nfs_clrpcops.c

/*-
 * SPDX-License-Identifier: BSD-3-Clause
 *
 * Copyright (c) 1989, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * Rick Macklem at The University of Guelph.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

/*
 * Rpc op calls, generally called from the vnode op calls or through the
 * buffer cache, for NFS v2, 3 and 4.
 * These do not normally make any changes to vnode arguments or use
 * structures that might change between the VFS variants. The returned
 * arguments are all at the end, after the NFSPROC_T *p one.
 */

#include "opt_inet6.h"

#include <fs/nfs/nfsport.h>
#include <fs/nfsclient/nfs.h>
#include <sys/extattr.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>

SYSCTL_DECL(_vfs_nfs);

static int	nfsignore_eexist = 0;
SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
    &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");

static int	nfscl_dssameconn = 0;
SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
    &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");

static uint64_t nfs_maxcopyrange = SSIZE_MAX;
SYSCTL_U64(_vfs_nfs, OID_AUTO, maxcopyrange, CTLFLAG_RW,
    &nfs_maxcopyrange, 0, "Max size of a Copy so RPC times reasonable");

/*
 * Global variables
 */
extern struct nfsstatsv1 nfsstatsv1;
extern int nfs_numnfscbd;
extern struct timeval nfsboottime;
extern u_int32_t newnfs_false, newnfs_true;
extern nfstype nfsv34_type[9];
extern int nfsrv_useacl;
extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
extern int nfscl_debuglevel;
extern int nfs_pnfsiothreads;
extern u_long sb_max_adj;
NFSCLSTATEMUTEX;
int nfstest_outofseq = 0;
int nfscl_assumeposixlocks = 1;
int nfscl_enablecallb = 0;
short nfsv4_cbport = NFSV4_CBPORT;
int nfstest_openallsetattr = 0;

#define	DIRHDSIZ	offsetof(struct dirent, d_name)

/*
 * nfscl_getsameserver() can return one of three values:
 * NFSDSP_USETHISSESSION - Use this session for the DS.
 * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
 *     session.
 * NFSDSP_NOTFOUND - No matching server was found.
 */
enum nfsclds_state {
	NFSDSP_USETHISSESSION = 0,
	NFSDSP_SEQTHISSESSION = 1,
	NFSDSP_NOTFOUND = 2,
};

/*
 * Do a write RPC on a DS data file, using this structure for the arguments,
 * so that this function can be executed by a separate kernel process.
 */
struct nfsclwritedsdorpc {
	int			done;
	int			inprog;
	struct task		tsk;
	struct vnode		*vp;
	int			iomode;
	int			must_commit;
	nfsv4stateid_t		*stateidp;
	struct nfsclds		*dsp;
	uint64_t		off;
	int			len;
#ifdef notyet
	int			advise;
#endif
	struct nfsfh		*fhp;
	struct mbuf		*m;
	int			vers;
	int			minorvers;
	struct ucred		*cred;
	NFSPROC_T		*p;
	int			err;
};

static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
    struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
    nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
    struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
    void *);
static int nfsrpc_deallocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
    struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
    nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
    struct nfsvattr *, struct nfsfh **, int *, int *, void *);
static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
    nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
    NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
    int *, void *, int *);
static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
    struct nfscllockowner *, u_int64_t, u_int64_t,
    u_int32_t, struct ucred *, NFSPROC_T *, int);
static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
    struct acl *, nfsv4stateid_t *, void *);
static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
    uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
    struct ucred *, NFSPROC_T *);
static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
    struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
    NFSPROC_T *);
static void nfscl_initsessionslots(struct nfsclsession *);
static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
    nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
    struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
    NFSPROC_T *);
static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
    nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
    struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
    struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
    struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
    struct ucred *, NFSPROC_T *);
static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
    nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
    struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
    struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
    struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
    struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
    struct ucred *, NFSPROC_T *);
static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
    struct nfsclds *, struct nfsclds **, uint32_t *);
static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
    struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
    NFSPROC_T *);
static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
    struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
#ifdef notyet
static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
    struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
    NFSPROC_T *);
static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
    struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
#endif
static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
    struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
    uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
    NFSPROC_T *);
static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
    nfsv4stateid_t *, int *, struct nfsclflayouthead *);
static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
    int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
    struct nfscldeleg **, struct ucred *, NFSPROC_T *);
static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
    nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
    struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
    struct nfsfh **, int *, int *, void *, int *);
static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
    int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
    struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
    struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
    nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
    struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
    struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
    int, int, int, int *, struct nfsclflayouthead *, int *);
static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
    uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
    struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
    int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
    struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
    nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
    struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
    int, struct nfsvattr *, int *, struct ucred *);
static struct mbuf *nfsm_split(struct mbuf *, uint64_t);

int nfs_pnfsio(task_fn_t *, void *);

