The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/vfs_syscalls.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*
    2  * Copyright (c) 1989, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  * (c) UNIX System Laboratories, Inc.
    5  * All or some portions of this file are derived from material licensed
    6  * to the University of California by American Telephone and Telegraph
    7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
    8  * the permission of UNIX System Laboratories, Inc.
    9  *
   10  * Redistribution and use in source and binary forms, with or without
   11  * modification, are permitted provided that the following conditions
   12  * are met:
   13  * 1. Redistributions of source code must retain the above copyright
   14  *    notice, this list of conditions and the following disclaimer.
   15  * 2. Redistributions in binary form must reproduce the above copyright
   16  *    notice, this list of conditions and the following disclaimer in the
   17  *    documentation and/or other materials provided with the distribution.
   18  * 3. All advertising materials mentioning features or use of this software
   19  *    must display the following acknowledgement:
   20  *      This product includes software developed by the University of
   21  *      California, Berkeley and its contributors.
   22  * 4. Neither the name of the University nor the names of its contributors
   23  *    may be used to endorse or promote products derived from this software
   24  *    without specific prior written permission.
   25  *
   26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   36  * SUCH DAMAGE.
   37  *
   38  *      @(#)vfs_syscalls.c      8.13 (Berkeley) 4/15/94
   39  * $FreeBSD: releng/5.0/sys/kern/vfs_syscalls.c 108822 2003-01-06 21:20:54Z nectar $
   40  */
   41 
   42 /* For 4.3 integer FS ID compatibility */
   43 #include "opt_compat.h"
   44 #include "opt_mac.h"
   45 
   46 #include <sys/param.h>
   47 #include <sys/systm.h>
   48 #include <sys/bio.h>
   49 #include <sys/buf.h>
   50 #include <sys/sysent.h>
   51 #include <sys/mac.h>
   52 #include <sys/malloc.h>
   53 #include <sys/mount.h>
   54 #include <sys/mutex.h>
   55 #include <sys/sysproto.h>
   56 #include <sys/namei.h>
   57 #include <sys/filedesc.h>
   58 #include <sys/kernel.h>
   59 #include <sys/fcntl.h>
   60 #include <sys/file.h>
   61 #include <sys/linker.h>
   62 #include <sys/stat.h>
   63 #include <sys/sx.h>
   64 #include <sys/unistd.h>
   65 #include <sys/vnode.h>
   66 #include <sys/proc.h>
   67 #include <sys/dirent.h>
   68 #include <sys/extattr.h>
   69 #include <sys/jail.h>
   70 #include <sys/syscallsubr.h>
   71 #include <sys/sysctl.h>
   72 
   73 #include <machine/limits.h>
   74 #include <machine/stdarg.h>
   75 
   76 #include <vm/vm.h>
   77 #include <vm/vm_object.h>
   78 #include <vm/vm_page.h>
   79 #include <vm/uma.h>
   80 
   81 static int change_dir(struct nameidata *ndp, struct thread *td);
   82 static int chroot_refuse_vdir_fds(struct filedesc *fdp);
   83 static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
   84 static int setfown(struct thread *td, struct vnode *, uid_t, gid_t);
   85 static int setfmode(struct thread *td, struct vnode *, int);
   86 static int setfflags(struct thread *td, struct vnode *, int);
   87 static int setutimes(struct thread *td, struct vnode *,
   88     const struct timespec *, int, int);
   89 static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
   90     struct thread *td);
   91 
   92 int (*union_dircheckp)(struct thread *td, struct vnode **, struct file *);
   93 int (*softdep_fsync_hook)(struct vnode *);
   94 
   95 /*
   96  * The module initialization routine for POSIX asynchronous I/O will
   97  * set this to the version of AIO that it implements.  (Zero means
   98  * that it is not implemented.)  This value is used here by pathconf()
   99  * and in kern_descrip.c by fpathconf().
  100  */
  101 int async_io_version;
  102 
  103 /*
  104  * Sync each mounted filesystem.
