Home | History | Annotate | Line # | Download | only in kern
vfs_vnops.c revision 1.191
      1  1.191      matt /*	$NetBSD: vfs_vnops.c,v 1.191 2014/09/05 09:20:59 matt Exp $	*/
      2  1.164        ad 
      3  1.164        ad /*-
      4  1.164        ad  * Copyright (c) 2009 The NetBSD Foundation, Inc.
      5  1.164        ad  * All rights reserved.
      6  1.164        ad  *
      7  1.164        ad  * This code is derived from software contributed to The NetBSD Foundation
      8  1.164        ad  * by Andrew Doran.
      9  1.164        ad  *
     10  1.164        ad  * Redistribution and use in source and binary forms, with or without
     11  1.164        ad  * modification, are permitted provided that the following conditions
     12  1.164        ad  * are met:
     13  1.164        ad  * 1. Redistributions of source code must retain the above copyright
     14  1.164        ad  *    notice, this list of conditions and the following disclaimer.
     15  1.164        ad  * 2. Redistributions in binary form must reproduce the above copyright
     16  1.164        ad  *    notice, this list of conditions and the following disclaimer in the
     17  1.164        ad  *    documentation and/or other materials provided with the distribution.
     18  1.164        ad  *
     19  1.164        ad  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  1.164        ad  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  1.164        ad  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  1.164        ad  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  1.164        ad  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  1.164        ad  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  1.164        ad  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  1.164        ad  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  1.164        ad  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  1.164        ad  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  1.164        ad  * POSSIBILITY OF SUCH DAMAGE.
     30  1.164        ad  */
     31   1.12       cgd 
     32   1.10       cgd /*
     33   1.11   mycroft  * Copyright (c) 1982, 1986, 1989, 1993
     34   1.11   mycroft  *	The Regents of the University of California.  All rights reserved.
     35   1.10       cgd  * (c) UNIX System Laboratories, Inc.
     36   1.10       cgd  * All or some portions of this file are derived from material licensed
     37   1.10       cgd  * to the University of California by American Telephone and Telegraph
     38   1.10       cgd  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     39   1.10       cgd  * the permission of UNIX System Laboratories, Inc.
     40   1.10       cgd  *
     41   1.10       cgd  * Redistribution and use in source and binary forms, with or without
     42   1.10       cgd  * modification, are permitted provided that the following conditions
     43   1.10       cgd  * are met:
     44   1.10       cgd  * 1. Redistributions of source code must retain the above copyright
     45   1.10       cgd  *    notice, this list of conditions and the following disclaimer.
     46   1.10       cgd  * 2. Redistributions in binary form must reproduce the above copyright
     47   1.10       cgd  *    notice, this list of conditions and the following disclaimer in the
     48   1.10       cgd  *    documentation and/or other materials provided with the distribution.
     49   1.73       agc  * 3. Neither the name of the University nor the names of its contributors
     50   1.10       cgd  *    may be used to endorse or promote products derived from this software
     51   1.10       cgd  *    without specific prior written permission.
     52   1.10       cgd  *
     53   1.10       cgd  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     54   1.10       cgd  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     55   1.10       cgd  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     56   1.10       cgd  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     57   1.10       cgd  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     58   1.10       cgd  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     59   1.10       cgd  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     60   1.10       cgd  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     61   1.10       cgd  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     62   1.10       cgd  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     63   1.10       cgd  * SUCH DAMAGE.
     64   1.10       cgd  *
     65   1.28      fvdl  *	@(#)vfs_vnops.c	8.14 (Berkeley) 6/15/95
     66   1.10       cgd  */
     67   1.52     lukem 
     68   1.52     lukem #include <sys/cdefs.h>
     69  1.191      matt __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.191 2014/09/05 09:20:59 matt Exp $");
     70   1.26       mrg 
     71  1.122    dogcow #include "veriexec.h"
     72   1.10       cgd 
     73   1.10       cgd #include <sys/param.h>
     74   1.10       cgd #include <sys/systm.h>
     75   1.10       cgd #include <sys/kernel.h>
     76   1.10       cgd #include <sys/file.h>
     77   1.10       cgd #include <sys/stat.h>
     78   1.10       cgd #include <sys/buf.h>
     79   1.10       cgd #include <sys/proc.h>
     80   1.10       cgd #include <sys/mount.h>
     81   1.10       cgd #include <sys/namei.h>
     82   1.10       cgd #include <sys/vnode.h>
     83   1.10       cgd #include <sys/ioctl.h>
     84   1.10       cgd #include <sys/tty.h>
     85   1.21   mycroft #include <sys/poll.h>
     86  1.111      elad #include <sys/kauth.h>
     87  1.128      elad #include <sys/syslog.h>
     88  1.141   hannken #include <sys/fstrans.h>
     89  1.145        ad #include <sys/atomic.h>
     90  1.151     pooka #include <sys/filedesc.h>
     91  1.159    simonb #include <sys/wapbl.h>
     92   1.11   mycroft 
     93   1.77   hannken #include <miscfs/specfs/specdev.h>
     94  1.170     pooka #include <miscfs/fifofs/fifo.h>
     95   1.77   hannken 
     96   1.25       mrg #include <uvm/uvm_extern.h>
     97  1.100      yamt #include <uvm/uvm_readahead.h>
     98   1.25       mrg 
     99   1.28      fvdl #ifdef UNION
    100   1.65  jdolecek #include <fs/union/union.h>
    101   1.28      fvdl #endif
    102   1.64  jdolecek 
    103  1.101  christos int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct lwp *);
    104   1.66  jdolecek 
    105   1.58     blymn #include <sys/verified_exec.h>
    106   1.28      fvdl 
    107  1.155        ad static int vn_read(file_t *fp, off_t *offset, struct uio *uio,
    108  1.111      elad 	    kauth_cred_t cred, int flags);
    109  1.155        ad static int vn_write(file_t *fp, off_t *offset, struct uio *uio,
    110  1.111      elad 	    kauth_cred_t cred, int flags);
    111  1.155        ad static int vn_closefile(file_t *fp);
    112  1.155        ad static int vn_poll(file_t *fp, int events);
    113  1.155        ad static int vn_fcntl(file_t *fp, u_int com, void *data);
    114  1.155        ad static int vn_statfile(file_t *fp, struct stat *sb);
    115  1.155        ad static int vn_ioctl(file_t *fp, u_long com, void *data);
    116   1.48  jdolecek 
    117   1.83  christos const struct fileops vnops = {
    118  1.164        ad 	.fo_read = vn_read,
    119  1.164        ad 	.fo_write = vn_write,
    120  1.164        ad 	.fo_ioctl = vn_ioctl,
    121  1.164        ad 	.fo_fcntl = vn_fcntl,
    122  1.164        ad 	.fo_poll = vn_poll,
    123  1.164        ad 	.fo_stat = vn_statfile,
    124  1.164        ad 	.fo_close = vn_closefile,
    125  1.164        ad 	.fo_kqfilter = vn_kqfilter,
    126  1.168       dsl 	.fo_restart = fnullop_restart,
    127   1.48  jdolecek };
    128   1.10       cgd 
    129   1.10       cgd /*
    130   1.10       cgd  * Common code for vnode open operations.
