xnu-6153.81.5.tar.gz

[apple/xnu.git] / bsd / miscfs / devfs / devfs_tree.c

/*
 * Copyright (c) 2000-2014 Apple Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 *
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. The rights granted to you under the License
 * may not be used to create, or enable the creation or redistribution of,
 * unlawful or unlicensed copies of an Apple operating system, or to
 * circumvent, violate, or enable the circumvention or violation of, any
 * terms of an Apple operating system software license agreement.
 *
 * Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this file.
 *
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 */

/*
 * Copyright 1997,1998 Julian Elischer.  All rights reserved.
 * julian@freebsd.org
 *
 * 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.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER ``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 HOLDER 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.
 *
 * devfs_tree.c
 */
/*
 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
 * support for mandatory and extensible security protections.  This notice
 * is included in support of clause 2.2 (b) of the Apple Public License,
 * Version 2.0.
 */

/*
 * HISTORY
 *  Dieter Siegmund (dieter@apple.com) Thu Apr  8 14:08:19 PDT 1999
 *  - removed mounting of "hidden" mountpoint
 *  - fixed problem in which devnode->dn_vn pointer was not
 *    updated with the vnode returned from checkalias()
 *  - replaced devfs_vntodn() with a macro VTODN()
 *  - rewrote dev_finddir() to not use recursion
 *  - added locking to avoid data structure corruption (DEVFS_(UN)LOCK())
 *  Dieter Siegmund (dieter@apple.com) Wed Jul 14 13:37:59 PDT 1999
 *  - fixed problem with devfs_dntovn() checking the v_id against the
 *    value cached in the device node; a union mount on top of us causes
 *    the v_id to get incremented thus, we would end up returning a new
 *    vnode instead of the existing one that has the mounted_here
 *    field filled in; the net effect was that the filesystem mounted
 *    on top of us would never show up
 *  - added devfs_stats to store how many data structures are actually
 *    allocated
 */

/* SPLIT_DEVS means each devfs uses a different devnode for the same device */
/* Otherwise the same device always ends up at the same vnode even if  */
/* reached througgh a different devfs instance. The practical difference */
/* is that with the same vnode, chmods and chowns show up on all instances of */
/* a device. (etc) */

#define SPLIT_DEVS 1 /* maybe make this an option */
/*#define SPLIT_DEVS 1*/

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/conf.h>
#include <sys/malloc.h>
#include <sys/mount_internal.h>
#include <sys/proc.h>
#include <sys/vnode_internal.h>
#include <stdarg.h>
#include <libkern/OSAtomic.h>
#include <os/refcnt.h>
#define BSD_KERNEL_PRIVATE      1       /* devfs_make_link() prototype */
#include "devfs.h"
#include "devfsdefs.h"

#if CONFIG_MACF
#include <security/mac_framework.h>
#endif

#if FDESC
#include "fdesc.h"
#endif

typedef struct devfs_vnode_event {
        vnode_t                 dve_vp;
        uint32_t                dve_vid;
        uint32_t                dve_events;
} *devfs_vnode_event_t;

/*
 * Size of stack buffer (fast path) for notifications.  If
 * the number of mounts is small, no need to malloc a buffer.
 */
#define NUM_STACK_ENTRIES 5

typedef struct devfs_event_log {
        size_t                  del_max;
        size_t                  del_used;
        devfs_vnode_event_t     del_entries;
} *devfs_event_log_t;


static void     dev_free_hier(devdirent_t *);
static int      devfs_propogate(devdirent_t *, devdirent_t *, devfs_event_log_t);
static int      dev_finddir(const char *, devnode_t *, int, devnode_t **, devfs_event_log_t);
static int      dev_dup_entry(devnode_t *, devdirent_t *, devdirent_t **, struct devfsmount *);
void            devfs_ref_node(devnode_t *);
void            devfs_rele_node(devnode_t *);
static void     devfs_record_event(devfs_event_log_t, devnode_t*, uint32_t);
static int      devfs_init_event_log(devfs_event_log_t, uint32_t, devfs_vnode_event_t);
static void     devfs_release_event_log(devfs_event_log_t, int);
static void     devfs_bulk_notify(devfs_event_log_t);
static devdirent_t *devfs_make_node_internal(dev_t, devfstype_t type, uid_t, gid_t, int,
    int (*clone)(dev_t dev, int action), const char *fmt, va_list ap);


lck_grp_t       * devfs_lck_grp;
lck_grp_attr_t  * devfs_lck_grp_attr;
lck_attr_t      * devfs_lck_attr;
lck_mtx_t         devfs_mutex;
lck_mtx_t         devfs_attr_mutex;

os_refgrp_decl(static, devfs_refgrp, "devfs", NULL);

devdirent_t *           dev_root = NULL;        /* root of backing tree */
struct devfs_stats      devfs_stats;            /* hold stats */

static ino_t            devfs_unique_fileno = 0;

#ifdef HIDDEN_MOUNTPOINT
static struct mount *devfs_hidden_mount;
#endif /* HIDDEN_MOINTPOINT */

static int devfs_ready = 0;
static uint32_t devfs_nmountplanes = 0; /* The first plane is not used for a mount */

#define DEVFS_NOCREATE  FALSE
#define DEVFS_CREATE    TRUE

/*
 * Set up the root directory node in the backing plane
 * This is happenning before the vfs system has been
 * set up yet, so be careful about what we reference..
 * Notice that the ops are by indirection.. as they haven't
 * been set up yet!
 * DEVFS has a hidden mountpoint that is used as the anchor point
 * for the internal 'blueprint' version of the dev filesystem tree.
 */
/*proto*/
int
devfs_sinit(void)
{
        int error;

        devfs_lck_grp_attr = lck_grp_attr_alloc_init();
        devfs_lck_grp = lck_grp_alloc_init("devfs_lock", devfs_lck_grp_attr);

        devfs_lck_attr = lck_attr_alloc_init();

        lck_mtx_init(&devfs_mutex, devfs_lck_grp, devfs_lck_attr);
        lck_mtx_init(&devfs_attr_mutex, devfs_lck_grp, devfs_lck_attr);

        DEVFS_LOCK();
        error = dev_add_entry("root", NULL, DEV_DIR, NULL, NULL, NULL, &dev_root);
        DEVFS_UNLOCK();

        if (error) {
                printf("devfs_sinit: dev_add_entry failed ");
                return ENOTSUP;
        }
#ifdef HIDDEN_MOUNTPOINT
        MALLOC(devfs_hidden_mount, struct mount *, sizeof(struct mount),
            M_MOUNT, M_WAITOK);
        bzero(devfs_hidden_mount, sizeof(struct mount));
        mount_lock_init(devfs_hidden_mount);
        TAILQ_INIT(&devfs_hidden_mount->mnt_vnodelist);
        TAILQ_INIT(&devfs_hidden_mount->mnt_workerqueue);
        TAILQ_INIT(&devfs_hidden_mount->mnt_newvnodes);
#if CONFIG_MACF
        mac_mount_label_init(devfs_hidden_mount);
        mac_mount_label_associate(vfs_context_kernel(), devfs_hidden_mount);
#endif

