vnode.c

Solaris操作系统下的过滤驱动程序, C源码程序.

/*
 * Copyright (c) 1997-2003 Erez Zadok
 * Copyright (c) 2001-2003 Stony Brook University
 * Copyright (c) 1997-2000 Columbia University
 *
 * For specific licensing information, see the COPYING file distributed with
 * this package, or get one from ftp://ftp.filesystems.org/pub/fist/COPYING.
 *
 * This Copyright notice must be kept intact and distributed with all
 * fistgen sources INCLUDING sources generated by fistgen.
 */
/*
 * 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.
 *
 *      @(#)wrapfs_vnops.c      8.6 (Berkeley) 5/27/95
 *
 * Ancestors:
 *      @(#)lofs_vnops.c        1.2 (Berkeley) 6/18/92
 *      $Id: vnode.c,v 1.4 2002/12/27 20:18:52 ezk Exp $
 *      ...and...
 *      @(#)wrapfs_vnodeops.c 1.20 92/07/07 UCLA Ficus project
 *
 * $Id: vnode.c,v 1.4 2002/12/27 20:18:52 ezk Exp $
 */

/*
 * wrapfs Layer
 *
 * (See mount_wrapfs(8) for more information.)
 *
 * The wrapfs 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 its "wrapfs-node"s stack above
 * all lower-layer vnodes, not just over directory vnodes.
 *
 * The wrapfs 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 wrapfs 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 wrapfs layer.
 *
 * The remainder of this man page examines the wrapfs layer as a basis
 * for constructing new layers.
 *
 *
 * INSTANTIATING NEW WRAPFS LAYERS
 *
 * New wrapfs layers are created with mount_wrapfs(8).
 * Mount_wrapfs(8) takes two arguments, the pathname
 * of the lower vfs (target-pn) and the pathname where the wrapfs
 * layer will appear in the namespace (alias-pn).  After
 * the wrapfs layer is put into place, the contents
 * of target-pn subtree will be aliased under alias-pn.
 *
 *
 * OPERATION OF A WRAPFS LAYER
 *
 * The wrapfs 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, through 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 wrapfs-nodes by their
 * lower-layer equivlants.  It then invokes the operation
 * on the lower layer.  Finally, it replaces the wrapfs-nodes
 * in the arguments and, if a vnode is return by the operation,
 * stacks a wrapfs-node on top of the returned vnode.
 *
 * Although bypass handles most operations, vop_getattr, vop_lock,
 * vop_unlock, vop_inactive, vop_reclaim, and vop_print are not
 * bypassed. Vop_getattr must change the fsid being returned.
 * Vop_lock and vop_unlock must handle any locking for the
 * current vnode as well as pass the lock request down.
 * Vop_inactive and vop_reclaim are not bypassed so that
 * they can handle freeing wrapfs-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 wrapfs 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 wrapfs layer.  All other vnode stacks
 * are created as a result of vnode operations on
 * this or other wrapfs vnode stacks.
 *
 * New vnode stacks come into existance as a result of
 * an operation which returns a vnode.
 * The bypass routine stacks a wrapfs-node above the new
 * vnode before returning it to the caller.
 *
 * For example, imagine mounting a wrapfs layer with
 * "mount_wrapfs /usr/include /dev/layer/wrapfs".
 * Changing directory to /dev/layer/wrapfs will assign
 * the root wrapfs-node (which was created when the wrapfs layer was mounted).
 * Now consider opening "sys".  A vop_lookup would be
 * done on the root wrapfs-node.  This operation would bypass through
 * to the lower layer which would return a vnode representing
 * the UFS "sys".  wrapfs_bypass then builds a wrapfs-node
 * aliasing the UFS "sys" and returns this to the caller.
 * Later operations on the wrapfs-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 wrapfs 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
 * wrapfs 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 handled on the lower layer.  It has the advantage
 * that the bypass routine already must do argument mapping.
 * An example of this is wrapfs_getattrs in the wrapfs layer.
 *
 * A second approach is to directly invoke 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 vnode arguments must be manualy mapped.
 *
 */

#ifdef HAVE_CONFIG_H
# include <config.h>
#endif /* HAVE_CONFIG_H */
#ifdef FISTGEN
# include "fist_wrapfs.h"
#endif /* FISTGEN */
#include <fist.h>
#include <wrapfs.h>

#if 0
#include "opt_debug.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <sys/namei.h>
#include <sys/malloc.h>
#include <sys/buf.h>
#include <sys/proc.h>
#include <sys/ioccom.h>
#include <sys/dirent.h>

#include <vm/vm.h>
#include <vm/vm_extern.h>
#include <vm/vm_zone.h>
#include <vm/vnode_pager.h>
#include <vm/vm_prot.h>
#include <vm/vm_page.h>
#include <vm/vm_pageout.h>
#include <vm/vm_object.h>
#include <vm/vm_pager.h>