    131   1.10       cgd  * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
    132   1.10       cgd  */
    133   1.20  christos int
    134   1.89   thorpej vn_open(struct nameidata *ndp, int fmode, int cmode)
    135   1.10       cgd {
    136   1.94       erh 	struct vnode *vp;
    137  1.148     pooka 	struct lwp *l = curlwp;
    138  1.117        ad 	kauth_cred_t cred = l->l_cred;
    139   1.19   mycroft 	struct vattr va;
    140   1.10       cgd 	int error;
    141  1.180  dholland 	const char *pathstring;
    142   1.75   hannken 
    143  1.178       chs 	if ((fmode & (O_CREAT | O_DIRECTORY)) == (O_CREAT | O_DIRECTORY))
    144  1.178       chs 		return EINVAL;
    145  1.178       chs 
    146  1.163     enami 	ndp->ni_cnd.cn_flags &= TRYEMULROOT | NOCHROOT;
    147  1.138       dsl 
    148   1.10       cgd 	if (fmode & O_CREAT) {
    149   1.11   mycroft 		ndp->ni_cnd.cn_nameiop = CREATE;
    150  1.138       dsl 		ndp->ni_cnd.cn_flags |= LOCKPARENT | LOCKLEAF;
    151   1.38    bouyer 		if ((fmode & O_EXCL) == 0 &&
    152   1.59  christos 		    ((fmode & O_NOFOLLOW) == 0))
    153   1.11   mycroft 			ndp->ni_cnd.cn_flags |= FOLLOW;
    154  1.130      elad 	} else {
    155  1.130      elad 		ndp->ni_cnd.cn_nameiop = LOOKUP;
    156  1.138       dsl 		ndp->ni_cnd.cn_flags |= LOCKLEAF;
    157  1.130      elad 		if ((fmode & O_NOFOLLOW) == 0)
    158  1.130      elad 			ndp->ni_cnd.cn_flags |= FOLLOW;
    159  1.130      elad 	}
    160  1.139  christos 
    161  1.180  dholland 	pathstring = pathbuf_stringcopy_get(ndp->ni_pathbuf);
    162  1.180  dholland 	if (pathstring == NULL) {
    163  1.180  dholland 		return ENOMEM;
    164  1.180  dholland 	}
    165  1.139  christos 
    166  1.130      elad 	error = namei(ndp);
    167  1.130      elad 	if (error)
    168  1.139  christos 		goto out;
    169   1.91      elad 
    170  1.130      elad 	vp = ndp->ni_vp;
    171   1.91      elad 
    172  1.130      elad #if NVERIEXEC > 0
    173  1.180  dholland 	error = veriexec_openchk(l, ndp->ni_vp, pathstring, fmode);
    174  1.190      maxv 	if (error) {
    175  1.190      maxv 		/* We have to release the locks ourselves */
    176  1.190      maxv 		if (fmode & O_CREAT) {
    177  1.190      maxv 			if (vp == NULL) {
    178  1.190      maxv 				vput(ndp->ni_dvp);
    179  1.190      maxv 			} else {
    180  1.190      maxv 				VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    181  1.190      maxv 				if (ndp->ni_dvp == ndp->ni_vp)
    182  1.190      maxv 					vrele(ndp->ni_dvp);
    183  1.190      maxv 				else
    184  1.190      maxv 					vput(ndp->ni_dvp);
    185  1.190      maxv 				ndp->ni_dvp = NULL;
    186  1.190      maxv 				vput(vp);
    187  1.190      maxv 			}
    188  1.190      maxv 		} else {
    189  1.190      maxv 			vput(vp);
    190  1.190      maxv 		}
    191  1.190      maxv 		goto out;
    192  1.190      maxv 	}
    193  1.115      elad #endif /* NVERIEXEC > 0 */
    194   1.91      elad 
    195  1.130      elad 	if (fmode & O_CREAT) {
    196  1.130      elad 		if (ndp->ni_vp == NULL) {
    197  1.169     pooka 			vattr_null(&va);
    198   1.19   mycroft 			va.va_type = VREG;
    199   1.19   mycroft 			va.va_mode = cmode;
    200   1.28      fvdl 			if (fmode & O_EXCL)
    201   1.28      fvdl 				 va.va_vaflags |= VA_EXCLUSIVE;
    202   1.20  christos 			error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
    203   1.20  christos 					   &ndp->ni_cnd, &va);
    204  1.187   hannken 			vput(ndp->ni_dvp);
    205   1.20  christos 			if (error)
    206  1.139  christos 				goto out;
    207   1.10       cgd 			fmode &= ~O_TRUNC;
    208   1.10       cgd 			vp = ndp->ni_vp;
    209  1.188   hannken 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    210   1.10       cgd 		} else {
    211   1.11   mycroft 			VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
    212   1.10       cgd 			if (ndp->ni_dvp == ndp->ni_vp)
    213   1.10       cgd 				vrele(ndp->ni_dvp);
    214   1.10       cgd 			else
    215   1.10       cgd 				vput(ndp->ni_dvp);
    216   1.10       cgd 			ndp->ni_dvp = NULL;
    217   1.10       cgd 			vp = ndp->ni_vp;
    218   1.10       cgd 			if (fmode & O_EXCL) {
    219   1.10       cgd 				error = EEXIST;
    220   1.38    bouyer 				goto bad;
    221   1.38    bouyer 			}
    222   1.10       cgd 			fmode &= ~O_CREAT;
    223   1.10       cgd 		}
    224   1.10       cgd 	} else {
    225   1.10       cgd 		vp = ndp->ni_vp;
    226   1.10       cgd 	}
    227   1.10       cgd 	if (vp->v_type == VSOCK) {
    228   1.10       cgd 		error = EOPNOTSUPP;
    229   1.74        cb 		goto bad;
    230   1.74        cb 	}
    231   1.74        cb 	if (ndp->ni_vp->v_type == VLNK) {
    232   1.74        cb 		error = EFTYPE;
    233   1.10       cgd 		goto bad;
    234   1.10       cgd 	}
    235   1.58     blymn 
    236   1.10       cgd 	if ((fmode & O_CREAT) == 0) {
    237  1.146      yamt 		error = vn_openchk(vp, cred, fmode);
    238  1.146      yamt 		if (error != 0)
    239  1.146      yamt 			goto bad;
    240   1.10       cgd 	}
    241   1.87     blymn 
    242   1.10       cgd 	if (fmode & O_TRUNC) {
    243  1.169     pooka 		vattr_null(&va);
    244   1.19   mycroft 		va.va_size = 0;
    245  1.144     pooka 		error = VOP_SETATTR(vp, &va, cred);
    246   1.75   hannken 		if (error != 0)
    247   1.10       cgd 			goto bad;
    248   1.10       cgd 	}
    249  1.144     pooka 	if ((error = VOP_OPEN(vp, fmode, cred)) != 0)
    250   1.10       cgd 		goto bad;
    251  1.143        ad 	if (fmode & FWRITE) {
    252  1.181     rmind 		mutex_enter(vp->v_interlock);
    253   1.10       cgd 		vp->v_writecount++;
    254  1.181     rmind 		mutex_exit(vp->v_interlock);
    255  1.143        ad 	}
    256   1.45       chs 
    257   1.10       cgd bad:
    258  1.127      elad 	if (error)
    259  1.127      elad 		vput(vp);
    260  1.139  christos out:
    261  1.180  dholland 	pathbuf_stringcopy_put(ndp->ni_pathbuf, pathstring);
    262   1.10       cgd 	return (error);
    263   1.10       cgd }
    264   1.10       cgd 
    265   1.10       cgd /*
    266   1.10       cgd  * Check for write permissions on the specified vnode.
