[dragonfly.git] / sys / kern / vfs_subr.c

/*
 * Copyright (c) 1989, 1993
 *      The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * 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. 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.
 *
 *      @(#)vfs_subr.c  8.31 (Berkeley) 5/26/95
 * $FreeBSD: src/sys/kern/vfs_subr.c,v 1.249.2.30 2003/04/04 20:35:57 tegge Exp $
 */

/*
 * External virtual filesystem routines
 */
#include "opt_ddb.h"
#include "opt_inet.h"
#include "opt_inet6.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/uio.h>
#include <sys/buf.h>
#include <sys/conf.h>
#include <sys/dirent.h>
#include <sys/endian.h>
#include <sys/eventhandler.h>
#include <sys/fcntl.h>
#include <sys/file.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/mount.h>
#include <sys/caps.h>
#include <sys/proc.h>
#include <sys/reboot.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/sysctl.h>
#include <sys/syslog.h>
#include <sys/unistd.h>
#include <sys/vmmeter.h>
#include <sys/vnode.h>

#include <machine/limits.h>

#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <vm/vm_kern.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
#include <vm/vm_page.h>
#include <vm/vm_pager.h>
#include <vm/vnode_pager.h>
#include <vm/vm_zone.h>

#include <sys/buf2.h>
#include <vm/vm_page2.h>

#include <netinet/in.h>

static MALLOC_DEFINE(M_NETCRED, "Export Host", "Export host address structure");

__read_mostly int numvnodes;
SYSCTL_INT(_debug, OID_AUTO, numvnodes, CTLFLAG_RD, &numvnodes, 0,
    "Number of vnodes allocated");
__read_mostly int verbose_reclaims;
SYSCTL_INT(_debug, OID_AUTO, verbose_reclaims, CTLFLAG_RD, &verbose_reclaims, 0,
    "Output filename of reclaimed vnode(s)");

__read_mostly enum vtype iftovt_tab[16] = {
        VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON,
        VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD,
};
__read_mostly int vttoif_tab[9] = {
        0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK,
        S_IFSOCK, S_IFIFO, S_IFMT,
};

static int reassignbufcalls;
SYSCTL_INT(_vfs, OID_AUTO, reassignbufcalls, CTLFLAG_RW, &reassignbufcalls,
    0, "Number of times buffers have been reassigned to the proper list");

__read_mostly static int check_buf_overlap = 2; /* invasive check */
SYSCTL_INT(_vfs, OID_AUTO, check_buf_overlap, CTLFLAG_RW, &check_buf_overlap,
    0, "Enable overlapping buffer checks");

int     nfs_mount_type = -1;
static struct lwkt_token spechash_token;
struct nfs_public nfs_pub;      /* publicly exported FS */

__read_mostly int maxvnodes;
SYSCTL_INT(_kern, KERN_MAXVNODES, maxvnodes, CTLFLAG_RW,
           &maxvnodes, 0, "Maximum number of vnodes");

static struct radix_node_head *vfs_create_addrlist_af(int af,
                    struct netexport *nep);
static void vclean_vxlocked(struct vnode *vp, int flags);

__read_mostly int prtactive = 0; /* 1 => print out reclaim of active vnodes */

/*
 * Red black tree functions
 */
static int rb_buf_compare(struct buf *b1, struct buf *b2);
RB_GENERATE2(buf_rb_tree, buf, b_rbnode, rb_buf_compare, off_t, b_loffset);
RB_GENERATE2(buf_rb_hash, buf, b_rbhash, rb_buf_compare, off_t, b_loffset);

static int
rb_buf_compare(struct buf *b1, struct buf *b2)
{
        if (b1->b_loffset < b2->b_loffset)
                return(-1);
        if (b1->b_loffset > b2->b_loffset)
                return(1);
        return(0);
}

/*
 * Initialize the vnode management data structures.
 *
 * Called from vfsinit()
 */
#define VNBREAKMEM1     (1L * 1024 * 1024 * 1024)
#define VNBREAKMEM2     (7L * 1024 * 1024 * 1024)
#define MINVNODES       2000
#define MAXVNODES       4000000

void
vfs_subr_init(void)
{
        int factor1;    /* Limit based on ram (x 2 above 1GB) */
        size_t freemem;

        /*
         * Size maxvnodes non-linearly to available memory.  Don't bloat
         * the count on low-memory systems.  Scale up for systems with
         * more than 1G and more than 8G of ram, but do so non-linearly
         * because the value of a large maxvnodes count diminishes
         * significantly beyond a certain point.
         *
         * The general minimum is maxproc * 8 (we want someone pushing
         * up maxproc a lot to also get more vnodes).  Usually maxproc
         * does not affect this calculation.  The KvaSize limitation also
         * typically does not affect this calculation (it is just in case
         * the kernel VM space is made much smaller than main memory, which
         * should no longer happen on 64-bit systems).
         *
         * There isn't much of a point allowing maxvnodes to exceed a
         * few million as modern filesystems cache pages in the
         * underlying block device and not so much hanging off of VM
         * objects.
         *
         * Also, VM objects, vnodes, and filesystem inode and other related
         * structures have gotten a lot larger in recent years and the kernel
         * memory use tends to scale with maxvnodes, so we don't want to bloat
         * it too much.  But neither do we want the max set too low because
         * systems with large amounts of memory and cores are capable of
         * doing a hell of a lot.
         */
        factor1 = 80 * (sizeof(struct vm_object) + sizeof(struct vnode));

        freemem = (int64_t)vmstats.v_page_count * PAGE_SIZE;

        maxvnodes = freemem / factor1;
        if (freemem > VNBREAKMEM1) {
                freemem -= VNBREAKMEM1;
                if (freemem < VNBREAKMEM2) {
                        maxvnodes += freemem / factor1 / 2;
                } else {
                        maxvnodes += VNBREAKMEM2 / factor1 / 2;
                        freemem -= VNBREAKMEM2;
                        maxvnodes += freemem / factor1 / 4;
                }
        }
        maxvnodes = imax(maxvnodes, maxproc * 8);
        maxvnodes = imin(maxvnodes, KvaSize / factor1);
        maxvnodes = imin(maxvnodes, MAXVNODES);
        maxvnodes = imax(maxvnodes, MINVNODES);

        lwkt_token_init(&spechash_token, "spechash");
}

/*
 * Knob to control the precision of file timestamps:
 *
 *   0 = seconds only; nanoseconds zeroed.
 *   1 = microseconds accurate to tick precision
 *   2 = microseconds accurate to tick precision        (default, hz >= 100)
 *   3 = nanoseconds accurate to tick precision
 *   4 = microseconds, maximum precision                (default, hz < 100)
 *   5 = nanoseconds, maximum precision
 *
 * Note that utimes() precision is microseconds because it takes a timeval
 * structure, so its probably best to default to USEC or USEC_PRECISE, and
 * not NSEC.
 */
enum { TSP_SEC, TSP_HZ, TSP_USEC, TSP_NSEC,
       TSP_USEC_PRECISE, TSP_NSEC_PRECISE };

__read_mostly static int timestamp_precision = -1;
SYSCTL_INT(_vfs, OID_AUTO, timestamp_precision, CTLFLAG_RW,
                &timestamp_precision, 0, "Precision of file timestamps");

/*
 * Get a current timestamp.
 *
 * MPSAFE
 */
void
vfs_timestamp(struct timespec *tsp)
{
        switch (timestamp_precision) {
        case TSP_SEC:           /* seconds precision */
                getnanotime(tsp);
                tsp->tv_nsec = 0;
                break;
        case TSP_HZ:            /* ticks precision (limit to microseconds) */
                getnanotime(tsp);
                tsp->tv_nsec -= tsp->tv_nsec % 1000;
                break;
        default:
        case TSP_USEC:          /* microseconds (ticks precision) */
                getnanotime(tsp);
                tsp->tv_nsec -= tsp->tv_nsec % 1000;
                break;
        case TSP_NSEC:          /* nanoseconds (ticks precision) */
                getnanotime(tsp);
                break;
        case TSP_USEC_PRECISE:  /* microseconds (high preceision) */
                nanotime(tsp);
                tsp->tv_nsec -= tsp->tv_nsec % 1000;
                break;
        case TSP_NSEC_PRECISE:  /* nanoseconds (high precision) */
                nanotime(tsp);
                break;
        }
}

/*
 * Set vnode attributes to VNOVAL
 */
void
vattr_null(struct vattr *vap)
{
        vap->va_type = VNON;
        vap->va_size = VNOVAL;
        vap->va_bytes = VNOVAL;
        vap->va_mode = VNOVAL;
        vap->va_nlink = VNOVAL;
        vap->va_uid = VNOVAL;
        vap->va_gid = VNOVAL;
        vap->va_fsid = VNOVAL;
        vap->va_fileid = VNOVAL;
        vap->va_blocksize = VNOVAL;
        vap->va_rmajor = VNOVAL;
        vap->va_rminor = VNOVAL;
        vap->va_atime.tv_sec = VNOVAL;
        vap->va_atime.tv_nsec = VNOVAL;
        vap->va_mtime.tv_sec = VNOVAL;
        vap->va_mtime.tv_nsec = VNOVAL;
        vap->va_ctime.tv_sec = VNOVAL;
        vap->va_ctime.tv_nsec = VNOVAL;
        vap->va_flags = VNOVAL;
        vap->va_gen = VNOVAL;
        vap->va_vaflags = 0;
        /* va_*_uuid fields are only valid if related flags are set */
}

/*
 * Flush out and invalidate all buffers associated with a vnode.
 *
 * vp must be locked.
 */
static int vinvalbuf_bp(struct buf *bp, void *data);

struct vinvalbuf_bp_info {
        struct vnode *vp;
        int slptimeo;
        int lkflags;
        int flags;
        int clean;
};

int
vinvalbuf(struct vnode *vp, int flags, int slpflag, int slptimeo)
{
        struct vinvalbuf_bp_info info;
        vm_object_t object;
        int error;

        lwkt_gettoken(&vp->v_token);

