
5273         struct nfsclflayout *flp;
5274         struct nfsclflayouthead flh;
5275         int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5276         nfsv4stateid_t stateid;
5277         struct nfsclsession *tsep;
5278
5279         *lypp = NULL;
5280         if (NFSHASFLEXFILE(nmp))
5281                 layouttype = NFSLAYOUT_FLEXFILE;
5282         else
5283                 layouttype = NFSLAYOUT_NFSV4_1_FILES;
5284         /*
5285          * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5286          * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5287          * flp == NULL.
5288          */
5289         lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5290             off, &flp, &recalled);
5291         islocked = 0;
5292         if (lyp == NULL || flp == NULL) {
5293                 if (recalled != 0)
5294                         return (EIO);
5295                 LIST_INIT(&flh);
5296                 tsep = nfsmnt_mdssession(nmp);
5297                 layoutlen = tsep->nfsess_maxcache -
5298                     (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5299                 if (lyp == NULL) {
5300                         stateid.seqid = 0;
5301                         stateid.other[0] = stateidp->other[0];
5302                         stateid.other[1] = stateidp->other[1];
5303                         stateid.other[2] = stateidp->other[2];
5304                         error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5305                             nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5306                             (uint64_t)0, layouttype, layoutlen, &stateid,
5307                             &retonclose, &flh, cred, p, NULL);
5308                 } else {
5309                         islocked = 1;
5310                         stateid.seqid = lyp->nfsly_stateid.seqid;
5311                         stateid.other[0] = lyp->nfsly_stateid.other[0];
5312                         stateid.other[1] = lyp->nfsly_stateid.other[1];
5313                         stateid.other[2] = lyp->nfsly_stateid.other[2];
5314                         error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5315                             nfhp->nfh_len, iomode, off, UINT64_MAX,
5316                             (uint64_t)0, layouttype, layoutlen, &stateid,
5317                             &retonclose, &flh, cred, p, NULL);
5318                 }
5319                 error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5320                     nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5321                     &flh, layouttype, error, NULL, cred, p);
5322                 if (error == 0)
5323                         *lypp = lyp;
5324                 else if (islocked != 0)
5325                         nfscl_rellayout(lyp, 1);
5326         } else
5327                 *lypp = lyp;
5328         return (error);
5329 }
5330
5331 /*
5332  * Do a TCP connection plus exchange id and create session.
5333  * If successful, a "struct nfsclds" is linked into the list for the
5334  * mount point and a pointer to it is returned.
5335  */
5336 static int
5337 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5338     struct sockaddr_in6 *sin6, sa_family_t af, int vers, struct nfsclds **dspp,
5339     NFSPROC_T *p)
5340 {
5341         struct sockaddr_in *msad, *sad;
5342         struct sockaddr_in6 *msad6, *sad6;
5343         struct nfsclclient *clp;
5344         struct nfssockreq *nrp;
5345         struct nfsclds *dsp, *tdsp;
5346         int error;
5347         enum nfsclds_state retv;
5348         uint32_t sequenceid;
5349
5350         KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5351             ("nfsrpc_fillsa: NULL nr_cred"));
5352         NFSLOCKCLSTATE();
5353         clp = nmp->nm_clp;
5354         NFSUNLOCKCLSTATE();
5355         if (clp == NULL)
5356                 return (EPERM);
5357         if (af == AF_INET) {
5358                 NFSLOCKMNT(nmp);
5359                 /*
5360                  * Check to see if we already have a session for this
5361                  * address that is usable for a DS.
5362                  * Note that the MDS's address is in a different place
5363                  * than the sessions already acquired for DS's.
5364                  */
5365                 msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5366                 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5367                 while (tdsp != NULL) {
5368                         if (msad != NULL && msad->sin_family == AF_INET &&
5369                             sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5370                             sin->sin_port == msad->sin_port &&
5371                             (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5372                             tdsp->nfsclds_sess.nfsess_defunct == 0) {
5373                                 *dspp = tdsp;
5374                                 NFSUNLOCKMNT(nmp);
5375                                 NFSCL_DEBUG(4, "fnd same addr\n");
5376                                 return (0);
5377                         }
5378                         tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5379                         if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5380                                 msad = (struct sockaddr_in *)
5381                                     tdsp->nfsclds_sockp->nr_nam;
5382                         else
5383                                 msad = NULL;
5384                 }
5385                 NFSUNLOCKMNT(nmp);
5386
5387                 /* No IP address match, so look for new/trunked one. */
5388                 sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5389                 sad->sin_len = sizeof(*sad);
5390                 sad->sin_family = AF_INET;
5391                 sad->sin_port = sin->sin_port;
5392                 sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5393                 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5394                 nrp->nr_nam = (struct sockaddr *)sad;
5395         } else if (af == AF_INET6) {
5396                 NFSLOCKMNT(nmp);
5397                 /*
5398                  * Check to see if we already have a session for this
5399                  * address that is usable for a DS.
5400                  * Note that the MDS's address is in a different place
5401                  * than the sessions already acquired for DS's.
5402                  */
5403                 msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5404                 tdsp = TAILQ_FIRST(&nmp->nm_sess);
5405                 while (tdsp != NULL) {
5406                         if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5407                             IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5408                             &msad6->sin6_addr) &&
5409                             sin6->sin6_port == msad6->sin6_port &&
5410                             (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5411                             tdsp->nfsclds_sess.nfsess_defunct == 0) {
5412                                 *dspp = tdsp;
5413                                 NFSUNLOCKMNT(nmp);
5414                                 return (0);
5415                         }
5416                         tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5417                         if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5418                                 msad6 = (struct sockaddr_in6 *)
5419                                     tdsp->nfsclds_sockp->nr_nam;
5420                         else
5421                                 msad6 = NULL;
5422                 }
5423                 NFSUNLOCKMNT(nmp);
5424
5425                 /* No IP address match, so look for new/trunked one. */
5426                 sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5427                 sad6->sin6_len = sizeof(*sad6);
5428                 sad6->sin6_family = AF_INET6;
5429                 sad6->sin6_port = sin6->sin6_port;
5430                 NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5431                     sizeof(struct in6_addr));
5432                 nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5433                 nrp->nr_nam = (struct sockaddr *)sad6;
5434         } else
5435                 return (EPERM);
5436
5437         nrp->nr_sotype = SOCK_STREAM;
5438         mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5439         nrp->nr_prog = NFS_PROG;
5440         nrp->nr_vers = vers;
5441
5442         /*
5443          * Use the credentials that were used for the mount, which are
5444          * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5445          * Ref. counting the credentials with crhold() is probably not
5446          * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5447          * unmount, but I did it anyhow.
5448          */
5449         nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5450         error = newnfs_connect(nmp, nrp, NULL, p, 0);
5451         NFSCL_DEBUG(3, "DS connect=%d\n", error);
5452
5453         dsp = NULL;
5454         /* Now, do the exchangeid and create session. */
5455         if (error == 0) {
5456                 if (vers == NFS_VER4) {
5457                         error = nfsrpc_exchangeid(nmp, clp, nrp,
5458                             NFSV4EXCH_USEPNFSDS, &dsp, nrp->nr_cred, p);
5459                         NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5460                         if (error != 0)
5461                                 newnfs_disconnect(nrp);
5462                 } else {
5463                         dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5464                             M_WAITOK | M_ZERO);
5465                         dsp->nfsclds_flags |= NFSCLDS_DS;
5466                         dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5467                         mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5468                         mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5469                             NULL, MTX_DEF);
5470                 }
5471         }
5472         if (error == 0) {
5473                 dsp->nfsclds_sockp = nrp;
5474                 if (vers == NFS_VER4) {
5475                         NFSLOCKMNT(nmp);
5476                         retv = nfscl_getsameserver(nmp, dsp, &tdsp);
5477                         NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5478                         if (retv == NFSDSP_USETHISSESSION) {
5479                                 NFSUNLOCKMNT(nmp);
5480                                 /*
5481                                  * If there is already a session for this
5482                                  * server, use it.
5483                                  */
5484                                 (void)newnfs_disconnect(nrp);
5485                                 nfscl_freenfsclds(dsp);
5486                                 *dspp = tdsp;
5487                                 return (0);
5488                         }
5489                         if (retv == NFSDSP_SEQTHISSESSION)
5490                                 sequenceid =
5491                                     tdsp->nfsclds_sess.nfsess_sequenceid;
5492                         else
5493                                 sequenceid =
5494                                     dsp->nfsclds_sess.nfsess_sequenceid;
5495                         NFSUNLOCKMNT(nmp);
5496                         error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5497                             nrp, sequenceid, 0, nrp->nr_cred, p);
5498                         NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5499                 }
5500         } else {
5501                 NFSFREECRED(nrp->nr_cred);
5502                 NFSFREEMUTEX(&nrp->nr_mtx);
5503                 free(nrp->nr_nam, M_SONAME);
5504                 free(nrp, M_NFSSOCKREQ);
5505         }
5506         if (error == 0) {
5507                 NFSCL_DEBUG(3, "add DS session\n");
5508                 /*
5509                  * Put it at the end of the list. That way the list
5510                  * is ordered by when the entry was added. This matters
5511                  * since the one done first is the one that should be
5512                  * used for sequencid'ing any subsequent create sessions.
5513                  */
5514                 NFSLOCKMNT(nmp);
5515                 TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5516                 NFSUNLOCKMNT(nmp);
5517                 *dspp = dsp;
5518         } else if (dsp != NULL) {
5519                 newnfs_disconnect(nrp);
5520                 nfscl_freenfsclds(dsp);
5521         }
5522         return (error);
5523 }
5524
5525 /*
5526  * Do the NFSv4.1 Reclaim Complete.
5527  */
5528 int
5529 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5530 {
5531         uint32_t *tl;
5532         struct nfsrv_descript nfsd;
5533         struct nfsrv_descript *nd = &nfsd;
5534         int error;
5535
5536         nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5537             0);
5538         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5539         *tl = newnfs_false;
5540         nd->nd_flag |= ND_USEGSSNAME;
5541         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5542             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5543         if (error != 0)
5544                 return (error);
5545         error = nd->nd_repstat;
5546         mbuf_freem(nd->nd_mrep);
5547         return (error);
5548 }
5549
5550 /*
5551  * Initialize the slot tables for a session.