/*
 * nfs null call from vfs.
 */
int
nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
{
	int error;
	struct nfsrv_descript nfsd, *nd = &nfsd;

	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
	error = nfscl_request(nd, vp, p, cred, NULL);
	if (nd->nd_repstat && !error)
		error = nd->nd_repstat;
	m_freem(nd->nd_mrep);
	return (error);
}

/*
 * nfs access rpc op.
 * For nfs version 3 and 4, use the access rpc to check accessibility. If file
 * modes are changed on the server, accesses might still fail later.
 */
int
nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
    NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
{
	int error;
	u_int32_t mode, rmode;

	if (acmode & VREAD)
		mode = NFSACCESS_READ;
	else
		mode = 0;
	if (vnode_vtype(vp) == VDIR) {
		if (acmode & VWRITE)
			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
				 NFSACCESS_DELETE);
		if (acmode & VEXEC)
			mode |= NFSACCESS_LOOKUP;
	} else {
		if (acmode & VWRITE)
			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
		if (acmode & VEXEC)
			mode |= NFSACCESS_EXECUTE;
	}

	/*
	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
	 */
	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
	    NULL);

	/*
	 * The NFS V3 spec does not clarify whether or not
	 * the returned access bits can be a superset of
	 * the ones requested, so...
	 */
	if (!error && (rmode & mode) != mode)
		error = EACCES;
	return (error);
}

/*
 * The actual rpc, separated out for Darwin.
 */
int
nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
    NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
    void *stuff)
{
	u_int32_t *tl;
	u_int32_t supported, rmode;
	int error;
	struct nfsrv_descript nfsd, *nd = &nfsd;
	nfsattrbit_t attrbits;

	*attrflagp = 0;
	supported = mode;
	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
	*tl = txdr_unsigned(mode);
	if (nd->nd_flag & ND_NFSV4) {
		/*
		 * And do a Getattr op.
		 */
		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
		*tl = txdr_unsigned(NFSV4OP_GETATTR);
		NFSGETATTR_ATTRBIT(&attrbits);
		(void) nfsrv_putattrbit(nd, &attrbits);
	}
	error = nfscl_request(nd, vp, p, cred, stuff);
	if (error)
		return (error);
	if (nd->nd_flag & ND_NFSV3) {
		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
		if (error)
			goto nfsmout;
	}
	if (!nd->nd_repstat) {
		if (nd->nd_flag & ND_NFSV4) {
			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
			supported = fxdr_unsigned(u_int32_t, *tl++);
		} else {
			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
		}
		rmode = fxdr_unsigned(u_int32_t, *tl);
		if (nd->nd_flag & ND_NFSV4)
			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);

		/*
		 * It's not obvious what should be done about
		 * unsupported access modes. For now, be paranoid
		 * and clear the unsupported ones.
		 */
		rmode &= supported;
		*rmodep = rmode;
	} else
		error = nd->nd_repstat;
nfsmout:
	m_freem(nd->nd_mrep);
	return (error);
}

/*
 * nfs open rpc
 */
int
nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
{
	struct nfsclopen *op;
	struct nfscldeleg *dp;
	struct nfsfh *nfhp;
	struct nfsnode *np = VTONFS(vp);
	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
	u_int32_t mode, clidrev;
	int ret, newone, error, expireret = 0, retrycnt;

	/*
	 * For NFSv4, Open Ops are only done on Regular Files.
	 */
	if (vnode_vtype(vp) != VREG)
		return (0);
	mode = 0;
	if (amode & FREAD)
		mode |= NFSV4OPEN_ACCESSREAD;
	if (amode & FWRITE)
		mode |= NFSV4OPEN_ACCESSWRITE;
	nfhp = np->n_fhp;

