Aren/NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
35ba1291c8
NetBSD/sys/kern/vfs_subr.c

2278 lines
53 KiB
C
Raw Blame History

/* $NetBSD: vfs_subr.c,v 1.107 1999/08/14 06:23:59 ross Exp $ */
/*-
* Copyright (c) 1997, 1998 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
* NASA Ames Research Center.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the NetBSD
* Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation 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 NETBSD FOUNDATION, INC. 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 FOUNDATION 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) 1989, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)vfs_subr.c 8.13 (Berkeley) 4/18/94
*/
/*
* External virtual filesystem routines
*/
#include "opt_compat_netbsd.h"
#include "opt_compat_43.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/mount.h>
#include <sys/time.h>
#include <sys/fcntl.h>
#include <sys/vnode.h>
#include <sys/stat.h>
#include <sys/namei.h>
#include <sys/ucred.h>
#include <sys/buf.h>
#include <sys/errno.h>
#include <sys/malloc.h>
#include <sys/domain.h>
#include <sys/mbuf.h>
#include <sys/syscallargs.h>
#include <sys/device.h>
#include <sys/dirent.h>
#include <vm/vm.h>
#include <sys/sysctl.h>
#include <miscfs/specfs/specdev.h>
#include <uvm/uvm_extern.h>
enum vtype iftovt_tab[16] = {
VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON,
VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD,
};
int vttoif_tab[9] = {
0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK,
S_IFSOCK, S_IFIFO, S_IFMT,
};
int doforce = 1; /* 1 => permit forcible unmounting */
int prtactive = 0; /* 1 => print out reclaim of active vnodes */
/*
* Insq/Remq for the vnode usage lists.
*/
#define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs)
#define bufremvn(bp) { \
LIST_REMOVE(bp, b_vnbufs); \
(bp)->b_vnbufs.le_next = NOLIST; \
}
TAILQ_HEAD(freelst, vnode) vnode_free_list = /* vnode free list */
TAILQ_HEAD_INITIALIZER(vnode_free_list);
struct mntlist mountlist = /* mounted filesystem list */
CIRCLEQ_HEAD_INITIALIZER(mountlist);
struct vfs_list_head vfs_list = /* vfs list */
LIST_HEAD_INITIALIZER(vfs_list);
struct nfs_public nfs_pub; /* publicly exported FS */
struct simplelock mountlist_slock;
static struct simplelock mntid_slock;
struct simplelock mntvnode_slock;
struct simplelock vnode_free_list_slock;
struct simplelock spechash_slock;
/*
* These define the root filesystem and device.
*/
struct mount *rootfs;
struct vnode *rootvnode;
struct device *root_device; /* root device */
struct pool vnode_pool; /* memory pool for vnodes */
/*
* Local declarations.
*/
void insmntque __P((struct vnode *, struct mount *));
int getdevvp __P((dev_t, struct vnode **, enum vtype));
void vgoneall __P((struct vnode *));
static int vfs_hang_addrlist __P((struct mount *, struct netexport *,
struct export_args *));
static int vfs_free_netcred __P((struct radix_node *, void *));
static void vfs_free_addrlist __P((struct netexport *));
#ifdef DEBUG
void printlockedvnodes __P((void));
#endif
/*
* Initialize the vnode management data structures.
*/
void
vntblinit()
{
simple_lock_init(&mntvnode_slock);
simple_lock_init(&mntid_slock);
simple_lock_init(&spechash_slock);
simple_lock_init(&vnode_free_list_slock);
pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodepl",
0, pool_page_alloc_nointr, pool_page_free_nointr, M_VNODE);
}
/*
* Mark a mount point as busy. Used to synchronize access and to delay
* unmounting. Interlock is not released on failure.
*/
int
vfs_busy(mp, flags, interlkp)
struct mount *mp;
int flags;
struct simplelock *interlkp;
{
int lkflags;
while (mp->mnt_flag & MNT_UNMOUNT) {
int gone;
if (flags & LK_NOWAIT)
return (ENOENT);
if (interlkp)
simple_unlock(interlkp);
/*
* Since all busy locks are shared except the exclusive
* lock granted when unmounting, the only place that a
* wakeup needs to be done is at the release of the
* exclusive lock at the end of dounmount.
*
* XXX MP: add spinlock protecting mnt_wcnt here once you
* can atomically unlock-and-sleep.
*/
mp->mnt_wcnt++;
sleep((caddr_t)mp, PVFS);
mp->mnt_wcnt--;
gone = mp->mnt_flag & MNT_GONE;
if (mp->mnt_wcnt == 0)
wakeup(&mp->mnt_wcnt);
if (interlkp)
simple_lock(interlkp);
if (gone)
return (ENOENT);
}
lkflags = LK_SHARED;
if (interlkp)
lkflags |= LK_INTERLOCK;
if (lockmgr(&mp->mnt_lock, lkflags, interlkp))
panic("vfs_busy: unexpected lock failure");
return (0);
}
/*
* Free a busy filesystem.
*/
void
vfs_unbusy(mp)
struct mount *mp;
{
lockmgr(&mp->mnt_lock, LK_RELEASE, NULL);
}
/*
* Lookup a filesystem type, and if found allocate and initialize
* a mount structure for it.
*
* Devname is usually updated by mount(8) after booting.
*/
int
vfs_rootmountalloc(fstypename, devname, mpp)
char *fstypename;
char *devname;
struct mount **mpp;
{
struct vfsops *vfsp = NULL;
struct mount *mp;
for (vfsp = LIST_FIRST(&vfs_list); vfsp != NULL;
vfsp = LIST_NEXT(vfsp, vfs_list))
if (!strncmp(vfsp->vfs_name, fstypename, MFSNAMELEN))
break;
if (vfsp == NULL)
return (ENODEV);
mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
memset((char *)mp, 0, (u_long)sizeof(struct mount));
lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
(void)vfs_busy(mp, LK_NOWAIT, 0);
LIST_INIT(&mp->mnt_vnodelist);
mp->mnt_op = vfsp;
mp->mnt_flag = MNT_RDONLY;
mp->mnt_vnodecovered = NULLVP;
vfsp->vfs_refcount++;
strncpy(mp->mnt_stat.f_fstypename, vfsp->vfs_name, MFSNAMELEN);
mp->mnt_stat.f_mntonname[0] = '/';
(void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0);
*mpp = mp;
return (0);
}
/*
* Lookup a mount point by filesystem identifier.
*/
struct mount *
vfs_getvfs(fsid)
fsid_t *fsid;
{
register struct mount *mp;
simple_lock(&mountlist_slock);
for (mp = mountlist.cqh_first; mp != (void *)&mountlist;
mp = mp->mnt_list.cqe_next) {
if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] &&
mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) {
simple_unlock(&mountlist_slock);
return (mp);
}
}
simple_unlock(&mountlist_slock);
return ((struct mount *)0);
}
/*
* Get a new unique fsid
*/
void
vfs_getnewfsid(mp, fstypename)
struct mount *mp;
char *fstypename;
{
static u_short xxxfs_mntid;
fsid_t tfsid;
int mtype;
simple_lock(&mntid_slock);
mtype = makefstype(fstypename);
mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0);
mp->mnt_stat.f_fsid.val[1] = mtype;
if (xxxfs_mntid == 0)
++xxxfs_mntid;
tfsid.val[0] = makedev((nblkdev + mtype) & 0xff, xxxfs_mntid);
tfsid.val[1] = mtype;
if (mountlist.cqh_first != (void *)&mountlist) {
while (vfs_getvfs(&tfsid)) {
tfsid.val[0]++;
xxxfs_mntid++;
}
}
mp->mnt_stat.f_fsid.val[0] = tfsid.val[0];
simple_unlock(&mntid_slock);
}
/*
* Make a 'unique' number from a mount type name.