        /*
         * If we are being asked to save, call fsync to ensure that the inode
         * is updated.
         */
        if (flags & V_SAVE) {
                error = bio_track_wait(&vp->v_track_write, slpflag, slptimeo);
                if (error)
                        goto done;
                if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
                        if ((error = VOP_FSYNC(vp, MNT_WAIT, 0)) != 0)
                                goto done;
#if 0
                        /*
                         * Dirty bufs may be left or generated via races
                         * in circumstances where vinvalbuf() is called on
                         * a vnode not undergoing reclamation.   Only
                         * panic if we are trying to reclaim the vnode.
                         */
                        if ((vp->v_flag & VRECLAIMED) &&
                            (bio_track_active(&vp->v_track_write) ||
                            !RB_EMPTY(&vp->v_rbdirty_tree))) {
                                panic("vinvalbuf: dirty bufs");
                        }
#endif
                }
        }
        info.slptimeo = slptimeo;
        info.lkflags = LK_EXCLUSIVE | LK_SLEEPFAIL;
        if (slpflag & PCATCH)
                info.lkflags |= LK_PCATCH;
        info.flags = flags;
        info.vp = vp;

        /*
         * Flush the buffer cache until nothing is left, wait for all I/O
         * to complete.  At least one pass is required.  We might block
         * in the pip code so we have to re-check.  Order is important.
         */
        do {
                /*
                 * Flush buffer cache
                 */
                if (!RB_EMPTY(&vp->v_rbclean_tree)) {
                        info.clean = 1;
                        error = RB_SCAN(buf_rb_tree, &vp->v_rbclean_tree,
                                        NULL, vinvalbuf_bp, &info);
                }
                if (!RB_EMPTY(&vp->v_rbdirty_tree)) {
                        info.clean = 0;
                        error = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree,
                                        NULL, vinvalbuf_bp, &info);
                }

                /*
                 * Wait for I/O completion.
                 */
                bio_track_wait(&vp->v_track_write, 0, 0);
                if ((object = vp->v_object) != NULL)
                        refcount_wait(&object->paging_in_progress, "vnvlbx");
        } while (bio_track_active(&vp->v_track_write) ||
                 !RB_EMPTY(&vp->v_rbclean_tree) ||
                 !RB_EMPTY(&vp->v_rbdirty_tree));

        /*
         * Destroy the copy in the VM cache, too.
         */
        if ((object = vp->v_object) != NULL) {
                vm_object_page_remove(object, 0, 0,
                        (flags & V_SAVE) ? TRUE : FALSE);
        }

        if (!RB_EMPTY(&vp->v_rbdirty_tree) || !RB_EMPTY(&vp->v_rbclean_tree))
                panic("vinvalbuf: flush failed");
        if (!RB_EMPTY(&vp->v_rbhash_tree))
                panic("vinvalbuf: flush failed, buffers still present");
        error = 0;
done:
        lwkt_reltoken(&vp->v_token);
        return (error);
}

static int
vinvalbuf_bp(struct buf *bp, void *data)
{
        struct vinvalbuf_bp_info *info = data;
        int error;

        if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
                atomic_add_int(&bp->b_refs, 1);
                error = BUF_TIMELOCK(bp, info->lkflags,
                                     "vinvalbuf", info->slptimeo);
                atomic_subtract_int(&bp->b_refs, 1);
                if (error == 0) {
                        BUF_UNLOCK(bp);
                        error = ENOLCK;
                }
                if (error == ENOLCK)
                        return(0);
                return (-error);
        }
        KKASSERT(bp->b_vp == info->vp);

        /*
         * Must check clean/dirty status after successfully locking as
         * it may race.
         */
        if ((info->clean && (bp->b_flags & B_DELWRI)) ||
            (info->clean == 0 && (bp->b_flags & B_DELWRI) == 0)) {
                BUF_UNLOCK(bp);
                return(0);
        }

        /*
         * NOTE:  NO B_LOCKED CHECK.  Also no buf_checkwrite()
         * check.  This code will write out the buffer, period.
         */
        bremfree(bp);
        if (((bp->b_flags & (B_DELWRI | B_INVAL)) == B_DELWRI) &&
            (info->flags & V_SAVE)) {
                cluster_awrite(bp);
        } else if (info->flags & V_SAVE) {
                /*
                 * Cannot set B_NOCACHE on a clean buffer as this will
                 * destroy the VM backing store which might actually
                 * be dirty (and unsynchronized).
                 */
                bp->b_flags |= (B_INVAL | B_RELBUF);
                brelse(bp);
        } else {
                bp->b_flags |= (B_INVAL | B_NOCACHE | B_RELBUF);
                brelse(bp);
        }
        return(0);
}

/*
 * Truncate a file's buffer and pages to a specified length.  This
 * is in lieu of the old vinvalbuf mechanism, which performed unneeded
 * sync activity.
 *
 * The vnode must be locked.
 */
static int vtruncbuf_bp_trunc_cmp(struct buf *bp, void *data);
static int vtruncbuf_bp_trunc(struct buf *bp, void *data);
static int vtruncbuf_bp_metasync_cmp(struct buf *bp, void *data);
static int vtruncbuf_bp_metasync(struct buf *bp, void *data);

struct vtruncbuf_info {
        struct vnode *vp;
        off_t   truncloffset;
        int     clean;
};

int
vtruncbuf(struct vnode *vp, off_t length, int blksize)
{
        struct vtruncbuf_info info;
        const char *filename;
        int count;

        /*
         * Round up to the *next* block, then destroy the buffers in question.
         * Since we are only removing some of the buffers we must rely on the
         * scan count to determine whether a loop is necessary.
         */
        if ((count = (int)(length % blksize)) != 0)
                info.truncloffset = length + (blksize - count);
        else
                info.truncloffset = length;
        info.vp = vp;

        lwkt_gettoken(&vp->v_token);
        do {
                info.clean = 1;
                count = RB_SCAN(buf_rb_tree, &vp->v_rbclean_tree,
                                vtruncbuf_bp_trunc_cmp,
                                vtruncbuf_bp_trunc, &info);
                info.clean = 0;
                count += RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree,
                                vtruncbuf_bp_trunc_cmp,
                                vtruncbuf_bp_trunc, &info);
        } while(count);

        /*
         * For safety, fsync any remaining metadata if the file is not being
         * truncated to 0.  Since the metadata does not represent the entire
         * dirty list we have to rely on the hit count to ensure that we get
         * all of it.
         */
        if (length > 0) {
                do {
                        count = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree,
                                        vtruncbuf_bp_metasync_cmp,
                                        vtruncbuf_bp_metasync, &info);
                } while (count);
        }

        /*
         * Clean out any left over VM backing store.
         *
         * It is possible to have in-progress I/O from buffers that were
         * not part of the truncation.  This should not happen if we
         * are truncating to 0-length.
         */
        vnode_pager_setsize(vp, length);
        bio_track_wait(&vp->v_track_write, 0, 0);

        /*
         * Debugging only
         */
        spin_lock(&vp->v_spin);
        filename = TAILQ_FIRST(&vp->v_namecache) ?
                   TAILQ_FIRST(&vp->v_namecache)->nc_name : "?";
        spin_unlock(&vp->v_spin);

        /*
         * Make sure no buffers were instantiated while we were trying
         * to clean out the remaining VM pages.  This could occur due
         * to busy dirty VM pages being flushed out to disk.
         */
        do {
                info.clean = 1;
                count = RB_SCAN(buf_rb_tree, &vp->v_rbclean_tree,
                                vtruncbuf_bp_trunc_cmp,
                                vtruncbuf_bp_trunc, &info);
                info.clean = 0;
                count += RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree,
                                vtruncbuf_bp_trunc_cmp,
                                vtruncbuf_bp_trunc, &info);
                if (count) {
                        kprintf("Warning: vtruncbuf():  Had to re-clean %d "
                               "left over buffers in %s\n", count, filename);
                }
        } while(count);

        lwkt_reltoken(&vp->v_token);

        return (0);
}

/*
 * The callback buffer is beyond the new file EOF and must be destroyed.
 * Note that the compare function must conform to the RB_SCAN's requirements.
 */
static
int
vtruncbuf_bp_trunc_cmp(struct buf *bp, void *data)
{
        struct vtruncbuf_info *info = data;

        if (bp->b_loffset >= info->truncloffset)
                return(0);
        return(-1);
}

static
int
vtruncbuf_bp_trunc(struct buf *bp, void *data)
{
        struct vtruncbuf_info *info = data;

        /*
         * Do not try to use a buffer we cannot immediately lock, but sleep
         * anyway to prevent a livelock.  The code will loop until all buffers
         * can be acted upon.
         *
         * We must always revalidate the buffer after locking it to deal
         * with MP races.
         */
        if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
                atomic_add_int(&bp->b_refs, 1);
                if (BUF_LOCK(bp, LK_EXCLUSIVE|LK_SLEEPFAIL) == 0)
                        BUF_UNLOCK(bp);
                atomic_subtract_int(&bp->b_refs, 1);
        } else if ((info->clean && (bp->b_flags & B_DELWRI)) ||
                   (info->clean == 0 && (bp->b_flags & B_DELWRI) == 0) ||
                   bp->b_vp != info->vp ||
                   vtruncbuf_bp_trunc_cmp(bp, data)) {
                BUF_UNLOCK(bp);
        } else {
                bremfree(bp);
                bp->b_flags |= (B_INVAL | B_RELBUF | B_NOCACHE);
                brelse(bp);
        }
        return(1);
}

/*
 * Fsync all meta-data after truncating a file to be non-zero.  Only metadata
 * blocks (with a negative loffset) are scanned.
 * Note that the compare function must conform to the RB_SCAN's requirements.
 */
static int
vtruncbuf_bp_metasync_cmp(struct buf *bp, void *data __unused)
{
        if (bp->b_loffset < 0)
                return(0);
        return(1);
}

static int
vtruncbuf_bp_metasync(struct buf *bp, void *data)
{
        struct vtruncbuf_info *info = data;

        if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
                atomic_add_int(&bp->b_refs, 1);
                if (BUF_LOCK(bp, LK_EXCLUSIVE|LK_SLEEPFAIL) == 0)
                        BUF_UNLOCK(bp);
                atomic_subtract_int(&bp->b_refs, 1);
        } else if ((bp->b_flags & B_DELWRI) == 0 ||
                   bp->b_vp != info->vp ||
                   vtruncbuf_bp_metasync_cmp(bp, data)) {
                BUF_UNLOCK(bp);
        } else {
                bremfree(bp);
                if (bp->b_vp == info->vp)
                        bawrite(bp);
                else
                        bwrite(bp);
        }
        return(1);
}