    267   1.28      fvdl  * Prototype text segments cannot be written.
    268   1.10       cgd  */
    269   1.20  christos int
    270   1.89   thorpej vn_writechk(struct vnode *vp)
    271   1.10       cgd {
    272   1.10       cgd 
    273   1.10       cgd 	/*
    274   1.45       chs 	 * If the vnode is in use as a process's text,
    275   1.45       chs 	 * we can't allow writing.
    276   1.10       cgd 	 */
    277  1.143        ad 	if (vp->v_iflag & VI_TEXT)
    278   1.25       mrg 		return (ETXTBSY);
    279   1.10       cgd 	return (0);
    280   1.42       chs }
    281   1.42       chs 
    282  1.146      yamt int
    283  1.146      yamt vn_openchk(struct vnode *vp, kauth_cred_t cred, int fflags)
    284  1.146      yamt {
    285  1.146      yamt 	int permbits = 0;
    286  1.146      yamt 	int error;
    287  1.146      yamt 
    288  1.178       chs 	if ((fflags & O_DIRECTORY) != 0 && vp->v_type != VDIR)
    289  1.178       chs 		return ENOTDIR;
    290  1.178       chs 
    291  1.146      yamt 	if ((fflags & FREAD) != 0) {
    292  1.146      yamt 		permbits = VREAD;
    293  1.146      yamt 	}
    294  1.146      yamt 	if ((fflags & (FWRITE | O_TRUNC)) != 0) {
    295  1.146      yamt 		permbits |= VWRITE;
    296  1.146      yamt 		if (vp->v_type == VDIR) {
    297  1.146      yamt 			error = EISDIR;
    298  1.146      yamt 			goto bad;
    299  1.146      yamt 		}
    300  1.146      yamt 		error = vn_writechk(vp);
    301  1.146      yamt 		if (error != 0)
    302  1.146      yamt 			goto bad;
    303  1.146      yamt 	}
    304  1.146      yamt 	error = VOP_ACCESS(vp, permbits, cred);
    305  1.146      yamt bad:
    306  1.146      yamt 	return error;
    307  1.146      yamt }
    308  1.146      yamt 
    309   1.42       chs /*
    310   1.51   thorpej  * Mark a vnode as having executable mappings.
    311   1.42       chs  */
    312   1.42       chs void
    313   1.89   thorpej vn_markexec(struct vnode *vp)
    314   1.42       chs {
    315  1.143        ad 
    316  1.158        ad 	if ((vp->v_iflag & VI_EXECMAP) != 0) {
    317  1.158        ad 		/* Safe unlocked, as long as caller holds a reference. */
    318  1.158        ad 		return;
    319  1.158        ad 	}
    320  1.143        ad 
    321  1.181     rmind 	mutex_enter(vp->v_interlock);
    322  1.143        ad 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
    323  1.145        ad 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
    324  1.145        ad 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
    325  1.158        ad 		vp->v_iflag |= VI_EXECMAP;
    326   1.46       chs 	}
    327  1.181     rmind 	mutex_exit(vp->v_interlock);
    328   1.55       chs }
    329   1.55       chs 
    330   1.55       chs /*
    331   1.55       chs  * Mark a vnode as being the text of a process.
    332   1.55       chs  * Fail if the vnode is currently writable.
