Home | History | Annotate | Line # | Download | only in kern
vfs_syscalls.c revision 1.383
      1  1.383  christos /*	$NetBSD: vfs_syscalls.c,v 1.383 2009/01/11 02:45:53 christos Exp $	*/
      2  1.345        ad 
      3  1.345        ad /*-
      4  1.345        ad  * Copyright (c) 2008 The NetBSD Foundation, Inc.
      5  1.345        ad  * All rights reserved.
      6  1.345        ad  *
      7  1.345        ad  * Redistribution and use in source and binary forms, with or without
      8  1.345        ad  * modification, are permitted provided that the following conditions
      9  1.345        ad  * are met:
     10  1.345        ad  * 1. Redistributions of source code must retain the above copyright
     11  1.345        ad  *    notice, this list of conditions and the following disclaimer.
     12  1.345        ad  * 2. Redistributions in binary form must reproduce the above copyright
     13  1.345        ad  *    notice, this list of conditions and the following disclaimer in the
     14  1.345        ad  *    documentation and/or other materials provided with the distribution.
     15  1.345        ad  *
     16  1.345        ad  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     17  1.345        ad  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     18  1.345        ad  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     19  1.345        ad  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     20  1.345        ad  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     21  1.345        ad  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     22  1.345        ad  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     23  1.345        ad  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     24  1.345        ad  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     25  1.345        ad  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     26  1.345        ad  * POSSIBILITY OF SUCH DAMAGE.
     27  1.345        ad  */
     28   1.31       cgd 
     29   1.31       cgd /*
     30   1.31       cgd  * Copyright (c) 1989, 1993
     31   1.31       cgd  *	The Regents of the University of California.  All rights reserved.
     32   1.31       cgd  * (c) UNIX System Laboratories, Inc.
     33   1.31       cgd  * All or some portions of this file are derived from material licensed
     34   1.31       cgd  * to the University of California by American Telephone and Telegraph
     35   1.31       cgd  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     36   1.31       cgd  * the permission of UNIX System Laboratories, Inc.
     37   1.31       cgd  *
     38   1.31       cgd  * Redistribution and use in source and binary forms, with or without
     39   1.31       cgd  * modification, are permitted provided that the following conditions
     40   1.31       cgd  * are met:
     41   1.31       cgd  * 1. Redistributions of source code must retain the above copyright
     42   1.31       cgd  *    notice, this list of conditions and the following disclaimer.
     43   1.31       cgd  * 2. Redistributions in binary form must reproduce the above copyright
     44   1.31       cgd  *    notice, this list of conditions and the following disclaimer in the
     45   1.31       cgd  *    documentation and/or other materials provided with the distribution.
     46  1.191       agc  * 3. Neither the name of the University nor the names of its contributors
     47   1.31       cgd  *    may be used to endorse or promote products derived from this software
     48   1.31       cgd  *    without specific prior written permission.
     49   1.31       cgd  *
     50   1.31       cgd  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     51   1.31       cgd  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     52   1.31       cgd  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     53   1.31       cgd  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     54   1.31       cgd  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     55   1.31       cgd  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     56   1.31       cgd  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     57   1.31       cgd  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     58   1.31       cgd  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     59   1.31       cgd  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     60   1.31       cgd  * SUCH DAMAGE.
     61   1.31       cgd  *
     62  1.113      fvdl  *	@(#)vfs_syscalls.c	8.42 (Berkeley) 7/31/95
     63   1.31       cgd  */
     64  1.173     lukem 
     65  1.173     lukem #include <sys/cdefs.h>
     66  1.383  christos __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.383 2009/01/11 02:45:53 christos Exp $");
     67  1.111       mrg 
     68  1.378        ad #ifdef _KERNEL_OPT
     69  1.258      elad #include "opt_fileassoc.h"
     70  1.261    dogcow #include "veriexec.h"
     71  1.378        ad #endif
     72   1.31       cgd 
     73   1.31       cgd #include <sys/param.h>
     74   1.31       cgd #include <sys/systm.h>
     75   1.31       cgd #include <sys/namei.h>
     76   1.31       cgd #include <sys/filedesc.h>
     77   1.31       cgd #include <sys/kernel.h>
     78   1.31       cgd #include <sys/file.h>
     79   1.31       cgd #include <sys/stat.h>
     80   1.31       cgd #include <sys/vnode.h>
     81   1.31       cgd #include <sys/mount.h>
     82   1.31       cgd #include <sys/proc.h>
     83   1.31       cgd #include <sys/uio.h>
     84   1.31       cgd #include <sys/malloc.h>
     85  1.248      yamt #include <sys/kmem.h>
     86   1.31       cgd #include <sys/dirent.h>
     87  1.172    simonb #include <sys/sysctl.h>
     88   1.35       cgd #include <sys/syscallargs.h>
     89  1.306       dsl #include <sys/vfs_syscalls.h>
     90  1.192  drochner #include <sys/ktrace.h>
     91  1.251      elad #ifdef FILEASSOC
     92  1.251      elad #include <sys/fileassoc.h>
     93  1.251      elad #endif /* FILEASSOC */
     94  1.219     blymn #include <sys/verified_exec.h>
     95  1.242      elad #include <sys/kauth.h>
     96  1.346        ad #include <sys/atomic.h>
     97  1.360        ad #include <sys/module.h>
     98  1.380     pooka #include <sys/buf.h>
     99   1.35       cgd 
    100  1.148      fvdl #include <miscfs/genfs/genfs.h>
    101  1.148      fvdl #include <miscfs/syncfs/syncfs.h>
    102  1.342        ad #include <miscfs/specfs/specdev.h>
    103  1.181   thorpej 
    104  1.232      jmmv #include <nfs/rpcv2.h>
    105  1.232      jmmv #include <nfs/nfsproto.h>
    106  1.232      jmmv #include <nfs/nfs.h>
    107  1.232      jmmv #include <nfs/nfs_var.h>
    108  1.232      jmmv 
    109  1.181   thorpej MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct");
    110  1.110       mrg 
    111  1.234  christos static int change_dir(struct nameidata *, struct lwp *);
    112  1.234  christos static int change_flags(struct vnode *, u_long, struct lwp *);
    113  1.234  christos static int change_mode(struct vnode *, int, struct lwp *l);
    114  1.234  christos static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int);
    115   1.63  christos 
    116  1.205  junyoung void checkdirs(struct vnode *);
    117   1.31       cgd 
    118  1.150      fvdl int dovfsusermount = 0;
    119  1.150      fvdl 
    120   1.31       cgd /*
    121   1.31       cgd  * Virtual File System System Calls
    122   1.31       cgd  */
    123   1.31       cgd 
    124   1.31       cgd /*
    125   1.31       cgd  * Mount a file system.
    126   1.31       cgd  */
    127   1.99   thorpej 
    128   1.99   thorpej /*
    129   1.99   thorpej  * This table is used to maintain compatibility with 4.3BSD
    130  1.378        ad  * and NetBSD 0.9 mount syscalls - and possibly other systems.
    131  1.378        ad  * Note, the order is important!
    132  1.124   thorpej  *
    133  1.167  jdolecek  * Do not modify this table. It should only contain filesystems
    134  1.167  jdolecek  * supported by NetBSD 0.9 and 4.3BSD.
    135   1.99   thorpej  */
    136  1.167  jdolecek const char * const mountcompatnames[] = {
    137   1.99   thorpej 	NULL,		/* 0 = MOUNT_NONE */
    138  1.167  jdolecek 	MOUNT_FFS,	/* 1 = MOUNT_UFS */
    139   1.99   thorpej 	MOUNT_NFS,	/* 2 */
    140   1.99   thorpej 	MOUNT_MFS,	/* 3 */
    141   1.99   thorpej 	MOUNT_MSDOS,	/* 4 */
    142  1.167  jdolecek 	MOUNT_CD9660,	/* 5 = MOUNT_ISOFS */
    143  1.167  jdolecek 	MOUNT_FDESC,	/* 6 */
    144  1.167  jdolecek 	MOUNT_KERNFS,	/* 7 */
    145  1.167  jdolecek 	NULL,		/* 8 = MOUNT_DEVFS */
    146  1.167  jdolecek 	MOUNT_AFS,	/* 9 */
    147   1.99   thorpej };
    148  1.113      fvdl const int nmountcompatnames = sizeof(mountcompatnames) /
    149   1.99   thorpej     sizeof(mountcompatnames[0]);
    150   1.99   thorpej 
    151  1.285      elad static int
    152  1.283      elad mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags,
    153  1.324     pooka     void *data, size_t *data_len)
    154   1.56   thorpej {
    155  1.113      fvdl 	struct mount *mp;
    156  1.283      elad 	int error = 0, saved_flags;
    157  1.283      elad 
    158  1.283      elad 	mp = vp->v_mount;
    159  1.283      elad 	saved_flags = mp->mnt_flag;
    160  1.283      elad 
    161  1.329        ad 	/* We can operate only on VV_ROOT nodes. */
    162  1.329        ad 	if ((vp->v_vflag & VV_ROOT) == 0) {
    163  1.329        ad 		error = EINVAL;
    164  1.329        ad 		goto out;
    165  1.329        ad 	}
    166   1.31       cgd 
    167  1.218      yamt 	/*
    168  1.283      elad 	 * We only allow the filesystem to be reloaded if it
    169  1.373        ad 	 * is currently mounted read-only.  Additionally, we
    170  1.373        ad 	 * prevent read-write to read-only downgrades.
    171  1.283      elad 	 */
    172  1.373        ad 	if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 &&
    173  1.373        ad 	    (mp->mnt_flag & MNT_RDONLY) == 0) {
    174  1.329        ad 		error = EOPNOTSUPP;	/* Needs translation */
    175  1.329        ad 		goto out;
    176  1.329        ad 	}
    177  1.292      elad 
    178  1.292      elad 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
    179  1.292      elad 	    KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data);
    180  1.292      elad 	if (error)
    181  1.329        ad 		goto out;
    182  1.292      elad 
    183  1.358        ad 	if (vfs_busy(mp, NULL)) {
    184  1.329        ad 		error = EPERM;
    185  1.329        ad 		goto out;
    186  1.329        ad 	}
    187  1.283      elad 
    188  1.358        ad 	mutex_enter(&mp->mnt_updating);
    189  1.358        ad 
    190  1.286      yamt 	mp->mnt_flag &= ~MNT_OP_FLAGS;
    191  1.286      yamt 	mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE);
    192  1.286      yamt 
    193  1.283      elad 	/*
    194  1.283      elad 	 * Set the mount level flags.
    195  1.283      elad 	 */
    196  1.283      elad 	if (flags & MNT_RDONLY)
    197  1.283      elad 		mp->mnt_flag |= MNT_RDONLY;
    198  1.283      elad 	else if (mp->mnt_flag & MNT_RDONLY)
    199  1.283      elad 		mp->mnt_iflag |= IMNT_WANTRDWR;
    200  1.283      elad 	mp->mnt_flag &=
    201  1.283      elad 	  ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
    202  1.283      elad 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
    203  1.370    simonb 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
    204  1.370    simonb 	    MNT_LOG);
    205  1.283      elad 	mp->mnt_flag |= flags &
    206  1.283      elad 	   (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
    207  1.283      elad 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
    208  1.283      elad 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
    209  1.370    simonb 	    MNT_LOG | MNT_IGNORE);
    210  1.283      elad 
    211  1.332     pooka 	error = VFS_MOUNT(mp, path, data, data_len);
    212  1.283      elad 
    213  1.320       dsl 	if (error && data != NULL) {
    214  1.283      elad 		int error2;
    215  1.283      elad 
    216  1.378        ad 		/*
    217  1.378        ad 		 * Update failed; let's try and see if it was an
    218  1.379        ad 		 * export request.  For compat with 3.0 and earlier.
    219  1.378        ad 		 */
    220  1.381        ad 		error2 = vfs_hooks_reexport(mp, path, data);
    221  1.283      elad 
    222  1.381        ad 		/*
    223  1.381        ad 		 * Only update error code if the export request was
    224  1.283      elad 		 * understood but some problem occurred while
    225  1.381        ad 		 * processing it.
    226  1.381        ad 		 */
    227  1.283      elad 		if (error2 != EJUSTRETURN)
    228  1.283      elad 			error = error2;
    229  1.283      elad 	}
    230  1.381        ad 
    231  1.283      elad 	if (mp->mnt_iflag & IMNT_WANTRDWR)
    232  1.283      elad 		mp->mnt_flag &= ~MNT_RDONLY;
    233  1.283      elad 	if (error)
    234  1.283      elad 		mp->mnt_flag = saved_flags;
    235  1.286      yamt 	mp->mnt_flag &= ~MNT_OP_FLAGS;
    236  1.286      yamt 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
    237  1.283      elad 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) {
    238  1.283      elad 		if (mp->mnt_syncer == NULL)
    239  1.283      elad 			error = vfs_allocate_syncvnode(mp);
    240  1.125       tls 	} else {
    241  1.283      elad 		if (mp->mnt_syncer != NULL)
    242  1.283      elad 			vfs_deallocate_syncvnode(mp);
    243  1.283      elad 	}
    244  1.358        ad 	mutex_exit(&mp->mnt_updating);
    245  1.354        ad 	vfs_unbusy(mp, false, NULL);
    246  1.283      elad 
    247  1.329        ad  out:
    248  1.283      elad 	return (error);
    249  1.283      elad }
    250  1.283      elad 
    251  1.285      elad static int
    252  1.320       dsl mount_get_vfsops(const char *fstype, struct vfsops **vfsops)
    253  1.283      elad {
    254  1.322  christos 	char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)];
    255  1.283      elad 	int error;
    256  1.283      elad 
    257  1.320       dsl 	/* Copy file-system type from userspace.  */
    258  1.322  christos 	error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL);
    259   1.63  christos 	if (error) {
    260   1.54       cgd 		/*
    261  1.227      jmmv 		 * Historically, filesystem types were identified by numbers.
    262   1.54       cgd 		 * If we get an integer for the filesystem type instead of a
    263   1.54       cgd 		 * string, we check to see if it matches one of the historic
    264   1.54       cgd 		 * filesystem types.
    265  1.205  junyoung 		 */
    266  1.283      elad 		u_long fsindex = (u_long)fstype;
    267   1.99   thorpej 		if (fsindex >= nmountcompatnames ||
    268  1.320       dsl 		    mountcompatnames[fsindex] == NULL)
    269  1.320       dsl 			return ENODEV;
    270  1.331     pooka 		strlcpy(fstypename, mountcompatnames[fsindex],
    271  1.331     pooka 		    sizeof(fstypename));
    272   1.31       cgd 	}
    273  1.283      elad 
    274  1.378        ad 	/* Accept `ufs' as an alias for `ffs', for compatibility. */
    275  1.320       dsl 	if (strcmp(fstypename, "ufs") == 0)
    276  1.320       dsl 		fstypename[0] = 'f';
    277  1.285      elad 
    278  1.360        ad 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
    279  1.360        ad 		return 0;
    280  1.360        ad 
    281  1.360        ad 	/* If we can autoload a vfs module, try again */
    282  1.376        ad 	mutex_enter(&module_lock);
    283  1.376        ad 	(void)module_autoload(fstype, MODULE_CLASS_VFS);
    284  1.376        ad 	mutex_exit(&module_lock);
    285  1.360        ad 
    286  1.360        ad 	if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL)
    287  1.360        ad 		return 0;
    288  1.360        ad 
    289  1.360        ad 	return ENODEV;
    290  1.320       dsl }
    291  1.320       dsl 
    292  1.320       dsl static int
    293  1.320       dsl mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops,
    294  1.337        ad     const char *path, int flags, void *data, size_t *data_len, u_int recurse)
    295  1.320       dsl {
    296  1.366    simonb 	struct mount *mp;
    297  1.320       dsl 	struct vnode *vp = *vpp;
    298  1.320       dsl 	struct vattr va;
    299  1.320       dsl 	int error;
    300  1.320       dsl 
    301  1.320       dsl 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
    302  1.320       dsl 	    KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data);
    303  1.320       dsl 	if (error)
    304  1.320       dsl 		return error;
    305  1.320       dsl 
    306  1.320       dsl 	/* Can't make a non-dir a mount-point (from here anyway). */
    307  1.320       dsl 	if (vp->v_type != VDIR)
    308  1.320       dsl 		return ENOTDIR;
    309  1.320       dsl 
    310  1.320       dsl 	/*
    311  1.320       dsl 	 * If the user is not root, ensure that they own the directory
    312  1.320       dsl 	 * onto which we are attempting to mount.
    313  1.320       dsl 	 */
    314  1.332     pooka 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 ||
    315  1.320       dsl 	    (va.va_uid != kauth_cred_geteuid(l->l_cred) &&
    316  1.320       dsl 	    (error = kauth_authorize_generic(l->l_cred,
    317  1.320       dsl 	    KAUTH_GENERIC_ISSUSER, NULL)) != 0)) {
    318  1.320       dsl 		return error;
    319  1.283      elad 	}
    320  1.283      elad 
    321  1.320       dsl 	if (flags & MNT_EXPORTED)
    322  1.320       dsl 		return EINVAL;
    323  1.320       dsl 
    324  1.320       dsl 	if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
    325  1.320       dsl 		return error;
    326  1.320       dsl 
    327  1.283      elad 	/*
    328  1.283      elad 	 * Check if a file-system is not already mounted on this vnode.
    329  1.283      elad 	 */
    330  1.320       dsl 	if (vp->v_mountedhere != NULL)
    331  1.320       dsl 		return EBUSY;
    332  1.283      elad 
    333  1.345        ad 	mp = kmem_zalloc(sizeof(*mp), KM_SLEEP);
    334  1.345        ad 	if (mp == NULL)
    335  1.345        ad 		return ENOMEM;
    336  1.283      elad 
    337  1.320       dsl 	mp->mnt_op = vfsops;
    338  1.345        ad 	mp->mnt_refcnt = 1;
    339   1.31       cgd 
    340  1.296     pooka 	TAILQ_INIT(&mp->mnt_vnodelist);
    341  1.358        ad 	rw_init(&mp->mnt_unmounting);
    342  1.344  dholland  	mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE);
    343  1.358        ad 	mutex_init(&mp->mnt_updating, MUTEX_DEFAULT, IPL_NONE);
    344  1.358        ad 	error = vfs_busy(mp, NULL);
    345  1.358        ad 	KASSERT(error == 0);
    346  1.358        ad 	mutex_enter(&mp->mnt_updating);
    347  1.283      elad 
    348  1.129      fvdl 	mp->mnt_vnodecovered = vp;
    349  1.257        ad 	mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred);
    350  1.277   hannken 	mount_initspecific(mp);
    351  1.195   thorpej 
    352  1.195   thorpej 	/*
    353  1.195   thorpej 	 * The underlying file system may refuse the mount for
    354  1.198   thorpej 	 * various reasons.  Allow the user to force it to happen.
    355  1.286      yamt 	 *
    356  1.284      elad 	 * Set the mount level flags.
    357  1.284      elad 	 */
    358  1.286      yamt 	mp->mnt_flag = flags &
    359  1.286      yamt 	   (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV |
    360  1.284      elad 	    MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP |
    361  1.284      elad 	    MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP |
    362  1.370    simonb 	    MNT_LOG | MNT_IGNORE | MNT_RDONLY);
    363  1.284      elad 
    364  1.332     pooka 	error = VFS_MOUNT(mp, path, data, data_len);
    365  1.286      yamt 	mp->mnt_flag &= ~MNT_OP_FLAGS;
    366  1.232      jmmv 
    367   1.31       cgd 	/*
    368   1.31       cgd 	 * Put the new filesystem on the mount list after root.
    369   1.31       cgd 	 */
    370   1.31       cgd 	cache_purge(vp);
    371  1.320       dsl 	if (error != 0) {
    372  1.283      elad 		vp->v_mountedhere = NULL;
    373  1.358        ad 		mutex_exit(&mp->mnt_updating);
    374  1.354        ad 		vfs_unbusy(mp, false, NULL);
    375  1.358        ad 		vfs_destroy(mp);
    376  1.320       dsl 		return error;
    377   1.31       cgd 	}
    378  1.283      elad 
    379  1.320       dsl 	mp->mnt_iflag &= ~IMNT_WANTRDWR;
    380  1.345        ad 	mutex_enter(&mountlist_lock);
    381  1.320       dsl 	vp->v_mountedhere = mp;
    382  1.320       dsl 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
    383  1.329        ad 	mutex_exit(&mountlist_lock);
    384  1.337        ad     	vn_restorerecurse(vp, recurse);
    385  1.320       dsl 	VOP_UNLOCK(vp, 0);
    386  1.320       dsl 	checkdirs(vp);
    387  1.320       dsl 	if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
    388  1.320       dsl 		error = vfs_allocate_syncvnode(mp);
    389  1.347        ad 	/* Hold an additional reference to the mount across VFS_START(). */
    390  1.358        ad 	mutex_exit(&mp->mnt_updating);
    391  1.354        ad 	vfs_unbusy(mp, true, NULL);
    392  1.332     pooka 	(void) VFS_STATVFS(mp, &mp->mnt_stat);
    393  1.332     pooka 	error = VFS_START(mp, 0);
    394  1.372        ad 	if (error)
    395  1.320       dsl 		vrele(vp);
    396  1.347        ad 	/* Drop reference held for VFS_START(). */
    397  1.358        ad 	vfs_destroy(mp);
    398  1.320       dsl 	*vpp = NULL;
    399  1.320       dsl 	return error;
    400  1.283      elad }
    401  1.283      elad 
    402  1.283      elad static int
    403  1.283      elad mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags,
    404  1.324     pooka     void *data, size_t *data_len)
    405  1.283      elad {
    406  1.283      elad 	struct mount *mp;
    407  1.283      elad 	int error;
    408  1.283      elad 
    409  1.289      elad 	/* If MNT_GETARGS is specified, it should be the only flag. */
    410  1.320       dsl 	if (flags & ~MNT_GETARGS)
    411  1.320       dsl 		return EINVAL;
    412  1.289      elad 
    413  1.283      elad 	mp = vp->v_mount;
    414  1.283      elad 
    415  1.292      elad 	/* XXX: probably some notion of "can see" here if we want isolation. */
    416  1.292      elad 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
    417  1.292      elad 	    KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL);
    418  1.292      elad 	if (error)
    419  1.320       dsl 		return error;
    420  1.292      elad 
    421  1.329        ad 	if ((vp->v_vflag & VV_ROOT) == 0)
    422  1.320       dsl 		return EINVAL;
    423  1.283      elad 
    424  1.358        ad 	if (vfs_busy(mp, NULL))
    425  1.320       dsl 		return EPERM;
    426  1.283      elad 
    427  1.358        ad 	mutex_enter(&mp->mnt_updating);
    428  1.286      yamt 	mp->mnt_flag &= ~MNT_OP_FLAGS;
    429  1.286      yamt 	mp->mnt_flag |= MNT_GETARGS;
    430  1.332     pooka 	error = VFS_MOUNT(mp, path, data, data_len);
    431  1.286      yamt 	mp->mnt_flag &= ~MNT_OP_FLAGS;
    432  1.358        ad 	mutex_exit(&mp->mnt_updating);
    433  1.283      elad 
    434  1.354        ad 	vfs_unbusy(mp, false, NULL);
    435  1.283      elad 	return (error);
    436  1.283      elad }
    437  1.283      elad 
    438  1.321       dsl int
    439  1.335       dsl sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval)
    440  1.321       dsl {
    441  1.335       dsl 	/* {
    442  1.321       dsl 		syscallarg(const char *) type;
    443  1.321       dsl 		syscallarg(const char *) path;
    444  1.321       dsl 		syscallarg(int) flags;
    445  1.321       dsl 		syscallarg(void *) data;
    446  1.321       dsl 		syscallarg(size_t) data_len;
    447  1.335       dsl 	} */
    448  1.321       dsl 
    449  1.321       dsl 	return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path),
    450  1.321       dsl 	    SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE,
    451  1.321       dsl 	    SCARG(uap, data_len), retval);
    452  1.321       dsl }
    453  1.320       dsl 
    454  1.320       dsl int
    455  1.320       dsl do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type,
    456  1.320       dsl     const char *path, int flags, void *data, enum uio_seg data_seg,
    457  1.320       dsl     size_t data_len, register_t *retval)
    458  1.320       dsl {
    459  1.283      elad 	struct vnode *vp;
    460  1.283      elad 	struct nameidata nd;
    461  1.320       dsl 	void *data_buf = data;
    462  1.337        ad 	u_int recurse;
    463  1.283      elad 	int error;
    464  1.283      elad 
    465  1.283      elad 	/*
    466  1.283      elad 	 * Get vnode to be covered
    467  1.283      elad 	 */
    468  1.334     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
    469  1.283      elad 	if ((error = namei(&nd)) != 0)
    470  1.283      elad 		return (error);
    471  1.283      elad 	vp = nd.ni_vp;
    472  1.283      elad 
    473  1.283      elad 	/*
    474  1.283      elad 	 * A lookup in VFS_MOUNT might result in an attempt to
    475  1.283      elad 	 * lock this vnode again, so make the lock recursive.