	retrycnt = 0;
#ifdef notdef
{ char name[100]; int namel;
namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
bcopy(NFS4NODENAME(np->n_v4), name, namel);
name[namel] = '\0';
printf("rpcopen p=0x%x name=%s",p->p_pid,name);
if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
else printf(" fhl=0\n");
}
#endif
	do {
	    dp = NULL;
	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
		cred, p, NULL, &op, &newone, &ret, 1, true);
	    if (error) {
		return (error);
	    }
	    if (nmp->nm_clp != NULL)
		clidrev = nmp->nm_clp->nfsc_clientidrev;
	    else
		clidrev = 0;
	    if (ret == NFSCLOPEN_DOOPEN) {
		if (np->n_v4 != NULL) {
			/*
			 * For the first attempt, try and get a layout, if
			 * pNFS is enabled for the mount.
			 */
			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
			    nfs_numnfscbd == 0 ||
			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
				error = nfsrpc_openrpc(nmp, vp,
				    np->n_v4->n4_data,
				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
				    np->n_fhp->nfh_len, mode, op,
				    NFS4NODENAME(np->n_v4),
				    np->n_v4->n4_namelen,
				    &dp, 0, 0x0, cred, p, 0, 0);
			else
				error = nfsrpc_getopenlayout(nmp, vp,
				    np->n_v4->n4_data,
				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
				    np->n_fhp->nfh_len, mode, op,
				    NFS4NODENAME(np->n_v4),
				    np->n_v4->n4_namelen, &dp, cred, p);
			if (dp != NULL) {
#ifdef APPLE
				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
#else
				NFSLOCKNODE(np);
				np->n_flag &= ~NDELEGMOD;
				/*
				 * Invalidate the attribute cache, so that
				 * attributes that pre-date the issue of a
				 * delegation are not cached, since the
				 * cached attributes will remain valid while
				 * the delegation is held.
				 */
				NFSINVALATTRCACHE(np);
				NFSUNLOCKNODE(np);
#endif
				(void) nfscl_deleg(nmp->nm_mountp,
				    op->nfso_own->nfsow_clp,
				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
			}
		} else {
			error = EIO;
		}
		newnfs_copyincred(cred, &op->nfso_cred);
	    } else if (ret == NFSCLOPEN_SETCRED)
		/*
		 * This is a new local open on a delegation. It needs
		 * to have credentials so that an open can be done
		 * against the server during recovery.
		 */
		newnfs_copyincred(cred, &op->nfso_cred);

	    /*
	     * nfso_opencnt is the count of how many VOP_OPEN()s have
	     * been done on this Open successfully and a VOP_CLOSE()
	     * is expected for each of these.
	     * If error is non-zero, don't increment it, since the Open
	     * hasn't succeeded yet.
	     */
	    if (!error) {
		op->nfso_opencnt++;
		if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
		    NFSLOCKNODE(np);
		    np->n_openstateid = op;
		    NFSUNLOCKNODE(np);
		}
	    }
	    nfscl_openrelease(nmp, op, error, newone);
	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
		error == NFSERR_BADSESSION) {
		(void) nfs_catnap(PZERO, error, "nfs_open");
	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
		&& clidrev != 0) {
		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
		retrycnt++;
	    }
	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
	    error == NFSERR_BADSESSION ||
	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
	     expireret == 0 && clidrev != 0 && retrycnt < 4));
	if (error && retrycnt >= 4)
		error = EIO;
	return (error);
}

/*
 * the actual open rpc
 */
int
nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
    u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
    u_int8_t *name, int namelen, struct nfscldeleg **dpp,
    int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
    int syscred, int recursed)
{
	u_int32_t *tl;
	struct nfsrv_descript nfsd, *nd = &nfsd;
	struct nfscldeleg *dp, *ndp = NULL;
	struct nfsvattr nfsva;
	u_int32_t rflags, deleg;
	nfsattrbit_t attrbits;
	int error, ret, acesize, limitby;
	struct nfsclsession *tsep;

