xref: /illumos-gate/usr/src/uts/common/fs/nfs/nfs4_vfsops.c (revision e010bda9)

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright 2015 Nexenta Systems, Inc.  All rights reserved.
 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
 */

/*
 *	Copyright (c) 1983,1984,1985,1986,1987,1988,1989  AT&T.
 *	All Rights Reserved
 */

#include <sys/param.h>
#include <sys/types.h>
#include <sys/systm.h>
#include <sys/cred.h>
#include <sys/vfs.h>
#include <sys/vfs_opreg.h>
#include <sys/vnode.h>
#include <sys/pathname.h>
#include <sys/sysmacros.h>
#include <sys/kmem.h>
#include <sys/mkdev.h>
#include <sys/mount.h>
#include <sys/statvfs.h>
#include <sys/errno.h>
#include <sys/debug.h>
#include <sys/cmn_err.h>
#include <sys/utsname.h>
#include <sys/bootconf.h>
#include <sys/modctl.h>
#include <sys/acl.h>
#include <sys/flock.h>
#include <sys/time.h>
#include <sys/disp.h>
#include <sys/policy.h>
#include <sys/socket.h>
#include <sys/netconfig.h>
#include <sys/dnlc.h>
#include <sys/list.h>
#include <sys/mntent.h>
#include <sys/tsol/label.h>

#include <rpc/types.h>
#include <rpc/auth.h>
#include <rpc/rpcsec_gss.h>
#include <rpc/clnt.h>

#include <nfs/nfs.h>
#include <nfs/nfs_clnt.h>
#include <nfs/mount.h>
#include <nfs/nfs_acl.h>

#include <fs/fs_subr.h>

#include <nfs/nfs4.h>
#include <nfs/rnode4.h>
#include <nfs/nfs4_clnt.h>
#include <sys/fs/autofs.h>

#include <sys/sdt.h>


/*
 * Arguments passed to thread to free data structures from forced unmount.
 */

typedef struct {
	vfs_t	*fm_vfsp;
	int	fm_flag;
	cred_t	*fm_cr;
} freemountargs_t;

static void	async_free_mount(vfs_t *, int, cred_t *);
static void	nfs4_free_mount(vfs_t *, int, cred_t *);
static void	nfs4_free_mount_thread(freemountargs_t *);
static int nfs4_chkdup_servinfo4(servinfo4_t *, servinfo4_t *);

/*
 * From rpcsec module (common/rpcsec).
 */
extern int sec_clnt_loadinfo(struct sec_data *, struct sec_data **, model_t);
extern void sec_clnt_freeinfo(struct sec_data *);

/*
 * The order and contents of this structure must be kept in sync with that of
 * rfsreqcnt_v4_tmpl in nfs_stats.c
 */
static char *rfsnames_v4[] = {
	"null", "compound", "reserved",	"access", "close", "commit", "create",
	"delegpurge", "delegreturn", "getattr",	"getfh", "link", "lock",
	"lockt", "locku", "lookup", "lookupp", "nverify", "open", "openattr",
	"open_confirm",	"open_downgrade", "putfh", "putpubfh", "putrootfh",
	"read", "readdir", "readlink", "remove", "rename", "renew",
	"restorefh", "savefh", "secinfo", "setattr", "setclientid",
	"setclientid_confirm", "verify", "write"
};

/*
 * nfs4_max_mount_retry is the number of times the client will redrive
 * a mount compound before giving up and returning failure.  The intent
 * is to redrive mount compounds which fail NFS4ERR_STALE so that
 * if a component of the server path being mounted goes stale, it can
 * "recover" by redriving the mount compund (LOOKUP ops).  This recovery
 * code is needed outside of the recovery framework because mount is a
 * special case.  The client doesn't create vnodes/rnodes for components
 * of the server path being mounted.  The recovery code recovers real
 * client objects, not STALE FHs which map to components of the server
 * path being mounted.
 *
 * We could just fail the mount on the first time, but that would
 * instantly trigger failover (from nfs4_mount), and the client should
 * try to re-lookup the STALE FH before doing failover.  The easiest
 * way to "re-lookup" is to simply redrive the mount compound.
 */
static int nfs4_max_mount_retry = 2;

