[apple/xnu.git] / bsd / miscfs / nullfs / null_vnops.c

/*
 * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_LICENSE_OSREFERENCE_HEADER_START@
 * 
 * This file contains Original Code and/or Modifications of Original Code 
 * as defined in and that are subject to the Apple Public Source License 
 * Version 2.0 (the 'License'). You may not use this file except in 
 * compliance with the License.  The rights granted to you under the 
 * License may not be used to create, or enable the creation or 
 * redistribution of, unlawful or unlicensed copies of an Apple operating 
 * system, or to circumvent, violate, or enable the circumvention or 
 * violation of, any terms of an Apple operating system software license 
 * agreement.
 *
 * Please obtain a copy of the License at 
 * http://www.opensource.apple.com/apsl/ and read it before using this 
 * file.
 *
 * The Original Code and all software distributed under the License are 
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER 
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, 
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, 
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. 
 * Please see the License for the specific language governing rights and 
 * limitations under the License.
 *
 * @APPLE_LICENSE_OSREFERENCE_HEADER_END@
 */
/* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
/*
 * Copyright (c) 1992, 1993
 *      The Regents of the University of California.  All rights reserved.
 *
 * This code is derived from software contributed to Berkeley by
 * John Heidemann of the UCLA Ficus project.
 *
 * 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.
 *
 *      @(#)null_vnops.c        8.6 (Berkeley) 5/27/95
 *
 * Ancestors:
 *      @(#)lofs_vnops.c        1.2 (Berkeley) 6/18/92
 *      ...and...
 *      @(#)null_vnodeops.c 1.20 92/07/07 UCLA Ficus project
 */

/*
 * Null Layer
 *
 * (See mount_null(8) for more information.)
 *
 * The null layer duplicates a portion of the file system
 * name space under a new name.  In this respect, it is
 * similar to the loopback file system.  It differs from
 * the loopback fs in two respects:  it is implemented using
 * a stackable layers techniques, and it's "null-node"s stack above
 * all lower-layer vnodes, not just over directory vnodes.
 *
 * The null layer has two purposes.  First, it serves as a demonstration
 * of layering by proving a layer which does nothing.  (It actually
 * does everything the loopback file system does, which is slightly
 * more than nothing.)  Second, the null layer can serve as a prototype
 * layer.  Since it provides all necessary layer framework,
 * new file system layers can be created very easily be starting
 * with a null layer.
 *
 * The remainder of this man page examines the null layer as a basis
 * for constructing new layers.
 *
 *
 * INSTANTIATING NEW NULL LAYERS
 *
 * New null layers are created with mount_null(8).
 * Mount_null(8) takes two arguments, the pathname
 * of the lower vfs (target-pn) and the pathname where the null
 * layer will appear in the namespace (alias-pn).  After
 * the null layer is put into place, the contents
 * of target-pn subtree will be aliased under alias-pn.
 *
 *
 * OPERATION OF A NULL LAYER
 *
 * The null layer is the minimum file system layer,
 * simply bypassing all possible operations to the lower layer
 * for processing there.  The majority of its activity centers
 * on the bypass routine, though which nearly all vnode operations
 * pass.
 *
 * The bypass routine accepts arbitrary vnode operations for
 * handling by the lower layer.  It begins by examing vnode
 * operation arguments and replacing any null-nodes by their
 * lower-layer equivlants.  It then invokes the operation
 * on the lower layer.  Finally, it replaces the null-nodes
 * in the arguments and, if a vnode is return by the operation,
 * stacks a null-node on top of the returned vnode.
 *
 * Although bypass handles most operations, vnop_getattr, vnop_lock,
 * vnop_unlock, vnop_inactive, vnop_reclaim, and vnop_print are not
 * bypassed. Vop_getattr must change the fsid being returned.
 * Vop_lock and vnop_unlock must handle any locking for the
 * current vnode as well as pass the lock request down.
 * Vop_inactive and vnop_reclaim are not bypassed so that
 * they can handle freeing null-layer specific data. Vop_print
 * is not bypassed to avoid excessive debugging information.
 * Also, certain vnode operations change the locking state within
 * the operation (create, mknod, remove, link, rename, mkdir, rmdir,
 * and symlink). Ideally these operations should not change the
 * lock state, but should be changed to let the caller of the
 * function unlock them. Otherwise all intermediate vnode layers
 * (such as union, umapfs, etc) must catch these functions to do
 * the necessary locking at their layer.
 *
 *
 * INSTANTIATING VNODE STACKS
 *
 * Mounting associates the null layer with a lower layer,
 * effect stacking two VFSes.  Vnode stacks are instead
 * created on demand as files are accessed.
 *
 * The initial mount creates a single vnode stack for the
 * root of the new null layer.  All other vnode stacks
 * are created as a result of vnode operations on
 * this or other null vnode stacks.
 *
 * New vnode stacks come into existance as a result of
 * an operation which returns a vnode.  
 * The bypass routine stacks a null-node above the new
 * vnode before returning it to the caller.
 *
 * For example, imagine mounting a null layer with
 * "mount_null /usr/include /dev/layer/null".
 * Changing directory to /dev/layer/null will assign
 * the root null-node (which was created when the null layer was mounted).
 * Now consider opening "sys".  A vnop_lookup would be
 * done on the root null-node.  This operation would bypass through
 * to the lower layer which would return a vnode representing 
 * the UFS "sys".  Null_bypass then builds a null-node
 * aliasing the UFS "sys" and returns this to the caller.
 * Later operations on the null-node "sys" will repeat this
 * process when constructing other vnode stacks.
 *
 *
 * CREATING OTHER FILE SYSTEM LAYERS
 *
 * One of the easiest ways to construct new file system layers is to make
 * a copy of the null layer, rename all files and variables, and
 * then begin modifing the copy.  Sed can be used to easily rename
 * all variables.
 *
 * The umap layer is an example of a layer descended from the 
 * null layer.
 *
 *
 * INVOKING OPERATIONS ON LOWER LAYERS
 *
 * There are two techniques to invoke operations on a lower layer 
 * when the operation cannot be completely bypassed.  Each method
 * is appropriate in different situations.  In both cases,
 * it is the responsibility of the aliasing layer to make
 * the operation arguments "correct" for the lower layer
 * by mapping an vnode arguments to the lower layer.
 *
 * The first approach is to call the aliasing layer's bypass routine.
 * This method is most suitable when you wish to invoke the operation
 * currently being hanldled on the lower layer.  It has the advantage
 * that the bypass routine already must do argument mapping.
 * An example of this is null_getattrs in the null layer.
 *
 * A second approach is to directly invoked vnode operations on
 * the lower layer with the VOP_OPERATIONNAME interface.
 * The advantage of this method is that it is easy to invoke
 * arbitrary operations on the lower layer.  The disadvantage
 * is that vnodes arguments must be manualy mapped.
 *
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/kauth.h>
#include <sys/time.h>
#include <sys/types.h>
#include <sys/vnode.h>
#include <sys/mount_internal.h>
#include <sys/namei.h>
#include <sys/malloc.h>
#include <sys/buf.h>
#include <miscfs/nullfs/null.h>