	dp = *dpp;
	*dpp = NULL;
	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
	tsep = nfsmnt_mdssession(nmp);
	*tl++ = tsep->nfsess_clientid.lval[0];
	*tl = tsep->nfsess_clientid.lval[1];
	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
	if (reclaim) {
		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
		*tl = txdr_unsigned(delegtype);
	} else {
		if (dp != NULL) {
			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
			if (NFSHASNFSV4N(nmp))
				*tl++ = 0;
			else
				*tl++ = dp->nfsdl_stateid.seqid;
			*tl++ = dp->nfsdl_stateid.other[0];
			*tl++ = dp->nfsdl_stateid.other[1];
			*tl = dp->nfsdl_stateid.other[2];
		} else {
			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
		}
		(void) nfsm_strtom(nd, name, namelen);
	}
	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
	*tl = txdr_unsigned(NFSV4OP_GETATTR);
	NFSZERO_ATTRBIT(&attrbits);
	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
	(void) nfsrv_putattrbit(nd, &attrbits);
	if (syscred)
		nd->nd_flag |= ND_USEGSSNAME;
	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
	if (error)
		return (error);
	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
	if (!nd->nd_repstat) {
		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
		    6 * NFSX_UNSIGNED);
		op->nfso_stateid.seqid = *tl++;
		op->nfso_stateid.other[0] = *tl++;
		op->nfso_stateid.other[1] = *tl++;
		op->nfso_stateid.other[2] = *tl;
		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
		if (error)
			goto nfsmout;
		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
		deleg = fxdr_unsigned(u_int32_t, *tl);
		if (deleg == NFSV4OPEN_DELEGATEREAD ||
		    deleg == NFSV4OPEN_DELEGATEWRITE) {
			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
			      NFSCLFLAGS_FIRSTDELEG))
				op->nfso_own->nfsow_clp->nfsc_flags |=
				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
			ndp = malloc(
			    sizeof (struct nfscldeleg) + newfhlen,
			    M_NFSCLDELEG, M_WAITOK);
			LIST_INIT(&ndp->nfsdl_owner);
			LIST_INIT(&ndp->nfsdl_lock);
			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
			ndp->nfsdl_fhlen = newfhlen;
			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
			newnfs_copyincred(cred, &ndp->nfsdl_cred);
			nfscl_lockinit(&ndp->nfsdl_rwlock);
			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
			    NFSX_UNSIGNED);
			ndp->nfsdl_stateid.seqid = *tl++;
			ndp->nfsdl_stateid.other[0] = *tl++;
			ndp->nfsdl_stateid.other[1] = *tl++;
			ndp->nfsdl_stateid.other[2] = *tl++;
			ret = fxdr_unsigned(int, *tl);
			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
				ndp->nfsdl_flags = NFSCLDL_WRITE;
				/*
				 * Indicates how much the file can grow.
				 */
				NFSM_DISSECT(tl, u_int32_t *,
				    3 * NFSX_UNSIGNED);
				limitby = fxdr_unsigned(int, *tl++);
				switch (limitby) {
				case NFSV4OPEN_LIMITSIZE:
					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
					break;
				case NFSV4OPEN_LIMITBLOCKS:
					ndp->nfsdl_sizelimit =
					    fxdr_unsigned(u_int64_t, *tl++);
					ndp->nfsdl_sizelimit *=
					    fxdr_unsigned(u_int64_t, *tl);
					break;
				default:
					error = NFSERR_BADXDR;
					goto nfsmout;
				}
			} else {
				ndp->nfsdl_flags = NFSCLDL_READ;
			}
			if (ret)
				ndp->nfsdl_flags |= NFSCLDL_RECALL;
			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
			    &acesize, p);
			if (error)
				goto nfsmout;
		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
			error = NFSERR_BADXDR;
			goto nfsmout;
		}
		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
		    NULL, NULL, NULL, p, cred);
		if (error)
			goto nfsmout;
		if (ndp != NULL) {
			ndp->nfsdl_change = nfsva.na_filerev;
			ndp->nfsdl_modtime = nfsva.na_mtime;
			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
		}
		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
		    do {
			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
			    cred, p);
			if (ret == NFSERR_DELAY)
			    (void) nfs_catnap(PZERO, ret, "nfs_open");
		    } while (ret == NFSERR_DELAY);
		    error = ret;
		}
		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
		    nfscl_assumeposixlocks)
		    op->nfso_posixlock = 1;
		else
		    op->nfso_posixlock = 0;

		/*
		 * If the server is handing out delegations, but we didn't
		 * get one because an OpenConfirm was required, try the
		 * Open again, to get a delegation. This is a harmless no-op,
		 * from a server's point of view.
		 */
		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
		    && !error && dp == NULL && ndp == NULL && !recursed) {
		    do {
			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
			    cred, p, syscred, 1);
			if (ret == NFSERR_DELAY)
			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
		    } while (ret == NFSERR_DELAY);
		    if (ret) {
			if (ndp != NULL) {
				free(ndp, M_NFSCLDELEG);
				ndp = NULL;
			}
			if (ret == NFSERR_STALECLIENTID ||
			    ret == NFSERR_STALEDONTRECOVER ||
			    ret == NFSERR_BADSESSION)
				error = ret;
		    }
		}
	}
	if (nd->nd_repstat != 0 && error == 0)
		error = nd->nd_repstat;
	if (error == NFSERR_STALECLIENTID)
		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
	if (!error)
		*dpp = ndp;
	else if (ndp != NULL)
		free(ndp, M_NFSCLDELEG);
	m_freem(nd->nd_mrep);
	return (error);
}