    476  1.283      elad 	 */
    477  1.320       dsl 	if (vfsops == NULL) {
    478  1.361        ad 		if (flags & (MNT_GETARGS | MNT_UPDATE)) {
    479  1.361        ad 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    480  1.361        ad 			recurse = vn_setrecurse(vp);
    481  1.320       dsl 			vfsops = vp->v_mount->mnt_op;
    482  1.361        ad 		} else {
    483  1.320       dsl 			/* 'type' is userspace */
    484  1.320       dsl 			error = mount_get_vfsops(type, &vfsops);
    485  1.361        ad 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    486  1.361        ad 			recurse = vn_setrecurse(vp);
    487  1.320       dsl 			if (error != 0)
    488  1.320       dsl 				goto done;
    489  1.320       dsl 		}
    490  1.361        ad 	} else {
    491  1.361        ad 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    492  1.361        ad 		recurse = vn_setrecurse(vp);
    493  1.320       dsl 	}
    494  1.320       dsl 
    495  1.320       dsl 	if (data != NULL && data_seg == UIO_USERSPACE) {
    496  1.320       dsl 		if (data_len == 0) {
    497  1.320       dsl 			/* No length supplied, use default for filesystem */
    498  1.320       dsl 			data_len = vfsops->vfs_min_mount_data;
    499  1.320       dsl 			if (data_len > VFS_MAX_MOUNT_DATA) {
    500  1.320       dsl 				error = EINVAL;
    501  1.320       dsl 				goto done;
    502  1.320       dsl 			}
    503  1.378        ad 			/*
    504  1.378        ad 			 * Hopefully a longer buffer won't make copyin() fail.
    505  1.378        ad 			 * For compatibility with 3.0 and earlier.
    506  1.378        ad 			 */
    507  1.320       dsl 			if (flags & MNT_UPDATE
    508  1.320       dsl 			    && data_len < sizeof (struct mnt_export_args30))
    509  1.320       dsl 				data_len = sizeof (struct mnt_export_args30);
    510  1.320       dsl 		}
    511  1.320       dsl 		data_buf = malloc(data_len, M_TEMP, M_WAITOK);
    512  1.283      elad 
    513  1.320       dsl 		/* NFS needs the buffer even for mnt_getargs .... */
    514  1.320       dsl 		error = copyin(data, data_buf, data_len);
    515  1.320       dsl 		if (error != 0)
    516  1.320       dsl 			goto done;
    517  1.320       dsl 	}
    518  1.320       dsl 
    519  1.320       dsl 	if (flags & MNT_GETARGS) {
    520  1.320       dsl 		if (data_len == 0) {
    521  1.320       dsl 			error = EINVAL;
    522  1.320       dsl 			goto done;
    523  1.320       dsl 		}
    524  1.324     pooka 		error = mount_getargs(l, vp, path, flags, data_buf, &data_len);
    525  1.320       dsl 		if (error != 0)
    526  1.320       dsl 			goto done;
    527  1.320       dsl 		if (data_seg == UIO_USERSPACE)
    528  1.320       dsl 			error = copyout(data_buf, data, data_len);
    529  1.320       dsl 		*retval = data_len;
    530  1.320       dsl 	} else if (flags & MNT_UPDATE) {
    531  1.324     pooka 		error = mount_update(l, vp, path, flags, data_buf, &data_len);
    532  1.283      elad 	} else {
    533  1.283      elad 		/* Locking is handled internally in mount_domount(). */
    534  1.320       dsl 		error = mount_domount(l, &vp, vfsops, path, flags, data_buf,
    535  1.337        ad 		    &data_len, recurse);
    536  1.283      elad 	}
    537  1.283      elad 
    538  1.320       dsl     done:
    539  1.337        ad     	if (vp != NULL) {
    540  1.337        ad 	    	vn_restorerecurse(vp, recurse);
    541  1.337        ad 	    	vput(vp);
    542  1.337        ad 	}
    543  1.320       dsl 	if (data_buf != data)
    544  1.320       dsl 		free(data_buf, M_TEMP);
    545   1.31       cgd 	return (error);
    546   1.31       cgd }
    547   1.31       cgd 
    548   1.31       cgd /*
    549   1.43   mycroft  * Scan all active processes to see if any of them have a current
    550   1.43   mycroft  * or root directory onto which the new filesystem has just been
    551   1.43   mycroft  * mounted. If so, replace them with the new mount point.
    552   1.43   mycroft  */
    553   1.63  christos void
    554  1.221   thorpej checkdirs(struct vnode *olddp)
    555   1.43   mycroft {
    556  1.134   thorpej 	struct cwdinfo *cwdi;
    557  1.355        ad 	struct vnode *newdp, *rele1, *rele2;
    558   1.43   mycroft 	struct proc *p;
    559  1.355        ad 	bool retry;
    560   1.43   mycroft 
    561   1.43   mycroft 	if (olddp->v_usecount == 1)
    562   1.43   mycroft 		return;
    563  1.189   thorpej 	if (VFS_ROOT(olddp->v_mountedhere, &newdp))
    564   1.43   mycroft 		panic("mount: lost mount");
    565  1.355        ad 
    566  1.355        ad 	do {
    567  1.355        ad 		retry = false;
    568  1.355        ad 		mutex_enter(proc_lock);
    569  1.355        ad 		PROCLIST_FOREACH(p, &allproc) {
    570  1.355        ad 			if ((p->p_flag & PK_MARKER) != 0)
    571  1.355        ad 				continue;
    572  1.355        ad 			if ((cwdi = p->p_cwdi) == NULL)
    573  1.355        ad 				continue;
    574  1.355        ad 			/*
    575  1.355        ad 			 * Can't change to the old directory any more,
    576  1.355        ad 			 * so even if we see a stale value it's not a
    577  1.355        ad 			 * problem.
    578  1.355        ad 			 */
    579  1.355        ad 			if (cwdi->cwdi_cdir != olddp &&
    580  1.355        ad 			    cwdi->cwdi_rdir != olddp)
    581  1.355        ad 			    	continue;
    582  1.355        ad 			retry = true;
    583  1.355        ad 			rele1 = NULL;
    584  1.355        ad 			rele2 = NULL;
    585  1.355        ad 			atomic_inc_uint(&cwdi->cwdi_refcnt);
    586  1.355        ad 			mutex_exit(proc_lock);
    587  1.355        ad 			rw_enter(&cwdi->cwdi_lock, RW_WRITER);
    588  1.355        ad 			if (cwdi->cwdi_cdir == olddp) {
    589  1.355        ad 				rele1 = cwdi->cwdi_cdir;
    590  1.355        ad 				VREF(newdp);
    591  1.355        ad 				cwdi->cwdi_cdir = newdp;
    592  1.355        ad 			}
    593  1.355        ad 			if (cwdi->cwdi_rdir == olddp) {
    594  1.355        ad 				rele2 = cwdi->cwdi_rdir;
    595  1.355        ad 				VREF(newdp);
    596  1.355        ad 				cwdi->cwdi_rdir = newdp;
    597  1.355        ad 			}
    598  1.355        ad 			rw_exit(&cwdi->cwdi_lock);
    599  1.355        ad 			cwdfree(cwdi);
    600  1.355        ad 			if (rele1 != NULL)
    601  1.355        ad 				vrele(rele1);
    602  1.355        ad 			if (rele2 != NULL)
    603  1.355        ad 				vrele(rele2);
    604  1.355        ad 			mutex_enter(proc_lock);
    605  1.355        ad 			break;
    606   1.43   mycroft 		}
    607  1.355        ad 		mutex_exit(proc_lock);
    608  1.355        ad 	} while (retry);
    609  1.355        ad 
    610   1.43   mycroft 	if (rootvnode == olddp) {
    611   1.43   mycroft 		vrele(rootvnode);
    612   1.43   mycroft 		VREF(newdp);
    613   1.43   mycroft 		rootvnode = newdp;
    614   1.43   mycroft 	}
    615   1.43   mycroft 	vput(newdp);
    616   1.43   mycroft }
    617   1.43   mycroft 
    618   1.43   mycroft /*
    619   1.31       cgd  * Unmount a file system.
    620   1.31       cgd  *
    621   1.31       cgd  * Note: unmount takes a path to the vnode mounted on as argument,
    622   1.31       cgd  * not special file (as before).
    623   1.31       cgd  */
    624   1.31       cgd /* ARGSUSED */
    625   1.63  christos int
    626  1.335       dsl sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval)
    627   1.56   thorpej {
    628  1.335       dsl 	/* {
    629   1.74       cgd 		syscallarg(const char *) path;
    630   1.35       cgd 		syscallarg(int) flags;
    631  1.335       dsl 	} */
    632  1.155  augustss 	struct vnode *vp;
    633   1.31       cgd 	struct mount *mp;
    634   1.31       cgd 	int error;
    635   1.31       cgd 	struct nameidata nd;
    636   1.31       cgd 
    637  1.310       dsl 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
    638  1.334     pooka 	    SCARG(uap, path));
    639   1.63  christos 	if ((error = namei(&nd)) != 0)
    640   1.31       cgd 		return (error);
    641   1.31       cgd 	vp = nd.ni_vp;
    642   1.43   mycroft 	mp = vp->v_mount;
    643  1.358        ad 	atomic_inc_uint(&mp->mnt_refcnt);
    644  1.345        ad 	VOP_UNLOCK(vp, 0);
    645   1.31       cgd 
    646  1.295      elad 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
    647  1.295      elad 	    KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL);
    648  1.295      elad 	if (error) {
    649  1.345        ad 		vrele(vp);
    650  1.358        ad 		vfs_destroy(mp);
    651   1.31       cgd 		return (error);
    652   1.31       cgd 	}
    653   1.31       cgd 
    654   1.31       cgd 	/*
    655   1.47   mycroft 	 * Don't allow unmounting the root file system.
    656   1.47   mycroft 	 */
    657   1.47   mycroft 	if (mp->mnt_flag & MNT_ROOTFS) {
    658  1.345        ad 		vrele(vp);
    659  1.358        ad 		vfs_destroy(mp);
    660   1.47   mycroft 		return (EINVAL);
    661   1.47   mycroft 	}
    662   1.47   mycroft 
    663   1.47   mycroft 	/*
    664   1.31       cgd 	 * Must be the root of the filesystem
    665   1.31       cgd 	 */
    666  1.329        ad 	if ((vp->v_vflag & VV_ROOT) == 0) {
    667  1.345        ad 		vrele(vp);
    668  1.358        ad 		vfs_destroy(mp);
    669   1.31       cgd 		return (EINVAL);
    670   1.31       cgd 	}
    671   1.78      fvdl 
    672  1.359        ad 	vrele(vp);
    673  1.358        ad 	error = dounmount(mp, SCARG(uap, flags), l);
    674  1.374        ad 	vfs_destroy(mp);
    675  1.358        ad 	return error;
    676   1.31       cgd }
    677   1.31       cgd 
    678   1.31       cgd /*
    679  1.358        ad  * Do the actual file system unmount.  File system is assumed to have
    680  1.358        ad  * been locked by the caller.
    681  1.358        ad  *
    682  1.375       wiz  * => Caller hold reference to the mount, explicitly for dounmount().
    683   1.31       cgd  */
    684   1.63  christos int
    685  1.234  christos dounmount(struct mount *mp, int flags, struct lwp *l)
    686   1.31       cgd {
    687   1.31       cgd 	struct vnode *coveredvp;
    688   1.31       cgd 	int error;
    689  1.140  sommerfe 	int async;
    690  1.162      fvdl 	int used_syncer;
    691   1.31       cgd 
    692  1.256      elad #if NVERIEXEC > 0
    693  1.279      elad 	error = veriexec_unmountchk(mp);
    694  1.279      elad 	if (error)
    695  1.279      elad 		return (error);
    696  1.256      elad #endif /* NVERIEXEC > 0 */
    697  1.251      elad 
    698  1.358        ad 	/*
    699  1.358        ad 	 * XXX Freeze syncer.  Must do this before locking the
    700  1.358        ad 	 * mount point.  See dounmount() for details.
    701  1.358        ad 	 */
    702  1.358        ad 	mutex_enter(&syncer_mutex);
    703  1.358        ad 	rw_enter(&mp->mnt_unmounting, RW_WRITER);
    704  1.358        ad 	if ((mp->mnt_iflag & IMNT_GONE) != 0) {
    705  1.358        ad 		rw_exit(&mp->mnt_unmounting);
    706  1.358        ad 		mutex_exit(&syncer_mutex);
    707  1.358        ad 		return ENOENT;
    708  1.358        ad 	}
    709  1.358        ad 
    710  1.162      fvdl 	used_syncer = (mp->mnt_syncer != NULL);
    711  1.162      fvdl 
    712  1.148      fvdl 	/*
    713  1.165   thorpej 	 * XXX Syncer must be frozen when we get here.  This should really
    714  1.165   thorpej 	 * be done on a per-mountpoint basis, but especially the softdep
    715  1.227      jmmv 	 * code possibly called from the syncer doesn't exactly work on a
    716  1.165   thorpej 	 * per-mountpoint basis, so the softdep code would become a maze
    717  1.165   thorpej 	 * of vfs_busy() calls.
    718  1.165   thorpej 	 *
    719  1.300        ad 	 * The caller of dounmount() must acquire syncer_mutex because
    720  1.300        ad 	 * the syncer itself acquires locks in syncer_mutex -> vfs_busy
    721  1.165   thorpej 	 * order, and we must preserve that order to avoid deadlock.
    722  1.165   thorpej 	 *
    723  1.165   thorpej 	 * So, if the file system did not use the syncer, now is
    724  1.300        ad 	 * the time to release the syncer_mutex.
    725  1.148      fvdl 	 */
    726  1.165   thorpej 	if (used_syncer == 0)
    727  1.300        ad 		mutex_exit(&syncer_mutex);
    728  1.148      fvdl 
    729  1.196       dbj 	mp->mnt_iflag |= IMNT_UNMOUNT;
    730  1.141  sommerfe 	async = mp->mnt_flag & MNT_ASYNC;
    731  1.151   mycroft 	mp->mnt_flag &= ~MNT_ASYNC;
    732   1.31       cgd 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
    733  1.160   mycroft 	if (mp->mnt_syncer != NULL)
    734  1.160   mycroft 		vfs_deallocate_syncvnode(mp);
    735  1.209   hannken 	error = 0;
    736  1.209   hannken 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
    737  1.371   hannken 		error = VFS_SYNC(mp, MNT_WAIT, l->l_cred);
    738  1.209   hannken 	}
    739  1.342        ad 	vfs_scrubvnlist(mp);
    740  1.209   hannken 	if (error == 0 || (flags & MNT_FORCE))
    741  1.332     pooka 		error = VFS_UNMOUNT(mp, flags);
    742   1.31       cgd 	if (error) {
    743  1.151   mycroft 		if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0)
    744  1.148      fvdl 			(void) vfs_allocate_syncvnode(mp);
    745  1.196       dbj 		mp->mnt_iflag &= ~IMNT_UNMOUNT;
    746  1.140  sommerfe 		mp->mnt_flag |= async;
    747  1.358        ad 		rw_exit(&mp->mnt_unmounting);
    748  1.162      fvdl 		if (used_syncer)
    749  1.300        ad 			mutex_exit(&syncer_mutex);
    750  1.142  sommerfe 		return (error);
    751  1.113      fvdl 	}
    752  1.338        ad 	vfs_scrubvnlist(mp);
    753  1.358        ad 	mutex_enter(&mountlist_lock);
    754  1.297   hannken 	if ((coveredvp = mp->mnt_vnodecovered) != NULLVP)
    755  1.113      fvdl 		coveredvp->v_mountedhere = NULL;
    756  1.358        ad 	CIRCLEQ_REMOVE(&mountlist, mp, mnt_list);
    757  1.358        ad 	mp->mnt_iflag |= IMNT_GONE;
    758  1.358        ad 	mutex_exit(&mountlist_lock);
    759  1.273   reinoud 	if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL)
    760  1.113      fvdl 		panic("unmount: dangling vnode");
    761  1.162      fvdl 	if (used_syncer)
    762  1.300        ad 		mutex_exit(&syncer_mutex);
    763  1.231      jmmv 	vfs_hooks_unmount(mp);
    764  1.358        ad 	rw_exit(&mp->mnt_unmounting);
    765  1.374        ad 	vfs_destroy(mp);	/* reference from mount() */
    766  1.345        ad 	if (coveredvp != NULLVP)
    767  1.345        ad 		vrele(coveredvp);
    768  1.113      fvdl 	return (0);
    769   1.31       cgd }
    770   1.31       cgd 
    771   1.31       cgd /*
    772   1.31       cgd  * Sync each mounted filesystem.
    773   1.31       cgd  */
    774   1.31       cgd #ifdef DEBUG
    775   1.31       cgd int syncprt = 0;
    776   1.31       cgd struct ctldebug debug0 = { "syncprt", &syncprt };
    777   1.31       cgd #endif
    778   1.31       cgd 
    779   1.31       cgd /* ARGSUSED */
    780   1.63  christos int
    781  1.335       dsl sys_sync(struct lwp *l, const void *v, register_t *retval)
    782   1.31       cgd {
    783  1.155  augustss 	struct mount *mp, *nmp;
    784   1.31       cgd 	int asyncflag;
    785  1.257        ad 
    786  1.257        ad 	if (l == NULL)
    787  1.257        ad 		l = &lwp0;
    788   1.31       cgd 
    789  1.329        ad 	mutex_enter(&mountlist_lock);
    790  1.353        ad 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
    791  1.353        ad 	     mp = nmp) {
    792  1.358        ad 		if (vfs_busy(mp, &nmp)) {
    793  1.113      fvdl 			continue;
    794  1.113      fvdl 		}
    795  1.358        ad 		mutex_enter(&mp->mnt_updating);
    796  1.307   hannken 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
    797   1.31       cgd 			asyncflag = mp->mnt_flag & MNT_ASYNC;
    798   1.31       cgd 			mp->mnt_flag &= ~MNT_ASYNC;
    799  1.332     pooka 			VFS_SYNC(mp, MNT_NOWAIT, l->l_cred);
    800   1.31       cgd 			if (asyncflag)
    801  1.113      fvdl 				 mp->mnt_flag |= MNT_ASYNC;
    802   1.31       cgd 		}
    803  1.358        ad 		mutex_exit(&mp->mnt_updating);
    804  1.354        ad 		vfs_unbusy(mp, false, &nmp);
    805   1.31       cgd 	}
    806  1.329        ad 	mutex_exit(&mountlist_lock);
    807   1.31       cgd #ifdef DEBUG
    808   1.31       cgd 	if (syncprt)
    809   1.31       cgd 		vfs_bufstats();
    810   1.31       cgd #endif /* DEBUG */
    811   1.31       cgd 	return (0);
    812   1.31       cgd }
    813   1.31       cgd 
    814   1.31       cgd /*
    815   1.31       cgd  * Change filesystem quotas.
    816   1.31       cgd  */
    817   1.31       cgd /* ARGSUSED */
    818   1.63  christos int
    819  1.335       dsl sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval)
    820   1.56   thorpej {
    821  1.335       dsl 	/* {
    822   1.74       cgd 		syscallarg(const char *) path;
    823   1.35       cgd 		syscallarg(int) cmd;
    824   1.35       cgd 		syscallarg(int) uid;
    825  1.272  christos 		syscallarg(void *) arg;
    826  1.335       dsl 	} */
    827  1.155  augustss 	struct mount *mp;
    828   1.31       cgd 	int error;
    829   1.31       cgd 	struct nameidata nd;
    830   1.31       cgd 
    831  1.326     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
    832  1.334     pooka 	    SCARG(uap, path));
    833   1.63  christos 	if ((error = namei(&nd)) != 0)
    834   1.31       cgd 		return (error);
    835  1.307   hannken 	mp = nd.ni_vp->v_mount;
    836  1.197   hannken 	error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid),
    837  1.332     pooka 	    SCARG(uap, arg));
    838  1.326     pooka 	vrele(nd.ni_vp);
    839  1.197   hannken 	return (error);
    840   1.31       cgd }
    841   1.31       cgd 
    842  1.206  christos int
    843  1.234  christos dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags,
    844  1.185  christos     int root)
    845  1.185  christos {
    846  1.234  christos 	struct cwdinfo *cwdi = l->l_proc->p_cwdi;
    847  1.185  christos 	int error = 0;
    848  1.185  christos 
    849  1.185  christos 	/*
    850  1.185  christos 	 * If MNT_NOWAIT or MNT_LAZY is specified, do not
    851  1.204       dbj 	 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY
    852  1.185  christos 	 * overrides MNT_NOWAIT.
    853  1.185  christos 	 */
    854  1.185  christos 	if (flags == MNT_NOWAIT	|| flags == MNT_LAZY ||
    855  1.185  christos 	    (flags != MNT_WAIT && flags != 0)) {
    856  1.185  christos 		memcpy(sp, &mp->mnt_stat, sizeof(*sp));
    857  1.185  christos 		goto done;
    858  1.185  christos 	}
    859  1.205  junyoung 
    860  1.211  jdolecek 	/* Get the filesystem stats now */
    861  1.211  jdolecek 	memset(sp, 0, sizeof(*sp));
    862  1.332     pooka 	if ((error = VFS_STATVFS(mp, sp)) != 0) {
    863  1.185  christos 		return error;
    864  1.185  christos 	}
    865  1.185  christos 
    866  1.185  christos 	if (cwdi->cwdi_rdir == NULL)
    867  1.185  christos 		(void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat));
    868  1.185  christos done:
    869  1.185  christos 	if (cwdi->cwdi_rdir != NULL) {
    870  1.185  christos 		size_t len;
    871  1.185  christos 		char *bp;
    872  1.364  christos 		char c;
    873  1.236      yamt 		char *path = PNBUF_GET();
    874  1.185  christos 
    875  1.185  christos 		bp = path + MAXPATHLEN;
    876  1.185  christos 		*--bp = '\0';
    877  1.328        ad 		rw_enter(&cwdi->cwdi_lock, RW_READER);
    878  1.185  christos 		error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path,
    879  1.234  christos 		    MAXPATHLEN / 2, 0, l);
    880  1.328        ad 		rw_exit(&cwdi->cwdi_lock);
    881  1.185  christos 		if (error) {
    882  1.236      yamt 			PNBUF_PUT(path);
    883  1.185  christos 			return error;
    884  1.185  christos 		}
    885  1.185  christos 		len = strlen(bp);
    886  1.185  christos 		/*
    887  1.185  christos 		 * for mount points that are below our root, we can see
    888  1.185  christos 		 * them, so we fix up the pathname and return them. The
    889  1.185  christos 		 * rest we cannot see, so we don't allow viewing the
    890  1.185  christos 		 * data.