5552  */
5553 static void
5554 nfscl_initsessionslots(struct nfsclsession *sep)
5555 {
5556         int i;
5557
5558         for (i = 0; i < NFSV4_CBSLOTS; i++) {
5559                 if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5560                         m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5561                 NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5562         }
5563         for (i = 0; i < 64; i++)
5564                 sep->nfsess_slotseq[i] = 0;
5565         sep->nfsess_slots = 0;
5566 }
5567
5568 /*
5569  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5570  */
5571 int
5572 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5573     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5574 {
5575         struct nfsnode *np = VTONFS(vp);
5576         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5577         struct nfscllayout *layp;
5578         struct nfscldevinfo *dip;
5579         struct nfsclflayout *rflp;
5580         struct mbuf *m;
5581         struct nfsclwritedsdorpc *drpc, *tdrpc;
5582         nfsv4stateid_t stateid;
5583         struct ucred *newcred;
5584         uint64_t lastbyte, len, off, oresid, xfer;
5585         int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5586         void *lckp;
5587         uint8_t *dev;
5588         void *iovbase;
5589         size_t iovlen;
5590         off_t offs;
5591         ssize_t resid;
5592
5593         if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5594             (np->n_flag & NNOLAYOUT) != 0)
5595                 return (EIO);
5596         /* Now, get a reference cnt on the clientid for this mount. */
5597         if (nfscl_getref(nmp) == 0)
5598                 return (EIO);
5599
5600         /* Find an appropriate stateid. */
5601         newcred = NFSNEWCRED(cred);
5602         error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5603             rwaccess, 1, newcred, p, &stateid, &lckp);
5604         if (error != 0) {
5605                 NFSFREECRED(newcred);
5606                 nfscl_relref(nmp);
5607                 return (error);
5608         }
5609         /* Search for a layout for this file. */
5610         off = uiop->uio_offset;
5611         layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5612             np->n_fhp->nfh_len, off, &rflp, &recalled);
5613         if (layp == NULL || rflp == NULL) {
5614                 if (recalled != 0) {
5615                         NFSFREECRED(newcred);
5616                         nfscl_relref(nmp);
5617                         return (EIO);
5618                 }
5619                 if (layp != NULL) {
5620                         nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5621                         layp = NULL;
5622                 }
5623                 /* Try and get a Layout, if it is supported. */
5624                 if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5625                     (np->n_flag & NWRITEOPENED) != 0)
5626                         iolaymode = NFSLAYOUTIOMODE_RW;
5627                 else
5628                         iolaymode = NFSLAYOUTIOMODE_READ;
5629                 error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5630                     NULL, &stateid, off, &layp, newcred, p);
5631                 if (error != 0) {
5632                         NFSLOCKNODE(np);
5633                         np->n_flag |= NNOLAYOUT;
5634                         NFSUNLOCKNODE(np);
5635                         if (lckp != NULL)
5636                                 nfscl_lockderef(lckp);
5637                         NFSFREECRED(newcred);
5638                         if (layp != NULL)
5639                                 nfscl_rellayout(layp, 0);
5640                         nfscl_relref(nmp);
5641                         return (error);
5642                 }
5643         }
5644
5645         /*
5646          * Loop around finding a layout that works for the first part of
5647          * this I/O operation, and then call the function that actually
5648          * does the RPC.
5649          */
5650         eof = 0;
5651         len = (uint64_t)uiop->uio_resid;
5652         while (len > 0 && error == 0 && eof == 0) {
5653                 off = uiop->uio_offset;
5654                 error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5655                 if (error == 0) {
5656                         oresid = xfer = (uint64_t)uiop->uio_resid;
5657                         if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5658                                 xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5659                         /*
5660                          * For Flex File layout with mirrored DSs, select one
5661                          * of them at random for reads. For writes and commits,
5662                          * do all mirrors.
5663                          */
5664                         m = NULL;
5665                         drpc = NULL;
5666                         firstmirror = 0;
5667                         mirrorcnt = 1;
5668                         if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5669                             (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5670                                 if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5671                                         firstmirror = arc4random() % mirrorcnt;
5672                                         mirrorcnt = firstmirror + 1;
5673                                 } else {
5674                                         if (docommit == 0) {
5675                                                 /*
5676                                                  * Save values, so uiop can be
5677                                                  * rolled back upon a write
5678                                                  * error.
5679                                                  */
5680                                                 offs = uiop->uio_offset;
5681                                                 resid = uiop->uio_resid;
5682                                                 iovbase =
5683                                                     uiop->uio_iov->iov_base;
5684                                                 iovlen = uiop->uio_iov->iov_len;
5685                                                 m = nfsm_uiombuflist(uiop, len,
5686                                                     NULL, NULL);
5687                                         }
5688                                         tdrpc = drpc = malloc(sizeof(*drpc) *
5689                                             (mirrorcnt - 1), M_TEMP, M_WAITOK |
5690                                             M_ZERO);
5691                                 }
5692                         }
5693                         for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5694                                 if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0)
5695                                         dev = rflp->nfsfl_ffm[i].dev;
5696                                 else
5697                                         dev = rflp->nfsfl_dev;
5698                                 dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5699                                     rflp->nfsfl_devp);
5700                                 if (dip != NULL) {
5701                                         if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5702                                             != 0)
5703                                                 error = nfscl_dofflayoutio(vp,
5704                                                     uiop, iomode, must_commit,
5705                                                     &eof, &stateid, rwaccess,
5706                                                     dip, layp, rflp, off, xfer,
5707                                                     i, docommit, m, tdrpc,
5708                                                     newcred, p);
5709                                         else
5710                                                 error = nfscl_doflayoutio(vp,
5711                                                     uiop, iomode, must_commit,
5712                                                     &eof, &stateid, rwaccess,
5713                                                     dip, layp, rflp, off, xfer,
5714                                                     docommit, newcred, p);
5715                                         nfscl_reldevinfo(dip);
5716                                 } else
5717                                         error = EIO;
5718                                 tdrpc++;
5719                         }
5720                         if (m != NULL)
5721                                 m_freem(m);
5722                         tdrpc = drpc;
5723                         timo = hz / 50;         /* Wait for 20msec. */
5724                         if (timo < 1)
5725                                 timo = 1;
5726                         for (i = firstmirror; i < mirrorcnt - 1 &&
5727                             tdrpc != NULL; i++, tdrpc++) {
5728                                 /*
5729                                  * For the unused drpc entries, both inprog and
5730                                  * err == 0, so this loop won't break.
5731                                  */
5732                                 while (tdrpc->inprog != 0 && tdrpc->done == 0)
5733                                         tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5734                                             timo);
5735                                 if (error == 0 && tdrpc->err != 0)
5736                                         error = tdrpc->err;
5737                         }
5738                         free(drpc, M_TEMP);
5739                         if (error == 0) {
5740                                 if (mirrorcnt > 1 && rwaccess ==
5741                                     NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5742                                         NFSLOCKCLSTATE();
5743                                         layp->nfsly_flags |= NFSLY_WRITTEN;
5744                                         NFSUNLOCKCLSTATE();
5745                                 }
5746                                 lastbyte = off + xfer - 1;
5747                                 NFSLOCKCLSTATE();
5748                                 if (lastbyte > layp->nfsly_lastbyte)
5749                                         layp->nfsly_lastbyte = lastbyte;
5750                                 NFSUNLOCKCLSTATE();
5751                         } else if (error == NFSERR_OPENMODE &&
5752                             rwaccess == NFSV4OPEN_ACCESSREAD) {
5753                                 NFSLOCKMNT(nmp);
5754                                 nmp->nm_state |= NFSSTA_OPENMODE;
5755                                 NFSUNLOCKMNT(nmp);
5756                         } else
5757                                 error = EIO;
5758                         if (error == 0)
5759                                 len -= (oresid - (uint64_t)uiop->uio_resid);
5760                         else if (mirrorcnt > 1 && rwaccess ==
5761                             NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5762                                 /*
5763                                  * In case the rpc gets retried, roll the
5764                                  * uio fields changed by nfsm_uiombuflist()
5765                                  * back.
5766                                  */
5767                                 uiop->uio_offset = offs;
5768                                 uiop->uio_resid = resid;
5769                                 uiop->uio_iov->iov_base = iovbase;
5770                                 uiop->uio_iov->iov_len = iovlen;
5771                         }
5772                 }
5773         }
5774         if (lckp != NULL)
5775                 nfscl_lockderef(lckp);
5776         NFSFREECRED(newcred);
5777         nfscl_rellayout(layp, 0);
5778         nfscl_relref(nmp);
5779         return (error);
5780 }
5781
5782 /*
5783  * Make a copy of the mbuf chain and add an mbuf for null padding, as required.
5784  */
5785 static struct mbuf *
5786 nfsm_copym(struct mbuf *m, int off, int xfer)
5787 {
5788         struct mbuf *m2, *m3, *m4;
5789         uint32_t *tl;
5790         int rem;
5791
5792         m2 = m_copym(m, off, xfer, M_WAITOK);
5793         rem = NFSM_RNDUP(xfer) - xfer;
5794         if (rem > 0) {
5795                 /*
5796                  * The zero padding to a multiple of 4 bytes is required by
5797                  * the XDR. So that the mbufs copied by reference aren't
5798                  * modified, add an mbuf with the zero'd bytes to the list.
5799                  * rem will be a maximum of 3, so one zero'd uint32_t is
5800                  * sufficient.
5801                  */
5802                 m3 = m2;
5803                 while (m3->m_next != NULL)
5804                         m3 = m3->m_next;
5805                 NFSMGET(m4);
5806                 tl = NFSMTOD(m4, uint32_t *);
5807                 *tl = 0;
5808                 mbuf_setlen(m4, rem);
5809                 mbuf_setnext(m3, m4);
5810         }
5811         return (m2);
5812 }
5813
5814 /*
5815  * Find a file layout that will handle the first bytes of the requested
5816  * range and return the information from it needed to to the I/O operation.