    333   1.55       chs  */
    334   1.55       chs int
    335   1.89   thorpej vn_marktext(struct vnode *vp)
    336   1.55       chs {
    337   1.55       chs 
    338  1.157        ad 	if ((vp->v_iflag & (VI_TEXT|VI_EXECMAP)) == (VI_TEXT|VI_EXECMAP)) {
    339  1.157        ad 		/* Safe unlocked, as long as caller holds a reference. */
    340  1.158        ad 		return (0);
    341  1.157        ad 	}
    342  1.157        ad 
    343  1.181     rmind 	mutex_enter(vp->v_interlock);
    344   1.55       chs 	if (vp->v_writecount != 0) {
    345  1.143        ad 		KASSERT((vp->v_iflag & VI_TEXT) == 0);
    346  1.181     rmind 		mutex_exit(vp->v_interlock);
    347   1.55       chs 		return (ETXTBSY);
    348   1.55       chs 	}
    349  1.158        ad 	if ((vp->v_iflag & VI_EXECMAP) == 0) {
    350  1.158        ad 		atomic_add_int(&uvmexp.filepages, -vp->v_uobj.uo_npages);
    351  1.158        ad 		atomic_add_int(&uvmexp.execpages, vp->v_uobj.uo_npages);
    352  1.158        ad 	}
    353  1.158        ad 	vp->v_iflag |= (VI_TEXT | VI_EXECMAP);
    354  1.181     rmind 	mutex_exit(vp->v_interlock);
    355   1.55       chs 	return (0);
    356   1.10       cgd }
    357   1.10       cgd 
    358   1.10       cgd /*
    359   1.10       cgd  * Vnode close call
    360   1.32  wrstuden  *
    361   1.32  wrstuden  * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
    362   1.10       cgd  */
    363   1.20  christos int
    364  1.155        ad vn_close(struct vnode *vp, int flags, kauth_cred_t cred)
    365   1.10       cgd {
    366   1.10       cgd 	int error;
    367   1.10       cgd 
    368  1.158        ad 	if (flags & FWRITE) {
    369  1.181     rmind 		mutex_enter(vp->v_interlock);
    370  1.182      yamt 		KASSERT(vp->v_writecount > 0);
    371   1.10       cgd 		vp->v_writecount--;
    372  1.181     rmind 		mutex_exit(vp->v_interlock);
    373  1.158        ad 	}
    374  1.158        ad 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    375  1.144     pooka 	error = VOP_CLOSE(vp, flags, cred);
    376   1.32  wrstuden 	vput(vp);
    377   1.10       cgd 	return (error);
    378   1.10       cgd }
    379   1.10       cgd 
    380  1.171     pooka static int
    381  1.171     pooka enforce_rlimit_fsize(struct vnode *vp, struct uio *uio, int ioflag)
    382  1.171     pooka {
    383  1.171     pooka 	struct lwp *l = curlwp;
    384  1.171     pooka 	off_t testoff;
    385  1.171     pooka 
    386  1.171     pooka 	if (uio->uio_rw != UIO_WRITE || vp->v_type != VREG)
    387  1.171     pooka 		return 0;
    388  1.171     pooka 
    389  1.171     pooka 	KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
    390  1.171     pooka 	if (ioflag & IO_APPEND)
    391  1.171     pooka 		testoff = vp->v_size;
    392  1.171     pooka 	else
    393  1.171     pooka 		testoff = uio->uio_offset;
    394  1.171     pooka 
    395  1.171     pooka 	if (testoff + uio->uio_resid >
    396  1.171     pooka 	    l->l_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
    397  1.171     pooka 		mutex_enter(proc_lock);
    398  1.171     pooka 		psignal(l->l_proc, SIGXFSZ);
    399  1.171     pooka 		mutex_exit(proc_lock);
    400  1.171     pooka 		return EFBIG;
    401  1.171     pooka 	}
    402  1.171     pooka 
    403  1.171     pooka 	return 0;
    404  1.171     pooka }
    405  1.171     pooka 
    406   1.10       cgd /*
    407   1.10       cgd  * Package up an I/O request on a vnode into a uio and do it.
    408   1.10       cgd  */
    409   1.20  christos int
    410  1.134  christos vn_rdwr(enum uio_rw rw, struct vnode *vp, void *base, int len, off_t offset,
    411  1.111      elad     enum uio_seg segflg, int ioflg, kauth_cred_t cred, size_t *aresid,
    412  1.101  christos     struct lwp *l)
    413   1.10       cgd {
    414   1.10       cgd 	struct uio auio;
    415   1.10       cgd 	struct iovec aiov;
    416   1.10       cgd 	int error;
    417   1.10       cgd 
    418   1.50       chs 	if ((ioflg & IO_NODELOCKED) == 0) {
    419   1.50       chs 		if (rw == UIO_READ) {
    420   1.50       chs 			vn_lock(vp, LK_SHARED | LK_RETRY);
    421   1.75   hannken 		} else /* UIO_WRITE */ {
    422   1.50       chs 			vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    423   1.50       chs 		}
    424   1.50       chs 	}
    425   1.10       cgd 	auio.uio_iov = &aiov;
    426   1.10       cgd 	auio.uio_iovcnt = 1;
    427   1.10       cgd 	aiov.iov_base = base;
    428   1.10       cgd 	aiov.iov_len = len;
    429   1.10       cgd 	auio.uio_resid = len;
    430   1.10       cgd 	auio.uio_offset = offset;
    431   1.10       cgd 	auio.uio_rw = rw;
    432  1.106      yamt 	if (segflg == UIO_SYSSPACE) {
    433  1.106      yamt 		UIO_SETUP_SYSSPACE(&auio);
    434  1.106      yamt 	} else {
    435  1.106      yamt 		auio.uio_vmspace = l->l_proc->p_vmspace;
    436  1.106      yamt 	}
    437  1.171     pooka 
    438  1.171     pooka 	if ((error = enforce_rlimit_fsize(vp, &auio, ioflg)) != 0)
    439  1.171     pooka 		goto out;
    440  1.171     pooka 
    441   1.11   mycroft 	if (rw == UIO_READ) {
    442   1.10       cgd 		error = VOP_READ(vp, &auio, ioflg, cred);
    443   1.11   mycroft 	} else {
    444   1.10       cgd 		error = VOP_WRITE(vp, &auio, ioflg, cred);
    445   1.11   mycroft 	}
    446  1.171     pooka 
    447   1.10       cgd 	if (aresid)
    448   1.10       cgd 		*aresid = auio.uio_resid;
    449   1.10       cgd 	else
    450   1.10       cgd 		if (auio.uio_resid && error == 0)
    451   1.10       cgd 			error = EIO;
    452  1.171     pooka 
    453  1.171     pooka  out:
    454   1.75   hannken 	if ((ioflg & IO_NODELOCKED) == 0) {
    455  1.174   hannken 		VOP_UNLOCK(vp);
    456   1.75   hannken 	}
    457   1.10       cgd 	return (error);
    458   1.23      fvdl }
    459   1.23      fvdl 
    460   1.23      fvdl int
    461  1.155        ad vn_readdir(file_t *fp, char *bf, int segflg, u_int count, int *done,
    462  1.101  christos     struct lwp *l, off_t **cookies, int *ncookies)
    463   1.23      fvdl {
    464  1.191      matt 	struct vnode *vp = fp->f_vnode;
    465   1.23      fvdl 	struct iovec aiov;
    466   1.23      fvdl 	struct uio auio;
    467   1.23      fvdl 	int error, eofflag;
    468   1.23      fvdl 
    469  1.112    simonb 	/* Limit the size on any kernel buffers used by VOP_READDIR */