        /* Initialize the default IO constraints */
        mp->mnt_maxreadcnt = mp->mnt_maxwritecnt = MAXPHYS;
        mp->mnt_segreadcnt = mp->mnt_segwritecnt = 32;
        mp->mnt_ioflags = 0;
        mp->mnt_realrootvp = NULLVP;
        mp->mnt_authcache_ttl = CACHED_LOOKUP_RIGHT_TTL;

        devfs_mount(devfs_hidden_mount, "dummy", NULL, NULL, NULL);
        dev_root->de_dnp->dn_dvm
                = (struct devfsmount *)devfs_hidden_mount->mnt_data;
#endif /* HIDDEN_MOUNTPOINT */
#if CONFIG_MACF
        mac_devfs_label_associate_directory("/", strlen("/"),
            dev_root->de_dnp, "/");
#endif
        devfs_ready = 1;
        return 0;
}

/***********************************************************************\
*************************************************************************
*       Routines used to find our way to a point in the tree            *
*************************************************************************
\***********************************************************************/



/***************************************************************
* Search down the linked list off a dir to find "name"
* return the devnode_t * for that node.
*
* called with DEVFS_LOCK held
***************************************************************/
devdirent_t *
dev_findname(devnode_t * dir, const char *name)
{
        devdirent_t * newfp;
        if (dir->dn_type != DEV_DIR) {
                return 0;                     /*XXX*/ /* printf?*/
        }
        if (name[0] == '.') {
                if (name[1] == 0) {
                        return dir->dn_typeinfo.Dir.myname;
                }
                if ((name[1] == '.') && (name[2] == 0)) {
                        /* for root, .. == . */
                        return dir->dn_typeinfo.Dir.parent->dn_typeinfo.Dir.myname;
                }
        }
        newfp = dir->dn_typeinfo.Dir.dirlist;

        while (newfp) {
                if (!(strncmp(name, newfp->de_name, sizeof(newfp->de_name)))) {
                        return newfp;
                }
                newfp = newfp->de_next;
        }
        return NULL;
}

/***********************************************************************
* Given a starting node (0 for root) and a pathname, return the node
* for the end item on the path. It MUST BE A DIRECTORY. If the 'DEVFS_CREATE'
* option is true, then create any missing nodes in the path and create
* and return the final node as well.
* This is used to set up a directory, before making nodes in it..
*
* called with DEVFS_LOCK held
***********************************************************************/
static int
dev_finddir(const char * path,
    devnode_t * dirnode,
    int create,
    devnode_t * * dn_pp,
    devfs_event_log_t delp)
{
        devnode_t *     dnp = NULL;
        int             error = 0;
        const char *            scan;
#if CONFIG_MACF
        char            fullpath[DEVMAXPATHSIZE];
#endif


        if (!dirnode) { /* dirnode == NULL means start at root */
                dirnode = dev_root->de_dnp;
        }

        if (dirnode->dn_type != DEV_DIR) {
                return ENOTDIR;
        }

        if (strlen(path) > (DEVMAXPATHSIZE - 1)) {
                return ENAMETOOLONG;
        }

#if CONFIG_MACF
        strlcpy(fullpath, path, DEVMAXPATHSIZE);
#endif
        scan = path;

        while (*scan == '/') {
                scan++;
        }

        *dn_pp = NULL;

        while (1) {
                char                component[DEVMAXPATHSIZE];
                devdirent_t *       dirent_p;
                const char *        start;

                if (*scan == 0) {
                        /* we hit the end of the string, we're done */
                        *dn_pp = dirnode;
                        break;
                }
                start = scan;
                while (*scan != '/' && *scan) {
                        scan++;
                }

                strlcpy(component, start, (scan - start) + 1);
                if (*scan == '/') {
                        scan++;
                }

                dirent_p = dev_findname(dirnode, component);
                if (dirent_p) {
                        dnp = dirent_p->de_dnp;
                        if (dnp->dn_type != DEV_DIR) {
                                error = ENOTDIR;
                                break;
                        }
                } else {
                        if (!create) {
                                error = ENOENT;
                                break;
                        }
                        error = dev_add_entry(component, dirnode,
                            DEV_DIR, NULL, NULL, NULL, &dirent_p);
                        if (error) {
                                break;
                        }
                        dnp = dirent_p->de_dnp;
#if CONFIG_MACF
                        mac_devfs_label_associate_directory(
                                dirnode->dn_typeinfo.Dir.myname->de_name,
                                strlen(dirnode->dn_typeinfo.Dir.myname->de_name),
                                dnp, fullpath);
#endif
                        devfs_propogate(dirnode->dn_typeinfo.Dir.myname, dirent_p, delp);
                }
                dirnode = dnp; /* continue relative to this directory */
        }
        return error;
}


/***********************************************************************
* Add a new NAME element to the devfs
* If we're creating a root node, then dirname is NULL
* Basically this creates a new namespace entry for the device node
*
* Creates a name node, and links it to the supplied node
*
* called with DEVFS_LOCK held
***********************************************************************/
int
dev_add_name(const char * name, devnode_t * dirnode, __unused devdirent_t * back,
    devnode_t * dnp, devdirent_t * *dirent_pp)
{
        devdirent_t *   dirent_p = NULL;

        if (dirnode != NULL) {
                if (dirnode->dn_type != DEV_DIR) {
                        return ENOTDIR;
                }

                if (dev_findname(dirnode, name)) {
                        return EEXIST;
                }
        }
        /*
         * make sure the name is legal
         * slightly misleading in the case of NULL
         */
        if (!name || (strlen(name) > (DEVMAXNAMESIZE - 1))) {
                return ENAMETOOLONG;
        }

        /*
         * Allocate and fill out a new directory entry
         */
        MALLOC(dirent_p, devdirent_t *, sizeof(devdirent_t),
            M_DEVFSNAME, M_WAITOK);
        if (!dirent_p) {
                return ENOMEM;
        }
        bzero(dirent_p, sizeof(devdirent_t));

        /* inherrit our parent's mount info */ /*XXX*/
        /* a kludge but.... */
        if (dirnode && (dnp->dn_dvm == NULL)) {
                dnp->dn_dvm = dirnode->dn_dvm;
                /* if(!dnp->dn_dvm) printf("parent had null dvm "); */
        }

        /*
         * Link the two together
         * include the implicit link in the count of links to the devnode..
         * this stops it from being accidentally freed later.
         */
        dirent_p->de_dnp = dnp;
        dnp->dn_links++;  /* implicit from our own name-node */

        /*
         * Make sure that we can find all the links that reference a node
         * so that we can get them all if we need to zap the node.
         */
        if (dnp->dn_linklist) {
                dirent_p->de_nextlink = dnp->dn_linklist;
                dirent_p->de_prevlinkp = dirent_p->de_nextlink->de_prevlinkp;
                dirent_p->de_nextlink->de_prevlinkp = &(dirent_p->de_nextlink);
                *dirent_p->de_prevlinkp = dirent_p;
        } else {
                dirent_p->de_nextlink = dirent_p;
                dirent_p->de_prevlinkp = &(dirent_p->de_nextlink);
        }
        dnp->dn_linklist = dirent_p;