*/
long
makefstype(type)
char *type;
{
long rv;
for (rv = 0; *type; type++) {
rv <<= 2;
rv ^= *type;
}
return rv;
}
/*
* Set vnode attributes to VNOVAL
*/
void
vattr_null(vap)
register struct vattr *vap;
{
vap->va_type = VNON;
/*
* Assign individually so that it is safe even if size and
* sign of each member are varied.
*/
vap->va_mode = VNOVAL;
vap->va_nlink = VNOVAL;
vap->va_uid = VNOVAL;
vap->va_gid = VNOVAL;
vap->va_fsid = VNOVAL;
vap->va_fileid = VNOVAL;
vap->va_size = VNOVAL;
vap->va_blocksize = VNOVAL;
vap->va_atime.tv_sec =
vap->va_mtime.tv_sec =
vap->va_ctime.tv_sec = VNOVAL;
vap->va_atime.tv_nsec =
vap->va_mtime.tv_nsec =
vap->va_ctime.tv_nsec = VNOVAL;
vap->va_gen = VNOVAL;
vap->va_flags = VNOVAL;
vap->va_rdev = VNOVAL;
vap->va_bytes = VNOVAL;
vap->va_vaflags = 0;
}
/*
* Routines having to do with the management of the vnode table.
*/
extern int (**dead_vnodeop_p) __P((void *));
long numvnodes;
/*
* Return the next vnode from the free list.
*/
int
getnewvnode(tag, mp, vops, vpp)
enum vtagtype tag;
struct mount *mp;
int (**vops) __P((void *));
struct vnode **vpp;
{
struct proc *p = curproc; /* XXX */
struct vnode *vp;
int error;
#ifdef DIAGNOSTIC
int s;
#endif
if (mp) {
/*
* Mark filesystem busy while we're creating a vnode.
* If unmount is in progress, this will wait; if the
* unmount succeeds (only if umount -f), this will
* return an error. If the unmount fails, we'll keep
* going afterwards.
* (This puts the per-mount vnode list logically under
* the protection of the vfs_busy lock).
*/
error = vfs_busy(mp, 0, 0);
if (error)
return error;
}
simple_lock(&vnode_free_list_slock);
if ((vnode_free_list.tqh_first == NULL &&
numvnodes < 2 * desiredvnodes) ||
numvnodes < desiredvnodes) {
simple_unlock(&vnode_free_list_slock);
vp = pool_get(&vnode_pool, PR_WAITOK);
memset((char *)vp, 0, sizeof(*vp));
simple_lock_init(&vp->v_interlock);
numvnodes++;
} else {
for (vp = vnode_free_list.tqh_first;
vp != NULLVP; vp = vp->v_freelist.tqe_next) {
if (simple_lock_try(&vp->v_interlock)) {
if ((vp->v_flag & VLAYER) == 0) {
break;
}
if (VOP_ISLOCKED(vp) == 0)
break;
else
simple_unlock(&vp->v_interlock);
}
}
/*
* Unless this is a bad time of the month, at most
* the first NCPUS items on the free list are
* locked, so this is close enough to being empty.
*/
if (vp == NULLVP) {
simple_unlock(&vnode_free_list_slock);
if (mp) vfs_unbusy(mp);
tablefull("vnode");
*vpp = 0;
return (ENFILE);
}
if (vp->v_usecount)
panic("free vnode isn't");
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
/* see comment on why 0xdeadb is set at end of vgone (below) */
vp->v_freelist.tqe_prev = (struct vnode **)0xdeadb;
simple_unlock(&vnode_free_list_slock);
vp->v_lease = NULL;
if (vp->v_type != VBAD)
vgonel(vp, p);
else
simple_unlock(&vp->v_interlock);
#ifdef DIAGNOSTIC
if (vp->v_data)
panic("cleaned vnode isn't");
s = splbio();
if (vp->v_numoutput)
panic("Clean vnode has pending I/O's");
splx(s);
#endif
vp->v_flag = 0;
vp->v_lastr = 0;
vp->v_ralen = 0;
vp->v_maxra = 0;
vp->v_lastw = 0;
vp->v_lasta = 0;
vp->v_cstart = 0;
vp->v_clen = 0;
vp->v_socket = 0;
}
vp->v_type = VNON;
vp->v_vnlock = &vp->v_lock;
lockinit(vp->v_vnlock, PVFS, "vnlock", 0, 0);
cache_purge(vp);
vp->v_tag = tag;
vp->v_op = vops;
insmntque(vp, mp);
*vpp = vp;
vp->v_usecount = 1;
vp->v_data = 0;
simple_lock_init(&vp->v_uvm.u_obj.vmobjlock);
if (mp) vfs_unbusy(mp);
return (0);
}
/*
* Move a vnode from one mount queue to another.
*/
void
insmntque(vp, mp)
register struct vnode *vp;
register struct mount *mp;
{
#ifdef DIAGNOSTIC
if ((mp != NULL) &&
(mp->mnt_flag & MNT_UNMOUNT)) {
panic("insmntque into dying filesystem");
}
#endif
simple_lock(&mntvnode_slock);
/*
* Delete from old mount point vnode list, if on one.
*/
if (vp->v_mount != NULL)
LIST_REMOVE(vp, v_mntvnodes);
/*
* Insert into list of vnodes for the new mount point, if available.
*/
if ((vp->v_mount = mp) != NULL)
LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes);
simple_unlock(&mntvnode_slock);
}
/*
* Update outstanding I/O count and do wakeup if requested.
*/
void
vwakeup(bp)
register struct buf *bp;
{
register struct vnode *vp;
bp->b_flags &= ~B_WRITEINPROG;
if ((vp = bp->b_vp) != NULL) {
if (--vp->v_numoutput < 0)
panic("vwakeup: neg numoutput");
if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) {
vp->v_flag &= ~VBWAIT;
wakeup((caddr_t)&vp->v_numoutput);
}
}
}
/*
* Flush out and invalidate all buffers associated with a vnode.
* Called with the underlying object locked.
*/
int
vinvalbuf(vp, flags, cred, p, slpflag, slptimeo)
register struct vnode *vp;
int flags;
struct ucred *cred;
struct proc *p;
int slpflag, slptimeo;
{
register struct buf *bp;
struct buf *nbp, *blist;
int s, error;
if (flags & V_SAVE) {
error = VOP_FSYNC(vp, cred, FSYNC_WAIT|FSYNC_RECLAIM, p);
if (error != 0)
return (error);
if (vp->v_dirtyblkhd.lh_first != NULL)
panic("vinvalbuf: dirty bufs");
}
for (;;) {
if ((blist = vp->v_cleanblkhd.lh_first) && flags & V_SAVEMETA)
while (blist && blist->b_lblkno < 0)
blist = blist->b_vnbufs.le_next;
if (!blist && (blist = vp->v_dirtyblkhd.lh_first) &&
(flags & V_SAVEMETA))
while (blist && blist->b_lblkno < 0)
blist = blist->b_vnbufs.le_next;
if (!blist)
break;
for (bp = blist; bp; bp = nbp) {
nbp = bp->b_vnbufs.le_next;
if (flags & V_SAVEMETA && bp->b_lblkno < 0)
continue;
s = splbio();
if (bp->b_flags & B_BUSY) {
bp->b_flags |= B_WANTED;
error = tsleep((caddr_t)bp,
slpflag | (PRIBIO + 1), "vinvalbuf",
slptimeo);
splx(s);
if (error)
return (error);
break;
}
bp->b_flags |= B_BUSY | B_VFLUSH;
splx(s);
/*
* XXX Since there are no node locks for NFS, I believe
* there is a slight chance that a delayed write will
* occur while sleeping just above, so check for it.