5817  */
5818 int
5819 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5820     struct nfsclflayout **retflpp)
5821 {
5822         struct nfsclflayout *flp, *nflp, *rflp;
5823         uint32_t rw;
5824
5825         rflp = NULL;
5826         rw = rwaccess;
5827         /* For reading, do the Read list first and then the Write list. */
5828         do {
5829                 if (rw == NFSV4OPEN_ACCESSREAD)
5830                         flp = LIST_FIRST(&lyp->nfsly_flayread);
5831                 else
5832                         flp = LIST_FIRST(&lyp->nfsly_flayrw);
5833                 while (flp != NULL) {
5834                         nflp = LIST_NEXT(flp, nfsfl_list);
5835                         if (flp->nfsfl_off > off)
5836                                 break;
5837                         if (flp->nfsfl_end > off &&
5838                             (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5839                                 rflp = flp;
5840                         flp = nflp;
5841                 }
5842                 if (rw == NFSV4OPEN_ACCESSREAD)
5843                         rw = NFSV4OPEN_ACCESSWRITE;
5844                 else
5845                         rw = 0;
5846         } while (rw != 0);
5847         if (rflp != NULL) {
5848                 /* This one covers the most bytes starting at off. */
5849                 *retflpp = rflp;
5850                 return (0);
5851         }
5852         return (EIO);
5853 }
5854
5855 /*
5856  * Do I/O using an NFSv4.1 file layout.
5857  */
5858 static int
5859 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5860     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5861     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5862     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
5863 {
5864         uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
5865         int commit_thru_mds, error, stripe_index, stripe_pos;
5866         struct nfsnode *np;
5867         struct nfsfh *fhp;
5868         struct nfsclds **dspp;
5869
5870         np = VTONFS(vp);
5871         rel_off = off - flp->nfsfl_patoff;
5872         stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
5873         stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
5874             dp->nfsdi_stripecnt;
5875         transfer = stripe_unit_size - (rel_off % stripe_unit_size);
5876         error = 0;
5877
5878         /* Loop around, doing I/O for each stripe unit. */
5879         while (len > 0 && error == 0) {
5880                 stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
5881                 dspp = nfsfldi_addr(dp, stripe_index);
5882                 if (len > transfer && docommit == 0)
5883                         xfer = transfer;
5884                 else
5885                         xfer = len;
5886                 if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
5887                         /* Dense layout. */
5888                         if (stripe_pos >= flp->nfsfl_fhcnt)
5889                                 return (EIO);
5890                         fhp = flp->nfsfl_fh[stripe_pos];
5891                         io_off = (rel_off / (stripe_unit_size *
5892                             dp->nfsdi_stripecnt)) * stripe_unit_size +
5893                             rel_off % stripe_unit_size;
5894                 } else {
5895                         /* Sparse layout. */
5896                         if (flp->nfsfl_fhcnt > 1) {
5897                                 if (stripe_index >= flp->nfsfl_fhcnt)
5898                                         return (EIO);
5899                                 fhp = flp->nfsfl_fh[stripe_index];
5900                         } else if (flp->nfsfl_fhcnt == 1)
5901                                 fhp = flp->nfsfl_fh[0];
5902                         else
5903                                 fhp = np->n_fhp;
5904                         io_off = off;
5905                 }
5906                 if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
5907                         commit_thru_mds = 1;
5908                         if (docommit != 0)
5909                                 error = EIO;
5910                 } else {
5911                         commit_thru_mds = 0;
5912                         mtx_lock(&np->n_mtx);
5913                         np->n_flag |= NDSCOMMIT;
5914                         mtx_unlock(&np->n_mtx);
5915                 }
5916                 if (docommit != 0) {
5917                         if (error == 0)
5918                                 error = nfsrpc_commitds(vp, io_off, xfer,
5919                                     *dspp, fhp, 0, 0, cred, p);
5920                         if (error == 0) {
5921                                 /*
5922                                  * Set both eof and uio_resid = 0 to end any
5923                                  * loops.
5924                                  */
5925                                 *eofp = 1;
5926                                 uiop->uio_resid = 0;
5927                         } else {
5928                                 mtx_lock(&np->n_mtx);
5929                                 np->n_flag &= ~NDSCOMMIT;
5930                                 mtx_unlock(&np->n_mtx);
5931                         }
5932                 } else if (rwflag == NFSV4OPEN_ACCESSREAD)
5933                         error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
5934                             io_off, xfer, fhp, 0, 0, 0, cred, p);
5935                 else {
5936                         error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
5937                             stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
5938                             0, 0, 0, cred, p);
5939                         if (error == 0) {
5940                                 NFSLOCKCLSTATE();
5941                                 lyp->nfsly_flags |= NFSLY_WRITTEN;
5942                                 NFSUNLOCKCLSTATE();
5943                         }
5944                 }
5945                 if (error == 0) {
5946                         transfer = stripe_unit_size;
5947                         stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
5948                         len -= xfer;
5949                         off += xfer;
5950                 }
5951         }
5952         return (error);
5953 }
5954
5955 /*
5956  * Do I/O using an NFSv4.1 flex file layout.
5957  */
5958 static int
5959 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5960     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5961     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5962     uint64_t len, int mirror, int docommit, struct mbuf *mp,
5963     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
5964 {
5965         uint64_t transfer, xfer;
5966         int error, rel_off;
5967         struct nfsnode *np;
5968         struct nfsfh *fhp;
5969         struct nfsclds **dspp;
5970         struct ucred *tcred;
5971         struct mbuf *m;
5972
5973         np = VTONFS(vp);
5974         error = 0;
5975         rel_off = 0;
5976         NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
5977             (uintmax_t)len);
5978         /* Loop around, doing I/O for each stripe unit. */
5979         while (len > 0 && error == 0) {
5980                 dspp = nfsfldi_addr(dp, 0);
5981                 fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
5982                 stateidp = &flp->nfsfl_ffm[mirror].st;
5983                 NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
5984                     mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
5985                 if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
5986                         tcred = NFSNEWCRED(cred);
5987                         tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
5988                         tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
5989                         tcred->cr_ngroups = 1;
5990                 } else
5991                         tcred = cred;
5992                 if (rwflag == NFSV4OPEN_ACCESSREAD)
5993                         transfer = dp->nfsdi_rsize;
5994                 else
5995                         transfer = dp->nfsdi_wsize;
5996                 mtx_lock(&np->n_mtx);
5997                 np->n_flag |= NDSCOMMIT;
5998                 mtx_unlock(&np->n_mtx);
5999                 if (len > transfer && docommit == 0)
6000                         xfer = transfer;
6001                 else
6002                         xfer = len;
6003                 if (docommit != 0) {
6004                         if (error == 0) {
6005                                 /*
6006                                  * Do last mirrored DS commit with this thread.
6007                                  */
6008                                 if (mirror < flp->nfsfl_mirrorcnt - 1)
6009                                         error = nfsio_commitds(vp, off, xfer,
6010                                             *dspp, fhp, dp->nfsdi_vers,
6011                                             dp->nfsdi_minorvers, drpc, tcred,
6012                                             p);
6013                                 else
6014                                         error = nfsrpc_commitds(vp, off, xfer,
6015                                             *dspp, fhp, dp->nfsdi_vers,
6016                                             dp->nfsdi_minorvers, tcred, p);
6017                         }
6018                         NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6019                         if (error == 0) {
6020                                 /*
6021                                  * Set both eof and uio_resid = 0 to end any
6022                                  * loops.
6023                                  */
6024                                 *eofp = 1;
6025                                 uiop->uio_resid = 0;
6026                         } else {
6027                                 mtx_lock(&np->n_mtx);
6028                                 np->n_flag &= ~NDSCOMMIT;
6029                                 mtx_unlock(&np->n_mtx);
6030                         }
6031                 } else if (rwflag == NFSV4OPEN_ACCESSREAD)
6032                         error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6033                             off, xfer, fhp, 1, dp->nfsdi_vers,
6034                             dp->nfsdi_minorvers, tcred, p);
6035                 else {
6036                         if (flp->nfsfl_mirrorcnt == 1) {
6037                                 error = nfsrpc_writeds(vp, uiop, iomode,
6038                                     must_commit, stateidp, *dspp, off, xfer,
6039                                     fhp, 0, 1, dp->nfsdi_vers,
6040                                     dp->nfsdi_minorvers, tcred, p);
6041                                 if (error == 0) {
6042                                         NFSLOCKCLSTATE();
6043                                         lyp->nfsly_flags |= NFSLY_WRITTEN;
6044                                         NFSUNLOCKCLSTATE();
6045                                 }
6046                         } else {
6047                                 m = nfsm_copym(mp, rel_off, xfer);
6048                                 NFSCL_DEBUG(4, "mcopy reloff=%d xfer=%jd\n",
6049                                     rel_off, (uintmax_t)xfer);
6050                                 /*
6051                                  * Do last write to a mirrored DS with this
6052                                  * thread.
6053                                  */
6054                                 if (mirror < flp->nfsfl_mirrorcnt - 1)
6055                                         error = nfsio_writedsmir(vp, iomode,
6056                                             must_commit, stateidp, *dspp, off,
6057                                             xfer, fhp, m, dp->nfsdi_vers,
6058                                             dp->nfsdi_minorvers, drpc, tcred,
6059                                             p);
6060                                 else
6061                                         error = nfsrpc_writedsmir(vp, iomode,
6062                                             must_commit, stateidp, *dspp, off,
6063                                             xfer, fhp, m, dp->nfsdi_vers,
6064                                             dp->nfsdi_minorvers, tcred, p);
6065                                 NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6066                         }
6067                 }
6068                 NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6069                 if (error == 0) {
6070                         len -= xfer;
6071                         off += xfer;
6072                         rel_off += xfer;
6073                 }
6074                 if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6075                         NFSFREECRED(tcred);
6076         }
6077         NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6078         return (error);
6079 }
6080
6081 /*
6082  * The actual read RPC done to a DS.
6083  */
6084 static int
6085 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6086     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6087     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6088 {
6089         uint32_t *tl;
6090         int attrflag, error, retlen;
6091         struct nfsrv_descript nfsd;
6092         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6093         struct nfsrv_descript *nd = &nfsd;
6094         struct nfssockreq *nrp;
6095         struct nfsvattr na;
6096
6097         nd->nd_mrep = NULL;
6098         if (vers == 0 || vers == NFS_VER4) {
6099                 nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6100                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6101                 vers = NFS_VER4;
6102                 NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6103                 if (flex != 0)
6104                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6105                 else
6106                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6107         } else {
6108                 nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6109                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6110                 NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6111         }
6112         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6113         txdr_hyper(io_off, tl);
6114         *(tl + 2) = txdr_unsigned(len);
6115         nrp = dsp->nfsclds_sockp;
6116         NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6117         if (nrp == NULL)
6118                 /* If NULL, use the MDS socket. */
6119                 nrp = &nmp->nm_sockreq;
6120         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6121             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6122         NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6123             error);
6124         if (error != 0)
6125                 return (error);
6126         if (vers == NFS_VER3) {
6127                 error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6128                 NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6129                 if (error != 0)
6130                         goto nfsmout;
6131         }
6132         if (nd->nd_repstat != 0) {
6133                 error = nd->nd_repstat;
6134                 goto nfsmout;
6135         }
6136         if (vers == NFS_VER3) {
6137                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6138                 *eofp = fxdr_unsigned(int, *(tl + 1));
6139         } else {
6140                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6141                 *eofp = fxdr_unsigned(int, *tl);
6142         }
6143         NFSM_STRSIZ(retlen, len);
6144         NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6145         error = nfsm_mbufuio(nd, uiop, retlen);
6146 nfsmout:
6147         if (nd->nd_mrep != NULL)
6148                 mbuf_freem(nd->nd_mrep);
6149         return (error);
6150 }
6151
6152 /*
6153  * The actual write RPC done to a DS.
