kernel - Remove SMP bottlenecks on uidinfo, descriptors, and lockf

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

/*
 * Copyright (c) 2005-2018 The DragonFly Project.  All rights reserved.
 * 
 * This code is derived from software contributed to The DragonFly Project
 * by Jeffrey Hsu and Matthew Dillon.
 * 
 * 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 DragonFly Project 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 COPYRIGHT HOLDERS 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
 * COPYRIGHT HOLDERS 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.
 *
 *
 * Copyright (c) 1982, 1986, 1989, 1991, 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.
 *
 *      @(#)kern_descrip.c      8.6 (Berkeley) 4/19/94
 * $FreeBSD: src/sys/kern/kern_descrip.c,v 1.81.2.19 2004/02/28 00:43:31 tegge Exp $
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/sysproto.h>
#include <sys/conf.h>
#include <sys/device.h>
#include <sys/file.h>
#include <sys/filedesc.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/vnode.h>
#include <sys/proc.h>
#include <sys/nlookup.h>
#include <sys/stat.h>
#include <sys/filio.h>
#include <sys/fcntl.h>
#include <sys/unistd.h>
#include <sys/resourcevar.h>
#include <sys/event.h>
#include <sys/kern_syscall.h>
#include <sys/kcore.h>
#include <sys/kinfo.h>
#include <sys/un.h>
#include <sys/objcache.h>

#include <vm/vm.h>
#include <vm/vm_extern.h>

#include <sys/thread2.h>
#include <sys/file2.h>
#include <sys/spinlock2.h>

static void fsetfd_locked(struct filedesc *fdp, struct file *fp, int fd);
static void fdreserve_locked (struct filedesc *fdp, int fd0, int incr);
static struct file *funsetfd_locked (struct filedesc *fdp, int fd);
static void ffree(struct file *fp);

static MALLOC_DEFINE(M_FILEDESC, "file desc", "Open file descriptor table");
static MALLOC_DEFINE(M_FILEDESC_TO_LEADER, "file desc to leader",
                     "file desc to leader structures");
MALLOC_DEFINE(M_FILE, "file", "Open file structure");
static MALLOC_DEFINE(M_SIGIO, "sigio", "sigio structures");

static struct krate krate_uidinfo = { .freq = 1 };

static   d_open_t  fdopen;
#define NUMFDESC 64

#define CDEV_MAJOR 22
static struct dev_ops fildesc_ops = {
        { "FD", 0, 0 },
        .d_open =       fdopen,
};

/*
 * Descriptor management.
 */
#ifndef NFILELIST_HEADS
#define NFILELIST_HEADS         257     /* primary number */
#endif

struct filelist_head {
        struct spinlock         spin;
        struct filelist         list;
} __cachealign;

static struct filelist_head     filelist_heads[NFILELIST_HEADS];

static int nfiles;              /* actual number of open files */
extern int cmask;       

struct lwkt_token revoke_token = LWKT_TOKEN_INITIALIZER(revoke_token);

static struct objcache          *file_objcache;

static struct objcache_malloc_args file_malloc_args = {
        .objsize        = sizeof(struct file),
        .mtype          = M_FILE
};

/*
 * Fixup fd_freefile and fd_lastfile after a descriptor has been cleared.
 *
 * must be called with fdp->fd_spin exclusively held
 */
static __inline
void
fdfixup_locked(struct filedesc *fdp, int fd)
{
        if (fd < fdp->fd_freefile) {
               fdp->fd_freefile = fd;
        }
        while (fdp->fd_lastfile >= 0 &&
               fdp->fd_files[fdp->fd_lastfile].fp == NULL &&
               fdp->fd_files[fdp->fd_lastfile].reserved == 0
        ) {
                --fdp->fd_lastfile;
        }
}

/*
 * Clear the fd thread caches for this fdnode.
 *
 * If match_fdc is NULL, all thread caches of fdn will be cleared.
 * The caller must hold fdp->fd_spin exclusively.  The threads caching
 * the descriptor do not have to be the current thread.  The (status)
 * argument is ignored.
 *
 * If match_fdc is not NULL, only the match_fdc's cache will be cleared.
 * The caller must hold fdp->fd_spin shared and match_fdc must match a
 * fdcache entry in curthread.  match_fdc has been locked by the caller
 * and had the specified (status).
 *
 * Since we are matching against a fp in the fdp (which must still be present
 * at this time), fp will have at least two refs on any match and we can
 * decrement the count trivially.
 */
static
void
fclearcache(struct fdnode *fdn, struct fdcache *match_fdc, int status)
{
        struct fdcache *fdc;
        struct file *fp;
        int i;

        /*
         * match_fdc == NULL    We are cleaning out all tdcache entries
         *                      for the fdn and hold fdp->fd_spin exclusively.
         *                      This can race against the target threads
         *                      cleaning out specific entries.
         *
         * match_fdc != NULL    We are cleaning out a specific tdcache
         *                      entry on behalf of the owning thread
         *                      and hold fdp->fd_spin shared.  The thread
         *                      has already locked the entry.  This cannot
         *                      race.
         */
        fp = fdn->fp;
        for (i = 0; i < NTDCACHEFD; ++i) {
                if ((fdc = fdn->tdcache[i]) == NULL)
                        continue;

                /*
                 * If match_fdc is non-NULL we are being asked to
                 * clear a specific fdc owned by curthread.  There must
                 * be exactly one match.  The caller has already locked
                 * the cache entry and will dispose of the lock after
                 * we return.
                 *
                 * Since we also have a shared lock on fdp, we
                 * can do this without atomic ops.
                 */
                if (match_fdc) {
                        if (fdc != match_fdc)
                                continue;
                        fdn->tdcache[i] = NULL;
                        KASSERT(fp == fdc->fp,
                                ("fclearcache(1): fp mismatch %p/%p\n",
                                fp, fdc->fp));
                        fdc->fp = NULL;
                        fdc->fd = -1;

                        /*
                         * status can be 0 or 2.  If 2 the ref is borrowed,
                         * if 0 the ref is not borrowed and we have to drop
                         * it.
                         */
                        if (status == 0)
                                atomic_add_int(&fp->f_count, -1);
                        fdn->isfull = 0;        /* heuristic */
                        return;
                }

                /*
                 * Otherwise we hold an exclusive spin-lock and can only
                 * race thread consumers borrowing cache entries.
                 *
                 * Acquire the lock and dispose of the entry.  We have to
                 * spin until we get the lock.
                 */
                for (;;) {
                        status = atomic_swap_int(&fdc->locked, 1);
                        if (status == 1) {      /* foreign lock, retry */
                                cpu_pause();
                                continue;
                        }
                        fdn->tdcache[i] = NULL;
                        KASSERT(fp == fdc->fp,
                                ("fclearcache(2): fp mismatch %p/%p\n",
                                fp, fdc->fp));
                        fdc->fp = NULL;
                        fdc->fd = -1;
                        if (status == 0)
                                atomic_add_int(&fp->f_count, -1);
                        fdn->isfull = 0;        /* heuristic */
                        atomic_swap_int(&fdc->locked, 0);
                        break;
                }
        }
        KKASSERT(match_fdc == NULL);
}

/*
 * Retrieve the fp for the specified fd given the specified file descriptor
 * table.  The fdp does not have to be owned by the current process.
 * If flags != -1, fp->f_flag must contain at least one of the flags.
 *
 * This function is not able to cache the fp.
 */
struct file *
holdfp_fdp(struct filedesc *fdp, int fd, int flag)
{
        struct file *fp;

        spin_lock_shared(&fdp->fd_spin);
        if (((u_int)fd) < fdp->fd_nfiles) {
                fp = fdp->fd_files[fd].fp;      /* can be NULL */
                if (fp) {
                        if ((fp->f_flag & flag) == 0 && flag != -1) {
                                fp = NULL;
                        } else {
                                fhold(fp);
                        }
                }
        } else {
                fp = NULL;
        }
        spin_unlock_shared(&fdp->fd_spin);

        return fp;
}

struct file *
holdfp_fdp_locked(struct filedesc *fdp, int fd, int flag)
{
        struct file *fp;

        if (((u_int)fd) < fdp->fd_nfiles) {
                fp = fdp->fd_files[fd].fp;      /* can be NULL */
                if (fp) {
                        if ((fp->f_flag & flag) == 0 && flag != -1) {
                                fp = NULL;
                        } else {
                                fhold(fp);
                        }
                }
        } else {
                fp = NULL;
        }
        return fp;
}

/*
 * Acquire the fp for the specified file descriptor, using the thread
 * cache if possible and caching it if possible.
 *
 * td must be the curren thread.
 */
static
struct file *
_holdfp_cache(thread_t td, int fd)
{
        struct filedesc *fdp;
        struct fdcache *fdc;
        struct fdcache *best;
        struct fdnode *fdn;
        struct file *fp;
        int status;
        int delta;
        int i;

        /*
         * Fast
         */
        for (fdc = &td->td_fdcache[0]; fdc < &td->td_fdcache[NFDCACHE]; ++fdc) {
                if (fdc->fd != fd || fdc->fp == NULL)
                        continue;
                status = atomic_swap_int(&fdc->locked, 1);

                /*
                 * If someone else has locked our cache entry they are in
                 * the middle of clearing it, skip the entry.
                 */
                if (status == 1)
                        continue;

                /*
                 * We have locked the entry, but if it no longer matches
                 * restore the previous state (0 or 2) and skip the entry.
                 */
                if (fdc->fd != fd || fdc->fp == NULL) {
                        atomic_swap_int(&fdc->locked, status);
                        continue;
                }

                /*
                 * We have locked a valid entry.  We can borrow the ref
                 * for a mode 0 entry.  We can get a valid fp for a mode
                 * 2 entry but not borrow the ref.
                 */
                if (status == 0) {
                        fp = fdc->fp;
                        fdc->lru = ++td->td_fdcache_lru;
                        atomic_swap_int(&fdc->locked, 2);

                        return fp;
                }
                if (status == 2) {
                        fp = fdc->fp;
                        fhold(fp);
                        fdc->lru = ++td->td_fdcache_lru;
                        atomic_swap_int(&fdc->locked, 2);

                        return fp;
                }
                KKASSERT(0);
        }

        /*
         * Lookup the descriptor the slow way.  This can contend against
         * modifying operations in a multi-threaded environment and cause
         * cache line ping ponging otherwise.
         */
        fdp = td->td_proc->p_fd;
        spin_lock_shared(&fdp->fd_spin);

        if (((u_int)fd) < fdp->fd_nfiles) {
                fp = fdp->fd_files[fd].fp;      /* can be NULL */
                if (fp) {
                        fhold(fp);
                        if (fdp->fd_files[fd].isfull == 0)
                                goto enter;
                }
        } else {
                fp = NULL;
        }
        spin_unlock_shared(&fdp->fd_spin);

        return fp;

        /*
         * We found a valid fp and held it, fdp is still shared locked.
         * Enter the fp into the per-thread cache.  Find the oldest entry
         * via lru, or an empty entry.
         *
         * Because fdp's spinlock is held (shared is fine), no other
         * thread should be in the middle of clearing our selected entry.
         */
enter:
        best = &td->td_fdcache[0];
        for (fdc = &td->td_fdcache[0]; fdc < &td->td_fdcache[NFDCACHE]; ++fdc) {
                if (fdc->fp == NULL) {
                        best = fdc;
                        break;
                }
                delta = fdc->lru - best->lru;
                if (delta < 0)
                        best = fdc;
        }

        /*
         * Replace best
         *
         * Don't enter into the cache if we cannot get the lock.
         */
        status = atomic_swap_int(&best->locked, 1);
        if (status == 1)
                goto done;

        /*
         * Clear the previous cache entry if present
         */
        if (best->fp) {
                KKASSERT(best->fd >= 0);
                fclearcache(&fdp->fd_files[best->fd], best, status);
        }

        /*
         * Create our new cache entry.  This entry is 'safe' until we tie
         * into the fdnode.  If we cannot tie in, we will clear the entry.
         */
        best->fd = fd;
        best->fp = fp;
        best->lru = ++td->td_fdcache_lru;
        best->locked = 2;                       /* borrowed ref */

        fdn = &fdp->fd_files[fd];
        for (i = 0; i < NTDCACHEFD; ++i) {
                if (fdn->tdcache[i] == NULL &&
                    atomic_cmpset_ptr((void **)&fdn->tdcache[i], NULL, best)) {
                        goto done;
                }
        }
        fdn->isfull = 1;                        /* no space */
        best->fd = -1;
        best->fp = NULL;
        best->locked = 0;
done:
        spin_unlock_shared(&fdp->fd_spin);

        return fp;
}

/*
 * Drop the file pointer and return to the thread cache if possible.
 *
 * Caller must not hold fdp's spin lock.
 * td must be the current thread.
 */
void
dropfp(thread_t td, int fd, struct file *fp)
{
        struct filedesc *fdp;
        struct fdcache *fdc;
        int status;

        fdp = td->td_proc->p_fd;