int null_bug_bypass = 0;   /* for debugging: enables bypass printf'ing */

/*
 * This is the 10-Apr-92 bypass routine.
 *    This version has been optimized for speed, throwing away some
 * safety checks.  It should still always work, but it's not as
 * robust to programmer errors.
 *    Define SAFETY to include some error checking code.
 *
 * In general, we map all vnodes going down and unmap them on the way back.
 * As an exception to this, vnodes can be marked "unmapped" by setting
 * the Nth bit in operation's vdesc_flags.
 *
 * Also, some BSD vnode operations have the side effect of node_put'ing
 * their arguments.  With stacking, the reference counts are held
 * by the upper node, not the lower one, so we must handle these
 * side-effects here.  This is not of concern in Sun-derived systems
 * since there are no such side-effects.
 *
 * This makes the following assumptions:
 * - only one returned vpp
 * - no INOUT vpp's (Sun's vnop_open has one of these)
 * - the vnode operation vector of the first vnode should be used
 *   to determine what implementation of the op should be invoked
 * - all mapped vnodes are of our vnode-type (NEEDSWORK:
 *   problems on rmdir'ing mount points and renaming?)
 */ 
int
null_bypass(ap)
        struct vnop_generic_args /* {
                struct vnodeop_desc *a_desc;
                <other random data follows, presumably>
        } */ *ap;
{
        extern int (**null_vnodeop_p)(void *);  /* not extern, really "forward" */
        register struct vnode **this_vp_p;
        int error;
        struct vnode *old_vps[VDESC_MAX_VPS];
        struct vnode **vps_p[VDESC_MAX_VPS];
        struct vnode ***vppp;
        struct vnodeop_desc *descp = ap->a_desc;
        int reles, i;

        if (null_bug_bypass)
                printf ("null_bypass: %s\n", descp->vdesc_name);

#ifdef SAFETY
        /*
         * We require at least one vp.
         */
        if (descp->vdesc_vp_offsets == NULL ||
            descp->vdesc_vp_offsets[0] == VDESC_NO_OFFSET)
                panic ("null_bypass: no vp's in map.\n");
#endif