/*
 * vfsync - implements a multipass fsync on a file which understands
 * dependancies and meta-data.  The passed vnode must be locked.  The
 * waitfor argument may be MNT_WAIT or MNT_NOWAIT, or MNT_LAZY.
 *
 * When fsyncing data asynchronously just do one consolidated pass starting
 * with the most negative block number.  This may not get all the data due
 * to dependancies.
 *
 * When fsyncing data synchronously do a data pass, then a metadata pass,
 * then do additional data+metadata passes to try to get all the data out.
 *
 * Caller must ref the vnode but does not have to lock it.
 */
static int vfsync_wait_output(struct vnode *vp,
                            int (*waitoutput)(struct vnode *, struct thread *));
static int vfsync_dummy_cmp(struct buf *bp __unused, void *data __unused);
static int vfsync_data_only_cmp(struct buf *bp, void *data);
static int vfsync_meta_only_cmp(struct buf *bp, void *data);
static int vfsync_lazy_range_cmp(struct buf *bp, void *data);
static int vfsync_bp(struct buf *bp, void *data);

struct vfsync_info {
        struct vnode *vp;
        int fastpass;
        int synchronous;
        int syncdeps;
        int lazycount;
        int lazylimit;
        int skippedbufs;
        int (*checkdef)(struct buf *);
        int (*cmpfunc)(struct buf *, void *);
};

int
vfsync(struct vnode *vp, int waitfor, int passes,
        int (*checkdef)(struct buf *),
        int (*waitoutput)(struct vnode *, struct thread *))
{
        struct vfsync_info info;
        int error;

        bzero(&info, sizeof(info));
        info.vp = vp;
        if ((info.checkdef = checkdef) == NULL)
                info.syncdeps = 1;

        lwkt_gettoken(&vp->v_token);

        switch(waitfor) {
        case MNT_LAZY | MNT_NOWAIT:
        case MNT_LAZY:
                /*
                 * Lazy (filesystem syncer typ) Asynchronous plus limit the
                 * number of data (not meta) pages we try to flush to 1MB.
                 * A non-zero return means that lazy limit was reached.
                 */
                info.lazylimit = 1024 * 1024;
                info.syncdeps = 1;
                info.cmpfunc = vfsync_lazy_range_cmp;
                error = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree,
                                vfsync_lazy_range_cmp, vfsync_bp, &info);
                info.cmpfunc = vfsync_meta_only_cmp;
                RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree,
                        vfsync_meta_only_cmp, vfsync_bp, &info);
                if (error == 0)
                        vp->v_lazyw = 0;
                else if (!RB_EMPTY(&vp->v_rbdirty_tree))
                        vn_syncer_add(vp, 1);
                error = 0;
                break;
        case MNT_NOWAIT:
                /*
                 * Asynchronous.  Do a data-only pass and a meta-only pass.
                 */
                info.syncdeps = 1;
                info.cmpfunc = vfsync_data_only_cmp;
                RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, vfsync_data_only_cmp,
                        vfsync_bp, &info);
                info.cmpfunc = vfsync_meta_only_cmp;
                RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, vfsync_meta_only_cmp,
                        vfsync_bp, &info);
                error = 0;
                break;
        default:
                /*
                 * Synchronous.  Do a data-only pass, then a meta-data+data
                 * pass, then additional integrated passes to try to get
                 * all the dependancies flushed.
                 */
                info.cmpfunc = vfsync_data_only_cmp;
                info.fastpass = 1;
                RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, vfsync_data_only_cmp,
                        vfsync_bp, &info);
                info.fastpass = 0;
                error = vfsync_wait_output(vp, waitoutput);
                if (error == 0) {
                        info.skippedbufs = 0;
                        info.cmpfunc = vfsync_dummy_cmp;
                        RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
                                vfsync_bp, &info);
                        error = vfsync_wait_output(vp, waitoutput);
                        if (info.skippedbufs) {
                                kprintf("Warning: vfsync skipped %d dirty "
                                        "buf%s in pass2!\n",
                                        info.skippedbufs,
                                        ((info.skippedbufs > 1) ? "s" : ""));
                        }
                }
                while (error == 0 && passes > 0 &&
                       !RB_EMPTY(&vp->v_rbdirty_tree)
                ) {
                        info.skippedbufs = 0;
                        if (--passes == 0) {
                                info.synchronous = 1;
                                info.syncdeps = 1;
                        }
                        info.cmpfunc = vfsync_dummy_cmp;
                        error = RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
                                        vfsync_bp, &info);
                        if (error < 0)
                                error = -error;
                        info.syncdeps = 1;
                        if (error == 0)
                                error = vfsync_wait_output(vp, waitoutput);
                        if (info.skippedbufs && passes == 0) {
                                kprintf("Warning: vfsync skipped %d dirty "
                                        "buf%s in final pass!\n",
                                        info.skippedbufs,
                                        ((info.skippedbufs > 1) ? "s" : ""));
                        }
                }
#if 0
                /*
                 * This case can occur normally because vnode lock might
                 * not be held.
                 */
                if (!RB_EMPTY(&vp->v_rbdirty_tree))
                        kprintf("dirty bufs left after final pass\n");
#endif
                break;
        }
        lwkt_reltoken(&vp->v_token);

        return(error);
}

static int
vfsync_wait_output(struct vnode *vp,
                   int (*waitoutput)(struct vnode *, struct thread *))
{
        int error;

        error = bio_track_wait(&vp->v_track_write, 0, 0);
        if (waitoutput)
                error = waitoutput(vp, curthread);
        return(error);
}

static int
vfsync_dummy_cmp(struct buf *bp __unused, void *data __unused)
{
        return(0);
}

static int
vfsync_data_only_cmp(struct buf *bp, void *data)
{
        if (bp->b_loffset < 0)
                return(-1);
        return(0);
}

static int
vfsync_meta_only_cmp(struct buf *bp, void *data)
{
        if (bp->b_loffset < 0)
                return(0);
        return(1);
}

static int
vfsync_lazy_range_cmp(struct buf *bp, void *data)
{
        struct vfsync_info *info = data;

        if (bp->b_loffset < info->vp->v_lazyw)
                return(-1);
        return(0);
}

static int
vfsync_bp(struct buf *bp, void *data)
{
        struct vfsync_info *info = data;
        struct vnode *vp = info->vp;
        int error;

        if (info->fastpass) {
                /*
                 * Ignore buffers that we cannot immediately lock.
                 */
                if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT)) {
                        /*
                         * Removed BUF_TIMELOCK(..., 1), even a 1-tick
                         * delay can mess up performance
                         *
                         * Another reason is that during a dirty-buffer
                         * scan a clustered write can start I/O on buffers
                         * ahead of the scan, causing the scan to not
                         * get a lock here.  Usually this means the write
                         * is already in progress so, in fact, we *want*
                         * to skip the buffer.
                         */
                        ++info->skippedbufs;
                        return(0);
                }
        } else if (info->synchronous == 0) {
                /*
                 * Normal pass, give the buffer a little time to become
                 * available to us.
                 */
                if (BUF_TIMELOCK(bp, LK_EXCLUSIVE, "bflst2", hz / 10)) {
                        ++info->skippedbufs;
                        return(0);
                }
        } else {
                /*
                 * Synchronous pass, give the buffer a lot of time before
                 * giving up.
                 */
                if (BUF_TIMELOCK(bp, LK_EXCLUSIVE, "bflst3", hz * 10)) {
                        ++info->skippedbufs;
                        return(0);
                }
        }

        /*
         * We must revalidate the buffer after locking.
         */
        if ((bp->b_flags & B_DELWRI) == 0 ||
            bp->b_vp != info->vp ||
            info->cmpfunc(bp, data)) {
                BUF_UNLOCK(bp);
                return(0);
        }

        /*
         * If syncdeps is not set we do not try to write buffers which have
         * dependancies.
         */
        if (!info->synchronous && info->syncdeps == 0 && info->checkdef(bp)) {
                BUF_UNLOCK(bp);
                return(0);
        }

        /*
         * B_NEEDCOMMIT (primarily used by NFS) is a state where the buffer
         * has been written but an additional handshake with the device
         * is required before we can dispose of the buffer.  We have no idea
         * how to do this so we have to skip these buffers.
         */
        if (bp->b_flags & B_NEEDCOMMIT) {
                BUF_UNLOCK(bp);
                return(0);
        }

        /*
         * Ask bioops if it is ok to sync.  If not the VFS may have
         * set B_LOCKED so we have to cycle the buffer.
         */
        if (LIST_FIRST(&bp->b_dep) != NULL && buf_checkwrite(bp)) {
                bremfree(bp);
                brelse(bp);
                return(0);
        }

        if (info->synchronous) {
                /*
                 * Synchronous flush.  An error may be returned and will
                 * stop the scan.
                 */
                bremfree(bp);
                error = bwrite(bp);
        } else {
                /*
                 * Asynchronous flush.  We use the error return to support
                 * MNT_LAZY flushes.
                 *
                 * In low-memory situations we revert to synchronous
                 * operation.  This should theoretically prevent the I/O
                 * path from exhausting memory in a non-recoverable way.
                 */
                vp->v_lazyw = bp->b_loffset;
                bremfree(bp);
                if (vm_paging_min()) {
                        /* low memory */
                        info->lazycount += bp->b_bufsize;
                        bwrite(bp);
                } else {
                        /* normal */
                        info->lazycount += cluster_awrite(bp);
                        waitrunningbufspace();
                        /*vm_wait_nominal();*/
                }
                if (info->lazylimit && info->lazycount >= info->lazylimit)
                        error = 1;
                else
                        error = 0;
        }
        return(-error);
}

/*
 * Associate a buffer with a vnode.
 *
 * MPSAFE
 */
int
bgetvp(struct vnode *vp, struct buf *bp, int testsize)
{
        KASSERT(bp->b_vp == NULL, ("bgetvp: not free"));
        KKASSERT((bp->b_flags & (B_HASHED|B_DELWRI|B_VNCLEAN|B_VNDIRTY)) == 0);

        /*
         * Insert onto list for new vnode.
         */
        lwkt_gettoken(&vp->v_token);

        if (buf_rb_hash_RB_INSERT(&vp->v_rbhash_tree, bp)) {
                lwkt_reltoken(&vp->v_token);
                return (EEXIST);
        }