        /*
         * If our placeholder is still present we can re-cache the ref.
         *
         * Note that we can race an fclearcache().
         */
        for (fdc = &td->td_fdcache[0]; fdc < &td->td_fdcache[NFDCACHE]; ++fdc) {
                if (fdc->fp != fp || fdc->fd != fd)
                        continue;
                status = atomic_swap_int(&fdc->locked, 1);
                switch(status) {
                case 0:
                        /*
                         * Not in mode 2, fdrop fp without caching.
                         */
                        atomic_swap_int(&fdc->locked, 0);
                        break;
                case 1:
                        /*
                         * Not in mode 2, locked by someone else.
                         * fdrop fp without caching.
                         */
                        break;
                case 2:
                        /*
                         * Intact borrowed ref, return to mode 0
                         * indicating that we have returned the ref.
                         *
                         * Return the borrowed ref (2->1->0)
                         */
                        if (fdc->fp == fp && fdc->fd == fd) {
                                atomic_swap_int(&fdc->locked, 0);
                                return;
                        }
                        atomic_swap_int(&fdc->locked, 2);
                        break;
                }
        }

        /*
         * Failed to re-cache, drop the fp without caching.
         */
        fdrop(fp);
}

/*
 * Clear all descriptors cached in the per-thread fd cache for
 * the specified thread.
 *
 * Caller must not hold p_fd->spin.  This function will temporarily
 * obtain a shared spin lock.
 */
void
fexitcache(thread_t td)
{
        struct filedesc *fdp;
        struct fdcache *fdc;
        int status;
        int i;

        if (td->td_proc == NULL)
                return;
        fdp = td->td_proc->p_fd;
        if (fdp == NULL)
                return;

        /*
         * A shared lock is sufficient as the caller controls td and we
         * are only clearing td's cache.
         */
        spin_lock_shared(&fdp->fd_spin);
        for (i = 0; i < NFDCACHE; ++i) {
                fdc = &td->td_fdcache[i];
                if (fdc->fp) {
                        status = atomic_swap_int(&fdc->locked, 1);
                        if (status == 1) {
                                cpu_pause();
                                --i;
                                continue;
                        }
                        if (fdc->fp) {
                                KKASSERT(fdc->fd >= 0);
                                fclearcache(&fdp->fd_files[fdc->fd], fdc,
                                            status);
                        }
                        atomic_swap_int(&fdc->locked, 0);
                }
        }
        spin_unlock_shared(&fdp->fd_spin);
}

static __inline struct filelist_head *
fp2filelist(const struct file *fp)
{
        u_int i;

        i = (u_int)(uintptr_t)fp % NFILELIST_HEADS;
        return &filelist_heads[i];
}

static __inline
struct plimit *
readplimits(struct proc *p)
{
        thread_t td = curthread;
        struct plimit *limit;

        limit = td->td_limit;
        if (limit != p->p_limit) {
                spin_lock_shared(&p->p_spin);
                limit = p->p_limit;
                atomic_add_int(&limit->p_refcnt, 1);
                spin_unlock_shared(&p->p_spin);
                if (td->td_limit)
                        plimit_free(td->td_limit);
                td->td_limit = limit;
        }
        return limit;
}

/*
 * System calls on descriptors.
 */
int
sys_getdtablesize(struct getdtablesize_args *uap) 
{
        struct proc *p = curproc;
        struct plimit *limit = readplimits(p);
        int dtsize;

        if (limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur > INT_MAX)
                dtsize = INT_MAX;
        else
                dtsize = (int)limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur;

        if (dtsize > maxfilesperproc)
                dtsize = maxfilesperproc;
        if (dtsize < minfilesperproc)
                dtsize = minfilesperproc;
        if (p->p_ucred->cr_uid && dtsize > maxfilesperuser)
                dtsize = maxfilesperuser;
        uap->sysmsg_result = dtsize;
        return (0);
}

/*
 * Duplicate a file descriptor to a particular value.
 *
 * note: keep in mind that a potential race condition exists when closing
 * descriptors from a shared descriptor table (via rfork).
 */
int
sys_dup2(struct dup2_args *uap)
{
        int error;
        int fd = 0;

        error = kern_dup(DUP_FIXED, uap->from, uap->to, &fd);
        uap->sysmsg_fds[0] = fd;

        return (error);
}

/*
 * Duplicate a file descriptor.
 */
int
sys_dup(struct dup_args *uap)
{
        int error;
        int fd = 0;

        error = kern_dup(DUP_VARIABLE, uap->fd, 0, &fd);
        uap->sysmsg_fds[0] = fd;

        return (error);
}

/*
 * MPALMOSTSAFE - acquires mplock for fp operations
 */
int
kern_fcntl(int fd, int cmd, union fcntl_dat *dat, struct ucred *cred)
{
        struct thread *td = curthread;
        struct proc *p = td->td_proc;
        struct file *fp;
        struct vnode *vp;
        u_int newmin;
        u_int oflags;
        u_int nflags;
        int closedcounter;
        int tmp, error, flg = F_POSIX;

        KKASSERT(p);

        /*
         * Operations on file descriptors that do not require a file pointer.
         */
        switch (cmd) {
        case F_GETFD:
                error = fgetfdflags(p->p_fd, fd, &tmp);
                if (error == 0)
                        dat->fc_cloexec = (tmp & UF_EXCLOSE) ? FD_CLOEXEC : 0;
                return (error);

        case F_SETFD:
                if (dat->fc_cloexec & FD_CLOEXEC)
                        error = fsetfdflags(p->p_fd, fd, UF_EXCLOSE);
                else
                        error = fclrfdflags(p->p_fd, fd, UF_EXCLOSE);
                return (error);
        case F_DUPFD:
                newmin = dat->fc_fd;
                error = kern_dup(DUP_VARIABLE | DUP_FCNTL, fd, newmin,
                    &dat->fc_fd);
                return (error);
        case F_DUPFD_CLOEXEC:
                newmin = dat->fc_fd;
                error = kern_dup(DUP_VARIABLE | DUP_CLOEXEC | DUP_FCNTL,
                    fd, newmin, &dat->fc_fd);
                return (error);
        case F_DUP2FD:
                newmin = dat->fc_fd;
                error = kern_dup(DUP_FIXED, fd, newmin, &dat->fc_fd);
                return (error);
        case F_DUP2FD_CLOEXEC:
                newmin = dat->fc_fd;
                error = kern_dup(DUP_FIXED | DUP_CLOEXEC, fd, newmin,
                                 &dat->fc_fd);
                return (error);
        default:
                break;
        }

        /*
         * Operations on file pointers
         */
        closedcounter = p->p_fd->fd_closedcounter;
        if ((fp = holdfp(td, fd, -1)) == NULL)
                return (EBADF);

        switch (cmd) {
        case F_GETFL:
                dat->fc_flags = OFLAGS(fp->f_flag);
                error = 0;
                break;

        case F_SETFL:
                oflags = fp->f_flag;
                nflags = FFLAGS(dat->fc_flags & ~O_ACCMODE) & FCNTLFLAGS;
                nflags |= oflags & ~FCNTLFLAGS;

                error = 0;
                if (((nflags ^ oflags) & O_APPEND) && (oflags & FAPPENDONLY))
                        error = EINVAL;
                if (error == 0 && ((nflags ^ oflags) & FASYNC)) {
                        tmp = nflags & FASYNC;
                        error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp,
                                         cred, NULL);
                }

                /*
                 * If no error, must be atomically set.
                 */
                while (error == 0) {
                        oflags = fp->f_flag;
                        cpu_ccfence();
                        nflags = (oflags & ~FCNTLFLAGS) | (nflags & FCNTLFLAGS);
                        if (atomic_cmpset_int(&fp->f_flag, oflags, nflags))
                                break;
                        cpu_pause();
                }
                break;

        case F_GETOWN:
                error = fo_ioctl(fp, FIOGETOWN, (caddr_t)&dat->fc_owner,
                                 cred, NULL);
                break;

        case F_SETOWN:
                error = fo_ioctl(fp, FIOSETOWN, (caddr_t)&dat->fc_owner,
                                 cred, NULL);
                break;

        case F_SETLKW:
                flg |= F_WAIT;
                /* Fall into F_SETLK */

        case F_SETLK:
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        break;
                }
                vp = (struct vnode *)fp->f_data;

                /*
                 * copyin/lockop may block
                 */
                if (dat->fc_flock.l_whence == SEEK_CUR)
                        dat->fc_flock.l_start += fp->f_offset;

                switch (dat->fc_flock.l_type) {
                case F_RDLCK:
                        if ((fp->f_flag & FREAD) == 0) {
                                error = EBADF;
                                break;
                        }
                        if (p->p_leader->p_advlock_flag == 0)
                                p->p_leader->p_advlock_flag = 1;
                        error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_SETLK,
                                            &dat->fc_flock, flg);
                        break;
                case F_WRLCK:
                        if ((fp->f_flag & FWRITE) == 0) {
                                error = EBADF;
                                break;
                        }
                        if (p->p_leader->p_advlock_flag == 0)
                                p->p_leader->p_advlock_flag = 1;
                        error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_SETLK,
                                            &dat->fc_flock, flg);
                        break;
                case F_UNLCK:
                        error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_UNLCK,
                                            &dat->fc_flock, F_POSIX);
                        break;
                default:
                        error = EINVAL;
                        break;
                }

                /*
                 * It is possible to race a close() on the descriptor while
                 * we were blocked getting the lock.  If this occurs the
                 * close might not have caught the lock.
                 */
                if (checkfdclosed(td, p->p_fd, fd, fp, closedcounter)) {
                        dat->fc_flock.l_whence = SEEK_SET;
                        dat->fc_flock.l_start = 0;
                        dat->fc_flock.l_len = 0;
                        dat->fc_flock.l_type = F_UNLCK;
                        VOP_ADVLOCK(vp, (caddr_t)p->p_leader,
                                    F_UNLCK, &dat->fc_flock, F_POSIX);
                }
                break;

        case F_GETLK:
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        break;
                }
                vp = (struct vnode *)fp->f_data;
                /*
                 * copyin/lockop may block
                 */
                if (dat->fc_flock.l_type != F_RDLCK &&
                    dat->fc_flock.l_type != F_WRLCK &&
                    dat->fc_flock.l_type != F_UNLCK) {
                        error = EINVAL;
                        break;
                }
                if (dat->fc_flock.l_whence == SEEK_CUR)
                        dat->fc_flock.l_start += fp->f_offset;
                error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_GETLK,
                                    &dat->fc_flock, F_POSIX);
                break;
        default:
                error = EINVAL;
                break;
        }

        fdrop(fp);
        return (error);
}

/*
 * The file control system call.
 */
int
sys_fcntl(struct fcntl_args *uap)
{
        union fcntl_dat dat;
        int error;

        switch (uap->cmd) {
        case F_DUPFD:
        case F_DUP2FD:
        case F_DUPFD_CLOEXEC:
        case F_DUP2FD_CLOEXEC:
                dat.fc_fd = uap->arg;
                break;
        case F_SETFD:
                dat.fc_cloexec = uap->arg;
                break;
        case F_SETFL:
                dat.fc_flags = uap->arg;
                break;
        case F_SETOWN:
                dat.fc_owner = uap->arg;
                break;
        case F_SETLKW:
        case F_SETLK:
        case F_GETLK:
                error = copyin((caddr_t)uap->arg, &dat.fc_flock,
                               sizeof(struct flock));
                if (error)
                        return (error);
                break;
        }

        error = kern_fcntl(uap->fd, uap->cmd, &dat, curthread->td_ucred);

        if (error == 0) {
                switch (uap->cmd) {
                case F_DUPFD:
                case F_DUP2FD:
                case F_DUPFD_CLOEXEC:
                case F_DUP2FD_CLOEXEC:
                        uap->sysmsg_result = dat.fc_fd;
                        break;
                case F_GETFD:
                        uap->sysmsg_result = dat.fc_cloexec;
                        break;
                case F_GETFL:
                        uap->sysmsg_result = dat.fc_flags;
                        break;
                case F_GETOWN:
                        uap->sysmsg_result = dat.fc_owner;
                        break;
                case F_GETLK:
                        error = copyout(&dat.fc_flock, (caddr_t)uap->arg,
                            sizeof(struct flock));
                        break;
                }
        }

        return (error);
}

/*
 * Common code for dup, dup2, and fcntl(F_DUPFD).
 *
 * There are four type flags: DUP_FCNTL, DUP_FIXED, DUP_VARIABLE, and
 * DUP_CLOEXEC.
 *
 * DUP_FCNTL is for handling EINVAL vs. EBADF differences between
 * fcntl()'s F_DUPFD and F_DUPFD_CLOEXEC and dup2() (per POSIX).
 * The next two flags are mutually exclusive, and the fourth is optional.
 * DUP_FIXED tells kern_dup() to destructively dup over an existing file
 * descriptor if "new" is already open.  DUP_VARIABLE tells kern_dup()
 * to find the lowest unused file descriptor that is greater than or
 * equal to "new".  DUP_CLOEXEC, which works with either of the first
 * two flags, sets the close-on-exec flag on the "new" file descriptor.
 */
int
kern_dup(int flags, int old, int new, int *res)
{
        struct thread *td = curthread;
        struct proc *p = td->td_proc;
        struct plimit *limit = readplimits(p);
        struct filedesc *fdp = p->p_fd;
        struct file *fp;
        struct file *delfp;
        int oldflags;
        int holdleaders;
        int dtsize;
        int error, newfd;