    891  1.185  christos 		 */
    892  1.364  christos 		if (strncmp(bp, sp->f_mntonname, len) == 0 &&
    893  1.364  christos 		    ((c = sp->f_mntonname[len]) == '/' || c == '\0')) {
    894  1.364  christos 			(void)strlcpy(sp->f_mntonname, &sp->f_mntonname[len],
    895  1.187    itojun 			    sizeof(sp->f_mntonname));
    896  1.185  christos 			if (sp->f_mntonname[0] == '\0')
    897  1.187    itojun 				(void)strlcpy(sp->f_mntonname, "/",
    898  1.187    itojun 				    sizeof(sp->f_mntonname));
    899  1.185  christos 		} else {
    900  1.185  christos 			if (root)
    901  1.187    itojun 				(void)strlcpy(sp->f_mntonname, "/",
    902  1.187    itojun 				    sizeof(sp->f_mntonname));
    903  1.185  christos 			else
    904  1.185  christos 				error = EPERM;
    905  1.185  christos 		}
    906  1.236      yamt 		PNBUF_PUT(path);
    907  1.185  christos 	}
    908  1.206  christos 	sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK;
    909  1.185  christos 	return error;
    910  1.185  christos }
    911  1.185  christos 
    912   1.31       cgd /*
    913  1.311       dsl  * Get filesystem statistics by path.
    914   1.31       cgd  */
    915  1.311       dsl int
    916  1.311       dsl do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb)
    917  1.311       dsl {
    918  1.311       dsl 	struct mount *mp;
    919  1.311       dsl 	int error;
    920  1.311       dsl 	struct nameidata nd;
    921  1.311       dsl 
    922  1.334     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path);
    923  1.311       dsl 	if ((error = namei(&nd)) != 0)
    924  1.311       dsl 		return error;
    925  1.311       dsl 	mp = nd.ni_vp->v_mount;
    926  1.311       dsl 	error = dostatvfs(mp, sb, l, flags, 1);
    927  1.311       dsl 	vrele(nd.ni_vp);
    928  1.311       dsl 	return error;
    929  1.311       dsl }
    930  1.311       dsl 
    931   1.31       cgd /* ARGSUSED */
    932   1.63  christos int
    933  1.335       dsl sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval)
    934   1.56   thorpej {
    935  1.335       dsl 	/* {
    936   1.74       cgd 		syscallarg(const char *) path;
    937  1.206  christos 		syscallarg(struct statvfs *) buf;
    938  1.206  christos 		syscallarg(int) flags;
    939  1.335       dsl 	} */
    940  1.241      yamt 	struct statvfs *sb;
    941   1.31       cgd 	int error;
    942   1.31       cgd 
    943  1.241      yamt 	sb = STATVFSBUF_GET();
    944  1.311       dsl 	error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb);
    945  1.311       dsl 	if (error == 0)
    946  1.241      yamt 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
    947  1.241      yamt 	STATVFSBUF_PUT(sb);
    948  1.241      yamt 	return error;
    949   1.31       cgd }
    950   1.31       cgd 
    951   1.31       cgd /*
    952  1.311       dsl  * Get filesystem statistics by fd.
    953   1.31       cgd  */
    954  1.311       dsl int
    955  1.311       dsl do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb)
    956  1.311       dsl {
    957  1.346        ad 	file_t *fp;
    958  1.311       dsl 	struct mount *mp;
    959  1.311       dsl 	int error;
    960  1.311       dsl 
    961  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
    962  1.346        ad 	if ((error = fd_getvnode(fd, &fp)) != 0)
    963  1.311       dsl 		return (error);
    964  1.311       dsl 	mp = ((struct vnode *)fp->f_data)->v_mount;
    965  1.346        ad 	error = dostatvfs(mp, sb, curlwp, flags, 1);
    966  1.346        ad 	fd_putfile(fd);
    967  1.311       dsl 	return error;
    968  1.311       dsl }
    969  1.311       dsl 
    970   1.31       cgd /* ARGSUSED */
    971   1.63  christos int
    972  1.335       dsl sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval)
    973   1.56   thorpej {
    974  1.335       dsl 	/* {
    975   1.35       cgd 		syscallarg(int) fd;
    976  1.206  christos 		syscallarg(struct statvfs *) buf;
    977  1.206  christos 		syscallarg(int) flags;
    978  1.335       dsl 	} */
    979  1.241      yamt 	struct statvfs *sb;
    980   1.31       cgd 	int error;
    981   1.31       cgd 
    982  1.241      yamt 	sb = STATVFSBUF_GET();
    983  1.311       dsl 	error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb);
    984  1.316       dsl 	if (error == 0)
    985  1.311       dsl 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
    986  1.241      yamt 	STATVFSBUF_PUT(sb);
    987  1.185  christos 	return error;
    988   1.31       cgd }
    989   1.31       cgd 
    990  1.185  christos 
    991   1.31       cgd /*
    992   1.31       cgd  * Get statistics on all filesystems.
    993   1.31       cgd  */
    994   1.63  christos int
    995  1.311       dsl do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags,
    996  1.311       dsl     int (*copyfn)(const void *, void *, size_t), size_t entry_sz,
    997  1.311       dsl     register_t *retval)
    998   1.56   thorpej {
    999  1.185  christos 	int root = 0;
   1000  1.179   thorpej 	struct proc *p = l->l_proc;
   1001  1.155  augustss 	struct mount *mp, *nmp;
   1002  1.241      yamt 	struct statvfs *sb;
   1003  1.206  christos 	size_t count, maxcount;
   1004  1.206  christos 	int error = 0;
   1005   1.31       cgd 
   1006  1.241      yamt 	sb = STATVFSBUF_GET();
   1007  1.311       dsl 	maxcount = bufsize / entry_sz;
   1008  1.329        ad 	mutex_enter(&mountlist_lock);
   1009  1.113      fvdl 	count = 0;
   1010  1.176      matt 	for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
   1011  1.176      matt 	     mp = nmp) {
   1012  1.358        ad 		if (vfs_busy(mp, &nmp)) {
   1013  1.113      fvdl 			continue;
   1014  1.113      fvdl 		}
   1015  1.113      fvdl 		if (sfsp && count < maxcount) {
   1016  1.311       dsl 			error = dostatvfs(mp, sb, l, flags, 0);
   1017  1.185  christos 			if (error) {
   1018  1.354        ad 				vfs_unbusy(mp, false, &nmp);
   1019  1.365  christos 				error = 0;
   1020   1.31       cgd 				continue;
   1021  1.113      fvdl 			}
   1022  1.311       dsl 			error = copyfn(sb, sfsp, entry_sz);
   1023  1.133   mycroft 			if (error) {
   1024  1.354        ad 				vfs_unbusy(mp, false, NULL);
   1025  1.241      yamt 				goto out;
   1026  1.133   mycroft 			}
   1027  1.311       dsl 			sfsp = (char *)sfsp + entry_sz;
   1028  1.241      yamt 			root |= strcmp(sb->f_mntonname, "/") == 0;
   1029   1.31       cgd 		}
   1030   1.31       cgd 		count++;
   1031  1.354        ad 		vfs_unbusy(mp, false, &nmp);
   1032   1.31       cgd 	}
   1033  1.358        ad 	mutex_exit(&mountlist_lock);
   1034  1.331     pooka 
   1035  1.185  christos 	if (root == 0 && p->p_cwdi->cwdi_rdir) {
   1036  1.185  christos 		/*
   1037  1.185  christos 		 * fake a root entry
   1038  1.185  christos 		 */
   1039  1.331     pooka 		error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount,
   1040  1.331     pooka 		    sb, l, flags, 1);
   1041  1.311       dsl 		if (error != 0)
   1042  1.241      yamt 			goto out;
   1043  1.365  christos 		if (sfsp) {
   1044  1.311       dsl 			error = copyfn(sb, sfsp, entry_sz);
   1045  1.365  christos 			if (error != 0)
   1046  1.365  christos 				goto out;
   1047  1.365  christos 		}
   1048  1.185  christos 		count++;
   1049  1.185  christos 	}
   1050   1.31       cgd 	if (sfsp && count > maxcount)
   1051   1.31       cgd 		*retval = maxcount;
   1052   1.31       cgd 	else
   1053   1.31       cgd 		*retval = count;
   1054  1.241      yamt out:
   1055  1.241      yamt 	STATVFSBUF_PUT(sb);
   1056  1.185  christos 	return error;
   1057   1.31       cgd }
   1058   1.31       cgd 
   1059  1.311       dsl int
   1060  1.335       dsl sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval)
   1061  1.311       dsl {
   1062  1.335       dsl 	/* {
   1063  1.311       dsl 		syscallarg(struct statvfs *) buf;
   1064  1.311       dsl 		syscallarg(size_t) bufsize;
   1065  1.311       dsl 		syscallarg(int) flags;
   1066  1.335       dsl 	} */
   1067  1.311       dsl 
   1068  1.311       dsl 	return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize),
   1069  1.311       dsl 	    SCARG(uap, flags), copyout, sizeof (struct statvfs), retval);
   1070  1.311       dsl }
   1071  1.311       dsl 
   1072   1.31       cgd /*
   1073   1.31       cgd  * Change current working directory to a given file descriptor.
   1074   1.31       cgd  */
   1075   1.31       cgd /* ARGSUSED */
   1076   1.63  christos int
   1077  1.335       dsl sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval)
   1078   1.56   thorpej {
   1079  1.335       dsl 	/* {
   1080   1.35       cgd 		syscallarg(int) fd;
   1081  1.335       dsl 	} */
   1082  1.179   thorpej 	struct proc *p = l->l_proc;
   1083  1.328        ad 	struct cwdinfo *cwdi;
   1084   1.43   mycroft 	struct vnode *vp, *tdp;
   1085   1.43   mycroft 	struct mount *mp;
   1086  1.346        ad 	file_t *fp;
   1087  1.346        ad 	int error, fd;
   1088   1.31       cgd 
   1089  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   1090  1.346        ad 	fd = SCARG(uap, fd);
   1091  1.346        ad 	if ((error = fd_getvnode(fd, &fp)) != 0)
   1092   1.31       cgd 		return (error);
   1093  1.346        ad 	vp = fp->f_data;
   1094  1.131  sommerfe 
   1095   1.43   mycroft 	VREF(vp);
   1096  1.113      fvdl 	vn_lock(vp,  LK_EXCLUSIVE | LK_RETRY);
   1097   1.31       cgd 	if (vp->v_type != VDIR)
   1098   1.31       cgd 		error = ENOTDIR;
   1099   1.31       cgd 	else
   1100  1.332     pooka 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
   1101  1.303     pooka 	if (error) {
   1102  1.303     pooka 		vput(vp);
   1103  1.303     pooka 		goto out;
   1104  1.303     pooka 	}
   1105  1.304     pooka 	while ((mp = vp->v_mountedhere) != NULL) {
   1106  1.358        ad 		error = vfs_busy(mp, NULL);
   1107  1.298       chs 		vput(vp);
   1108  1.357   xtraeme 		if (error != 0)
   1109  1.356        ad 			goto out;
   1110  1.189   thorpej 		error = VFS_ROOT(mp, &tdp);
   1111  1.354        ad 		vfs_unbusy(mp, false, NULL);
   1112  1.113      fvdl 		if (error)
   1113  1.303     pooka 			goto out;
   1114   1.43   mycroft 		vp = tdp;
   1115   1.43   mycroft 	}
   1116  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   1117  1.131  sommerfe 
   1118  1.131  sommerfe 	/*
   1119  1.131  sommerfe 	 * Disallow changing to a directory not under the process's
   1120  1.131  sommerfe 	 * current root directory (if there is one).
   1121  1.131  sommerfe 	 */
   1122  1.328        ad 	cwdi = p->p_cwdi;
   1123  1.328        ad 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
   1124  1.234  christos 	if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) {
   1125  1.131  sommerfe 		vrele(vp);
   1126  1.135   thorpej 		error = EPERM;	/* operation not permitted */
   1127  1.328        ad 	} else {
   1128  1.328        ad 		vrele(cwdi->cwdi_cdir);
   1129  1.328        ad 		cwdi->cwdi_cdir = vp;
   1130  1.131  sommerfe 	}
   1131  1.328        ad 	rw_exit(&cwdi->cwdi_lock);
   1132  1.205  junyoung 
   1133  1.135   thorpej  out:
   1134  1.346        ad 	fd_putfile(fd);
   1135  1.135   thorpej 	return (error);
   1136   1.31       cgd }
   1137   1.31       cgd 
   1138   1.31       cgd /*
   1139  1.221   thorpej  * Change this process's notion of the root directory to a given file
   1140  1.221   thorpej  * descriptor.
   1141  1.131  sommerfe  */
   1142  1.131  sommerfe int
   1143  1.335       dsl sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval)
   1144  1.131  sommerfe {
   1145  1.179   thorpej 	struct proc *p = l->l_proc;
   1146  1.328        ad 	struct cwdinfo *cwdi;
   1147  1.131  sommerfe 	struct vnode	*vp;
   1148  1.346        ad 	file_t	*fp;
   1149  1.346        ad 	int		 error, fd = SCARG(uap, fd);
   1150  1.131  sommerfe 
   1151  1.268      elad 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
   1152  1.268      elad  	    KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0)
   1153  1.131  sommerfe 		return error;
   1154  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   1155  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   1156  1.131  sommerfe 		return error;
   1157  1.346        ad 	vp = fp->f_data;
   1158  1.131  sommerfe 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   1159  1.131  sommerfe 	if (vp->v_type != VDIR)
   1160  1.131  sommerfe 		error = ENOTDIR;
   1161  1.131  sommerfe 	else
   1162  1.332     pooka 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
   1163  1.131  sommerfe 	VOP_UNLOCK(vp, 0);
   1164  1.131  sommerfe 	if (error)
   1165  1.135   thorpej 		goto out;
   1166  1.131  sommerfe 	VREF(vp);
   1167  1.131  sommerfe 
   1168  1.131  sommerfe 	/*
   1169  1.131  sommerfe 	 * Prevent escaping from chroot by putting the root under
   1170  1.131  sommerfe 	 * the working directory.  Silently chdir to / if we aren't
   1171  1.131  sommerfe 	 * already there.
   1172  1.131  sommerfe 	 */
   1173  1.328        ad 	cwdi = p->p_cwdi;
   1174  1.328        ad 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
   1175  1.234  christos 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
   1176  1.131  sommerfe 		/*
   1177  1.131  sommerfe 		 * XXX would be more failsafe to change directory to a
   1178  1.131  sommerfe 		 * deadfs node here instead
   1179  1.131  sommerfe 		 */
   1180  1.134   thorpej 		vrele(cwdi->cwdi_cdir);
   1181  1.131  sommerfe 		VREF(vp);
   1182  1.134   thorpej 		cwdi->cwdi_cdir = vp;
   1183  1.131  sommerfe 	}
   1184  1.205  junyoung 
   1185  1.134   thorpej 	if (cwdi->cwdi_rdir != NULL)
   1186  1.134   thorpej 		vrele(cwdi->cwdi_rdir);
   1187  1.134   thorpej 	cwdi->cwdi_rdir = vp;
   1188  1.328        ad 	rw_exit(&cwdi->cwdi_lock);
   1189  1.328        ad 
   1190  1.135   thorpej  out:
   1191  1.346        ad 	fd_putfile(fd);
   1192  1.135   thorpej 	return (error);
   1193  1.131  sommerfe }
   1194  1.131  sommerfe 
   1195  1.131  sommerfe /*
   1196   1.31       cgd  * Change current working directory (``.'').
   1197   1.31       cgd  */
   1198   1.31       cgd /* ARGSUSED */
   1199   1.63  christos int
   1200  1.335       dsl sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval)
   1201   1.56   thorpej {
   1202  1.335       dsl 	/* {
   1203   1.74       cgd 		syscallarg(const char *) path;
   1204  1.335       dsl 	} */
   1205  1.179   thorpej 	struct proc *p = l->l_proc;
   1206  1.328        ad 	struct cwdinfo *cwdi;
   1207   1.31       cgd 	int error;
   1208   1.31       cgd 	struct nameidata nd;
   1209   1.31       cgd 
   1210  1.310       dsl 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
   1211  1.334     pooka 	    SCARG(uap, path));
   1212  1.234  christos 	if ((error = change_dir(&nd, l)) != 0)
   1213   1.31       cgd 		return (error);
   1214  1.328        ad 	cwdi = p->p_cwdi;
   1215  1.328        ad 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
   1216  1.134   thorpej 	vrele(cwdi->cwdi_cdir);
   1217  1.134   thorpej 	cwdi->cwdi_cdir = nd.ni_vp;
   1218  1.328        ad 	rw_exit(&cwdi->cwdi_lock);
   1219   1.31       cgd 	return (0);
   1220   1.31       cgd }
   1221   1.31       cgd 
   1222   1.31       cgd /*
   1223   1.31       cgd  * Change notion of root (``/'') directory.
   1224   1.31       cgd  */
   1225   1.31       cgd /* ARGSUSED */
   1226   1.63  christos int
   1227  1.335       dsl sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval)
   1228   1.56   thorpej {
   1229  1.335       dsl 	/* {
   1230   1.74       cgd 		syscallarg(const char *) path;
   1231  1.335       dsl 	} */
   1232  1.179   thorpej 	struct proc *p = l->l_proc;
   1233  1.328        ad 	struct cwdinfo *cwdi;
   1234  1.131  sommerfe 	struct vnode *vp;
   1235   1.31       cgd 	int error;
   1236   1.31       cgd 	struct nameidata nd;
   1237   1.31       cgd 
   1238  1.268      elad 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT,
   1239  1.268      elad 	    KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0)
   1240   1.31       cgd 		return (error);
   1241  1.310       dsl 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
   1242  1.334     pooka 	    SCARG(uap, path));
   1243  1.234  christos 	if ((error = change_dir(&nd, l)) != 0)
   1244   1.31       cgd 		return (error);
   1245  1.328        ad 
   1246  1.328        ad 	cwdi = p->p_cwdi;
   1247  1.328        ad 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
   1248  1.134   thorpej 	if (cwdi->cwdi_rdir != NULL)
   1249  1.134   thorpej 		vrele(cwdi->cwdi_rdir);
   1250  1.131  sommerfe 	vp = nd.ni_vp;
   1251  1.134   thorpej 	cwdi->cwdi_rdir = vp;
   1252  1.131  sommerfe 
   1253  1.131  sommerfe 	/*
   1254  1.131  sommerfe 	 * Prevent escaping from chroot by putting the root under
   1255  1.131  sommerfe 	 * the working directory.  Silently chdir to / if we aren't
   1256  1.131  sommerfe 	 * already there.
   1257  1.131  sommerfe 	 */
   1258  1.234  christos 	if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) {
   1259  1.131  sommerfe 		/*
   1260  1.131  sommerfe 		 * XXX would be more failsafe to change directory to a
   1261  1.131  sommerfe 		 * deadfs node here instead
   1262  1.131  sommerfe 		 */
   1263  1.134   thorpej 		vrele(cwdi->cwdi_cdir);
   1264  1.131  sommerfe 		VREF(vp);
   1265  1.134   thorpej 		cwdi->cwdi_cdir = vp;
   1266  1.131  sommerfe 	}
   1267  1.328        ad 	rw_exit(&cwdi->cwdi_lock);
   1268  1.205  junyoung 
   1269   1.31       cgd 	return (0);
   1270   1.31       cgd }
   1271   1.31       cgd 
   1272   1.31       cgd /*
   1273   1.31       cgd  * Common routine for chroot and chdir.
   1274   1.31       cgd  */
   1275   1.31       cgd static int
   1276  1.234  christos change_dir(struct nameidata *ndp, struct lwp *l)
   1277   1.31       cgd {
   1278   1.31       cgd 	struct vnode *vp;
   1279   1.31       cgd 	int error;
   1280   1.31       cgd 
   1281   1.63  christos 	if ((error = namei(ndp)) != 0)
   1282   1.31       cgd 		return (error);
   1283   1.31       cgd 	vp = ndp->ni_vp;
   1284   1.31       cgd 	if (vp->v_type != VDIR)
   1285   1.31       cgd 		error = ENOTDIR;
   1286   1.31       cgd 	else
   1287  1.332     pooka 		error = VOP_ACCESS(vp, VEXEC, l->l_cred);
   1288  1.205  junyoung 
   1289   1.31       cgd 	if (error)
   1290  1.113      fvdl 		vput(vp);
   1291  1.113      fvdl 	else
   1292  1.113      fvdl 		VOP_UNLOCK(vp, 0);
   1293   1.31       cgd 	return (error);
   1294   1.31       cgd }
   1295   1.31       cgd 
   1296   1.31       cgd /*
   1297   1.31       cgd  * Check permissions, allocate an open file structure,
   1298   1.31       cgd  * and call the device open routine if any.