*/
if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) {
(void) VOP_BWRITE(bp);
break;
}
bp->b_flags |= B_INVAL;
brelse(bp);
}
}
if (!(flags & V_SAVEMETA) &&
(vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first))
panic("vinvalbuf: flush failed");
return (0);
}
void
vflushbuf(vp, sync)
register struct vnode *vp;
int sync;
{
register struct buf *bp, *nbp;
int s;
loop:
s = splbio();
for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) {
nbp = bp->b_vnbufs.le_next;
if ((bp->b_flags & B_BUSY))
continue;
if ((bp->b_flags & B_DELWRI) == 0)
panic("vflushbuf: not dirty");
bp->b_flags |= B_BUSY | B_VFLUSH;
splx(s);
/*
* Wait for I/O associated with indirect blocks to complete,
* since there is no way to quickly wait for them below.
*/
if (bp->b_vp == vp || sync == 0)
(void) bawrite(bp);
else
(void) bwrite(bp);
goto loop;
}
if (sync == 0) {
splx(s);
return;
}
while (vp->v_numoutput) {
vp->v_flag |= VBWAIT;
tsleep((caddr_t)&vp->v_numoutput, PRIBIO + 1, "vflushbuf", 0);
}
splx(s);
if (vp->v_dirtyblkhd.lh_first != NULL) {
vprint("vflushbuf: dirty", vp);
goto loop;
}
}
/*
* Associate a buffer with a vnode.
*/
void
bgetvp(vp, bp)
register struct vnode *vp;
register struct buf *bp;
{
if (bp->b_vp)
panic("bgetvp: not free");
VHOLD(vp);
bp->b_vp = vp;
if (vp->v_type == VBLK || vp->v_type == VCHR)
bp->b_dev = vp->v_rdev;
else
bp->b_dev = NODEV;
/*
* Insert onto list for new vnode.
*/
bufinsvn(bp, &vp->v_cleanblkhd);
}
/*
* Disassociate a buffer from a vnode.
*/
void
brelvp(bp)
register struct buf *bp;
{
struct vnode *vp;
if (bp->b_vp == (struct vnode *) 0)
panic("brelvp: NULL");
/*
* Delete from old vnode list, if on one.
*/
if (bp->b_vnbufs.le_next != NOLIST)
bufremvn(bp);
vp = bp->b_vp;
bp->b_vp = (struct vnode *) 0;
HOLDRELE(vp);
}
/*
* Reassign a buffer from one vnode to another.
* Used to assign file specific control information
* (indirect blocks) to the vnode to which they belong.
*/
void
reassignbuf(bp, newvp)
register struct buf *bp;
register struct vnode *newvp;
{
register struct buflists *listheadp;
if (newvp == NULL) {
printf("reassignbuf: NULL");
return;
}
/*
* Delete from old vnode list, if on one.
*/
if (bp->b_vnbufs.le_next != NOLIST)
bufremvn(bp);
/*
* If dirty, put on list of dirty buffers;
* otherwise insert onto list of clean buffers.
*/
if (bp->b_flags & B_DELWRI)
listheadp = &newvp->v_dirtyblkhd;
else
listheadp = &newvp->v_cleanblkhd;
bufinsvn(bp, listheadp);
}
/*
* Create a vnode for a block device.
* Used for root filesystem and swap areas.
* Also used for memory file system special devices.
*/
int
bdevvp(dev, vpp)
dev_t dev;
struct vnode **vpp;
{
return (getdevvp(dev, vpp, VBLK));
}
/*
* Create a vnode for a character device.
* Used for kernfs and some console handling.
*/
int
cdevvp(dev, vpp)
dev_t dev;
struct vnode **vpp;
{
return (getdevvp(dev, vpp, VCHR));
}
/*
* Create a vnode for a device.
* Used by bdevvp (block device) for root file system etc.,
* and by cdevvp (character device) for console and kernfs.
*/
int
getdevvp(dev, vpp, type)
dev_t dev;
struct vnode **vpp;
enum vtype type;
{
register struct vnode *vp;
struct vnode *nvp;
int error;
if (dev == NODEV) {
*vpp = NULLVP;
return (0);
}
error = getnewvnode(VT_NON, NULL, spec_vnodeop_p, &nvp);
if (error) {
*vpp = NULLVP;
return (error);
}
vp = nvp;
vp->v_type = type;
if ((nvp = checkalias(vp, dev, NULL)) != 0) {
vput(vp);
vp = nvp;
}
*vpp = vp;
return (0);
}
/*
* Check to see if the new vnode represents a special device
* for which we already have a vnode (either because of
* bdevvp() or because of a different vnode representing
* the same block device). If such an alias exists, deallocate
* the existing contents and return the aliased vnode. The
* caller is responsible for filling it with its new contents.
*/
struct vnode *
checkalias(nvp, nvp_rdev, mp)
register struct vnode *nvp;
dev_t nvp_rdev;
struct mount *mp;
{
struct proc *p = curproc; /* XXX */
register struct vnode *vp;
struct vnode **vpp;
if (nvp->v_type != VBLK && nvp->v_type != VCHR)
return (NULLVP);
vpp = &speclisth[SPECHASH(nvp_rdev)];
loop:
simple_lock(&spechash_slock);
for (vp = *vpp; vp; vp = vp->v_specnext) {
if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type)
continue;
/*
* Alias, but not in use, so flush it out.
*/
simple_lock(&vp->v_interlock);
if (vp->v_usecount == 0) {
simple_unlock(&spechash_slock);
vgonel(vp, p);
goto loop;
}
if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK)) {
simple_unlock(&spechash_slock);
goto loop;
}
break;
}
if (vp == NULL || vp->v_tag != VT_NON || vp->v_type != VBLK) {
MALLOC(nvp->v_specinfo, struct specinfo *,
sizeof(struct specinfo), M_VNODE, M_WAITOK);
nvp->v_rdev = nvp_rdev;
nvp->v_hashchain = vpp;
nvp->v_specnext = *vpp;
nvp->v_specflags = 0;
simple_unlock(&spechash_slock);
nvp->v_speclockf = NULL;
*vpp = nvp;
if (vp != NULLVP) {
nvp->v_flag |= VALIASED;
vp->v_flag |= VALIASED;
vput(vp);
}
return (NULLVP);
}
simple_unlock(&spechash_slock);
VOP_UNLOCK(vp, 0);
simple_lock(&vp->v_interlock);
vclean(vp, 0, p);
vp->v_op = nvp->v_op;
vp->v_tag = nvp->v_tag;
vp->v_vnlock = &vp->v_lock;
lockinit(vp->v_vnlock, PVFS, "vnlock", 0, 0);
nvp->v_type = VNON;
insmntque(vp, mp);
return (vp);
}
/*
* Grab a particular vnode from the free list, increment its
* reference count and lock it. If the vnode lock bit is set the
* vnode is being eliminated in vgone. In that case, we can not
* grab the vnode, so the process is awakened when the transition is
* completed, and an error returned to indicate that the vnode is no
* longer usable (possibly having been changed to a new file system type).