        /*
         * Verify that we have a valid descriptor to dup from and
         * possibly to dup to. When the new descriptor is out of
         * bounds, fcntl()'s F_DUPFD and F_DUPFD_CLOEXEC must
         * return EINVAL, while dup2() returns EBADF in
         * this case.
         *
         * NOTE: maxfilesperuser is not applicable to dup()
         */
retry:
        if (limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur > INT_MAX)
                dtsize = INT_MAX;
        else
                dtsize = (int)limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur;
        if (dtsize > maxfilesperproc)
                dtsize = maxfilesperproc;
        if (dtsize < minfilesperproc)
                dtsize = minfilesperproc;

        if (new < 0 || new > dtsize)
                return (flags & DUP_FCNTL ? EINVAL : EBADF);

        spin_lock(&fdp->fd_spin);
        if ((unsigned)old >= fdp->fd_nfiles || fdp->fd_files[old].fp == NULL) {
                spin_unlock(&fdp->fd_spin);
                return (EBADF);
        }
        if ((flags & DUP_FIXED) && old == new) {
                *res = new;
                if (flags & DUP_CLOEXEC)
                        fdp->fd_files[new].fileflags |= UF_EXCLOSE;
                spin_unlock(&fdp->fd_spin);
                return (0);
        }
        fp = fdp->fd_files[old].fp;
        oldflags = fdp->fd_files[old].fileflags;
        fhold(fp);

        /*
         * Allocate a new descriptor if DUP_VARIABLE, or expand the table
         * if the requested descriptor is beyond the current table size.
         *
         * This can block.  Retry if the source descriptor no longer matches
         * or if our expectation in the expansion case races.
         *
         * If we are not expanding or allocating a new decriptor, then reset
         * the target descriptor to a reserved state so we have a uniform
         * setup for the next code block.
         */
        if ((flags & DUP_VARIABLE) || new >= fdp->fd_nfiles) {
                spin_unlock(&fdp->fd_spin);
                error = fdalloc(p, new, &newfd);
                spin_lock(&fdp->fd_spin);
                if (error) {
                        spin_unlock(&fdp->fd_spin);
                        fdrop(fp);
                        return (error);
                }
                /*
                 * Check for ripout
                 */
                if (old >= fdp->fd_nfiles || fdp->fd_files[old].fp != fp) {
                        fsetfd_locked(fdp, NULL, newfd);
                        spin_unlock(&fdp->fd_spin);
                        fdrop(fp);
                        goto retry;
                }
                /*
                 * Check for expansion race
                 */
                if ((flags & DUP_VARIABLE) == 0 && new != newfd) {
                        fsetfd_locked(fdp, NULL, newfd);
                        spin_unlock(&fdp->fd_spin);
                        fdrop(fp);
                        goto retry;
                }
                /*
                 * Check for ripout, newfd reused old (this case probably
                 * can't occur).
                 */
                if (old == newfd) {
                        fsetfd_locked(fdp, NULL, newfd);
                        spin_unlock(&fdp->fd_spin);
                        fdrop(fp);
                        goto retry;
                }
                new = newfd;
                delfp = NULL;
        } else {
                if (fdp->fd_files[new].reserved) {
                        spin_unlock(&fdp->fd_spin);
                        fdrop(fp);
                        kprintf("Warning: dup(): target descriptor %d is reserved, waiting for it to be resolved\n", new);
                        tsleep(fdp, 0, "fdres", hz);
                        goto retry;
                }

                /*
                 * If the target descriptor was never allocated we have
                 * to allocate it.  If it was we have to clean out the
                 * old descriptor.  delfp inherits the ref from the 
                 * descriptor table.
                 */
                ++fdp->fd_closedcounter;
                fclearcache(&fdp->fd_files[new], NULL, 0);
                ++fdp->fd_closedcounter;
                delfp = fdp->fd_files[new].fp;
                fdp->fd_files[new].fp = NULL;
                fdp->fd_files[new].reserved = 1;
                if (delfp == NULL) {
                        fdreserve_locked(fdp, new, 1);
                        if (new > fdp->fd_lastfile)
                                fdp->fd_lastfile = new;
                }

        }

        /*
         * NOTE: still holding an exclusive spinlock
         */

        /*
         * If a descriptor is being overwritten we may hve to tell 
         * fdfree() to sleep to ensure that all relevant process
         * leaders can be traversed in closef().
         */
        if (delfp != NULL && p->p_fdtol != NULL) {
                fdp->fd_holdleaderscount++;
                holdleaders = 1;
        } else {
                holdleaders = 0;
        }
        KASSERT(delfp == NULL || (flags & DUP_FIXED),
                ("dup() picked an open file"));

        /*
         * Duplicate the source descriptor, update lastfile.  If the new
         * descriptor was not allocated and we aren't replacing an existing
         * descriptor we have to mark the descriptor as being in use.
         *
         * The fd_files[] array inherits fp's hold reference.
         */
        fsetfd_locked(fdp, fp, new);
        if ((flags & DUP_CLOEXEC) != 0)
                fdp->fd_files[new].fileflags = oldflags | UF_EXCLOSE;
        else
                fdp->fd_files[new].fileflags = oldflags & ~UF_EXCLOSE;
        spin_unlock(&fdp->fd_spin);
        fdrop(fp);
        *res = new;

        /*
         * If we dup'd over a valid file, we now own the reference to it
         * and must dispose of it using closef() semantics (as if a
         * close() were performed on it).
         */
        if (delfp) {
                if (SLIST_FIRST(&delfp->f_klist))
                        knote_fdclose(delfp, fdp, new);
                closef(delfp, p);
                if (holdleaders) {
                        spin_lock(&fdp->fd_spin);
                        fdp->fd_holdleaderscount--;
                        if (fdp->fd_holdleaderscount == 0 &&
                            fdp->fd_holdleaderswakeup != 0) {
                                fdp->fd_holdleaderswakeup = 0;
                                spin_unlock(&fdp->fd_spin);
                                wakeup(&fdp->fd_holdleaderscount);
                        } else {
                                spin_unlock(&fdp->fd_spin);
                        }
                }
        }
        return (0);
}

/*
 * If sigio is on the list associated with a process or process group,
 * disable signalling from the device, remove sigio from the list and
 * free sigio.
 */
void
funsetown(struct sigio **sigiop)
{
        struct pgrp *pgrp;
        struct proc *p;
        struct sigio *sigio;

        if ((sigio = *sigiop) != NULL) {
                lwkt_gettoken(&sigio_token);    /* protect sigio */
                KKASSERT(sigiop == sigio->sio_myref);
                sigio = *sigiop;
                *sigiop = NULL;
                lwkt_reltoken(&sigio_token);
        }
        if (sigio == NULL)
                return;

        if (sigio->sio_pgid < 0) {
                pgrp = sigio->sio_pgrp;
                sigio->sio_pgrp = NULL;
                lwkt_gettoken(&pgrp->pg_token);
                SLIST_REMOVE(&pgrp->pg_sigiolst, sigio, sigio, sio_pgsigio);
                lwkt_reltoken(&pgrp->pg_token);
                pgrel(pgrp);
        } else /* if ((*sigiop)->sio_pgid > 0) */ {
                p = sigio->sio_proc;
                sigio->sio_proc = NULL;
                PHOLD(p);
                lwkt_gettoken(&p->p_token);
                SLIST_REMOVE(&p->p_sigiolst, sigio, sigio, sio_pgsigio);
                lwkt_reltoken(&p->p_token);
                PRELE(p);
        }
        crfree(sigio->sio_ucred);
        sigio->sio_ucred = NULL;
        kfree(sigio, M_SIGIO);
}

/*
 * Free a list of sigio structures.  Caller is responsible for ensuring
 * that the list is MPSAFE.
 */
void
funsetownlst(struct sigiolst *sigiolst)
{
        struct sigio *sigio;

        while ((sigio = SLIST_FIRST(sigiolst)) != NULL)
                funsetown(sigio->sio_myref);
}

/*
 * This is common code for FIOSETOWN ioctl called by fcntl(fd, F_SETOWN, arg).
 *
 * After permission checking, add a sigio structure to the sigio list for
 * the process or process group.
 */
int
fsetown(pid_t pgid, struct sigio **sigiop)
{
        struct proc *proc = NULL;
        struct pgrp *pgrp = NULL;
        struct sigio *sigio;
        int error;

        if (pgid == 0) {
                funsetown(sigiop);
                return (0);
        }

        if (pgid > 0) {
                proc = pfind(pgid);
                if (proc == NULL) {
                        error = ESRCH;
                        goto done;
                }

                /*
                 * Policy - Don't allow a process to FSETOWN a process
                 * in another session.
                 *
                 * Remove this test to allow maximum flexibility or
                 * restrict FSETOWN to the current process or process
                 * group for maximum safety.
                 */
                if (proc->p_session != curproc->p_session) {
                        error = EPERM;
                        goto done;
                }
        } else /* if (pgid < 0) */ {
                pgrp = pgfind(-pgid);
                if (pgrp == NULL) {
                        error = ESRCH;
                        goto done;
                }

                /*
                 * Policy - Don't allow a process to FSETOWN a process
                 * in another session.
                 *
                 * Remove this test to allow maximum flexibility or
                 * restrict FSETOWN to the current process or process
                 * group for maximum safety.
                 */
                if (pgrp->pg_session != curproc->p_session) {
                        error = EPERM;
                        goto done;
                }
        }
        sigio = kmalloc(sizeof(struct sigio), M_SIGIO, M_WAITOK | M_ZERO);
        if (pgid > 0) {
                KKASSERT(pgrp == NULL);
                lwkt_gettoken(&proc->p_token);
                SLIST_INSERT_HEAD(&proc->p_sigiolst, sigio, sio_pgsigio);
                sigio->sio_proc = proc;
                lwkt_reltoken(&proc->p_token);
        } else {
                KKASSERT(proc == NULL);
                lwkt_gettoken(&pgrp->pg_token);
                SLIST_INSERT_HEAD(&pgrp->pg_sigiolst, sigio, sio_pgsigio);
                sigio->sio_pgrp = pgrp;
                lwkt_reltoken(&pgrp->pg_token);
                pgrp = NULL;
        }
        sigio->sio_pgid = pgid;
        sigio->sio_ucred = crhold(curthread->td_ucred);
        /* It would be convenient if p_ruid was in ucred. */
        sigio->sio_ruid = sigio->sio_ucred->cr_ruid;
        sigio->sio_myref = sigiop;

        lwkt_gettoken(&sigio_token);
        while (*sigiop)
                funsetown(sigiop);
        *sigiop = sigio;
        lwkt_reltoken(&sigio_token);
        error = 0;
done:
        if (pgrp)
                pgrel(pgrp);
        if (proc)
                PRELE(proc);
        return (error);
}

/*
 * This is common code for FIOGETOWN ioctl called by fcntl(fd, F_GETOWN, arg).
 */
pid_t
fgetown(struct sigio **sigiop)
{
        struct sigio *sigio;
        pid_t own;

        lwkt_gettoken_shared(&sigio_token);
        sigio = *sigiop;
        own = (sigio != NULL ? sigio->sio_pgid : 0);
        lwkt_reltoken(&sigio_token);

        return (own);
}

/*
 * Close many file descriptors.
 */
int
sys_closefrom(struct closefrom_args *uap)
{
        return(kern_closefrom(uap->fd));
}

/*
 * Close all file descriptors greater then or equal to fd
 */
int
kern_closefrom(int fd)
{
        struct thread *td = curthread;
        struct proc *p = td->td_proc;
        struct filedesc *fdp;

        KKASSERT(p);
        fdp = p->p_fd;

        if (fd < 0)
                return (EINVAL);

        /*
         * NOTE: This function will skip unassociated descriptors and
         * reserved descriptors that have not yet been assigned.  
         * fd_lastfile can change as a side effect of kern_close().
         */
        spin_lock(&fdp->fd_spin);
        while (fd <= fdp->fd_lastfile) {
                if (fdp->fd_files[fd].fp != NULL) {
                        spin_unlock(&fdp->fd_spin);
                        /* ok if this races another close */
                        if (kern_close(fd) == EINTR)
                                return (EINTR);
                        spin_lock(&fdp->fd_spin);
                }
                ++fd;
        }
        spin_unlock(&fdp->fd_spin);
        return (0);
}

