fs/nfs/nfs3_srv.c

/*
 * 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 2008 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

/* 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/buf.h>
#include <sys/vfs.h>
#include <sys/vnode.h>
#include <sys/uio.h>
#include <sys/errno.h>
#include <sys/sysmacros.h>
#include <sys/statvfs.h>
#include <sys/kmem.h>
#include <sys/dirent.h>
#include <sys/cmn_err.h>
#include <sys/debug.h>
#include <sys/systeminfo.h>
#include <sys/flock.h>
#include <sys/nbmlock.h>
#include <sys/policy.h>
#include <sys/sdt.h>

#include <rpc/types.h>
#include <rpc/auth.h>
#include <rpc/svc.h>
#include <rpc/rpc_rdma.h>

#include <nfs/nfs.h>
#include <nfs/export.h>

#include <sys/strsubr.h>

#include <sys/tsol/label.h>
#include <sys/tsol/tndb.h>

#include <inet/ip.h>
#include <inet/ip6.h>

/*
 * These are the interface routines for the server side of the
 * Network File System.  See the NFS version 3 protocol specification
 * for a description of this interface.
 */

#ifdef DEBUG
int rfs3_do_pre_op_attr = 1;
int rfs3_do_post_op_attr = 1;
int rfs3_do_post_op_fh3 = 1;
#endif

static writeverf3 write3verf;

static int	sattr3_to_vattr(sattr3 *, struct vattr *);
static int	vattr_to_fattr3(struct vattr *, fattr3 *);
static int	vattr_to_wcc_attr(struct vattr *, wcc_attr *);
static void	vattr_to_pre_op_attr(struct vattr *, pre_op_attr *);
static void	vattr_to_wcc_data(struct vattr *, struct vattr *, wcc_data *);
static int	rdma_setup_read_data3(READ3args *, READ3resok *);

u_longlong_t nfs3_srv_caller_id;

/* ARGSUSED */
void
rfs3_getattr(GETATTR3args *args, GETATTR3res *resp, struct exportinfo *exi,
	struct svc_req *req, cred_t *cr)
{
	int error;
	vnode_t *vp;
	struct vattr va;

	vp = nfs3_fhtovp(&args->object, exi);

	DTRACE_NFSV3_4(op__getattr__start, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, GETATTR3args *, args);

	if (vp == NULL) {
		error = ESTALE;
		goto out;
	}

	va.va_mask = AT_ALL;
	error = rfs4_delegated_getattr(vp, &va, 0, cr);

	if (!error) {
		/* overflow error if time or size is out of range */
		error = vattr_to_fattr3(&va, &resp->resok.obj_attributes);
		if (error)
			goto out;
		resp->status = NFS3_OK;

		DTRACE_NFSV3_4(op__getattr__done, struct svc_req *, req,
		    cred_t *, cr, vnode_t *, vp, GETATTR3res *, resp);

		VN_RELE(vp);

		return;
	}

out:
	if (curthread->t_flag & T_WOULDBLOCK) {
		curthread->t_flag &= ~T_WOULDBLOCK;
		resp->status = NFS3ERR_JUKEBOX;
	} else
		resp->status = puterrno3(error);

	DTRACE_NFSV3_4(op__getattr__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, GETATTR3res *, resp);

	if (vp != NULL)
		VN_RELE(vp);
}

void *
rfs3_getattr_getfh(GETATTR3args *args)
{

	return (&args->object);
}

void
rfs3_setattr(SETATTR3args *args, SETATTR3res *resp, struct exportinfo *exi,
	struct svc_req *req, cred_t *cr)
{
	int error;
	vnode_t *vp;
	struct vattr *bvap;
	struct vattr bva;
	struct vattr *avap;
	struct vattr ava;
	int flag;
	int in_crit = 0;
	struct flock64 bf;
	caller_context_t ct;

	bvap = NULL;
	avap = NULL;

	vp = nfs3_fhtovp(&args->object, exi);