*/
int
vget(vp, flags)
struct vnode *vp;
int flags;
{
int error;
/*
* If the vnode is in the process of being cleaned out for
* another use, we wait for the cleaning to finish and then
* return failure. Cleaning is determined by checking that
* the VXLOCK flag is set.
*/
if ((flags & LK_INTERLOCK) == 0)
simple_lock(&vp->v_interlock);
if (vp->v_flag & VXLOCK) {
vp->v_flag |= VXWANT;
simple_unlock(&vp->v_interlock);
tsleep((caddr_t)vp, PINOD, "vget", 0);
return (ENOENT);
}
if (vp->v_usecount == 0) {
simple_lock(&vnode_free_list_slock);
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
simple_unlock(&vnode_free_list_slock);
}
vp->v_usecount++;
if (flags & LK_TYPE_MASK) {
if ((error = vn_lock(vp, flags | LK_INTERLOCK)))
vrele(vp);
return (error);
}
simple_unlock(&vp->v_interlock);
return (0);
}
/*
* vput(), just unlock and vrele()
*/
void
vput(vp)
struct vnode *vp;
{
struct proc *p = curproc; /* XXX */
#ifdef DIGANOSTIC
if (vp == NULL)
panic("vput: null vp");
#endif
simple_lock(&vp->v_interlock);
vp->v_usecount--;
if (vp->v_usecount > 0) {
simple_unlock(&vp->v_interlock);
VOP_UNLOCK(vp, 0);
return;
}
#ifdef DIAGNOSTIC
if (vp->v_usecount < 0 || vp->v_writecount != 0) {
vprint("vput: bad ref count", vp);
panic("vput: ref cnt");
}
#endif
/*
* Insert at tail of LRU list.
*/
simple_lock(&vnode_free_list_slock);
TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
simple_unlock(&vnode_free_list_slock);
simple_unlock(&vp->v_interlock);
VOP_INACTIVE(vp, p);
}
/*
* Vnode release.
* If count drops to zero, call inactive routine and return to freelist.
*/
void
vrele(vp)
struct vnode *vp;
{
struct proc *p = curproc; /* XXX */
#ifdef DIAGNOSTIC
if (vp == NULL)
panic("vrele: null vp");
#endif
simple_lock(&vp->v_interlock);
vp->v_usecount--;
if (vp->v_usecount > 0) {
simple_unlock(&vp->v_interlock);
return;
}
#ifdef DIAGNOSTIC
if (vp->v_usecount < 0 || vp->v_writecount != 0) {
vprint("vrele: bad ref count", vp);
panic("vrele: ref cnt");
}
#endif
/*
* Insert at tail of LRU list.
*/
simple_lock(&vnode_free_list_slock);
TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
simple_unlock(&vnode_free_list_slock);
if (vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK) == 0)
VOP_INACTIVE(vp, p);
}
#ifdef DIAGNOSTIC
/*
* Page or buffer structure gets a reference.
*/
void
vhold(vp)
register struct vnode *vp;
{
simple_lock(&vp->v_interlock);
vp->v_holdcnt++;
simple_unlock(&vp->v_interlock);
}
/*
* Page or buffer structure frees a reference.
*/
void
holdrele(vp)
register struct vnode *vp;
{
simple_lock(&vp->v_interlock);
if (vp->v_holdcnt <= 0)
panic("holdrele: holdcnt");
vp->v_holdcnt--;
simple_unlock(&vp->v_interlock);
}
/*
* Vnode reference.
*/
void
vref(vp)
struct vnode *vp;
{
simple_lock(&vp->v_interlock);
if (vp->v_usecount <= 0)
panic("vref used where vget required");
vp->v_usecount++;
simple_unlock(&vp->v_interlock);
}
#endif /* DIAGNOSTIC */
/*
* Remove any vnodes in the vnode table belonging to mount point mp.
*
* If MNT_NOFORCE is specified, there should not be any active ones,
* return error if any are found (nb: this is a user error, not a
* system error). If MNT_FORCE is specified, detach any active vnodes
* that are found.
*/
#ifdef DEBUG
int busyprt = 0; /* print out busy vnodes */
struct ctldebug debug1 = { "busyprt", &busyprt };
#endif
int
vflush(mp, skipvp, flags)
struct mount *mp;
struct vnode *skipvp;
int flags;
{
struct proc *p = curproc; /* XXX */
register struct vnode *vp, *nvp;
int busy = 0;
simple_lock(&mntvnode_slock);
loop:
for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
if (vp->v_mount != mp)
goto loop;
nvp = vp->v_mntvnodes.le_next;
/*
* Skip over a selected vnode.
*/
if (vp == skipvp)
continue;
simple_lock(&vp->v_interlock);
/*
* Skip over a vnodes marked VSYSTEM.
*/
if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) {
simple_unlock(&vp->v_interlock);
continue;
}
/*
* If WRITECLOSE is set, only flush out regular file
* vnodes open for writing.
*/
if ((flags & WRITECLOSE) &&
(vp->v_writecount == 0 || vp->v_type != VREG)) {
simple_unlock(&vp->v_interlock);
continue;
}
/*
* With v_usecount == 0, all we need to do is clear
* out the vnode data structures and we are done.
*/
if (vp->v_usecount == 0) {
simple_unlock(&mntvnode_slock);
vgonel(vp, p);
simple_lock(&mntvnode_slock);
continue;
}
/*
* If FORCECLOSE is set, forcibly close the vnode.
* For block or character devices, revert to an
* anonymous device. For all other files, just kill them.
*/
if (flags & FORCECLOSE) {
simple_unlock(&mntvnode_slock);
if (vp->v_type != VBLK && vp->v_type != VCHR) {
vgonel(vp, p);
} else {
vclean(vp, 0, p);
vp->v_op = spec_vnodeop_p;
insmntque(vp, (struct mount *)0);
}
simple_lock(&mntvnode_slock);
continue;
}
#ifdef DEBUG
if (busyprt)
vprint("vflush: busy vnode", vp);
#endif
simple_unlock(&vp->v_interlock);
busy++;
}
simple_unlock(&mntvnode_slock);
if (busy)
return (EBUSY);
return (0);
}
/*
* Disassociate the underlying file system from a vnode.
*/
void
vclean(vp, flags, p)
register struct vnode *vp;
int flags;
struct proc *p;
{
int active;
/*
* Check to see if the vnode is in use.
* If so we have to reference it before we clean it out
* so that its count cannot fall to zero and generate a
* race against ourselves to recycle it.
*/
if ((active = vp->v_usecount) != 0)
/* We have the vnode interlock. */
vp->v_usecount++;
/*
* Prevent the vnode from being recycled or
* brought into use while we clean it out.
*/
if (vp->v_flag & VXLOCK)
panic("vclean: deadlock");
vp->v_flag |= VXLOCK;
/*
* Even if the count is zero, the VOP_INACTIVE routine may still
* have the object locked while it cleans it out. The VOP_LOCK
* ensures that the VOP_INACTIVE routine is done with its work.
