union_vnops.c

来自「早期freebsd实现」· C语言 代码 · 共 1,496 行 · 第 1/3 页

/*
 * Copyright (c) 1992, 1993, 1994 The Regents of the University of California.
 * Copyright (c) 1992, 1993, 1994 Jan-Simon Pendry.
 * All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * Jan-Simon Pendry.
 *
 * 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. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. 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.
 *
 *	@(#)union_vnops.c	8.6 (Berkeley) 2/17/94
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/file.h>
#include <sys/time.h>
#include <sys/types.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <sys/namei.h>
#include <sys/malloc.h>
#include <sys/buf.h>
#include <sys/queue.h>
#include <miscfs/union/union.h>

#define FIXUP(un) { \
	if (((un)->un_flags & UN_ULOCK) == 0) { \
		union_fixup(un); \
	} \
}

static void
union_fixup(un)
	struct union_node *un;
{

	VOP_LOCK(un->un_uppervp);
	un->un_flags |= UN_ULOCK;
}

static int
union_lookup1(udvp, dvp, vpp, cnp)
	struct vnode *udvp;
	struct vnode *dvp;
	struct vnode **vpp;
	struct componentname *cnp;
{
	int error;
	struct vnode *tdvp;
	struct mount *mp;

	/*
	 * If stepping up the directory tree, check for going
	 * back across the mount point, in which case do what
	 * lookup would do by stepping back down the mount
	 * hierarchy.
	 */
	if (cnp->cn_flags & ISDOTDOT) {
		for (;;) {
			/*
			 * Don't do the NOCROSSMOUNT check
			 * at this level.  By definition,
			 * union fs deals with namespaces, not
			 * filesystems.
			 */
			if ((dvp->v_flag & VROOT) == 0)
				break;

			tdvp = dvp;
			dvp = dvp->v_mount->mnt_vnodecovered;
			vput(tdvp);
			VREF(dvp);
			VOP_LOCK(dvp);
		}
	}

        error = VOP_LOOKUP(dvp, &tdvp, cnp);
	if (error)
		return (error);

	/*
	 * The parent directory will have been unlocked, unless lookup
	 * found the last component.  In which case, re-lock the node
	 * here to allow it to be unlocked again (phew) in union_lookup.
	 */
	if (dvp != tdvp && !(cnp->cn_flags & ISLASTCN))
		VOP_LOCK(dvp);

	dvp = tdvp;

	/*
	 * Lastly check if the current node is a mount point in
	 * which case walk up the mount hierarchy making sure not to
	 * bump into the root of the mount tree (ie. dvp != udvp).
	 */
	while (dvp != udvp && (dvp->v_type == VDIR) &&
	       (mp = dvp->v_mountedhere)) {

		if (mp->mnt_flag & MNT_MLOCK) {
			mp->mnt_flag |= MNT_MWAIT;
			sleep((caddr_t) mp, PVFS);
			continue;
		}

		if (error = VFS_ROOT(mp, &tdvp)) {
			vput(dvp);
			return (error);
		}

		vput(dvp);
		dvp = tdvp;
	}

	*vpp = dvp;
	return (0);
}

int
union_lookup(ap)
	struct vop_lookup_args /* {
		struct vnodeop_desc *a_desc;
		struct vnode *a_dvp;
		struct vnode **a_vpp;
		struct componentname *a_cnp;
	} */ *ap;
{
	int error;
	int uerror, lerror;
	struct vnode *uppervp, *lowervp;
	struct vnode *upperdvp, *lowerdvp;
	struct vnode *dvp = ap->a_dvp;
	struct union_node *dun = VTOUNION(dvp);
	struct componentname *cnp = ap->a_cnp;
	int lockparent = cnp->cn_flags & LOCKPARENT;
	int rdonly = cnp->cn_flags & RDONLY;
	struct union_mount *um = MOUNTTOUNIONMOUNT(dvp->v_mount);
	struct ucred *saved_cred;

	cnp->cn_flags |= LOCKPARENT;

	upperdvp = dun->un_uppervp;
	lowerdvp = dun->un_lowervp;
	uppervp = NULLVP;
	lowervp = NULLVP;

	/*
	 * do the lookup in the upper level.
	 * if that level comsumes additional pathnames,
	 * then assume that something special is going
	 * on and just return that vnode.
	 */
	if (upperdvp) {
		FIXUP(dun);
		uerror = union_lookup1(um->um_uppervp, upperdvp,
					&uppervp, cnp);
		/*if (uppervp == upperdvp)
			dun->un_flags |= UN_KLOCK;*/

		if (cnp->cn_consume != 0) {
			*ap->a_vpp = uppervp;
			if (!lockparent)
				cnp->cn_flags &= ~LOCKPARENT;
			return (uerror);
		}
	} else {
		uerror = ENOENT;
	}