6154  */
6155 static int
6156 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6157     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6158     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6159     struct ucred *cred, NFSPROC_T *p)
6160 {
6161         uint32_t *tl;
6162         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6163         int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6164         int32_t backup;
6165         struct nfsrv_descript nfsd;
6166         struct nfsrv_descript *nd = &nfsd;
6167         struct nfssockreq *nrp;
6168         struct nfsvattr na;
6169
6170         KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6171         nd->nd_mrep = NULL;
6172         if (vers == 0 || vers == NFS_VER4) {
6173                 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6174                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6175                 NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6176                 vers = NFS_VER4;
6177                 if (flex != 0)
6178                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6179                 else
6180                         nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6181                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6182         } else {
6183                 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6184                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6185                 NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6186                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6187         }
6188         txdr_hyper(io_off, tl);
6189         tl += 2;
6190         if (vers == NFS_VER3)
6191                 *tl++ = txdr_unsigned(len);
6192         *tl++ = txdr_unsigned(*iomode);
6193         *tl = txdr_unsigned(len);
6194         nfsm_uiombuf(nd, uiop, len);
6195         nrp = dsp->nfsclds_sockp;
6196         if (nrp == NULL)
6197                 /* If NULL, use the MDS socket. */
6198                 nrp = &nmp->nm_sockreq;
6199         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6200             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6201         NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6202             nd->nd_repstat);
6203         if (error != 0)
6204                 return (error);
6205         if (nd->nd_repstat != 0) {
6206                 /*
6207                  * In case the rpc gets retried, roll
6208                  * the uio fileds changed by nfsm_uiombuf()
6209                  * back.
6210                  */
6211                 uiop->uio_offset -= len;
6212                 uio_uio_resid_add(uiop, len);
6213                 uio_iov_base_add(uiop, -len);
6214                 uio_iov_len_add(uiop, len);
6215                 error = nd->nd_repstat;
6216         } else {
6217                 if (vers == NFS_VER3) {
6218                         error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6219                             NULL);
6220                         NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6221                         if (error != 0)
6222                                 goto nfsmout;
6223                 }
6224                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6225                 rlen = fxdr_unsigned(int, *tl++);
6226                 NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6227                 if (rlen == 0) {
6228                         error = NFSERR_IO;
6229                         goto nfsmout;
6230                 } else if (rlen < len) {
6231                         backup = len - rlen;
6232                         uio_iov_base_add(uiop, -(backup));
6233                         uio_iov_len_add(uiop, backup);
6234                         uiop->uio_offset -= backup;
6235                         uio_uio_resid_add(uiop, backup);
6236                         len = rlen;
6237                 }
6238                 commit = fxdr_unsigned(int, *tl++);
6239
6240                 /*
6241                  * Return the lowest commitment level
6242                  * obtained by any of the RPCs.
6243                  */
6244                 if (committed == NFSWRITE_FILESYNC)
6245                         committed = commit;
6246                 else if (committed == NFSWRITE_DATASYNC &&
6247                     commit == NFSWRITE_UNSTABLE)
6248                         committed = commit;
6249                 if (commit_thru_mds != 0) {
6250                         NFSLOCKMNT(nmp);
6251                         if (!NFSHASWRITEVERF(nmp)) {
6252                                 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6253                                 NFSSETWRITEVERF(nmp);
6254                         } else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6255                                 *must_commit = 1;
6256                                 NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6257                         }
6258                         NFSUNLOCKMNT(nmp);
6259                 } else {
6260                         NFSLOCKDS(dsp);
6261                         if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6262                                 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6263                                 dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6264                         } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6265                                 *must_commit = 1;
6266                                 NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6267                         }
6268                         NFSUNLOCKDS(dsp);
6269                 }
6270         }
6271 nfsmout:
6272         if (nd->nd_mrep != NULL)
6273                 mbuf_freem(nd->nd_mrep);
6274         *iomode = committed;
6275         if (nd->nd_repstat != 0 && error == 0)
6276                 error = nd->nd_repstat;
6277         return (error);
6278 }
6279
6280 /*
6281  * The actual write RPC done to a DS.
6282  * This variant is called from a separate kernel process for mirrors.
6283  * Any short write is considered an IO error.
6284  */
6285 static int
6286 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6287     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6288     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6289     struct ucred *cred, NFSPROC_T *p)
6290 {
6291         uint32_t *tl;
6292         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6293         int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6294         struct nfsrv_descript nfsd;
6295         struct nfsrv_descript *nd = &nfsd;
6296         struct nfssockreq *nrp;
6297         struct nfsvattr na;
6298
6299         nd->nd_mrep = NULL;
6300         if (vers == 0 || vers == NFS_VER4) {
6301                 nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6302                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6303                 vers = NFS_VER4;
6304                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6305                     minorvers);
6306                 nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6307                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6308         } else {
6309                 nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6310                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6311                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6312                 NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6313         }
6314         txdr_hyper(io_off, tl);
6315         tl += 2;
6316         if (vers == NFS_VER3)
6317                 *tl++ = txdr_unsigned(len);
6318         *tl++ = txdr_unsigned(*iomode);
6319         *tl = txdr_unsigned(len);
6320         if (len > 0) {
6321                 /* Put data in mbuf chain. */
6322                 nd->nd_mb->m_next = m;
6323                 /* Set nd_mb and nd_bpos to end of data. */
6324                 while (m->m_next != NULL)
6325                         m = m->m_next;
6326                 nd->nd_mb = m;
6327                 nd->nd_bpos = mtod(m, char *) + m->m_len;
6328                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: lastmb len=%d\n", m->m_len);
6329         }
6330         nrp = dsp->nfsclds_sockp;
6331         if (nrp == NULL)
6332                 /* If NULL, use the MDS socket. */
6333                 nrp = &nmp->nm_sockreq;
6334         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6335             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6336         NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6337             nd->nd_repstat);
6338         if (error != 0)
6339                 return (error);
6340         if (nd->nd_repstat != 0)
6341                 error = nd->nd_repstat;
6342         else {
6343                 if (vers == NFS_VER3) {
6344                         error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6345                             NULL);
6346                         NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6347                             error);
6348                         if (error != 0)
6349                                 goto nfsmout;
6350                 }
6351                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6352                 rlen = fxdr_unsigned(int, *tl++);
6353                 NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6354                     rlen);
6355                 if (rlen != len) {
6356                         error = NFSERR_IO;
6357                         NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6358                             len, rlen);
6359                         goto nfsmout;
6360                 }
6361                 commit = fxdr_unsigned(int, *tl++);
6362
6363                 /*
6364                  * Return the lowest commitment level
6365                  * obtained by any of the RPCs.
6366                  */
6367                 if (committed == NFSWRITE_FILESYNC)
6368                         committed = commit;
6369                 else if (committed == NFSWRITE_DATASYNC &&
6370                     commit == NFSWRITE_UNSTABLE)
6371                         committed = commit;
6372                 NFSLOCKDS(dsp);
6373                 if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6374                         NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6375                         dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6376                 } else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6377                         *must_commit = 1;
6378                         NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6379                 }
6380                 NFSUNLOCKDS(dsp);
6381         }
6382 nfsmout:
6383         if (nd->nd_mrep != NULL)
6384                 mbuf_freem(nd->nd_mrep);
6385         *iomode = committed;
6386         if (nd->nd_repstat != 0 && error == 0)
6387                 error = nd->nd_repstat;
6388         return (error);
6389 }
6390
6391 /*
6392  * Start up the thread that will execute nfsrpc_writedsmir().
6393  */
6394 static void
6395 start_writedsmir(void *arg, int pending)
6396 {
6397         struct nfsclwritedsdorpc *drpc;
6398
6399         drpc = (struct nfsclwritedsdorpc *)arg;
6400         drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6401             &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6402             drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6403             drpc->p);
6404         drpc->done = 1;
6405         NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6406 }
6407
6408 /*
6409  * Set up the write DS mirror call for the pNFS I/O thread.
6410  */
6411 static int
6412 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6413     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6414     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6415     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6416 {
6417         int error, ret;
6418
6419         error = 0;
6420         drpc->done = 0;
6421         drpc->vp = vp;
6422         drpc->iomode = *iomode;
6423         drpc->must_commit = *must_commit;
6424         drpc->stateidp = stateidp;
6425         drpc->dsp = dsp;
6426         drpc->off = off;
6427         drpc->len = len;
6428         drpc->fhp = fhp;
6429         drpc->m = m;
6430         drpc->vers = vers;
6431         drpc->minorvers = minorvers;
6432         drpc->cred = cred;
6433         drpc->p = p;
6434         drpc->inprog = 0;
6435         ret = EIO;
6436         if (nfs_pnfsiothreads > 0) {
6437                 ret = nfs_pnfsio(start_writedsmir, drpc);
6438                 NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6439         }
6440         if (ret != 0)
6441                 error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6442                     dsp, off, len, fhp, m, vers, minorvers, cred, p);
6443         NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6444         return (error);
6445 }
6446
6447 /*
6448  * Free up the nfsclds structure.