        /*
         * If the node is a directory, then we need to handle the
         * creation of the .. link.
         * A NULL dirnode indicates a root node, so point to ourself.
         */
        if (dnp->dn_type == DEV_DIR) {
                dnp->dn_typeinfo.Dir.myname = dirent_p;
                /*
                 * If we are unlinking from an old dir, decrement its links
                 * as we point our '..' elsewhere
                 * Note: it's up to the calling code to remove the
                 * us from the original directory's list
                 */
                if (dnp->dn_typeinfo.Dir.parent) {
                        dnp->dn_typeinfo.Dir.parent->dn_links--;
                }
                if (dirnode) {
                        dnp->dn_typeinfo.Dir.parent = dirnode;
                } else {
                        dnp->dn_typeinfo.Dir.parent = dnp;
                }
                dnp->dn_typeinfo.Dir.parent->dn_links++; /* account for the new '..' */
        }

        /*
         * put the name into the directory entry.
         */
        strlcpy(dirent_p->de_name, name, DEVMAXNAMESIZE);


        /*
         * Check if we are not making a root node..
         * (i.e. have parent)
         */
        if (dirnode) {
                /*
                 * Put it on the END of the linked list of directory entries
                 */
                dirent_p->de_parent = dirnode; /* null for root */
                dirent_p->de_prevp = dirnode->dn_typeinfo.Dir.dirlast;
                dirent_p->de_next = *(dirent_p->de_prevp); /* should be NULL */
                                                           /*right?*/
                *(dirent_p->de_prevp) = dirent_p;
                dirnode->dn_typeinfo.Dir.dirlast = &(dirent_p->de_next);
                dirnode->dn_typeinfo.Dir.entrycount++;
                dirnode->dn_len += strlen(name) + 8;/*ok, ok?*/
        }

        *dirent_pp = dirent_p;
        DEVFS_INCR_ENTRIES();
        return 0;
}


/***********************************************************************
* Add a new element to the devfs plane.
*
* Creates a new dev_node to go with it if the prototype should not be
* reused. (Is a DIR, or we select SPLIT_DEVS at compile time)
* typeinfo gives us info to make our node if we don't have a prototype.
* If typeinfo is null and proto exists, then the typeinfo field of
* the proto is used intead in the DEVFS_CREATE case.
* note the 'links' count is 0 (except if a dir)
* but it is only cleared on a transition
* so this is ok till we link it to something
* Even in SPLIT_DEVS mode,
* if the node already exists on the wanted plane, just return it
*
* called with DEVFS_LOCK held
***********************************************************************/
int
dev_add_node(int entrytype, devnode_type_t * typeinfo, devnode_t * proto,
    devnode_t * *dn_pp, struct devfsmount *dvm)
{
        devnode_t *     dnp = NULL;
        int     error = 0;

#if defined SPLIT_DEVS
        /*
         * If we have a prototype, then check if there is already a sibling
         * on the mount plane we are looking at, if so, just return it.
         */
        if (proto) {
                dnp = proto->dn_nextsibling;
                while (dnp != proto) {
                        if (dnp->dn_dvm == dvm) {
                                *dn_pp = dnp;
                                return 0;
                        }
                        dnp = dnp->dn_nextsibling;
                }
                if (typeinfo == NULL) {
                        typeinfo = &(proto->dn_typeinfo);
                }
        }
#else   /* SPLIT_DEVS */
        if (proto) {
                switch (proto->type) {
                case DEV_BDEV:
                case DEV_CDEV:
                        *dn_pp = proto;
                        return 0;
                }
        }
#endif  /* SPLIT_DEVS */
        MALLOC(dnp, devnode_t *, sizeof(devnode_t), M_DEVFSNODE, M_WAITOK);
        if (!dnp) {
                return ENOMEM;
        }

        /*
         * If we have a proto, that means that we are duplicating some
         * other device, which can only happen if we are not at the back plane
         */
        if (proto) {
                bcopy(proto, dnp, sizeof(devnode_t));
                dnp->dn_links = 0;
                dnp->dn_linklist = NULL;
                dnp->dn_vn = NULL;
                dnp->dn_len = 0;
                /* add to END of siblings list */
                dnp->dn_prevsiblingp = proto->dn_prevsiblingp;
                *(dnp->dn_prevsiblingp) = dnp;
                dnp->dn_nextsibling = proto;
                proto->dn_prevsiblingp = &(dnp->dn_nextsibling);
#if CONFIG_MACF
                mac_devfs_label_init(dnp);
                mac_devfs_label_copy(proto->dn_label, dnp->dn_label);
#endif
        } else {
                struct timeval tv;

                /*
                 * We have no prototype, so start off with a clean slate
                 */
                microtime(&tv);
                bzero(dnp, sizeof(devnode_t));
                dnp->dn_type = entrytype;
                dnp->dn_nextsibling = dnp;
                dnp->dn_prevsiblingp = &(dnp->dn_nextsibling);
                dnp->dn_atime.tv_sec = tv.tv_sec;
                dnp->dn_mtime.tv_sec = tv.tv_sec;
                dnp->dn_ctime.tv_sec = tv.tv_sec;
#if CONFIG_MACF
                mac_devfs_label_init(dnp);
#endif
        }
        dnp->dn_dvm = dvm;

        /* Note: this inits the reference count to 1, this is considered unreferenced */
        os_ref_init_raw(&dnp->dn_refcount, &devfs_refgrp);
        dnp->dn_ino = devfs_unique_fileno;
        devfs_unique_fileno++;

        /*
         * fill out the dev node according to type
         */
        switch (entrytype) {
        case DEV_DIR:
                /*
                 * As it's a directory, make sure
                 * it has a null entries list
                 */
                dnp->dn_typeinfo.Dir.dirlast = &(dnp->dn_typeinfo.Dir.dirlist);
                dnp->dn_typeinfo.Dir.dirlist = (devdirent_t *)0;
                dnp->dn_typeinfo.Dir.entrycount = 0;
                /*  until we know better, it has a null parent pointer*/
                dnp->dn_typeinfo.Dir.parent = NULL;
                dnp->dn_links++; /* for .*/
                dnp->dn_typeinfo.Dir.myname = NULL;
                /*
                 * make sure that the ops associated with it are the ops
                 * that we use (by default) for directories
                 */
                dnp->dn_ops = &devfs_vnodeop_p;
                dnp->dn_mode |= 0555;   /* default perms */
                break;
        case DEV_SLNK:
                /*
                 * As it's a symlink allocate and store the link info
                 * Symlinks should only ever be created by the user,
                 * so they are not on the back plane and should not be
                 * propogated forward.. a bit like directories in that way..
                 * A symlink only exists on one plane and has its own
                 * node.. therefore we might be on any random plane.
                 */
                MALLOC(dnp->dn_typeinfo.Slnk.name, char *,
                    typeinfo->Slnk.namelen + 1,
                    M_DEVFSNODE, M_WAITOK);
                if (!dnp->dn_typeinfo.Slnk.name) {
                        error = ENOMEM;
                        break;
                }
                strlcpy(dnp->dn_typeinfo.Slnk.name, typeinfo->Slnk.name,
                    typeinfo->Slnk.namelen + 1);
                dnp->dn_typeinfo.Slnk.namelen = typeinfo->Slnk.namelen;
                DEVFS_INCR_STRINGSPACE(dnp->dn_typeinfo.Slnk.namelen + 1);
                dnp->dn_ops = &devfs_vnodeop_p;
                dnp->dn_mode |= 0555;   /* default perms */
                break;
        case DEV_CDEV:
        case DEV_BDEV:
                /*
                 * Make sure it has DEVICE type ops
                 * and device specific fields are correct
                 */
                dnp->dn_ops = &devfs_spec_vnodeop_p;
                dnp->dn_typeinfo.dev = typeinfo->dev;
                break;