	/*
	 * in a similar way to the upper layer, do the lookup
	 * in the lower layer.   this time, if there is some
	 * component magic going on, then vput whatever we got
	 * back from the upper layer and return the lower vnode
	 * instead.
	 */
	if (lowerdvp) {
		int nameiop;

		VOP_LOCK(lowerdvp);

		/*
		 * Only do a LOOKUP on the bottom node, since
		 * we won't be making changes to it anyway.
		 */
		nameiop = cnp->cn_nameiop;
		cnp->cn_nameiop = LOOKUP;
		if (um->um_op == UNMNT_BELOW) {
			saved_cred = cnp->cn_cred;
			cnp->cn_cred = um->um_cred;
		}
		lerror = union_lookup1(um->um_lowervp, lowerdvp,
				&lowervp, cnp);
		if (um->um_op == UNMNT_BELOW)
			cnp->cn_cred = saved_cred;
		cnp->cn_nameiop = nameiop;

		if (lowervp != lowerdvp)
			VOP_UNLOCK(lowerdvp);

		if (cnp->cn_consume != 0) {
			if (uppervp) {
				if (uppervp == upperdvp)
					vrele(uppervp);
				else
					vput(uppervp);
				uppervp = NULLVP;
			}
			*ap->a_vpp = lowervp;
			if (!lockparent)
				cnp->cn_flags &= ~LOCKPARENT;
			return (lerror);
		}
	} else {
		lerror = ENOENT;
	}

	if (!lockparent)
		cnp->cn_flags &= ~LOCKPARENT;

	/*
	 * at this point, we have uerror and lerror indicating
	 * possible errors with the lookups in the upper and lower
	 * layers.  additionally, uppervp and lowervp are (locked)
	 * references to existing vnodes in the upper and lower layers.
	 *
	 * there are now three cases to consider.
	 * 1. if both layers returned an error, then return whatever
	 *    error the upper layer generated.
	 *
	 * 2. if the top layer failed and the bottom layer succeeded
	 *    then two subcases occur.
	 *    a.  the bottom vnode is not a directory, in which
	 *	  case just return a new union vnode referencing
	 *	  an empty top layer and the existing bottom layer.
	 *    b.  the bottom vnode is a directory, in which case
	 *	  create a new directory in the top-level and
	 *	  continue as in case 3.
	 *
	 * 3. if the top layer succeeded then return a new union
	 *    vnode referencing whatever the new top layer and
	 *    whatever the bottom layer returned.
	 */

	*ap->a_vpp = NULLVP;

	/* case 1. */
	if ((uerror != 0) && (lerror != 0)) {
		return (uerror);
	}

	/* case 2. */
	if (uerror != 0 /* && (lerror == 0) */ ) {
		if (lowervp->v_type == VDIR) { /* case 2b. */
			dun->un_flags &= ~UN_ULOCK;
			VOP_UNLOCK(upperdvp);
			uerror = union_mkshadow(um, upperdvp, cnp, &uppervp);
			VOP_LOCK(upperdvp);
			dun->un_flags |= UN_ULOCK;

			if (uerror) {
				if (lowervp) {
					vput(lowervp);
					lowervp = NULLVP;
				}
				return (uerror);
			}
		}
	}

	if (lowervp)
		VOP_UNLOCK(lowervp);

	error = union_allocvp(ap->a_vpp, dvp->v_mount, dvp, upperdvp, cnp,
			      uppervp, lowervp);

	if (error) {
		if (uppervp)
			vput(uppervp);
		if (lowervp)
			vrele(lowervp);
	} else {
		if (*ap->a_vpp != dvp)
			if (!lockparent || !(cnp->cn_flags & ISLASTCN))
				VOP_UNLOCK(dvp);
	}

	return (error);
}

int
union_create(ap)
	struct vop_create_args /* {
		struct vnode *a_dvp;
		struct vnode **a_vpp;
		struct componentname *a_cnp;
		struct vattr *a_vap;
	} */ *ap;
{
	struct union_node *un = VTOUNION(ap->a_dvp);
	struct vnode *dvp = un->un_uppervp;

	if (dvp) {
		int error;
		struct vnode *vp;

		FIXUP(un);