/*
 * nfs4 vfs operations.
 */
int		nfs4_mount(vfs_t *, vnode_t *, struct mounta *, cred_t *);
static int	nfs4_unmount(vfs_t *, int, cred_t *);
static int	nfs4_root(vfs_t *, vnode_t **);
static int	nfs4_statvfs(vfs_t *, struct statvfs64 *);
static int	nfs4_sync(vfs_t *, short, cred_t *);
static int	nfs4_vget(vfs_t *, vnode_t **, fid_t *);
static int	nfs4_mountroot(vfs_t *, whymountroot_t);
static void	nfs4_freevfs(vfs_t *);

static int	nfs4rootvp(vnode_t **, vfs_t *, struct servinfo4 *,
		    int, cred_t *, zone_t *);

vfsops_t	*nfs4_vfsops;

int nfs4_vfsinit(void);
void nfs4_vfsfini(void);
static void nfs4setclientid_init(void);
static void nfs4setclientid_fini(void);
static void nfs4setclientid_otw(mntinfo4_t *, servinfo4_t *,  cred_t *,
		struct nfs4_server *, nfs4_error_t *, int *);
static void	destroy_nfs4_server(nfs4_server_t *);
static void	remove_mi(nfs4_server_t *, mntinfo4_t *);

extern void nfs4_ephemeral_init(void);
extern void nfs4_ephemeral_fini(void);

/* referral related routines */
static servinfo4_t *copy_svp(servinfo4_t *);
static void free_knconf_contents(struct knetconfig *k);
static char *extract_referral_point(const char *, int);
static void setup_newsvpath(servinfo4_t *, int);
static void update_servinfo4(servinfo4_t *, fs_location4 *,
		struct nfs_fsl_info *, char *, int);

/*
 * Initialize the vfs structure
 */

static int nfs4fstyp;


/*
 * Debug variable to check for rdma based
 * transport startup and cleanup. Controlled
 * through /etc/system. Off by default.
 */
extern int rdma_debug;

int
nfs4init(int fstyp, char *name)
{
	static const fs_operation_def_t nfs4_vfsops_template[] = {
		VFSNAME_MOUNT,		{ .vfs_mount = nfs4_mount },
		VFSNAME_UNMOUNT,	{ .vfs_unmount = nfs4_unmount },
		VFSNAME_ROOT,		{ .vfs_root = nfs4_root },
		VFSNAME_STATVFS,	{ .vfs_statvfs = nfs4_statvfs },
		VFSNAME_SYNC,		{ .vfs_sync = nfs4_sync },
		VFSNAME_VGET,		{ .vfs_vget = nfs4_vget },
		VFSNAME_MOUNTROOT,	{ .vfs_mountroot = nfs4_mountroot },
		VFSNAME_FREEVFS,	{ .vfs_freevfs = nfs4_freevfs },
		NULL,			NULL
	};
	int error;

	nfs4_vfsops = NULL;
	nfs4_vnodeops = NULL;
	nfs4_trigger_vnodeops = NULL;

	error = vfs_setfsops(fstyp, nfs4_vfsops_template, &nfs4_vfsops);
	if (error != 0) {
		zcmn_err(GLOBAL_ZONEID, CE_WARN,
		    "nfs4init: bad vfs ops template");
		goto out;
	}

	error = vn_make_ops(name, nfs4_vnodeops_template, &nfs4_vnodeops);
	if (error != 0) {
		zcmn_err(GLOBAL_ZONEID, CE_WARN,
		    "nfs4init: bad vnode ops template");
		goto out;
	}

	error = vn_make_ops("nfs4_trigger", nfs4_trigger_vnodeops_template,
	    &nfs4_trigger_vnodeops);
	if (error != 0) {
		zcmn_err(GLOBAL_ZONEID, CE_WARN,
		    "nfs4init: bad trigger vnode ops template");
		goto out;
	}

	nfs4fstyp = fstyp;
	(void) nfs4_vfsinit();
	(void) nfs4_init_dot_entries();

out:
	if (error) {
		if (nfs4_trigger_vnodeops != NULL)
			vn_freevnodeops(nfs4_trigger_vnodeops);

		if (nfs4_vnodeops != NULL)
			vn_freevnodeops(nfs4_vnodeops);