    470  1.112    simonb 	count = min(MAXBSIZE, count);
    471  1.112    simonb 
    472   1.23      fvdl unionread:
    473   1.23      fvdl 	if (vp->v_type != VDIR)
    474   1.23      fvdl 		return (EINVAL);
    475   1.88  christos 	aiov.iov_base = bf;
    476   1.23      fvdl 	aiov.iov_len = count;
    477   1.23      fvdl 	auio.uio_iov = &aiov;
    478   1.23      fvdl 	auio.uio_iovcnt = 1;
    479   1.23      fvdl 	auio.uio_rw = UIO_READ;
    480  1.106      yamt 	if (segflg == UIO_SYSSPACE) {
    481  1.106      yamt 		UIO_SETUP_SYSSPACE(&auio);
    482  1.106      yamt 	} else {
    483  1.106      yamt 		KASSERT(l == curlwp);
    484  1.106      yamt 		auio.uio_vmspace = l->l_proc->p_vmspace;
    485  1.106      yamt 	}
    486   1.23      fvdl 	auio.uio_resid = count;
    487   1.50       chs 	vn_lock(vp, LK_SHARED | LK_RETRY);
    488   1.23      fvdl 	auio.uio_offset = fp->f_offset;
    489   1.28      fvdl 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
    490   1.23      fvdl 		    ncookies);
    491  1.166      yamt 	mutex_enter(&fp->f_lock);
    492   1.23      fvdl 	fp->f_offset = auio.uio_offset;
    493  1.166      yamt 	mutex_exit(&fp->f_lock);
    494  1.174   hannken 	VOP_UNLOCK(vp);
    495   1.23      fvdl 	if (error)
    496   1.23      fvdl 		return (error);
    497   1.23      fvdl 
    498   1.66  jdolecek 	if (count == auio.uio_resid && vn_union_readdir_hook) {
    499   1.66  jdolecek 		struct vnode *ovp = vp;
    500   1.23      fvdl 
    501  1.101  christos 		error = (*vn_union_readdir_hook)(&vp, fp, l);
    502   1.66  jdolecek 		if (error)
    503   1.66  jdolecek 			return (error);
    504   1.66  jdolecek 		if (vp != ovp)
    505   1.23      fvdl 			goto unionread;
    506   1.23      fvdl 	}
    507   1.23      fvdl 
    508  1.143        ad 	if (count == auio.uio_resid && (vp->v_vflag & VV_ROOT) &&
    509   1.23      fvdl 	    (vp->v_mount->mnt_flag & MNT_UNION)) {
    510   1.23      fvdl 		struct vnode *tvp = vp;
    511   1.23      fvdl 		vp = vp->v_mount->mnt_vnodecovered;
    512  1.169     pooka 		vref(vp);
    513  1.166      yamt 		mutex_enter(&fp->f_lock);
    514  1.191      matt 		fp->f_vnode = vp;
    515   1.23      fvdl 		fp->f_offset = 0;
    516  1.166      yamt 		mutex_exit(&fp->f_lock);
    517   1.23      fvdl 		vrele(tvp);
    518   1.23      fvdl 		goto unionread;
    519   1.23      fvdl 	}
    520   1.23      fvdl 	*done = count - auio.uio_resid;
    521   1.23      fvdl 	return error;
    522   1.10       cgd }
    523   1.10       cgd 
    524   1.10       cgd /*
    525   1.10       cgd  * File table vnode read routine.
    526   1.10       cgd  */
    527   1.56  gmcgarry static int
    528  1.155        ad vn_read(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
    529   1.89   thorpej     int flags)
    530   1.10       cgd {
    531  1.191      matt 	struct vnode *vp = fp->f_vnode;
    532  1.185  dholland 	int error, ioflag, fflag;
    533  1.185  dholland 	size_t count;
    534   1.10       cgd 
    535  1.100      yamt 	ioflag = IO_ADV_ENCODE(fp->f_advice);
    536  1.155        ad 	fflag = fp->f_flag;
    537  1.155        ad 	if (fflag & FNONBLOCK)
    538   1.31    kleink 		ioflag |= IO_NDELAY;
    539  1.155        ad 	if ((fflag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
    540   1.31    kleink 		ioflag |= IO_SYNC;
    541  1.155        ad 	if (fflag & FALTIO)
    542   1.37  wrstuden 		ioflag |= IO_ALTSEMANTICS;
    543  1.155        ad 	if (fflag & FDIRECT)
    544  1.125       chs 		ioflag |= IO_DIRECT;
    545   1.50       chs 	vn_lock(vp, LK_SHARED | LK_RETRY);
    546   1.29   thorpej 	uio->uio_offset = *offset;
    547   1.10       cgd 	count = uio->uio_resid;
    548   1.31    kleink 	error = VOP_READ(vp, uio, ioflag, cred);
    549   1.29   thorpej 	if (flags & FOF_UPDATE_OFFSET)
    550   1.29   thorpej 		*offset += count - uio->uio_resid;
    551  1.174   hannken 	VOP_UNLOCK(vp);
    552   1.10       cgd 	return (error);
    553   1.10       cgd }
    554   1.10       cgd 
    555   1.10       cgd /*
    556   1.10       cgd  * File table vnode write routine.
    557   1.10       cgd  */
    558   1.56  gmcgarry static int
    559  1.155        ad vn_write(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
    560   1.89   thorpej     int flags)
    561   1.10       cgd {
    562  1.191      matt 	struct vnode *vp = fp->f_vnode;
    563  1.185  dholland 	int error, ioflag, fflag;
    564  1.185  dholland 	size_t count;
    565   1.10       cgd 
    566  1.151     pooka 	ioflag = IO_ADV_ENCODE(fp->f_advice) | IO_UNIT;
    567  1.155        ad 	fflag = fp->f_flag;
    568  1.155        ad 	if (vp->v_type == VREG && (fflag & O_APPEND))
    569   1.10       cgd 		ioflag |= IO_APPEND;
    570  1.155        ad 	if (fflag & FNONBLOCK)
    571   1.10       cgd 		ioflag |= IO_NDELAY;
    572  1.155        ad 	if (fflag & FFSYNC ||
    573   1.24   thorpej 	    (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
    574   1.24   thorpej 		ioflag |= IO_SYNC;
    575  1.155        ad 	else if (fflag & FDSYNC)
    576   1.31    kleink 		ioflag |= IO_DSYNC;
    577  1.155        ad 	if (fflag & FALTIO)
    578   1.37  wrstuden 		ioflag |= IO_ALTSEMANTICS;
    579  1.155        ad 	if (fflag & FDIRECT)
    580  1.125       chs 		ioflag |= IO_DIRECT;
    581   1.28      fvdl 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    582   1.29   thorpej 	uio->uio_offset = *offset;
    583   1.10       cgd 	count = uio->uio_resid;
    584  1.171     pooka 
    585  1.171     pooka 	if ((error = enforce_rlimit_fsize(vp, uio, ioflag)) != 0)
    586  1.171     pooka 		goto out;
    587  1.171     pooka 
    588   1.10       cgd 	error = VOP_WRITE(vp, uio, ioflag, cred);
    589  1.171     pooka 
    590   1.29   thorpej 	if (flags & FOF_UPDATE_OFFSET) {
    591  1.160  christos 		if (ioflag & IO_APPEND) {
    592  1.160  christos 			/*
    593  1.160  christos 			 * SUSv3 describes behaviour for count = 0 as following:
    594  1.160  christos 			 * "Before any action ... is taken, and if nbyte is zero
    595  1.160  christos 			 * and the file is a regular file, the write() function
    596  1.160  christos 			 * ... in the absence of errors ... shall return zero
    597  1.160  christos 			 * and have no other results."