	DTRACE_NFSV3_4(op__setattr__start, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, SETATTR3args *, args);

	if (vp == NULL) {
		error = ESTALE;
		goto out;
	}

	error = sattr3_to_vattr(&args->new_attributes, &ava);
	if (error)
		goto out;

	if (is_system_labeled()) {
		bslabel_t *clabel = req->rq_label;

		ASSERT(clabel != NULL);
		DTRACE_PROBE2(tx__rfs3__log__info__opsetattr__clabel, char *,
		    "got client label from request(1)", struct svc_req *, req);

		if (!blequal(&l_admin_low->tsl_label, clabel)) {
			if (!do_rfs_label_check(clabel, vp, EQUALITY_CHECK)) {
				resp->status = NFS3ERR_ACCES;
				goto out1;
			}
		}
	}

	/*
	 * We need to specially handle size changes because of
	 * possible conflicting NBMAND locks. Get into critical
	 * region before VOP_GETATTR, so the size attribute is
	 * valid when checking conflicts.
	 *
	 * Also, check to see if the v4 side of the server has
	 * delegated this file.  If so, then we return JUKEBOX to
	 * allow the client to retrasmit its request.
	 */
	if (vp->v_type == VREG && (ava.va_mask & AT_SIZE)) {
		if (nbl_need_check(vp)) {
			nbl_start_crit(vp, RW_READER);
			in_crit = 1;
		}
	}

	bva.va_mask = AT_ALL;
	error = rfs4_delegated_getattr(vp, &bva, 0, cr);

	/*
	 * If we can't get the attributes, then we can't do the
	 * right access checking.  So, we'll fail the request.
	 */
	if (error)
		goto out;

#ifdef DEBUG
	if (rfs3_do_pre_op_attr)
		bvap = &bva;
#else
	bvap = &bva;
#endif

	if (rdonly(exi, req) || vn_is_readonly(vp)) {
		resp->status = NFS3ERR_ROFS;
		goto out1;
	}

	if (args->guard.check &&
	    (args->guard.obj_ctime.seconds != bva.va_ctime.tv_sec ||
	    args->guard.obj_ctime.nseconds != bva.va_ctime.tv_nsec)) {
		resp->status = NFS3ERR_NOT_SYNC;
		goto out1;
	}

	if (args->new_attributes.mtime.set_it == SET_TO_CLIENT_TIME)
		flag = ATTR_UTIME;
	else
		flag = 0;

	/*
	 * If the filesystem is exported with nosuid, then mask off
	 * the setuid and setgid bits.
	 */
	if ((ava.va_mask & AT_MODE) && vp->v_type == VREG &&
	    (exi->exi_export.ex_flags & EX_NOSUID))
		ava.va_mode &= ~(VSUID | VSGID);

	ct.cc_sysid = 0;
	ct.cc_pid = 0;
	ct.cc_caller_id = nfs3_srv_caller_id;
	ct.cc_flags = CC_DONTBLOCK;

	/*
	 * We need to specially handle size changes because it is
	 * possible for the client to create a file with modes
	 * which indicate read-only, but with the file opened for
	 * writing.  If the client then tries to set the size of
	 * the file, then the normal access checking done in
	 * VOP_SETATTR would prevent the client from doing so,
	 * although it should be legal for it to do so.  To get
	 * around this, we do the access checking for ourselves
	 * and then use VOP_SPACE which doesn't do the access
	 * checking which VOP_SETATTR does. VOP_SPACE can only
	 * operate on VREG files, let VOP_SETATTR handle the other
	 * extremely rare cases.
	 * Also the client should not be allowed to change the
	 * size of the file if there is a conflicting non-blocking
	 * mandatory lock in the region the change.
	 */
	if (vp->v_type == VREG && (ava.va_mask & AT_SIZE)) {
		if (in_crit) {
			u_offset_t offset;
			ssize_t length;