		(void) vfs_freevfsops_by_type(fstyp);
	}

	return (error);
}

void
nfs4fini(void)
{
	(void) nfs4_destroy_dot_entries();
	nfs4_vfsfini();
}

/*
 * Create a new sec_data structure to store AUTH_DH related data:
 * netname, syncaddr, knetconfig. There is no AUTH_F_RPCTIMESYNC
 * flag set for NFS V4 since we are avoiding to contact the rpcbind
 * daemon and is using the IP time service (IPPORT_TIMESERVER).
 *
 * sec_data can be freed by sec_clnt_freeinfo().
 */
static struct sec_data *
create_authdh_data(char *netname, int nlen, struct netbuf *syncaddr,
    struct knetconfig *knconf)
{
	struct sec_data *secdata;
	dh_k4_clntdata_t *data;
	char *pf, *p;

	if (syncaddr == NULL || syncaddr->buf == NULL || nlen == 0)
		return (NULL);

	secdata = kmem_alloc(sizeof (*secdata), KM_SLEEP);
	secdata->flags = 0;

	data = kmem_alloc(sizeof (*data), KM_SLEEP);

	data->syncaddr.maxlen = syncaddr->maxlen;
	data->syncaddr.len = syncaddr->len;
	data->syncaddr.buf = (char *)kmem_alloc(syncaddr->len, KM_SLEEP);
	bcopy(syncaddr->buf, data->syncaddr.buf, syncaddr->len);

	/*
	 * duplicate the knconf information for the
	 * new opaque data.
	 */
	data->knconf = kmem_alloc(sizeof (*knconf), KM_SLEEP);
	*data->knconf = *knconf;
	pf = kmem_alloc(KNC_STRSIZE, KM_SLEEP);
	p = kmem_alloc(KNC_STRSIZE, KM_SLEEP);
	bcopy(knconf->knc_protofmly, pf, KNC_STRSIZE);
	bcopy(knconf->knc_proto, p, KNC_STRSIZE);
	data->knconf->knc_protofmly = pf;
	data->knconf->knc_proto = p;

	/* move server netname to the sec_data structure */
	data->netname = kmem_alloc(nlen, KM_SLEEP);
	bcopy(netname, data->netname, nlen);
	data->netnamelen = (int)nlen;

	secdata->secmod = AUTH_DH;
	secdata->rpcflavor = AUTH_DH;
	secdata->data = (caddr_t)data;

	return (secdata);
}

/*
 * Returns (deep) copy of sec_data_t. Allocates all memory required; caller
 * is responsible for freeing.
 */
sec_data_t *
copy_sec_data(sec_data_t *fsecdata)
{
	sec_data_t *tsecdata;

	if (fsecdata == NULL)
		return (NULL);

	if (fsecdata->rpcflavor == AUTH_DH) {
		dh_k4_clntdata_t *fdata = (dh_k4_clntdata_t *)fsecdata->data;

		if (fdata == NULL)
			return (NULL);

		tsecdata = (sec_data_t *)create_authdh_data(fdata->netname,
		    fdata->netnamelen, &fdata->syncaddr, fdata->knconf);

		return (tsecdata);
	}

	tsecdata = kmem_zalloc(sizeof (sec_data_t), KM_SLEEP);

	tsecdata->secmod = fsecdata->secmod;
	tsecdata->rpcflavor = fsecdata->rpcflavor;
	tsecdata->flags = fsecdata->flags;
	tsecdata->uid = fsecdata->uid;

	if (fsecdata->rpcflavor == RPCSEC_GSS) {
		gss_clntdata_t *gcd = (gss_clntdata_t *)fsecdata->data;

		tsecdata->data = (caddr_t)copy_sec_data_gss(gcd);
	} else {
		tsecdata->data = NULL;
	}

	return (tsecdata);
}

gss_clntdata_t *
copy_sec_data_gss(gss_clntdata_t *fdata)
{
	gss_clntdata_t *tdata;

	if (fdata == NULL)
		return (NULL);

	tdata = kmem_zalloc(sizeof (gss_clntdata_t), KM_SLEEP);

	tdata->mechanism.length = fdata->mechanism.length;
	tdata->mechanism.elements = kmem_zalloc(fdata->mechanism.length,
	    KM_SLEEP);
	bcopy(fdata->mechanism.elements, tdata->mechanism.elements,
	    fdata->mechanism.length);

	tdata->service = fdata->service;