   1299   1.31       cgd  */
   1300   1.63  christos int
   1301  1.335       dsl sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval)
   1302   1.56   thorpej {
   1303  1.335       dsl 	/* {
   1304   1.74       cgd 		syscallarg(const char *) path;
   1305   1.35       cgd 		syscallarg(int) flags;
   1306   1.35       cgd 		syscallarg(int) mode;
   1307  1.335       dsl 	} */
   1308  1.179   thorpej 	struct proc *p = l->l_proc;
   1309  1.134   thorpej 	struct cwdinfo *cwdi = p->p_cwdi;
   1310  1.346        ad 	file_t *fp;
   1311  1.135   thorpej 	struct vnode *vp;
   1312   1.31       cgd 	int flags, cmode;
   1313   1.31       cgd 	int type, indx, error;
   1314   1.31       cgd 	struct flock lf;
   1315   1.31       cgd 	struct nameidata nd;
   1316   1.31       cgd 
   1317  1.104   mycroft 	flags = FFLAGS(SCARG(uap, flags));
   1318  1.104   mycroft 	if ((flags & (FREAD | FWRITE)) == 0)
   1319  1.104   mycroft 		return (EINVAL);
   1320  1.346        ad 	if ((error = fd_allocfile(&fp, &indx)) != 0)
   1321   1.31       cgd 		return (error);
   1322  1.328        ad 	/* We're going to read cwdi->cwdi_cmask unlocked here. */
   1323  1.134   thorpej 	cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT;
   1324  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   1325  1.334     pooka 	    SCARG(uap, path));
   1326  1.194  jdolecek 	l->l_dupfd = -indx - 1;			/* XXX check for fdopen */
   1327   1.63  christos 	if ((error = vn_open(&nd, flags, cmode)) != 0) {
   1328  1.346        ad 		fd_abort(p, fp, indx);
   1329  1.213  christos 		if ((error == EDUPFD || error == EMOVEFD) &&
   1330  1.194  jdolecek 		    l->l_dupfd >= 0 &&			/* XXX from fdopen */
   1331   1.45   mycroft 		    (error =
   1332  1.346        ad 			fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) {
   1333   1.45   mycroft 			*retval = indx;
   1334   1.45   mycroft 			return (0);
   1335   1.45   mycroft 		}
   1336   1.45   mycroft 		if (error == ERESTART)
   1337   1.44   mycroft 			error = EINTR;
   1338   1.31       cgd 		return (error);
   1339   1.31       cgd 	}
   1340  1.331     pooka 
   1341  1.194  jdolecek 	l->l_dupfd = 0;
   1342   1.31       cgd 	vp = nd.ni_vp;
   1343   1.31       cgd 	fp->f_flag = flags & FMASK;
   1344   1.31       cgd 	fp->f_type = DTYPE_VNODE;
   1345   1.31       cgd 	fp->f_ops = &vnops;
   1346  1.184       dsl 	fp->f_data = vp;
   1347   1.31       cgd 	if (flags & (O_EXLOCK | O_SHLOCK)) {
   1348   1.31       cgd 		lf.l_whence = SEEK_SET;
   1349   1.31       cgd 		lf.l_start = 0;
   1350   1.31       cgd 		lf.l_len = 0;
   1351   1.31       cgd 		if (flags & O_EXLOCK)
   1352   1.31       cgd 			lf.l_type = F_WRLCK;
   1353   1.31       cgd 		else
   1354   1.31       cgd 			lf.l_type = F_RDLCK;
   1355   1.31       cgd 		type = F_FLOCK;
   1356   1.31       cgd 		if ((flags & FNONBLOCK) == 0)
   1357   1.31       cgd 			type |= F_WAIT;
   1358  1.113      fvdl 		VOP_UNLOCK(vp, 0);
   1359  1.184       dsl 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
   1360   1.63  christos 		if (error) {
   1361  1.346        ad 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
   1362  1.346        ad 			fd_abort(p, fp, indx);
   1363   1.31       cgd 			return (error);
   1364   1.31       cgd 		}
   1365  1.113      fvdl 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   1366  1.346        ad 		atomic_or_uint(&fp->f_flag, FHASLOCK);
   1367   1.31       cgd 	}
   1368  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   1369   1.31       cgd 	*retval = indx;
   1370  1.346        ad 	fd_affix(p, fp, indx);
   1371   1.31       cgd 	return (0);
   1372  1.137  wrstuden }
   1373  1.137  wrstuden 
   1374  1.249      yamt static void
   1375  1.249      yamt vfs__fhfree(fhandle_t *fhp)
   1376  1.249      yamt {
   1377  1.249      yamt 	size_t fhsize;
   1378  1.249      yamt 
   1379  1.249      yamt 	if (fhp == NULL) {
   1380  1.249      yamt 		return;
   1381  1.249      yamt 	}
   1382  1.249      yamt 	fhsize = FHANDLE_SIZE(fhp);
   1383  1.249      yamt 	kmem_free(fhp, fhsize);
   1384  1.249      yamt }
   1385  1.249      yamt 
   1386  1.137  wrstuden /*
   1387  1.243      yamt  * vfs_composefh: compose a filehandle.
   1388  1.243      yamt  */
   1389  1.243      yamt 
   1390  1.243      yamt int
   1391  1.244    martin vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size)
   1392  1.243      yamt {
   1393  1.243      yamt 	struct mount *mp;
   1394  1.250      yamt 	struct fid *fidp;
   1395  1.243      yamt 	int error;
   1396  1.250      yamt 	size_t needfhsize;
   1397  1.250      yamt 	size_t fidsize;
   1398  1.243      yamt 
   1399  1.243      yamt 	mp = vp->v_mount;
   1400  1.250      yamt 	fidp = NULL;
   1401  1.252    martin 	if (*fh_size < FHANDLE_SIZE_MIN) {
   1402  1.250      yamt 		fidsize = 0;
   1403  1.244    martin 	} else {
   1404  1.250      yamt 		fidsize = *fh_size - offsetof(fhandle_t, fh_fid);
   1405  1.250      yamt 		if (fhp != NULL) {
   1406  1.250      yamt 			memset(fhp, 0, *fh_size);
   1407  1.250      yamt 			fhp->fh_fsid = mp->mnt_stat.f_fsidx;
   1408  1.250      yamt 			fidp = &fhp->fh_fid;
   1409  1.250      yamt 		}
   1410  1.244    martin 	}
   1411  1.250      yamt 	error = VFS_VPTOFH(vp, fidp, &fidsize);
   1412  1.250      yamt 	needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
   1413  1.250      yamt 	if (error == 0 && *fh_size < needfhsize) {
   1414  1.250      yamt 		error = E2BIG;
   1415  1.250      yamt 	}
   1416  1.250      yamt 	*fh_size = needfhsize;
   1417  1.243      yamt 	return error;
   1418  1.243      yamt }
   1419  1.243      yamt 
   1420  1.249      yamt int
   1421  1.249      yamt vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp)
   1422  1.249      yamt {
   1423  1.249      yamt 	struct mount *mp;
   1424  1.249      yamt 	fhandle_t *fhp;
   1425  1.249      yamt 	size_t fhsize;
   1426  1.249      yamt 	size_t fidsize;
   1427  1.249      yamt 	int error;
   1428  1.249      yamt 
   1429  1.249      yamt 	*fhpp = NULL;
   1430  1.249      yamt 	mp = vp->v_mount;
   1431  1.255  christos 	fidsize = 0;
   1432  1.249      yamt 	error = VFS_VPTOFH(vp, NULL, &fidsize);
   1433  1.249      yamt 	KASSERT(error != 0);
   1434  1.249      yamt 	if (error != E2BIG) {
   1435  1.249      yamt 		goto out;
   1436  1.249      yamt 	}
   1437  1.250      yamt 	fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize);
   1438  1.249      yamt 	fhp = kmem_zalloc(fhsize, KM_SLEEP);
   1439  1.249      yamt 	if (fhp == NULL) {
   1440  1.249      yamt 		error = ENOMEM;
   1441  1.249      yamt 		goto out;
   1442  1.249      yamt 	}
   1443  1.249      yamt 	fhp->fh_fsid = mp->mnt_stat.f_fsidx;
   1444  1.249      yamt 	error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize);
   1445  1.249      yamt 	if (error == 0) {
   1446  1.249      yamt 		KASSERT((FHANDLE_SIZE(fhp) == fhsize &&
   1447  1.249      yamt 		    FHANDLE_FILEID(fhp)->fid_len == fidsize));
   1448  1.249      yamt 		*fhpp = fhp;
   1449  1.249      yamt 	} else {
   1450  1.249      yamt 		kmem_free(fhp, fhsize);
   1451  1.249      yamt 	}
   1452  1.249      yamt out:
   1453  1.249      yamt 	return error;
   1454  1.249      yamt }
   1455  1.249      yamt 
   1456  1.249      yamt void
   1457  1.249      yamt vfs_composefh_free(fhandle_t *fhp)
   1458  1.249      yamt {
   1459  1.249      yamt 
   1460  1.249      yamt 	vfs__fhfree(fhp);
   1461  1.249      yamt }
   1462  1.249      yamt 
   1463  1.243      yamt /*
   1464  1.248      yamt  * vfs_fhtovp: lookup a vnode by a filehandle.
   1465  1.248      yamt  */
   1466  1.248      yamt 
   1467  1.248      yamt int
   1468  1.248      yamt vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp)
   1469  1.248      yamt {
   1470  1.248      yamt 	struct mount *mp;
   1471  1.248      yamt 	int error;
   1472  1.248      yamt 
   1473  1.248      yamt 	*vpp = NULL;
   1474  1.248      yamt 	mp = vfs_getvfs(FHANDLE_FSID(fhp));
   1475  1.248      yamt 	if (mp == NULL) {
   1476  1.248      yamt 		error = ESTALE;
   1477  1.248      yamt 		goto out;
   1478  1.248      yamt 	}
   1479  1.248      yamt 	if (mp->mnt_op->vfs_fhtovp == NULL) {
   1480  1.248      yamt 		error = EOPNOTSUPP;
   1481  1.248      yamt 		goto out;
   1482  1.248      yamt 	}
   1483  1.248      yamt 	error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp);
   1484  1.248      yamt out:
   1485  1.248      yamt 	return error;
   1486  1.248      yamt }
   1487  1.248      yamt 
   1488  1.248      yamt /*
   1489  1.265      yamt  * vfs_copyinfh_alloc: allocate and copyin a filehandle, given
   1490  1.263    martin  * the needed size.
   1491  1.263    martin  */
   1492  1.263    martin 
   1493  1.263    martin int
   1494  1.265      yamt vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp)
   1495  1.263    martin {
   1496  1.263    martin 	fhandle_t *fhp;
   1497  1.263    martin 	int error;
   1498  1.263    martin 
   1499  1.263    martin 	*fhpp = NULL;
   1500  1.250      yamt 	if (fhsize > FHANDLE_SIZE_MAX) {
   1501  1.250      yamt 		return EINVAL;
   1502  1.250      yamt 	}
   1503  1.265      yamt 	if (fhsize < FHANDLE_SIZE_MIN) {
   1504  1.265      yamt 		return EINVAL;
   1505  1.265      yamt 	}
   1506  1.267      yamt again:
   1507  1.248      yamt 	fhp = kmem_alloc(fhsize, KM_SLEEP);
   1508  1.248      yamt 	if (fhp == NULL) {
   1509  1.248      yamt 		return ENOMEM;
   1510  1.248      yamt 	}
   1511  1.248      yamt 	error = copyin(ufhp, fhp, fhsize);
   1512  1.248      yamt 	if (error == 0) {
   1513  1.265      yamt 		/* XXX this check shouldn't be here */
   1514  1.265      yamt 		if (FHANDLE_SIZE(fhp) == fhsize) {
   1515  1.263    martin 			*fhpp = fhp;
   1516  1.263    martin 			return 0;
   1517  1.267      yamt 		} else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) {
   1518  1.267      yamt 			/*
   1519  1.267      yamt 			 * a kludge for nfsv2 padded handles.
   1520  1.267      yamt 			 */
   1521  1.267      yamt 			size_t sz;
   1522  1.267      yamt 
   1523  1.267      yamt 			sz = FHANDLE_SIZE(fhp);
   1524  1.267      yamt 			kmem_free(fhp, fhsize);
   1525  1.267      yamt 			fhsize = sz;
   1526  1.267      yamt 			goto again;
   1527  1.264      yamt 		} else {
   1528  1.265      yamt 			/*
   1529  1.265      yamt 			 * userland told us wrong size.
   1530  1.265      yamt 			 */
   1531  1.263    martin 		    	error = EINVAL;
   1532  1.264      yamt 		}
   1533  1.248      yamt 	}
   1534  1.263    martin 	kmem_free(fhp, fhsize);
   1535  1.248      yamt 	return error;
   1536  1.248      yamt }
   1537  1.248      yamt 
   1538  1.248      yamt void
   1539  1.248      yamt vfs_copyinfh_free(fhandle_t *fhp)
   1540  1.248      yamt {
   1541  1.248      yamt 
   1542  1.249      yamt 	vfs__fhfree(fhp);
   1543  1.248      yamt }
   1544  1.248      yamt 
   1545  1.248      yamt /*
   1546  1.137  wrstuden  * Get file handle system call
   1547  1.137  wrstuden  */
   1548  1.137  wrstuden int
   1549  1.335       dsl sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval)
   1550  1.137  wrstuden {
   1551  1.335       dsl 	/* {
   1552  1.137  wrstuden 		syscallarg(char *) fname;
   1553  1.137  wrstuden 		syscallarg(fhandle_t *) fhp;
   1554  1.244    martin 		syscallarg(size_t *) fh_size;
   1555  1.335       dsl 	} */
   1556  1.155  augustss 	struct vnode *vp;
   1557  1.244    martin 	fhandle_t *fh;
   1558  1.137  wrstuden 	int error;
   1559  1.137  wrstuden 	struct nameidata nd;
   1560  1.244    martin 	size_t sz;
   1561  1.249      yamt 	size_t usz;
   1562  1.137  wrstuden 
   1563  1.137  wrstuden 	/*
   1564  1.137  wrstuden 	 * Must be super user
   1565  1.137  wrstuden 	 */
   1566  1.268      elad 	error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
   1567  1.268      elad 	    0, NULL, NULL, NULL);
   1568  1.137  wrstuden 	if (error)
   1569  1.137  wrstuden 		return (error);
   1570  1.310       dsl 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
   1571  1.334     pooka 	    SCARG(uap, fname));
   1572  1.137  wrstuden 	error = namei(&nd);
   1573  1.137  wrstuden 	if (error)
   1574  1.137  wrstuden 		return (error);
   1575  1.137  wrstuden 	vp = nd.ni_vp;
   1576  1.249      yamt 	error = vfs_composefh_alloc(vp, &fh);
   1577  1.247      yamt 	vput(vp);
   1578  1.249      yamt 	if (error != 0) {
   1579  1.249      yamt 		goto out;
   1580  1.249      yamt 	}
   1581  1.249      yamt 	error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t));
   1582  1.249      yamt 	if (error != 0) {
   1583  1.249      yamt 		goto out;
   1584  1.249      yamt 	}
   1585  1.249      yamt 	sz = FHANDLE_SIZE(fh);
   1586  1.249      yamt 	error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t));
   1587  1.249      yamt 	if (error != 0) {
   1588  1.249      yamt 		goto out;
   1589  1.244    martin 	}
   1590  1.249      yamt 	if (usz >= sz) {
   1591  1.249      yamt 		error = copyout(fh, SCARG(uap, fhp), sz);
   1592  1.249      yamt 	} else {
   1593  1.249      yamt 		error = E2BIG;
   1594  1.244    martin 	}
   1595  1.249      yamt out:
   1596  1.249      yamt 	vfs_composefh_free(fh);
   1597  1.137  wrstuden 	return (error);
   1598  1.137  wrstuden }
   1599  1.137  wrstuden 
   1600  1.137  wrstuden /*
   1601  1.137  wrstuden  * Open a file given a file handle.
   1602  1.137  wrstuden  *
   1603  1.137  wrstuden  * Check permissions, allocate an open file structure,
   1604  1.137  wrstuden  * and call the device open routine if any.
   1605  1.137  wrstuden  */
   1606  1.265      yamt 
   1607  1.137  wrstuden int
   1608  1.265      yamt dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags,
   1609  1.265      yamt     register_t *retval)
   1610  1.137  wrstuden {
   1611  1.346        ad 	file_t *fp;
   1612  1.139  wrstuden 	struct vnode *vp = NULL;
   1613  1.257        ad 	kauth_cred_t cred = l->l_cred;
   1614  1.346        ad 	file_t *nfp;
   1615  1.137  wrstuden 	int type, indx, error=0;
   1616  1.137  wrstuden 	struct flock lf;
   1617  1.137  wrstuden 	struct vattr va;
   1618  1.248      yamt 	fhandle_t *fh;
   1619  1.265      yamt 	int flags;
   1620  1.346        ad 	proc_t *p;
   1621  1.346        ad 
   1622  1.346        ad 	p = curproc;
   1623  1.137  wrstuden 
   1624  1.137  wrstuden 	/*
   1625  1.137  wrstuden 	 * Must be super user
   1626  1.137  wrstuden 	 */
   1627  1.269      elad 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
   1628  1.270      elad 	    0, NULL, NULL, NULL)))
   1629  1.137  wrstuden 		return (error);
   1630  1.137  wrstuden 
   1631  1.265      yamt 	flags = FFLAGS(oflags);
   1632  1.137  wrstuden 	if ((flags & (FREAD | FWRITE)) == 0)
   1633  1.137  wrstuden 		return (EINVAL);
   1634  1.137  wrstuden 	if ((flags & O_CREAT))
   1635  1.137  wrstuden 		return (EINVAL);
   1636  1.346        ad 	if ((error = fd_allocfile(&nfp, &indx)) != 0)
   1637  1.137  wrstuden 		return (error);
   1638  1.137  wrstuden 	fp = nfp;
   1639  1.265      yamt 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
   1640  1.248      yamt 	if (error != 0) {
   1641  1.139  wrstuden 		goto bad;
   1642  1.139  wrstuden 	}
   1643  1.248      yamt 	error = vfs_fhtovp(fh, &vp);
   1644  1.248      yamt 	if (error != 0) {
   1645  1.139  wrstuden 		goto bad;
   1646  1.139  wrstuden 	}
   1647  1.137  wrstuden 
   1648  1.137  wrstuden 	/* Now do an effective vn_open */
   1649  1.137  wrstuden 
   1650  1.137  wrstuden 	if (vp->v_type == VSOCK) {
   1651  1.137  wrstuden 		error = EOPNOTSUPP;
   1652  1.137  wrstuden 		goto bad;
   1653  1.137  wrstuden 	}
   1654  1.333      yamt 	error = vn_openchk(vp, cred, flags);
   1655  1.333      yamt 	if (error != 0)
   1656  1.333      yamt 		goto bad;
   1657  1.137  wrstuden 	if (flags & O_TRUNC) {
   1658  1.137  wrstuden 		VOP_UNLOCK(vp, 0);			/* XXX */
   1659  1.137  wrstuden 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);   /* XXX */
   1660  1.137  wrstuden 		VATTR_NULL(&va);
   1661  1.137  wrstuden 		va.va_size = 0;
   1662  1.332     pooka 		error = VOP_SETATTR(vp, &va, cred);
   1663  1.197   hannken 		if (error)
   1664  1.137  wrstuden 			goto bad;
   1665  1.137  wrstuden 	}
   1666  1.332     pooka 	if ((error = VOP_OPEN(vp, flags, cred)) != 0)
   1667  1.137  wrstuden 		goto bad;
   1668  1.346        ad 	if (flags & FWRITE) {
   1669  1.346        ad 		mutex_enter(&vp->v_interlock);
   1670  1.137  wrstuden 		vp->v_writecount++;
   1671  1.346        ad 		mutex_exit(&vp->v_interlock);
   1672  1.346        ad 	}
   1673  1.137  wrstuden 
   1674  1.137  wrstuden 	/* done with modified vn_open, now finish what sys_open does. */
   1675  1.137  wrstuden 
   1676  1.137  wrstuden 	fp->f_flag = flags & FMASK;
   1677  1.137  wrstuden 	fp->f_type = DTYPE_VNODE;
   1678  1.137  wrstuden 	fp->f_ops = &vnops;
   1679  1.184       dsl 	fp->f_data = vp;
   1680  1.137  wrstuden 	if (flags & (O_EXLOCK | O_SHLOCK)) {
   1681  1.137  wrstuden 		lf.l_whence = SEEK_SET;
   1682  1.137  wrstuden 		lf.l_start = 0;
   1683  1.137  wrstuden 		lf.l_len = 0;
   1684  1.137  wrstuden 		if (flags & O_EXLOCK)
   1685  1.137  wrstuden 			lf.l_type = F_WRLCK;
   1686  1.137  wrstuden 		else
   1687  1.137  wrstuden 			lf.l_type = F_RDLCK;
   1688  1.137  wrstuden 		type = F_FLOCK;
   1689  1.137  wrstuden 		if ((flags & FNONBLOCK) == 0)
   1690  1.137  wrstuden 			type |= F_WAIT;
   1691  1.137  wrstuden 		VOP_UNLOCK(vp, 0);
   1692  1.184       dsl 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type);
   1693  1.137  wrstuden 		if (error) {
   1694  1.346        ad 			(void) vn_close(vp, fp->f_flag, fp->f_cred);
   1695  1.346        ad 			fd_abort(p, fp, indx);
   1696  1.137  wrstuden 			return (error);
   1697  1.137  wrstuden 		}
   1698  1.137  wrstuden 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   1699  1.346        ad 		atomic_or_uint(&fp->f_flag, FHASLOCK);
   1700  1.137  wrstuden 	}
   1701  1.137  wrstuden 	VOP_UNLOCK(vp, 0);
   1702  1.137  wrstuden 	*retval = indx;
   1703  1.346        ad 	fd_affix(p, fp, indx);
   1704  1.248      yamt 	vfs_copyinfh_free(fh);
   1705  1.137  wrstuden 	return (0);
   1706  1.139  wrstuden 
   1707  1.137  wrstuden bad:
   1708  1.346        ad 	fd_abort(p, fp, indx);
   1709  1.139  wrstuden 	if (vp != NULL)
   1710  1.139  wrstuden 		vput(vp);
   1711  1.248      yamt 	vfs_copyinfh_free(fh);
   1712  1.137  wrstuden 	return (error);
   1713  1.137  wrstuden }
   1714  1.137  wrstuden 
   1715  1.137  wrstuden int
   1716  1.335       dsl sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval)
   1717  1.137  wrstuden {
   1718  1.335       dsl 	/* {
   1719  1.263    martin 		syscallarg(const void *) fhp;
   1720  1.263    martin 		syscallarg(size_t) fh_size;
   1721  1.265      yamt 		syscallarg(int) flags;
   1722  1.335       dsl 	} */
   1723  1.265      yamt 
   1724  1.265      yamt 	return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size),
   1725  1.265      yamt 	    SCARG(uap, flags), retval);
   1726  1.265      yamt }
   1727  1.265      yamt 
   1728  1.306       dsl int
   1729  1.306       dsl do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb)
   1730  1.306       dsl {
   1731  1.137  wrstuden 	int error;
   1732  1.248      yamt 	fhandle_t *fh;
   1733  1.137  wrstuden 	struct vnode *vp;
   1734  1.137  wrstuden 
   1735  1.137  wrstuden 	/*
   1736  1.137  wrstuden 	 * Must be super user
   1737  1.137  wrstuden 	 */
   1738  1.268      elad 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
   1739  1.270      elad 	    0, NULL, NULL, NULL)))
   1740  1.137  wrstuden 		return (error);
   1741  1.137  wrstuden 
   1742  1.265      yamt 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
   1743  1.306       dsl 	if (error != 0)
   1744  1.306       dsl 		return error;
   1745  1.306       dsl 
   1746  1.248      yamt 	error = vfs_fhtovp(fh, &vp);
   1747  1.306       dsl 	vfs_copyinfh_free(fh);
   1748  1.306       dsl 	if (error != 0)
   1749  1.306       dsl 		return error;
   1750  1.306       dsl 
   1751  1.346        ad 	error = vn_stat(vp, sb);
   1752  1.137  wrstuden 	vput(vp);
   1753  1.248      yamt 	return error;
   1754  1.137  wrstuden }
   1755  1.137  wrstuden 
   1756  1.265      yamt 
   1757  1.137  wrstuden /* ARGSUSED */
   1758  1.137  wrstuden int
   1759  1.383  christos sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval)
   1760  1.137  wrstuden {
   1761  1.335       dsl 	/* {
   1762  1.265      yamt 		syscallarg(const void *) fhp;
   1763  1.263    martin 		syscallarg(size_t) fh_size;
   1764  1.265      yamt 		syscallarg(struct stat *) sb;
   1765  1.335       dsl 	} */
   1766  1.306       dsl 	struct stat sb;
   1767  1.306       dsl 	int error;
   1768  1.265      yamt 
   1769  1.306       dsl 	error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb);
   1770  1.306       dsl 	if (error)
   1771  1.306       dsl 		return error;
   1772  1.306       dsl 	return copyout(&sb, SCARG(uap, sb), sizeof(sb));
   1773  1.265      yamt }
   1774  1.265      yamt 
   1775  1.306       dsl int
   1776  1.306       dsl do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb,
   1777  1.306       dsl     int flags)
   1778  1.306       dsl {
   1779  1.248      yamt 	fhandle_t *fh;
   1780  1.137  wrstuden 	struct mount *mp;
   1781  1.137  wrstuden 	struct vnode *vp;
   1782  1.137  wrstuden 	int error;
   1783  1.137  wrstuden 
   1784  1.137  wrstuden 	/*
   1785  1.137  wrstuden 	 * Must be super user
   1786  1.137  wrstuden 	 */
   1787  1.268      elad 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE,
   1788  1.270      elad 	    0, NULL, NULL, NULL)))
   1789  1.206  christos 		return error;
   1790  1.137  wrstuden 
   1791  1.265      yamt 	error = vfs_copyinfh_alloc(ufhp, fhsize, &fh);
   1792  1.306       dsl 	if (error != 0)
   1793  1.306       dsl 		return error;
   1794  1.306       dsl 
   1795  1.248      yamt 	error = vfs_fhtovp(fh, &vp);
   1796  1.306       dsl 	vfs_copyinfh_free(fh);
   1797  1.306       dsl 	if (error != 0)
   1798  1.306       dsl 		return error;
   1799  1.306       dsl 
   1800  1.137  wrstuden 	mp = vp->v_mount;
   1801  1.306       dsl 	error = dostatvfs(mp, sb, l, flags, 1);
   1802  1.137  wrstuden 	vput(vp);
   1803  1.241      yamt 	return error;
   1804   1.31       cgd }
   1805   1.31       cgd 
   1806  1.265      yamt /* ARGSUSED */
   1807  1.265      yamt int
   1808  1.335       dsl sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval)
   1809  1.265      yamt {
   1810  1.335       dsl 	/* {
   1811  1.266      yamt 		syscallarg(const void *) fhp;
   1812  1.265      yamt 		syscallarg(size_t) fh_size;
   1813  1.265      yamt 		syscallarg(struct statvfs *) buf;
   1814  1.265      yamt 		syscallarg(int)	flags;
   1815  1.335       dsl 	} */
   1816  1.306       dsl 	struct statvfs *sb = STATVFSBUF_GET();
   1817  1.306       dsl 	int error;
   1818  1.265      yamt 
   1819  1.306       dsl 	error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb,
   1820  1.306       dsl 	    SCARG(uap, flags));
   1821  1.306       dsl 	if (error == 0)
   1822  1.306       dsl 		error = copyout(sb, SCARG(uap, buf), sizeof(*sb));
   1823  1.306       dsl 	STATVFSBUF_PUT(sb);
   1824  1.306       dsl 	return error;
   1825  1.265      yamt }
   1826  1.265      yamt 
   1827   1.31       cgd /*
   1828   1.31       cgd  * Create a special file.