#include <machine/cputypes.h>

#include <wrapfs.h>
#endif

#define SAFETY
// #define HAVE_BMAP
#define DO_WLOCK
#define WRITE_SYNC


static int wrapfs_bug_bypass = 0;	/* for debugging: enables bypass
					 * printf'ing */
SYSCTL_INT(_debug, OID_AUTO, wrapfs_bug_bypass, CTLFLAG_RW,
	   &wrapfs_bug_bypass, 0, "");

static int wrapfs_access __P((struct vop_access_args * ap));
static int wrapfs_bwrite __P((struct vop_bwrite_args * ap));
static int wrapfs_getattr __P((struct vop_getattr_args * ap));
static int wrapfs_inactive __P((struct vop_inactive_args * ap));
static int wrapfs_lock __P((struct vop_lock_args * ap));
static int wrapfs_lookup __P((struct vop_lookup_args * ap));
static int wrapfs_print __P((struct vop_print_args * ap));
static int wrapfs_reclaim __P((struct vop_reclaim_args * ap));
static int wrapfs_setattr __P((struct vop_setattr_args * ap));
static int wrapfs_strategy __P((struct vop_strategy_args * ap));
static int wrapfs_unlock __P((struct vop_unlock_args * ap));

/*
 * 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 vrele'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 vop_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
wrapfs_bypass(ap)
     struct vop_generic_args /* {
				struct vnodeop_desc *a_desc;
				<other random data follows, presumably>
				} */ *ap;
{
  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;

  fist_dprint(4, "FXN=%s FILE=%s LINE=%d FOR=%s\n",
	      __FUNCTION__,__FILE__,__LINE__,descp->vdesc_name);

  if (wrapfs_bug_bypass)
    fist_dprint(5, "wrapfs_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("wrapfs_bypass: no vp's in map.");
#endif /* not SAFETY */

  /*
   * 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 == NULLVP ||
	      (*this_vp_p)->v_op != wrapfs_vnodeop_p)) {
      old_vps[i] = NULLVP;
    } else {
      old_vps[i] = *this_vp_p;
      *(vps_p[i]) = WRAPFS_VP_TO_LOWERVP(*this_vp_p);
      /*
       * XXX - Several operations have the side effect of vrele'ing their
       * vp's.  We must account for that.  (This should go away in the
       * future.)
       */
      if (reles & 1)
	VREF(*this_vp_p);
    }
  } /* end of "for (i = 0;" loop */

  /*
   * 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)
	vrele(*(vps_p[i]));
    }
  }

  /*
   * Map the possible out-going vpp (Assumes that the lower layer always
   * returns a VREF'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 vrele 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);
    if (*vppp)
      error = wrapfs_node_create(old_vps[0]->v_mount, **vppp, *vppp);
  }
out:
  return (error);
}

/*
 * We have to carry on the locking protocol on the wrapfs layer vnodes
 * as we progress through the tree. We also have to enforce read-only
 * if this layer is mounted read-only.
 */
static int
wrapfs_lookup(ap)
     struct vop_lookup_args /* {
			       struct vnode * a_dvp;
			       struct vnode ** a_vpp;
			       struct componentname * a_cnp;
			       } */ *ap;
{
  struct componentname *cnp = ap->a_cnp;
  char *name = ap->a_cnp->cn_nameptr;
  struct proc *p = cnp->cn_proc;
  cred_t *cr = curproc->p_cred->pc_ucred;
  int flags = cnp->cn_flags;
  struct vop_lock_args lockargs;
  struct vop_unlock_args unlockargs;
  struct vnode *dvp, *vp;
  vnode_t *this_dir = ap->a_dvp;
  vnode_t *this_vnode = ap->a_dvp;
  int error;
  static int counter;
  CNP_VARS;

  fist_dprint(2, "FXN=%s FILE=%s LINE=%d\n",__FUNCTION__,__FILE__,__LINE__);

  FIST_OP_LOOKUP_PRECALL;

  if ((flags & ISLASTCN) && (ap->a_dvp->v_mount->mnt_flag & MNT_RDONLY) &&
      (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
    return (EROFS);

  // fist_print_cn("LOOKUP0", ap->a_cnp);

  counter++;
  fist_dprint(1, "LK0 %d (dvp=0x%x)\n", counter, ap->a_dvp);
  CNP_BEFORE(ap->a_dvp);
  // fist_print_cn("LOOKUP1", ap->a_cnp);
  error = wrapfs_bypass((struct vop_generic_args *) ap);

  FIST_OP_LOOKUP_POSTCALL;

#if 0
  if ((thiscnp->cn_flags & PARAMASK) != (lowercnp->cn_flags & PARAMASK)) {
    char buf[256];
    strcpy(buf, fist_cn_flags(thiscnp->cn_flags));
    panic("LK: this 0x%x <%s>, lower 0x%x <%s>",
	  thiscnp->cn_flags,
	  buf,
	  lowercnp->cn_flags,
	  fist_cn_flags(lowercnp->cn_flags));
  }
#endif
#ifdef FIST_FILTER_NAME
  thiscnp->cn_flags = lowercnp->cn_flags;
#endif /* FIST_FILTER_NAME */