* For active vnodes, it ensures that no other activity can
* occur while the underlying object is being cleaned out.
*/
VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK);
/*
* clean out any VM data associated with the vnode.
*/
uvm_vnp_terminate(vp);
/*
* Clean out any buffers associated with the vnode.
*/
if (flags & DOCLOSE)
vinvalbuf(vp, V_SAVE, NOCRED, p, 0, 0);
/*
* If purging an active vnode, it must be closed and
* deactivated before being reclaimed. Note that the
* VOP_INACTIVE will unlock the vnode.
*/
if (active) {
if (flags & DOCLOSE)
VOP_CLOSE(vp, FNONBLOCK, NOCRED, NULL);
VOP_INACTIVE(vp, p);
} else {
/*
* Any other processes trying to obtain this lock must first
* wait for VXLOCK to clear, then call the new lock operation.
*/
VOP_UNLOCK(vp, 0);
}
/*
* Reclaim the vnode.
*/
if (VOP_RECLAIM(vp, p))
panic("vclean: cannot reclaim");
if (active) {
/*
* Inline copy of vrele() since VOP_INACTIVE
* has already been called.
*/
simple_lock(&vp->v_interlock);
if (--vp->v_usecount <= 0) {
#ifdef DIAGNOSTIC
if (vp->v_usecount < 0 || vp->v_writecount != 0) {
vprint("vclean: bad ref count", vp);
panic("vclean: ref cnt");
}
#endif
/*
* Insert at tail of LRU list.
*/
simple_lock(&vnode_free_list_slock);
#ifdef DIAGNOSTIC
if (vp->v_vnlock) {
if ((vp->v_vnlock->lk_flags & LK_DRAINED) == 0)
vprint("vclean: lock not drained", vp);
}
#endif
TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
simple_unlock(&vnode_free_list_slock);
}
simple_unlock(&vp->v_interlock);
}
cache_purge(vp);
/*
* Done with purge, notify sleepers of the grim news.
*/
vp->v_op = dead_vnodeop_p;
vp->v_tag = VT_NON;
vp->v_flag &= ~VXLOCK;
if (vp->v_flag & VXWANT) {
vp->v_flag &= ~VXWANT;
wakeup((caddr_t)vp);
}
}
/*
* Recycle an unused vnode to the front of the free list.
* Release the passed interlock if the vnode will be recycled.
*/
int
vrecycle(vp, inter_lkp, p)
struct vnode *vp;
struct simplelock *inter_lkp;
struct proc *p;
{
simple_lock(&vp->v_interlock);
if (vp->v_usecount == 0) {
if (inter_lkp)
simple_unlock(inter_lkp);
vgonel(vp, p);
return (1);
}
simple_unlock(&vp->v_interlock);
return (0);
}
/*
* Eliminate all activity associated with a vnode
* in preparation for reuse.
*/
void
vgone(vp)
struct vnode *vp;
{
struct proc *p = curproc; /* XXX */
simple_lock(&vp->v_interlock);
vgonel(vp, p);
}
/*
* vgone, with the vp interlock held.
*/
void
vgonel(vp, p)
register struct vnode *vp;
struct proc *p;
{
struct vnode *vq;
struct vnode *vx;
/*
* If a vgone (or vclean) is already in progress,
* wait until it is done and return.
*/
if (vp->v_flag & VXLOCK) {
vp->v_flag |= VXWANT;
simple_unlock(&vp->v_interlock);
tsleep((caddr_t)vp, PINOD, "vgone", 0);
return;
}
/*
* Clean out the filesystem specific data.
*/
vclean(vp, DOCLOSE, p);
/*
* Delete from old mount point vnode list, if on one.
*/
if (vp->v_mount != NULL)
insmntque(vp, (struct mount *)0);
/*
* If special device, remove it from special device alias list.
* if it is on one.
*/
if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) {
simple_lock(&spechash_slock);
if (*vp->v_hashchain == vp) {
*vp->v_hashchain = vp->v_specnext;
} else {
for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) {
if (vq->v_specnext != vp)
continue;
vq->v_specnext = vp->v_specnext;
break;
}
if (vq == NULL)
panic("missing bdev");
}
if (vp->v_flag & VALIASED) {
vx = NULL;
for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) {
if (vq->v_rdev != vp->v_rdev ||
vq->v_type != vp->v_type)
continue;
if (vx)
break;
vx = vq;
}
if (vx == NULL)
panic("missing alias");
if (vq == NULL)
vx->v_flag &= ~VALIASED;
vp->v_flag &= ~VALIASED;
}
simple_unlock(&spechash_slock);
FREE(vp->v_specinfo, M_VNODE);
vp->v_specinfo = NULL;
}
/*
* If it is on the freelist and not already at the head,
* move it to the head of the list. The test of the back
* pointer and the reference count of zero is because
* it will be removed from the free list by getnewvnode,
* but will not have its reference count incremented until
* after calling vgone. If the reference count were
* incremented first, vgone would (incorrectly) try to
* close the previous instance of the underlying object.
* So, the back pointer is explicitly set to `0xdeadb' in
* getnewvnode after removing it from the freelist to ensure
* that we do not try to move it here.
*/
if (vp->v_usecount == 0) {
simple_lock(&vnode_free_list_slock);
if (vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb &&
vnode_free_list.tqh_first != vp) {
TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist);
}
simple_unlock(&vnode_free_list_slock);
}
vp->v_type = VBAD;
}
/*
* Lookup a vnode by device number.
*/
int
vfinddev(dev, type, vpp)
dev_t dev;
enum vtype type;
struct vnode **vpp;
{
struct vnode *vp;
int rc = 0;
simple_lock(&spechash_slock);
for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) {
if (dev != vp->v_rdev || type != vp->v_type)
continue;
*vpp = vp;
rc = 1;
break;
}
simple_unlock(&spechash_slock);
return (rc);
}
/*
* Revoke all the vnodes corresponding to the specified minor number
* range (endpoints inclusive) of the specified major.
*/
void
vdevgone(maj, minl, minh, type)
int maj, minl, minh;
enum vtype type;
{
struct vnode *vp;
int mn;
for (mn = minl; mn <= minh; mn++)
if (vfinddev(makedev(maj, mn), type, &vp))
VOP_REVOKE(vp, REVOKEALL);
}
/*
* Calculate the total number of references to a special device.
*/
int
vcount(vp)
register struct vnode *vp;
{
register struct vnode *vq, *vnext;
int count;
loop:
if ((vp->v_flag & VALIASED) == 0)
return (vp->v_usecount);
simple_lock(&spechash_slock);
for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) {
vnext = vq->v_specnext;
if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type)
continue;
/*
* Alias, but not in use, so flush it out.
*/
if (vq->v_usecount == 0 && vq != vp) {
simple_unlock(&spechash_slock);
vgone(vq);
goto loop;
}
count += vq->v_usecount;
}
simple_unlock(&spechash_slock);
return (count);
}
/*
* Print out a description of a vnode.