   1829   1.31       cgd  */
   1830   1.31       cgd /* ARGSUSED */
   1831   1.63  christos int
   1832  1.383  christos sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap,
   1833  1.383  christos     register_t *retval)
   1834   1.56   thorpej {
   1835  1.335       dsl 	/* {
   1836   1.74       cgd 		syscallarg(const char *) path;
   1837  1.383  christos 		syscallarg(mode_t) mode;
   1838  1.383  christos 		syscallarg(dev_t) dev;
   1839  1.335       dsl 	} */
   1840  1.383  christos 	return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode),
   1841  1.383  christos 	    SCARG(uap, dev), retval);
   1842  1.383  christos }
   1843  1.383  christos 
   1844  1.383  christos int
   1845  1.383  christos do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev,
   1846  1.383  christos     register_t *retval)
   1847  1.383  christos {
   1848  1.179   thorpej 	struct proc *p = l->l_proc;
   1849  1.155  augustss 	struct vnode *vp;
   1850   1.31       cgd 	struct vattr vattr;
   1851  1.301     pooka 	int error, optype;
   1852  1.315  christos 	struct nameidata nd;
   1853  1.314  christos 	char *path;
   1854  1.315  christos 	const char *cpath;
   1855  1.315  christos 	enum uio_seg seg = UIO_USERSPACE;
   1856   1.31       cgd 
   1857  1.268      elad 	if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD,
   1858  1.268      elad 	    0, NULL, NULL, NULL)) != 0)
   1859   1.31       cgd 		return (error);
   1860  1.301     pooka 
   1861  1.301     pooka 	optype = VOP_MKNOD_DESCOFFSET;
   1862  1.315  christos 
   1863  1.383  christos 	VERIEXEC_PATH_GET(pathname, seg, cpath, path);
   1864  1.334     pooka 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath);
   1865  1.315  christos 
   1866   1.63  christos 	if ((error = namei(&nd)) != 0)
   1867  1.314  christos 		goto out;
   1868   1.31       cgd 	vp = nd.ni_vp;
   1869   1.43   mycroft 	if (vp != NULL)
   1870   1.31       cgd 		error = EEXIST;
   1871   1.43   mycroft 	else {
   1872   1.43   mycroft 		VATTR_NULL(&vattr);
   1873  1.328        ad 		/* We will read cwdi->cwdi_cmask unlocked. */
   1874  1.383  christos 		vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
   1875  1.383  christos 		vattr.va_rdev = dev;
   1876   1.43   mycroft 
   1877  1.383  christos 		switch (mode & S_IFMT) {
   1878   1.43   mycroft 		case S_IFMT:	/* used by badsect to flag bad sectors */
   1879   1.43   mycroft 			vattr.va_type = VBAD;
   1880   1.43   mycroft 			break;
   1881   1.43   mycroft 		case S_IFCHR:
   1882   1.43   mycroft 			vattr.va_type = VCHR;
   1883   1.43   mycroft 			break;
   1884   1.43   mycroft 		case S_IFBLK:
   1885   1.43   mycroft 			vattr.va_type = VBLK;
   1886   1.43   mycroft 			break;
   1887   1.43   mycroft 		case S_IFWHT:
   1888  1.301     pooka 			optype = VOP_WHITEOUT_DESCOFFSET;
   1889  1.301     pooka 			break;
   1890  1.301     pooka 		case S_IFREG:
   1891  1.314  christos #if NVERIEXEC > 0
   1892  1.314  christos 			error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp,
   1893  1.314  christos 			    O_CREAT);
   1894  1.314  christos #endif /* NVERIEXEC > 0 */
   1895  1.301     pooka 			vattr.va_type = VREG;
   1896  1.302     pooka 			vattr.va_rdev = VNOVAL;
   1897  1.301     pooka 			optype = VOP_CREATE_DESCOFFSET;
   1898   1.43   mycroft 			break;
   1899   1.43   mycroft 		default:
   1900   1.43   mycroft 			error = EINVAL;
   1901   1.43   mycroft 			break;
   1902   1.43   mycroft 		}
   1903   1.31       cgd 	}
   1904   1.43   mycroft 	if (!error) {
   1905  1.301     pooka 		switch (optype) {
   1906  1.301     pooka 		case VOP_WHITEOUT_DESCOFFSET:
   1907   1.43   mycroft 			error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
   1908   1.43   mycroft 			if (error)
   1909   1.43   mycroft 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   1910   1.43   mycroft 			vput(nd.ni_dvp);
   1911  1.301     pooka 			break;
   1912  1.301     pooka 
   1913  1.301     pooka 		case VOP_MKNOD_DESCOFFSET:
   1914   1.43   mycroft 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
   1915   1.43   mycroft 						&nd.ni_cnd, &vattr);
   1916  1.168     assar 			if (error == 0)
   1917  1.168     assar 				vput(nd.ni_vp);
   1918  1.301     pooka 			break;
   1919  1.301     pooka 
   1920  1.301     pooka 		case VOP_CREATE_DESCOFFSET:
   1921  1.301     pooka 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp,
   1922  1.301     pooka 						&nd.ni_cnd, &vattr);
   1923  1.301     pooka 			if (error == 0)
   1924  1.301     pooka 				vput(nd.ni_vp);
   1925  1.301     pooka 			break;
   1926   1.43   mycroft 		}
   1927   1.43   mycroft 	} else {
   1928   1.43   mycroft 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   1929   1.43   mycroft 		if (nd.ni_dvp == vp)
   1930   1.43   mycroft 			vrele(nd.ni_dvp);
   1931   1.43   mycroft 		else
   1932   1.43   mycroft 			vput(nd.ni_dvp);
   1933   1.43   mycroft 		if (vp)
   1934   1.43   mycroft 			vrele(vp);
   1935   1.31       cgd 	}
   1936  1.314  christos out:
   1937  1.315  christos 	VERIEXEC_PATH_PUT(path);
   1938   1.31       cgd 	return (error);
   1939   1.31       cgd }
   1940   1.31       cgd 
   1941   1.31       cgd /*
   1942   1.31       cgd  * Create a named pipe.
   1943   1.31       cgd  */
   1944   1.31       cgd /* ARGSUSED */
   1945   1.63  christos int
   1946  1.335       dsl sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval)
   1947   1.56   thorpej {
   1948  1.335       dsl 	/* {
   1949   1.74       cgd 		syscallarg(const char *) path;
   1950   1.35       cgd 		syscallarg(int) mode;
   1951  1.335       dsl 	} */
   1952  1.179   thorpej 	struct proc *p = l->l_proc;
   1953   1.31       cgd 	struct vattr vattr;
   1954   1.31       cgd 	int error;
   1955   1.31       cgd 	struct nameidata nd;
   1956   1.31       cgd 
   1957  1.334     pooka 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
   1958  1.334     pooka 	    SCARG(uap, path));
   1959   1.63  christos 	if ((error = namei(&nd)) != 0)
   1960   1.31       cgd 		return (error);
   1961   1.31       cgd 	if (nd.ni_vp != NULL) {
   1962   1.31       cgd 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   1963   1.31       cgd 		if (nd.ni_dvp == nd.ni_vp)
   1964   1.31       cgd 			vrele(nd.ni_dvp);
   1965   1.31       cgd 		else
   1966   1.31       cgd 			vput(nd.ni_dvp);
   1967   1.31       cgd 		vrele(nd.ni_vp);
   1968   1.31       cgd 		return (EEXIST);
   1969   1.31       cgd 	}
   1970   1.31       cgd 	VATTR_NULL(&vattr);
   1971   1.31       cgd 	vattr.va_type = VFIFO;
   1972  1.328        ad 	/* We will read cwdi->cwdi_cmask unlocked. */
   1973  1.134   thorpej 	vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask;
   1974  1.168     assar 	error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
   1975  1.168     assar 	if (error == 0)
   1976  1.168     assar 		vput(nd.ni_vp);
   1977  1.168     assar 	return (error);
   1978   1.31       cgd }
   1979   1.31       cgd 
   1980   1.31       cgd /*
   1981   1.31       cgd  * Make a hard file link.
   1982   1.31       cgd  */
   1983   1.31       cgd /* ARGSUSED */
   1984   1.63  christos int
   1985  1.335       dsl sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval)
   1986   1.56   thorpej {
   1987  1.335       dsl 	/* {
   1988   1.74       cgd 		syscallarg(const char *) path;
   1989   1.74       cgd 		syscallarg(const char *) link;
   1990  1.335       dsl 	} */
   1991  1.155  augustss 	struct vnode *vp;
   1992   1.31       cgd 	struct nameidata nd;
   1993   1.31       cgd 	int error;
   1994   1.31       cgd 
   1995  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   1996  1.334     pooka 	    SCARG(uap, path));
   1997   1.63  christos 	if ((error = namei(&nd)) != 0)
   1998   1.31       cgd 		return (error);
   1999   1.31       cgd 	vp = nd.ni_vp;
   2000  1.331     pooka 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
   2001  1.334     pooka 	    SCARG(uap, link));
   2002   1.66   mycroft 	if ((error = namei(&nd)) != 0)
   2003   1.66   mycroft 		goto out;
   2004   1.66   mycroft 	if (nd.ni_vp) {
   2005   1.66   mycroft 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   2006   1.66   mycroft 		if (nd.ni_dvp == nd.ni_vp)
   2007   1.66   mycroft 			vrele(nd.ni_dvp);
   2008   1.66   mycroft 		else
   2009   1.66   mycroft 			vput(nd.ni_dvp);
   2010   1.66   mycroft 		vrele(nd.ni_vp);
   2011   1.66   mycroft 		error = EEXIST;
   2012   1.66   mycroft 		goto out;
   2013   1.31       cgd 	}
   2014   1.66   mycroft 	error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
   2015   1.66   mycroft out:
   2016   1.31       cgd 	vrele(vp);
   2017   1.31       cgd 	return (error);
   2018   1.31       cgd }
   2019   1.31       cgd 
   2020   1.31       cgd /*
   2021   1.31       cgd  * Make a symbolic link.
   2022   1.31       cgd  */
   2023   1.31       cgd /* ARGSUSED */
   2024   1.63  christos int
   2025  1.335       dsl sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval)
   2026   1.56   thorpej {
   2027  1.335       dsl 	/* {
   2028   1.74       cgd 		syscallarg(const char *) path;
   2029   1.74       cgd 		syscallarg(const char *) link;
   2030  1.335       dsl 	} */
   2031  1.179   thorpej 	struct proc *p = l->l_proc;
   2032   1.31       cgd 	struct vattr vattr;
   2033   1.31       cgd 	char *path;
   2034   1.31       cgd 	int error;
   2035   1.31       cgd 	struct nameidata nd;
   2036   1.31       cgd 
   2037  1.161   thorpej 	path = PNBUF_GET();
   2038   1.63  christos 	error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL);
   2039   1.63  christos 	if (error)
   2040   1.43   mycroft 		goto out;
   2041  1.331     pooka 	NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE,
   2042  1.334     pooka 	    SCARG(uap, link));
   2043   1.63  christos 	if ((error = namei(&nd)) != 0)
   2044   1.43   mycroft 		goto out;
   2045   1.31       cgd 	if (nd.ni_vp) {
   2046   1.31       cgd 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   2047   1.31       cgd 		if (nd.ni_dvp == nd.ni_vp)
   2048   1.31       cgd 			vrele(nd.ni_dvp);
   2049   1.31       cgd 		else
   2050   1.31       cgd 			vput(nd.ni_dvp);
   2051   1.31       cgd 		vrele(nd.ni_vp);
   2052   1.31       cgd 		error = EEXIST;
   2053   1.43   mycroft 		goto out;
   2054   1.31       cgd 	}
   2055   1.31       cgd 	VATTR_NULL(&vattr);
   2056  1.230      jmmv 	vattr.va_type = VLNK;
   2057  1.328        ad 	/* We will read cwdi->cwdi_cmask unlocked. */
   2058  1.134   thorpej 	vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask;
   2059   1.31       cgd 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path);
   2060  1.168     assar 	if (error == 0)
   2061  1.168     assar 		vput(nd.ni_vp);
   2062   1.43   mycroft out:
   2063  1.161   thorpej 	PNBUF_PUT(path);
   2064   1.31       cgd 	return (error);
   2065   1.31       cgd }
   2066   1.31       cgd 
   2067   1.31       cgd /*
   2068   1.43   mycroft  * Delete a whiteout from the filesystem.
   2069   1.43   mycroft  */
   2070   1.43   mycroft /* ARGSUSED */
   2071   1.63  christos int
   2072  1.335       dsl sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval)
   2073   1.56   thorpej {
   2074  1.335       dsl 	/* {
   2075   1.74       cgd 		syscallarg(const char *) path;
   2076  1.335       dsl 	} */
   2077   1.43   mycroft 	int error;
   2078   1.43   mycroft 	struct nameidata nd;
   2079   1.43   mycroft 
   2080  1.331     pooka 	NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT,
   2081  1.334     pooka 	    UIO_USERSPACE, SCARG(uap, path));
   2082   1.43   mycroft 	error = namei(&nd);
   2083   1.43   mycroft 	if (error)
   2084   1.43   mycroft 		return (error);
   2085   1.43   mycroft 
   2086   1.43   mycroft 	if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
   2087   1.43   mycroft 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   2088   1.43   mycroft 		if (nd.ni_dvp == nd.ni_vp)
   2089   1.43   mycroft 			vrele(nd.ni_dvp);
   2090   1.43   mycroft 		else
   2091   1.43   mycroft 			vput(nd.ni_dvp);
   2092   1.43   mycroft 		if (nd.ni_vp)
   2093   1.43   mycroft 			vrele(nd.ni_vp);
   2094   1.43   mycroft 		return (EEXIST);
   2095   1.43   mycroft 	}
   2096   1.63  christos 	if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0)
   2097   1.43   mycroft 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   2098   1.43   mycroft 	vput(nd.ni_dvp);
   2099   1.43   mycroft 	return (error);
   2100   1.43   mycroft }
   2101   1.43   mycroft 
   2102   1.43   mycroft /*
   2103   1.31       cgd  * Delete a name from the filesystem.
   2104   1.31       cgd  */
   2105   1.31       cgd /* ARGSUSED */
   2106   1.63  christos int
   2107  1.335       dsl sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval)
   2108   1.56   thorpej {
   2109  1.335       dsl 	/* {
   2110   1.74       cgd 		syscallarg(const char *) path;
   2111  1.335       dsl 	} */
   2112  1.336        ad 
   2113  1.336        ad 	return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE);
   2114  1.336        ad }
   2115  1.336        ad 
   2116  1.336        ad int
   2117  1.336        ad do_sys_unlink(const char *arg, enum uio_seg seg)
   2118  1.336        ad {
   2119  1.155  augustss 	struct vnode *vp;
   2120   1.31       cgd 	int error;
   2121   1.31       cgd 	struct nameidata nd;
   2122  1.336        ad 	kauth_cred_t cred;
   2123  1.315  christos 	char *path;
   2124  1.315  christos 	const char *cpath;
   2125   1.31       cgd 
   2126  1.336        ad 	VERIEXEC_PATH_GET(arg, seg, cpath, path);
   2127  1.334     pooka 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath);
   2128  1.313      elad 
   2129   1.63  christos 	if ((error = namei(&nd)) != 0)
   2130  1.313      elad 		goto out;
   2131   1.31       cgd 	vp = nd.ni_vp;
   2132   1.31       cgd 
   2133   1.66   mycroft 	/*
   2134   1.66   mycroft 	 * The root of a mounted filesystem cannot be deleted.
   2135   1.66   mycroft 	 */
   2136  1.329        ad 	if (vp->v_vflag & VV_ROOT) {
   2137   1.43   mycroft 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   2138   1.43   mycroft 		if (nd.ni_dvp == vp)
   2139   1.43   mycroft 			vrele(nd.ni_dvp);
   2140   1.43   mycroft 		else
   2141   1.43   mycroft 			vput(nd.ni_dvp);
   2142   1.66   mycroft 		vput(vp);
   2143   1.66   mycroft 		error = EBUSY;
   2144   1.66   mycroft 		goto out;
   2145   1.31       cgd 	}
   2146  1.219     blymn 
   2147  1.256      elad #if NVERIEXEC > 0
   2148  1.222      elad 	/* Handle remove requests for veriexec entries. */
   2149  1.336        ad 	if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) {
   2150  1.219     blymn 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   2151  1.219     blymn 		if (nd.ni_dvp == vp)
   2152  1.219     blymn 			vrele(nd.ni_dvp);
   2153  1.219     blymn 		else
   2154  1.219     blymn 			vput(nd.ni_dvp);
   2155  1.219     blymn 		vput(vp);
   2156  1.219     blymn 		goto out;
   2157  1.219     blymn 	}
   2158  1.256      elad #endif /* NVERIEXEC > 0 */
   2159  1.256      elad 
   2160  1.336        ad 	cred = kauth_cred_get();
   2161  1.253      elad #ifdef FILEASSOC
   2162  1.278      elad 	(void)fileassoc_file_delete(vp);
   2163  1.253      elad #endif /* FILEASSOC */
   2164   1.66   mycroft 	error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
   2165   1.66   mycroft out:
   2166  1.315  christos 	VERIEXEC_PATH_PUT(path);
   2167   1.31       cgd 	return (error);
   2168   1.31       cgd }
   2169   1.31       cgd 
   2170   1.31       cgd /*
   2171   1.31       cgd  * Reposition read/write file offset.
   2172   1.31       cgd  */
   2173   1.63  christos int
   2174  1.335       dsl sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval)
   2175   1.56   thorpej {
   2176  1.335       dsl 	/* {
   2177   1.35       cgd 		syscallarg(int) fd;
   2178   1.35       cgd 		syscallarg(int) pad;
   2179   1.35       cgd 		syscallarg(off_t) offset;
   2180   1.35       cgd 		syscallarg(int) whence;
   2181  1.335       dsl 	} */
   2182  1.257        ad 	kauth_cred_t cred = l->l_cred;
   2183  1.346        ad 	file_t *fp;
   2184   1.84    kleink 	struct vnode *vp;
   2185   1.31       cgd 	struct vattr vattr;
   2186  1.155  augustss 	off_t newoff;
   2187  1.346        ad 	int error, fd;
   2188  1.346        ad 
   2189  1.346        ad 	fd = SCARG(uap, fd);
   2190   1.31       cgd 
   2191  1.346        ad 	if ((fp = fd_getfile(fd)) == NULL)
   2192   1.31       cgd 		return (EBADF);
   2193   1.84    kleink 
   2194  1.346        ad 	vp = fp->f_data;
   2195  1.135   thorpej 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
   2196  1.135   thorpej 		error = ESPIPE;
   2197  1.135   thorpej 		goto out;
   2198  1.135   thorpej 	}
   2199   1.84    kleink 
   2200   1.35       cgd 	switch (SCARG(uap, whence)) {
   2201   1.92    kleink 	case SEEK_CUR:
   2202   1.84    kleink 		newoff = fp->f_offset + SCARG(uap, offset);
   2203   1.31       cgd 		break;
   2204   1.92    kleink 	case SEEK_END:
   2205  1.332     pooka 		error = VOP_GETATTR(vp, &vattr, cred);
   2206  1.328        ad 		if (error) {
   2207  1.135   thorpej 			goto out;
   2208  1.328        ad 		}
   2209   1.84    kleink 		newoff = SCARG(uap, offset) + vattr.va_size;
   2210   1.31       cgd 		break;
   2211   1.92    kleink 	case SEEK_SET:
   2212   1.84    kleink 		newoff = SCARG(uap, offset);
   2213   1.31       cgd 		break;
   2214   1.31       cgd 	default:
   2215  1.135   thorpej 		error = EINVAL;
   2216  1.135   thorpej 		goto out;
   2217   1.31       cgd 	}
   2218  1.328        ad 	if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) {
   2219  1.328        ad 		*(off_t *)retval = fp->f_offset = newoff;
   2220  1.328        ad 	}
   2221  1.135   thorpej  out:
   2222  1.346        ad  	fd_putfile(fd);
   2223  1.135   thorpej 	return (error);
   2224  1.120   thorpej }
   2225  1.120   thorpej 
   2226  1.120   thorpej /*
   2227  1.120   thorpej  * Positional read system call.