6449  */
6450 void
6451 nfscl_freenfsclds(struct nfsclds *dsp)
6452 {
6453         int i;
6454
6455         if (dsp == NULL)
6456                 return;
6457         if (dsp->nfsclds_sockp != NULL) {
6458                 NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6459                 NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6460                 free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6461                 free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6462         }
6463         NFSFREEMUTEX(&dsp->nfsclds_mtx);
6464         NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6465         for (i = 0; i < NFSV4_CBSLOTS; i++) {
6466                 if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6467                         m_freem(
6468                             dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6469         }
6470         free(dsp, M_NFSCLDS);
6471 }
6472
6473 static enum nfsclds_state
6474 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6475     struct nfsclds **retdspp)
6476 {
6477         struct nfsclds *dsp, *cur_dsp;
6478
6479         /*
6480          * Search the list of nfsclds structures for one with the same
6481          * server.
6482          */
6483         cur_dsp = NULL;
6484         TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6485                 if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6486                     dsp->nfsclds_servownlen != 0 &&
6487                     !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6488                     dsp->nfsclds_servownlen) &&
6489                     dsp->nfsclds_sess.nfsess_defunct == 0) {
6490                         NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6491                             TAILQ_FIRST(&nmp->nm_sess), dsp,
6492                             dsp->nfsclds_flags);
6493                         /* Server major id matches. */
6494                         if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6495                                 *retdspp = dsp;
6496                                 return (NFSDSP_USETHISSESSION);
6497                         }
6498
6499                         /*
6500                          * Note the first match, so it can be used for
6501                          * sequence'ing new sessions.
6502                          */
6503                         if (cur_dsp == NULL)
6504                                 cur_dsp = dsp;
6505                 }
6506         }
6507         if (cur_dsp != NULL) {
6508                 *retdspp = cur_dsp;
6509                 return (NFSDSP_SEQTHISSESSION);
6510         }
6511         return (NFSDSP_NOTFOUND);
6512 }
6513
6514 /*
6515  * NFS commit rpc to a NFSv4.1 DS.
6516  */
6517 static int
6518 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6519     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6520     NFSPROC_T *p)
6521 {
6522         uint32_t *tl;
6523         struct nfsrv_descript nfsd, *nd = &nfsd;
6524         struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6525         struct nfssockreq *nrp;
6526         struct nfsvattr na;
6527         int attrflag, error;
6528
6529         nd->nd_mrep = NULL;
6530         if (vers == 0 || vers == NFS_VER4) {
6531                 nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6532                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6533                 vers = NFS_VER4;
6534         } else
6535                 nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6536                     fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6537         NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6538             minorvers);
6539         NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6540         txdr_hyper(offset, tl);
6541         tl += 2;
6542         *tl = txdr_unsigned(cnt);
6543         nrp = dsp->nfsclds_sockp;
6544         if (nrp == NULL)
6545                 /* If NULL, use the MDS socket. */
6546                 nrp = &nmp->nm_sockreq;
6547         error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6548             NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6549         NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6550             nd->nd_repstat);
6551         if (error != 0)
6552                 return (error);
6553         if (nd->nd_repstat == 0) {
6554                 if (vers == NFS_VER3) {
6555                         error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6556                             NULL);
6557                         NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6558                         if (error != 0)
6559                                 goto nfsmout;
6560                 }
6561                 NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6562                 NFSLOCKDS(dsp);
6563                 if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6564                         NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6565                         error = NFSERR_STALEWRITEVERF;
6566                 }
6567                 NFSUNLOCKDS(dsp);
6568         }
6569 nfsmout:
6570         if (error == 0 && nd->nd_repstat != 0)
6571                 error = nd->nd_repstat;
6572         mbuf_freem(nd->nd_mrep);
6573         return (error);
6574 }
6575
6576 /*
6577  * Start up the thread that will execute nfsrpc_commitds().
6578  */
6579 static void
6580 start_commitds(void *arg, int pending)
6581 {
6582         struct nfsclwritedsdorpc *drpc;
6583
6584         drpc = (struct nfsclwritedsdorpc *)arg;
6585         drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6586             drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6587             drpc->p);
6588         drpc->done = 1;
6589         NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6590 }
6591
6592 /*
6593  * Set up the commit DS mirror call for the pNFS I/O thread.
6594  */
6595 static int
6596 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6597     struct nfsfh *fhp, int vers, int minorvers,
6598     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6599 {
6600         int error, ret;
6601
6602         error = 0;
6603         drpc->done = 0;
6604         drpc->vp = vp;
6605         drpc->off = offset;
6606         drpc->len = cnt;
6607         drpc->dsp = dsp;
6608         drpc->fhp = fhp;
6609         drpc->vers = vers;
6610         drpc->minorvers = minorvers;
6611         drpc->cred = cred;
6612         drpc->p = p;
6613         drpc->inprog = 0;
6614         ret = EIO;
6615         if (nfs_pnfsiothreads > 0) {
6616                 ret = nfs_pnfsio(start_commitds, drpc);
6617                 NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6618         }
6619         if (ret != 0)
6620                 error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6621                     minorvers, cred, p);
6622         NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6623         return (error);
6624 }
6625
6626 /*
6627  * Set up the XDR arguments for the LayoutGet operation.
6628  */
6629 static void
6630 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
6631     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
6632     int layoutlen, int usecurstateid)
6633 {
6634         uint32_t *tl;
6635
6636         NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
6637             NFSX_STATEID);
6638         *tl++ = newnfs_false;           /* Don't signal availability. */
6639         *tl++ = txdr_unsigned(layouttype);
6640         *tl++ = txdr_unsigned(iomode);
6641         txdr_hyper(offset, tl);
6642         tl += 2;
6643         txdr_hyper(len, tl);
6644         tl += 2;
6645         txdr_hyper(minlen, tl);
6646         tl += 2;
6647         if (usecurstateid != 0) {
6648                 /* Special stateid for Current stateid. */
6649                 *tl++ = txdr_unsigned(1);
6650                 *tl++ = 0;
6651                 *tl++ = 0;
6652                 *tl++ = 0;
6653         } else {
6654                 *tl++ = txdr_unsigned(stateidp->seqid);
6655                 NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
6656                 *tl++ = stateidp->other[0];
6657                 *tl++ = stateidp->other[1];
6658                 *tl++ = stateidp->other[2];
6659         }
6660         *tl = txdr_unsigned(layoutlen);
6661 }
6662
6663 /*
6664  * Parse the reply for a successful LayoutGet operation.
6665  */
6666 static int
6667 nfsrv_parselayoutget(struct nfsrv_descript *nd, nfsv4stateid_t *stateidp,
6668     int *retonclosep, struct nfsclflayouthead *flhp)
6669 {
6670         uint32_t *tl;
6671         struct nfsclflayout *flp, *prevflp, *tflp;
6672         int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
6673         int m, mirrorcnt;
6674         uint64_t retlen, off;
6675         struct nfsfh *nfhp;
6676         uint8_t *cp;
6677         uid_t user;
6678         gid_t grp;
6679
6680         NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
6681         error = 0;
6682         flp = NULL;
6683         gotiomode = -1;
6684         NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
6685         if (*tl++ != 0)
6686                 *retonclosep = 1;
6687         else
6688                 *retonclosep = 0;
6689         stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
6690         NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
6691             (int)stateidp->seqid);
6692         stateidp->other[0] = *tl++;
6693         stateidp->other[1] = *tl++;
6694         stateidp->other[2] = *tl++;
6695         cnt = fxdr_unsigned(int, *tl);
6696         NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
6697         if (cnt <= 0 || cnt > 10000) {
6698                 /* Don't accept more than 10000 layouts in reply. */
6699                 error = NFSERR_BADXDR;
6700                 goto nfsmout;
6701         }
6702         for (i = 0; i < cnt; i++) {
6703                 /* Dissect to the layout type. */
6704                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
6705                     3 * NFSX_UNSIGNED);
6706                 off = fxdr_hyper(tl); tl += 2;
6707                 retlen = fxdr_hyper(tl); tl += 2;
6708                 iomode = fxdr_unsigned(int, *tl++);
6709                 laytype = fxdr_unsigned(int, *tl);
6710                 NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
6711                     (uintmax_t)off, (uintmax_t)retlen, iomode);
6712                 /* Ignore length of layout body for now. */
6713                 if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
6714                         /* Parse the File layout up to fhcnt. */
6715                         NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
6716                             NFSX_HYPER + NFSX_V4DEVICEID);
6717                         fhcnt = fxdr_unsigned(int, *(tl + 4 +
6718                             NFSX_V4DEVICEID / NFSX_UNSIGNED));
6719                         NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6720                         if (fhcnt < 0 || fhcnt > 100) {
6721                                 /* Don't accept more than 100 file handles. */
6722                                 error = NFSERR_BADXDR;
6723                                 goto nfsmout;
6724                         }
6725                         if (fhcnt > 0)
6726                                 flp = malloc(sizeof(*flp) + fhcnt *
6727                                     sizeof(struct nfsfh *), M_NFSFLAYOUT,
6728                                     M_WAITOK);
6729                         else
6730                                 flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
6731                                     M_WAITOK);
6732                         flp->nfsfl_flags = NFSFL_FILE;
6733                         flp->nfsfl_fhcnt = 0;
6734                         flp->nfsfl_devp = NULL;
6735                         flp->nfsfl_off = off;
6736                         if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6737                                 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6738                         else
6739                                 flp->nfsfl_end = flp->nfsfl_off + retlen;
6740                         flp->nfsfl_iomode = iomode;
6741                         if (gotiomode == -1)
6742                                 gotiomode = flp->nfsfl_iomode;
6743                         /* Ignore layout body length for now. */
6744                         NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
6745                         tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
6746                         flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
6747                         NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
6748                         flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
6749                         flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
6750                         NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
6751                             flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
6752                         for (j = 0; j < fhcnt; j++) {
6753                                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6754                                 nfhlen = fxdr_unsigned(int, *tl);
6755                                 if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
6756                                         error = NFSERR_BADXDR;
6757                                         goto nfsmout;
6758                                 }
6759                                 nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
6760                                     M_NFSFH, M_WAITOK);
6761                                 flp->nfsfl_fh[j] = nfhp;
6762                                 flp->nfsfl_fhcnt++;
6763                                 nfhp->nfh_len = nfhlen;
6764                                 NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
6765                                 NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
6766                         }
6767                 } else if (laytype == NFSLAYOUT_FLEXFILE) {
6768                         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
6769                             NFSX_HYPER);
6770                         mirrorcnt = fxdr_unsigned(int, *(tl + 2));
6771                         NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
6772                         if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
6773                                 error = NFSERR_BADXDR;
6774                                 goto nfsmout;
6775                         }
6776                         flp = malloc(sizeof(*flp) + mirrorcnt *
6777                             sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