		VREF(dvp);
		un->un_flags |= UN_KLOCK;
		vput(ap->a_dvp);
		error = VOP_CREATE(dvp, &vp, ap->a_cnp, ap->a_vap);
		if (error)
			return (error);

		error = union_allocvp(
				ap->a_vpp,
				ap->a_dvp->v_mount,
				ap->a_dvp,
				NULLVP,
				ap->a_cnp,
				vp,
				NULLVP);
		if (error)
			vput(vp);
		return (error);
	}

	vput(ap->a_dvp);
	return (EROFS);
}

int
union_mknod(ap)
	struct vop_mknod_args /* {
		struct vnode *a_dvp;
		struct vnode **a_vpp;
		struct componentname *a_cnp;
		struct vattr *a_vap;
	} */ *ap;
{
	struct union_node *un = VTOUNION(ap->a_dvp);
	struct vnode *dvp = un->un_uppervp;

	if (dvp) {
		int error;
		struct vnode *vp;

		FIXUP(un);

		VREF(dvp);
		un->un_flags |= UN_KLOCK;
		vput(ap->a_dvp);
		error = VOP_MKNOD(dvp, &vp, ap->a_cnp, ap->a_vap);
		if (error)
			return (error);

		if (vp) {
			error = union_allocvp(
					ap->a_vpp,
					ap->a_dvp->v_mount,
					ap->a_dvp,
					NULLVP,
					ap->a_cnp,
					vp,
					NULLVP);
			if (error)
				vput(vp);
		}
		return (error);
	}

	vput(ap->a_dvp);
	return (EROFS);
}

int
union_open(ap)
	struct vop_open_args /* {
		struct vnodeop_desc *a_desc;
		struct vnode *a_vp;
		int a_mode;
		struct ucred *a_cred;
		struct proc *a_p;
	} */ *ap;
{
	struct union_node *un = VTOUNION(ap->a_vp);
	struct vnode *tvp;
	int mode = ap->a_mode;
	struct ucred *cred = ap->a_cred;
	struct proc *p = ap->a_p;
	int error;

	/*
	 * If there is an existing upper vp then simply open that.
	 */
	tvp = un->un_uppervp;
	if (tvp == NULLVP) {
		/*
		 * If the lower vnode is being opened for writing, then
		 * copy the file contents to the upper vnode and open that,
		 * otherwise can simply open the lower vnode.
		 */
		tvp = un->un_lowervp;
		if ((ap->a_mode & FWRITE) && (tvp->v_type == VREG)) {
			struct vnode *vp;
			int i;

			/*
			 * Open the named file in the upper layer.  Note that
			 * the file may have come into existence *since* the
			 * lookup was done, since the upper layer may really
			 * be a loopback mount of some other filesystem...
			 * so open the file with exclusive create and barf if
			 * it already exists.
			 * XXX - perhaps should re-lookup the node (once more
			 * with feeling) and simply open that.  Who knows.
			 */
			error = union_vn_create(&vp, un, p);
			if (error)
				return (error);

			/* at this point, uppervp is locked */
			union_newupper(un, vp);
			un->un_flags |= UN_ULOCK;

			/*
			 * Now, if the file is being opened with truncation,
			 * then the (new) upper vnode is ready to fly,
			 * otherwise the data from the lower vnode must be
			 * copied to the upper layer first.  This only works
			 * for regular files (check is made above).
			 */
			if ((mode & O_TRUNC) == 0) {
				/*
				 * XXX - should not ignore errors
				 * from VOP_CLOSE
				 */
				VOP_LOCK(tvp);
				error = VOP_OPEN(tvp, FREAD, cred, p);
				if (error == 0) {
					error = union_copyfile(p, cred,
						       tvp, un->un_uppervp);
					VOP_UNLOCK(tvp);
					(void) VOP_CLOSE(tvp, FREAD);
				} else {
					VOP_UNLOCK(tvp);
				}

#ifdef UNION_DIAGNOSTIC
				if (!error)
					uprintf("union: copied up %s\n",
								un->un_path);
#endif
			}

			un->un_flags &= ~UN_ULOCK;
			VOP_UNLOCK(un->un_uppervp);
			union_vn_close(un->un_uppervp, FWRITE, cred, p);
			VOP_LOCK(un->un_uppervp);
			un->un_flags |= UN_ULOCK;

			/*
			 * Subsequent IOs will go to the top layer, so
			 * call close on the lower vnode and open on the
			 * upper vnode to ensure that the filesystem keeps
			 * its references counts right.  This doesn't do
			 * the right thing with (cred) and (FREAD) though.
			 * Ignoring error returns is not righ, either.
			 */
			for (i = 0; i < un->un_openl; i++) {
				(void) VOP_CLOSE(tvp, FREAD);