			if (ava.va_size < bva.va_size) {
				offset = ava.va_size;
				length = bva.va_size - ava.va_size;
			} else {
				offset = bva.va_size;
				length = ava.va_size - bva.va_size;
			}
			if (nbl_conflict(vp, NBL_WRITE, offset, length, 0,
			    NULL)) {
				error = EACCES;
				goto out;
			}
		}

		if (crgetuid(cr) == bva.va_uid && ava.va_size != bva.va_size) {
			ava.va_mask &= ~AT_SIZE;
			bf.l_type = F_WRLCK;
			bf.l_whence = 0;
			bf.l_start = (off64_t)ava.va_size;
			bf.l_len = 0;
			bf.l_sysid = 0;
			bf.l_pid = 0;
			error = VOP_SPACE(vp, F_FREESP, &bf, FWRITE,
			    (offset_t)ava.va_size, cr, &ct);
		}
	}

	if (!error && ava.va_mask)
		error = VOP_SETATTR(vp, &ava, flag, cr, &ct);

	/* check if a monitor detected a delegation conflict */
	if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) {
		resp->status = NFS3ERR_JUKEBOX;
		goto out1;
	}

#ifdef DEBUG
	if (rfs3_do_post_op_attr) {
		ava.va_mask = AT_ALL;
		avap = rfs4_delegated_getattr(vp, &ava, 0, cr) ? NULL : &ava;
	} else
		avap = NULL;
#else
	ava.va_mask = AT_ALL;
	avap = rfs4_delegated_getattr(vp, &ava, 0, cr) ? NULL : &ava;
#endif

	/*
	 * Force modified metadata out to stable storage.
	 */
	(void) VOP_FSYNC(vp, FNODSYNC, cr, &ct);

	if (error)
		goto out;

	if (in_crit)
		nbl_end_crit(vp);

	resp->status = NFS3_OK;
	vattr_to_wcc_data(bvap, avap, &resp->resok.obj_wcc);

	DTRACE_NFSV3_4(op__setattr__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, SETATTR3res *, resp);

	VN_RELE(vp);

	return;

out:
	if (curthread->t_flag & T_WOULDBLOCK) {
		curthread->t_flag &= ~T_WOULDBLOCK;
		resp->status = NFS3ERR_JUKEBOX;
	} else
		resp->status = puterrno3(error);
out1:
	DTRACE_NFSV3_4(op__setattr__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, SETATTR3res *, resp);

	if (vp != NULL) {
		if (in_crit)
			nbl_end_crit(vp);
		VN_RELE(vp);
	}
	vattr_to_wcc_data(bvap, avap, &resp->resfail.obj_wcc);
}

void *
rfs3_setattr_getfh(SETATTR3args *args)
{

	return (&args->object);
}

/* ARGSUSED */
void
rfs3_lookup(LOOKUP3args *args, LOOKUP3res *resp, struct exportinfo *exi,
	struct svc_req *req, cred_t *cr)
{
	int error;
	vnode_t *vp;
	vnode_t *dvp;
	struct vattr *vap;
	struct vattr va;
	struct vattr *dvap;
	struct vattr dva;
	nfs_fh3 *fhp;
	struct sec_ol sec = {0, 0};
	bool_t publicfh_flag = FALSE, auth_weak = FALSE;

	dvap = NULL;

	/*
	 * Allow lookups from the root - the default
	 * location of the public filehandle.
	 */
	if (exi != NULL && (exi->exi_export.ex_flags & EX_PUBLIC)) {
		dvp = rootdir;
		VN_HOLD(dvp);

		DTRACE_NFSV3_4(op__lookup__start, struct svc_req *, req,
		    cred_t *, cr, vnode_t *, dvp, LOOKUP3args *, args);
	} else {
		dvp = nfs3_fhtovp(&args->what.dir, exi);