   2228  1.120   thorpej  */
   2229  1.120   thorpej int
   2230  1.335       dsl sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval)
   2231  1.120   thorpej {
   2232  1.335       dsl 	/* {
   2233  1.120   thorpej 		syscallarg(int) fd;
   2234  1.120   thorpej 		syscallarg(void *) buf;
   2235  1.120   thorpej 		syscallarg(size_t) nbyte;
   2236  1.120   thorpej 		syscallarg(off_t) offset;
   2237  1.335       dsl 	} */
   2238  1.346        ad 	file_t *fp;
   2239  1.120   thorpej 	struct vnode *vp;
   2240  1.120   thorpej 	off_t offset;
   2241  1.120   thorpej 	int error, fd = SCARG(uap, fd);
   2242  1.120   thorpej 
   2243  1.346        ad 	if ((fp = fd_getfile(fd)) == NULL)
   2244  1.166   thorpej 		return (EBADF);
   2245  1.166   thorpej 
   2246  1.183        pk 	if ((fp->f_flag & FREAD) == 0) {
   2247  1.346        ad 		fd_putfile(fd);
   2248  1.120   thorpej 		return (EBADF);
   2249  1.183        pk 	}
   2250  1.120   thorpej 
   2251  1.346        ad 	vp = fp->f_data;
   2252  1.135   thorpej 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
   2253  1.135   thorpej 		error = ESPIPE;
   2254  1.135   thorpej 		goto out;
   2255  1.135   thorpej 	}
   2256  1.120   thorpej 
   2257  1.120   thorpej 	offset = SCARG(uap, offset);
   2258  1.120   thorpej 
   2259  1.120   thorpej 	/*
   2260  1.120   thorpej 	 * XXX This works because no file systems actually
   2261  1.120   thorpej 	 * XXX take any action on the seek operation.
   2262  1.120   thorpej 	 */
   2263  1.120   thorpej 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
   2264  1.135   thorpej 		goto out;
   2265  1.120   thorpej 
   2266  1.135   thorpej 	/* dofileread() will unuse the descriptor for us */
   2267  1.328        ad 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
   2268  1.120   thorpej 	    &offset, 0, retval));
   2269  1.135   thorpej 
   2270  1.135   thorpej  out:
   2271  1.346        ad 	fd_putfile(fd);
   2272  1.135   thorpej 	return (error);
   2273  1.120   thorpej }
   2274  1.120   thorpej 
   2275  1.120   thorpej /*
   2276  1.120   thorpej  * Positional scatter read system call.
   2277  1.120   thorpej  */
   2278  1.120   thorpej int
   2279  1.335       dsl sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval)
   2280  1.120   thorpej {
   2281  1.335       dsl 	/* {
   2282  1.120   thorpej 		syscallarg(int) fd;
   2283  1.120   thorpej 		syscallarg(const struct iovec *) iovp;
   2284  1.120   thorpej 		syscallarg(int) iovcnt;
   2285  1.120   thorpej 		syscallarg(off_t) offset;
   2286  1.335       dsl 	} */
   2287  1.335       dsl 	off_t offset = SCARG(uap, offset);
   2288  1.120   thorpej 
   2289  1.328        ad 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
   2290  1.335       dsl 	    SCARG(uap, iovcnt), &offset, 0, retval);
   2291  1.120   thorpej }
   2292  1.120   thorpej 
   2293  1.120   thorpej /*
   2294  1.120   thorpej  * Positional write system call.
   2295  1.120   thorpej  */
   2296  1.120   thorpej int
   2297  1.335       dsl sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval)
   2298  1.120   thorpej {
   2299  1.335       dsl 	/* {
   2300  1.120   thorpej 		syscallarg(int) fd;
   2301  1.120   thorpej 		syscallarg(const void *) buf;
   2302  1.120   thorpej 		syscallarg(size_t) nbyte;
   2303  1.120   thorpej 		syscallarg(off_t) offset;
   2304  1.335       dsl 	} */
   2305  1.346        ad 	file_t *fp;
   2306  1.120   thorpej 	struct vnode *vp;
   2307  1.120   thorpej 	off_t offset;
   2308  1.120   thorpej 	int error, fd = SCARG(uap, fd);
   2309  1.120   thorpej 
   2310  1.346        ad 	if ((fp = fd_getfile(fd)) == NULL)
   2311  1.166   thorpej 		return (EBADF);
   2312  1.166   thorpej 
   2313  1.183        pk 	if ((fp->f_flag & FWRITE) == 0) {
   2314  1.346        ad 		fd_putfile(fd);
   2315  1.120   thorpej 		return (EBADF);
   2316  1.183        pk 	}
   2317  1.120   thorpej 
   2318  1.346        ad 	vp = fp->f_data;
   2319  1.135   thorpej 	if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
   2320  1.135   thorpej 		error = ESPIPE;
   2321  1.135   thorpej 		goto out;
   2322  1.135   thorpej 	}
   2323  1.120   thorpej 
   2324  1.120   thorpej 	offset = SCARG(uap, offset);
   2325  1.120   thorpej 
   2326  1.120   thorpej 	/*
   2327  1.120   thorpej 	 * XXX This works because no file systems actually
   2328  1.120   thorpej 	 * XXX take any action on the seek operation.
   2329  1.120   thorpej 	 */
   2330  1.120   thorpej 	if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0)
   2331  1.135   thorpej 		goto out;
   2332  1.120   thorpej 
   2333  1.135   thorpej 	/* dofilewrite() will unuse the descriptor for us */
   2334  1.328        ad 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
   2335  1.120   thorpej 	    &offset, 0, retval));
   2336  1.135   thorpej 
   2337  1.135   thorpej  out:
   2338  1.346        ad 	fd_putfile(fd);
   2339  1.135   thorpej 	return (error);
   2340  1.120   thorpej }
   2341  1.120   thorpej 
   2342  1.120   thorpej /*
   2343  1.120   thorpej  * Positional gather write system call.
   2344  1.120   thorpej  */
   2345  1.120   thorpej int
   2346  1.335       dsl sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval)
   2347  1.120   thorpej {
   2348  1.335       dsl 	/* {
   2349  1.120   thorpej 		syscallarg(int) fd;
   2350  1.120   thorpej 		syscallarg(const struct iovec *) iovp;
   2351  1.120   thorpej 		syscallarg(int) iovcnt;
   2352  1.120   thorpej 		syscallarg(off_t) offset;
   2353  1.335       dsl 	} */
   2354  1.335       dsl 	off_t offset = SCARG(uap, offset);
   2355  1.120   thorpej 
   2356  1.328        ad 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
   2357  1.335       dsl 	    SCARG(uap, iovcnt), &offset, 0, retval);
   2358   1.31       cgd }
   2359   1.31       cgd 
   2360   1.31       cgd /*
   2361   1.31       cgd  * Check access permissions.
   2362   1.31       cgd  */
   2363   1.63  christos int
   2364  1.335       dsl sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval)
   2365   1.56   thorpej {
   2366  1.335       dsl 	/* {
   2367   1.74       cgd 		syscallarg(const char *) path;
   2368   1.35       cgd 		syscallarg(int) flags;
   2369  1.335       dsl 	} */
   2370  1.242      elad 	kauth_cred_t cred;
   2371  1.155  augustss 	struct vnode *vp;
   2372  1.169  christos 	int error, flags;
   2373   1.31       cgd 	struct nameidata nd;
   2374   1.31       cgd 
   2375  1.257        ad 	cred = kauth_cred_dup(l->l_cred);
   2376  1.257        ad 	kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred));
   2377  1.257        ad 	kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred));
   2378  1.310       dsl 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
   2379  1.334     pooka 	    SCARG(uap, path));
   2380  1.169  christos 	/* Override default credentials */
   2381  1.169  christos 	nd.ni_cnd.cn_cred = cred;
   2382   1.63  christos 	if ((error = namei(&nd)) != 0)
   2383  1.169  christos 		goto out;
   2384   1.31       cgd 	vp = nd.ni_vp;
   2385   1.31       cgd 
   2386   1.31       cgd 	/* Flags == 0 means only check for existence. */
   2387   1.35       cgd 	if (SCARG(uap, flags)) {
   2388   1.31       cgd 		flags = 0;
   2389   1.35       cgd 		if (SCARG(uap, flags) & R_OK)
   2390   1.31       cgd 			flags |= VREAD;
   2391   1.35       cgd 		if (SCARG(uap, flags) & W_OK)
   2392   1.31       cgd 			flags |= VWRITE;
   2393   1.89   mycroft 		if (SCARG(uap, flags) & X_OK)
   2394   1.89   mycroft 			flags |= VEXEC;
   2395  1.138        is 
   2396  1.332     pooka 		error = VOP_ACCESS(vp, flags, cred);
   2397  1.138        is 		if (!error && (flags & VWRITE))
   2398  1.138        is 			error = vn_writechk(vp);
   2399   1.31       cgd 	}
   2400   1.31       cgd 	vput(vp);
   2401  1.169  christos out:
   2402  1.242      elad 	kauth_cred_free(cred);
   2403   1.31       cgd 	return (error);
   2404   1.31       cgd }
   2405   1.31       cgd 
   2406   1.31       cgd /*
   2407  1.306       dsl  * Common code for all sys_stat functions, including compat versions.
   2408  1.306       dsl  */
   2409  1.306       dsl int
   2410  1.346        ad do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb)
   2411  1.306       dsl {
   2412  1.306       dsl 	int error;
   2413  1.306       dsl 	struct nameidata nd;
   2414  1.306       dsl 
   2415  1.331     pooka 	NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT,
   2416  1.334     pooka 	    UIO_USERSPACE, path);
   2417  1.306       dsl 	error = namei(&nd);
   2418  1.306       dsl 	if (error != 0)
   2419  1.306       dsl 		return error;
   2420  1.346        ad 	error = vn_stat(nd.ni_vp, sb);
   2421  1.306       dsl 	vput(nd.ni_vp);
   2422  1.306       dsl 	return error;
   2423  1.306       dsl }
   2424  1.306       dsl 
   2425  1.306       dsl /*
   2426   1.31       cgd  * Get file status; this version follows links.
   2427   1.31       cgd  */
   2428   1.31       cgd /* ARGSUSED */
   2429   1.63  christos int
   2430  1.383  christos sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval)
   2431   1.56   thorpej {
   2432  1.335       dsl 	/* {
   2433   1.74       cgd 		syscallarg(const char *) path;
   2434   1.35       cgd 		syscallarg(struct stat *) ub;
   2435  1.335       dsl 	} */
   2436   1.31       cgd 	struct stat sb;
   2437   1.31       cgd 	int error;
   2438   1.31       cgd 
   2439  1.346        ad 	error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb);
   2440   1.31       cgd 	if (error)
   2441  1.306       dsl 		return error;
   2442  1.306       dsl 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
   2443   1.31       cgd }
   2444   1.31       cgd 
   2445   1.31       cgd /*
   2446   1.31       cgd  * Get file status; this version does not follow links.
   2447   1.31       cgd  */
   2448   1.31       cgd /* ARGSUSED */
   2449   1.63  christos int
   2450  1.383  christos sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval)
   2451   1.56   thorpej {
   2452  1.335       dsl 	/* {
   2453   1.74       cgd 		syscallarg(const char *) path;
   2454   1.35       cgd 		syscallarg(struct stat *) ub;
   2455  1.335       dsl 	} */
   2456   1.65   mycroft 	struct stat sb;
   2457   1.31       cgd 	int error;
   2458   1.31       cgd 
   2459  1.346        ad 	error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb);
   2460   1.64       jtc 	if (error)
   2461  1.306       dsl 		return error;
   2462  1.306       dsl 	return copyout(&sb, SCARG(uap, ub), sizeof(sb));
   2463   1.31       cgd }
   2464   1.31       cgd 
   2465   1.31       cgd /*
   2466   1.31       cgd  * Get configurable pathname variables.
   2467   1.31       cgd  */
   2468   1.31       cgd /* ARGSUSED */
   2469   1.63  christos int
   2470  1.335       dsl sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval)
   2471   1.56   thorpej {
   2472  1.335       dsl 	/* {
   2473   1.74       cgd 		syscallarg(const char *) path;
   2474   1.35       cgd 		syscallarg(int) name;
   2475  1.335       dsl 	} */
   2476   1.31       cgd 	int error;
   2477   1.31       cgd 	struct nameidata nd;
   2478   1.31       cgd 
   2479  1.310       dsl 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
   2480  1.334     pooka 	    SCARG(uap, path));
   2481   1.63  christos 	if ((error = namei(&nd)) != 0)
   2482   1.31       cgd 		return (error);
   2483   1.35       cgd 	error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval);
   2484   1.31       cgd 	vput(nd.ni_vp);
   2485   1.31       cgd 	return (error);
   2486   1.31       cgd }
   2487   1.31       cgd 
   2488   1.31       cgd /*
   2489   1.31       cgd  * Return target name of a symbolic link.
   2490   1.31       cgd  */
   2491   1.31       cgd /* ARGSUSED */
   2492   1.63  christos int
   2493  1.335       dsl sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval)
   2494   1.56   thorpej {
   2495  1.335       dsl 	/* {
   2496   1.74       cgd 		syscallarg(const char *) path;
   2497   1.35       cgd 		syscallarg(char *) buf;
   2498  1.115    kleink 		syscallarg(size_t) count;
   2499  1.335       dsl 	} */
   2500  1.155  augustss 	struct vnode *vp;
   2501   1.31       cgd 	struct iovec aiov;
   2502   1.31       cgd 	struct uio auio;
   2503   1.31       cgd 	int error;
   2504   1.31       cgd 	struct nameidata nd;
   2505   1.31       cgd 
   2506  1.310       dsl 	NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
   2507  1.334     pooka 	    SCARG(uap, path));
   2508   1.63  christos 	if ((error = namei(&nd)) != 0)
   2509   1.31       cgd 		return (error);
   2510   1.31       cgd 	vp = nd.ni_vp;
   2511   1.31       cgd 	if (vp->v_type != VLNK)
   2512   1.31       cgd 		error = EINVAL;
   2513  1.106     enami 	else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) ||
   2514  1.332     pooka 	    (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) {
   2515   1.35       cgd 		aiov.iov_base = SCARG(uap, buf);
   2516   1.35       cgd 		aiov.iov_len = SCARG(uap, count);
   2517   1.31       cgd 		auio.uio_iov = &aiov;
   2518   1.31       cgd 		auio.uio_iovcnt = 1;
   2519   1.31       cgd 		auio.uio_offset = 0;
   2520   1.31       cgd 		auio.uio_rw = UIO_READ;
   2521  1.238      yamt 		KASSERT(l == curlwp);
   2522  1.238      yamt 		auio.uio_vmspace = l->l_proc->p_vmspace;
   2523   1.35       cgd 		auio.uio_resid = SCARG(uap, count);
   2524  1.257        ad 		error = VOP_READLINK(vp, &auio, l->l_cred);
   2525   1.31       cgd 	}
   2526   1.31       cgd 	vput(vp);
   2527   1.35       cgd 	*retval = SCARG(uap, count) - auio.uio_resid;
   2528   1.31       cgd 	return (error);
   2529   1.31       cgd }
   2530   1.31       cgd 
   2531   1.31       cgd /*
   2532   1.31       cgd  * Change flags of a file given a path name.
   2533   1.31       cgd  */
   2534   1.31       cgd /* ARGSUSED */
   2535   1.63  christos int
   2536  1.335       dsl sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval)
   2537   1.56   thorpej {
   2538  1.335       dsl 	/* {
   2539   1.74       cgd 		syscallarg(const char *) path;
   2540   1.74       cgd 		syscallarg(u_long) flags;
   2541  1.335       dsl 	} */
   2542  1.155  augustss 	struct vnode *vp;
   2543   1.31       cgd 	int error;
   2544   1.31       cgd 	struct nameidata nd;
   2545   1.31       cgd 
   2546  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2547  1.334     pooka 	    SCARG(uap, path));
   2548   1.63  christos 	if ((error = namei(&nd)) != 0)
   2549   1.31       cgd 		return (error);
   2550   1.31       cgd 	vp = nd.ni_vp;
   2551  1.234  christos 	error = change_flags(vp, SCARG(uap, flags), l);
   2552   1.31       cgd 	vput(vp);
   2553   1.31       cgd 	return (error);
   2554   1.31       cgd }
   2555   1.31       cgd 
   2556   1.31       cgd /*
   2557   1.31       cgd  * Change flags of a file given a file descriptor.
   2558   1.31       cgd  */
   2559   1.31       cgd /* ARGSUSED */
   2560   1.63  christos int
   2561  1.335       dsl sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval)
   2562   1.56   thorpej {
   2563  1.335       dsl 	/* {
   2564   1.35       cgd 		syscallarg(int) fd;
   2565   1.74       cgd 		syscallarg(u_long) flags;
   2566  1.335       dsl 	} */
   2567   1.31       cgd 	struct vnode *vp;
   2568  1.346        ad 	file_t *fp;
   2569   1.31       cgd 	int error;
   2570   1.31       cgd 
   2571  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   2572  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   2573   1.31       cgd 		return (error);
   2574  1.346        ad 	vp = fp->f_data;
   2575  1.234  christos 	error = change_flags(vp, SCARG(uap, flags), l);
   2576  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   2577  1.346        ad 	fd_putfile(SCARG(uap, fd));
   2578  1.156       mrg 	return (error);
   2579  1.156       mrg }
   2580  1.156       mrg 
   2581  1.156       mrg /*
   2582  1.163     enami  * Change flags of a file given a path name; this version does
   2583  1.156       mrg  * not follow links.
   2584  1.156       mrg  */
   2585  1.156       mrg int
   2586  1.335       dsl sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval)
   2587  1.156       mrg {
   2588  1.335       dsl 	/* {
   2589  1.156       mrg 		syscallarg(const char *) path;
   2590  1.156       mrg 		syscallarg(u_long) flags;
   2591  1.335       dsl 	} */
   2592  1.163     enami 	struct vnode *vp;
   2593  1.156       mrg 	int error;
   2594  1.156       mrg 	struct nameidata nd;
   2595  1.156       mrg 
   2596  1.331     pooka 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2597  1.334     pooka 	    SCARG(uap, path));
   2598  1.156       mrg 	if ((error = namei(&nd)) != 0)
   2599  1.156       mrg 		return (error);
   2600  1.156       mrg 	vp = nd.ni_vp;
   2601  1.234  christos 	error = change_flags(vp, SCARG(uap, flags), l);
   2602  1.163     enami 	vput(vp);
   2603  1.163     enami 	return (error);
   2604  1.163     enami }
   2605  1.163     enami 
   2606  1.163     enami /*
   2607  1.163     enami  * Common routine to change flags of a file.
   2608  1.163     enami  */
   2609  1.163     enami int
   2610  1.234  christos change_flags(struct vnode *vp, u_long flags, struct lwp *l)
   2611  1.163     enami {
   2612  1.163     enami 	struct vattr vattr;
   2613  1.163     enami 	int error;
   2614  1.163     enami 
   2615  1.156       mrg 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   2616  1.163     enami 	/*
   2617  1.163     enami 	 * Non-superusers cannot change the flags on devices, even if they
   2618  1.163     enami 	 * own them.
   2619  1.163     enami 	 */
   2620  1.294      elad 	if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
   2621  1.332     pooka 		if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
   2622  1.156       mrg 			goto out;
   2623  1.156       mrg 		if (vattr.va_type == VCHR || vattr.va_type == VBLK) {
   2624  1.156       mrg 			error = EINVAL;
   2625  1.156       mrg 			goto out;
   2626  1.156       mrg 		}
   2627  1.156       mrg 	}
   2628  1.156       mrg 	VATTR_NULL(&vattr);
   2629  1.163     enami 	vattr.va_flags = flags;
   2630  1.332     pooka 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
   2631  1.156       mrg out:
   2632   1.31       cgd 	return (error);
   2633   1.31       cgd }
   2634   1.31       cgd 
   2635   1.31       cgd /*
   2636   1.98     enami  * Change mode of a file given path name; this version follows links.
   2637   1.31       cgd  */
   2638   1.31       cgd /* ARGSUSED */
   2639   1.63  christos int
   2640  1.335       dsl sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval)
   2641   1.56   thorpej {
   2642  1.335       dsl 	/* {
   2643   1.74       cgd 		syscallarg(const char *) path;
   2644   1.35       cgd 		syscallarg(int) mode;
   2645  1.335       dsl 	} */
   2646   1.31       cgd 	int error;
   2647   1.31       cgd 	struct nameidata nd;
   2648   1.31       cgd 
   2649  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2650  1.334     pooka 	    SCARG(uap, path));
   2651   1.63  christos 	if ((error = namei(&nd)) != 0)
   2652   1.31       cgd 		return (error);
   2653   1.97     enami 
   2654  1.234  christos 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
   2655   1.97     enami 
   2656   1.97     enami 	vrele(nd.ni_vp);
   2657   1.31       cgd 	return (error);
   2658   1.31       cgd }
   2659   1.31       cgd 
   2660   1.31       cgd /*
   2661   1.31       cgd  * Change mode of a file given a file descriptor.
   2662   1.31       cgd  */
   2663   1.31       cgd /* ARGSUSED */
   2664   1.63  christos int
   2665  1.335       dsl sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval)
   2666   1.56   thorpej {
   2667  1.335       dsl 	/* {
   2668   1.35       cgd 		syscallarg(int) fd;
   2669   1.35       cgd 		syscallarg(int) mode;
   2670  1.335       dsl 	} */
   2671  1.346        ad 	file_t *fp;
   2672   1.31       cgd 	int error;
   2673   1.31       cgd 
   2674  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   2675  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   2676   1.31       cgd 		return (error);
   2677  1.346        ad 	error = change_mode(fp->f_data, SCARG(uap, mode), l);
   2678  1.346        ad 	fd_putfile(SCARG(uap, fd));
   2679  1.135   thorpej 	return (error);
   2680   1.97     enami }
   2681   1.97     enami 
   2682   1.97     enami /*
   2683   1.98     enami  * Change mode of a file given path name; this version does not follow links.
   2684   1.98     enami  */
   2685   1.98     enami /* ARGSUSED */
   2686   1.98     enami int
   2687  1.335       dsl sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval)
   2688   1.98     enami {
   2689  1.335       dsl 	/* {
   2690   1.98     enami 		syscallarg(const char *) path;
   2691   1.98     enami 		syscallarg(int) mode;
   2692  1.335       dsl 	} */
   2693   1.98     enami 	int error;
   2694   1.98     enami 	struct nameidata nd;
   2695   1.98     enami 
   2696  1.331     pooka 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2697  1.334     pooka 	    SCARG(uap, path));
   2698   1.98     enami 	if ((error = namei(&nd)) != 0)
   2699   1.98     enami 		return (error);
   2700   1.98     enami 
   2701  1.234  christos 	error = change_mode(nd.ni_vp, SCARG(uap, mode), l);
   2702   1.98     enami 
   2703   1.98     enami 	vrele(nd.ni_vp);
   2704   1.98     enami 	return (error);
   2705   1.98     enami }
   2706   1.98     enami 
   2707   1.98     enami /*
   2708   1.97     enami  * Common routine to set mode given a vnode.