        /*
         * Map the vnodes going in.
         * Later, we'll invoke the operation based on
         * the first mapped vnode's operation vector.
         */
        reles = descp->vdesc_flags;
        for (i = 0; i < VDESC_MAX_VPS; reles >>= 1, i++) {
                if (descp->vdesc_vp_offsets[i] == VDESC_NO_OFFSET)
                        break;   /* bail out at end of list */
                vps_p[i] = this_vp_p = 
                        VOPARG_OFFSETTO(struct vnode**,descp->vdesc_vp_offsets[i],ap);
                /*
                 * We're not guaranteed that any but the first vnode
                 * are of our type.  Check for and don't map any
                 * that aren't.  (We must always map first vp or vclean fails.)
                 */
                if (i && (*this_vp_p == NULL ||
                    (*this_vp_p)->v_op != null_vnodeop_p)) {
                        old_vps[i] = NULL;
                } else {
                        old_vps[i] = *this_vp_p;
                        *(vps_p[i]) = NULLVPTOLOWERVP(*this_vp_p);
                        /*
                         * XXX - Several operations have the side effect
                         * of vnode_put'ing their vp's.  We must account for
                         * that.  (This should go away in the future.)
                         */
                        if (reles & 1)
                                vnode_get(*this_vp_p);
                }
                        
        }

        /*
         * Call the operation on the lower layer
         * with the modified argument structure.
         */
        error = VCALL(*(vps_p[0]), descp->vdesc_offset, ap);

        /*
         * Maintain the illusion of call-by-value
         * by restoring vnodes in the argument structure
         * to their original value.
         */
        reles = descp->vdesc_flags;
        for (i = 0; i < VDESC_MAX_VPS; reles >>= 1, i++) {
                if (descp->vdesc_vp_offsets[i] == VDESC_NO_OFFSET)
                        break;   /* bail out at end of list */
                if (old_vps[i]) {
                        *(vps_p[i]) = old_vps[i];
                        if (reles & 1)
                                vnode_put(*(vps_p[i]));
                }
        }

        /*
         * Map the possible out-going vpp
         * (Assumes that the lower layer always returns
         * a vnode_get'ed vpp unless it gets an error.)
         */
        if (descp->vdesc_vpp_offset != VDESC_NO_OFFSET &&
            !(descp->vdesc_flags & VDESC_NOMAP_VPP) &&
            !error) {
                /*
                 * XXX - even though some ops have vpp returned vp's,
                 * several ops actually vnode_put this before returning.
                 * We must avoid these ops.
                 * (This should go away when these ops are regularized.)
                 */
                if (descp->vdesc_flags & VDESC_VPP_WILLRELE)
                        goto out;
                vppp = VOPARG_OFFSETTO(struct vnode***,
                                 descp->vdesc_vpp_offset,ap);
                error = null_node_create(old_vps[0]->v_mount, **vppp, *vppp);
        }

 out:
        return (error);
}

/*
 * We have to carry on the locking protocol on the null layer vnodes
 * as we progress through the tree. We also have to enforce read-only
 * if this layer is mounted read-only.
 */
null_lookup(ap)
        struct vnop_lookup_args /* {
                struct vnode * a_dvp;
                struct vnode ** a_vpp;
                struct componentname * a_cnp;
                vfs_context_t a_context;
        } */ *ap;
{
        struct componentname *cnp = ap->a_cnp;
        struct proc *p = cnp->cn_proc;
        int flags = cnp->cn_flags;
        struct vnode *dvp, *vp;
        int error;

        error = null_bypass(ap);

        /*
         * We must do the same locking and unlocking at this layer as 
         * is done in the layers below us. We could figure this out 
         * based on the error return and the LASTCN, LOCKPARENT, and
         * LOCKLEAF flags. However, it is more expidient to just find 
         * out the state of the lower level vnodes and set ours to the
         * same state.
         */
        dvp = ap->a_dvp;
        vp = *ap->a_vpp;
        if (dvp == vp)
                return (error);
        return (error);
}

/*
 * Setattr call.
 */
int
null_setattr(
        struct vnop_setattr_args /* {
                struct vnodeop_desc *a_desc;
                struct vnode *a_vp;
                struct vnode_attr *a_vap;
                kauth_cred_t a_cred;
                struct proc *a_p;
        } */ *ap)
{
        struct vnode *vp = ap->a_vp;
        struct vnode_attr *vap = ap->a_vap;

        if (VATTR_IS_ACTIVE(vap, va_data_size)) {
                switch (vp->v_type) {
                case VDIR:
                        return (EISDIR);
                case VCHR:
                case VBLK:
                case VSOCK:
                case VFIFO:
                        return (0);
                case VREG:
                case VLNK:
                default:
                }
        }
        return (null_bypass(ap));
}