		DTRACE_NFSV3_4(op__lookup__start, struct svc_req *, req,
		    cred_t *, cr, vnode_t *, dvp, LOOKUP3args *, args);

		if (dvp == NULL) {
			error = ESTALE;
			goto out;
		}
	}

#ifdef DEBUG
	if (rfs3_do_pre_op_attr) {
		dva.va_mask = AT_ALL;
		dvap = VOP_GETATTR(dvp, &dva, 0, cr, NULL) ? NULL : &dva;
	}
#else
	dva.va_mask = AT_ALL;
	dvap = VOP_GETATTR(dvp, &dva, 0, cr, NULL) ? NULL : &dva;
#endif

	if (args->what.name == nfs3nametoolong) {
		resp->status = NFS3ERR_NAMETOOLONG;
		goto out1;
	}

	if (args->what.name == NULL || *(args->what.name) == '\0') {
		resp->status = NFS3ERR_ACCES;
		goto out1;
	}

	fhp = &args->what.dir;
	if (strcmp(args->what.name, "..") == 0 &&
	    EQFID(&exi->exi_fid, FH3TOFIDP(fhp))) {
		resp->status = NFS3ERR_NOENT;
		goto out1;
	}

	/*
	 * If the public filehandle is used then allow
	 * a multi-component lookup
	 */
	if (PUBLIC_FH3(&args->what.dir)) {
		publicfh_flag = TRUE;
		error = rfs_publicfh_mclookup(args->what.name, dvp, cr, &vp,
		    &exi, &sec);
		if (error && exi != NULL)
			exi_rele(exi); /* See comment below Re: publicfh_flag */
		/*
		 * Since WebNFS may bypass MOUNT, we need to ensure this
		 * request didn't come from an unlabeled admin_low client.
		 */
		if (is_system_labeled() && error == 0) {
			struct sockaddr *ca;
			int		addr_type;
			void		*ipaddr;
			tsol_tpc_t	*tp;

			ca = (struct sockaddr *)svc_getrpccaller(
			    req->rq_xprt)->buf;
			if (ca->sa_family == AF_INET) {
				addr_type = IPV4_VERSION;
				ipaddr = &((struct sockaddr_in *)ca)->sin_addr;
			} else if (ca->sa_family == AF_INET6) {
				addr_type = IPV6_VERSION;
				ipaddr = &((struct sockaddr_in6 *)
				    ca)->sin6_addr;
			}
			tp = find_tpc(ipaddr, addr_type, B_FALSE);
			if (tp == NULL || tp->tpc_tp.tp_doi !=
			    l_admin_low->tsl_doi || tp->tpc_tp.host_type !=
			    SUN_CIPSO) {
				if (exi != NULL)
					exi_rele(exi);
				VN_RELE(vp);
				resp->status = NFS3ERR_ACCES;
				error = 1;
			}
			if (tp != NULL)
				TPC_RELE(tp);
		}
	} else {
		error = VOP_LOOKUP(dvp, args->what.name, &vp,
		    NULL, 0, NULL, cr, NULL, NULL, NULL);
	}

	if (is_system_labeled() && error == 0) {
		bslabel_t *clabel = req->rq_label;

		ASSERT(clabel != NULL);
		DTRACE_PROBE2(tx__rfs3__log__info__oplookup__clabel, char *,
		    "got client label from request(1)", struct svc_req *, req);

		if (!blequal(&l_admin_low->tsl_label, clabel)) {
			if (!do_rfs_label_check(clabel, dvp,
			    DOMINANCE_CHECK)) {
				if (publicfh_flag && exi != NULL)
					exi_rele(exi);
				VN_RELE(vp);
				resp->status = NFS3ERR_ACCES;
				error = 1;
			}
		}
	}