   2709   1.97     enami  */
   2710   1.97     enami static int
   2711  1.234  christos change_mode(struct vnode *vp, int mode, struct lwp *l)
   2712   1.97     enami {
   2713   1.97     enami 	struct vattr vattr;
   2714   1.97     enami 	int error;
   2715   1.97     enami 
   2716  1.113      fvdl 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   2717  1.113      fvdl 	VATTR_NULL(&vattr);
   2718  1.113      fvdl 	vattr.va_mode = mode & ALLPERMS;
   2719  1.332     pooka 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
   2720  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   2721   1.31       cgd 	return (error);
   2722   1.31       cgd }
   2723   1.31       cgd 
   2724   1.31       cgd /*
   2725   1.98     enami  * Set ownership given a path name; this version follows links.
   2726   1.31       cgd  */
   2727   1.31       cgd /* ARGSUSED */
   2728   1.63  christos int
   2729  1.335       dsl sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval)
   2730   1.56   thorpej {
   2731  1.335       dsl 	/* {
   2732   1.74       cgd 		syscallarg(const char *) path;
   2733   1.74       cgd 		syscallarg(uid_t) uid;
   2734   1.74       cgd 		syscallarg(gid_t) gid;
   2735  1.335       dsl 	} */
   2736   1.31       cgd 	int error;
   2737   1.31       cgd 	struct nameidata nd;
   2738   1.31       cgd 
   2739  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2740  1.334     pooka 	    SCARG(uap, path));
   2741   1.63  christos 	if ((error = namei(&nd)) != 0)
   2742   1.31       cgd 		return (error);
   2743   1.86    kleink 
   2744  1.234  christos 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
   2745  1.112    kleink 
   2746  1.112    kleink 	vrele(nd.ni_vp);
   2747  1.112    kleink 	return (error);
   2748  1.112    kleink }
   2749  1.112    kleink 
   2750  1.112    kleink /*
   2751  1.112    kleink  * Set ownership given a path name; this version follows links.
   2752  1.112    kleink  * Provides POSIX semantics.
   2753  1.112    kleink  */
   2754  1.112    kleink /* ARGSUSED */
   2755  1.112    kleink int
   2756  1.335       dsl sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval)
   2757  1.112    kleink {
   2758  1.335       dsl 	/* {
   2759  1.112    kleink 		syscallarg(const char *) path;
   2760  1.112    kleink 		syscallarg(uid_t) uid;
   2761  1.112    kleink 		syscallarg(gid_t) gid;
   2762  1.335       dsl 	} */
   2763  1.112    kleink 	int error;
   2764  1.112    kleink 	struct nameidata nd;
   2765  1.112    kleink 
   2766  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2767  1.334     pooka 	    SCARG(uap, path));
   2768  1.112    kleink 	if ((error = namei(&nd)) != 0)
   2769  1.112    kleink 		return (error);
   2770  1.112    kleink 
   2771  1.234  christos 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
   2772   1.86    kleink 
   2773   1.86    kleink 	vrele(nd.ni_vp);
   2774   1.31       cgd 	return (error);
   2775   1.31       cgd }
   2776   1.31       cgd 
   2777   1.31       cgd /*
   2778   1.31       cgd  * Set ownership given a file descriptor.
   2779   1.31       cgd  */
   2780   1.31       cgd /* ARGSUSED */
   2781   1.63  christos int
   2782  1.335       dsl sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval)
   2783   1.56   thorpej {
   2784  1.335       dsl 	/* {
   2785   1.35       cgd 		syscallarg(int) fd;
   2786   1.74       cgd 		syscallarg(uid_t) uid;
   2787   1.74       cgd 		syscallarg(gid_t) gid;
   2788  1.335       dsl 	} */
   2789   1.71   mycroft 	int error;
   2790  1.346        ad 	file_t *fp;
   2791   1.31       cgd 
   2792  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   2793  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   2794   1.31       cgd 		return (error);
   2795  1.346        ad 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
   2796  1.346        ad 	    l, 0);
   2797  1.346        ad 	fd_putfile(SCARG(uap, fd));
   2798  1.135   thorpej 	return (error);
   2799  1.112    kleink }
   2800  1.112    kleink 
   2801  1.112    kleink /*
   2802  1.112    kleink  * Set ownership given a file descriptor, providing POSIX/XPG semantics.
   2803  1.112    kleink  */
   2804  1.112    kleink /* ARGSUSED */
   2805  1.112    kleink int
   2806  1.335       dsl sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval)
   2807  1.112    kleink {
   2808  1.335       dsl 	/* {
   2809  1.112    kleink 		syscallarg(int) fd;
   2810  1.112    kleink 		syscallarg(uid_t) uid;
   2811  1.112    kleink 		syscallarg(gid_t) gid;
   2812  1.335       dsl 	} */
   2813  1.112    kleink 	int error;
   2814  1.346        ad 	file_t *fp;
   2815  1.112    kleink 
   2816  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   2817  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   2818  1.112    kleink 		return (error);
   2819  1.346        ad 	error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid),
   2820  1.346        ad 	    l, 1);
   2821  1.346        ad 	fd_putfile(SCARG(uap, fd));
   2822  1.135   thorpej 	return (error);
   2823   1.86    kleink }
   2824   1.86    kleink 
   2825   1.86    kleink /*
   2826   1.98     enami  * Set ownership given a path name; this version does not follow links.
   2827   1.98     enami  */
   2828   1.98     enami /* ARGSUSED */
   2829   1.98     enami int
   2830  1.335       dsl sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval)
   2831   1.98     enami {
   2832  1.335       dsl 	/* {
   2833   1.98     enami 		syscallarg(const char *) path;
   2834   1.98     enami 		syscallarg(uid_t) uid;
   2835   1.98     enami 		syscallarg(gid_t) gid;
   2836  1.335       dsl 	} */
   2837   1.98     enami 	int error;
   2838   1.98     enami 	struct nameidata nd;
   2839   1.98     enami 
   2840  1.331     pooka 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2841  1.334     pooka 	    SCARG(uap, path));
   2842   1.98     enami 	if ((error = namei(&nd)) != 0)
   2843   1.98     enami 		return (error);
   2844   1.98     enami 
   2845  1.234  christos 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0);
   2846  1.112    kleink 
   2847  1.112    kleink 	vrele(nd.ni_vp);
   2848  1.112    kleink 	return (error);
   2849  1.112    kleink }
   2850  1.112    kleink 
   2851  1.112    kleink /*
   2852  1.112    kleink  * Set ownership given a path name; this version does not follow links.
   2853  1.112    kleink  * Provides POSIX/XPG semantics.
   2854  1.112    kleink  */
   2855  1.112    kleink /* ARGSUSED */
   2856  1.112    kleink int
   2857  1.335       dsl sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval)
   2858  1.112    kleink {
   2859  1.335       dsl 	/* {
   2860  1.112    kleink 		syscallarg(const char *) path;
   2861  1.112    kleink 		syscallarg(uid_t) uid;
   2862  1.112    kleink 		syscallarg(gid_t) gid;
   2863  1.335       dsl 	} */
   2864  1.112    kleink 	int error;
   2865  1.112    kleink 	struct nameidata nd;
   2866  1.112    kleink 
   2867  1.331     pooka 	NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE,
   2868  1.334     pooka 	    SCARG(uap, path));
   2869  1.112    kleink 	if ((error = namei(&nd)) != 0)
   2870  1.112    kleink 		return (error);
   2871  1.112    kleink 
   2872  1.234  christos 	error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1);
   2873   1.98     enami 
   2874   1.98     enami 	vrele(nd.ni_vp);
   2875   1.98     enami 	return (error);
   2876   1.98     enami }
   2877   1.98     enami 
   2878   1.98     enami /*
   2879  1.205  junyoung  * Common routine to set ownership given a vnode.
   2880   1.86    kleink  */
   2881   1.86    kleink static int
   2882  1.234  christos change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l,
   2883  1.221   thorpej     int posix_semantics)
   2884   1.86    kleink {
   2885   1.86    kleink 	struct vattr vattr;
   2886  1.112    kleink 	mode_t newmode;
   2887   1.86    kleink 	int error;
   2888   1.86    kleink 
   2889  1.113      fvdl 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   2890  1.332     pooka 	if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0)
   2891   1.86    kleink 		goto out;
   2892   1.86    kleink 
   2893  1.175   thorpej #define CHANGED(x) ((int)(x) != -1)
   2894  1.112    kleink 	newmode = vattr.va_mode;
   2895  1.112    kleink 	if (posix_semantics) {
   2896  1.112    kleink 		/*
   2897  1.114    kleink 		 * POSIX/XPG semantics: if the caller is not the super-user,
   2898  1.114    kleink 		 * clear set-user-id and set-group-id bits.  Both POSIX and
   2899  1.114    kleink 		 * the XPG consider the behaviour for calls by the super-user
   2900  1.114    kleink 		 * implementation-defined; we leave the set-user-id and set-
   2901  1.114    kleink 		 * group-id settings intact in that case.
   2902  1.112    kleink 		 */
   2903  1.257        ad 		if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
   2904  1.242      elad 				      NULL) != 0)
   2905  1.112    kleink 			newmode &= ~(S_ISUID | S_ISGID);
   2906  1.112    kleink 	} else {
   2907  1.112    kleink 		/*
   2908  1.112    kleink 		 * NetBSD semantics: when changing owner and/or group,
   2909  1.112    kleink 		 * clear the respective bit(s).
   2910  1.112    kleink 		 */
   2911  1.112    kleink 		if (CHANGED(uid))
   2912  1.112    kleink 			newmode &= ~S_ISUID;
   2913  1.112    kleink 		if (CHANGED(gid))
   2914  1.112    kleink 			newmode &= ~S_ISGID;
   2915  1.112    kleink 	}
   2916  1.112    kleink 	/* Update va_mode iff altered. */
   2917  1.112    kleink 	if (vattr.va_mode == newmode)
   2918  1.112    kleink 		newmode = VNOVAL;
   2919  1.205  junyoung 
   2920   1.86    kleink 	VATTR_NULL(&vattr);
   2921  1.175   thorpej 	vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL;
   2922  1.175   thorpej 	vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL;
   2923   1.86    kleink 	vattr.va_mode = newmode;
   2924  1.332     pooka 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
   2925  1.112    kleink #undef CHANGED
   2926  1.205  junyoung 
   2927   1.86    kleink out:
   2928  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   2929   1.31       cgd 	return (error);
   2930   1.31       cgd }
   2931   1.31       cgd 
   2932   1.31       cgd /*
   2933   1.98     enami  * Set the access and modification times given a path name; this
   2934   1.98     enami  * version follows links.
   2935   1.31       cgd  */
   2936   1.31       cgd /* ARGSUSED */
   2937   1.63  christos int
   2938  1.383  christos sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap,
   2939  1.383  christos     register_t *retval)
   2940   1.56   thorpej {
   2941  1.335       dsl 	/* {
   2942   1.74       cgd 		syscallarg(const char *) path;
   2943   1.74       cgd 		syscallarg(const struct timeval *) tptr;
   2944  1.335       dsl 	} */
   2945   1.31       cgd 
   2946  1.312       dsl 	return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW,
   2947  1.346        ad 	    SCARG(uap, tptr), UIO_USERSPACE);
   2948   1.71   mycroft }
   2949   1.71   mycroft 
   2950   1.71   mycroft /*
   2951   1.71   mycroft  * Set the access and modification times given a file descriptor.
   2952   1.71   mycroft  */
   2953   1.71   mycroft /* ARGSUSED */
   2954   1.71   mycroft int
   2955  1.383  christos sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap,
   2956  1.383  christos     register_t *retval)
   2957   1.71   mycroft {
   2958  1.335       dsl 	/* {
   2959   1.71   mycroft 		syscallarg(int) fd;
   2960   1.74       cgd 		syscallarg(const struct timeval *) tptr;
   2961  1.335       dsl 	} */
   2962   1.71   mycroft 	int error;
   2963  1.346        ad 	file_t *fp;
   2964   1.71   mycroft 
   2965  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   2966  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   2967   1.96     enami 		return (error);
   2968  1.346        ad 	error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr),
   2969  1.346        ad 	    UIO_USERSPACE);
   2970  1.346        ad 	fd_putfile(SCARG(uap, fd));
   2971  1.135   thorpej 	return (error);
   2972   1.98     enami }
   2973   1.98     enami 
   2974   1.98     enami /*
   2975   1.98     enami  * Set the access and modification times given a path name; this
   2976   1.98     enami  * version does not follow links.
   2977   1.98     enami  */
   2978   1.98     enami int
   2979  1.383  christos sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap,
   2980  1.383  christos     register_t *retval)
   2981   1.98     enami {
   2982  1.335       dsl 	/* {
   2983   1.98     enami 		syscallarg(const char *) path;
   2984   1.98     enami 		syscallarg(const struct timeval *) tptr;
   2985  1.335       dsl 	} */
   2986   1.98     enami 
   2987  1.312       dsl 	return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW,
   2988  1.346        ad 	    SCARG(uap, tptr), UIO_USERSPACE);
   2989   1.97     enami }
   2990   1.97     enami 
   2991   1.97     enami /*
   2992   1.97     enami  * Common routine to set access and modification times given a vnode.
   2993   1.97     enami  */
   2994  1.312       dsl int
   2995  1.312       dsl do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag,
   2996  1.312       dsl     const struct timeval *tptr, enum uio_seg seg)
   2997   1.97     enami {
   2998   1.97     enami 	struct vattr vattr;
   2999  1.312       dsl 	struct nameidata nd;
   3000   1.97     enami 	int error;
   3001  1.367  christos 	bool vanull, setbirthtime;
   3002  1.367  christos 	struct timespec ts[2];
   3003   1.97     enami 
   3004   1.97     enami 	if (tptr == NULL) {
   3005  1.367  christos 		vanull = true;
   3006  1.367  christos 		nanotime(&ts[0]);
   3007  1.367  christos 		ts[1] = ts[0];
   3008   1.71   mycroft 	} else {
   3009  1.233      yamt 		struct timeval tv[2];
   3010  1.233      yamt 
   3011  1.367  christos 		vanull = false;
   3012  1.312       dsl 		if (seg != UIO_SYSSPACE) {
   3013  1.383  christos 			error = copyin(tptr, tv, sizeof (tv));
   3014  1.312       dsl 			if (error != 0)
   3015  1.312       dsl 				return error;
   3016  1.312       dsl 			tptr = tv;
   3017  1.312       dsl 		}
   3018  1.367  christos 		TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]);
   3019  1.367  christos 		TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]);
   3020   1.71   mycroft 	}
   3021  1.312       dsl 
   3022  1.312       dsl 	if (vp == NULL) {
   3023  1.334     pooka 		NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path);
   3024  1.312       dsl 		if ((error = namei(&nd)) != 0)
   3025  1.367  christos 			return error;
   3026  1.312       dsl 		vp = nd.ni_vp;
   3027  1.312       dsl 	} else
   3028  1.312       dsl 		nd.ni_vp = NULL;
   3029  1.312       dsl 
   3030  1.113      fvdl 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   3031  1.367  christos 	setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 &&
   3032  1.367  christos 	    timespeccmp(&ts[1], &vattr.va_birthtime, <));
   3033  1.367  christos 	VATTR_NULL(&vattr);
   3034  1.368  christos 	vattr.va_atime = ts[0];
   3035  1.368  christos 	vattr.va_mtime = ts[1];
   3036  1.367  christos 	if (setbirthtime)
   3037  1.367  christos 		vattr.va_birthtime = ts[1];
   3038  1.367  christos 	if (vanull)
   3039  1.367  christos 		vattr.va_flags |= VA_UTIMES_NULL;
   3040  1.332     pooka 	error = VOP_SETATTR(vp, &vattr, l->l_cred);
   3041  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   3042  1.312       dsl 
   3043  1.312       dsl 	if (nd.ni_vp != NULL)
   3044  1.312       dsl 		vrele(nd.ni_vp);
   3045  1.312       dsl 
   3046  1.367  christos 	return error;
   3047   1.31       cgd }
   3048   1.31       cgd 
   3049   1.31       cgd /*
   3050   1.31       cgd  * Truncate a file given its path name.
   3051   1.31       cgd  */
   3052   1.31       cgd /* ARGSUSED */
   3053   1.63  christos int
   3054  1.335       dsl sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval)
   3055   1.56   thorpej {
   3056  1.335       dsl 	/* {
   3057   1.74       cgd 		syscallarg(const char *) path;
   3058   1.35       cgd 		syscallarg(int) pad;
   3059   1.35       cgd 		syscallarg(off_t) length;
   3060  1.335       dsl 	} */
   3061  1.155  augustss 	struct vnode *vp;
   3062   1.31       cgd 	struct vattr vattr;
   3063   1.31       cgd 	int error;
   3064   1.31       cgd 	struct nameidata nd;
   3065   1.31       cgd 
   3066  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   3067  1.334     pooka 	    SCARG(uap, path));
   3068   1.63  christos 	if ((error = namei(&nd)) != 0)
   3069   1.31       cgd 		return (error);
   3070   1.31       cgd 	vp = nd.ni_vp;
   3071  1.113      fvdl 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   3072   1.31       cgd 	if (vp->v_type == VDIR)
   3073   1.31       cgd 		error = EISDIR;
   3074   1.31       cgd 	else if ((error = vn_writechk(vp)) == 0 &&
   3075  1.332     pooka 	    (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) {
   3076   1.31       cgd 		VATTR_NULL(&vattr);
   3077   1.35       cgd 		vattr.va_size = SCARG(uap, length);
   3078  1.332     pooka 		error = VOP_SETATTR(vp, &vattr, l->l_cred);
   3079   1.31       cgd 	}
   3080   1.31       cgd 	vput(vp);
   3081   1.31       cgd 	return (error);
   3082   1.31       cgd }
   3083   1.31       cgd 
   3084   1.31       cgd /*
   3085   1.31       cgd  * Truncate a file given a file descriptor.
   3086   1.31       cgd  */
   3087   1.31       cgd /* ARGSUSED */
   3088   1.63  christos int
   3089  1.335       dsl sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval)
   3090   1.56   thorpej {
   3091  1.335       dsl 	/* {
   3092   1.35       cgd 		syscallarg(int) fd;
   3093   1.35       cgd 		syscallarg(int) pad;
   3094   1.35       cgd 		syscallarg(off_t) length;
   3095  1.335       dsl 	} */
   3096   1.31       cgd 	struct vattr vattr;
   3097   1.31       cgd 	struct vnode *vp;
   3098  1.346        ad 	file_t *fp;
   3099   1.31       cgd 	int error;
   3100   1.31       cgd 
   3101  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   3102  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   3103   1.31       cgd 		return (error);
   3104  1.135   thorpej 	if ((fp->f_flag & FWRITE) == 0) {
   3105  1.135   thorpej 		error = EINVAL;
   3106  1.135   thorpej 		goto out;
   3107  1.135   thorpej 	}
   3108  1.346        ad 	vp = fp->f_data;
   3109  1.113      fvdl 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   3110   1.31       cgd 	if (vp->v_type == VDIR)
   3111   1.31       cgd 		error = EISDIR;
   3112   1.31       cgd 	else if ((error = vn_writechk(vp)) == 0) {
   3113   1.31       cgd 		VATTR_NULL(&vattr);
   3114   1.35       cgd 		vattr.va_size = SCARG(uap, length);
   3115  1.332     pooka 		error = VOP_SETATTR(vp, &vattr, fp->f_cred);
   3116   1.31       cgd 	}
   3117  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   3118  1.135   thorpej  out:
   3119  1.346        ad 	fd_putfile(SCARG(uap, fd));
   3120   1.31       cgd 	return (error);
   3121   1.31       cgd }
   3122   1.31       cgd 
   3123   1.31       cgd /*
   3124   1.31       cgd  * Sync an open file.
   3125   1.31       cgd  */
   3126   1.31       cgd /* ARGSUSED */
   3127   1.63  christos int
   3128  1.335       dsl sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval)
   3129   1.56   thorpej {
   3130  1.335       dsl 	/* {
   3131   1.35       cgd 		syscallarg(int) fd;
   3132  1.335       dsl 	} */
   3133  1.155  augustss 	struct vnode *vp;
   3134  1.346        ad 	file_t *fp;
   3135   1.31       cgd 	int error;
   3136   1.31       cgd 
   3137  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   3138  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   3139   1.31       cgd 		return (error);
   3140  1.346        ad 	vp = fp->f_data;
   3141  1.113      fvdl 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   3142  1.332     pooka 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0);
   3143  1.327     pooka 	if (error == 0 && bioopsp != NULL &&
   3144  1.148      fvdl 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
   3145  1.329        ad 		(*bioopsp->io_fsync)(vp, 0);
   3146  1.113      fvdl 	VOP_UNLOCK(vp, 0);
   3147  1.346        ad 	fd_putfile(SCARG(uap, fd));
   3148  1.201   thorpej 	return (error);
   3149  1.201   thorpej }
   3150  1.201   thorpej 
   3151  1.201   thorpej /*
   3152  1.201   thorpej  * Sync a range of file data.  API modeled after that found in AIX.
   3153  1.201   thorpej  *
   3154  1.201   thorpej  * FDATASYNC indicates that we need only save enough metadata to be able
   3155  1.201   thorpej  * to re-read the written data.  Note we duplicate AIX's requirement that
   3156  1.201   thorpej  * the file be open for writing.
   3157  1.201   thorpej  */
   3158  1.201   thorpej /* ARGSUSED */
   3159  1.201   thorpej int
   3160  1.335       dsl sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval)
   3161  1.201   thorpej {
   3162  1.335       dsl 	/* {
   3163  1.201   thorpej 		syscallarg(int) fd;
   3164  1.201   thorpej 		syscallarg(int) flags;
   3165  1.201   thorpej 		syscallarg(off_t) start;
   3166  1.225      cube 		syscallarg(off_t) length;
   3167  1.335       dsl 	} */
   3168  1.201   thorpej 	struct vnode *vp;
   3169  1.346        ad 	file_t *fp;
   3170  1.201   thorpej 	int flags, nflags;
   3171  1.201   thorpej 	off_t s, e, len;
   3172  1.201   thorpej 	int error;
   3173  1.201   thorpej 
   3174  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   3175  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   3176  1.201   thorpej 		return (error);
   3177  1.201   thorpej 
   3178  1.201   thorpej 	if ((fp->f_flag & FWRITE) == 0) {
   3179  1.293  wrstuden 		error = EBADF;
   3180  1.293  wrstuden 		goto out;
   3181  1.201   thorpej 	}
   3182  1.201   thorpej 
   3183  1.201   thorpej 	flags = SCARG(uap, flags);
   3184  1.201   thorpej 	if (((flags & (FDATASYNC | FFILESYNC)) == 0) ||
   3185  1.201   thorpej 	    ((~flags & (FDATASYNC | FFILESYNC)) == 0)) {
   3186  1.293  wrstuden 		error = EINVAL;
   3187  1.293  wrstuden 		goto out;
   3188  1.201   thorpej 	}
   3189  1.201   thorpej 	/* Now set up the flags for value(s) to pass to VOP_FSYNC() */
   3190  1.201   thorpej 	if (flags & FDATASYNC)
   3191  1.201   thorpej 		nflags = FSYNC_DATAONLY | FSYNC_WAIT;
   3192  1.201   thorpej 	else
   3193  1.201   thorpej 		nflags = FSYNC_WAIT;
   3194  1.216  wrstuden 	if (flags & FDISKSYNC)
   3195  1.216  wrstuden 		nflags |= FSYNC_CACHE;
   3196  1.201   thorpej 
   3197  1.201   thorpej 	len = SCARG(uap, length);
   3198  1.201   thorpej 	/* If length == 0, we do the whole file, and s = l = 0 will do that */
   3199  1.201   thorpej 	if (len) {
   3200  1.201   thorpej 		s = SCARG(uap, start);
   3201  1.201   thorpej 		e = s + len;
   3202  1.201   thorpej 		if (e < s) {
   3203  1.293  wrstuden 			error = EINVAL;
   3204  1.293  wrstuden 			goto out;
   3205  1.201   thorpej 		}
   3206  1.201   thorpej 	} else {
   3207  1.201   thorpej 		e = 0;
   3208  1.201   thorpej 		s = 0;
   3209  1.201   thorpej 	}
   3210  1.201   thorpej 
   3211  1.346        ad 	vp = fp->f_data;
   3212  1.201   thorpej 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   3213  1.332     pooka 	error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e);
   3214  1.201   thorpej 
   3215  1.327     pooka 	if (error == 0 && bioopsp != NULL &&
   3216  1.201   thorpej 	    vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP))
   3217  1.329        ad 		(*bioopsp->io_fsync)(vp, nflags);
   3218  1.201   thorpej 
   3219  1.201   thorpej 	VOP_UNLOCK(vp, 0);
   3220  1.293  wrstuden out:
   3221  1.346        ad 	fd_putfile(SCARG(uap, fd));
   3222   1.31       cgd 	return (error);
   3223   1.31       cgd }
   3224   1.31       cgd 
   3225   1.31       cgd /*
   3226  1.117    kleink  * Sync the data of an open file.