/*
 * open downgrade rpc
 */
int
nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
    struct ucred *cred, NFSPROC_T *p)
{
	u_int32_t *tl;
	struct nfsrv_descript nfsd, *nd = &nfsd;
	int error;

	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
		*tl++ = 0;
	else
		*tl++ = op->nfso_stateid.seqid;
	*tl++ = op->nfso_stateid.other[0];
	*tl++ = op->nfso_stateid.other[1];
	*tl++ = op->nfso_stateid.other[2];
	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
	error = nfscl_request(nd, vp, p, cred, NULL);
	if (error)
		return (error);
	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
	if (!nd->nd_repstat) {
		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
		op->nfso_stateid.seqid = *tl++;
		op->nfso_stateid.other[0] = *tl++;
		op->nfso_stateid.other[1] = *tl++;
		op->nfso_stateid.other[2] = *tl;
	}
	if (nd->nd_repstat && error == 0)
		error = nd->nd_repstat;
	if (error == NFSERR_STALESTATEID)
		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
	m_freem(nd->nd_mrep);
	return (error);
}

/*
 * V4 Close operation.
 */
int
nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
{
	struct nfsclclient *clp;
	int error;

	if (vnode_vtype(vp) != VREG)
		return (0);
	if (doclose)
		error = nfscl_doclose(vp, &clp, p);
	else {
		error = nfscl_getclose(vp, &clp);
		if (error == 0)
			nfscl_clientrelease(clp);
	}
	return (error);
}

/*
 * Close the open.
 */
int
nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
    bool loop_on_delayed, bool freeop)
{
	struct nfsrv_descript nfsd, *nd = &nfsd;
	struct nfscllockowner *lp, *nlp;
	struct nfscllock *lop, *nlop;
	struct ucred *tcred;
	u_int64_t off = 0, len = 0;
	u_int32_t type = NFSV4LOCKT_READ;
	int error, do_unlock, trycnt;

	tcred = newnfs_getcred();
	newnfs_copycred(&op->nfso_cred, tcred);
	/*
	 * (Theoretically this could be done in the same
	 *  compound as the close, but having multiple
	 *  sequenced Ops in the same compound might be
	 *  too scary for some servers.)
	 */
	if (op->nfso_posixlock) {
		off = 0;
		len = NFS64BITSSET;
		type = NFSV4LOCKT_READ;
	}

	/*
	 * Since this function is only called from VOP_INACTIVE(), no
	 * other thread will be manipulating this Open. As such, the
	 * lock lists are not being changed by other threads, so it should
	 * be safe to do this without locking.
	 */
	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
		do_unlock = 1;
		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
			if (op->nfso_posixlock == 0) {
				off = lop->nfslo_first;
				len = lop->nfslo_end - lop->nfslo_first;
				if (lop->nfslo_type == F_WRLCK)
					type = NFSV4LOCKT_WRITE;
				else
					type = NFSV4LOCKT_READ;
			}
			if (do_unlock) {
				trycnt = 0;
				do {
					error = nfsrpc_locku(nd, nmp, lp, off,
					    len, type, tcred, p, 0);
					if ((nd->nd_repstat == NFSERR_GRACE ||
					    nd->nd_repstat == NFSERR_DELAY) &&
					    error == 0)
						(void) nfs_catnap(PZERO,
						    (int)nd->nd_repstat,
						    "nfs_close");
				} while ((nd->nd_repstat == NFSERR_GRACE ||
				    nd->nd_repstat == NFSERR_DELAY) &&
				    error == 0 && trycnt++ < 5);
				if (op->nfso_posixlock)
					do_unlock = 0;
			}
			nfscl_freelock(lop, 0);
		}
		/*
		 * Do a ReleaseLockOwner.
		 * The lock owner name nfsl_owner may be used by other opens for
		 * other files but the lock_owner4 name that nfsrpc_rellockown()
		 * puts on the wire has the file handle for this file appended
		 * to it, so it can be done now.
		 */
		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
		    lp->nfsl_open->nfso_fhlen, tcred, p);
	}

	/*
	 * There could be other Opens for different files on the same
	 * OpenOwner, so locking is required.
	 */
	NFSLOCKCLSTATE();
	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
	NFSUNLOCKCLSTATE();
	do {
		error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
		if (error == NFSERR_GRACE)
			(void) nfs_catnap(PZERO, error, "nfs_close");
	} while (error == NFSERR_GRACE);
	NFSLOCKCLSTATE();
	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);