#ifdef DEBUG
	if (rfs3_do_post_op_attr) {
		dva.va_mask = AT_ALL;
		dvap = VOP_GETATTR(dvp, &dva, 0, cr, NULL) ? NULL : &dva;
	} else
		dvap = NULL;
#else
	dva.va_mask = AT_ALL;
	dvap = VOP_GETATTR(dvp, &dva, 0, cr, NULL) ? NULL : &dva;
#endif

	if (error)
		goto out;

	if (sec.sec_flags & SEC_QUERY) {
		error = makefh3_ol(&resp->resok.object, exi, sec.sec_index);
	} else {
		error = makefh3(&resp->resok.object, vp, exi);
		if (!error && publicfh_flag && !chk_clnt_sec(exi, req))
			auth_weak = TRUE;
	}

	if (error) {
		VN_RELE(vp);
		goto out;
	}

	/*
	 * If publicfh_flag is true then we have called rfs_publicfh_mclookup
	 * and have obtained a new exportinfo in exi which needs to be
	 * released. Note the the original exportinfo pointed to by exi
	 * will be released by the caller, common_dispatch.
	 */
	if (publicfh_flag)
		exi_rele(exi);

#ifdef DEBUG
	if (rfs3_do_post_op_attr) {
		va.va_mask = AT_ALL;
		vap = rfs4_delegated_getattr(vp, &va, 0, cr) ? NULL : &va;
	} else
		vap = NULL;
#else
	va.va_mask = AT_ALL;
	vap = rfs4_delegated_getattr(vp, &va, 0, cr) ? NULL : &va;
#endif

	VN_RELE(vp);

	resp->status = NFS3_OK;
	vattr_to_post_op_attr(vap, &resp->resok.obj_attributes);
	vattr_to_post_op_attr(dvap, &resp->resok.dir_attributes);

	/*
	 * If it's public fh, no 0x81, and client's flavor is
	 * invalid, set WebNFS status to WNFSERR_CLNT_FLAVOR now.
	 * Then set RPC status to AUTH_TOOWEAK in common_dispatch.
	 */
	if (auth_weak)
		resp->status = (enum nfsstat3)WNFSERR_CLNT_FLAVOR;

	DTRACE_NFSV3_4(op__lookup__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, dvp, LOOKUP3res *, resp);
	VN_RELE(dvp);

	return;

out:
	if (curthread->t_flag & T_WOULDBLOCK) {
		curthread->t_flag &= ~T_WOULDBLOCK;
		resp->status = NFS3ERR_JUKEBOX;
	} else
		resp->status = puterrno3(error);
out1:
	DTRACE_NFSV3_4(op__lookup__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, dvp, LOOKUP3res *, resp);

	if (dvp != NULL)
		VN_RELE(dvp);
	vattr_to_post_op_attr(dvap, &resp->resfail.dir_attributes);

}

void *
rfs3_lookup_getfh(LOOKUP3args *args)
{

	return (&args->what.dir);
}

/* ARGSUSED */
void
rfs3_access(ACCESS3args *args, ACCESS3res *resp, struct exportinfo *exi,
	struct svc_req *req, cred_t *cr)
{
	int error;
	vnode_t *vp;
	struct vattr *vap;
	struct vattr va;
	int checkwriteperm;
	boolean_t dominant_label = B_FALSE;
	boolean_t equal_label = B_FALSE;
	boolean_t admin_low_client;

	vap = NULL;

	vp = nfs3_fhtovp(&args->object, exi);

	DTRACE_NFSV3_4(op__access__start, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, ACCESS3args *, args);

	if (vp == NULL) {
		error = ESTALE;
		goto out;
	}

	/*
	 * If the file system is exported read only, it is not appropriate
	 * to check write permissions for regular files and directories.
	 * Special files are interpreted by the client, so the underlying
	 * permissions are sent back to the client for interpretation.
	 */
	if (rdonly(exi, req) && (vp->v_type == VREG || vp->v_type == VDIR))
		checkwriteperm = 0;
	else
		checkwriteperm = 1;