   3227  1.117    kleink  */
   3228  1.117    kleink /* ARGSUSED */
   3229  1.117    kleink int
   3230  1.335       dsl sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval)
   3231  1.117    kleink {
   3232  1.335       dsl 	/* {
   3233  1.117    kleink 		syscallarg(int) fd;
   3234  1.335       dsl 	} */
   3235  1.117    kleink 	struct vnode *vp;
   3236  1.346        ad 	file_t *fp;
   3237  1.117    kleink 	int error;
   3238  1.117    kleink 
   3239  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   3240  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   3241  1.117    kleink 		return (error);
   3242  1.199    kleink 	if ((fp->f_flag & FWRITE) == 0) {
   3243  1.346        ad 		fd_putfile(SCARG(uap, fd));
   3244  1.199    kleink 		return (EBADF);
   3245  1.199    kleink 	}
   3246  1.346        ad 	vp = fp->f_data;
   3247  1.117    kleink 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
   3248  1.332     pooka 	error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
   3249  1.117    kleink 	VOP_UNLOCK(vp, 0);
   3250  1.346        ad 	fd_putfile(SCARG(uap, fd));
   3251  1.117    kleink 	return (error);
   3252  1.117    kleink }
   3253  1.117    kleink 
   3254  1.117    kleink /*
   3255   1.90    kleink  * Rename files, (standard) BSD semantics frontend.
   3256   1.31       cgd  */
   3257   1.31       cgd /* ARGSUSED */
   3258   1.63  christos int
   3259  1.335       dsl sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval)
   3260   1.56   thorpej {
   3261  1.335       dsl 	/* {
   3262   1.74       cgd 		syscallarg(const char *) from;
   3263   1.74       cgd 		syscallarg(const char *) to;
   3264  1.335       dsl 	} */
   3265   1.90    kleink 
   3266  1.336        ad 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0));
   3267   1.90    kleink }
   3268   1.90    kleink 
   3269   1.90    kleink /*
   3270   1.90    kleink  * Rename files, POSIX semantics frontend.
   3271   1.90    kleink  */
   3272   1.90    kleink /* ARGSUSED */
   3273   1.90    kleink int
   3274  1.335       dsl sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval)
   3275   1.90    kleink {
   3276  1.335       dsl 	/* {
   3277   1.90    kleink 		syscallarg(const char *) from;
   3278   1.90    kleink 		syscallarg(const char *) to;
   3279  1.335       dsl 	} */
   3280   1.90    kleink 
   3281  1.336        ad 	return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1));
   3282   1.90    kleink }
   3283   1.90    kleink 
   3284   1.90    kleink /*
   3285   1.90    kleink  * Rename files.  Source and destination must either both be directories,
   3286   1.90    kleink  * or both not be directories.  If target is a directory, it must be empty.
   3287   1.90    kleink  * If `from' and `to' refer to the same object, the value of the `retain'
   3288   1.90    kleink  * argument is used to determine whether `from' will be
   3289   1.90    kleink  *
   3290   1.90    kleink  * (retain == 0)	deleted unless `from' and `to' refer to the same
   3291   1.90    kleink  *			object in the file system's name space (BSD).
   3292   1.90    kleink  * (retain == 1)	always retained (POSIX).
   3293   1.90    kleink  */
   3294  1.336        ad int
   3295  1.336        ad do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain)
   3296   1.90    kleink {
   3297  1.155  augustss 	struct vnode *tvp, *fvp, *tdvp;
   3298   1.31       cgd 	struct nameidata fromnd, tond;
   3299  1.344  dholland 	struct mount *fs;
   3300  1.336        ad 	struct lwp *l = curlwp;
   3301  1.234  christos 	struct proc *p;
   3302  1.344  dholland 	uint32_t saveflag;
   3303   1.31       cgd 	int error;
   3304   1.31       cgd 
   3305  1.331     pooka 	NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT,
   3306  1.336        ad 	    seg, from);
   3307   1.63  christos 	if ((error = namei(&fromnd)) != 0)
   3308   1.31       cgd 		return (error);
   3309  1.291     pooka 	if (fromnd.ni_dvp != fromnd.ni_vp)
   3310  1.291     pooka 		VOP_UNLOCK(fromnd.ni_dvp, 0);
   3311   1.31       cgd 	fvp = fromnd.ni_vp;
   3312  1.344  dholland 
   3313  1.344  dholland 	fs = fvp->v_mount;
   3314  1.344  dholland 	error = VFS_RENAMELOCK_ENTER(fs);
   3315  1.344  dholland 	if (error) {
   3316  1.344  dholland 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
   3317  1.344  dholland 		vrele(fromnd.ni_dvp);
   3318  1.344  dholland 		vrele(fvp);
   3319  1.344  dholland 		goto out1;
   3320  1.344  dholland 	}
   3321  1.344  dholland 
   3322  1.344  dholland 	/*
   3323  1.344  dholland 	 * close, partially, yet another race - ideally we should only
   3324  1.344  dholland 	 * go as far as getting fromnd.ni_dvp before getting the per-fs
   3325  1.344  dholland 	 * lock, and then continue to get fromnd.ni_vp, but we can't do
   3326  1.344  dholland 	 * that with namei as it stands.
   3327  1.344  dholland 	 *
   3328  1.344  dholland 	 * This still won't prevent rmdir from nuking fromnd.ni_vp
   3329  1.344  dholland 	 * under us. The real fix is to get the locks in the right
   3330  1.344  dholland 	 * order and do the lookups in the right places, but that's a
   3331  1.344  dholland 	 * major rototill.
   3332  1.344  dholland 	 *
   3333  1.344  dholland 	 * Preserve the SAVESTART in cn_flags, because who knows what
   3334  1.344  dholland 	 * might happen if we don't.
   3335  1.344  dholland 	 *
   3336  1.344  dholland 	 * Note: this logic (as well as this whole function) is cloned
   3337  1.344  dholland 	 * in nfs_serv.c. Proceed accordingly.
   3338  1.344  dholland 	 */
   3339  1.344  dholland 	vrele(fvp);
   3340  1.348  dholland 	if ((fromnd.ni_cnd.cn_namelen == 1 &&
   3341  1.348  dholland 	     fromnd.ni_cnd.cn_nameptr[0] == '.') ||
   3342  1.348  dholland 	    (fromnd.ni_cnd.cn_namelen == 2 &&
   3343  1.348  dholland 	     fromnd.ni_cnd.cn_nameptr[0] == '.' &&
   3344  1.348  dholland 	     fromnd.ni_cnd.cn_nameptr[1] == '.')) {
   3345  1.348  dholland 		error = EINVAL;
   3346  1.348  dholland 		VFS_RENAMELOCK_EXIT(fs);
   3347  1.348  dholland 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
   3348  1.348  dholland 		vrele(fromnd.ni_dvp);
   3349  1.348  dholland 		goto out1;
   3350  1.348  dholland 	}
   3351  1.344  dholland 	saveflag = fromnd.ni_cnd.cn_flags & SAVESTART;
   3352  1.344  dholland 	fromnd.ni_cnd.cn_flags &= ~SAVESTART;
   3353  1.344  dholland 	vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY);
   3354  1.344  dholland 	error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd);
   3355  1.344  dholland 	fromnd.ni_cnd.cn_flags |= saveflag;
   3356  1.344  dholland 	if (error) {
   3357  1.344  dholland 		VOP_UNLOCK(fromnd.ni_dvp, 0);
   3358  1.344  dholland 		VFS_RENAMELOCK_EXIT(fs);
   3359  1.344  dholland 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
   3360  1.344  dholland 		vrele(fromnd.ni_dvp);
   3361  1.344  dholland 		goto out1;
   3362  1.344  dholland 	}
   3363  1.344  dholland 	VOP_UNLOCK(fromnd.ni_vp, 0);
   3364  1.344  dholland 	if (fromnd.ni_dvp != fromnd.ni_vp)
   3365  1.344  dholland 		VOP_UNLOCK(fromnd.ni_dvp, 0);
   3366  1.344  dholland 	fvp = fromnd.ni_vp;
   3367  1.344  dholland 
   3368  1.331     pooka 	NDINIT(&tond, RENAME,
   3369  1.331     pooka 	    LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT
   3370  1.331     pooka 	      | (fvp->v_type == VDIR ? CREATEDIR : 0),
   3371  1.336        ad 	    seg, to);
   3372   1.63  christos 	if ((error = namei(&tond)) != 0) {
   3373  1.344  dholland 		VFS_RENAMELOCK_EXIT(fs);
   3374   1.31       cgd 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
   3375   1.31       cgd 		vrele(fromnd.ni_dvp);
   3376   1.31       cgd 		vrele(fvp);
   3377   1.31       cgd 		goto out1;
   3378   1.31       cgd 	}
   3379   1.31       cgd 	tdvp = tond.ni_dvp;
   3380   1.31       cgd 	tvp = tond.ni_vp;
   3381   1.90    kleink 
   3382   1.31       cgd 	if (tvp != NULL) {
   3383   1.31       cgd 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
   3384   1.31       cgd 			error = ENOTDIR;
   3385   1.31       cgd 			goto out;
   3386   1.31       cgd 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
   3387   1.31       cgd 			error = EISDIR;
   3388   1.31       cgd 			goto out;
   3389   1.31       cgd 		}
   3390   1.31       cgd 	}
   3391   1.90    kleink 
   3392   1.31       cgd 	if (fvp == tdvp)
   3393   1.31       cgd 		error = EINVAL;
   3394   1.90    kleink 
   3395   1.82    kleink 	/*
   3396   1.90    kleink 	 * Source and destination refer to the same object.
   3397   1.82    kleink 	 */
   3398   1.90    kleink 	if (fvp == tvp) {
   3399   1.90    kleink 		if (retain)
   3400   1.90    kleink 			error = -1;
   3401   1.90    kleink 		else if (fromnd.ni_dvp == tdvp &&
   3402   1.90    kleink 		    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
   3403  1.123     perry 		    !memcmp(fromnd.ni_cnd.cn_nameptr,
   3404   1.90    kleink 		          tond.ni_cnd.cn_nameptr,
   3405   1.90    kleink 		          fromnd.ni_cnd.cn_namelen))
   3406   1.82    kleink 		error = -1;
   3407   1.90    kleink 	}
   3408   1.90    kleink 
   3409  1.256      elad #if NVERIEXEC > 0
   3410  1.282      elad 	if (!error) {
   3411  1.313      elad 		char *f1, *f2;
   3412  1.313      elad 
   3413  1.313      elad 		f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
   3414  1.382      elad 		strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen + 1);
   3415  1.313      elad 
   3416  1.313      elad 		f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK);
   3417  1.382      elad 		strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen + 1);
   3418  1.282      elad 
   3419  1.315  christos 		error = veriexec_renamechk(l, fvp, f1, tvp, f2);
   3420  1.282      elad 
   3421  1.313      elad 		free(f1, M_TEMP);
   3422  1.313      elad 		free(f2, M_TEMP);
   3423  1.282      elad 	}
   3424  1.256      elad #endif /* NVERIEXEC > 0 */
   3425  1.229      elad 
   3426   1.31       cgd out:
   3427  1.234  christos 	p = l->l_proc;
   3428   1.31       cgd 	if (!error) {
   3429   1.31       cgd 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
   3430   1.31       cgd 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
   3431  1.344  dholland 		VFS_RENAMELOCK_EXIT(fs);
   3432   1.31       cgd 	} else {
   3433   1.31       cgd 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
   3434   1.31       cgd 		if (tdvp == tvp)
   3435   1.31       cgd 			vrele(tdvp);
   3436   1.31       cgd 		else
   3437   1.31       cgd 			vput(tdvp);
   3438   1.31       cgd 		if (tvp)
   3439   1.31       cgd 			vput(tvp);
   3440  1.344  dholland 		VFS_RENAMELOCK_EXIT(fs);
   3441   1.31       cgd 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
   3442   1.31       cgd 		vrele(fromnd.ni_dvp);
   3443   1.31       cgd 		vrele(fvp);
   3444   1.31       cgd 	}
   3445   1.31       cgd 	vrele(tond.ni_startdir);
   3446  1.161   thorpej 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
   3447   1.31       cgd out1:
   3448   1.31       cgd 	if (fromnd.ni_startdir)
   3449   1.31       cgd 		vrele(fromnd.ni_startdir);
   3450  1.161   thorpej 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
   3451   1.80    kleink 	return (error == -1 ? 0 : error);
   3452   1.31       cgd }
   3453   1.31       cgd 
   3454   1.31       cgd /*
   3455   1.31       cgd  * Make a directory file.
   3456   1.31       cgd  */
   3457   1.31       cgd /* ARGSUSED */
   3458   1.63  christos int
   3459  1.335       dsl sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval)
   3460   1.56   thorpej {
   3461  1.335       dsl 	/* {
   3462   1.74       cgd 		syscallarg(const char *) path;
   3463   1.35       cgd 		syscallarg(int) mode;
   3464  1.335       dsl 	} */
   3465  1.179   thorpej 	struct proc *p = l->l_proc;
   3466  1.155  augustss 	struct vnode *vp;
   3467   1.31       cgd 	struct vattr vattr;
   3468   1.31       cgd 	int error;
   3469   1.31       cgd 	struct nameidata nd;
   3470   1.31       cgd 
   3471  1.310       dsl 	NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE,
   3472  1.334     pooka 	    SCARG(uap, path));
   3473   1.63  christos 	if ((error = namei(&nd)) != 0)
   3474   1.31       cgd 		return (error);
   3475   1.31       cgd 	vp = nd.ni_vp;
   3476   1.31       cgd 	if (vp != NULL) {
   3477   1.31       cgd 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   3478   1.31       cgd 		if (nd.ni_dvp == vp)
   3479   1.31       cgd 			vrele(nd.ni_dvp);
   3480   1.31       cgd 		else
   3481   1.31       cgd 			vput(nd.ni_dvp);
   3482   1.31       cgd 		vrele(vp);
   3483   1.31       cgd 		return (EEXIST);
   3484   1.31       cgd 	}
   3485   1.31       cgd 	VATTR_NULL(&vattr);
   3486   1.31       cgd 	vattr.va_type = VDIR;
   3487  1.328        ad 	/* We will read cwdi->cwdi_cmask unlocked. */
   3488  1.134   thorpej 	vattr.va_mode =
   3489  1.134   thorpej 	    (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask;
   3490   1.31       cgd 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
   3491   1.31       cgd 	if (!error)
   3492   1.31       cgd 		vput(nd.ni_vp);
   3493   1.31       cgd 	return (error);
   3494   1.31       cgd }
   3495   1.31       cgd 
   3496   1.31       cgd /*
   3497   1.31       cgd  * Remove a directory file.
   3498   1.31       cgd  */
   3499   1.31       cgd /* ARGSUSED */
   3500   1.63  christos int
   3501  1.335       dsl sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval)
   3502   1.56   thorpej {
   3503  1.335       dsl 	/* {
   3504   1.74       cgd 		syscallarg(const char *) path;
   3505  1.335       dsl 	} */
   3506  1.155  augustss 	struct vnode *vp;
   3507   1.31       cgd 	int error;
   3508   1.31       cgd 	struct nameidata nd;
   3509   1.31       cgd 
   3510  1.310       dsl 	NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE,
   3511  1.334     pooka 	    SCARG(uap, path));
   3512   1.63  christos 	if ((error = namei(&nd)) != 0)
   3513   1.31       cgd 		return (error);
   3514   1.31       cgd 	vp = nd.ni_vp;
   3515   1.31       cgd 	if (vp->v_type != VDIR) {
   3516   1.31       cgd 		error = ENOTDIR;
   3517   1.31       cgd 		goto out;
   3518   1.31       cgd 	}
   3519   1.31       cgd 	/*
   3520   1.31       cgd 	 * No rmdir "." please.
   3521   1.31       cgd 	 */
   3522   1.31       cgd 	if (nd.ni_dvp == vp) {
   3523   1.31       cgd 		error = EINVAL;
   3524   1.31       cgd 		goto out;
   3525   1.31       cgd 	}
   3526   1.31       cgd 	/*
   3527   1.31       cgd 	 * The root of a mounted filesystem cannot be deleted.
   3528   1.31       cgd 	 */
   3529  1.350     joerg 	if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) {
   3530   1.31       cgd 		error = EBUSY;
   3531  1.197   hannken 		goto out;
   3532  1.197   hannken 	}
   3533  1.197   hannken 	error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
   3534  1.197   hannken 	return (error);
   3535  1.197   hannken 
   3536  1.197   hannken out:
   3537  1.197   hannken 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
   3538  1.197   hannken 	if (nd.ni_dvp == vp)
   3539  1.197   hannken 		vrele(nd.ni_dvp);
   3540  1.197   hannken 	else
   3541  1.197   hannken 		vput(nd.ni_dvp);
   3542  1.197   hannken 	vput(vp);
   3543   1.31       cgd 	return (error);
   3544   1.31       cgd }
   3545   1.31       cgd 
   3546   1.31       cgd /*
   3547   1.31       cgd  * Read a block of directory entries in a file system independent format.
   3548   1.31       cgd  */
   3549   1.63  christos int
   3550  1.335       dsl sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval)
   3551   1.56   thorpej {
   3552  1.335       dsl 	/* {
   3553   1.35       cgd 		syscallarg(int) fd;
   3554   1.35       cgd 		syscallarg(char *) buf;
   3555  1.101      fvdl 		syscallarg(size_t) count;
   3556  1.335       dsl 	} */
   3557  1.346        ad 	file_t *fp;
   3558  1.101      fvdl 	int error, done;
   3559   1.31       cgd 
   3560  1.346        ad 	/* fd_getvnode() will use the descriptor for us */
   3561  1.346        ad 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
   3562   1.31       cgd 		return (error);
   3563  1.135   thorpej 	if ((fp->f_flag & FREAD) == 0) {
   3564  1.135   thorpej 		error = EBADF;
   3565  1.135   thorpej 		goto out;
   3566  1.135   thorpej 	}
   3567  1.101      fvdl 	error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE,
   3568  1.234  christos 			SCARG(uap, count), &done, l, 0, 0);
   3569  1.325        ad 	ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error);
   3570  1.101      fvdl 	*retval = done;
   3571  1.135   thorpej  out:
   3572  1.346        ad 	fd_putfile(SCARG(uap, fd));
   3573   1.31       cgd 	return (error);
   3574   1.31       cgd }
   3575   1.31       cgd 
   3576   1.31       cgd /*
   3577   1.31       cgd  * Set the mode mask for creation of filesystem nodes.
   3578   1.31       cgd  */
   3579   1.56   thorpej int
   3580  1.335       dsl sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval)
   3581   1.56   thorpej {
   3582  1.335       dsl 	/* {
   3583  1.103   mycroft 		syscallarg(mode_t) newmask;
   3584  1.335       dsl 	} */
   3585  1.179   thorpej 	struct proc *p = l->l_proc;
   3586  1.134   thorpej 	struct cwdinfo *cwdi;
   3587   1.31       cgd 
   3588  1.328        ad 	/*
   3589  1.328        ad 	 * cwdi->cwdi_cmask will be read unlocked elsewhere.  What's
   3590  1.328        ad 	 * important is that we serialize changes to the mask.  The
   3591  1.328        ad 	 * rw_exit() will issue a write memory barrier on our behalf,
   3592  1.328        ad 	 * and force the changes out to other CPUs (as it must use an
   3593  1.328        ad 	 * atomic operation, draining the local CPU's store buffers).
   3594  1.328        ad 	 */
   3595  1.134   thorpej 	cwdi = p->p_cwdi;
   3596  1.328        ad 	rw_enter(&cwdi->cwdi_lock, RW_WRITER);
   3597  1.134   thorpej 	*retval = cwdi->cwdi_cmask;
   3598  1.134   thorpej 	cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS;
   3599  1.328        ad 	rw_exit(&cwdi->cwdi_lock);
   3600  1.328        ad 
   3601   1.31       cgd 	return (0);
   3602   1.31       cgd }
   3603   1.31       cgd 
   3604  1.340      elad int
   3605  1.340      elad dorevoke(struct vnode *vp, kauth_cred_t cred)
   3606  1.340      elad {
   3607  1.340      elad 	struct vattr vattr;
   3608  1.340      elad 	int error;
   3609  1.340      elad 
   3610  1.340      elad 	if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0)
   3611  1.342        ad 		return error;
   3612  1.340      elad 	if (kauth_cred_geteuid(cred) != vattr.va_uid &&
   3613  1.340      elad 	    (error = kauth_authorize_generic(cred,
   3614  1.342        ad 	    KAUTH_GENERIC_ISSUSER, NULL)) == 0)
   3615  1.340      elad 		VOP_REVOKE(vp, REVOKEALL);
   3616  1.340      elad 	return (error);
   3617  1.340      elad }
   3618  1.340      elad 
   3619   1.31       cgd /*
   3620   1.31       cgd  * Void all references to file by ripping underlying filesystem
   3621   1.31       cgd  * away from vnode.
   3622   1.31       cgd  */
   3623   1.31       cgd /* ARGSUSED */
   3624   1.63  christos int
   3625  1.335       dsl sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval)
   3626   1.56   thorpej {
   3627  1.335       dsl 	/* {
   3628   1.74       cgd 		syscallarg(const char *) path;
   3629  1.335       dsl 	} */
   3630  1.155  augustss 	struct vnode *vp;
   3631   1.31       cgd 	int error;
   3632   1.31       cgd 	struct nameidata nd;
   3633   1.31       cgd 
   3634  1.331     pooka 	NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE,
   3635  1.334     pooka 	    SCARG(uap, path));
   3636   1.63  christos 	if ((error = namei(&nd)) != 0)
   3637   1.31       cgd 		return (error);
   3638   1.31       cgd 	vp = nd.ni_vp;
   3639  1.340      elad 	error = dorevoke(vp, l->l_cred);
   3640   1.31       cgd 	vrele(vp);
   3641   1.31       cgd 	return (error);
   3642   1.31       cgd }
   3643