        #if FDESC
        /* /dev/fd is special */
        case DEV_DEVFD:
                dnp->dn_ops = &devfs_devfd_vnodeop_p;
                dnp->dn_mode |= 0555;   /* default perms */
                break;

        #endif /* FDESC */
        default:
                error = EINVAL;
        }

        if (error) {
                FREE(dnp, M_DEVFSNODE);
        } else {
                *dn_pp = dnp;
                DEVFS_INCR_NODES();
        }

        return error;
}


/***********************************************************************
 * called with DEVFS_LOCK held
 **********************************************************************/
void
devnode_free(devnode_t * dnp)
{
#if CONFIG_MACF
        mac_devfs_label_destroy(dnp);
#endif
        if (dnp->dn_type == DEV_SLNK) {
                DEVFS_DECR_STRINGSPACE(dnp->dn_typeinfo.Slnk.namelen + 1);
                FREE(dnp->dn_typeinfo.Slnk.name, M_DEVFSNODE);
        }
        DEVFS_DECR_NODES();
        FREE(dnp, M_DEVFSNODE);
}


/***********************************************************************
 * called with DEVFS_LOCK held
 **********************************************************************/
static void
devfs_dn_free(devnode_t * dnp)
{
        if (--dnp->dn_links <= 0) { /* can be -1 for initial free, on error */
                /*probably need to do other cleanups XXX */
                if (dnp->dn_nextsibling != dnp) {
                        devnode_t * *   prevp = dnp->dn_prevsiblingp;
                        *prevp = dnp->dn_nextsibling;
                        dnp->dn_nextsibling->dn_prevsiblingp = prevp;
                }

                /* Can only free if there are no references; otherwise, wait for last vnode to be reclaimed */
                os_ref_count_t rc = os_ref_get_count_raw(&dnp->dn_refcount);
                if (rc == 1) {
                        /* release final reference from dev_add_node */
                        (void) os_ref_release_locked_raw(&dnp->dn_refcount, &devfs_refgrp);
                        devnode_free(dnp);
                } else {
                        dnp->dn_lflags |= DN_DELETE;
                }
        }
}

/***********************************************************************\
*       Front Node Operations                                           *
*       Add or delete a chain of front nodes                            *
\***********************************************************************/


/***********************************************************************
* Given a directory backing node, and a child backing node, add the
* appropriate front nodes to the front nodes of the directory to
* represent the child node to the user
*
* on failure, front nodes will either be correct or not exist for each
* front dir, however dirs completed will not be stripped of completed
* frontnodes on failure of a later frontnode
*
* This allows a new node to be propogated through all mounted planes
*
* called with DEVFS_LOCK held
***********************************************************************/
static int
devfs_propogate(devdirent_t * parent, devdirent_t * child, devfs_event_log_t delp)
{
        int     error;
        devdirent_t * newnmp;
        devnode_t *     dnp = child->de_dnp;
        devnode_t *     pdnp = parent->de_dnp;
        devnode_t *     adnp = parent->de_dnp;
        int type = child->de_dnp->dn_type;
        uint32_t events;

        events = (dnp->dn_type == DEV_DIR ? VNODE_EVENT_DIR_CREATED : VNODE_EVENT_FILE_CREATED);
        if (delp != NULL) {
                devfs_record_event(delp, pdnp, events);
        }

        /***********************************************
        * Find the other instances of the parent node
        ***********************************************/
        for (adnp = pdnp->dn_nextsibling;
            adnp != pdnp;
            adnp = adnp->dn_nextsibling) {
                /*
                 * Make the node, using the original as a prototype)
                 * if the node already exists on that plane it won't be
                 * re-made..
                 */
                if ((error = dev_add_entry(child->de_name, adnp, type,
                    NULL, dnp, adnp->dn_dvm,
                    &newnmp)) != 0) {
                        printf("duplicating %s failed\n", child->de_name);
                } else {
                        if (delp != NULL) {
                                devfs_record_event(delp, adnp, events);

                                /*
                                 * Slightly subtle.  We're guaranteed that there will
                                 * only be a vnode hooked into this devnode if we're creating
                                 * a new link to an existing node; otherwise, the devnode is new
                                 * and no one can have looked it up yet. If we're making a link,
                                 * then the buffer is large enough for two nodes in each
                                 * plane; otherwise, there's no vnode and this call will
                                 * do nothing.
                                 */
                                devfs_record_event(delp, newnmp->de_dnp, VNODE_EVENT_LINK);
                        }
                }
        }
        return 0;       /* for now always succeed */
}

static uint32_t
remove_notify_count(devnode_t *dnp)
{
        uint32_t notify_count = 0;
        devnode_t *dnp2;

        /*
         * Could need to notify for one removed node on each mount and
         * one parent for each such node.
         */
        notify_count = devfs_nmountplanes;
        notify_count += dnp->dn_links;
        for (dnp2 = dnp->dn_nextsibling; dnp2 != dnp; dnp2 = dnp2->dn_nextsibling) {
                notify_count += dnp2->dn_links;
        }

        return notify_count;
}

/***********************************************************************
* remove all instances of this devicename [for backing nodes..]
* note.. if there is another link to the node (non dir nodes only)
* then the devfs_node will still exist as the ref count will be non-0
* removing a directory node will remove all sup-nodes on all planes (ZAP)
*
* Used by device drivers to remove nodes that are no longer relevant
* The argument is the 'cookie' they were given when they created the node
* this function is exported.. see devfs.h
***********************************************************************/
void
devfs_remove(void *dirent_p)
{
        devnode_t * dnp = ((devdirent_t *)dirent_p)->de_dnp;
        devnode_t * dnp2;
        boolean_t   lastlink;
        struct devfs_event_log event_log;
        uint32_t    log_count = 0;
        int         do_notify = 0;
        int         need_free = 0;
        struct devfs_vnode_event stackbuf[NUM_STACK_ENTRIES];