    598  1.160  christos 			 */
    599  1.160  christos 			if (count)
    600  1.160  christos 				*offset = uio->uio_offset;
    601  1.160  christos 		} else
    602   1.29   thorpej 			*offset += count - uio->uio_resid;
    603   1.29   thorpej 	}
    604  1.171     pooka 
    605  1.171     pooka  out:
    606  1.174   hannken 	VOP_UNLOCK(vp);
    607   1.10       cgd 	return (error);
    608   1.10       cgd }
    609   1.10       cgd 
    610   1.10       cgd /*
    611   1.10       cgd  * File table vnode stat routine.
    612   1.10       cgd  */
    613   1.48  jdolecek static int
    614  1.155        ad vn_statfile(file_t *fp, struct stat *sb)
    615   1.48  jdolecek {
    616  1.191      matt 	struct vnode *vp = fp->f_vnode;
    617  1.165  christos 	int error;
    618   1.48  jdolecek 
    619  1.165  christos 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    620  1.165  christos 	error = vn_stat(vp, sb);
    621  1.174   hannken 	VOP_UNLOCK(vp);
    622  1.165  christos 	return error;
    623   1.48  jdolecek }
    624   1.48  jdolecek 
    625   1.20  christos int
    626  1.155        ad vn_stat(struct vnode *vp, struct stat *sb)
    627   1.10       cgd {
    628   1.19   mycroft 	struct vattr va;
    629   1.10       cgd 	int error;
    630   1.35  wrstuden 	mode_t mode;
    631   1.10       cgd 
    632  1.175     pooka 	memset(&va, 0, sizeof(va));
    633  1.155        ad 	error = VOP_GETATTR(vp, &va, kauth_cred_get());
    634   1.10       cgd 	if (error)
    635   1.10       cgd 		return (error);
    636   1.10       cgd 	/*
    637   1.10       cgd 	 * Copy from vattr table
    638   1.10       cgd 	 */
    639  1.179     pooka 	memset(sb, 0, sizeof(*sb));
    640   1.19   mycroft 	sb->st_dev = va.va_fsid;
    641   1.19   mycroft 	sb->st_ino = va.va_fileid;
    642   1.19   mycroft 	mode = va.va_mode;
    643   1.10       cgd 	switch (vp->v_type) {
    644   1.10       cgd 	case VREG:
    645   1.10       cgd 		mode |= S_IFREG;
    646   1.10       cgd 		break;
    647   1.10       cgd 	case VDIR:
    648   1.10       cgd 		mode |= S_IFDIR;
    649   1.10       cgd 		break;
    650   1.10       cgd 	case VBLK:
    651   1.10       cgd 		mode |= S_IFBLK;
    652   1.10       cgd 		break;
    653   1.10       cgd 	case VCHR:
    654   1.10       cgd 		mode |= S_IFCHR;
    655   1.10       cgd 		break;
    656   1.10       cgd 	case VLNK:
    657   1.10       cgd 		mode |= S_IFLNK;
    658   1.10       cgd 		break;
    659   1.10       cgd 	case VSOCK:
    660   1.10       cgd 		mode |= S_IFSOCK;
    661   1.10       cgd 		break;
    662   1.10       cgd 	case VFIFO:
    663   1.10       cgd 		mode |= S_IFIFO;
    664   1.10       cgd 		break;
    665   1.10       cgd 	default:
    666   1.10       cgd 		return (EBADF);
    667   1.10       cgd 	};
    668   1.10       cgd 	sb->st_mode = mode;
    669   1.19   mycroft 	sb->st_nlink = va.va_nlink;
    670   1.19   mycroft 	sb->st_uid = va.va_uid;
    671   1.19   mycroft 	sb->st_gid = va.va_gid;
    672   1.19   mycroft 	sb->st_rdev = va.va_rdev;
    673   1.19   mycroft 	sb->st_size = va.va_size;
    674   1.19   mycroft 	sb->st_atimespec = va.va_atime;
    675   1.19   mycroft 	sb->st_mtimespec = va.va_mtime;
    676   1.19   mycroft 	sb->st_ctimespec = va.va_ctime;
    677   1.69      fvdl 	sb->st_birthtimespec = va.va_birthtime;
    678   1.19   mycroft 	sb->st_blksize = va.va_blocksize;
    679   1.19   mycroft 	sb->st_flags = va.va_flags;
    680   1.22   mycroft 	sb->st_gen = 0;
    681   1.19   mycroft 	sb->st_blocks = va.va_bytes / S_BLKSIZE;
    682   1.10       cgd 	return (0);
    683   1.37  wrstuden }
    684   1.37  wrstuden 
    685   1.37  wrstuden /*
    686   1.37  wrstuden  * File table vnode fcntl routine.
    687   1.37  wrstuden  */
    688   1.56  gmcgarry static int
    689  1.155        ad vn_fcntl(file_t *fp, u_int com, void *data)
    690   1.37  wrstuden {
    691  1.191      matt 	struct vnode *vp = fp->f_vnode;
    692   1.37  wrstuden 	int error;
    693   1.37  wrstuden 
    694  1.155        ad 	error = VOP_FCNTL(vp, com, data, fp->f_flag, kauth_cred_get());
    695   1.37  wrstuden 	return (error);
    696   1.10       cgd }
    697   1.10       cgd 
    698   1.10       cgd /*
    699   1.10       cgd  * File table vnode ioctl routine.