/*
 * Close a file descriptor.
 */
int
sys_close(struct close_args *uap)
{
        return(kern_close(uap->fd));
}

/*
 * close() helper
 */
int
kern_close(int fd)
{
        struct thread *td = curthread;
        struct proc *p = td->td_proc;
        struct filedesc *fdp;
        struct file *fp;
        int error;
        int holdleaders;

        KKASSERT(p);
        fdp = p->p_fd;

        /*
         * funsetfd*() also clears the fd cache
         */
        spin_lock(&fdp->fd_spin);
        if ((fp = funsetfd_locked(fdp, fd)) == NULL) {
                spin_unlock(&fdp->fd_spin);
                return (EBADF);
        }
        holdleaders = 0;
        if (p->p_fdtol != NULL) {
                /*
                 * Ask fdfree() to sleep to ensure that all relevant
                 * process leaders can be traversed in closef().
                 */
                fdp->fd_holdleaderscount++;
                holdleaders = 1;
        }

        /*
         * we now hold the fp reference that used to be owned by the descriptor
         * array.
         */
        spin_unlock(&fdp->fd_spin);
        if (SLIST_FIRST(&fp->f_klist))
                knote_fdclose(fp, fdp, fd);
        error = closef(fp, p);
        if (holdleaders) {
                spin_lock(&fdp->fd_spin);
                fdp->fd_holdleaderscount--;
                if (fdp->fd_holdleaderscount == 0 &&
                    fdp->fd_holdleaderswakeup != 0) {
                        fdp->fd_holdleaderswakeup = 0;
                        spin_unlock(&fdp->fd_spin);
                        wakeup(&fdp->fd_holdleaderscount);
                } else {
                        spin_unlock(&fdp->fd_spin);
                }
        }
        return (error);
}

/*
 * shutdown_args(int fd, int how)
 */
int
kern_shutdown(int fd, int how)
{
        struct thread *td = curthread;
        struct file *fp;
        int error;

        if ((fp = holdfp(td, fd, -1)) == NULL)
                return (EBADF);
        error = fo_shutdown(fp, how);
        fdrop(fp);

        return (error);
}

/*
 * MPALMOSTSAFE
 */
int
sys_shutdown(struct shutdown_args *uap)
{
        int error;

        error = kern_shutdown(uap->s, uap->how);

        return (error);
}

/*
 * fstat() helper
 */
int
kern_fstat(int fd, struct stat *ub)
{
        struct thread *td = curthread;
        struct file *fp;
        int error;

        if ((fp = holdfp(td, fd, -1)) == NULL)
                return (EBADF);
        error = fo_stat(fp, ub, td->td_ucred);
        fdrop(fp);

        return (error);
}

/*
 * Return status information about a file descriptor.
 */
int
sys_fstat(struct fstat_args *uap)
{
        struct stat st;
        int error;

        error = kern_fstat(uap->fd, &st);

        if (error == 0)
                error = copyout(&st, uap->sb, sizeof(st));
        return (error);
}

/*
 * Return pathconf information about a file descriptor.
 *
 * MPALMOSTSAFE
 */
int
sys_fpathconf(struct fpathconf_args *uap)
{
        struct thread *td = curthread;
        struct file *fp;
        struct vnode *vp;
        int error = 0;

        if ((fp = holdfp(td, uap->fd, -1)) == NULL)
                return (EBADF);

        switch (fp->f_type) {
        case DTYPE_PIPE:
        case DTYPE_SOCKET:
                if (uap->name != _PC_PIPE_BUF) {
                        error = EINVAL;
                } else {
                        uap->sysmsg_result = PIPE_BUF;
                        error = 0;
                }
                break;
        case DTYPE_FIFO:
        case DTYPE_VNODE:
                vp = (struct vnode *)fp->f_data;
                error = VOP_PATHCONF(vp, uap->name, &uap->sysmsg_reg);
                break;
        default:
                error = EOPNOTSUPP;
                break;
        }
        fdrop(fp);
        return(error);
}

/*
 * Grow the file table so it can hold through descriptor (want).
 *
 * The fdp's spinlock must be held exclusively on entry and may be held
 * exclusively on return.  The spinlock may be cycled by the routine.
 */
static void
fdgrow_locked(struct filedesc *fdp, int want)
{
        struct fdnode *newfiles;
        struct fdnode *oldfiles;
        int nf, extra;

        nf = fdp->fd_nfiles;
        do {
                /* nf has to be of the form 2^n - 1 */
                nf = 2 * nf + 1;
        } while (nf <= want);

        spin_unlock(&fdp->fd_spin);
        newfiles = kmalloc(nf * sizeof(struct fdnode), M_FILEDESC, M_WAITOK);
        spin_lock(&fdp->fd_spin);

        /*
         * We could have raced another extend while we were not holding
         * the spinlock.
         */
        if (fdp->fd_nfiles >= nf) {
                spin_unlock(&fdp->fd_spin);
                kfree(newfiles, M_FILEDESC);
                spin_lock(&fdp->fd_spin);
                return;
        }
        /*
         * Copy the existing ofile and ofileflags arrays
         * and zero the new portion of each array.
         */
        extra = nf - fdp->fd_nfiles;
        bcopy(fdp->fd_files, newfiles, fdp->fd_nfiles * sizeof(struct fdnode));
        bzero(&newfiles[fdp->fd_nfiles], extra * sizeof(struct fdnode));

        oldfiles = fdp->fd_files;
        fdp->fd_files = newfiles;
        fdp->fd_nfiles = nf;

        if (oldfiles != fdp->fd_builtin_files) {
                spin_unlock(&fdp->fd_spin);
                kfree(oldfiles, M_FILEDESC);
                spin_lock(&fdp->fd_spin);
        }
}

/*
 * Number of nodes in right subtree, including the root.
 */
static __inline int
right_subtree_size(int n)
{
        return (n ^ (n | (n + 1)));
}

/*
 * Bigger ancestor.
 */
static __inline int
right_ancestor(int n)
{
        return (n | (n + 1));
}

/*
 * Smaller ancestor.
 */
static __inline int
left_ancestor(int n)
{
        return ((n & (n + 1)) - 1);
}

/*
 * Traverse the in-place binary tree buttom-up adjusting the allocation
 * count so scans can determine where free descriptors are located.
 *
 * caller must be holding an exclusive spinlock on fdp
 */
static
void
fdreserve_locked(struct filedesc *fdp, int fd, int incr)
{
        while (fd >= 0) {
                fdp->fd_files[fd].allocated += incr;
                KKASSERT(fdp->fd_files[fd].allocated >= 0);
                fd = left_ancestor(fd);
        }
}

/*
 * Reserve a file descriptor for the process.  If no error occurs, the
 * caller MUST at some point call fsetfd() or assign a file pointer
 * or dispose of the reservation.
 */
int
fdalloc(struct proc *p, int want, int *result)
{
        struct plimit *limit = readplimits(p);
        struct filedesc *fdp = p->p_fd;
        struct uidinfo *uip;
        int fd, rsize, rsum, node, lim;

        /*
         * Check dtable size limit
         */
        *result = -1;   /* avoid gcc warnings */
        if (limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur > INT_MAX)
                lim = INT_MAX;
        else
                lim = (int)limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur;

        if (lim > maxfilesperproc)
                lim = maxfilesperproc;
        if (lim < minfilesperproc)
                lim = minfilesperproc;
        if (want >= lim)
                return (EMFILE);

        /*
         * Check that the user has not run out of descriptors (non-root only).
         * As a safety measure the dtable is allowed to have at least
         * minfilesperproc open fds regardless of the maxfilesperuser limit.
         *
         * This isn't as loose a spec as ui_posixlocks, so we use atomic
         * ops to force synchronize and recheck if we would otherwise
         * error.
         */
        if (p->p_ucred->cr_uid && fdp->fd_nfiles >= minfilesperproc) {
                uip = p->p_ucred->cr_uidinfo;
                if (uip->ui_openfiles > maxfilesperuser) {
                        int n;
                        int count;

                        for (n = 0; n < ncpus; ++n) {
                                count = atomic_swap_int(
                                            &uip->ui_pcpu[n].pu_openfiles, 0);
                                atomic_add_int(&uip->ui_openfiles, count);
                        }
                        if (uip->ui_openfiles > maxfilesperuser) {
                                krateprintf(&krate_uidinfo,
                                            "Warning: user %d pid %d (%s) "
                                            "ran out of file descriptors "
                                            "(%d/%d)\n",
                                            p->p_ucred->cr_uid, (int)p->p_pid,
                                            p->p_comm,
                                            uip->ui_openfiles, maxfilesperuser);
                                return(ENFILE);
                        }
                }
        }

        /*
         * Grow the dtable if necessary
         */
        spin_lock(&fdp->fd_spin);
        if (want >= fdp->fd_nfiles)
                fdgrow_locked(fdp, want);

        /*
         * Search for a free descriptor starting at the higher
         * of want or fd_freefile.  If that fails, consider
         * expanding the ofile array.
         *
         * NOTE! the 'allocated' field is a cumulative recursive allocation
         * count.  If we happen to see a value of 0 then we can shortcut
         * our search.  Otherwise we run through through the tree going
         * down branches we know have free descriptor(s) until we hit a
         * leaf node.  The leaf node will be free but will not necessarily
         * have an allocated field of 0.
         */
retry:
        /* move up the tree looking for a subtree with a free node */
        for (fd = max(want, fdp->fd_freefile); fd < min(fdp->fd_nfiles, lim);
             fd = right_ancestor(fd)) {
                if (fdp->fd_files[fd].allocated == 0)
                        goto found;

                rsize = right_subtree_size(fd);
                if (fdp->fd_files[fd].allocated == rsize)
                        continue;       /* right subtree full */

                /*
                 * Free fd is in the right subtree of the tree rooted at fd.
                 * Call that subtree R.  Look for the smallest (leftmost)
                 * subtree of R with an unallocated fd: continue moving
                 * down the left branch until encountering a full left
                 * subtree, then move to the right.
                 */
                for (rsum = 0, rsize /= 2; rsize > 0; rsize /= 2) {
                        node = fd + rsize;
                        rsum += fdp->fd_files[node].allocated;
                        if (fdp->fd_files[fd].allocated == rsum + rsize) {
                                fd = node;      /* move to the right */
                                if (fdp->fd_files[node].allocated == 0)
                                        goto found;
                                rsum = 0;
                        }
                }
                goto found;
        }

        /*
         * No space in current array.  Expand?
         */
        if (fdp->fd_nfiles >= lim) {
                spin_unlock(&fdp->fd_spin);
                return (EMFILE);
        }
        fdgrow_locked(fdp, want);
        goto retry;

found:
        KKASSERT(fd < fdp->fd_nfiles);
        if (fd > fdp->fd_lastfile)
                fdp->fd_lastfile = fd;
        if (want <= fdp->fd_freefile)
                fdp->fd_freefile = fd;
        *result = fd;
        KKASSERT(fdp->fd_files[fd].fp == NULL);
        KKASSERT(fdp->fd_files[fd].reserved == 0);
        fdp->fd_files[fd].fileflags = 0;
        fdp->fd_files[fd].reserved = 1;
        fdreserve_locked(fdp, fd, 1);
        spin_unlock(&fdp->fd_spin);
        return (0);
}

/*
 * Check to see whether n user file descriptors
 * are available to the process p.
 */
int
fdavail(struct proc *p, int n)
{
        struct plimit *limit = readplimits(p);
        struct filedesc *fdp = p->p_fd;
        struct fdnode *fdnode;
        int i, lim, last;

        if (limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur > INT_MAX)
                lim = INT_MAX;
        else
                lim = (int)limit->pl_rlimit[RLIMIT_NOFILE].rlim_cur;

        if (lim > maxfilesperproc)
                lim = maxfilesperproc;
        if (lim < minfilesperproc)
                lim = minfilesperproc;

        spin_lock(&fdp->fd_spin);
        if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0) {
                spin_unlock(&fdp->fd_spin);
                return (1);
        }
        last = min(fdp->fd_nfiles, lim);
        fdnode = &fdp->fd_files[fdp->fd_freefile];
        for (i = last - fdp->fd_freefile; --i >= 0; ++fdnode) {
                if (fdnode->fp == NULL && --n <= 0) {
                        spin_unlock(&fdp->fd_spin);
                        return (1);
                }
        }
        spin_unlock(&fdp->fd_spin);
        return (0);
}