*/
static char *typename[] =
{ "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" };
void
vprint(label, vp)
char *label;
register struct vnode *vp;
{
char buf[64];
if (label != NULL)
printf("%s: ", label);
printf("type %s, usecount %d, writecount %d, refcount %ld,",
typename[vp->v_type], vp->v_usecount, vp->v_writecount,
vp->v_holdcnt);
buf[0] = '\0';
if (vp->v_flag & VROOT)
strcat(buf, "|VROOT");
if (vp->v_flag & VTEXT)
strcat(buf, "|VTEXT");
if (vp->v_flag & VSYSTEM)
strcat(buf, "|VSYSTEM");
if (vp->v_flag & VXLOCK)
strcat(buf, "|VXLOCK");
if (vp->v_flag & VXWANT)
strcat(buf, "|VXWANT");
if (vp->v_flag & VBWAIT)
strcat(buf, "|VBWAIT");
if (vp->v_flag & VALIASED)
strcat(buf, "|VALIASED");
if (buf[0] != '\0')
printf(" flags (%s)", &buf[1]);
if (vp->v_data == NULL) {
printf("\n");
} else {
printf("\n\t");
VOP_PRINT(vp);
}
}
#ifdef DEBUG
/*
* List all of the locked vnodes in the system.
* Called when debugging the kernel.
*/
void
printlockedvnodes()
{
struct mount *mp, *nmp;
struct vnode *vp;
printf("Locked vnodes\n");
simple_lock(&mountlist_slock);
for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
nmp = mp->mnt_list.cqe_next;
continue;
}
for (vp = mp->mnt_vnodelist.lh_first;
vp != NULL;
vp = vp->v_mntvnodes.le_next) {
if (VOP_ISLOCKED(vp))
vprint((char *)0, vp);
}
simple_lock(&mountlist_slock);
nmp = mp->mnt_list.cqe_next;
vfs_unbusy(mp);
}
simple_unlock(&mountlist_slock);
}
#endif
extern const char *mountcompatnames[];
extern const int nmountcompatnames;
/*
* Top level filesystem related information gathering.
*/
int
vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
int *name;
u_int namelen;
void *oldp;
size_t *oldlenp;
void *newp;
size_t newlen;
struct proc *p;
{
#if defined(COMPAT_09) || defined(COMPAT_43) || defined(COMPAT_44)
struct vfsconf vfc;
#endif
struct vfsops *vfsp;
/* all sysctl names at this level are at least name and field */
if (namelen < 2)
return (ENOTDIR); /* overloaded */
/* Not generic: goes to file system. */
if (name[0] != VFS_GENERIC) {
if (name[0] >= nmountcompatnames || name[0] < 0 ||
mountcompatnames[name[0]] == NULL)
return (EOPNOTSUPP);
vfsp = vfs_getopsbyname(mountcompatnames[name[0]]);
if (vfsp == NULL || vfsp->vfs_sysctl == NULL)
return (EOPNOTSUPP);
return ((*vfsp->vfs_sysctl)(&name[1], namelen - 1,
oldp, oldlenp, newp, newlen, p));
}
/* The rest are generic vfs sysctls. */
switch (name[1]) {
#if defined(COMPAT_09) || defined(COMPAT_43) || defined(COMPAT_44)
case VFS_MAXTYPENUM:
/*
* Provided for 4.4BSD-Lite2 compatibility.
*/
return (sysctl_rdint(oldp, oldlenp, newp, nmountcompatnames));
case VFS_CONF:
/*
* Special: a node, next is a file system name.
* Provided for 4.4BSD-Lite2 compatibility.
*/
if (namelen < 3)
return (ENOTDIR); /* overloaded */
if (name[2] >= nmountcompatnames || name[2] < 0 ||
mountcompatnames[name[2]] == NULL)
return (EOPNOTSUPP);
vfsp = vfs_getopsbyname(mountcompatnames[name[2]]);
if (vfsp == NULL)
return (EOPNOTSUPP);
vfc.vfc_vfsops = vfsp;
strncpy(vfc.vfc_name, vfsp->vfs_name, MFSNAMELEN);
vfc.vfc_typenum = name[2];
vfc.vfc_refcount = vfsp->vfs_refcount;
vfc.vfc_flags = 0;
vfc.vfc_mountroot = vfsp->vfs_mountroot;
vfc.vfc_next = NULL;
return (sysctl_rdstruct(oldp, oldlenp, newp, &vfc,
sizeof(struct vfsconf)));
#endif
default:
break;
}
return (EOPNOTSUPP);
}
int kinfo_vdebug = 1;
int kinfo_vgetfailed;
#define KINFO_VNODESLOP 10
/*
* Dump vnode list (via sysctl).
* Copyout address of vnode followed by vnode.
*/
/* ARGSUSED */
int
sysctl_vnode(where, sizep, p)
char *where;
size_t *sizep;
struct proc *p;
{
struct mount *mp, *nmp;
struct vnode *nvp, *vp;
char *bp = where, *savebp;
char *ewhere;
int error;
#define VPTRSZ sizeof(struct vnode *)
#define VNODESZ sizeof(struct vnode)
if (where == NULL) {
*sizep = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ);
return (0);
}
ewhere = where + *sizep;
simple_lock(&mountlist_slock);
for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) {
nmp = mp->mnt_list.cqe_next;
continue;
}
savebp = bp;
again:
simple_lock(&mntvnode_slock);
for (vp = mp->mnt_vnodelist.lh_first;
vp != NULL;
vp = nvp) {
/*
* Check that the vp is still associated with
* this filesystem. RACE: could have been
* recycled onto the same filesystem.
*/
if (vp->v_mount != mp) {
simple_unlock(&mntvnode_slock);
if (kinfo_vdebug)
printf("kinfo: vp changed\n");
bp = savebp;
goto again;
}
nvp = vp->v_mntvnodes.le_next;
if (bp + VPTRSZ + VNODESZ > ewhere) {
simple_unlock(&mntvnode_slock);
*sizep = bp - where;
return (ENOMEM);
}
simple_unlock(&mntvnode_slock);
if ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) ||
(error = copyout((caddr_t)vp, bp + VPTRSZ, VNODESZ)))
return (error);
bp += VPTRSZ + VNODESZ;
simple_lock(&mntvnode_slock);
}
simple_unlock(&mntvnode_slock);
simple_lock(&mountlist_slock);
nmp = mp->mnt_list.cqe_next;
vfs_unbusy(mp);
}
simple_unlock(&mountlist_slock);
*sizep = bp - where;
return (0);
}
/*
* Check to see if a filesystem is mounted on a block device.
*/
int
vfs_mountedon(vp)
struct vnode *vp;
{
struct vnode *vq;
int error = 0;
if (vp->v_specflags & SI_MOUNTEDON)
return (EBUSY);
if (vp->v_flag & VALIASED) {
simple_lock(&spechash_slock);
for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) {
if (vq->v_rdev != vp->v_rdev ||
vq->v_type != vp->v_type)
continue;
if (vq->v_specflags & SI_MOUNTEDON) {
error = EBUSY;
break;
}
}
simple_unlock(&spechash_slock);
}
return (error);
}
/*
* Build hash lists of net addresses and hang them off the mount point.
* Called by ufs_mount() to set up the lists of export addresses.