        DEVFS_LOCK();

        if (!devfs_ready) {
                printf("devfs_remove: not ready for devices!\n");
                goto out;
        }

        log_count = remove_notify_count(dnp);

        if (log_count > NUM_STACK_ENTRIES) {
                uint32_t new_count;
wrongsize:
                DEVFS_UNLOCK();
                if (devfs_init_event_log(&event_log, log_count, NULL) == 0) {
                        do_notify = 1;
                        need_free = 1;
                }
                DEVFS_LOCK();

                new_count = remove_notify_count(dnp);
                if (need_free && (new_count > log_count)) {
                        devfs_release_event_log(&event_log, 1);
                        need_free = 0;
                        do_notify = 0;
                        log_count = log_count * 2;
                        goto wrongsize;
                }
        } else {
                if (devfs_init_event_log(&event_log, NUM_STACK_ENTRIES, &stackbuf[0]) == 0) {
                        do_notify = 1;
                }
        }

        /* This file has been deleted */
        if (do_notify != 0) {
                devfs_record_event(&event_log, dnp, VNODE_EVENT_DELETE);
        }

        /* keep removing the next sibling till only we exist. */
        while ((dnp2 = dnp->dn_nextsibling) != dnp) {
                /*
                 * Keep removing the next front node till no more exist
                 */
                dnp->dn_nextsibling = dnp2->dn_nextsibling;
                dnp->dn_nextsibling->dn_prevsiblingp = &(dnp->dn_nextsibling);
                dnp2->dn_nextsibling = dnp2;
                dnp2->dn_prevsiblingp = &(dnp2->dn_nextsibling);

                /* This file has been deleted in this plane */
                if (do_notify != 0) {
                        devfs_record_event(&event_log, dnp2, VNODE_EVENT_DELETE);
                }

                if (dnp2->dn_linklist) {
                        do {
                                lastlink = (1 == dnp2->dn_links);
                                /* Each parent of a link to this file has lost a child in this plane */
                                if (do_notify != 0) {
                                        devfs_record_event(&event_log, dnp2->dn_linklist->de_parent, VNODE_EVENT_FILE_REMOVED);
                                }
                                dev_free_name(dnp2->dn_linklist);
                        } while (!lastlink);
                }
        }

        /*
         * then free the main node
         * If we are not running in SPLIT_DEVS mode, then
         * THIS is what gets rid of the propogated nodes.
         */
        if (dnp->dn_linklist) {
                do {
                        lastlink = (1 == dnp->dn_links);
                        /* Each parent of a link to this file has lost a child */
                        if (do_notify != 0) {
                                devfs_record_event(&event_log, dnp->dn_linklist->de_parent, VNODE_EVENT_FILE_REMOVED);
                        }
                        dev_free_name(dnp->dn_linklist);
                } while (!lastlink);
        }
out:
        DEVFS_UNLOCK();
        if (do_notify != 0) {
                devfs_bulk_notify(&event_log);
                devfs_release_event_log(&event_log, need_free);
        }

        return;
}



/***************************************************************
 * duplicate the backing tree into a tree of nodes hung off the
 * mount point given as the argument. Do this by
 * calling dev_dup_entry which recurses all the way
 * up the tree..
 *
 * called with DEVFS_LOCK held
 **************************************************************/
int
dev_dup_plane(struct devfsmount *devfs_mp_p)
{
        devdirent_t *   new;
        int             error = 0;

        if ((error = dev_dup_entry(NULL, dev_root, &new, devfs_mp_p))) {
                return error;
        }
        devfs_mp_p->plane_root = new;
        devfs_nmountplanes++;
        return error;
}



/***************************************************************
* Free a whole plane
*
* called with DEVFS_LOCK held
***************************************************************/
void
devfs_free_plane(struct devfsmount *devfs_mp_p)
{
        devdirent_t * dirent_p;

        dirent_p = devfs_mp_p->plane_root;
        if (dirent_p) {
                dev_free_hier(dirent_p);
                dev_free_name(dirent_p);
        }
        devfs_mp_p->plane_root = NULL;
        devfs_nmountplanes--;

        if (devfs_nmountplanes > (devfs_nmountplanes + 1)) {
                panic("plane count wrapped around.\n");
        }
}


/***************************************************************
* Create and link in a new front element..
* Parent can be 0 for a root node
* Not presently usable to make a symlink XXX
* (Ok, symlinks don't propogate)
* recursively will create subnodes corresponding to equivalent
* child nodes in the base level
*
* called with DEVFS_LOCK held
***************************************************************/
static int
dev_dup_entry(devnode_t * parent, devdirent_t * back, devdirent_t * *dnm_pp,
    struct devfsmount *dvm)
{
        devdirent_t *   entry_p = NULL;
        devdirent_t *   newback;
        devdirent_t *   newfront;
        int     error;
        devnode_t *     dnp = back->de_dnp;
        int type = dnp->dn_type;

        /*
         * go get the node made (if we need to)
         * use the back one as a prototype
         */
        error = dev_add_entry(back->de_name, parent, type, NULL, dnp,
            parent?parent->dn_dvm:dvm, &entry_p);
        if (!error && (entry_p == NULL)) {
                error = ENOMEM; /* Really can't happen, but make static analyzer happy */
        }
        if (error != 0) {
                printf("duplicating %s failed\n", back->de_name);
                goto out;
        }

        /*
         * If we have just made the root, then insert the pointer to the
         * mount information
         */
        if (dvm) {
                entry_p->de_dnp->dn_dvm = dvm;
        }

        /*
         * If it is a directory, then recurse down all the other
         * subnodes in it....
         * note that this time we don't pass on the mount info..
         */
        if (type == DEV_DIR) {
                for (newback = back->de_dnp->dn_typeinfo.Dir.dirlist;
                    newback; newback = newback->de_next) {
                        if ((error = dev_dup_entry(entry_p->de_dnp,
                            newback, &newfront, NULL)) != 0) {
                                break; /* back out with an error */
                        }
                }
        }
out:
        *dnm_pp = entry_p;
        return error;
}


/***************************************************************
* Free a name node
* remember that if there are other names pointing to the
* dev_node then it may not get freed yet
* can handle if there is no dnp
*
* called with DEVFS_LOCK held
***************************************************************/

int
dev_free_name(devdirent_t * dirent_p)
{
        devnode_t *     parent = dirent_p->de_parent;
        devnode_t *     dnp = dirent_p->de_dnp;

        if (dnp) {
                if (dnp->dn_type == DEV_DIR) {
                        devnode_t * p;

                        if (dnp->dn_typeinfo.Dir.dirlist) {
                                return ENOTEMPTY;
                        }
                        p = dnp->dn_typeinfo.Dir.parent;
                        devfs_dn_free(dnp);     /* account for '.' */
                        devfs_dn_free(p);       /* '..' */
                }
                /*
                 * unlink us from the list of links for this node
                 * If we are the only link, it's easy!
                 * if we are a DIR of course there should not be any
                 * other links.
                 */
                if (dirent_p->de_nextlink == dirent_p) {
                        dnp->dn_linklist = NULL;
                } else {
                        if (dnp->dn_linklist == dirent_p) {
                                dnp->dn_linklist = dirent_p->de_nextlink;
                        }
                }
                devfs_dn_free(dnp);
        }

        dirent_p->de_nextlink->de_prevlinkp = dirent_p->de_prevlinkp;
        *(dirent_p->de_prevlinkp) = dirent_p->de_nextlink;