        /*
         * Diagnostics (mainly for HAMMER debugging).  Check for
         * overlapping buffers.
         */
        if (check_buf_overlap) {
                struct buf *bx;
                bx = buf_rb_hash_RB_PREV(bp);
                if (bx) {
                        if (bx->b_loffset + bx->b_bufsize > bp->b_loffset) {
                                kprintf("bgetvp: overlapl %016jx/%d %016jx "
                                        "bx %p bp %p\n",
                                        (intmax_t)bx->b_loffset,
                                        bx->b_bufsize,
                                        (intmax_t)bp->b_loffset,
                                        bx, bp);
                                if (check_buf_overlap > 1)
                                        panic("bgetvp - overlapping buffer");
                        }
                }
                bx = buf_rb_hash_RB_NEXT(bp);
                if (bx) {
                        if (bp->b_loffset + testsize > bx->b_loffset) {
                                kprintf("bgetvp: overlapr %016jx/%d %016jx "
                                        "bp %p bx %p\n",
                                        (intmax_t)bp->b_loffset,
                                        testsize,
                                        (intmax_t)bx->b_loffset,
                                        bp, bx);
                                if (check_buf_overlap > 1)
                                        panic("bgetvp - overlapping buffer");
                        }
                }
        }
        bp->b_vp = vp;
        bp->b_flags |= B_HASHED;
        bp->b_flags |= B_VNCLEAN;
        if (buf_rb_tree_RB_INSERT(&vp->v_rbclean_tree, bp))
                panic("reassignbuf: dup lblk/clean vp %p bp %p", vp, bp);
        /*vhold(vp);*/
        lwkt_reltoken(&vp->v_token);
        return(0);
}

/*
 * Disassociate a buffer from a vnode.
 *
 * MPSAFE
 */
void
brelvp(struct buf *bp)
{
        struct vnode *vp;

        KASSERT(bp->b_vp != NULL, ("brelvp: NULL"));

        /*
         * Delete from old vnode list, if on one.
         */
        vp = bp->b_vp;
        lwkt_gettoken(&vp->v_token);
        if (bp->b_flags & (B_VNDIRTY | B_VNCLEAN)) {
                if (bp->b_flags & B_VNDIRTY)
                        buf_rb_tree_RB_REMOVE(&vp->v_rbdirty_tree, bp);
                else
                        buf_rb_tree_RB_REMOVE(&vp->v_rbclean_tree, bp);
                bp->b_flags &= ~(B_VNDIRTY | B_VNCLEAN);
        }
        if (bp->b_flags & B_HASHED) {
                buf_rb_hash_RB_REMOVE(&vp->v_rbhash_tree, bp);
                bp->b_flags &= ~B_HASHED;
        }

        /*
         * Only remove from synclist when no dirty buffers are left AND
         * the VFS has not flagged the vnode's inode as being dirty.
         */
        if ((vp->v_flag & (VONWORKLST | VISDIRTY | VOBJDIRTY)) == VONWORKLST &&
            RB_EMPTY(&vp->v_rbdirty_tree)) {
                vn_syncer_remove(vp, 0);
        }
        bp->b_vp = NULL;

        lwkt_reltoken(&vp->v_token);

        /*vdrop(vp);*/
}

/*
 * Reassign the buffer to the proper clean/dirty list based on B_DELWRI.
 * This routine is called when the state of the B_DELWRI bit is changed.
 *
 * Must be called with vp->v_token held.
 * MPSAFE
 */
void
reassignbuf(struct buf *bp)
{
        struct vnode *vp = bp->b_vp;
        int delay;

        ASSERT_LWKT_TOKEN_HELD(&vp->v_token);
        ++reassignbufcalls;

        /*
         * B_PAGING flagged buffers cannot be reassigned because their vp
         * is not fully linked in.
         */
        if (bp->b_flags & B_PAGING)
                panic("cannot reassign paging buffer");

        if (bp->b_flags & B_DELWRI) {
                /*
                 * Move to the dirty list, add the vnode to the worklist
                 */
                if (bp->b_flags & B_VNCLEAN) {
                        buf_rb_tree_RB_REMOVE(&vp->v_rbclean_tree, bp);
                        bp->b_flags &= ~B_VNCLEAN;
                }
                if ((bp->b_flags & B_VNDIRTY) == 0) {
                        if (buf_rb_tree_RB_INSERT(&vp->v_rbdirty_tree, bp)) {
                                panic("reassignbuf: dup lblk vp %p bp %p",
                                      vp, bp);
                        }
                        bp->b_flags |= B_VNDIRTY;
                }
                if ((vp->v_flag & VONWORKLST) == 0) {
                        switch (vp->v_type) {
                        case VDIR:
                                delay = dirdelay;
                                break;
                        case VCHR:
                        case VBLK:
                                if (vp->v_rdev &&
                                    vp->v_rdev->si_mountpoint != NULL) {
                                        delay = metadelay;
                                        break;
                                }
                                /* fall through */
                        default:
                                delay = filedelay;
                        }
                        vn_syncer_add(vp, delay);
                }
        } else {
                /*
                 * Move to the clean list, remove the vnode from the worklist
                 * if no dirty blocks remain.
                 */
                if (bp->b_flags & B_VNDIRTY) {
                        buf_rb_tree_RB_REMOVE(&vp->v_rbdirty_tree, bp);
                        bp->b_flags &= ~B_VNDIRTY;
                }
                if ((bp->b_flags & B_VNCLEAN) == 0) {
                        if (buf_rb_tree_RB_INSERT(&vp->v_rbclean_tree, bp)) {
                                panic("reassignbuf: dup lblk vp %p bp %p",
                                      vp, bp);
                        }
                        bp->b_flags |= B_VNCLEAN;
                }

                /*
                 * Only remove from synclist when no dirty buffers are left
                 * AND the VFS has not flagged the vnode's inode as being
                 * dirty.
                 */
                if ((vp->v_flag & (VONWORKLST | VISDIRTY | VOBJDIRTY)) ==
                     VONWORKLST &&
                    RB_EMPTY(&vp->v_rbdirty_tree)) {
                        vn_syncer_remove(vp, 0);
                }
        }
}

/*
 * Create a vnode for a block device.  Used for mounting the root file
 * system.
 *
 * A vref()'d vnode is returned.
 */
extern struct vop_ops *devfs_vnode_dev_vops_p;
int
bdevvp(cdev_t dev, struct vnode **vpp)
{
        struct vnode *vp;
        struct vnode *nvp;
        int error;

        if (dev == NULL) {
                *vpp = NULLVP;
                return (ENXIO);
        }
        error = getspecialvnode(VT_NON, NULL, &devfs_vnode_dev_vops_p,
                                &nvp, 0, 0);
        if (error) {
                *vpp = NULLVP;
                return (error);
        }
        vp = nvp;
        vp->v_type = VCHR;
#if 0
        vp->v_rdev = dev;
#endif
        v_associate_rdev(vp, dev);
        vp->v_umajor = dev->si_umajor;
        vp->v_uminor = dev->si_uminor;
        vx_unlock(vp);
        *vpp = vp;
        return (0);
}

int
v_associate_rdev(struct vnode *vp, cdev_t dev)
{
        if (dev == NULL)
                return(ENXIO);
        if (dev_is_good(dev) == 0)
                return(ENXIO);
        KKASSERT(vp->v_rdev == NULL);
        vp->v_rdev = reference_dev(dev);
        lwkt_gettoken(&spechash_token);
        SLIST_INSERT_HEAD(&dev->si_hlist, vp, v_cdevnext);
        lwkt_reltoken(&spechash_token);
        return(0);
}

void
v_release_rdev(struct vnode *vp)
{
        cdev_t dev;

        if ((dev = vp->v_rdev) != NULL) {
                lwkt_gettoken(&spechash_token);
                SLIST_REMOVE(&dev->si_hlist, vp, vnode, v_cdevnext);
                vp->v_rdev = NULL;
                release_dev(dev);
                lwkt_reltoken(&spechash_token);
        }
}

/*
 * Add a vnode to the alias list hung off the cdev_t.  We only associate
 * the device number with the vnode.  The actual device is not associated
 * until the vnode is opened (usually in spec_open()), and will be
 * disassociated on last close.
 */
void
addaliasu(struct vnode *nvp, int x, int y)
{
        if (nvp->v_type != VBLK && nvp->v_type != VCHR)
                panic("addaliasu on non-special vnode");
        nvp->v_umajor = x;
        nvp->v_uminor = y;
}

/*
 * Simple call that a filesystem can make to try to get rid of a
 * vnode.  It will fail if anyone is referencing the vnode (including
 * the caller).
 *
 * The filesystem can check whether its in-memory inode structure still
 * references the vp on return.
 *
 * May only be called if the vnode is in a known state (i.e. being prevented
 * from being deallocated by some other condition such as a vfs inode hold).
 *
 * This call might not succeed.
 */
void
vclean_unlocked(struct vnode *vp)
{
        vx_get(vp);
        if (VREFCNT(vp) <= 1)
                vgone_vxlocked(vp);
        vx_put(vp);
}

/*
 * Disassociate a vnode from its underlying filesystem.
 *
 * The vnode must be VX locked and referenced.  In all normal situations
 * there are no active references.  If vclean_vxlocked() is called while
 * there are active references, the vnode is being ripped out and we have
 * to call VOP_CLOSE() as appropriate before we can reclaim it.
 */
static void
vclean_vxlocked(struct vnode *vp, int flags)
{
        int active;
        int n;
        vm_object_t object;
        struct namecache *ncp;

        /*
         * If the vnode has already been reclaimed we have nothing to do.
         */
        if (vp->v_flag & VRECLAIMED)
                return;

        /*
         * Set flag to interlock operation, flag finalization to ensure
         * that the vnode winds up on the inactive list, and set v_act to 0.
         */
        vsetflags(vp, VRECLAIMED);
        atomic_set_int(&vp->v_refcnt, VREF_FINALIZE);
        vp->v_act = 0;

        if (verbose_reclaims) {
                if ((ncp = TAILQ_FIRST(&vp->v_namecache)) != NULL)
                        kprintf("Debug: reclaim %p %s\n", vp, ncp->nc_name);
        }

        /*
         * Scrap the vfs cache
         */
        while (cache_inval_vp(vp, 0) != 0) {
                kprintf("Warning: vnode %p clean/cache_resolution "
                        "race detected\n", vp);
                tsleep(vp, 0, "vclninv", 2);
        }