/*
 * Revoke open descriptors referencing (f_data, f_type)
 *
 * Any revoke executed within a prison is only able to
 * revoke descriptors for processes within that prison.
 *
 * Returns 0 on success or an error code.
 */
struct fdrevoke_info {
        void *data;
        short type;
        short unused;
        int found;
        struct ucred *cred;
        struct file *nfp;
};

static int fdrevoke_check_callback(struct file *fp, void *vinfo);
static int fdrevoke_proc_callback(struct proc *p, void *vinfo);

int
fdrevoke(void *f_data, short f_type, struct ucred *cred)
{
        struct fdrevoke_info info;
        int error;

        bzero(&info, sizeof(info));
        info.data = f_data;
        info.type = f_type;
        info.cred = cred;
        error = falloc(NULL, &info.nfp, NULL);
        if (error)
                return (error);

        /*
         * Scan the file pointer table once.  dups do not dup file pointers,
         * only descriptors, so there is no leak.  Set FREVOKED on the fps
         * being revoked.
         *
         * Any fps sent over unix-domain sockets will be revoked by the
         * socket code checking for FREVOKED when the fps are externialized.
         * revoke_token is used to make sure that fps marked FREVOKED and
         * externalized will be picked up by the following allproc_scan().
         */
        lwkt_gettoken(&revoke_token);
        allfiles_scan_exclusive(fdrevoke_check_callback, &info);
        lwkt_reltoken(&revoke_token);

        /*
         * If any fps were marked track down the related descriptors
         * and close them.  Any dup()s at this point will notice
         * the FREVOKED already set in the fp and do the right thing.
         */
        if (info.found)
                allproc_scan(fdrevoke_proc_callback, &info, 0);
        fdrop(info.nfp);
        return(0);
}

/*
 * Locate matching file pointers directly.
 *
 * WARNING: allfiles_scan_exclusive() holds a spinlock through these calls!
 */
static int
fdrevoke_check_callback(struct file *fp, void *vinfo)
{
        struct fdrevoke_info *info = vinfo;

        /*
         * File pointers already flagged for revokation are skipped.
         */
        if (fp->f_flag & FREVOKED)
                return(0);

        /*
         * If revoking from a prison file pointers created outside of
         * that prison, or file pointers without creds, cannot be revoked.
         */
        if (info->cred->cr_prison &&
            (fp->f_cred == NULL ||
             info->cred->cr_prison != fp->f_cred->cr_prison)) {
                return(0);
        }

        /*
         * If the file pointer matches then mark it for revocation.  The
         * flag is currently only used by unp_revoke_gc().
         *
         * info->found is a heuristic and can race in a SMP environment.
         */
        if (info->data == fp->f_data && info->type == fp->f_type) {
                atomic_set_int(&fp->f_flag, FREVOKED);
                info->found = 1;
        }
        return(0);
}

/*
 * Locate matching file pointers via process descriptor tables.
 */
static int
fdrevoke_proc_callback(struct proc *p, void *vinfo)
{
        struct fdrevoke_info *info = vinfo;
        struct filedesc *fdp;
        struct file *fp;
        int n;

        if (p->p_stat == SIDL || p->p_stat == SZOMB)
                return(0);
        if (info->cred->cr_prison &&
            info->cred->cr_prison != p->p_ucred->cr_prison) {
                return(0);
        }

        /*
         * If the controlling terminal of the process matches the
         * vnode being revoked we clear the controlling terminal.
         *
         * The normal spec_close() may not catch this because it
         * uses curproc instead of p.
         */
        if (p->p_session && info->type == DTYPE_VNODE &&
            info->data == p->p_session->s_ttyvp) {
                p->p_session->s_ttyvp = NULL;
                vrele(info->data);
        }

        /*
         * Softref the fdp to prevent it from being destroyed
         */
        spin_lock(&p->p_spin);
        if ((fdp = p->p_fd) == NULL) {
                spin_unlock(&p->p_spin);
                return(0);
        }
        atomic_add_int(&fdp->fd_softrefs, 1);
        spin_unlock(&p->p_spin);

        /*
         * Locate and close any matching file descriptors, replacing
         * them with info->nfp.
         */
        spin_lock(&fdp->fd_spin);
        for (n = 0; n < fdp->fd_nfiles; ++n) {
                if ((fp = fdp->fd_files[n].fp) == NULL)
                        continue;
                if (fp->f_flag & FREVOKED) {
                        ++fdp->fd_closedcounter;
                        fclearcache(&fdp->fd_files[n], NULL, 0);
                        ++fdp->fd_closedcounter;
                        fhold(info->nfp);
                        fdp->fd_files[n].fp = info->nfp;
                        spin_unlock(&fdp->fd_spin);
                        knote_fdclose(fp, fdp, n);      /* XXX */
                        closef(fp, p);
                        spin_lock(&fdp->fd_spin);
                }
        }
        spin_unlock(&fdp->fd_spin);
        atomic_subtract_int(&fdp->fd_softrefs, 1);
        return(0);
}

/*
 * falloc:
 *      Create a new open file structure and reserve a file decriptor
 *      for the process that refers to it.
 *
 *      Root creds are checked using lp, or assumed if lp is NULL.  If
 *      resultfd is non-NULL then lp must also be non-NULL.  No file
 *      descriptor is reserved (and no process context is needed) if
 *      resultfd is NULL.
 *
 *      A file pointer with a refcount of 1 is returned.  Note that the
 *      file pointer is NOT associated with the descriptor.  If falloc
 *      returns success, fsetfd() MUST be called to either associate the
 *      file pointer or clear the reservation.
 */
int
falloc(struct lwp *lp, struct file **resultfp, int *resultfd)
{
        static struct timeval lastfail;
        static int curfail;
        struct filelist_head *head;
        struct file *fp;
        struct ucred *cred = lp ? lp->lwp_thread->td_ucred : proc0.p_ucred;
        int error;

        fp = NULL;

        /*
         * Handle filetable full issues and root overfill.
         */
        if (nfiles >= maxfiles - maxfilesrootres &&
            (cred->cr_ruid != 0 || nfiles >= maxfiles)) {
                if (ppsratecheck(&lastfail, &curfail, 1)) {
                        kprintf("kern.maxfiles limit exceeded by uid %d, "
                                "please see tuning(7).\n",
                                cred->cr_ruid);
                }
                error = ENFILE;
                goto done;
        }

        /*
         * Allocate a new file descriptor.
         */
        fp = objcache_get(file_objcache, M_WAITOK);
        bzero(fp, sizeof(*fp));
        spin_init(&fp->f_spin, "falloc");
        SLIST_INIT(&fp->f_klist);
        fp->f_count = 1;
        fp->f_ops = &badfileops;
        fp->f_seqcount = 1;
        fsetcred(fp, cred);
        atomic_add_int(&nfiles, 1);

        head = fp2filelist(fp);
        spin_lock(&head->spin);
        LIST_INSERT_HEAD(&head->list, fp, f_list);
        spin_unlock(&head->spin);

        if (resultfd) {
                if ((error = fdalloc(lp->lwp_proc, 0, resultfd)) != 0) {
                        fdrop(fp);
                        fp = NULL;
                }
        } else {
                error = 0;
        }
done:
        *resultfp = fp;
        return (error);
}

/*
 * Check for races against a file descriptor by determining that the
 * file pointer is still associated with the specified file descriptor,
 * and a close is not currently in progress.
 */
int
checkfdclosed(thread_t td, struct filedesc *fdp, int fd, struct file *fp,
              int closedcounter)
{
        struct fdcache *fdc;
        int error;

        cpu_lfence();
        if (fdp->fd_closedcounter == closedcounter)
                return 0;

        if (td->td_proc && td->td_proc->p_fd == fdp) {
                for (fdc = &td->td_fdcache[0];
                     fdc < &td->td_fdcache[NFDCACHE]; ++fdc) {
                        if (fdc->fd == fd && fdc->fp == fp)
                                return 0;
                }
        }

        spin_lock_shared(&fdp->fd_spin);
        if ((unsigned)fd >= fdp->fd_nfiles || fp != fdp->fd_files[fd].fp)
                error = EBADF;
        else
                error = 0;
        spin_unlock_shared(&fdp->fd_spin);
        return (error);
}

/*
 * Associate a file pointer with a previously reserved file descriptor.
 * This function always succeeds.
 *
 * If fp is NULL, the file descriptor is returned to the pool.
 *
 * Caller must hold an exclusive spinlock on fdp->fd_spin.
 */
static void
fsetfd_locked(struct filedesc *fdp, struct file *fp, int fd)
{
        KKASSERT((unsigned)fd < fdp->fd_nfiles);
        KKASSERT(fdp->fd_files[fd].reserved != 0);
        if (fp) {
                fhold(fp);
                fclearcache(&fdp->fd_files[fd], NULL, 0);
                fdp->fd_files[fd].fp = fp;
                fdp->fd_files[fd].reserved = 0;
        } else {
                fdp->fd_files[fd].reserved = 0;
                fdreserve_locked(fdp, fd, -1);
                fdfixup_locked(fdp, fd);
        }
}

/*
 * Caller must hold an exclusive spinlock on fdp->fd_spin.
 */
void
fsetfd(struct filedesc *fdp, struct file *fp, int fd)
{
        spin_lock(&fdp->fd_spin);
        fsetfd_locked(fdp, fp, fd);
        spin_unlock(&fdp->fd_spin);
}

/*
 * Caller must hold an exclusive spinlock on fdp->fd_spin.
 */
static 
struct file *
funsetfd_locked(struct filedesc *fdp, int fd)
{
        struct file *fp;

        if ((unsigned)fd >= fdp->fd_nfiles)
                return (NULL);
        if ((fp = fdp->fd_files[fd].fp) == NULL)
                return (NULL);
        ++fdp->fd_closedcounter;
        fclearcache(&fdp->fd_files[fd], NULL, 0);
        fdp->fd_files[fd].fp = NULL;
        fdp->fd_files[fd].fileflags = 0;
        ++fdp->fd_closedcounter;

        fdreserve_locked(fdp, fd, -1);
        fdfixup_locked(fdp, fd);

        return(fp);
}

/*
 * WARNING: May not be called before initial fsetfd().
 */
int
fgetfdflags(struct filedesc *fdp, int fd, int *flagsp)
{
        int error;

        spin_lock(&fdp->fd_spin);
        if (((u_int)fd) >= fdp->fd_nfiles) {
                error = EBADF;
        } else if (fdp->fd_files[fd].fp == NULL) {
                error = EBADF;
        } else {
                *flagsp = fdp->fd_files[fd].fileflags;
                error = 0;
        }
        spin_unlock(&fdp->fd_spin);
        return (error);
}

/*
 * WARNING: May not be called before initial fsetfd().
 */
int
fsetfdflags(struct filedesc *fdp, int fd, int add_flags)
{
        int error;

        spin_lock(&fdp->fd_spin);
        if (((u_int)fd) >= fdp->fd_nfiles) {
                error = EBADF;
        } else if (fdp->fd_files[fd].fp == NULL) {
                error = EBADF;
        } else {
                fdp->fd_files[fd].fileflags |= add_flags;
                error = 0;
        }
        spin_unlock(&fdp->fd_spin);
        return (error);
}

/*
 * WARNING: May not be called before initial fsetfd().
 */
int
fclrfdflags(struct filedesc *fdp, int fd, int rem_flags)
{
        int error;

        spin_lock(&fdp->fd_spin);
        if (((u_int)fd) >= fdp->fd_nfiles) {
                error = EBADF;
        } else if (fdp->fd_files[fd].fp == NULL) {
                error = EBADF;
        } else {
                fdp->fd_files[fd].fileflags &= ~rem_flags;
                error = 0;
        }
        spin_unlock(&fdp->fd_spin);
        return (error);
}

/*
 * Set/Change/Clear the creds for a fp and synchronize the uidinfo.
 */
void
fsetcred(struct file *fp, struct ucred *ncr)
{
        struct ucred *ocr;
        struct uidinfo *uip;
        struct uidcount *pup;
        int cpu = mycpuid;
        int count;

        ocr = fp->f_cred;
        if (ocr == NULL || ncr == NULL || ocr->cr_uidinfo != ncr->cr_uidinfo) {
                if (ocr) {
                        uip = ocr->cr_uidinfo;
                        pup = &uip->ui_pcpu[cpu];
                        atomic_add_int(&pup->pu_openfiles, -1);
                        if (pup->pu_openfiles < -PUP_LIMIT ||
                            pup->pu_openfiles > PUP_LIMIT) {
                                count = atomic_swap_int(&pup->pu_openfiles, 0);
                                atomic_add_int(&uip->ui_openfiles, count);
                        }
                }
                if (ncr) {
                        uip = ncr->cr_uidinfo;
                        pup = &uip->ui_pcpu[cpu];
                        atomic_add_int(&pup->pu_openfiles, 1);
                        if (pup->pu_openfiles < -PUP_LIMIT ||
                            pup->pu_openfiles > PUP_LIMIT) {
                                count = atomic_swap_int(&pup->pu_openfiles, 0);
                                atomic_add_int(&uip->ui_openfiles, count);
                        }
                }
        }
        if (ncr)
                crhold(ncr);
        fp->f_cred = ncr;
        if (ocr)
                crfree(ocr);
}