	(void) strcpy(tdata->uname, fdata->uname);
	(void) strcpy(tdata->inst, fdata->inst);
	(void) strcpy(tdata->realm, fdata->realm);

	tdata->qop = fdata->qop;

	return (tdata);
}

static int
nfs4_chkdup_servinfo4(servinfo4_t *svp_head, servinfo4_t *svp)
{
	servinfo4_t *si;

	/*
	 * Iterate over the servinfo4 list to make sure
	 * we do not have a duplicate. Skip any servinfo4
	 * that has been marked "NOT IN USE"
	 */
	for (si = svp_head; si; si = si->sv_next) {
		(void) nfs_rw_enter_sig(&si->sv_lock, RW_READER, 0);
		if (si->sv_flags & SV4_NOTINUSE) {
			nfs_rw_exit(&si->sv_lock);
			continue;
		}
		nfs_rw_exit(&si->sv_lock);
		if (si == svp)
			continue;
		if (si->sv_addr.len == svp->sv_addr.len &&
		    strcmp(si->sv_knconf->knc_protofmly,
		    svp->sv_knconf->knc_protofmly) == 0 &&
		    bcmp(si->sv_addr.buf, svp->sv_addr.buf,
		    si->sv_addr.len) == 0) {
			/* it's a duplicate */
			return (1);
		}
	}
	/* it's not a duplicate */
	return (0);
}

void
nfs4_free_args(struct nfs_args *nargs)
{
	if (nargs->knconf) {
		if (nargs->knconf->knc_protofmly)
			kmem_free(nargs->knconf->knc_protofmly,
			    KNC_STRSIZE);
		if (nargs->knconf->knc_proto)
			kmem_free(nargs->knconf->knc_proto, KNC_STRSIZE);
		kmem_free(nargs->knconf, sizeof (*nargs->knconf));
		nargs->knconf = NULL;
	}

	if (nargs->fh) {
		kmem_free(nargs->fh, strlen(nargs->fh) + 1);
		nargs->fh = NULL;
	}

	if (nargs->hostname) {
		kmem_free(nargs->hostname, strlen(nargs->hostname) + 1);
		nargs->hostname = NULL;
	}

	if (nargs->addr) {
		if (nargs->addr->buf) {
			ASSERT(nargs->addr->len);
			kmem_free(nargs->addr->buf, nargs->addr->len);
		}
		kmem_free(nargs->addr, sizeof (struct netbuf));
		nargs->addr = NULL;
	}

	if (nargs->syncaddr) {
		ASSERT(nargs->syncaddr->len);
		if (nargs->syncaddr->buf) {
			ASSERT(nargs->syncaddr->len);
			kmem_free(nargs->syncaddr->buf, nargs->syncaddr->len);
		}
		kmem_free(nargs->syncaddr, sizeof (struct netbuf));
		nargs->syncaddr = NULL;
	}

	if (nargs->netname) {
		kmem_free(nargs->netname, strlen(nargs->netname) + 1);
		nargs->netname = NULL;
	}

	if (nargs->nfs_ext_u.nfs_extA.secdata) {
		sec_clnt_freeinfo(
		    nargs->nfs_ext_u.nfs_extA.secdata);
		nargs->nfs_ext_u.nfs_extA.secdata = NULL;
	}
}