	/*
	 * We need the mode so that we can correctly determine access
	 * permissions relative to a mandatory lock file.  Access to
	 * mandatory lock files is denied on the server, so it might
	 * as well be reflected to the server during the open.
	 */
	va.va_mask = AT_MODE;
	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
	if (error)
		goto out;

#ifdef DEBUG
	if (rfs3_do_post_op_attr)
		vap = &va;
#else
	vap = &va;
#endif

	resp->resok.access = 0;

	if (is_system_labeled()) {
		bslabel_t *clabel = req->rq_label;

		ASSERT(clabel != NULL);
		DTRACE_PROBE2(tx__rfs3__log__info__opaccess__clabel, char *,
		    "got client label from request(1)", struct svc_req *, req);

		if (!blequal(&l_admin_low->tsl_label, clabel)) {
			if ((equal_label = do_rfs_label_check(clabel, vp,
			    EQUALITY_CHECK)) == B_FALSE) {
				dominant_label = do_rfs_label_check(clabel,
				    vp, DOMINANCE_CHECK);
			} else
				dominant_label = B_TRUE;
			admin_low_client = B_FALSE;
		} else
			admin_low_client = B_TRUE;
	}

	if (args->access & ACCESS3_READ) {
		error = VOP_ACCESS(vp, VREAD, 0, cr, NULL);
		if (error) {
			if (curthread->t_flag & T_WOULDBLOCK)
				goto out;
		} else if (!MANDLOCK(vp, va.va_mode) &&
		    (!is_system_labeled() || admin_low_client ||
		    dominant_label))
			resp->resok.access |= ACCESS3_READ;
	}
	if ((args->access & ACCESS3_LOOKUP) && vp->v_type == VDIR) {
		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
		if (error) {
			if (curthread->t_flag & T_WOULDBLOCK)
				goto out;
		} else if (!is_system_labeled() || admin_low_client ||
		    dominant_label)
			resp->resok.access |= ACCESS3_LOOKUP;
	}
	if (checkwriteperm &&
	    (args->access & (ACCESS3_MODIFY|ACCESS3_EXTEND))) {
		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
		if (error) {
			if (curthread->t_flag & T_WOULDBLOCK)
				goto out;
		} else if (!MANDLOCK(vp, va.va_mode) &&
		    (!is_system_labeled() || admin_low_client || equal_label)) {
			resp->resok.access |=
			    (args->access & (ACCESS3_MODIFY|ACCESS3_EXTEND));
		}
	}
	if (checkwriteperm &&
	    (args->access & ACCESS3_DELETE) && vp->v_type == VDIR) {
		error = VOP_ACCESS(vp, VWRITE, 0, cr, NULL);
		if (error) {
			if (curthread->t_flag & T_WOULDBLOCK)
				goto out;
		} else if (!is_system_labeled() || admin_low_client ||
		    equal_label)
			resp->resok.access |= ACCESS3_DELETE;
	}
	if (args->access & ACCESS3_EXECUTE) {
		error = VOP_ACCESS(vp, VEXEC, 0, cr, NULL);
		if (error) {
			if (curthread->t_flag & T_WOULDBLOCK)
				goto out;
		} else if (!MANDLOCK(vp, va.va_mode) &&
		    (!is_system_labeled() || admin_low_client ||
		    dominant_label))
			resp->resok.access |= ACCESS3_EXECUTE;
	}

#ifdef DEBUG
	if (rfs3_do_post_op_attr) {
		va.va_mask = AT_ALL;
		vap = rfs4_delegated_getattr(vp, &va, 0, cr) ? NULL : &va;
	} else
		vap = NULL;
#else
	va.va_mask = AT_ALL;
	vap = rfs4_delegated_getattr(vp, &va, 0, cr) ? NULL : &va;
#endif

	resp->status = NFS3_OK;
	vattr_to_post_op_attr(vap, &resp->resok.obj_attributes);