  105  */
  106 #ifndef _SYS_SYSPROTO_H_
  107 struct sync_args {
  108         int     dummy;
  109 };
  110 #endif
  111 
  112 #ifdef DEBUG
  113 static int syncprt = 0;
  114 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, "");
  115 #endif
  116 
  117 /* ARGSUSED */
  118 int
  119 sync(td, uap)
  120         struct thread *td;
  121         struct sync_args *uap;
  122 {
  123         struct mount *mp, *nmp;
  124         int asyncflag;
  125 
  126         mtx_lock(&mountlist_mtx);
  127         for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
  128                 if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
  129                         nmp = TAILQ_NEXT(mp, mnt_list);
  130                         continue;
  131                 }
  132                 if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
  133                     vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
  134                         asyncflag = mp->mnt_flag & MNT_ASYNC;
  135                         mp->mnt_flag &= ~MNT_ASYNC;
  136                         vfs_msync(mp, MNT_NOWAIT);
  137                         VFS_SYNC(mp, MNT_NOWAIT,
  138                             ((td != NULL) ? td->td_ucred : NOCRED), td);
  139                         mp->mnt_flag |= asyncflag;
  140                         vn_finished_write(mp);
  141                 }
  142                 mtx_lock(&mountlist_mtx);
  143                 nmp = TAILQ_NEXT(mp, mnt_list);
  144                 vfs_unbusy(mp, td);
  145         }
  146         mtx_unlock(&mountlist_mtx);
  147 #if 0
  148 /*
  149  * XXX don't call vfs_bufstats() yet because that routine
  150  * was not imported in the Lite2 merge.
  151  */
  152 #ifdef DIAGNOSTIC
  153         if (syncprt)
  154                 vfs_bufstats();
  155 #endif /* DIAGNOSTIC */
  156 #endif
  157         return (0);
  158 }
  159 
  160 /* XXX PRISON: could be per prison flag */
  161 static int prison_quotas;
  162 #if 0
  163 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, "");
  164 #endif
  165 
  166 /*
  167  * Change filesystem quotas.
  168  */
  169 #ifndef _SYS_SYSPROTO_H_
  170 struct quotactl_args {
  171         char *path;
  172         int cmd;
  173         int uid;
  174         caddr_t arg;
  175 };
  176 #endif
  177 /* ARGSUSED */
  178 int
  179 quotactl(td, uap)
  180         struct thread *td;
  181         register struct quotactl_args /* {
  182                 char *path;
  183                 int cmd;
  184                 int uid;
  185                 caddr_t arg;
  186         } */ *uap;
  187 {
  188         struct mount *mp;
  189         int error;
  190         struct nameidata nd;
  191 
  192         if (jailed(td->td_ucred) && !prison_quotas)
  193                 return (EPERM);
  194         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
  195         if ((error = namei(&nd)) != 0)
  196                 return (error);
  197         NDFREE(&nd, NDF_ONLY_PNBUF);
  198         error = vn_start_write(nd.ni_vp, &mp, V_WAIT | PCATCH);
  199         vrele(nd.ni_vp);
  200         if (error)
  201                 return (error);
  202         error = VFS_QUOTACTL(mp, uap->cmd, uap->uid,
  203             uap->arg, td);
  204         vn_finished_write(mp);
  205         return (error);
  206 }
  207 
  208 /*
  209  * Get filesystem statistics.
  210  */
  211 #ifndef _SYS_SYSPROTO_H_
  212 struct statfs_args {
  213         char *path;
  214         struct statfs *buf;
  215 };
  216 #endif
  217 /* ARGSUSED */
  218 int
  219 statfs(td, uap)
  220         struct thread *td;
  221         register struct statfs_args /* {
  222                 char *path;
  223                 struct statfs *buf;
  224         } */ *uap;
  225 {
  226         register struct mount *mp;
  227         register struct statfs *sp;
  228         int error;
  229         struct nameidata nd;
  230         struct statfs sb;
  231 
  232         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
  233         if ((error = namei(&nd)) != 0)
  234                 return (error);
  235         mp = nd.ni_vp->v_mount;
  236         sp = &mp->mnt_stat;
  237         NDFREE(&nd, NDF_ONLY_PNBUF);
  238         vrele(nd.ni_vp);
  239 #ifdef MAC
  240         error = mac_check_mount_stat(td->td_ucred, mp);
  241         if (error)
  242                 return (error);
  243 #endif
  244         error = VFS_STATFS(mp, sp, td);
  245         if (error)
  246                 return (error);
  247         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  248         if (suser(td)) {
  249                 bcopy(sp, &sb, sizeof(sb));
  250                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
  251                 sp = &sb;
  252         }
  253         return (copyout(sp, uap->buf, sizeof(*sp)));
  254 }
  255 
  256 /*
  257  * Get filesystem statistics.