int
nfs4_copyin(char *data, int datalen, struct nfs_args *nargs)
{

	int error;
	size_t hlen;			/* length of hostname */
	size_t nlen;			/* length of netname */
	char netname[MAXNETNAMELEN+1];	/* server's netname */
	struct netbuf addr;		/* server's address */
	struct netbuf syncaddr;		/* AUTH_DES time sync addr */
	struct knetconfig *knconf;		/* transport structure */
	struct sec_data *secdata = NULL;	/* security data */
	STRUCT_DECL(nfs_args, args);		/* nfs mount arguments */
	STRUCT_DECL(knetconfig, knconf_tmp);
	STRUCT_DECL(netbuf, addr_tmp);
	int flags;
	char *p, *pf;
	struct pathname pn;
	char *userbufptr;


	bzero(nargs, sizeof (*nargs));

	STRUCT_INIT(args, get_udatamodel());
	bzero(STRUCT_BUF(args), SIZEOF_STRUCT(nfs_args, DATAMODEL_NATIVE));
	if (copyin(data, STRUCT_BUF(args), MIN(datalen,
	    STRUCT_SIZE(args))))
		return (EFAULT);

	nargs->wsize = STRUCT_FGET(args, wsize);
	nargs->rsize = STRUCT_FGET(args, rsize);
	nargs->timeo = STRUCT_FGET(args, timeo);
	nargs->retrans = STRUCT_FGET(args, retrans);
	nargs->acregmin = STRUCT_FGET(args, acregmin);
	nargs->acregmax = STRUCT_FGET(args, acregmax);
	nargs->acdirmin = STRUCT_FGET(args, acdirmin);
	nargs->acdirmax = STRUCT_FGET(args, acdirmax);

	flags = STRUCT_FGET(args, flags);
	nargs->flags = flags;

	addr.buf = NULL;
	syncaddr.buf = NULL;


	/*
	 * Allocate space for a knetconfig structure and
	 * its strings and copy in from user-land.
	 */
	knconf = kmem_zalloc(sizeof (*knconf), KM_SLEEP);
	STRUCT_INIT(knconf_tmp, get_udatamodel());
	if (copyin(STRUCT_FGETP(args, knconf), STRUCT_BUF(knconf_tmp),
	    STRUCT_SIZE(knconf_tmp))) {
		kmem_free(knconf, sizeof (*knconf));
		return (EFAULT);
	}

	knconf->knc_semantics = STRUCT_FGET(knconf_tmp, knc_semantics);
	knconf->knc_protofmly = STRUCT_FGETP(knconf_tmp, knc_protofmly);
	knconf->knc_proto = STRUCT_FGETP(knconf_tmp, knc_proto);
	if (get_udatamodel() != DATAMODEL_LP64) {
		knconf->knc_rdev = expldev(STRUCT_FGET(knconf_tmp, knc_rdev));
	} else {
		knconf->knc_rdev = STRUCT_FGET(knconf_tmp, knc_rdev);
	}

	pf = kmem_alloc(KNC_STRSIZE, KM_SLEEP);
	p = kmem_alloc(KNC_STRSIZE, KM_SLEEP);
	error = copyinstr(knconf->knc_protofmly, pf, KNC_STRSIZE, NULL);
	if (error) {
		kmem_free(pf, KNC_STRSIZE);
		kmem_free(p, KNC_STRSIZE);
		kmem_free(knconf, sizeof (*knconf));
		return (error);
	}

	error = copyinstr(knconf->knc_proto, p, KNC_STRSIZE, NULL);
	if (error) {
		kmem_free(pf, KNC_STRSIZE);
		kmem_free(p, KNC_STRSIZE);
		kmem_free(knconf, sizeof (*knconf));
		return (error);
	}


	knconf->knc_protofmly = pf;
	knconf->knc_proto = p;

	nargs->knconf = knconf;

	/*
	 * Get server address
	 */
	STRUCT_INIT(addr_tmp, get_udatamodel());
	if (copyin(STRUCT_FGETP(args, addr), STRUCT_BUF(addr_tmp),
	    STRUCT_SIZE(addr_tmp))) {
		error = EFAULT;
		goto errout;
	}

	nargs->addr = kmem_zalloc(sizeof (struct netbuf), KM_SLEEP);
	userbufptr = STRUCT_FGETP(addr_tmp, buf);
	addr.len = STRUCT_FGET(addr_tmp, len);
	addr.buf = kmem_alloc(addr.len, KM_SLEEP);
	addr.maxlen = addr.len;
	if (copyin(userbufptr, addr.buf, addr.len)) {
		kmem_free(addr.buf, addr.len);
		error = EFAULT;
		goto errout;
	}
	bcopy(&addr, nargs->addr, sizeof (struct netbuf));