/*
 * Free a file descriptor.
 */
static
void
ffree(struct file *fp)
{
        KASSERT((fp->f_count == 0), ("ffree: fp_fcount not 0!"));
        fsetcred(fp, NULL);
        if (fp->f_nchandle.ncp)
            cache_drop(&fp->f_nchandle);
        objcache_put(file_objcache, fp);
}

/*
 * called from init_main, initialize filedesc0 for proc0.
 */
void
fdinit_bootstrap(struct proc *p0, struct filedesc *fdp0, int cmask)
{
        p0->p_fd = fdp0;
        p0->p_fdtol = NULL;
        fdp0->fd_refcnt = 1;
        fdp0->fd_cmask = cmask;
        fdp0->fd_files = fdp0->fd_builtin_files;
        fdp0->fd_nfiles = NDFILE;
        fdp0->fd_lastfile = -1;
        spin_init(&fdp0->fd_spin, "fdinitbootstrap");
}

/*
 * Build a new filedesc structure.
 */
struct filedesc *
fdinit(struct proc *p)
{
        struct filedesc *newfdp;
        struct filedesc *fdp = p->p_fd;

        newfdp = kmalloc(sizeof(struct filedesc), M_FILEDESC, M_WAITOK|M_ZERO);
        spin_lock(&fdp->fd_spin);
        if (fdp->fd_cdir) {
                newfdp->fd_cdir = fdp->fd_cdir;
                vref(newfdp->fd_cdir);
                cache_copy(&fdp->fd_ncdir, &newfdp->fd_ncdir);
        }

        /*
         * rdir may not be set in e.g. proc0 or anything vm_fork'd off of
         * proc0, but should unconditionally exist in other processes.
         */
        if (fdp->fd_rdir) {
                newfdp->fd_rdir = fdp->fd_rdir;
                vref(newfdp->fd_rdir);
                cache_copy(&fdp->fd_nrdir, &newfdp->fd_nrdir);
        }
        if (fdp->fd_jdir) {
                newfdp->fd_jdir = fdp->fd_jdir;
                vref(newfdp->fd_jdir);
                cache_copy(&fdp->fd_njdir, &newfdp->fd_njdir);
        }
        spin_unlock(&fdp->fd_spin);

        /* Create the file descriptor table. */
        newfdp->fd_refcnt = 1;
        newfdp->fd_cmask = cmask;
        newfdp->fd_files = newfdp->fd_builtin_files;
        newfdp->fd_nfiles = NDFILE;
        newfdp->fd_lastfile = -1;
        spin_init(&newfdp->fd_spin, "fdinit");

        return (newfdp);
}

/*
 * Share a filedesc structure.
 */
struct filedesc *
fdshare(struct proc *p)
{
        struct filedesc *fdp;

        fdp = p->p_fd;
        spin_lock(&fdp->fd_spin);
        fdp->fd_refcnt++;
        spin_unlock(&fdp->fd_spin);
        return (fdp);
}

/*
 * Copy a filedesc structure.
 */
int
fdcopy(struct proc *p, struct filedesc **fpp)
{
        struct filedesc *fdp = p->p_fd;
        struct filedesc *newfdp;
        struct fdnode *fdnode;
        int i;
        int ni;

        /*
         * Certain daemons might not have file descriptors. 
         */
        if (fdp == NULL)
                return (0);

        /*
         * Allocate the new filedesc and fd_files[] array.  This can race
         * with operations by other threads on the fdp so we have to be
         * careful.
         */
        newfdp = kmalloc(sizeof(struct filedesc), 
                         M_FILEDESC, M_WAITOK | M_ZERO | M_NULLOK);
        if (newfdp == NULL) {
                *fpp = NULL;
                return (-1);
        }
again:
        spin_lock(&fdp->fd_spin);
        if (fdp->fd_lastfile < NDFILE) {
                newfdp->fd_files = newfdp->fd_builtin_files;
                i = NDFILE;
        } else {
                /*
                 * We have to allocate (N^2-1) entries for our in-place
                 * binary tree.  Allow the table to shrink.
                 */
                i = fdp->fd_nfiles;
                ni = (i - 1) / 2;
                while (ni > fdp->fd_lastfile && ni > NDFILE) {
                        i = ni;
                        ni = (i - 1) / 2;
                }
                spin_unlock(&fdp->fd_spin);
                newfdp->fd_files = kmalloc(i * sizeof(struct fdnode),
                                          M_FILEDESC, M_WAITOK | M_ZERO);

                /*
                 * Check for race, retry
                 */
                spin_lock(&fdp->fd_spin);
                if (i <= fdp->fd_lastfile) {
                        spin_unlock(&fdp->fd_spin);
                        kfree(newfdp->fd_files, M_FILEDESC);
                        goto again;
                }
        }

        /*
         * Dup the remaining fields. vref() and cache_hold() can be
         * safely called while holding the read spinlock on fdp.
         *
         * The read spinlock on fdp is still being held.
         *
         * NOTE: vref and cache_hold calls for the case where the vnode
         * or cache entry already has at least one ref may be called
         * while holding spin locks.
         */
        if ((newfdp->fd_cdir = fdp->fd_cdir) != NULL) {
                vref(newfdp->fd_cdir);
                cache_copy(&fdp->fd_ncdir, &newfdp->fd_ncdir);
        }
        /*
         * We must check for fd_rdir here, at least for now because
         * the init process is created before we have access to the
         * rootvode to take a reference to it.
         */
        if ((newfdp->fd_rdir = fdp->fd_rdir) != NULL) {
                vref(newfdp->fd_rdir);
                cache_copy(&fdp->fd_nrdir, &newfdp->fd_nrdir);
        }
        if ((newfdp->fd_jdir = fdp->fd_jdir) != NULL) {
                vref(newfdp->fd_jdir);
                cache_copy(&fdp->fd_njdir, &newfdp->fd_njdir);
        }
        newfdp->fd_refcnt = 1;
        newfdp->fd_nfiles = i;
        newfdp->fd_lastfile = fdp->fd_lastfile;
        newfdp->fd_freefile = fdp->fd_freefile;
        newfdp->fd_cmask = fdp->fd_cmask;
        spin_init(&newfdp->fd_spin, "fdcopy");

        /*
         * Copy the descriptor table through (i).  This also copies the
         * allocation state.   Then go through and ref the file pointers
         * and clean up any KQ descriptors.
         *
         * kq descriptors cannot be copied.  Since we haven't ref'd the
         * copied files yet we can ignore the return value from funsetfd().
         *
         * The read spinlock on fdp is still being held.
         *
         * Be sure to clean out fdnode->tdcache, otherwise bad things will
         * happen.
         */
        bcopy(fdp->fd_files, newfdp->fd_files, i * sizeof(struct fdnode));
        for (i = 0 ; i < newfdp->fd_nfiles; ++i) {
                fdnode = &newfdp->fd_files[i];
                if (fdnode->reserved) {
                        fdreserve_locked(newfdp, i, -1);
                        fdnode->reserved = 0;
                        fdfixup_locked(newfdp, i);
                } else if (fdnode->fp) {
                        bzero(&fdnode->tdcache, sizeof(fdnode->tdcache));
                        if (fdnode->fp->f_type == DTYPE_KQUEUE) {
                                (void)funsetfd_locked(newfdp, i);
                        } else {
                                fhold(fdnode->fp);
                        }
                }
        }
        spin_unlock(&fdp->fd_spin);
        *fpp = newfdp;
        return (0);
}

/*
 * Release a filedesc structure.
 *
 * NOT MPSAFE (MPSAFE for refs > 1, but the final cleanup code is not MPSAFE)
 */
void
fdfree(struct proc *p, struct filedesc *repl)
{
        struct filedesc *fdp;
        struct fdnode *fdnode;
        int i;
        struct filedesc_to_leader *fdtol;
        struct file *fp;
        struct vnode *vp;
        struct flock lf;

        /*
         * Before destroying or replacing p->p_fd we must be sure to
         * clean out the cache of the last thread, which should be
         * curthread.
         */
        fexitcache(curthread);

        /*
         * Certain daemons might not have file descriptors.
         */
        fdp = p->p_fd;
        if (fdp == NULL) {
                p->p_fd = repl;
                return;
        }

        /*
         * Severe messing around to follow.
         */
        spin_lock(&fdp->fd_spin);

        /* Check for special need to clear POSIX style locks */
        fdtol = p->p_fdtol;
        if (fdtol != NULL) {
                KASSERT(fdtol->fdl_refcount > 0,
                        ("filedesc_to_refcount botch: fdl_refcount=%d",
                         fdtol->fdl_refcount));
                if (fdtol->fdl_refcount == 1 && p->p_leader->p_advlock_flag) {
                        for (i = 0; i <= fdp->fd_lastfile; ++i) {
                                fdnode = &fdp->fd_files[i];
                                if (fdnode->fp == NULL ||
                                    fdnode->fp->f_type != DTYPE_VNODE) {
                                        continue;
                                }
                                fp = fdnode->fp;
                                fhold(fp);
                                spin_unlock(&fdp->fd_spin);

                                lf.l_whence = SEEK_SET;
                                lf.l_start = 0;
                                lf.l_len = 0;
                                lf.l_type = F_UNLCK;
                                vp = (struct vnode *)fp->f_data;
                                VOP_ADVLOCK(vp, (caddr_t)p->p_leader,
                                            F_UNLCK, &lf, F_POSIX);
                                fdrop(fp);
                                spin_lock(&fdp->fd_spin);
                        }
                }
        retry:
                if (fdtol->fdl_refcount == 1) {
                        if (fdp->fd_holdleaderscount > 0 &&
                            p->p_leader->p_advlock_flag) {
                                /*
                                 * close() or do_dup() has cleared a reference
                                 * in a shared file descriptor table.
                                 */
                                fdp->fd_holdleaderswakeup = 1;
                                ssleep(&fdp->fd_holdleaderscount,
                                       &fdp->fd_spin, 0, "fdlhold", 0);
                                goto retry;
                        }
                        if (fdtol->fdl_holdcount > 0) {
                                /* 
                                 * Ensure that fdtol->fdl_leader
                                 * remains valid in closef().
                                 */
                                fdtol->fdl_wakeup = 1;
                                ssleep(fdtol, &fdp->fd_spin, 0, "fdlhold", 0);
                                goto retry;
                        }
                }
                fdtol->fdl_refcount--;
                if (fdtol->fdl_refcount == 0 &&
                    fdtol->fdl_holdcount == 0) {
                        fdtol->fdl_next->fdl_prev = fdtol->fdl_prev;
                        fdtol->fdl_prev->fdl_next = fdtol->fdl_next;
                } else {
                        fdtol = NULL;
                }
                p->p_fdtol = NULL;
                if (fdtol != NULL) {
                        spin_unlock(&fdp->fd_spin);
                        kfree(fdtol, M_FILEDESC_TO_LEADER);
                        spin_lock(&fdp->fd_spin);
                }
        }
        if (--fdp->fd_refcnt > 0) {
                spin_unlock(&fdp->fd_spin);
                spin_lock(&p->p_spin);
                p->p_fd = repl;
                spin_unlock(&p->p_spin);
                return;
        }

        /*
         * Even though we are the last reference to the structure allproc
         * scans may still reference the structure.  Maintain proper
         * locks until we can replace p->p_fd.
         *
         * Also note that kqueue's closef still needs to reference the
         * fdp via p->p_fd, so we have to close the descriptors before
         * we replace p->p_fd.
         */
        for (i = 0; i <= fdp->fd_lastfile; ++i) {
                if (fdp->fd_files[i].fp) {
                        fp = funsetfd_locked(fdp, i);
                        if (fp) {
                                spin_unlock(&fdp->fd_spin);
                                if (SLIST_FIRST(&fp->f_klist))
                                        knote_fdclose(fp, fdp, i);
                                closef(fp, p);
                                spin_lock(&fdp->fd_spin);
                        }
                }
        }
        spin_unlock(&fdp->fd_spin);

        /*
         * Interlock against an allproc scan operations (typically frevoke).
         */
        spin_lock(&p->p_spin);
        p->p_fd = repl;
        spin_unlock(&p->p_spin);