*/
static int
vfs_hang_addrlist(mp, nep, argp)
struct mount *mp;
struct netexport *nep;
struct export_args *argp;
{
register struct netcred *np, *enp;
register struct radix_node_head *rnh;
register int i;
struct radix_node *rn;
struct sockaddr *saddr, *smask = 0;
struct domain *dom;
int error;
if (argp->ex_addrlen == 0) {
if (mp->mnt_flag & MNT_DEFEXPORTED)
return (EPERM);
np = &nep->ne_defexported;
np->netc_exflags = argp->ex_flags;
np->netc_anon = argp->ex_anon;
np->netc_anon.cr_ref = 1;
mp->mnt_flag |= MNT_DEFEXPORTED;
return (0);
}
i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen;
np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK);
memset((caddr_t)np, 0, i);
saddr = (struct sockaddr *)(np + 1);
error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen);
if (error)
goto out;
if (saddr->sa_len > argp->ex_addrlen)
saddr->sa_len = argp->ex_addrlen;
if (argp->ex_masklen) {
smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen);
error = copyin(argp->ex_mask, (caddr_t)smask, argp->ex_masklen);
if (error)
goto out;
if (smask->sa_len > argp->ex_masklen)
smask->sa_len = argp->ex_masklen;
}
i = saddr->sa_family;
if ((rnh = nep->ne_rtable[i]) == 0) {
/*
* Seems silly to initialize every AF when most are not
* used, do so on demand here
*/
for (dom = domains; dom; dom = dom->dom_next)
if (dom->dom_family == i && dom->dom_rtattach) {
dom->dom_rtattach((void **)&nep->ne_rtable[i],
dom->dom_rtoffset);
break;
}
if ((rnh = nep->ne_rtable[i]) == 0) {
error = ENOBUFS;
goto out;
}
}
rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh,
np->netc_rnodes);
if (rn == 0 || np != (struct netcred *)rn) { /* already exists */
if (rn == 0) {
enp = (struct netcred *)(*rnh->rnh_lookup)(saddr,
smask, rnh);
if (enp == 0) {
error = EPERM;
goto out;
}
} else
enp = (struct netcred *)rn;
if (enp->netc_exflags != argp->ex_flags ||
enp->netc_anon.cr_uid != argp->ex_anon.cr_uid ||
enp->netc_anon.cr_gid != argp->ex_anon.cr_gid ||
enp->netc_anon.cr_ngroups != argp->ex_anon.cr_ngroups ||
memcmp(&enp->netc_anon.cr_groups, &argp->ex_anon.cr_groups,
enp->netc_anon.cr_ngroups))
error = EPERM;
else
error = 0;
goto out;
}
np->netc_exflags = argp->ex_flags;
np->netc_anon = argp->ex_anon;
np->netc_anon.cr_ref = 1;
return (0);
out:
free(np, M_NETADDR);
return (error);
}
/* ARGSUSED */
static int
vfs_free_netcred(rn, w)
struct radix_node *rn;
void *w;
{
register struct radix_node_head *rnh = (struct radix_node_head *)w;
(*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh);
free((caddr_t)rn, M_NETADDR);
return (0);
}
/*
* Free the net address hash lists that are hanging off the mount points.
*/
static void
vfs_free_addrlist(nep)
struct netexport *nep;
{
register int i;
register struct radix_node_head *rnh;
for (i = 0; i <= AF_MAX; i++)
if ((rnh = nep->ne_rtable[i]) != NULL) {
(*rnh->rnh_walktree)(rnh, vfs_free_netcred, rnh);
free((caddr_t)rnh, M_RTABLE);
nep->ne_rtable[i] = 0;
}
}
int
vfs_export(mp, nep, argp)
struct mount *mp;
struct netexport *nep;
struct export_args *argp;
{
int error;
if (argp->ex_flags & MNT_DELEXPORT) {
if (mp->mnt_flag & MNT_EXPUBLIC) {
vfs_setpublicfs(NULL, NULL, NULL);
mp->mnt_flag &= ~MNT_EXPUBLIC;
}
vfs_free_addrlist(nep);
mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED);
}
if (argp->ex_flags & MNT_EXPORTED) {
if (argp->ex_flags & MNT_EXPUBLIC) {
if ((error = vfs_setpublicfs(mp, nep, argp)) != 0)
return (error);
mp->mnt_flag |= MNT_EXPUBLIC;
}
if ((error = vfs_hang_addrlist(mp, nep, argp)) != 0)
return (error);
mp->mnt_flag |= MNT_EXPORTED;
}
return (0);
}
/*
* Set the publicly exported filesystem (WebNFS). Currently, only
* one public filesystem is possible in the spec (RFC 2054 and 2055)
*/
int
vfs_setpublicfs(mp, nep, argp)
struct mount *mp;
struct netexport *nep;
struct export_args *argp;
{
int error;
struct vnode *rvp;
char *cp;
/*
* mp == NULL -> invalidate the current info, the FS is
* no longer exported. May be called from either vfs_export
* or unmount, so check if it hasn't already been done.
*/
if (mp == NULL) {
if (nfs_pub.np_valid) {
nfs_pub.np_valid = 0;
if (nfs_pub.np_index != NULL) {
FREE(nfs_pub.np_index, M_TEMP);
nfs_pub.np_index = NULL;
}
}
return (0);
}
/*
* Only one allowed at a time.
*/
if (nfs_pub.np_valid != 0 && mp != nfs_pub.np_mount)
return (EBUSY);
/*
* Get real filehandle for root of exported FS.
*/
memset((caddr_t)&nfs_pub.np_handle, 0, sizeof(nfs_pub.np_handle));
nfs_pub.np_handle.fh_fsid = mp->mnt_stat.f_fsid;
if ((error = VFS_ROOT(mp, &rvp)))
return (error);
if ((error = VFS_VPTOFH(rvp, &nfs_pub.np_handle.fh_fid)))
return (error);
vput(rvp);
/*
* If an indexfile was specified, pull it in.
*/
if (argp->ex_indexfile != NULL) {
MALLOC(nfs_pub.np_index, char *, MAXNAMLEN + 1, M_TEMP,
M_WAITOK);
error = copyinstr(argp->ex_indexfile, nfs_pub.np_index,
MAXNAMLEN, (size_t *)0);
if (!error) {
/*
* Check for illegal filenames.
*/
for (cp = nfs_pub.np_index; *cp; cp++) {
if (*cp == '/') {
error = EINVAL;
break;
}
}
}
if (error) {
FREE(nfs_pub.np_index, M_TEMP);
return (error);
}
}
nfs_pub.np_mount = mp;
nfs_pub.np_valid = 1;
return (0);
}
struct netcred *
vfs_export_lookup(mp, nep, nam)
register struct mount *mp;
struct netexport *nep;
struct mbuf *nam;
{
register struct netcred *np;
register struct radix_node_head *rnh;
struct sockaddr *saddr;
np = NULL;
if (mp->mnt_flag & MNT_EXPORTED) {
/*
* Lookup in the export list first.
*/
if (nam != NULL) {
saddr = mtod(nam, struct sockaddr *);
rnh = nep->ne_rtable[saddr->sa_family];
if (rnh != NULL) {
np = (struct netcred *)
(*rnh->rnh_matchaddr)((caddr_t)saddr,
rnh);
if (np && np->netc_rnodes->rn_flags & RNF_ROOT)
np = NULL;
}
}
/*
* If no address match, use the default if it exists.
*/
if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED)
np = &nep->ne_defexported;
}
return (np);
}
/*
* Do the usual access checking.
* file_mode, uid and gid are from the vnode in question,
* while acc_mode and cred are from the VOP_ACCESS parameter list
*/
int
vaccess(type, file_mode, uid, gid, acc_mode, cred)
enum vtype type;
mode_t file_mode;
uid_t uid;
gid_t gid;
mode_t acc_mode;
struct ucred *cred;
{
mode_t mask;
/*
* Super-user always gets read/write access, but execute access depends
* on at least one execute bit being set.