	/*
	 * Get the root fhandle
	 */
	error = pn_get(STRUCT_FGETP(args, fh), UIO_USERSPACE, &pn);
	if (error)
		goto errout;

	/* Volatile fh: keep server paths, so use actual-size strings */
	nargs->fh = kmem_alloc(pn.pn_pathlen + 1, KM_SLEEP);
	bcopy(pn.pn_path, nargs->fh, pn.pn_pathlen);
	nargs->fh[pn.pn_pathlen] = '\0';
	pn_free(&pn);


	/*
	 * Get server's hostname
	 */
	if (flags & NFSMNT_HOSTNAME) {
		error = copyinstr(STRUCT_FGETP(args, hostname),
		    netname, sizeof (netname), &hlen);
		if (error)
			goto errout;
		nargs->hostname = kmem_zalloc(hlen, KM_SLEEP);
		(void) strcpy(nargs->hostname, netname);

	} else {
		nargs->hostname = NULL;
	}


	/*
	 * If there are syncaddr and netname data, load them in. This is
	 * to support data needed for NFSV4 when AUTH_DH is the negotiated
	 * flavor via SECINFO. (instead of using MOUNT protocol in V3).
	 */
	netname[0] = '\0';
	if (flags & NFSMNT_SECURE) {

		/* get syncaddr */
		STRUCT_INIT(addr_tmp, get_udatamodel());
		if (copyin(STRUCT_FGETP(args, syncaddr), STRUCT_BUF(addr_tmp),
		    STRUCT_SIZE(addr_tmp))) {
			error = EINVAL;
			goto errout;
		}
		userbufptr = STRUCT_FGETP(addr_tmp, buf);
		syncaddr.len = STRUCT_FGET(addr_tmp, len);
		syncaddr.buf = kmem_alloc(syncaddr.len, KM_SLEEP);
		syncaddr.maxlen = syncaddr.len;
		if (copyin(userbufptr, syncaddr.buf, syncaddr.len)) {
			kmem_free(syncaddr.buf, syncaddr.len);
			error = EFAULT;
			goto errout;
		}

		nargs->syncaddr = kmem_alloc(sizeof (struct netbuf), KM_SLEEP);
		bcopy(&syncaddr, nargs->syncaddr, sizeof (struct netbuf));

		/* get server's netname */
		if (copyinstr(STRUCT_FGETP(args, netname), netname,
		    sizeof (netname), &nlen)) {
			error = EFAULT;
			goto errout;
		}

		netname[nlen] = '\0';
		nargs->netname = kmem_zalloc(nlen, KM_SLEEP);
		(void) strcpy(nargs->netname, netname);
	}

	/*
	 * Get the extention data which has the security data structure.
	 * This includes data for AUTH_SYS as well.
	 */
	if (flags & NFSMNT_NEWARGS) {
		nargs->nfs_args_ext = STRUCT_FGET(args, nfs_args_ext);
		if (nargs->nfs_args_ext == NFS_ARGS_EXTA ||
		    nargs->nfs_args_ext == NFS_ARGS_EXTB) {
			/*
			 * Indicating the application is using the new
			 * sec_data structure to pass in the security
			 * data.
			 */
			if (STRUCT_FGETP(args,
			    nfs_ext_u.nfs_extA.secdata) != NULL) {
				error = sec_clnt_loadinfo(
				    (struct sec_data *)STRUCT_FGETP(args,
				    nfs_ext_u.nfs_extA.secdata),
				    &secdata, get_udatamodel());
			}
			nargs->nfs_ext_u.nfs_extA.secdata = secdata;
		}
	}

	if (error)
		goto errout;