        /*
         * Check to see if the vnode is in use. If so we have to reference it
         * before we clean it out so that its count cannot fall to zero and
         * generate a race against ourselves to recycle it.
         */
        active = (VREFCNT(vp) > 0);

        /*
         * Clean out any buffers associated with the vnode and destroy its
         * object, if it has one.
         */
        vinvalbuf(vp, V_SAVE, 0, 0);

        /*
         * If purging an active vnode (typically during a forced unmount
         * or reboot), it must be closed and deactivated before being
         * reclaimed.  This isn't really all that safe, but what can
         * we do? XXX.
         *
         * Note that neither of these routines unlocks the vnode.
         */
        if (active && (flags & DOCLOSE)) {
                while ((n = vp->v_opencount) != 0) {
                        if (vp->v_writecount)
                                VOP_CLOSE(vp, FWRITE|FNONBLOCK, NULL);
                        else
                                VOP_CLOSE(vp, FNONBLOCK, NULL);
                        if (vp->v_opencount == n) {
                                kprintf("Warning: unable to force-close"
                                       " vnode %p\n", vp);
                                break;
                        }
                }
        }

        /*
         * If the vnode has not been deactivated, deactivated it.  Deactivation
         * can create new buffers and VM pages so we have to call vinvalbuf()
         * again to make sure they all get flushed.
         *
         * This can occur if a file with a link count of 0 needs to be
         * truncated.
         *
         * If the vnode is already dead don't try to deactivate it.
         */
        if ((vp->v_flag & VINACTIVE) == 0) {
                vsetflags(vp, VINACTIVE);
                if (vp->v_mount)
                        VOP_INACTIVE(vp);
                vinvalbuf(vp, V_SAVE, 0, 0);
        }

        /*
         * If the vnode has an object, destroy it.
         */
        while ((object = vp->v_object) != NULL) {
                vm_object_hold(object);
                if (object == vp->v_object)
                        break;
                vm_object_drop(object);
        }

        if (object != NULL) {
                if (object->ref_count == 0) {
                        if ((object->flags & OBJ_DEAD) == 0)
                                vm_object_terminate(object);
                        vm_object_drop(object);
                        vclrflags(vp, VOBJBUF);
                } else {
                        vm_pager_deallocate(object);
                        vclrflags(vp, VOBJBUF);
                        vm_object_drop(object);
                }
        }
        KKASSERT((vp->v_flag & VOBJBUF) == 0);

        if (vp->v_flag & VOBJDIRTY)
                vclrobjdirty(vp);

        /*
         * Reclaim the vnode if not already dead.
         */
        if (vp->v_mount && VOP_RECLAIM(vp))
                panic("vclean: cannot reclaim");

        /*
         * Done with purge, notify sleepers of the grim news.
         */
        vp->v_ops = &dead_vnode_vops_p;
        vn_gone(vp);
        vp->v_tag = VT_NON;

        /*
         * If we are destroying an active vnode, reactivate it now that
         * we have reassociated it with deadfs.  This prevents the system
         * from crashing on the vnode due to it being unexpectedly marked
         * as inactive or reclaimed.
         */
        if (active && (flags & DOCLOSE)) {
                vclrflags(vp, VINACTIVE | VRECLAIMED);
        }
}

/*
 * Eliminate all activity associated with the requested vnode
 * and with all vnodes aliased to the requested vnode.
 *
 * The vnode must be referenced but should not be locked.
 */
int
vrevoke(struct vnode *vp, struct ucred *cred)
{
        struct vnode *vq;
        struct vnode *vqn;
        cdev_t dev;
        int error;

        /*
         * If the vnode has a device association, scrap all vnodes associated
         * with the device.  Don't let the device disappear on us while we
         * are scrapping the vnodes.
         *
         * The passed vp will probably show up in the list, do not VX lock
         * it twice!
         *
         * Releasing the vnode's rdev here can mess up specfs's call to
         * device close, so don't do it.  The vnode has been disassociated
         * and the device will be closed after the last ref on the related
         * fp goes away (if not still open by e.g. the kernel).
         */
        if (vp->v_type != VCHR) {
                error = fdrevoke(vp, DTYPE_VNODE, cred);
                return (error);
        }
        if ((dev = vp->v_rdev) == NULL) {
                return(0);
        }
        reference_dev(dev);
        lwkt_gettoken(&spechash_token);

restart:
        vqn = SLIST_FIRST(&dev->si_hlist);
        if (vqn)
                vhold(vqn);
        while ((vq = vqn) != NULL) {
                if (VREFCNT(vq) > 0) {
                        vref(vq);
                        fdrevoke(vq, DTYPE_VNODE, cred);
                        /*v_release_rdev(vq);*/
                        vrele(vq);
                        if (vq->v_rdev != dev) {
                                vdrop(vq);
                                goto restart;
                        }
                }
                vqn = SLIST_NEXT(vq, v_cdevnext);
                if (vqn)
                        vhold(vqn);
                vdrop(vq);
        }
        lwkt_reltoken(&spechash_token);
        dev_drevoke(dev);
        release_dev(dev);
        return (0);
}

/*
 * This is called when the object underlying a vnode is being destroyed,
 * such as in a remove().  Try to recycle the vnode immediately if the
 * only active reference is our reference.
 *
 * Directory vnodes in the namecache with children cannot be immediately
 * recycled because numerous VOP_N*() ops require them to be stable.
 *
 * To avoid recursive recycling from VOP_INACTIVE implemenetations this
 * function is a NOP if VRECLAIMED is already set.
 */
int
vrecycle(struct vnode *vp)
{
        if (VREFCNT(vp) <= 1 && (vp->v_flag & VRECLAIMED) == 0) {
                if (cache_inval_vp_nonblock(vp))
                        return(0);
                vgone_vxlocked(vp);
                return (1);
        }
        return (0);
}

/*
 * Return the maximum I/O size allowed for strategy calls on VP.
 *
 * If vp is VCHR or VBLK we dive the device, otherwise we use
 * the vp's mount info.
 *
 * The returned value is clamped at MAXPHYS as most callers cannot use
 * buffers larger than that size.
 */
int
vmaxiosize(struct vnode *vp)
{
        int maxiosize;

        if (vp->v_type == VBLK || vp->v_type == VCHR)
                maxiosize = vp->v_rdev->si_iosize_max;
        else
                maxiosize = vp->v_mount->mnt_iosize_max;

        if (maxiosize > MAXPHYS)
                maxiosize = MAXPHYS;
        return (maxiosize);
}

/*
 * Eliminate all activity associated with a vnode in preparation for
 * destruction.
 *
 * The vnode must be VX locked and refd and will remain VX locked and refd
 * on return.  This routine may be called with the vnode in any state, as
 * long as it is VX locked.  The vnode will be cleaned out and marked
 * VRECLAIMED but will not actually be reused until all existing refs and
 * holds go away.
 *
 * NOTE: This routine may be called on a vnode which has not yet been
 * already been deactivated (VOP_INACTIVE), or on a vnode which has
 * already been reclaimed.
 *
 * This routine is not responsible for placing us back on the freelist.
 * Instead, it happens automatically when the caller releases the VX lock
 * (assuming there aren't any other references).
 */
void
vgone_vxlocked(struct vnode *vp)
{
        /*
         * assert that the VX lock is held.  This is an absolute requirement
         * now for vgone_vxlocked() to be called.
         */
        KKASSERT(lockinuse(&vp->v_lock));

        /*
         * Clean out the filesystem specific data and set the VRECLAIMED
         * bit.  Also deactivate the vnode if necessary.
         *
         * The vnode should have automatically been removed from the syncer
         * list as syncer/dirty flags cleared during the cleaning.
         */
        vclean_vxlocked(vp, DOCLOSE);

        /*
         * Normally panic if the vnode is still dirty, unless we are doing
         * a forced unmount (tmpfs typically).
         */
        if (vp->v_flag & VONWORKLST) {
                if (vp->v_mount->mnt_kern_flag & MNTK_UNMOUNTF) {
                        /* force removal */
                        vn_syncer_remove(vp, 1);
                } else {
                        panic("vp %p still dirty in vgone after flush", vp);
                }
        }

        /*
         * Delete from old mount point vnode list, if on one.
         */
        if (vp->v_mount != NULL) {
                KKASSERT(vp->v_data == NULL);
                insmntque(vp, NULL);
        }

        /*
         * If special device, remove it from special device alias list
         * if it is on one.  This should normally only occur if a vnode is
         * being revoked as the device should otherwise have been released
         * naturally.
         */
        if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_rdev != NULL) {
                v_release_rdev(vp);
        }

        /*
         * Set us to VBAD
         */
        vp->v_type = VBAD;
}

/*
 * Calculate the total number of references to a special device.  This
 * routine may only be called for VBLK and VCHR vnodes since v_rdev is
 * an overloaded field.  Since dev_from_devid() can now return NULL, we
 * have to check for a NULL v_rdev.
 */
int
count_dev(cdev_t dev)
{
        struct vnode *vp;
        int count = 0;

        if (SLIST_FIRST(&dev->si_hlist)) {
                lwkt_gettoken(&spechash_token);
                SLIST_FOREACH(vp, &dev->si_hlist, v_cdevnext) {
                        count += vp->v_opencount;
                }
                lwkt_reltoken(&spechash_token);
        }
        return(count);
}

int
vcount(struct vnode *vp)
{
        if (vp->v_rdev == NULL)
                return(0);
        return(count_dev(vp->v_rdev));
}

/*
 * Initialize VMIO for a vnode.  This routine MUST be called before a
 * VFS can issue buffer cache ops on a vnode.  It is typically called
 * when a vnode is initialized from its inode.
 */
int
vinitvmio(struct vnode *vp, off_t filesize, int blksize, int boff)
{
        vm_object_t object;
        int error = 0;

        object = vp->v_object;
        if (object) {
                vm_object_hold(object);
                KKASSERT(vp->v_object == object);
        }

        if (object == NULL) {
                object = vnode_pager_alloc(vp, filesize, 0, 0, blksize, boff);