        /*
         * unlink ourselves from the directory on this plane
         */
        if (parent) { /* if not fs root */
                if ((*dirent_p->de_prevp = dirent_p->de_next)) {/* yes, assign */
                        dirent_p->de_next->de_prevp = dirent_p->de_prevp;
                } else {
                        parent->dn_typeinfo.Dir.dirlast
                                = dirent_p->de_prevp;
                }
                parent->dn_typeinfo.Dir.entrycount--;
                parent->dn_len -= strlen(dirent_p->de_name) + 8;
        }

        DEVFS_DECR_ENTRIES();
        FREE(dirent_p, M_DEVFSNAME);
        return 0;
}


/***************************************************************
* Free a hierarchy starting at a directory node name
* remember that if there are other names pointing to the
* dev_node then it may not get freed yet
* can handle if there is no dnp
* leave the node itself allocated.
*
* called with DEVFS_LOCK held
***************************************************************/

static void
dev_free_hier(devdirent_t * dirent_p)
{
        devnode_t *     dnp = dirent_p->de_dnp;

        if (dnp) {
                if (dnp->dn_type == DEV_DIR) {
                        while (dnp->dn_typeinfo.Dir.dirlist) {
                                dev_free_hier(dnp->dn_typeinfo.Dir.dirlist);
                                dev_free_name(dnp->dn_typeinfo.Dir.dirlist);
                        }
                }
        }
}


/***************************************************************
 * given a dev_node, find the appropriate vnode if one is already
 * associated, or get a new one and associate it with the dev_node
 *
 * called with DEVFS_LOCK held
 *
 * If an error is returned, then the dnp may have been freed (we
 * raced with a delete and lost).  A devnode should not be accessed
 * after devfs_dntovn() fails.
 ****************************************************************/
int
devfs_dntovn(devnode_t * dnp, struct vnode **vn_pp, __unused struct proc * p)
{
        struct vnode *vn_p;
        int error = 0;
        struct vnode_fsparam vfsp;
        enum vtype vtype = 0;
        int markroot = 0;
        int nretries = 0;
        int n_minor = DEVFS_CLONE_ALLOC; /* new minor number for clone device */

        /*
         * We should never come in and find that our devnode has been marked for delete.
         * The lookup should have held the lock from entry until now; it should not have
         * been able to find a removed entry. Any other pathway would have just created
         * the devnode and come here without dropping the devfs lock, so no one would
         * have a chance to delete.
         */
        if (dnp->dn_lflags & DN_DELETE) {
                panic("devfs_dntovn: DN_DELETE set on a devnode upon entry.");
        }

        devfs_ref_node(dnp);

retry:
        *vn_pp = NULL;
        vn_p = dnp->dn_vn;

        if (vn_p) { /* already has a vnode */
                uint32_t vid;

                vid = vnode_vid(vn_p);

                DEVFS_UNLOCK();

                /*
                 * We want to use the drainok variant of vnode_getwithvid
                 * because we _don't_ want to get an iocount if the vnode is
                 * is blocked in vnode_drain as it can cause infinite
                 * loops in vn_open_auth. While in use vnodes are typically
                 * only reclaimed on forced unmounts, In use devfs tty vnodes
                 * can  be quite frequently reclaimed by revoke(2) or by the
                 * exit of a controlling process.
                 */
                error = vnode_getwithvid_drainok(vn_p, vid);

                DEVFS_LOCK();

                if (dnp->dn_lflags & DN_DELETE) {
                        /*
                         * our BUSY node got marked for
                         * deletion while the DEVFS lock
                         * was dropped...
                         */
                        if (error == 0) {
                                /*
                                 * vnode_getwithvid returned a valid ref
                                 * which we need to drop
                                 */
                                vnode_put(vn_p);
                        }

                        /*
                         * This entry is no longer in the namespace.  This is only
                         * possible for lookup: no other path would not find an existing
                         * vnode.  Therefore, ENOENT is a valid result.
                         */
                        error = ENOENT;
                } else if (error == ENODEV) {
                        /*
                         * The Filesystem is getting unmounted.
                         */
                        error = ENOENT;
                } else if (error && (nretries < DEV_MAX_VNODE_RETRY)) {
                        /*
                         * If we got an error from vnode_getwithvid, it means
                         * we raced with a recycle and lost i.e. we asked for
                         * an iocount only after vnode_drain had been entered
                         * for the vnode and returned with an error only after
                         * devfs_reclaim was called on the vnode.  devfs_reclaim
                         * sets dn_vn to NULL but while we were waiting to
                         * reacquire DEVFS_LOCK, another vnode might have gotten
                         * associated with the dnp. In either case, we need to
                         * retry otherwise we will end up returning an ENOENT
                         * for this lookup but the next lookup will  succeed
                         * because it creates a new vnode (or a racing  lookup
                         * created a new vnode already).
                         */
                        error = 0;
                        nretries++;
                        goto retry;
                }
                if (!error) {
                        *vn_pp = vn_p;
                }

                goto out;
        }

        /*
         * If we get here, then we've beaten any deletes;
         * if someone sets DN_DELETE during a subsequent drop
         * of the devfs lock, we'll still vend a vnode.
         */

        if (dnp->dn_lflags & DN_CREATE) {
                dnp->dn_lflags |= DN_CREATEWAIT;
                msleep(&dnp->dn_lflags, &devfs_mutex, PRIBIO, 0, 0);
                goto retry;
        }

        dnp->dn_lflags |= DN_CREATE;

        switch (dnp->dn_type) {
        case    DEV_SLNK:
                vtype = VLNK;
                break;
        case    DEV_DIR:
                if (dnp->dn_typeinfo.Dir.parent == dnp) {
                        markroot = 1;
                }
                vtype = VDIR;
                break;
        case    DEV_BDEV:
        case    DEV_CDEV:
                vtype = (dnp->dn_type == DEV_BDEV) ? VBLK : VCHR;
                break;
#if FDESC
        case    DEV_DEVFD:
                vtype = VDIR;
                break;
#endif /* FDESC */
        }
        vfsp.vnfs_mp = dnp->dn_dvm->mount;
        vfsp.vnfs_vtype = vtype;
        vfsp.vnfs_str = "devfs";
        vfsp.vnfs_dvp = 0;
        vfsp.vnfs_fsnode = dnp;
        vfsp.vnfs_cnp = 0;
        vfsp.vnfs_vops = *(dnp->dn_ops);

        if (vtype == VBLK || vtype == VCHR) {
                /*
                 * Ask the clone minor number function for a new minor number
                 * to use for the next device instance.  If an administative
                 * limit has been reached, this function will return -1.
                 */
                if (dnp->dn_clone != NULL) {
                        int     n_major = major(dnp->dn_typeinfo.dev);

                        n_minor = (*dnp->dn_clone)(dnp->dn_typeinfo.dev, DEVFS_CLONE_ALLOC);
                        if (n_minor == -1) {
                                error = ENOMEM;
                                goto out;
                        }