	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
		nfscl_freelockowner(lp, 0);
	if (freeop && error != NFSERR_DELAY)
		nfscl_freeopen(op, 0, true);
	NFSUNLOCKCLSTATE();
	NFSFREECRED(tcred);
	return (error);
}

/*
 * The actual Close RPC.
 */
int
nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
    struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
    int syscred)
{
	u_int32_t *tl;
	int error;

	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
	    op->nfso_fhlen, NULL, NULL, 0, 0);
	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
	if (NFSHASNFSV4N(nmp)) {
		*tl++ = 0;
		*tl++ = 0;
	} else {
		*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
		*tl++ = op->nfso_stateid.seqid;
	}
	*tl++ = op->nfso_stateid.other[0];
	*tl++ = op->nfso_stateid.other[1];
	*tl = op->nfso_stateid.other[2];
	if (syscred)
		nd->nd_flag |= ND_USEGSSNAME;
	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
	if (error)
		return (error);
	if (!NFSHASNFSV4N(nmp))
		NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
	if (nd->nd_repstat == 0)
		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
	error = nd->nd_repstat;
	if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
	m_freem(nd->nd_mrep);
	return (error);
}

/*
 * V4 Open Confirm RPC.
 */
int
nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
    struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
{
	u_int32_t *tl;
	struct nfsrv_descript nfsd, *nd = &nfsd;
	struct nfsmount *nmp;
	int error;

	nmp = VFSTONFS(vp->v_mount);
	if (NFSHASNFSV4N(nmp))
		return (0);		/* No confirmation for NFSv4.1. */
	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
	    0, 0);
	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
	*tl++ = op->nfso_stateid.seqid;
	*tl++ = op->nfso_stateid.other[0];
	*tl++ = op->nfso_stateid.other[1];
	*tl++ = op->nfso_stateid.other[2];
	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
	error = nfscl_request(nd, vp, p, cred, NULL);
	if (error)
		return (error);
	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
	if (!nd->nd_repstat) {
		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
		op->nfso_stateid.seqid = *tl++;
		op->nfso_stateid.other[0] = *tl++;
		op->nfso_stateid.other[1] = *tl++;
		op->nfso_stateid.other[2] = *tl;
	}
	error = nd->nd_repstat;
	if (error == NFSERR_STALESTATEID)
		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
nfsmout:
	m_freem(nd->nd_mrep);
	return (error);
}

/*
 * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
 * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
 */
int
nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
    bool *retokp, struct ucred *cred, NFSPROC_T *p)
{
	u_int32_t *tl;
	struct nfsrv_descript nfsd;
	struct nfsrv_descript *nd = &nfsd;
	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
	u_short port;
	int error, isinet6 = 0, callblen;
	nfsquad_t confirm;
	static u_int32_t rev = 0;
	struct nfsclds *dsp, *odsp;
	struct in6_addr a6;
	struct nfsclsession *tsep;
	struct rpc_reconupcall recon;
	struct nfscl_reconarg *rcp;

	if (nfsboottime.tv_sec == 0)
		NFSSETBOOTTIME(nfsboottime);
	if (NFSHASNFSV4N(nmp)) {
		error = NFSERR_BADSESSION;
		odsp = dsp = NULL;
		if (retokp != NULL) {
			NFSLOCKMNT(nmp);
			odsp = TAILQ_FIRST(&nmp->nm_sess);
			NFSUNLOCKMNT(nmp);
		}
		if (odsp != NULL) {
			/*
			 * When a session already exists, first try a
			 * CreateSession with the extant ClientID.
			 */
			dsp = malloc(sizeof(struct nfsclds) +
			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
			    M_WAITOK | M_ZERO);
			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
			dsp->nfsclds_sess.nfsess_clientid =
			    odsp->nfsclds_sess.nfsess_clientid;
			dsp->nfsclds_sess.nfsess_sequenceid =
			    odsp->nfsclds_sess.nfsess_sequenceid;
			dsp->nfsclds_flags = odsp->nfsclds_flags;
			if (dsp->nfsclds_servownlen > 0)
				memcpy(dsp->nfsclds_serverown,
				    odsp->nfsclds_serverown,
				    dsp->nfsclds_servownlen + 1);
			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
			    NULL, MTX_DEF);
			nfscl_initsessionslots(&dsp->nfsclds_sess);
			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
			    &nmp->nm_sockreq, NULL,