        /*
         * Wait for any softrefs to go away.  This race rarely occurs so
         * we can use a non-critical-path style poll/sleep loop.  The
         * race only occurs against allproc scans.
         *
         * No new softrefs can occur with the fdp disconnected from the
         * process.
         */
        if (fdp->fd_softrefs) {
                kprintf("pid %d: Warning, fdp race avoided\n", p->p_pid);
                while (fdp->fd_softrefs)
                        tsleep(&fdp->fd_softrefs, 0, "fdsoft", 1);
        }

        if (fdp->fd_files != fdp->fd_builtin_files)
                kfree(fdp->fd_files, M_FILEDESC);
        if (fdp->fd_cdir) {
                cache_drop(&fdp->fd_ncdir);
                vrele(fdp->fd_cdir);
        }
        if (fdp->fd_rdir) {
                cache_drop(&fdp->fd_nrdir);
                vrele(fdp->fd_rdir);
        }
        if (fdp->fd_jdir) {
                cache_drop(&fdp->fd_njdir);
                vrele(fdp->fd_jdir);
        }
        kfree(fdp, M_FILEDESC);
}

/*
 * Retrieve and reference the file pointer associated with a descriptor.
 *
 * td must be the current thread.
 */
struct file *
holdfp(thread_t td, int fd, int flag)
{
        struct file *fp;

        fp = _holdfp_cache(td, fd);
        if (fp) {
                if ((fp->f_flag & flag) == 0 && flag != -1) {
                        fdrop(fp);
                        fp = NULL;
                }
        }
        return fp;
}

/*
 * holdsock() - load the struct file pointer associated
 * with a socket into *fpp.  If an error occurs, non-zero
 * will be returned and *fpp will be set to NULL.
 *
 * td must be the current thread.
 */
int
holdsock(thread_t td, int fd, struct file **fpp)
{
        struct file *fp;
        int error;

        /*
         * Lockless shortcut
         */
        fp = _holdfp_cache(td, fd);
        if (fp) {
                if (fp->f_type != DTYPE_SOCKET) {
                        fdrop(fp);
                        fp = NULL;
                        error = ENOTSOCK;
                } else {
                        error = 0;
                }
        } else {
                error = EBADF;
        }
        *fpp = fp;

        return (error);
}

/*
 * Convert a user file descriptor to a held file pointer.
 *
 * td must be the current thread.
 */
int
holdvnode(thread_t td, int fd, struct file **fpp)
{
        struct file *fp;
        int error;

        fp = _holdfp_cache(td, fd);
        if (fp) {
                if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO) {
                        fdrop(fp);
                        fp = NULL;
                        error = EINVAL;
                } else {
                        error = 0;
                }
        } else {
                error = EBADF;
        }
        *fpp = fp;

        return (error);
}

/*
 * For setugid programs, we don't want to people to use that setugidness
 * to generate error messages which write to a file which otherwise would
 * otherwise be off-limits to the process.
 *
 * This is a gross hack to plug the hole.  A better solution would involve
 * a special vop or other form of generalized access control mechanism.  We
 * go ahead and just reject all procfs file systems accesses as dangerous.
 *
 * Since setugidsafety calls this only for fd 0, 1 and 2, this check is
 * sufficient.  We also don't for check setugidness since we know we are.
 */
static int
is_unsafe(struct file *fp)
{
        if (fp->f_type == DTYPE_VNODE && 
            ((struct vnode *)(fp->f_data))->v_tag == VT_PROCFS)
                return (1);
        return (0);
}

/*
 * Make this setguid thing safe, if at all possible.
 *
 * NOT MPSAFE - scans fdp without spinlocks, calls knote_fdclose()
 */
void
setugidsafety(struct proc *p)
{
        struct filedesc *fdp = p->p_fd;
        int i;

        /* Certain daemons might not have file descriptors. */
        if (fdp == NULL)
                return;

        /*
         * note: fdp->fd_files may be reallocated out from under us while
         * we are blocked in a close.  Be careful!
         */
        for (i = 0; i <= fdp->fd_lastfile; i++) {
                if (i > 2)
                        break;
                if (fdp->fd_files[i].fp && is_unsafe(fdp->fd_files[i].fp)) {
                        struct file *fp;

                        /*
                         * NULL-out descriptor prior to close to avoid
                         * a race while close blocks.
                         */
                        if ((fp = funsetfd_locked(fdp, i)) != NULL) {
                                knote_fdclose(fp, fdp, i);
                                closef(fp, p);
                        }
                }
        }
}

/*
 * Close all CLOEXEC files on exec.
 *
 * Only a single thread remains for the current process.
 *
 * NOT MPSAFE - scans fdp without spinlocks, calls knote_fdclose()
 */
void
fdcloseexec(struct proc *p)
{
        struct filedesc *fdp = p->p_fd;
        int i;

        /* Certain daemons might not have file descriptors. */
        if (fdp == NULL)
                return;

        /*
         * We cannot cache fd_files since operations may block and rip
         * them out from under us.
         */
        for (i = 0; i <= fdp->fd_lastfile; i++) {
                if (fdp->fd_files[i].fp != NULL &&
                    (fdp->fd_files[i].fileflags & UF_EXCLOSE)) {
                        struct file *fp;

                        /*
                         * NULL-out descriptor prior to close to avoid
                         * a race while close blocks.
                         *
                         * (funsetfd*() also clears the fd cache)
                         */
                        if ((fp = funsetfd_locked(fdp, i)) != NULL) {
                                knote_fdclose(fp, fdp, i);
                                closef(fp, p);
                        }
                }
        }
}

/*
 * It is unsafe for set[ug]id processes to be started with file
 * descriptors 0..2 closed, as these descriptors are given implicit
 * significance in the Standard C library.  fdcheckstd() will create a
 * descriptor referencing /dev/null for each of stdin, stdout, and
 * stderr that is not already open.
 *
 * NOT MPSAFE - calls falloc, vn_open, etc
 */
int
fdcheckstd(struct lwp *lp)
{
        struct nlookupdata nd;
        struct filedesc *fdp;
        struct file *fp;
        int retval;
        int i, error, flags, devnull;

        fdp = lp->lwp_proc->p_fd;
        if (fdp == NULL)
                return (0);
        devnull = -1;
        error = 0;
        for (i = 0; i < 3; i++) {
                if (fdp->fd_files[i].fp != NULL)
                        continue;
                if (devnull < 0) {
                        if ((error = falloc(lp, &fp, &devnull)) != 0)
                                break;

                        error = nlookup_init(&nd, "/dev/null", UIO_SYSSPACE,
                                                NLC_FOLLOW|NLC_LOCKVP);
                        flags = FREAD | FWRITE;
                        if (error == 0)
                                error = vn_open(&nd, fp, flags, 0);
                        if (error == 0)
                                fsetfd(fdp, fp, devnull);
                        else
                                fsetfd(fdp, NULL, devnull);
                        fdrop(fp);
                        nlookup_done(&nd);
                        if (error)
                                break;
                        KKASSERT(i == devnull);
                } else {
                        error = kern_dup(DUP_FIXED, devnull, i, &retval);
                        if (error != 0)
                                break;
                }
        }
        return (error);
}

/*
 * Internal form of close.
 * Decrement reference count on file structure.
 * Note: td and/or p may be NULL when closing a file
 * that was being passed in a message.
 *
 * MPALMOSTSAFE - acquires mplock for VOP operations
 */
int
closef(struct file *fp, struct proc *p)
{
        struct vnode *vp;
        struct flock lf;
        struct filedesc_to_leader *fdtol;

        if (fp == NULL)
                return (0);

        /*
         * POSIX record locking dictates that any close releases ALL
         * locks owned by this process.  This is handled by setting
         * a flag in the unlock to free ONLY locks obeying POSIX
         * semantics, and not to free BSD-style file locks.
         * If the descriptor was in a message, POSIX-style locks
         * aren't passed with the descriptor.
         */
        if (p != NULL && fp->f_type == DTYPE_VNODE &&
            (((struct vnode *)fp->f_data)->v_flag & VMAYHAVELOCKS)
        ) {
                if (p->p_leader->p_advlock_flag) {
                        lf.l_whence = SEEK_SET;
                        lf.l_start = 0;
                        lf.l_len = 0;
                        lf.l_type = F_UNLCK;
                        vp = (struct vnode *)fp->f_data;
                        VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_UNLCK,
                                    &lf, F_POSIX);
                }
                fdtol = p->p_fdtol;
                if (fdtol != NULL) {
                        lwkt_gettoken(&p->p_token);

                        /*
                         * Handle special case where file descriptor table
                         * is shared between multiple process leaders.
                         */
                        for (fdtol = fdtol->fdl_next;
                             fdtol != p->p_fdtol;
                             fdtol = fdtol->fdl_next) {
                                if (fdtol->fdl_leader->p_advlock_flag == 0)
                                        continue;
                                fdtol->fdl_holdcount++;
                                lf.l_whence = SEEK_SET;
                                lf.l_start = 0;
                                lf.l_len = 0;
                                lf.l_type = F_UNLCK;
                                vp = (struct vnode *)fp->f_data;
                                VOP_ADVLOCK(vp, (caddr_t)fdtol->fdl_leader,
                                            F_UNLCK, &lf, F_POSIX);
                                fdtol->fdl_holdcount--;
                                if (fdtol->fdl_holdcount == 0 &&
                                    fdtol->fdl_wakeup != 0) {
                                        fdtol->fdl_wakeup = 0;
                                        wakeup(fdtol);
                                }
                        }
                        lwkt_reltoken(&p->p_token);
                }
        }
        return (fdrop(fp));
}

/*
 * fhold() can only be called if f_count is already at least 1 (i.e. the
 * caller of fhold() already has a reference to the file pointer in some
 * manner or other). 
 *
 * Atomic ops are used for incrementing and decrementing f_count before
 * the 1->0 transition.  f_count 1->0 transition is special, see the
 * comment in fdrop().
 */
void
fhold(struct file *fp)
{
        /* 0->1 transition will never work */
        KASSERT(fp->f_count > 0, ("fhold: invalid f_count %d", fp->f_count));
        atomic_add_int(&fp->f_count, 1);
}

/*
 * fdrop() - drop a reference to a descriptor
 */
int
fdrop(struct file *fp)
{
        struct flock lf;
        struct vnode *vp;
        int error, do_free = 0;

        /*
         * NOTE:
         * Simple atomic_fetchadd_int(f_count, -1) here will cause use-
         * after-free or double free (due to f_count 0->1 transition), if
         * fhold() is called on the fps found through filehead iteration.
         */
        for (;;) {
                int count = fp->f_count;

                cpu_ccfence();
                KASSERT(count > 0, ("fdrop: invalid f_count %d", count));
                if (count == 1) {
                        struct filelist_head *head = fp2filelist(fp);

                        /*
                         * About to drop the last reference, hold the
                         * filehead spin lock and drop it, so that no
                         * one could see this fp through filehead anymore,
                         * let alone fhold() this fp.
                         */
                        spin_lock(&head->spin);
                        if (atomic_cmpset_int(&fp->f_count, count, 0)) {
                                LIST_REMOVE(fp, f_list);
                                spin_unlock(&head->spin);
                                atomic_subtract_int(&nfiles, 1);
                                do_free = 1; /* free this fp */
                                break;
                        }
                        spin_unlock(&head->spin);
                        /* retry */
                } else if (atomic_cmpset_int(&fp->f_count, count, count - 1)) {
                        break;
                }
                /* retry */
        }
        if (!do_free)
                return (0);

        KKASSERT(SLIST_FIRST(&fp->f_klist) == NULL);

        /*
         * The last reference has gone away, we own the fp structure free
         * and clear.
         */
        if (fp->f_count < 0)
                panic("fdrop: count < 0");
        if ((fp->f_flag & FHASLOCK) && fp->f_type == DTYPE_VNODE &&
            (((struct vnode *)fp->f_data)->v_flag & VMAYHAVELOCKS)
        ) {
                lf.l_whence = SEEK_SET;
                lf.l_start = 0;
                lf.l_len = 0;
                lf.l_type = F_UNLCK;
                vp = (struct vnode *)fp->f_data;
                VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, 0);
        }
        if (fp->f_ops != &badfileops)
                error = fo_close(fp);
        else
                error = 0;
        ffree(fp);
        return (error);
}

/*
 * Apply an advisory lock on a file descriptor.
 *
 * Just attempt to get a record lock of the requested type on
 * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
 *
 * MPALMOSTSAFE
 */
int
sys_flock(struct flock_args *uap)
{
        thread_t td = curthread;
        struct file *fp;
        struct vnode *vp;
        struct flock lf;
        int error;

        if ((fp = holdfp(td, uap->fd, -1)) == NULL)
                return (EBADF);
        if (fp->f_type != DTYPE_VNODE) {
                error = EOPNOTSUPP;
                goto done;
        }
        vp = (struct vnode *)fp->f_data;
        lf.l_whence = SEEK_SET;
        lf.l_start = 0;
        lf.l_len = 0;
        if (uap->how & LOCK_UN) {
                lf.l_type = F_UNLCK;
                atomic_clear_int(&fp->f_flag, FHASLOCK); /* race ok */
                error = VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, 0);
                goto done;
        }
        if (uap->how & LOCK_EX)
                lf.l_type = F_WRLCK;
        else if (uap->how & LOCK_SH)
                lf.l_type = F_RDLCK;
        else {
                error = EBADF;
                goto done;
        }
        if (uap->how & LOCK_NB)
                error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, 0);
        else
                error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_WAIT);
        atomic_set_int(&fp->f_flag, FHASLOCK);  /* race ok */
done:
        fdrop(fp);
        return (error);
}