                        vfsp.vnfs_rdev = makedev(n_major, n_minor);;
                } else {
                        vfsp.vnfs_rdev = dnp->dn_typeinfo.dev;
                }
        } else {
                vfsp.vnfs_rdev = 0;
        }
        vfsp.vnfs_filesize = 0;
        vfsp.vnfs_flags = VNFS_NOCACHE | VNFS_CANTCACHE;
        /* Tag system files */
        vfsp.vnfs_marksystem = 0;
        vfsp.vnfs_markroot = markroot;

        DEVFS_UNLOCK();

        error = vnode_create(VNCREATE_FLAVOR, VCREATESIZE, &vfsp, &vn_p);

        /* Do this before grabbing the lock */
        if (error == 0) {
                vnode_setneedinactive(vn_p);
        }

        DEVFS_LOCK();

        if (error == 0) {
                vnode_settag(vn_p, VT_DEVFS);

                if ((dnp->dn_clone != NULL) && (dnp->dn_vn != NULLVP)) {
                        panic("devfs_dntovn: cloning device with a vnode?\n");
                }

                *vn_pp = vn_p;

                /*
                 * Another vnode that has this devnode as its v_data.
                 * This reference, unlike the one taken at the start
                 * of the function, persists until a VNOP_RECLAIM
                 * comes through for this vnode.
                 */
                devfs_ref_node(dnp);

                /*
                 * A cloned vnode is not hooked into the devnode; every lookup
                 * gets a new vnode.
                 */
                if (dnp->dn_clone == NULL) {
                        dnp->dn_vn = vn_p;
                }
        } else if (n_minor != DEVFS_CLONE_ALLOC) {
                /*
                 * If we failed the create, we need to release the cloned minor
                 * back to the free list.  In general, this is only useful if
                 * the clone function results in a state change in the cloned
                 * device for which the minor number was obtained.  If we get
                 * past this point withouth falling into this case, it's
                 * assumed that any state to be released will be released when
                 * the vnode is dropped, instead.
                 */
                (void)(*dnp->dn_clone)(dnp->dn_typeinfo.dev, DEVFS_CLONE_FREE);
        }

        dnp->dn_lflags &= ~DN_CREATE;
        if (dnp->dn_lflags & DN_CREATEWAIT) {
                dnp->dn_lflags &= ~DN_CREATEWAIT;
                wakeup(&dnp->dn_lflags);
        }

out:
        /*
         * Release the reference we took to prevent deletion while we weren't holding the lock.
         * If not returning success, then dropping this reference could delete the devnode;
         * no one should access a devnode after a call to devfs_dntovn fails.
         */
        devfs_rele_node(dnp);

        return error;
}

/*
 * Increment refcount on a devnode; prevents free of the node
 * while the devfs lock is not held.
 */
void
devfs_ref_node(devnode_t *dnp)
{
        os_ref_retain_locked_raw(&dnp->dn_refcount, &devfs_refgrp);
}

/*
 * Release a reference on a devnode.  If the devnode is marked for
 * free and the refcount is dropped to one, do the free.
 */
void
devfs_rele_node(devnode_t *dnp)
{
        os_ref_count_t rc = os_ref_release_locked_raw(&dnp->dn_refcount, &devfs_refgrp);
        if (rc < 1) {
                panic("devfs_rele_node: devnode without a refcount!\n");
        } else if ((rc == 1) && (dnp->dn_lflags & DN_DELETE)) {
                /* release final reference from dev_add_node */
                (void) os_ref_release_locked_raw(&dnp->dn_refcount, &devfs_refgrp);
                devnode_free(dnp);
        }
}

/***********************************************************************
* add a whole device, with no prototype.. make name element and node
* Used for adding the original device entries
*
* called with DEVFS_LOCK held
***********************************************************************/
int
dev_add_entry(const char *name, devnode_t * parent, int type, devnode_type_t * typeinfo,
    devnode_t * proto, struct devfsmount *dvm, devdirent_t * *nm_pp)
{
        devnode_t *     dnp;
        int     error = 0;

        if ((error = dev_add_node(type, typeinfo, proto, &dnp,
            (parent?parent->dn_dvm:dvm))) != 0) {
                printf("devfs: %s: base node allocation failed (Errno=%d)\n",
                    name, error);
                return error;
        }
        if ((error = dev_add_name(name, parent, NULL, dnp, nm_pp)) != 0) {
                devfs_dn_free(dnp); /* 1->0 for dir, 0->(-1) for other */
                printf("devfs: %s: name slot allocation failed (Errno=%d)\n",
                    name, error);
        }
        return error;
}

static void
devfs_bulk_notify(devfs_event_log_t delp)
{
        uint32_t i;
        for (i = 0; i < delp->del_used; i++) {
                devfs_vnode_event_t dvep = &delp->del_entries[i];
                if (vnode_getwithvid(dvep->dve_vp, dvep->dve_vid) == 0) {
                        vnode_notify(dvep->dve_vp, dvep->dve_events, NULL);
                        vnode_put(dvep->dve_vp);
                }
        }
}

static void
devfs_record_event(devfs_event_log_t delp, devnode_t *dnp, uint32_t events)
{
        if (delp->del_used >= delp->del_max) {
                panic("devfs event log overflowed.\n");
        }

        /* Can only notify for nodes that have an associated vnode */
        if (dnp->dn_vn != NULLVP && vnode_ismonitored(dnp->dn_vn)) {
                devfs_vnode_event_t dvep = &delp->del_entries[delp->del_used];
                dvep->dve_vp = dnp->dn_vn;
                dvep->dve_vid = vnode_vid(dnp->dn_vn);
                dvep->dve_events = events;
                delp->del_used++;
        }
}

static int
devfs_init_event_log(devfs_event_log_t delp, uint32_t count, devfs_vnode_event_t buf)
{
        devfs_vnode_event_t dvearr;

        if (buf == NULL) {
                MALLOC(dvearr, devfs_vnode_event_t, count * sizeof(struct devfs_vnode_event), M_TEMP, M_WAITOK | M_ZERO);
                if (dvearr == NULL) {
                        return ENOMEM;
                }
        } else {
                dvearr = buf;
        }

        delp->del_max = count;
        delp->del_used = 0;
        delp->del_entries = dvearr;
        return 0;
}

static void
devfs_release_event_log(devfs_event_log_t delp, int need_free)
{
        if (delp->del_entries == NULL) {
                panic("Free of devfs notify info that has not been intialized.\n");
        }

        if (need_free) {
                FREE(delp->del_entries, M_TEMP);
        }

        delp->del_entries = NULL;
}

/*
 * Function: devfs_make_node
 *
 * Purpose
 *   Create a device node with the given pathname in the devfs namespace.
 *
 * Parameters:
 *   dev        - the dev_t value to associate
 *   chrblk     - block or character device (DEVFS_CHAR or DEVFS_BLOCK)
 *   uid, gid   - ownership
 *   perms      - permissions
 *   clone      - minor number cloning function
 *   fmt, ...   - path format string with printf args to format the path name
 * Returns:
 *   A handle to a device node if successful, NULL otherwise.
 */
void *
devfs_make_node_clone(dev_t dev, int chrblk, uid_t uid,
    gid_t gid, int perms, int (*clone)(dev_t dev, int action),
    const char *fmt, ...)
{
        devdirent_t *   new_dev = NULL;
        devfstype_t     type;
        va_list ap;

        switch (chrblk) {
        case DEVFS_CHAR:
                type = DEV_CDEV;
                break;
        case DEVFS_BLOCK:
                type = DEV_BDEV;
                break;
        default:
                goto out;
        }

        va_start(ap, fmt);
        new_dev = devfs_make_node_internal(dev, type, uid, gid, perms, clone, fmt, ap);
        va_end(ap);
out:
        return new_dev;
}