                /*
                 * Dereference the reference we just created.  This assumes
                 * that the object is associated with the vp.  Allow it to
                 * have zero refs.  It cannot be destroyed as long as it
                 * is associated with the vnode.
                 */
                vm_object_hold(object);
                atomic_add_int(&object->ref_count, -1);
                vrele(vp);
        } else {
                KKASSERT((object->flags & OBJ_DEAD) == 0);
        }
        KASSERT(vp->v_object != NULL, ("vinitvmio: NULL object"));
        vsetflags(vp, VOBJBUF);
        vm_object_drop(object);

        return (error);
}


/*
 * Print out a description of a vnode.
 */
static char *typename[] =
{"VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD"};

void
vprint(char *label, struct vnode *vp)
{
        char buf[96];

        if (label != NULL)
                kprintf("%s: %p: ", label, (void *)vp);
        else
                kprintf("%p: ", (void *)vp);
        kprintf("type %s, refcnt %08x, writecount %d, holdcnt %d,",
                typename[vp->v_type],
                vp->v_refcnt, vp->v_writecount, vp->v_auxrefs);
        buf[0] = '\0';
        if (vp->v_flag & VROOT)
                strcat(buf, "|VROOT");
        if (vp->v_flag & VPFSROOT)
                strcat(buf, "|VPFSROOT");
        if (vp->v_flag & VTEXT)
                strcat(buf, "|VTEXT");
        if (vp->v_flag & VSYSTEM)
                strcat(buf, "|VSYSTEM");
        if (vp->v_flag & VOBJBUF)
                strcat(buf, "|VOBJBUF");
        if (buf[0] != '\0')
                kprintf(" flags (%s)", &buf[1]);
        if (vp->v_data == NULL) {
                kprintf("\n");
        } else {
                kprintf("\n\t");
                VOP_PRINT(vp);
        }
}

/*
 * Do the usual access checking.
 * file_mode, uid and gid are from the vnode in question,
 * while acc_mode and cred are from the VOP_ACCESS parameter list
 */
int
vaccess(enum vtype type, mode_t file_mode, uid_t uid, gid_t gid,
        mode_t acc_mode, struct ucred *cred)
{
        mode_t mask;
        int ismember;

        /*
         * Super-user always gets read/write access, but execute access depends
         * on at least one execute bit being set.
         */
        if (caps_priv_check(cred, SYSCAP_RESTRICTEDROOT) == 0) {
                if ((acc_mode & VEXEC) && type != VDIR &&
                    (file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0)
                        return (EACCES);
                return (0);
        }

        mask = 0;

        /* Otherwise, check the owner. */
        if (cred->cr_uid == uid) {
                if (acc_mode & VEXEC)
                        mask |= S_IXUSR;
                if (acc_mode & VREAD)
                        mask |= S_IRUSR;
                if (acc_mode & VWRITE)
                        mask |= S_IWUSR;
                return ((file_mode & mask) == mask ? 0 : EACCES);
        }

        /* Otherwise, check the groups. */
        ismember = groupmember(gid, cred);
        if (cred->cr_svgid == gid || ismember) {
                if (acc_mode & VEXEC)
                        mask |= S_IXGRP;
                if (acc_mode & VREAD)
                        mask |= S_IRGRP;
                if (acc_mode & VWRITE)
                        mask |= S_IWGRP;
                return ((file_mode & mask) == mask ? 0 : EACCES);
        }

        /* Otherwise, check everyone else. */
        if (acc_mode & VEXEC)
                mask |= S_IXOTH;
        if (acc_mode & VREAD)
                mask |= S_IROTH;
        if (acc_mode & VWRITE)
                mask |= S_IWOTH;
        return ((file_mode & mask) == mask ? 0 : EACCES);
}

#ifdef DDB
#include <ddb/ddb.h>

static int db_show_locked_vnodes(struct mount *mp, void *data);

/*
 * List all of the locked vnodes in the system.
 * Called when debugging the kernel.
 */
DB_SHOW_COMMAND(lockedvnodes, lockedvnodes)
{
        kprintf("Locked vnodes\n");
        mountlist_scan(db_show_locked_vnodes, NULL,
                        MNTSCAN_FORWARD|MNTSCAN_NOBUSY);
}

static int
db_show_locked_vnodes(struct mount *mp, void *data __unused)
{
        struct vnode *vp;

        TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes) {
                if (vn_islocked(vp))
                        vprint(NULL, vp);
        }
        return(0);
}
#endif

/*
 * Top level filesystem related information gathering.
 */
static int      sysctl_ovfs_conf (SYSCTL_HANDLER_ARGS);

static int
vfs_sysctl(SYSCTL_HANDLER_ARGS)
{
        int *name = (int *)arg1 - 1;    /* XXX */
        u_int namelen = arg2 + 1;       /* XXX */
        struct vfsconf *vfsp;
        int maxtypenum;

#if 1 || defined(COMPAT_PRELITE2)
        /* Resolve ambiguity between VFS_VFSCONF and VFS_GENERIC. */
        if (namelen == 1)
                return (sysctl_ovfs_conf(oidp, arg1, arg2, req));
#endif

#ifdef notyet
        /* all sysctl names at this level are at least name and field */
        if (namelen < 2)
                return (ENOTDIR);               /* overloaded */
        if (name[0] != VFS_GENERIC) {
                vfsp = vfsconf_find_by_typenum(name[0]);
                if (vfsp == NULL)
                        return (EOPNOTSUPP);
                return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1,
                    oldp, oldlenp, newp, newlen, p));
        }
#endif
        switch (name[1]) {
        case VFS_MAXTYPENUM:
                if (namelen != 2)
                        return (ENOTDIR);
                maxtypenum = vfsconf_get_maxtypenum();
                return (SYSCTL_OUT(req, &maxtypenum, sizeof(maxtypenum)));
        case VFS_CONF:
                if (namelen != 3)
                        return (ENOTDIR);       /* overloaded */
                vfsp = vfsconf_find_by_typenum(name[2]);
                if (vfsp == NULL)
                        return (EOPNOTSUPP);
                return (SYSCTL_OUT(req, vfsp, sizeof *vfsp));
        }
        return (EOPNOTSUPP);
}

SYSCTL_NODE(_vfs, VFS_GENERIC, generic, CTLFLAG_RD, vfs_sysctl,
        "Generic filesystem");

#if 1 || defined(COMPAT_PRELITE2)

static int
sysctl_ovfs_conf_iter(struct vfsconf *vfsp, void *data)
{
        int error;
        struct ovfsconf ovfs;
        struct sysctl_req *req = (struct sysctl_req*) data;

        bzero(&ovfs, sizeof(ovfs));
        ovfs.vfc_vfsops = vfsp->vfc_vfsops;     /* XXX used as flag */
        strcpy(ovfs.vfc_name, vfsp->vfc_name);
        ovfs.vfc_index = vfsp->vfc_typenum;
        ovfs.vfc_refcount = vfsp->vfc_refcount;
        ovfs.vfc_flags = vfsp->vfc_flags;
        error = SYSCTL_OUT(req, &ovfs, sizeof ovfs);
        if (error)
                return error; /* abort iteration with error code */
        else
                return 0; /* continue iterating with next element */
}

static int
sysctl_ovfs_conf(SYSCTL_HANDLER_ARGS)
{
        return vfsconf_each(sysctl_ovfs_conf_iter, (void*)req);
}

#endif /* 1 || COMPAT_PRELITE2 */

/*
 * Check to see if a filesystem is mounted on a block device.
 */
int
vfs_mountedon(struct vnode *vp)
{
        cdev_t dev;

        dev = vp->v_rdev;
        if (dev != NULL && dev->si_mountpoint)
                return (EBUSY);
        return (0);
}

/*
 * Unmount all filesystems. The list is traversed in reverse order
 * of mounting to avoid dependencies.
 *
 * We want the umountall to be able to break out of its loop if a
 * failure occurs, after scanning all possible mounts, so the callback
 * returns 0 on error.
 *
 * NOTE: Do not call mountlist_remove(mp) on error any more, this will
 *       confuse mountlist_scan()'s unbusy check.
 */
static int vfs_umountall_callback(struct mount *mp, void *data);

void
vfs_unmountall(int halting)
{
        int count;

        do {
                count = mountlist_scan(vfs_umountall_callback, &halting,
                                       MNTSCAN_REVERSE|MNTSCAN_NOBUSY);
        } while (count);
}

static
int
vfs_umountall_callback(struct mount *mp, void *data)
{
        int error;
        int halting = *(int *)data;

        /*
         * NOTE: When halting, dounmount will disconnect but leave
         *       certain mount points intact.  e.g. devfs.
         */
        error = dounmount(mp, MNT_FORCE, halting);
        if (error) {
                kprintf("unmount of filesystem mounted from %s failed (",
                        mp->mnt_stat.f_mntfromname);
                if (error == EBUSY)
                        kprintf("BUSY)\n");
                else
                        kprintf("%d)\n", error);
                return 0;
        } else {
                return 1;
        }
}

/*
 * Checks the mount flags for parameter mp and put the names comma-separated
 * into a string buffer buf with a size limit specified by len.
 *
 * It returns the number of bytes written into buf, and (*errorp) will be
 * set to 0, EINVAL (if passed length is 0), or ENOSPC (supplied buffer was
 * not large enough).  The buffer will be 0-terminated if len was not 0.
 */
size_t
vfs_flagstostr(int flags, const struct mountctl_opt *optp,
               char *buf, size_t len, int *errorp)
{
        static const struct mountctl_opt optnames[] = {
                { MNT_RDONLY,           "read-only" },
                { MNT_SYNCHRONOUS,      "synchronous" },
                { MNT_NOEXEC,           "noexec" },
                { MNT_NOSUID,           "nosuid" },
                { MNT_NODEV,            "nodev" },
                { MNT_AUTOMOUNTED,      "automounted" },
                { MNT_ASYNC,            "asynchronous" },
                { MNT_SUIDDIR,          "suiddir" },
                { MNT_SOFTDEP,          "soft-updates" },
                { MNT_NOSYMFOLLOW,      "nosymfollow" },
                { MNT_TRIM,             "trim" },
                { MNT_NOATIME,          "noatime" },
                { MNT_NOCLUSTERR,       "noclusterr" },
                { MNT_NOCLUSTERW,       "noclusterw" },
                { MNT_EXRDONLY,         "NFS read-only" },
                { MNT_EXPORTED,         "NFS exported" },
                /* Remaining NFS flags could come here */
                { MNT_LOCAL,            "local" },
                { MNT_QUOTA,            "with-quotas" },
                /* { MNT_ROOTFS,           "rootfs" }, */
                /* { MNT_IGNORE,           "ignore" }, */
                { 0,                    NULL}
        };
        int bwritten;
        int bleft;
        int optlen;
        int actsize;