/*
 * File Descriptor pseudo-device driver (/dev/fd/).
 *
 * Opening minor device N dup()s the file (if any) connected to file
 * descriptor N belonging to the calling process.  Note that this driver
 * consists of only the ``open()'' routine, because all subsequent
 * references to this file will be direct to the other driver.
 */
static int
fdopen(struct dev_open_args *ap)
{
        thread_t td = curthread;

        KKASSERT(td->td_lwp != NULL);

        /*
         * XXX Kludge: set curlwp->lwp_dupfd to contain the value of the
         * the file descriptor being sought for duplication. The error
         * return ensures that the vnode for this device will be released
         * by vn_open. Open will detect this special error and take the
         * actions in dupfdopen below. Other callers of vn_open or VOP_OPEN
         * will simply report the error.
         */
        td->td_lwp->lwp_dupfd = minor(ap->a_head.a_dev);
        return (ENODEV);
}

/*
 * The caller has reserved the file descriptor dfd for us.  On success we
 * must fsetfd() it.  On failure the caller will clean it up.
 */
int
dupfdopen(thread_t td, int dfd, int sfd, int mode, int error)
{
        struct filedesc *fdp;
        struct file *wfp;
        struct file *xfp;
        int werror;

        if ((wfp = holdfp(td, sfd, -1)) == NULL)
                return (EBADF);

        /*
         * Close a revoke/dup race.  Duping a descriptor marked as revoked
         * will dup a dummy descriptor instead of the real one.
         */
        if (wfp->f_flag & FREVOKED) {
                kprintf("Warning: attempt to dup() a revoked descriptor\n");
                fdrop(wfp);
                wfp = NULL;
                werror = falloc(NULL, &wfp, NULL);
                if (werror)
                        return (werror);
        }

        fdp = td->td_proc->p_fd;

        /*
         * There are two cases of interest here.
         *
         * For ENODEV simply dup sfd to file descriptor dfd and return.
         *
         * For ENXIO steal away the file structure from sfd and store it
         * dfd.  sfd is effectively closed by this operation.
         *
         * Any other error code is just returned.
         */
        switch (error) {
        case ENODEV:
                /*
                 * Check that the mode the file is being opened for is a
                 * subset of the mode of the existing descriptor.
                 */
                if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
                        error = EACCES;
                        break;
                }
                spin_lock(&fdp->fd_spin);
                fdp->fd_files[dfd].fileflags = fdp->fd_files[sfd].fileflags;
                fsetfd_locked(fdp, wfp, dfd);
                spin_unlock(&fdp->fd_spin);
                error = 0;
                break;
        case ENXIO:
                /*
                 * Steal away the file pointer from dfd, and stuff it into indx.
                 */
                spin_lock(&fdp->fd_spin);
                fdp->fd_files[dfd].fileflags = fdp->fd_files[sfd].fileflags;
                fsetfd(fdp, wfp, dfd);
                if ((xfp = funsetfd_locked(fdp, sfd)) != NULL) {
                        spin_unlock(&fdp->fd_spin);
                        fdrop(xfp);
                } else {
                        spin_unlock(&fdp->fd_spin);
                }
                error = 0;
                break;
        default:
                break;
        }
        fdrop(wfp);
        return (error);
}

/*
 * NOT MPSAFE - I think these refer to a common file descriptor table
 * and we need to spinlock that to link fdtol in.
 */
struct filedesc_to_leader *
filedesc_to_leader_alloc(struct filedesc_to_leader *old,
                         struct proc *leader)
{
        struct filedesc_to_leader *fdtol;
        
        fdtol = kmalloc(sizeof(struct filedesc_to_leader), 
                        M_FILEDESC_TO_LEADER, M_WAITOK | M_ZERO);
        fdtol->fdl_refcount = 1;
        fdtol->fdl_holdcount = 0;
        fdtol->fdl_wakeup = 0;
        fdtol->fdl_leader = leader;
        if (old != NULL) {
                fdtol->fdl_next = old->fdl_next;
                fdtol->fdl_prev = old;
                old->fdl_next = fdtol;
                fdtol->fdl_next->fdl_prev = fdtol;
        } else {
                fdtol->fdl_next = fdtol;
                fdtol->fdl_prev = fdtol;
        }
        return fdtol;
}

/*
 * Scan all file pointers in the system.  The callback is made with
 * the master list spinlock held exclusively.
 */
void
allfiles_scan_exclusive(int (*callback)(struct file *, void *), void *data)
{
        int i;

        for (i = 0; i < NFILELIST_HEADS; ++i) {
                struct filelist_head *head = &filelist_heads[i];
                struct file *fp;

                spin_lock(&head->spin);
                LIST_FOREACH(fp, &head->list, f_list) {
                        int res;

                        res = callback(fp, data);
                        if (res < 0)
                                break;
                }
                spin_unlock(&head->spin);
        }
}

/*
 * Get file structures.
 *
 * NOT MPSAFE - process list scan, SYSCTL_OUT (probably not mpsafe)
 */

struct sysctl_kern_file_info {
        int count;
        int error;
        struct sysctl_req *req;
};

static int sysctl_kern_file_callback(struct proc *p, void *data);

static int
sysctl_kern_file(SYSCTL_HANDLER_ARGS)
{
        struct sysctl_kern_file_info info;

        /*
         * Note: because the number of file descriptors is calculated
         * in different ways for sizing vs returning the data,
         * there is information leakage from the first loop.  However,
         * it is of a similar order of magnitude to the leakage from
         * global system statistics such as kern.openfiles.
         *
         * When just doing a count, note that we cannot just count
         * the elements and add f_count via the filehead list because 
         * threaded processes share their descriptor table and f_count might
         * still be '1' in that case.
         *
         * Since the SYSCTL op can block, we must hold the process to
         * prevent it being ripped out from under us either in the 
         * file descriptor loop or in the greater LIST_FOREACH.  The
         * process may be in varying states of disrepair.  If the process
         * is in SZOMB we may have caught it just as it is being removed
         * from the allproc list, we must skip it in that case to maintain
         * an unbroken chain through the allproc list.
         */
        info.count = 0;
        info.error = 0;
        info.req = req;
        allproc_scan(sysctl_kern_file_callback, &info, 0);

        /*
         * When just calculating the size, overestimate a bit to try to
         * prevent system activity from causing the buffer-fill call 
         * to fail later on.
         */
        if (req->oldptr == NULL) {
                info.count = (info.count + 16) + (info.count / 10);
                info.error = SYSCTL_OUT(req, NULL,
                                        info.count * sizeof(struct kinfo_file));
        }
        return (info.error);
}

static int
sysctl_kern_file_callback(struct proc *p, void *data)
{
        struct sysctl_kern_file_info *info = data;
        struct kinfo_file kf;
        struct filedesc *fdp;
        struct file *fp;
        uid_t uid;
        int n;

        if (p->p_stat == SIDL || p->p_stat == SZOMB)
                return(0);
        if (!(PRISON_CHECK(info->req->td->td_ucred, p->p_ucred) != 0))
                return(0);

        /*
         * Softref the fdp to prevent it from being destroyed
         */
        spin_lock(&p->p_spin);
        if ((fdp = p->p_fd) == NULL) {
                spin_unlock(&p->p_spin);
                return(0);
        }
        atomic_add_int(&fdp->fd_softrefs, 1);
        spin_unlock(&p->p_spin);

        /*
         * The fdp's own spinlock prevents the contents from being
         * modified.
         */
        spin_lock_shared(&fdp->fd_spin);
        for (n = 0; n < fdp->fd_nfiles; ++n) {
                if ((fp = fdp->fd_files[n].fp) == NULL)
                        continue;
                if (info->req->oldptr == NULL) {
                        ++info->count;
                } else {
                        uid = p->p_ucred ? p->p_ucred->cr_uid : -1;
                        kcore_make_file(&kf, fp, p->p_pid, uid, n);
                        spin_unlock_shared(&fdp->fd_spin);
                        info->error = SYSCTL_OUT(info->req, &kf, sizeof(kf));
                        spin_lock_shared(&fdp->fd_spin);
                        if (info->error)
                                break;
                }
        }
        spin_unlock_shared(&fdp->fd_spin);
        atomic_subtract_int(&fdp->fd_softrefs, 1);
        if (info->error)
                return(-1);
        return(0);
}

SYSCTL_PROC(_kern, KERN_FILE, file, CTLTYPE_OPAQUE|CTLFLAG_RD,
    0, 0, sysctl_kern_file, "S,file", "Entire file table");

SYSCTL_INT(_kern, OID_AUTO, minfilesperproc, CTLFLAG_RW,
    &minfilesperproc, 0, "Minimum files allowed open per process");
SYSCTL_INT(_kern, KERN_MAXFILESPERPROC, maxfilesperproc, CTLFLAG_RW, 
    &maxfilesperproc, 0, "Maximum files allowed open per process");
SYSCTL_INT(_kern, OID_AUTO, maxfilesperuser, CTLFLAG_RW,
    &maxfilesperuser, 0, "Maximum files allowed open per user");

SYSCTL_INT(_kern, KERN_MAXFILES, maxfiles, CTLFLAG_RW, 
    &maxfiles, 0, "Maximum number of files");

SYSCTL_INT(_kern, OID_AUTO, maxfilesrootres, CTLFLAG_RW, 
    &maxfilesrootres, 0, "Descriptors reserved for root use");

SYSCTL_INT(_kern, OID_AUTO, openfiles, CTLFLAG_RD, 
        &nfiles, 0, "System-wide number of open files");

static void
fildesc_drvinit(void *unused)
{
        int fd;

        for (fd = 0; fd < NUMFDESC; fd++) {
                make_dev(&fildesc_ops, fd,
                         UID_BIN, GID_BIN, 0666, "fd/%d", fd);
        }

        make_dev(&fildesc_ops, 0, UID_ROOT, GID_WHEEL, 0666, "stdin");
        make_dev(&fildesc_ops, 1, UID_ROOT, GID_WHEEL, 0666, "stdout");
        make_dev(&fildesc_ops, 2, UID_ROOT, GID_WHEEL, 0666, "stderr");
}

struct fileops badfileops = {
        .fo_read = badfo_readwrite,
        .fo_write = badfo_readwrite,
        .fo_ioctl = badfo_ioctl,
        .fo_kqfilter = badfo_kqfilter,
        .fo_stat = badfo_stat,
        .fo_close = badfo_close,
        .fo_shutdown = badfo_shutdown
};

int
badfo_readwrite(
        struct file *fp,
        struct uio *uio,
        struct ucred *cred,
        int flags
) {
        return (EBADF);
}

int
badfo_ioctl(struct file *fp, u_long com, caddr_t data,
            struct ucred *cred, struct sysmsg *msgv)
{
        return (EBADF);
}

/*
 * Must return an error to prevent registration, typically
 * due to a revoked descriptor (file_filtops assigned).
 */
int
badfo_kqfilter(struct file *fp, struct knote *kn)
{
        return (EOPNOTSUPP);
}

int
badfo_stat(struct file *fp, struct stat *sb, struct ucred *cred)
{
        return (EBADF);
}

int
badfo_close(struct file *fp)
{
        return (EBADF);
}

int
badfo_shutdown(struct file *fp, int how)
{
        return (EBADF);
}

int
nofo_shutdown(struct file *fp, int how)
{
        return (EOPNOTSUPP);
}

SYSINIT(fildescdev, SI_SUB_DRIVERS, SI_ORDER_MIDDLE + CDEV_MAJOR,
    fildesc_drvinit,NULL);

static void
filelist_heads_init(void *arg __unused)
{
        int i;

        for (i = 0; i < NFILELIST_HEADS; ++i) {
                struct filelist_head *head = &filelist_heads[i];

                spin_init(&head->spin, "filehead_spin");
                LIST_INIT(&head->list);
        }
}

SYSINIT(filelistheads, SI_BOOT1_LOCK, SI_ORDER_ANY,
    filelist_heads_init, NULL);

static void
file_objcache_init(void *dummy __unused)
{
        file_objcache = objcache_create("file", maxfiles, maxfiles / 8,
            NULL, NULL, NULL, /* TODO: ctor/dtor */
            objcache_malloc_alloc, objcache_malloc_free, &file_malloc_args);
}
SYSINIT(fpobjcache, SI_BOOT2_POST_SMP, SI_ORDER_ANY, file_objcache_init, NULL);