6778                         flp->nfsfl_flags = NFSFL_FLEXFILE;
6779                         flp->nfsfl_mirrorcnt = mirrorcnt;
6780                         flp->nfsfl_devp = NULL;
6781                         flp->nfsfl_off = off;
6782                         if (flp->nfsfl_off + retlen < flp->nfsfl_off)
6783                                 flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
6784                         else
6785                                 flp->nfsfl_end = flp->nfsfl_off + retlen;
6786                         flp->nfsfl_iomode = iomode;
6787                         if (gotiomode == -1)
6788                                 gotiomode = flp->nfsfl_iomode;
6789                         flp->nfsfl_stripeunit = fxdr_hyper(tl);
6790                         NFSCL_DEBUG(4, "stripeunit=%ju\n",
6791                             (uintmax_t)flp->nfsfl_stripeunit);
6792                         for (j = 0; j < mirrorcnt; j++) {
6793                                 NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6794                                 k = fxdr_unsigned(int, *tl);
6795                                 if (k < 1 || k > 128) {
6796                                         error = NFSERR_BADXDR;
6797                                         goto nfsmout;
6798                                 }
6799                                 NFSCL_DEBUG(4, "servercnt=%d\n", k);
6800                                 for (l = 0; l < k; l++) {
6801                                         NFSM_DISSECT(tl, uint32_t *,
6802                                             NFSX_V4DEVICEID + NFSX_STATEID +
6803                                             2 * NFSX_UNSIGNED);
6804                                         if (l == 0) {
6805                                                 /* Just use the first server. */
6806                                                 NFSBCOPY(tl,
6807                                                     flp->nfsfl_ffm[j].dev,
6808                                                     NFSX_V4DEVICEID);
6809                                                 tl += (NFSX_V4DEVICEID /
6810                                                     NFSX_UNSIGNED);
6811                                                 tl++;
6812                                                 flp->nfsfl_ffm[j].st.seqid =
6813                                                     *tl++;
6814                                                 flp->nfsfl_ffm[j].st.other[0] =
6815                                                     *tl++;
6816                                                 flp->nfsfl_ffm[j].st.other[1] =
6817                                                     *tl++;
6818                                                 flp->nfsfl_ffm[j].st.other[2] =
6819                                                     *tl++;
6820                                                 NFSCL_DEBUG(4, "st.seqid=%u "
6821                                                  "st.o0=0x%x st.o1=0x%x "
6822                                                  "st.o2=0x%x\n",
6823                                                  flp->nfsfl_ffm[j].st.seqid,
6824                                                  flp->nfsfl_ffm[j].st.other[0],
6825                                                  flp->nfsfl_ffm[j].st.other[1],
6826                                                  flp->nfsfl_ffm[j].st.other[2]);
6827                                         } else
6828                                                 tl += ((NFSX_V4DEVICEID +
6829                                                     NFSX_STATEID +
6830                                                     NFSX_UNSIGNED) /
6831                                                     NFSX_UNSIGNED);
6832                                         fhcnt = fxdr_unsigned(int, *tl);
6833                                         NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
6834                                         if (fhcnt < 1 ||
6835                                             fhcnt > NFSDEV_MAXVERS) {
6836                                                 error = NFSERR_BADXDR;
6837                                                 goto nfsmout;
6838                                         }
6839                                         for (m = 0; m < fhcnt; m++) {
6840                                                 NFSM_DISSECT(tl, uint32_t *,
6841                                                     NFSX_UNSIGNED);
6842                                                 nfhlen = fxdr_unsigned(int,
6843                                                     *tl);
6844                                                 NFSCL_DEBUG(4, "nfhlen=%d\n",
6845                                                     nfhlen);
6846                                                 if (nfhlen <= 0 || nfhlen >
6847                                                     NFSX_V4FHMAX) {
6848                                                         error = NFSERR_BADXDR;
6849                                                         goto nfsmout;
6850                                                 }
6851                                                 NFSM_DISSECT(cp, uint8_t *,
6852                                                     NFSM_RNDUP(nfhlen));
6853                                                 if (l == 0) {
6854                                                         flp->nfsfl_ffm[j].fhcnt
6855                                                             = fhcnt;
6856                                                         nfhp = malloc(
6857                                                             sizeof(*nfhp) +
6858                                                             nfhlen - 1, M_NFSFH,
6859                                                             M_WAITOK);
6860                                                         flp->nfsfl_ffm[j].fh[m]
6861                                                             = nfhp;
6862                                                         nfhp->nfh_len = nfhlen;
6863                                                         NFSBCOPY(cp,
6864                                                             nfhp->nfh_fh,
6865                                                             nfhlen);
6866                                                         NFSCL_DEBUG(4,
6867                                                             "got fh\n");
6868                                                 }
6869                                         }
6870                                         /* Now, get the ffsd_user/ffds_group. */
6871                                         error = nfsrv_parseug(nd, 0, &user,
6872                                             &grp, curthread);
6873                                         NFSCL_DEBUG(4, "after parseu=%d\n",
6874                                             error);
6875                                         if (error == 0)
6876                                                 error = nfsrv_parseug(nd, 1,
6877                                                     &user, &grp, curthread);
6878                                         NFSCL_DEBUG(4, "aft parseg=%d\n",
6879                                             grp);
6880                                         if (error != 0)
6881                                                 goto nfsmout;
6882                                         NFSCL_DEBUG(4, "user=%d group=%d\n",
6883                                             user, grp);
6884                                         if (l == 0) {
6885                                                 flp->nfsfl_ffm[j].user = user;
6886                                                 flp->nfsfl_ffm[j].group = grp;
6887                                                 NFSCL_DEBUG(4,
6888                                                     "usr=%d grp=%d\n", user,
6889                                                     grp);
6890                                         }
6891                                 }
6892                         }
6893                         NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6894                         flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
6895                         flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
6896                         NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
6897                             flp->nfsfl_fflags, flp->nfsfl_statshint);
6898                 } else {
6899                         error = NFSERR_BADXDR;
6900                         goto nfsmout;
6901                 }
6902                 if (flp->nfsfl_iomode == gotiomode) {
6903                         /* Keep the list in increasing offset order. */
6904                         tflp = LIST_FIRST(flhp);
6905                         prevflp = NULL;
6906                         while (tflp != NULL &&
6907                             tflp->nfsfl_off < flp->nfsfl_off) {
6908                                 prevflp = tflp;
6909                                 tflp = LIST_NEXT(tflp, nfsfl_list);
6910                         }
6911                         if (prevflp == NULL)
6912                                 LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
6913                         else
6914                                 LIST_INSERT_AFTER(prevflp, flp,
6915                                     nfsfl_list);
6916                         NFSCL_DEBUG(4, "flp inserted\n");
6917                 } else {
6918                         printf("nfscl_layoutget(): got wrong iomode\n");
6919                         nfscl_freeflayout(flp);
6920                 }
6921                 flp = NULL;
6922         }
6923 nfsmout:
6924         NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
6925         if (error != 0 && flp != NULL)
6926                 nfscl_freeflayout(flp);
6927         return (error);
6928 }
6929
6930 /*
6931  * Parse a user/group digit string.
6932  */
6933 static int
6934 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
6935     NFSPROC_T *p)
6936 {
6937         uint32_t *tl;
6938         char *cp, *str, str0[NFSV4_SMALLSTR + 1];
6939         uint32_t len = 0;
6940         int error = 0;
6941
6942         NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6943         len = fxdr_unsigned(uint32_t, *tl);
6944         if (len > NFSV4_OPAQUELIMIT) {
6945                 error = NFSERR_BADXDR;
6946                 goto nfsmout;
6947         }
6948         NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
6949         if (len == 0) {
6950                 if (dogrp != 0)
6951                         *gidp = GID_NOGROUP;
6952                 else
6953                         *uidp = UID_NOBODY;
6954                 return (0);
6955         }
6956         if (len > NFSV4_SMALLSTR)
6957                 str = malloc(len + 1, M_TEMP, M_WAITOK);
6958         else
6959                 str = str0;
6960         NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
6961         NFSBCOPY(cp, str, len);
6962         str[len] = '\0';
6963         NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
6964         if (dogrp != 0)
6965                 error = nfsv4_strtogid(nd, str, len, gidp, p);
6966         else
6967                 error = nfsv4_strtouid(nd, str, len, uidp, p);
6968 nfsmout:
6969         if (len > NFSV4_SMALLSTR)
6970                 free(str, M_TEMP);
6971         NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
6972         return (error);
6973 }
6974
6975 /*
6976  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
6977  * so that it does both an Open and a Layoutget.
6978  */
6979 static int
6980 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
6981     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
6982     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
6983     struct ucred *cred, NFSPROC_T *p)
6984 {
6985         struct nfscllayout *lyp;
6986         struct nfsclflayout *flp;
6987         struct nfsclflayouthead flh;
6988         int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
6989         int layouttype, laystat;
6990         nfsv4stateid_t stateid;
6991         struct nfsclsession *tsep;
6992
6993         error = 0;
6994         if (NFSHASFLEXFILE(nmp))
6995                 layouttype = NFSLAYOUT_FLEXFILE;
6996         else
6997                 layouttype = NFSLAYOUT_NFSV4_1_FILES;
6998         /*
6999          * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7000          * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7001          * flp == NULL.
7002          */
7003         lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7004             &recalled);
7005         NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7006         if (lyp == NULL)
7007                 islocked = 0;
7008         else if (flp != NULL)
7009                 islocked = 1;
7010         else
7011                 islocked = 2;
7012         if ((lyp == NULL || flp == NULL) && recalled == 0) {
7013                 LIST_INIT(&flh);
7014                 tsep = nfsmnt_mdssession(nmp);
7015                 layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7016                     3 * NFSX_UNSIGNED);
7017                 if (lyp == NULL)
7018                         usecurstateid = 1;
7019                 else {
7020                         usecurstateid = 0;
7021                         stateid.seqid = lyp->nfsly_stateid.seqid;
7022                         stateid.other[0] = lyp->nfsly_stateid.other[0];
7023                         stateid.other[1] = lyp->nfsly_stateid.other[1];
7024                         stateid.other[2] = lyp->nfsly_stateid.other[2];
7025                 }
7026                 error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7027                     newfhp, newfhlen, mode, op, name, namelen,
7028                     dpp, &stateid, usecurstateid, layouttype, layoutlen,
7029                     &retonclose, &flh, &laystat, cred, p);
7030                 NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7031                     laystat, error);
7032                 laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7033                     &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7034                     &islocked, cred, p);
7035         } else
7036                 error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7037                     mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7038         if (islocked == 2)
7039                 nfscl_rellayout(lyp, 1);
7040         else if (islocked == 1)
7041                 nfscl_rellayout(lyp, 0);
7042         return (error);
7043 }
7044
7045 /*
7046  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7047  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7048  * handled by nfsrpc_openrpc().