        *errorp = 0;
        bwritten = 0;
        bleft = len - 1;        /* leave room for trailing \0 */

        /*
         * Checks the size of the string. If it contains
         * any data, then we will append the new flags to
         * it.
         */
        actsize = strlen(buf);
        if (actsize > 0)
                buf += actsize;

        /* Default flags if no flags passed */
        if (optp == NULL)
                optp = optnames;

        if (bleft < 0) {        /* degenerate case, 0-length buffer */
                *errorp = EINVAL;
                return(0);
        }

        for (; flags && optp->o_opt; ++optp) {
                if ((flags & optp->o_opt) == 0)
                        continue;
                optlen = strlen(optp->o_name);
                if (bwritten || actsize > 0) {
                        if (bleft < 2) {
                                *errorp = ENOSPC;
                                break;
                        }
                        buf[bwritten++] = ',';
                        buf[bwritten++] = ' ';
                        bleft -= 2;
                }
                if (bleft < optlen) {
                        *errorp = ENOSPC;
                        break;
                }
                bcopy(optp->o_name, buf + bwritten, optlen);
                bwritten += optlen;
                bleft -= optlen;
                flags &= ~optp->o_opt;
        }

        /*
         * Space already reserved for trailing \0
         */
        buf[bwritten] = 0;
        return (bwritten);
}

/*
 * Build hash lists of net addresses and hang them off the mount point.
 * Called by ufs_mount() to set up the lists of export addresses.
 */
static int
vfs_hang_addrlist(struct mount *mp, struct netexport *nep,
                const struct export_args *argp)
{
        struct netcred *np;
        struct radix_node_head *rnh;
        int i;
        struct radix_node *rn;
        struct sockaddr *saddr, *smask = NULL;
        int error;

        if (argp->ex_addrlen == 0) {
                if (mp->mnt_flag & MNT_DEFEXPORTED)
                        return (EPERM);
                np = &nep->ne_defexported;
                np->netc_exflags = argp->ex_flags;
                np->netc_anon = argp->ex_anon;
                np->netc_anon.cr_ref = 1;
                mp->mnt_flag |= MNT_DEFEXPORTED;
                return (0);
        }

        if (argp->ex_addrlen < 0 || argp->ex_addrlen > MLEN)
                return (EINVAL);
        if (argp->ex_masklen < 0 || argp->ex_masklen > MLEN)
                return (EINVAL);

        i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen;
        np = (struct netcred *)kmalloc(i, M_NETCRED, M_WAITOK | M_ZERO);
        saddr = (struct sockaddr *) (np + 1);
        if ((error = copyin(argp->ex_addr, (caddr_t) saddr, argp->ex_addrlen)))
                goto out;
        if (saddr->sa_len > argp->ex_addrlen)
                saddr->sa_len = argp->ex_addrlen;
        if (argp->ex_masklen) {
                smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen);
                error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen);
                if (error)
                        goto out;
                if (smask->sa_len > argp->ex_masklen)
                        smask->sa_len = argp->ex_masklen;
        }
        NE_LOCK(nep);
        if (nep->ne_maskhead == NULL) {
                if (!rn_inithead(&nep->ne_maskhead, NULL, 0)) {
                        error = ENOBUFS;
                        goto out;
                }
        }
        if ((rnh = vfs_create_addrlist_af(saddr->sa_family, nep)) == NULL) {
                error = ENOBUFS;
                goto out;
        }
        rn = rnh->rnh_addaddr(saddr, smask, rnh, np->netc_rnodes);
        NE_UNLOCK(nep);
        if (rn == NULL || np != (struct netcred *)rn) { /* already exists */
                error = EPERM;
                goto out;
        }
        np->netc_exflags = argp->ex_flags;
        np->netc_anon = argp->ex_anon;
        np->netc_anon.cr_ref = 1;
        return (0);

out:
        kfree(np, M_NETCRED);
        return (error);
}

/*
 * Free netcred structures installed in the netexport
 */
static void
vfs_free_netcred(struct radix_node *rn)
{
        struct netcred *np;

        np = (struct netcred *)rn;
        kfree(np, M_NETCRED);
}

static struct radix_node_head *
vfs_create_addrlist_af(int af, struct netexport *nep)
{
        struct radix_node_head *rnh = NULL;
#if defined(INET) || defined(INET6)
        struct radix_node_head *maskhead = nep->ne_maskhead;
        int off;
#endif

        NE_ASSERT_LOCKED(nep);
#if defined(INET) || defined(INET6)
        KKASSERT(maskhead != NULL);
#endif
        switch (af) {
#ifdef INET
        case AF_INET:
                if ((rnh = nep->ne_inethead) == NULL) {
                        off = offsetof(struct sockaddr_in, sin_addr);
                        if (!rn_inithead(&rnh, maskhead, off))
                                return (NULL);
                        nep->ne_inethead = rnh;
                }
                break;
#endif
#ifdef INET6
        case AF_INET6:
                if ((rnh = nep->ne_inet6head) == NULL) {
                        off = offsetof(struct sockaddr_in6, sin6_addr);
                        if (!rn_inithead(&rnh, maskhead, off))
                                return (NULL);
                        nep->ne_inet6head = rnh;
                }
                break;
#endif
        }
        return (rnh);
}

/*
 * Free the net address hash lists that are hanging off the mount points.
 */
static void
vfs_free_addrlist(struct netexport *nep)
{
        NE_LOCK(nep);
        if (nep->ne_inethead != NULL) {
                rn_flush(nep->ne_inethead, vfs_free_netcred);
                rn_freehead(nep->ne_inethead);
                nep->ne_inethead = NULL;
        }
        if (nep->ne_inet6head != NULL) {
                rn_flush(nep->ne_inet6head, vfs_free_netcred);
                rn_freehead(nep->ne_inet6head);
                nep->ne_inet6head = NULL;
        }
        if (nep->ne_maskhead != NULL) {
                rn_flush(nep->ne_maskhead, rn_freemask);
                rn_freehead(nep->ne_maskhead);
                nep->ne_maskhead = NULL;
        }
        NE_UNLOCK(nep);
}

int
vfs_export(struct mount *mp, struct netexport *nep,
           const struct export_args *argp)
{
        int error;

        if (argp->ex_flags & MNT_DELEXPORT) {
                if (mp->mnt_flag & MNT_EXPUBLIC) {
                        vfs_setpublicfs(NULL, NULL, NULL);
                        mp->mnt_flag &= ~MNT_EXPUBLIC;
                }
                vfs_free_addrlist(nep);
                mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED);
        }
        if (argp->ex_flags & MNT_EXPORTED) {
                if (argp->ex_flags & MNT_EXPUBLIC) {
                        if ((error = vfs_setpublicfs(mp, nep, argp)) != 0)
                                return (error);
                        mp->mnt_flag |= MNT_EXPUBLIC;
                }
                if ((error = vfs_hang_addrlist(mp, nep, argp)))
                        return (error);
                mp->mnt_flag |= MNT_EXPORTED;
        }
        return (0);
}


/*
 * Set the publicly exported filesystem (WebNFS). Currently, only
 * one public filesystem is possible in the spec (RFC 2054 and 2055)
 */
int
vfs_setpublicfs(struct mount *mp, struct netexport *nep,
                const struct export_args *argp)
{
        int error;
        struct vnode *rvp;
        char *cp;

        /*
         * mp == NULL -> invalidate the current info, the FS is
         * no longer exported. May be called from either vfs_export
         * or unmount, so check if it hasn't already been done.
         */
        if (mp == NULL) {
                if (nfs_pub.np_valid) {
                        nfs_pub.np_valid = 0;
                        if (nfs_pub.np_index != NULL) {
                                kfree(nfs_pub.np_index, M_TEMP);
                                nfs_pub.np_index = NULL;
                        }
                }
                return (0);
        }

        /*
         * Only one allowed at a time.
         */
        if (nfs_pub.np_valid != 0 && mp != nfs_pub.np_mount)
                return (EBUSY);

        /*
         * Get real filehandle for root of exported FS.
         */
        bzero((caddr_t)&nfs_pub.np_handle, sizeof(nfs_pub.np_handle));
        nfs_pub.np_handle.fh_fsid = mp->mnt_stat.f_fsid;

        if ((error = VFS_ROOT(mp, &rvp)))
                return (error);

        if ((error = VFS_VPTOFH(rvp, &nfs_pub.np_handle.fh_fid)))
                return (error);

        vput(rvp);

        /*
         * If an indexfile was specified, pull it in.
         */
        if (argp->ex_indexfile != NULL) {
                int namelen;

                error = vn_get_namelen(rvp, &namelen);
                if (error)
                        return (error);
                nfs_pub.np_index = kmalloc(namelen, M_TEMP, M_WAITOK);
                error = copyinstr(argp->ex_indexfile, nfs_pub.np_index,
                    namelen, NULL);
                if (!error) {
                        /*
                         * Check for illegal filenames.
                         */
                        for (cp = nfs_pub.np_index; *cp; cp++) {
                                if (*cp == '/') {
                                        error = EINVAL;
                                        break;
                                }
                        }
                }
                if (error) {
                        kfree(nfs_pub.np_index, M_TEMP);
                        return (error);
                }
        }

        nfs_pub.np_mount = mp;
        nfs_pub.np_valid = 1;
        return (0);
}

struct netcred *
vfs_export_lookup(struct mount *mp, struct netexport *nep,
                struct sockaddr *nam)
{
        struct netcred *np;
        struct radix_node_head *rnh;
        struct sockaddr *saddr;

        np = NULL;
        if (mp->mnt_flag & MNT_EXPORTED) {
                /*
                 * Lookup in the export list first.
                 */
                NE_LOCK(nep);
                if (nam != NULL) {
                        saddr = nam;
                        switch (saddr->sa_family) {
#ifdef INET
                        case AF_INET:
                                rnh = nep->ne_inethead;
                                break;
#endif
#ifdef INET6
                        case AF_INET6:
                                rnh = nep->ne_inet6head;
                                break;
#endif
                        default:
                                rnh = NULL;
                        }
                        if (rnh != NULL) {
                                np = (struct netcred *)
                                        rnh->rnh_matchaddr(saddr, rnh);
                                if (np && np->netc_rnodes->rn_flags & RNF_ROOT)
                                        np = NULL;
                        }
                }
                NE_UNLOCK(nep);
                /*
                 * If no address match, use the default if it exists.
                 */
                if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED)
                        np = &nep->ne_defexported;
        }
        return (np);
}