/*
 *  We handle getattr only to change the fsid.
 */
int
null_getattr(ap)
        struct vnop_getattr_args /* {
                struct vnode *a_vp;
                struct vnode_attr *a_vap;
                vfs_context_t a_context;
        } */ *ap;
{
        int error;

        if (error = null_bypass(ap))
                return (error);
        /* Requires that arguments be restored. */
        VATTR_RETURN(ap->a_vap, va_fsid, ap->a_vp->v_mount->mnt_vfsstat.f_fsid.val[0]);
        return (0);
}

int
null_access(ap)
        struct vnop_access_args /* {
                struct vnode *a_vp;
                int  a_action;
                vfs_context_t a_context;
        } */ *ap;
{
        return (null_bypass(ap));
}

int
null_inactive(ap)
        struct vnop_inactive_args /* {
                struct vnode *a_vp;
                vfs_context_t a_context;
        } */ *ap;
{
        /*
         * Do nothing (and _don't_ bypass).
         * Wait to vnode_put lowervp until reclaim,
         * so that until then our null_node is in the
         * cache and reusable.
         *
         * NEEDSWORK: Someday, consider inactive'ing
         * the lowervp and then trying to reactivate it
         * with capabilities (v_id)
         * like they do in the name lookup cache code.
         * That's too much work for now.
         */
        return (0);
}

int
null_reclaim(ap)
        struct vnop_reclaim_args /* {
                struct vnode *a_vp;
                vfs_context_t a_context;
        } */ *ap;
{
        struct vnode *vp = ap->a_vp;
        struct null_node *xp = VTONULL(vp);
        struct vnode *lowervp = xp->null_lowervp;

        /*
         * Note: in vnop_reclaim, vp->v_op == dead_vnodeop_p,
         * so we can't call VOPs on ourself.
         */
        /* After this assignment, this node will not be re-used. */
        xp->null_lowervp = NULL;
        LIST_REMOVE(xp, null_hash);
        FREE(vp->v_data, M_TEMP);
        vp->v_data = NULL;
        vnode_put (lowervp);
        return (0);
}

/*
 * XXX - vnop_strategy must be hand coded because it has no
 * vnode in its arguments.
 * This goes away with a merged VM/buffer cache.
 */
int
null_strategy(ap)
        struct vnop_strategy_args /* {
                struct buf *a_bp;
        } */ *ap;
{
        struct buf *bp = ap->a_bp;
        int error;
        struct vnode *savedvp;

        savedvp = vnode(bp);
        buf_setvnode(bp, NULLVPTOLOWERVP(savedvp));

        error = VNOP_STRATEGY(bp);

        buf_setvnode(bp, savedvp);

        return (error);
}

/*
 * XXX - like vnop_strategy, vnop_bwrite must be hand coded because it has no
 * vnode in its arguments.
 * This goes away with a merged VM/buffer cache.
 */
int
null_bwrite(ap)
        struct vnop_bwrite_args /* {
                struct buf *a_bp;
        } */ *ap;
{
        struct buf *bp = ap->a_bp;
        int error;
        struct vnode *savedvp;

        savedvp = buf_vnode(bp);
        buf_setvnode(bp, NULLVPTOLOWERVP(savedvp));

        error = VNOP_BWRITE(bp);

        buf_setvnode(bp, savedvp);

        return (error);
}

/*
 * Global vfs data structures
 */

#define VOPFUNC int (*)(void *)

int (**null_vnodeop_p)(void *);
struct vnodeopv_entry_desc null_vnodeop_entries[] = {
        { &vnop_default_desc, (VOPFUNC)null_bypass },

        { &vnop_lookup_desc, (VOPFUNC)null_lookup },
        { &vnop_setattr_desc, (VOPFUNC)null_setattr },
        { &vnop_getattr_desc, (VOPFUNC)null_getattr },
        { &vnop_access_desc, (VOPFUNC)null_access },
        { &vnop_inactive_desc, (VOPFUNC)null_inactive },
        { &vnop_reclaim_desc, (VOPFUNC)null_reclaim },

        { &vnop_strategy_desc, (VOPFUNC)null_strategy },
        { &vnop_bwrite_desc, (VOPFUNC)null_bwrite },

        { (struct vnodeop_desc*)NULL, (int(*)())NULL }
};
struct vnodeopv_desc null_vnodeop_opv_desc =
        { &null_vnodeop_p, null_vnodeop_entries };