    700   1.10       cgd  */
    701   1.56  gmcgarry static int
    702  1.155        ad vn_ioctl(file_t *fp, u_long com, void *data)
    703   1.10       cgd {
    704  1.191      matt 	struct vnode *vp = fp->f_vnode, *ovp;
    705   1.10       cgd 	struct vattr vattr;
    706   1.10       cgd 	int error;
    707   1.10       cgd 
    708   1.10       cgd 	switch (vp->v_type) {
    709   1.10       cgd 
    710   1.10       cgd 	case VREG:
    711   1.10       cgd 	case VDIR:
    712   1.10       cgd 		if (com == FIONREAD) {
    713  1.183   hannken 			vn_lock(vp, LK_SHARED | LK_RETRY);
    714  1.183   hannken 			error = VOP_GETATTR(vp, &vattr, kauth_cred_get());
    715  1.183   hannken 			VOP_UNLOCK(vp);
    716   1.20  christos 			if (error)
    717   1.10       cgd 				return (error);
    718   1.10       cgd 			*(int *)data = vattr.va_size - fp->f_offset;
    719   1.10       cgd 			return (0);
    720   1.60    atatat 		}
    721   1.82  christos 		if ((com == FIONWRITE) || (com == FIONSPACE)) {
    722   1.81  wrstuden 			/*
    723   1.81  wrstuden 			 * Files don't have send queues, so there never
    724   1.81  wrstuden 			 * are any bytes in them, nor is there any
    725   1.81  wrstuden 			 * open space in them.
    726   1.81  wrstuden 			 */
    727   1.81  wrstuden 			*(int *)data = 0;
    728   1.81  wrstuden 			return (0);
    729   1.81  wrstuden 		}
    730   1.60    atatat 		if (com == FIOGETBMAP) {
    731   1.60    atatat 			daddr_t *block;
    732   1.60    atatat 
    733   1.62    atatat 			if (*(daddr_t *)data < 0)
    734   1.62    atatat 				return (EINVAL);
    735   1.60    atatat 			block = (daddr_t *)data;
    736   1.60    atatat 			return (VOP_BMAP(vp, *block, NULL, block, NULL));
    737   1.61      fvdl 		}
    738   1.61      fvdl 		if (com == OFIOGETBMAP) {
    739   1.61      fvdl 			daddr_t ibn, obn;
    740   1.61      fvdl 
    741   1.62    atatat 			if (*(int32_t *)data < 0)
    742   1.62    atatat 				return (EINVAL);
    743   1.61      fvdl 			ibn = (daddr_t)*(int32_t *)data;
    744   1.61      fvdl 			error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
    745   1.61      fvdl 			*(int32_t *)data = (int32_t)obn;
    746   1.61      fvdl 			return error;
    747   1.10       cgd 		}
    748   1.10       cgd 		if (com == FIONBIO || com == FIOASYNC)	/* XXX */
    749   1.10       cgd 			return (0);			/* XXX */
    750   1.10       cgd 		/* fall into ... */
    751   1.10       cgd 	case VFIFO:
    752   1.10       cgd 	case VCHR:
    753   1.10       cgd 	case VBLK:
    754  1.155        ad 		error = VOP_IOCTL(vp, com, data, fp->f_flag,
    755  1.155        ad 		    kauth_cred_get());
    756   1.10       cgd 		if (error == 0 && com == TIOCSCTTY) {
    757  1.169     pooka 			vref(vp);
    758  1.156        ad 			mutex_enter(proc_lock);
    759  1.155        ad 			ovp = curproc->p_session->s_ttyvp;
    760  1.155        ad 			curproc->p_session->s_ttyvp = vp;
    761  1.156        ad 			mutex_exit(proc_lock);
    762  1.132        ad 			if (ovp != NULL)
    763  1.132        ad 				vrele(ovp);
    764   1.10       cgd 		}
    765   1.10       cgd 		return (error);
    766   1.63  perseant 
    767   1.63  perseant 	default:
    768   1.63  perseant 		return (EPASSTHROUGH);
    769   1.10       cgd 	}
    770   1.10       cgd }
    771   1.10       cgd 
    772   1.10       cgd /*
    773   1.21   mycroft  * File table vnode poll routine.
    774   1.10       cgd  */
    775   1.56  gmcgarry static int
    776  1.155        ad vn_poll(file_t *fp, int events)
    777   1.10       cgd {
    778   1.10       cgd 
    779  1.191      matt 	return (VOP_POLL(fp->f_vnode, events));
    780   1.57  jdolecek }
    781   1.57  jdolecek 
    782   1.57  jdolecek /*
    783   1.57  jdolecek  * File table vnode kqfilter routine.
    784   1.57  jdolecek  */
    785   1.57  jdolecek int
    786  1.155        ad vn_kqfilter(file_t *fp, struct knote *kn)
    787   1.57  jdolecek {
    788   1.57  jdolecek 
    789  1.191      matt 	return (VOP_KQFILTER(fp->f_vnode, kn));
    790   1.28      fvdl }
    791   1.28      fvdl 
    792   1.28      fvdl /*
    793   1.28      fvdl  * Check that the vnode is still valid, and if so
    794   1.28      fvdl  * acquire requested lock.
    795   1.28      fvdl  */
    796   1.28      fvdl int
    797   1.89   thorpej vn_lock(struct vnode *vp, int flags)
    798   1.28      fvdl {
    799   1.28      fvdl 	int error;
    800   1.34  sommerfe 
    801   1.84      yamt #if 0
    802  1.176   hannken 	KASSERT(vp->v_usecount > 0 || (vp->v_iflag & VI_ONWORKLST) != 0);
    803   1.84      yamt #endif
    804  1.176   hannken 	KASSERT((flags & ~(LK_SHARED|LK_EXCLUSIVE|LK_NOWAIT|LK_RETRY)) == 0);
    805  1.181     rmind 	KASSERT(!mutex_owned(vp->v_interlock));
    806   1.80      yamt 
    807  1.159    simonb #ifdef DIAGNOSTIC
    808  1.159    simonb 	if (wapbl_vphaswapbl(vp))
    809  1.159    simonb 		WAPBL_JUNLOCK_ASSERT(wapbl_vptomp(vp));
    810  1.159    simonb #endif
    811  1.159    simonb 
    812  1.189   hannken 	error = VOP_LOCK(vp, flags);
    813  1.189   hannken 	if ((flags & LK_RETRY) != 0 && error == ENOENT)
    814  1.189   hannken 		error = VOP_LOCK(vp, flags);
    815  1.189   hannken 
    816  1.189   hannken 	KASSERT((flags & LK_RETRY) == 0 || (flags & LK_NOWAIT) != 0 ||
    817  1.189   hannken 	    error == 0);
    818  1.189   hannken 
    819  1.189   hannken 	return error;
    820   1.10       cgd }
    821   1.10       cgd 
    822   1.10       cgd /*
    823   1.10       cgd  * File table vnode close routine.
    824   1.10       cgd  */
    825   1.56  gmcgarry static int
    826  1.155        ad vn_closefile(file_t *fp)
    827   1.10       cgd {
    828   1.10       cgd 
    829  1.191      matt 	return vn_close(fp->f_vnode, fp->f_flag, fp->f_cred);
    830   1.39      fvdl }
    831   1.39      fvdl 
    832   1.39      fvdl /*
    833   1.85   thorpej  * Simplified in-kernel wrapper calls for extended attribute access.