7049  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7050  * can be NULL.
7051  */
7052 static int
7053 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7054     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7055     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7056     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7057     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7058     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7059 {
7060         uint32_t *tl;
7061         struct nfsrv_descript nfsd, *nd = &nfsd;
7062         struct nfscldeleg *ndp = NULL;
7063         struct nfsvattr nfsva;
7064         struct nfsclsession *tsep;
7065         uint32_t rflags, deleg;
7066         nfsattrbit_t attrbits;
7067         int error, ret, acesize, limitby, iomode;
7068
7069         *dpp = NULL;
7070         *laystatp = ENXIO;
7071         nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7072             0, 0);
7073         NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7074         *tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7075         *tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7076         *tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7077         tsep = nfsmnt_mdssession(nmp);
7078         *tl++ = tsep->nfsess_clientid.lval[0];
7079         *tl = tsep->nfsess_clientid.lval[1];
7080         nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7081         NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7082         *tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7083         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7084         nfsm_strtom(nd, name, namelen);
7085         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7086         *tl = txdr_unsigned(NFSV4OP_GETATTR);
7087         NFSZERO_ATTRBIT(&attrbits);
7088         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7089         NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7090         nfsrv_putattrbit(nd, &attrbits);
7091         NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7092         *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7093         if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7094                 iomode = NFSLAYOUTIOMODE_RW;
7095         else
7096                 iomode = NFSLAYOUTIOMODE_READ;
7097         nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7098             layouttype, layoutlen, usecurstateid);
7099         error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7100             NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7101         if (error != 0)
7102                 return (error);
7103         NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7104         if (nd->nd_repstat != 0)
7105                 *laystatp = nd->nd_repstat;
7106         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7107                 /* ND_NOMOREDATA will be set if the Open operation failed. */
7108                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7109                     6 * NFSX_UNSIGNED);
7110                 op->nfso_stateid.seqid = *tl++;
7111                 op->nfso_stateid.other[0] = *tl++;
7112                 op->nfso_stateid.other[1] = *tl++;
7113                 op->nfso_stateid.other[2] = *tl;
7114                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7115                 error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7116                 if (error != 0)
7117                         goto nfsmout;
7118                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7119                 deleg = fxdr_unsigned(u_int32_t, *tl);
7120                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7121                     deleg == NFSV4OPEN_DELEGATEWRITE) {
7122                         if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7123                               NFSCLFLAGS_FIRSTDELEG))
7124                                 op->nfso_own->nfsow_clp->nfsc_flags |=
7125                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7126                         ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7127                             M_NFSCLDELEG, M_WAITOK);
7128                         LIST_INIT(&ndp->nfsdl_owner);
7129                         LIST_INIT(&ndp->nfsdl_lock);
7130                         ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7131                         ndp->nfsdl_fhlen = newfhlen;
7132                         NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7133                         newnfs_copyincred(cred, &ndp->nfsdl_cred);
7134                         nfscl_lockinit(&ndp->nfsdl_rwlock);
7135                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7136                             NFSX_UNSIGNED);
7137                         ndp->nfsdl_stateid.seqid = *tl++;
7138                         ndp->nfsdl_stateid.other[0] = *tl++;
7139                         ndp->nfsdl_stateid.other[1] = *tl++;
7140                         ndp->nfsdl_stateid.other[2] = *tl++;
7141                         ret = fxdr_unsigned(int, *tl);
7142                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7143                                 ndp->nfsdl_flags = NFSCLDL_WRITE;
7144                                 /*
7145                                  * Indicates how much the file can grow.
7146                                  */
7147                                 NFSM_DISSECT(tl, u_int32_t *,
7148                                     3 * NFSX_UNSIGNED);
7149                                 limitby = fxdr_unsigned(int, *tl++);
7150                                 switch (limitby) {
7151                                 case NFSV4OPEN_LIMITSIZE:
7152                                         ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7153                                         break;
7154                                 case NFSV4OPEN_LIMITBLOCKS:
7155                                         ndp->nfsdl_sizelimit =
7156                                             fxdr_unsigned(u_int64_t, *tl++);
7157                                         ndp->nfsdl_sizelimit *=
7158                                             fxdr_unsigned(u_int64_t, *tl);
7159                                         break;
7160                                 default:
7161                                         error = NFSERR_BADXDR;
7162                                         goto nfsmout;
7163                                 };
7164                         } else
7165                                 ndp->nfsdl_flags = NFSCLDL_READ;
7166                         if (ret != 0)
7167                                 ndp->nfsdl_flags |= NFSCLDL_RECALL;
7168                         error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7169                             &acesize, p);
7170                         if (error != 0)
7171                                 goto nfsmout;
7172                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7173                         error = NFSERR_BADXDR;
7174                         goto nfsmout;
7175                 }
7176                 if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7177                     nfscl_assumeposixlocks)
7178                         op->nfso_posixlock = 1;
7179                 else
7180                         op->nfso_posixlock = 0;
7181                 NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7182                 /* If the 2nd element == NFS_OK, the Getattr succeeded. */
7183                 if (*++tl == 0) {
7184                         error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7185                             NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7186                             NULL, NULL, NULL, p, cred);
7187                         if (error != 0)
7188                                 goto nfsmout;
7189                         if (ndp != NULL) {
7190                                 ndp->nfsdl_change = nfsva.na_filerev;
7191                                 ndp->nfsdl_modtime = nfsva.na_mtime;
7192                                 ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7193                                 *dpp = ndp;
7194                                 ndp = NULL;
7195                         }
7196                         /*
7197                          * At this point, the Open has succeeded, so set
7198                          * nd_repstat = NFS_OK.  If the Layoutget failed,
7199                          * this function just won't return a layout.
7200                          */
7201                         if (nd->nd_repstat == 0) {
7202                                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7203                                 *laystatp = fxdr_unsigned(int, *++tl);
7204                                 if (*laystatp == 0) {
7205                                         error = nfsrv_parselayoutget(nd,
7206                                             stateidp, retonclosep, flhp);
7207                                         if (error != 0)
7208                                                 *laystatp = error;
7209                                 }
7210                         } else
7211                                 nd->nd_repstat = 0;     /* Return 0 for Open. */
7212                 }
7213         }
7214         if (nd->nd_repstat != 0 && error == 0)
7215                 error = nd->nd_repstat;
7216 nfsmout:
7217         free(ndp, M_NFSCLDELEG);
7218         mbuf_freem(nd->nd_mrep);
7219         return (error);
7220 }
7221
7222 /*
7223  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7224  * Used only for mounts with pNFS enabled.
7225  */
7226 static int
7227 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7228     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7229     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7230     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7231     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7232     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7233     struct nfsclflayouthead *flhp, int *laystatp)
7234 {
7235         uint32_t *tl;
7236         int error = 0, deleg, newone, ret, acesize, limitby;
7237         struct nfsrv_descript nfsd, *nd = &nfsd;
7238         struct nfsclopen *op;
7239         struct nfscldeleg *dp = NULL;
7240         struct nfsnode *np;
7241         struct nfsfh *nfhp;
7242         struct nfsclsession *tsep;
7243         nfsattrbit_t attrbits;
7244         nfsv4stateid_t stateid;
7245         uint32_t rflags;
7246         struct nfsmount *nmp;
7247
7248         nmp = VFSTONFS(dvp->v_mount);
7249         np = VTONFS(dvp);
7250         *laystatp = ENXIO;
7251         *unlockedp = 0;
7252         *nfhpp = NULL;
7253         *dpp = NULL;
7254         *attrflagp = 0;
7255         *dattrflagp = 0;
7256         if (namelen > NFS_MAXNAMLEN)
7257                 return (ENAMETOOLONG);
7258         NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7259         /*
7260          * For V4, this is actually an Open op.
7261          */
7262         NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7263         *tl++ = txdr_unsigned(owp->nfsow_seqid);
7264         *tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7265             NFSV4OPEN_ACCESSREAD);
7266         *tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7267         tsep = nfsmnt_mdssession(nmp);
7268         *tl++ = tsep->nfsess_clientid.lval[0];
7269         *tl = tsep->nfsess_clientid.lval[1];
7270         nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7271         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7272         *tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7273         if ((fmode & O_EXCL) != 0) {
7274                 if (NFSHASSESSPERSIST(nmp)) {
7275                         /* Use GUARDED for persistent sessions. */
7276                         *tl = txdr_unsigned(NFSCREATE_GUARDED);
7277                         nfscl_fillsattr(nd, vap, dvp, 0, 0);
7278                 } else {
7279                         /* Otherwise, use EXCLUSIVE4_1. */
7280                         *tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7281                         NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7282                         *tl++ = cverf.lval[0];
7283                         *tl = cverf.lval[1];
7284                         nfscl_fillsattr(nd, vap, dvp, 0, 0);
7285                 }
7286         } else {
7287                 *tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7288                 nfscl_fillsattr(nd, vap, dvp, 0, 0);
7289         }
7290         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7291         *tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7292         nfsm_strtom(nd, name, namelen);
7293         /* Get the new file's handle and attributes, plus save the FH. */
7294         NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7295         *tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7296         *tl++ = txdr_unsigned(NFSV4OP_GETFH);
7297         *tl = txdr_unsigned(NFSV4OP_GETATTR);
7298         NFSGETATTR_ATTRBIT(&attrbits);
7299         nfsrv_putattrbit(nd, &attrbits);
7300         /* Get the directory's post-op attributes. */
7301         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7302         *tl = txdr_unsigned(NFSV4OP_PUTFH);
7303         nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7304         NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7305         *tl = txdr_unsigned(NFSV4OP_GETATTR);
7306         nfsrv_putattrbit(nd, &attrbits);
7307         NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7308         *tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7309         *tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7310         nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7311             layouttype, layoutlen, usecurstateid);
7312         error = nfscl_request(nd, dvp, p, cred, dstuff);
7313         if (error != 0)
7314                 return (error);
7315         NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7316             error);
7317         if (nd->nd_repstat != 0)
7318                 *laystatp = nd->nd_repstat;
7319         NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7320         if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7321                 NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7322                 NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7323                     6 * NFSX_UNSIGNED);
7324                 stateid.seqid = *tl++;
7325                 stateid.other[0] = *tl++;
7326                 stateid.other[1] = *tl++;
7327                 stateid.other[2] = *tl;
7328                 rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7329                 nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7330                 NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7331                 deleg = fxdr_unsigned(int, *tl);
7332                 if (deleg == NFSV4OPEN_DELEGATEREAD ||
7333                     deleg == NFSV4OPEN_DELEGATEWRITE) {
7334                         if (!(owp->nfsow_clp->nfsc_flags &
7335                               NFSCLFLAGS_FIRSTDELEG))
7336                                 owp->nfsow_clp->nfsc_flags |=
7337                                   (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7338                         dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7339                             M_NFSCLDELEG, M_WAITOK);
7340                         LIST_INIT(&dp->nfsdl_owner);
7341                         LIST_INIT(&dp->nfsdl_lock);
7342                         dp->nfsdl_clp = owp->nfsow_clp;
7343                         newnfs_copyincred(cred, &dp->nfsdl_cred);
7344                         nfscl_lockinit(&dp->nfsdl_rwlock);
7345                         NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7346                             NFSX_UNSIGNED);
7347                         dp->nfsdl_stateid.seqid = *tl++;
7348                         dp->nfsdl_stateid.other[0] = *tl++;
7349                         dp->nfsdl_stateid.other[1] = *tl++;
7350                         dp->nfsdl_stateid.other[2] = *tl++;
7351                         ret = fxdr_unsigned(int, *tl);
7352                         if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7353                                 dp->nfsdl_flags = NFSCLDL_WRITE;
7354                                 /*
7355                                  * Indicates how much the file can grow.