*/
if (cred->cr_uid == 0) {
if ((acc_mode & VEXEC) && type != VDIR &&
(file_mode & (S_IXUSR|S_IXGRP|S_IXOTH)) == 0)
return (EACCES);
return (0);
}
mask = 0;
/* Otherwise, check the owner. */
if (cred->cr_uid == uid) {
if (acc_mode & VEXEC)
mask |= S_IXUSR;
if (acc_mode & VREAD)
mask |= S_IRUSR;
if (acc_mode & VWRITE)
mask |= S_IWUSR;
return ((file_mode & mask) == mask ? 0 : EACCES);
}
/* Otherwise, check the groups. */
if (cred->cr_gid == gid || groupmember(gid, cred)) {
if (acc_mode & VEXEC)
mask |= S_IXGRP;
if (acc_mode & VREAD)
mask |= S_IRGRP;
if (acc_mode & VWRITE)
mask |= S_IWGRP;
return ((file_mode & mask) == mask ? 0 : EACCES);
}
/* Otherwise, check everyone else. */
if (acc_mode & VEXEC)
mask |= S_IXOTH;
if (acc_mode & VREAD)
mask |= S_IROTH;
if (acc_mode & VWRITE)
mask |= S_IWOTH;
return ((file_mode & mask) == mask ? 0 : EACCES);
}
/*
* Unmount all file systems.
* We traverse the list in reverse order under the assumption that doing so
* will avoid needing to worry about dependencies.
*/
void
vfs_unmountall()
{
register struct mount *mp, *nmp;
int allerror, error;
struct proc *p = curproc; /* XXX */
/*
* Unmounting a file system blocks the requesting process.
* However, it's possible for this routine to be called when
* curproc is NULL (e.g. panic situation, or via the debugger).
* If we get stuck in this situation, just abort, since any
* attempts to sleep will fault.
*/
if (p == NULL) {
printf("vfs_unmountall: no context, aborting\n");
return;
}
for (allerror = 0,
mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
nmp = mp->mnt_list.cqe_prev;
#ifdef DEBUG
printf("unmounting %s (%s)...\n",
mp->mnt_stat.f_mntonname, mp->mnt_stat.f_mntfromname);
#endif
if (vfs_busy(mp, 0, 0))
continue;
if ((error = dounmount(mp, MNT_FORCE, p)) != 0) {
printf("unmount of %s failed with error %d\n",
mp->mnt_stat.f_mntonname, error);
allerror = 1;
}
}
if (allerror)
printf("WARNING: some file systems would not unmount\n");
}
/*
* Sync and unmount file systems before shutting down.
*/
void
vfs_shutdown()
{
register struct buf *bp;
int iter, nbusy;
printf("syncing disks... ");
/* XXX Should suspend scheduling. */
(void) spl0();
sys_sync(&proc0, (void *)0, (register_t *)0);
/* Wait for sync to finish. */
for (iter = 0; iter < 20; iter++) {
nbusy = 0;
for (bp = &buf[nbuf]; --bp >= buf; )
if ((bp->b_flags & (B_BUSY|B_INVAL)) == B_BUSY)
nbusy++;
if (nbusy == 0)
break;
printf("%d ", nbusy);
DELAY(40000 * iter);
}
if (nbusy) {
printf("giving up\n");
return;
} else
printf("done\n");
/*
* If we've panic'd, don't make the situation potentially
* worse by unmounting the file systems.
*/
if (panicstr != NULL)
return;
/* Release inodes held by texts before update. */
#ifdef notdef
vnshutdown();
#endif
/* Unmount file systems. */
vfs_unmountall();
}
/*
* Mount the root file system. If the operator didn't specify a
* file system to use, try all possible file systems until one
* succeeds.
*/
int
vfs_mountroot()
{
extern int (*mountroot) __P((void));
struct vfsops *v;
if (root_device == NULL)
panic("vfs_mountroot: root device unknown");
switch (root_device->dv_class) {
case DV_IFNET:
if (rootdev != NODEV)
panic("vfs_mountroot: rootdev set for DV_IFNET");
break;
case DV_DISK:
if (rootdev == NODEV)
panic("vfs_mountroot: rootdev not set for DV_DISK");
break;
default:
printf("%s: inappropriate for root file system\n",
root_device->dv_xname);
return (ENODEV);
}
/*
* If user specified a file system, use it.
*/
if (mountroot != NULL)
return ((*mountroot)());
/*
* Try each file system currently configured into the kernel.
*/
for (v = LIST_FIRST(&vfs_list); v != NULL; v = LIST_NEXT(v, vfs_list)) {
if (v->vfs_mountroot == NULL)
continue;
#ifdef DEBUG
printf("mountroot: trying %s...\n", v->vfs_name);
#endif
if ((*v->vfs_mountroot)() == 0) {
printf("root file system type: %s\n", v->vfs_name);
break;
}
}
if (v == NULL) {
printf("no file system for %s", root_device->dv_xname);
if (root_device->dv_class == DV_DISK)
printf(" (dev 0x%x)", rootdev);
printf("\n");
return (EFTYPE);
}
return (0);
}
/*
* Given a file system name, look up the vfsops for that
* file system, or return NULL if file system isn't present
* in the kernel.
*/
struct vfsops *
vfs_getopsbyname(name)
const char *name;
{
struct vfsops *v;
for (v = LIST_FIRST(&vfs_list); v != NULL; v = LIST_NEXT(v, vfs_list)) {
if (strcmp(v->vfs_name, name) == 0)
break;
}
return (v);
}
/*
* Establish a file system and initialize it.
*/
int
vfs_attach(vfs)
struct vfsops *vfs;
{
struct vfsops *v;
int error = 0;
/*
* Make sure this file system doesn't already exist.
*/
for (v = LIST_FIRST(&vfs_list); v != NULL; v = LIST_NEXT(v, vfs_list)) {
if (strcmp(vfs->vfs_name, v->vfs_name) == 0) {
error = EEXIST;
goto out;
}
}
/*
* Initialize the vnode operations for this file system.
*/
vfs_opv_init(vfs->vfs_opv_descs);
/*
* Now initialize the file system itself.
*/
(*vfs->vfs_init)();
/*
* ...and link it into the kernel's list.
*/
LIST_INSERT_HEAD(&vfs_list, vfs, vfs_list);
/*
* Sanity: make sure the reference count is 0.
*/
vfs->vfs_refcount = 0;
out:
return (error);
}
/*
* Remove a file system from the kernel.
*/
int
vfs_detach(vfs)
struct vfsops *vfs;
{
struct vfsops *v;
/*
* Make sure no one is using the filesystem.
*/
if (vfs->vfs_refcount != 0)
return (EBUSY);
/*
* ...and remove it from the kernel's list.
*/
for (v = LIST_FIRST(&vfs_list); v != NULL; v = LIST_NEXT(v, vfs_list)) {
if (v == vfs) {
LIST_REMOVE(v, vfs_list);
break;
}
}
if (v == NULL)
return (ESRCH);
/*
* Free the vnode operations vector.
*/
vfs_opv_free(vfs->vfs_opv_descs);
return (0);
}
Reference in New Issue View Git Blame Copy Permalink