/*
 * Function: devfs_make_node
 *
 * Purpose
 *   Create a device node with the given pathname in the devfs namespace.
 *
 * Parameters:
 *   dev        - the dev_t value to associate
 *   chrblk     - block or character device (DEVFS_CHAR or DEVFS_BLOCK)
 *   uid, gid   - ownership
 *   perms      - permissions
 *   fmt, ...   - path format string with printf args to format the path name
 * Returns:
 *   A handle to a device node if successful, NULL otherwise.
 */
void *
devfs_make_node(dev_t dev, int chrblk, uid_t uid,
    gid_t gid, int perms, const char *fmt, ...)
{
        devdirent_t *   new_dev = NULL;
        devfstype_t type;
        va_list ap;

        if (chrblk != DEVFS_CHAR && chrblk != DEVFS_BLOCK) {
                goto out;
        }

        type = (chrblk == DEVFS_BLOCK ? DEV_BDEV : DEV_CDEV);

        va_start(ap, fmt);
        new_dev = devfs_make_node_internal(dev, type, uid, gid, perms, NULL, fmt, ap);
        va_end(ap);

out:
        return new_dev;
}

static devdirent_t *
devfs_make_node_internal(dev_t dev, devfstype_t type, uid_t uid,
    gid_t gid, int perms, int (*clone)(dev_t dev, int action), const char *fmt, va_list ap)
{
        devdirent_t *   new_dev = NULL;
        devnode_t * dnp;
        devnode_type_t  typeinfo;

        char            *name, buf[256]; /* XXX */
        const char      *path;
#if CONFIG_MACF
        char buff[sizeof(buf)];
#endif
        int             i;
        uint32_t        log_count;
        struct devfs_event_log event_log;
        struct devfs_vnode_event stackbuf[NUM_STACK_ENTRIES];
        int             need_free = 0;

        vsnprintf(buf, sizeof(buf), fmt, ap);

#if CONFIG_MACF
        bcopy(buf, buff, sizeof(buff));
        buff[sizeof(buff) - 1] = 0;
#endif
        name = NULL;

        for (i = strlen(buf); i > 0; i--) {
                if (buf[i] == '/') {
                        name = &buf[i];
                        buf[i] = 0;
                        break;
                }
        }

        if (name) {
                *name++ = '\0';
                path = buf;
        } else {
                name = buf;
                path = "/";
        }

        log_count = devfs_nmountplanes;
        if (log_count > NUM_STACK_ENTRIES) {
wrongsize:
                need_free = 1;
                if (devfs_init_event_log(&event_log, log_count, NULL) != 0) {
                        return NULL;
                }
        } else {
                need_free = 0;
                log_count = NUM_STACK_ENTRIES;
                if (devfs_init_event_log(&event_log, log_count, &stackbuf[0]) != 0) {
                        return NULL;
                }
        }

        DEVFS_LOCK();
        if (log_count < devfs_nmountplanes) {
                DEVFS_UNLOCK();
                devfs_release_event_log(&event_log, need_free);
                log_count = log_count * 2;
                goto wrongsize;
        }

        if (!devfs_ready) {
                printf("devfs_make_node: not ready for devices!\n");
                goto out;
        }

        /* find/create directory path ie. mkdir -p */
        if (dev_finddir(path, NULL, DEVFS_CREATE, &dnp, &event_log) == 0) {
                typeinfo.dev = dev;
                if (dev_add_entry(name, dnp, type, &typeinfo, NULL, NULL, &new_dev) == 0) {
                        new_dev->de_dnp->dn_gid = gid;
                        new_dev->de_dnp->dn_uid = uid;
                        new_dev->de_dnp->dn_mode |= perms;
                        new_dev->de_dnp->dn_clone = clone;
#if CONFIG_MACF
                        mac_devfs_label_associate_device(dev, new_dev->de_dnp, buff);
#endif
                        devfs_propogate(dnp->dn_typeinfo.Dir.myname, new_dev, &event_log);
                }
        }

out:
        DEVFS_UNLOCK();

        devfs_bulk_notify(&event_log);
        devfs_release_event_log(&event_log, need_free);
        return new_dev;
}

/*
 * Function: devfs_make_link
 *
 * Purpose:
 *   Create a link to a previously created device node.
 *
 * Returns:
 *   0 if successful, -1 if failed
 */
int
devfs_make_link(void *original, char *fmt, ...)
{
        devdirent_t *   new_dev = NULL;
        devdirent_t *   orig = (devdirent_t *) original;
        devnode_t *     dirnode;        /* devnode for parent directory */
        struct devfs_event_log event_log;
        uint32_t        log_count;

        va_list ap;
        char *p, buf[256]; /* XXX */
        int i;

        DEVFS_LOCK();

        if (!devfs_ready) {
                DEVFS_UNLOCK();
                printf("devfs_make_link: not ready for devices!\n");
                return -1;
        }
        DEVFS_UNLOCK();

        va_start(ap, fmt);
        vsnprintf(buf, sizeof(buf), fmt, ap);
        va_end(ap);

        p = NULL;

        for (i = strlen(buf); i > 0; i--) {
                if (buf[i] == '/') {
                        p = &buf[i];
                        buf[i] = 0;
                        break;
                }
        }

        /*
         * One slot for each directory, one for each devnode
         * whose link count changes
         */
        log_count = devfs_nmountplanes * 2;
wrongsize:
        if (devfs_init_event_log(&event_log, log_count, NULL) != 0) {
                /* No lock held, no allocations done, can just return */
                return -1;
        }

        DEVFS_LOCK();

        if (log_count < devfs_nmountplanes) {
                DEVFS_UNLOCK();
                devfs_release_event_log(&event_log, 1);
                log_count = log_count * 2;
                goto wrongsize;
        }

        if (p) {
                *p++ = '\0';

                if (dev_finddir(buf, NULL, DEVFS_CREATE, &dirnode, &event_log)
                    || dev_add_name(p, dirnode, NULL, orig->de_dnp, &new_dev)) {
                        goto fail;
                }
        } else {
                if (dev_finddir("", NULL, DEVFS_CREATE, &dirnode, &event_log)
                    || dev_add_name(buf, dirnode, NULL, orig->de_dnp, &new_dev)) {
                        goto fail;
                }
        }
        devfs_propogate(dirnode->dn_typeinfo.Dir.myname, new_dev, &event_log);
fail:
        DEVFS_UNLOCK();
        devfs_bulk_notify(&event_log);
        devfs_release_event_log(&event_log, 1);

        return (new_dev != NULL) ? 0 : -1;
}