7356                                  */
7357                                 NFSM_DISSECT(tl, u_int32_t *,
7358                                     3 * NFSX_UNSIGNED);
7359                                 limitby = fxdr_unsigned(int, *tl++);
7360                                 switch (limitby) {
7361                                 case NFSV4OPEN_LIMITSIZE:
7362                                         dp->nfsdl_sizelimit = fxdr_hyper(tl);
7363                                         break;
7364                                 case NFSV4OPEN_LIMITBLOCKS:
7365                                         dp->nfsdl_sizelimit =
7366                                             fxdr_unsigned(u_int64_t, *tl++);
7367                                         dp->nfsdl_sizelimit *=
7368                                             fxdr_unsigned(u_int64_t, *tl);
7369                                         break;
7370                                 default:
7371                                         error = NFSERR_BADXDR;
7372                                         goto nfsmout;
7373                                 };
7374                         } else {
7375                                 dp->nfsdl_flags = NFSCLDL_READ;
7376                         }
7377                         if (ret != 0)
7378                                 dp->nfsdl_flags |= NFSCLDL_RECALL;
7379                         error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7380                             &acesize, p);
7381                         if (error != 0)
7382                                 goto nfsmout;
7383                 } else if (deleg != NFSV4OPEN_DELEGATENONE) {
7384                         error = NFSERR_BADXDR;
7385                         goto nfsmout;
7386                 }
7387
7388                 /* Now, we should have the status for the SaveFH. */
7389                 NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7390                 if (*++tl == 0) {
7391                         NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7392                         /*
7393                          * Now, process the GetFH and Getattr for the newly
7394                          * created file. nfscl_mtofh() will set
7395                          * ND_NOMOREDATA if these weren't successful.
7396                          */
7397                         error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7398                         NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7399                         if (error != 0)
7400                                 goto nfsmout;
7401                 } else
7402                         nd->nd_flag |= ND_NOMOREDATA;
7403                 /* Now we have the PutFH and Getattr for the directory. */
7404                 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7405                         NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7406                         if (*++tl != 0)
7407                                 nd->nd_flag |= ND_NOMOREDATA;
7408                         else {
7409                                 NFSM_DISSECT(tl, uint32_t *, 2 *
7410                                     NFSX_UNSIGNED);
7411                                 if (*++tl != 0)
7412                                         nd->nd_flag |= ND_NOMOREDATA;
7413                         }
7414                 }
7415                 if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7416                         /* Load the directory attributes. */
7417                         error = nfsm_loadattr(nd, dnap);
7418                         NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7419                         if (error != 0)
7420                                 goto nfsmout;
7421                         *dattrflagp = 1;
7422                         if (dp != NULL && *attrflagp != 0) {
7423                                 dp->nfsdl_change = nnap->na_filerev;
7424                                 dp->nfsdl_modtime = nnap->na_mtime;
7425                                 dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7426                         }
7427                         /*
7428                          * We can now complete the Open state.
7429                          */
7430                         nfhp = *nfhpp;
7431                         if (dp != NULL) {
7432                                 dp->nfsdl_fhlen = nfhp->nfh_len;
7433                                 NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7434                                     nfhp->nfh_len);
7435                         }
7436                         /*
7437                          * Get an Open structure that will be
7438                          * attached to the OpenOwner, acquired already.
7439                          */
7440                         error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7441                             (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7442                             cred, p, NULL, &op, &newone, NULL, 0);
7443                         if (error != 0)
7444                                 goto nfsmout;
7445                         op->nfso_stateid = stateid;
7446                         newnfs_copyincred(cred, &op->nfso_cred);
7447
7448                         nfscl_openrelease(nmp, op, error, newone);
7449                         *unlockedp = 1;
7450
7451                         /* Now, handle the RestoreFH and LayoutGet. */
7452                         if (nd->nd_repstat == 0) {
7453                                 NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7454                                 *laystatp = fxdr_unsigned(int, *(tl + 3));
7455                                 if (*laystatp == 0) {
7456                                         error = nfsrv_parselayoutget(nd,
7457                                             stateidp, retonclosep, flhp);
7458                                         if (error != 0)
7459                                                 *laystatp = error;
7460                                 }
7461                                 NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7462                                     error);
7463                         } else
7464                                 nd->nd_repstat = 0;
7465                 }
7466         }
7467         if (nd->nd_repstat != 0 && error == 0)
7468                 error = nd->nd_repstat;
7469         if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7470                 nfscl_initiate_recovery(owp->nfsow_clp);
7471 nfsmout:
7472         NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7473         if (error == 0)
7474                 *dpp = dp;
7475         else
7476                 free(dp, M_NFSCLDELEG);
7477         mbuf_freem(nd->nd_mrep);
7478         return (error);
7479 }
7480
7481 /*
7482  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7483  */
7484 static int
7485 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7486     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7487     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7488     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7489     int *dattrflagp, void *dstuff, int *unlockedp)
7490 {
7491         struct nfscllayout *lyp;
7492         struct nfsclflayouthead flh;
7493         struct nfsfh *nfhp;
7494         struct nfsclsession *tsep;
7495         struct nfsmount *nmp;
7496         nfsv4stateid_t stateid;
7497         int error, layoutlen, layouttype, retonclose, laystat;
7498
7499         error = 0;
7500         nmp = VFSTONFS(dvp->v_mount);
7501         if (NFSHASFLEXFILE(nmp))
7502                 layouttype = NFSLAYOUT_FLEXFILE;
7503         else
7504                 layouttype = NFSLAYOUT_NFSV4_1_FILES;
7505         LIST_INIT(&flh);
7506         tsep = nfsmnt_mdssession(nmp);
7507         layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7508         error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7509             owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7510             dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7511             &flh, &laystat);
7512         NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7513             laystat, error);
7514         lyp = NULL;
7515         if (laystat == 0) {
7516                 nfhp = *nfhpp;
7517                 laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
7518                     nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
7519                     layouttype, laystat, NULL, cred, p);
7520         } else
7521                 laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
7522                     retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
7523                     cred, p);
7524         if (laystat == 0)
7525                 nfscl_rellayout(lyp, 0);
7526         return (error);
7527 }
7528
7529 /*
7530  * Process the results of a layoutget() operation.
7531  */
7532 static int
7533 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
7534     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
7535     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
7536     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
7537 {
7538         struct nfsclflayout *tflp;
7539         struct nfscldevinfo *dip;
7540         uint8_t *dev;
7541
7542         if (laystat == NFSERR_UNKNLAYOUTTYPE) {
7543                 NFSLOCKMNT(nmp);
7544                 if (!NFSHASFLEXFILE(nmp)) {
7545                         /* Switch to using Flex File Layout. */
7546                         nmp->nm_state |= NFSSTA_FLEXFILE;
7547                 } else if (layouttype == NFSLAYOUT_FLEXFILE) {
7548                         /* Disable pNFS. */
7549                         NFSCL_DEBUG(1, "disable PNFS\n");
7550                         nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
7551                 }
7552                 NFSUNLOCKMNT(nmp);
7553         }
7554         if (laystat == 0) {
7555                 NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
7556                 LIST_FOREACH(tflp, flhp, nfsfl_list) {
7557                         laystat = nfscl_adddevinfo(nmp, NULL, tflp);
7558                         NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
7559                         if (laystat != 0) {
7560                                 if (layouttype == NFSLAYOUT_FLEXFILE)
7561                                         dev = tflp->nfsfl_ffm[0].dev;
7562                                 else
7563                                         dev = tflp->nfsfl_dev;
7564                                 laystat = nfsrpc_getdeviceinfo(nmp, dev,
7565                                     layouttype, notifybit, &dip, cred, p);
7566                                 NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
7567                                     laystat);
7568                                 if (laystat != 0)
7569                                         break;
7570                                 laystat = nfscl_adddevinfo(nmp, dip, tflp);
7571                                 if (laystat != 0)
7572                                         printf("getlayout: cannot add\n");
7573                         }
7574                 }
7575         }
7576         if (laystat == 0) {
7577                 /*
7578                  * nfscl_layout() always returns with the nfsly_lock
7579                  * set to a refcnt (shared lock).
7580                  * Passing in dvp is sufficient, since it is only used to
7581                  * get the fsid for the file system.
7582                  */
7583                 laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
7584                     layouttype, retonclose, flhp, lypp, cred, p);
7585                 NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
7586                     laystat);
7587                 if (laystat == 0 && islockedp != NULL)
7588                         *islockedp = 1;
7589         }
7590         return (laystat);
7591 }
7592