	/*
	 * Failover support:
	 *
	 * We may have a linked list of nfs_args structures,
	 * which means the user is looking for failover.  If
	 * the mount is either not "read-only" or "soft",
	 * we want to bail out with EINVAL.
	 */
	if (nargs->nfs_args_ext == NFS_ARGS_EXTB)
		nargs->nfs_ext_u.nfs_extB.next =
		    STRUCT_FGETP(args, nfs_ext_u.nfs_extB.next);

errout:
	if (error)
		nfs4_free_args(nargs);

	return (error);
}


/*
 * nfs mount vfsop
 * Set up mount info record and attach it to vfs struct.
 */
int
nfs4_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
{
	char *data = uap->dataptr;
	int error;
	vnode_t *rtvp;			/* the server's root */
	mntinfo4_t *mi;			/* mount info, pointed at by vfs */
	struct knetconfig *rdma_knconf;	/* rdma transport structure */
	rnode4_t *rp;
	struct servinfo4 *svp;		/* nfs server info */
	struct servinfo4 *svp_tail = NULL; /* previous nfs server info */
	struct servinfo4 *svp_head;	/* first nfs server info */
	struct servinfo4 *svp_2ndlast;	/* 2nd last in server info list */
	struct sec_data *secdata;	/* security data */
	struct nfs_args *args = NULL;
	int flags, addr_type, removed;
	zone_t *zone = nfs_zone();
	nfs4_error_t n4e;
	zone_t *mntzone = NULL;

	if (secpolicy_fs_mount(cr, mvp, vfsp) != 0)
		return (EPERM);
	if (mvp->v_type != VDIR)
		return (ENOTDIR);

	/*
	 * get arguments
	 *
	 * nfs_args is now versioned and is extensible, so
	 * uap->datalen might be different from sizeof (args)
	 * in a compatible situation.
	 */
more:
	if (!(uap->flags & MS_SYSSPACE)) {
		if (args == NULL)
			args = kmem_zalloc(sizeof (struct nfs_args), KM_SLEEP);
		else
			nfs4_free_args(args);
		error = nfs4_copyin(data, uap->datalen, args);
		if (error) {
			if (args) {
				kmem_free(args, sizeof (*args));
			}
			return (error);
		}
	} else {
		args = (struct nfs_args *)data;
	}

	flags = args->flags;

	/*
	 * If the request changes the locking type, disallow the remount,
	 * because it's questionable whether we can transfer the
	 * locking state correctly.
	 */
	if (uap->flags & MS_REMOUNT) {
		if (!(uap->flags & MS_SYSSPACE)) {
			nfs4_free_args(args);
			kmem_free(args, sizeof (*args));
		}
		if ((mi = VFTOMI4(vfsp)) != NULL) {
			uint_t new_mi_llock;
			uint_t old_mi_llock;
			new_mi_llock = (flags & NFSMNT_LLOCK) ? 1 : 0;
			old_mi_llock = (mi->mi_flags & MI4_LLOCK) ? 1 : 0;
			if (old_mi_llock != new_mi_llock)
				return (EBUSY);
		}
		return (0);
	}

	/*
	 * For ephemeral mount trigger stub vnodes, we have two problems
	 * to solve: racing threads will likely fail the v_count check, and
	 * we want only one to proceed with the mount.
	 *
	 * For stubs, if the mount has already occurred (via a racing thread),
	 * just return success. If not, skip the v_count check and proceed.
	 * Note that we are already serialised at this point.
	 */
	mutex_enter(&mvp->v_lock);
	if (vn_matchops(mvp, nfs4_trigger_vnodeops)) {
		/* mntpt is a v4 stub vnode */
		ASSERT(RP_ISSTUB(VTOR4(mvp)));
		ASSERT(!(uap->flags & MS_OVERLAY));
		ASSERT(!(mvp->v_flag & VROOT));
		if (vn_mountedvfs(mvp) != NULL) {
			/* ephemeral mount has already occurred */
			ASSERT(uap->flags & MS_SYSSPACE);
			mutex_exit(&mvp->v_lock);
			return (0);
		}
	} else {
		/* mntpt is a non-v4 or v4 non-stub vnode */
		if (!(uap->flags & MS_OVERLAY) &&
		    (mvp->v_count != 1 || (mvp->v_flag & VROOT))) {
			mutex_exit(&mvp->v_lock);
			if (!(uap->flags & MS_SYSSPACE)) {
				nfs4_free_args(args);
				kmem_free(args, sizeof (*args));
			}
			return (EBUSY);
		}
	}
	mutex_exit(&mvp->v_lock);