	DTRACE_NFSV3_4(op__access__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, ACCESS3res *, resp);

	VN_RELE(vp);

	return;

out:
	if (curthread->t_flag & T_WOULDBLOCK) {
		curthread->t_flag &= ~T_WOULDBLOCK;
		resp->status = NFS3ERR_JUKEBOX;
	} else
		resp->status = puterrno3(error);
	DTRACE_NFSV3_4(op__access__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, ACCESS3res *, resp);
	if (vp != NULL)
		VN_RELE(vp);
	vattr_to_post_op_attr(vap, &resp->resfail.obj_attributes);
}

void *
rfs3_access_getfh(ACCESS3args *args)
{

	return (&args->object);
}

/* ARGSUSED */
void
rfs3_readlink(READLINK3args *args, READLINK3res *resp, struct exportinfo *exi,
	struct svc_req *req, cred_t *cr)
{
	int error;
	vnode_t *vp;
	struct vattr *vap;
	struct vattr va;
	struct iovec iov;
	struct uio uio;
	char *data;

	vap = NULL;

	vp = nfs3_fhtovp(&args->symlink, exi);

	DTRACE_NFSV3_4(op__readlink__start, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, READLINK3args *, args);

	if (vp == NULL) {
		error = ESTALE;
		goto out;
	}

	va.va_mask = AT_ALL;
	error = VOP_GETATTR(vp, &va, 0, cr, NULL);
	if (error)
		goto out;

#ifdef DEBUG
	if (rfs3_do_post_op_attr)
		vap = &va;
#else
	vap = &va;
#endif

	if (vp->v_type != VLNK) {
		resp->status = NFS3ERR_INVAL;
		goto out1;
	}

	if (MANDLOCK(vp, va.va_mode)) {
		resp->status = NFS3ERR_ACCES;
		goto out1;
	}

	if (is_system_labeled()) {
		bslabel_t *clabel = req->rq_label;

		ASSERT(clabel != NULL);
		DTRACE_PROBE2(tx__rfs3__log__info__opreadlink__clabel, char *,
		    "got client label from request(1)", struct svc_req *, req);

		if (!blequal(&l_admin_low->tsl_label, clabel)) {
			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
				resp->status = NFS3ERR_ACCES;
				goto out1;
			}
		}
	}

	data = kmem_alloc(MAXPATHLEN + 1, KM_SLEEP);

	iov.iov_base = data;
	iov.iov_len = MAXPATHLEN;
	uio.uio_iov = &iov;
	uio.uio_iovcnt = 1;
	uio.uio_segflg = UIO_SYSSPACE;
	uio.uio_extflg = UIO_COPY_CACHED;
	uio.uio_loffset = 0;
	uio.uio_resid = MAXPATHLEN;

	error = VOP_READLINK(vp, &uio, cr, NULL);

#ifdef DEBUG
	if (rfs3_do_post_op_attr) {
		va.va_mask = AT_ALL;
		vap = VOP_GETATTR(vp, &va, 0, cr, NULL) ? NULL : &va;
	} else
		vap = NULL;
#else
	va.va_mask = AT_ALL;
	vap = VOP_GETATTR(vp, &va, 0, cr, NULL) ? NULL : &va;
#endif

#if 0 /* notyet */
	/*
	 * Don't do this.  It causes local disk writes when just
	 * reading the file and the overhead is deemed larger
	 * than the benefit.
	 */
	/*
	 * Force modified metadata out to stable storage.
	 */
	(void) VOP_FSYNC(vp, FNODSYNC, cr, NULL);
#endif

	if (error) {
		kmem_free(data, MAXPATHLEN + 1);
		goto out;
	}

	resp->status = NFS3_OK;
	vattr_to_post_op_attr(vap, &resp->resok.symlink_attributes);
	resp->resok.data = data;
	*(data + MAXPATHLEN - uio.uio_resid) = '\0';