/*
 * perform msync on all vnodes under a mount point.  The mount point must
 * be locked.  This code is also responsible for lazy-freeing unreferenced
 * vnodes whos VM objects no longer contain pages.
 *
 * NOTE: MNT_WAIT still skips vnodes in the VXLOCK state.
 *
 * NOTE: XXX VOP_PUTPAGES and friends requires that the vnode be locked,
 * but vnode_pager_putpages() doesn't lock the vnode.  We have to do it
 * way up in this high level function.
 */
static int vfs_msync_scan1(struct mount *mp, struct vnode *vp, void *data);
static int vfs_msync_scan2(struct mount *mp, struct vnode *vp, void *data);

void
vfs_msync(struct mount *mp, int flags)
{
        int vmsc_flags;

        /*
         * tmpfs sets this flag to prevent msync(), sync, and the
         * filesystem periodic syncer from trying to flush VM pages
         * to swap.  Only pure memory pressure flushes tmpfs VM pages
         * to swap.
         */
        if (mp->mnt_kern_flag & MNTK_NOMSYNC)
                return;

        /*
         * Ok, scan the vnodes for work.  If the filesystem is using the
         * syncer thread feature we can use vsyncscan() instead of
         * vmntvnodescan(), which is much faster.
         */
        vmsc_flags = VMSC_GETVP;
        if (flags != MNT_WAIT)
                vmsc_flags |= VMSC_NOWAIT;

        if (mp->mnt_kern_flag & MNTK_THR_SYNC) {
                vsyncscan(mp, vmsc_flags, vfs_msync_scan2,
                          (void *)(intptr_t)flags);
        } else {
                vmntvnodescan(mp, vmsc_flags,
                              vfs_msync_scan1, vfs_msync_scan2,
                              (void *)(intptr_t)flags);
        }
}

/*
 * scan1 is a fast pre-check.  There could be hundreds of thousands of
 * vnodes, we cannot afford to do anything heavy weight until we have a
 * fairly good indication that there is work to do.
 *
 * The new namecache holds the vnode for each v_namecache association
 * so allow these refs.
 */
static
int
vfs_msync_scan1(struct mount *mp, struct vnode *vp, void *data)
{
        int flags = (int)(intptr_t)data;

        if ((vp->v_flag & VRECLAIMED) == 0) {
                if (vp->v_auxrefs == vp->v_namecache_count &&
                    VREFCNT(vp) <= 0 && vp->v_object) {
                        return(0);      /* call scan2 */
                }
                if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
                    (vp->v_flag & VOBJDIRTY) &&
                    (flags == MNT_WAIT || vn_islocked(vp) == 0)) {
                        return(0);      /* call scan2 */
                }
        }

        /*
         * do not call scan2, continue the loop
         */
        return(-1);
}

/*
 * This callback is handed a locked vnode.
 */
static
int
vfs_msync_scan2(struct mount *mp, struct vnode *vp, void *data)
{
        vm_object_t obj;
        int flags = (int)(intptr_t)data;
        int opcflags;

        if (vp->v_flag & VRECLAIMED)
                return(0);

        if ((mp->mnt_flag & MNT_RDONLY) == 0 && (vp->v_flag & VOBJDIRTY)) {
                if ((obj = vp->v_object) != NULL) {
                        if (flags == MNT_WAIT) {
                                /*
                                 * VFS_MSYNC is called with MNT_WAIT when
                                 * unmounting.
                                 */
                                opcflags = OBJPC_SYNC;
                        } else if (vp->v_writecount || obj->ref_count) {
                                /*
                                 * VFS_MSYNC is otherwise called via the
                                 * periodic filesystem sync or the 'sync'
                                 * command.  Honor MADV_NOSYNC / MAP_NOSYNC
                                 * if the file is open for writing or memory
                                 * mapped.  Pages flagged PG_NOSYNC will not
                                 * be automatically flushed at this time.
                                 *
                                 * The obj->ref_count test is not perfect
                                 * since temporary refs may be present, but
                                 * the periodic filesystem sync will ultimately
                                 * catch it if the file is not open and not
                                 * mapped.
                                 */
                                opcflags = OBJPC_NOSYNC;
                        } else {
                                /*
                                 * If the file is no longer open for writing
                                 * and also no longer mapped, do not honor
                                 * MAP_NOSYNC.  That is, fully synchronize
                                 * the file.
                                 *
                                 * This still occurs on the periodic fs sync,
                                 * so frontend programs which turn the file
                                 * over quickly enough can still avoid the
                                 * sync, but ultimately we do want to flush
                                 * even MADV_NOSYNC pages once it is no longer
                                 * mapped or open for writing.
                                 */
                                opcflags = 0;
                        }
                        vm_object_page_clean(obj, 0, 0, opcflags);
                }
        }
        return(0);
}

/*
 * Wake up anyone interested in vp because it is being revoked.
 */
void
vn_gone(struct vnode *vp)
{
        lwkt_gettoken(&vp->v_token);
        KNOTE(&vp->v_pollinfo.vpi_kqinfo.ki_note, NOTE_REVOKE);
        lwkt_reltoken(&vp->v_token);
}

/*
 * extract the cdev_t from a VBLK or VCHR.  The vnode must have been opened
 * (or v_rdev might be NULL).
 */
cdev_t
vn_todev(struct vnode *vp)
{
        if (vp->v_type != VBLK && vp->v_type != VCHR)
                return (NULL);
        KKASSERT(vp->v_rdev != NULL);
        return (vp->v_rdev);
}

/*
 * Check if vnode represents a disk device.  The vnode does not need to be
 * opened.
 *
 * MPALMOSTSAFE
 */
int
vn_isdisk(struct vnode *vp, int *errp)
{
        cdev_t dev;

        if (vp->v_type != VCHR) {
                if (errp != NULL)
                        *errp = ENOTBLK;
                return (0);
        }

        dev = vp->v_rdev;

        if (dev == NULL) {
                if (errp != NULL)
                        *errp = ENXIO;
                return (0);
        }
        if (dev_is_good(dev) == 0) {
                if (errp != NULL)
                        *errp = ENXIO;
                return (0);
        }
        if ((dev_dflags(dev) & D_DISK) == 0) {
                if (errp != NULL)
                        *errp = ENOTBLK;
                return (0);
        }
        if (errp != NULL)
                *errp = 0;
        return (1);
}

int
vn_get_namelen(struct vnode *vp, int *namelen)
{
        int error;
        register_t retval[2];

        error = VOP_PATHCONF(vp, _PC_NAME_MAX, retval);
        if (error)
                return (error);
        *namelen = (int)retval[0];
        return (0);
}

int
vop_write_dirent(int *error, struct uio *uio, ino_t d_ino, uint8_t d_type,
                uint16_t d_namlen, const char *d_name)
{
        struct dirent *dp;
        size_t len;

        len = _DIRENT_RECLEN(d_namlen);
        if (len > uio->uio_resid)
                return(1);

        dp = kmalloc(len, M_TEMP, M_WAITOK | M_ZERO);

        dp->d_ino = d_ino;
        dp->d_namlen = d_namlen;
        dp->d_type = d_type;
        bcopy(d_name, dp->d_name, d_namlen);

        *error = uiomove((caddr_t)dp, len, uio);

        kfree(dp, M_TEMP);

        return(0);
}

void
vn_mark_atime(struct vnode *vp, struct thread *td)
{
        struct proc *p = td->td_proc;
        struct ucred *cred = p ? p->p_ucred : proc0.p_ucred;

        if ((vp->v_mount->mnt_flag & (MNT_NOATIME | MNT_RDONLY)) == 0) {
                VOP_MARKATIME(vp, cred);
        }
}

/*
 * Calculate the number of entries in an inode-related chained hash table.
 * With today's memory sizes, maxvnodes can wind up being a very large
 * number.  There is no reason to waste memory, so tolerate some stacking.
 */
int
vfs_inodehashsize(void)
{
        int hsize;

        hsize = 32;
        while (hsize < maxvnodes)
                hsize <<= 1;
        while (hsize > maxvnodes * 2)
                hsize >>= 1;            /* nominal 2x stacking */

        if (maxvnodes > 1024 * 1024)
                hsize >>= 1;            /* nominal 8x stacking */

        if (maxvnodes > 128 * 1024)
                hsize >>= 1;            /* nominal 4x stacking */

        if (hsize < 16)
                hsize = 16;

        return hsize;
}

union _qcvt {
        quad_t qcvt;
        int32_t val[2];
};

#define SETHIGH(q, h) { \
        union _qcvt tmp; \
        tmp.qcvt = (q); \
        tmp.val[_QUAD_HIGHWORD] = (h); \
        (q) = tmp.qcvt; \
}
#define SETLOW(q, l) { \
        union _qcvt tmp; \
        tmp.qcvt = (q); \
        tmp.val[_QUAD_LOWWORD] = (l); \
        (q) = tmp.qcvt; \
}

u_quad_t
init_va_filerev(void)
{
        struct timeval tv;
        u_quad_t ret = 0;

        getmicrouptime(&tv);
        SETHIGH(ret, tv.tv_sec);
        SETLOW(ret, tv.tv_usec * 4294);

        return ret;
}

/*
 * Set default timestamp_precision.  If hz is reasonably high we go for
 * performance and limit vfs timestamps to microseconds with tick resolution.
 * If hz is too low, however, we lose a bit of performance to get a more
 * precise timestamp, because the mtime/ctime granularity might just be too
 * rough otherwise (for make and Makefile's, for example).
 */
static void
vfs_ts_prec_init(void *dummy)
{
        if (timestamp_precision < 0) {
                if (hz >= 100)
                        timestamp_precision = TSP_USEC;
                else
                        timestamp_precision = TSP_USEC_PRECISE;
        }
}
SYSINIT(vfs_ts_prec_init, SI_SUB_VFS, SI_ORDER_ANY, vfs_ts_prec_init, NULL);