	/* make sure things are zeroed for errout: */
	rtvp = NULL;
	mi = NULL;
	secdata = NULL;

	/*
	 * A valid knetconfig structure is required.
	 */
	if (!(flags & NFSMNT_KNCONF) ||
	    args->knconf == NULL || args->knconf->knc_protofmly == NULL ||
	    args->knconf->knc_proto == NULL ||
	    (strcmp(args->knconf->knc_proto, NC_UDP) == 0)) {
		if (!(uap->flags & MS_SYSSPACE)) {
			nfs4_free_args(args);
			kmem_free(args, sizeof (*args));
		}
		return (EINVAL);
	}

	if ((strlen(args->knconf->knc_protofmly) >= KNC_STRSIZE) ||
	    (strlen(args->knconf->knc_proto) >= KNC_STRSIZE)) {
		if (!(uap->flags & MS_SYSSPACE)) {
			nfs4_free_args(args);
			kmem_free(args, sizeof (*args));
		}
		return (EINVAL);
	}

	/*
	 * Allocate a servinfo4 struct.
	 */
	svp = kmem_zalloc(sizeof (*svp), KM_SLEEP);
	nfs_rw_init(&svp->sv_lock, NULL, RW_DEFAULT, NULL);
	if (svp_tail) {
		svp_2ndlast = svp_tail;
		svp_tail->sv_next = svp;
	} else {
		svp_head = svp;
		svp_2ndlast = svp;
	}

	svp_tail = svp;
	svp->sv_knconf = args->knconf;
	args->knconf = NULL;

	/*
	 * Get server address
	 */
	if (args->addr == NULL || args->addr->buf == NULL) {
		error = EINVAL;
		goto errout;
	}

	svp->sv_addr.maxlen = args->addr->maxlen;
	svp->sv_addr.len = args->addr->len;
	svp->sv_addr.buf = args->addr->buf;
	args->addr->buf = NULL;

	/*
	 * Get the root fhandle
	 */
	if (args->fh == NULL || (strlen(args->fh) >= MAXPATHLEN)) {
		error = EINVAL;
		goto errout;
	}

	svp->sv_path = args->fh;
	svp->sv_pathlen = strlen(args->fh) + 1;
	args->fh = NULL;

	/*
	 * Get server's hostname
	 */
	if (flags & NFSMNT_HOSTNAME) {
		if (args->hostname == NULL || (strlen(args->hostname) >
		    MAXNETNAMELEN)) {
			error = EINVAL;
			goto errout;
		}
		svp->sv_hostnamelen = strlen(args->hostname) + 1;
		svp->sv_hostname = args->hostname;
		args->hostname = NULL;
	} else {
		char *p = "unknown-host";
		svp->sv_hostnamelen = strlen(p) + 1;
		svp->sv_hostname = kmem_zalloc(svp->sv_hostnamelen, KM_SLEEP);
		(void) strcpy(svp->sv_hostname, p);
	}

	/*
	 * RDMA MOUNT SUPPORT FOR NFS v4.
	 * Establish, is it possible to use RDMA, if so overload the
	 * knconf with rdma specific knconf and free the orignal knconf.
	 */
	if ((flags & NFSMNT_TRYRDMA) || (flags & NFSMNT_DORDMA)) {
		/*
		 * Determine the addr type for RDMA, IPv4 or v6.
		 */
		if (strcmp(svp->sv_knconf->knc_protofmly, NC_INET) == 0)
			addr_type = AF_INET;
		else if (strcmp(svp->sv_knconf->knc_protofmly, NC_INET6) == 0)
			addr_type = AF_INET6;