	DTRACE_NFSV3_4(op__readlink__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, READLINK3res *, resp);
	VN_RELE(vp);

	return;

out:
	if (curthread->t_flag & T_WOULDBLOCK) {
		curthread->t_flag &= ~T_WOULDBLOCK;
		resp->status = NFS3ERR_JUKEBOX;
	} else
		resp->status = puterrno3(error);
out1:
	DTRACE_NFSV3_4(op__readlink__done, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, READLINK3res *, resp);
	if (vp != NULL)
		VN_RELE(vp);
	vattr_to_post_op_attr(vap, &resp->resfail.symlink_attributes);
}

void *
rfs3_readlink_getfh(READLINK3args *args)
{

	return (&args->symlink);
}

void
rfs3_readlink_free(READLINK3res *resp)
{

	if (resp->status == NFS3_OK)
		kmem_free(resp->resok.data, MAXPATHLEN + 1);
}

/*
 * Server routine to handle read
 * May handle RDMA data as well as mblks
 */
/* ARGSUSED */
void
rfs3_read(READ3args *args, READ3res *resp, struct exportinfo *exi,
	struct svc_req *req, cred_t *cr)
{
	int error;
	vnode_t *vp;
	struct vattr *vap;
	struct vattr va;
	struct iovec iov;
	struct uio uio;
	u_offset_t offset;
	mblk_t *mp;
	int alloc_err = 0;
	int in_crit = 0;
	int need_rwunlock = 0;
	caller_context_t ct;

	vap = NULL;

	vp = nfs3_fhtovp(&args->file, exi);

	DTRACE_NFSV3_4(op__read__start, struct svc_req *, req,
	    cred_t *, cr, vnode_t *, vp, READ3args *, args);

	if (vp == NULL) {
		error = ESTALE;
		goto out;
	}

	if (is_system_labeled()) {
		bslabel_t *clabel = req->rq_label;

		ASSERT(clabel != NULL);
		DTRACE_PROBE2(tx__rfs3__log__info__opread__clabel, char *,
		    "got client label from request(1)", struct svc_req *, req);

		if (!blequal(&l_admin_low->tsl_label, clabel)) {
			if (!do_rfs_label_check(clabel, vp, DOMINANCE_CHECK)) {
				resp->status = NFS3ERR_ACCES;
				goto out1;
			}
		}
	}

	ct.cc_sysid = 0;
	ct.cc_pid = 0;
	ct.cc_caller_id = nfs3_srv_caller_id;
	ct.cc_flags = CC_DONTBLOCK;

	/*
	 * Enter the critical region before calling VOP_RWLOCK
	 * to avoid a deadlock with write requests.
	 */
	if (nbl_need_check(vp)) {
		nbl_start_crit(vp, RW_READER);
		in_crit = 1;
		if (nbl_conflict(vp, NBL_READ, args->offset, args->count, 0,
		    NULL)) {
			error = EACCES;
			goto out;
		}
	}

	error = VOP_RWLOCK(vp, V_WRITELOCK_FALSE, &ct);

	/* check if a monitor detected a delegation conflict */
	if (error == EAGAIN && (ct.cc_flags & CC_WOULDBLOCK)) {
		resp->status = NFS3ERR_JUKEBOX;
		goto out1;
	}

	need_rwunlock = 1;

	va.va_mask = AT_ALL;
	error = VOP_GETATTR(vp, &va, 0, cr, &ct);

	/*
	 * If we can't get the attributes, then we can't do the
	 * right access checking.  So, we'll fail the request.
	 */
	if (error)
		goto out;

#ifdef DEBUG
	if (rfs3_do_post_op_attr)
		vap = &va;
#else
	vap = &va;
#endif

	if (vp->v_type != VREG) {
		resp->status = NFS3ERR_INVAL;
		goto out1;
	}