    834   1.85   thorpej  * Both calls pass in a NULL credential, authorizing a "kernel" access.
    835   1.85   thorpej  * Set IO_NODELOCKED in ioflg if the vnode is already locked.
    836   1.85   thorpej  */
    837   1.85   thorpej int
    838   1.85   thorpej vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
    839  1.101  christos     const char *attrname, size_t *buflen, void *bf, struct lwp *l)
    840   1.85   thorpej {
    841   1.85   thorpej 	struct uio auio;
    842   1.85   thorpej 	struct iovec aiov;
    843   1.85   thorpej 	int error;
    844   1.85   thorpej 
    845   1.85   thorpej 	aiov.iov_len = *buflen;
    846   1.88  christos 	aiov.iov_base = bf;
    847   1.85   thorpej 
    848   1.85   thorpej 	auio.uio_iov = &aiov;
    849   1.85   thorpej 	auio.uio_iovcnt = 1;
    850   1.85   thorpej 	auio.uio_rw = UIO_READ;
    851   1.85   thorpej 	auio.uio_offset = 0;
    852   1.85   thorpej 	auio.uio_resid = *buflen;
    853  1.106      yamt 	UIO_SETUP_SYSSPACE(&auio);
    854   1.85   thorpej 
    855   1.85   thorpej 	if ((ioflg & IO_NODELOCKED) == 0)
    856   1.85   thorpej 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    857   1.86     perry 
    858  1.144     pooka 	error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL);
    859   1.86     perry 
    860   1.85   thorpej 	if ((ioflg & IO_NODELOCKED) == 0)
    861  1.174   hannken 		VOP_UNLOCK(vp);
    862   1.86     perry 
    863   1.85   thorpej 	if (error == 0)
    864   1.85   thorpej 		*buflen = *buflen - auio.uio_resid;
    865   1.86     perry 
    866   1.85   thorpej 	return (error);
    867   1.85   thorpej }
    868   1.85   thorpej 
    869   1.85   thorpej /*
    870   1.85   thorpej  * XXX Failure mode if partially written?
    871   1.85   thorpej  */
    872   1.85   thorpej int
    873   1.85   thorpej vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
    874  1.101  christos     const char *attrname, size_t buflen, const void *bf, struct lwp *l)
    875   1.85   thorpej {
    876   1.85   thorpej 	struct uio auio;
    877   1.85   thorpej 	struct iovec aiov;
    878   1.85   thorpej 	int error;
    879   1.85   thorpej 
    880   1.85   thorpej 	aiov.iov_len = buflen;
    881   1.88  christos 	aiov.iov_base = __UNCONST(bf);		/* XXXUNCONST kills const */
    882   1.85   thorpej 
    883   1.85   thorpej 	auio.uio_iov = &aiov;
    884   1.85   thorpej 	auio.uio_iovcnt = 1;
    885   1.85   thorpej 	auio.uio_rw = UIO_WRITE;
    886   1.85   thorpej 	auio.uio_offset = 0;
    887   1.85   thorpej 	auio.uio_resid = buflen;
    888  1.106      yamt 	UIO_SETUP_SYSSPACE(&auio);
    889   1.85   thorpej 
    890   1.85   thorpej 	if ((ioflg & IO_NODELOCKED) == 0) {
    891   1.85   thorpej 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    892   1.85   thorpej 	}
    893   1.85   thorpej 
    894  1.144     pooka 	error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL);
    895   1.85   thorpej 
    896   1.85   thorpej 	if ((ioflg & IO_NODELOCKED) == 0) {
    897  1.174   hannken 		VOP_UNLOCK(vp);
    898   1.85   thorpej 	}
    899   1.85   thorpej 
    900   1.85   thorpej 	return (error);
    901   1.85   thorpej }
    902   1.85   thorpej 
    903   1.85   thorpej int
    904   1.85   thorpej vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
    905  1.101  christos     const char *attrname, struct lwp *l)
    906   1.85   thorpej {
    907   1.85   thorpej 	int error;
    908   1.85   thorpej 
    909   1.85   thorpej 	if ((ioflg & IO_NODELOCKED) == 0) {
    910   1.85   thorpej 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
    911   1.85   thorpej 	}
    912   1.85   thorpej 
    913  1.144     pooka 	error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL);
    914   1.85   thorpej 	if (error == EOPNOTSUPP)
    915  1.144     pooka 		error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL);
    916   1.86     perry 
    917   1.85   thorpej 	if ((ioflg & IO_NODELOCKED) == 0) {
    918  1.174   hannken 		VOP_UNLOCK(vp);
    919   1.85   thorpej 	}
    920   1.85   thorpej 
    921   1.85   thorpej 	return (error);
    922   1.85   thorpej }
    923   1.95      yamt 
    924  1.100      yamt void
    925  1.100      yamt vn_ra_allocctx(struct vnode *vp)
    926  1.100      yamt {
    927  1.100      yamt 	struct uvm_ractx *ra = NULL;
    928  1.100      yamt 
    929  1.181     rmind 	KASSERT(mutex_owned(vp->v_interlock));
    930  1.149        ad 
    931  1.100      yamt 	if (vp->v_type != VREG) {
    932  1.100      yamt 		return;
    933  1.100      yamt 	}
    934  1.100      yamt 	if (vp->v_ractx != NULL) {
    935  1.100      yamt 		return;
    936  1.100      yamt 	}
    937  1.100      yamt 	if (vp->v_ractx == NULL) {
    938  1.181     rmind 		mutex_exit(vp->v_interlock);
    939  1.100      yamt 		ra = uvm_ra_allocctx();
    940  1.181     rmind 		mutex_enter(vp->v_interlock);
    941  1.100      yamt 		if (ra != NULL && vp->v_ractx == NULL) {
    942  1.100      yamt 			vp->v_ractx = ra;
    943  1.100      yamt 			ra = NULL;
    944  1.100      yamt 		}
    945  1.100      yamt 	}
    946  1.100      yamt 	if (ra != NULL) {
    947  1.100      yamt 		uvm_ra_freectx(ra);
    948  1.100      yamt 	}
    949  1.100      yamt }
    950  1.170     pooka 
    951  1.170     pooka int
    952  1.170     pooka vn_fifo_bypass(void *v)
    953  1.170     pooka {
    954  1.170     pooka 	struct vop_generic_args *ap = v;
    955  1.170     pooka 
    956  1.170     pooka 	return VOCALL(fifo_vnodeop_p, ap->a_desc->vdesc_offset, v);
    957  1.170     pooka }
    958