  258  */
  259 #ifndef _SYS_SYSPROTO_H_
  260 struct fstatfs_args {
  261         int fd;
  262         struct statfs *buf;
  263 };
  264 #endif
  265 /* ARGSUSED */
  266 int
  267 fstatfs(td, uap)
  268         struct thread *td;
  269         register struct fstatfs_args /* {
  270                 int fd;
  271                 struct statfs *buf;
  272         } */ *uap;
  273 {
  274         struct file *fp;
  275         struct mount *mp;
  276         register struct statfs *sp;
  277         int error;
  278         struct statfs sb;
  279 
  280         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
  281                 return (error);
  282         mp = ((struct vnode *)fp->f_data)->v_mount;
  283         fdrop(fp, td);
  284         if (mp == NULL)
  285                 return (EBADF);
  286 #ifdef MAC
  287         error = mac_check_mount_stat(td->td_ucred, mp);
  288         if (error)
  289                 return (error);
  290 #endif
  291         sp = &mp->mnt_stat;
  292         error = VFS_STATFS(mp, sp, td);
  293         if (error)
  294                 return (error);
  295         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  296         if (suser(td)) {
  297                 bcopy(sp, &sb, sizeof(sb));
  298                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
  299                 sp = &sb;
  300         }
  301         return (copyout(sp, uap->buf, sizeof(*sp)));
  302 }
  303 
  304 /*
  305  * Get statistics on all filesystems.
  306  */
  307 #ifndef _SYS_SYSPROTO_H_
  308 struct getfsstat_args {
  309         struct statfs *buf;
  310         long bufsize;
  311         int flags;
  312 };
  313 #endif
  314 int
  315 getfsstat(td, uap)
  316         struct thread *td;
  317         register struct getfsstat_args /* {
  318                 struct statfs *buf;
  319                 long bufsize;
  320                 int flags;
  321         } */ *uap;
  322 {
  323         register struct mount *mp, *nmp;
  324         register struct statfs *sp;
  325         caddr_t sfsp;
  326         long count, maxcount, error;
  327 
  328         maxcount = uap->bufsize / sizeof(struct statfs);
  329         sfsp = (caddr_t)uap->buf;
  330         count = 0;
  331         mtx_lock(&mountlist_mtx);
  332         for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
  333 #ifdef MAC
  334                 if (mac_check_mount_stat(td->td_ucred, mp) != 0) {
  335                         nmp = TAILQ_NEXT(mp, mnt_list);
  336                         continue;
  337                 }
  338 #endif
  339                 if (vfs_busy(mp, LK_NOWAIT, &mountlist_mtx, td)) {
  340                         nmp = TAILQ_NEXT(mp, mnt_list);
  341                         continue;
  342                 }
  343                 if (sfsp && count < maxcount) {
  344                         sp = &mp->mnt_stat;
  345                         /*
  346                          * If MNT_NOWAIT or MNT_LAZY is specified, do not
  347                          * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
  348                          * overrides MNT_WAIT.
  349                          */
  350                         if (((uap->flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
  351                             (uap->flags & MNT_WAIT)) &&
  352                             (error = VFS_STATFS(mp, sp, td))) {
  353                                 mtx_lock(&mountlist_mtx);
  354                                 nmp = TAILQ_NEXT(mp, mnt_list);
  355                                 vfs_unbusy(mp, td);
  356                                 continue;
  357                         }
  358                         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
  359                         error = copyout(sp, sfsp, sizeof(*sp));
  360                         if (error) {
  361                                 vfs_unbusy(mp, td);
  362                                 return (error);
  363                         }
  364                         sfsp += sizeof(*sp);
  365                 }
  366                 count++;
  367                 mtx_lock(&mountlist_mtx);
  368                 nmp = TAILQ_NEXT(mp, mnt_list);
  369                 vfs_unbusy(mp, td);
  370         }
  371         mtx_unlock(&mountlist_mtx);
  372         if (sfsp && count > maxcount)
  373                 td->td_retval[0] = maxcount;
  374         else
  375                 td->td_retval[0] = count;
  376         return (0);
  377 }
  378 
  379 /*
  380  * Change current working directory to a given file descriptor.
  381  */
  382 #ifndef _SYS_SYSPROTO_H_
  383 struct fchdir_args {
  384         int     fd;
  385 };
  386 #endif
  387 /* ARGSUSED */
  388 int
  389 fchdir(td, uap)
  390         struct thread *td;
  391         struct fchdir_args /* {
  392                 int fd;
  393         } */ *uap;
  394 {
  395         register struct filedesc *fdp = td->td_proc->p_fd;
  396         struct vnode *vp, *tdp, *vpold;
  397         struct mount *mp;
  398         struct file *fp;
  399         int error;
  400 
  401         if ((error = getvnode(fdp, uap->fd, &fp)) != 0)
  402                 return (error);
  403         vp = (struct vnode *)fp->f_data;
  404         VREF(vp);
  405         fdrop(fp, td);
  406         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
  407         if (vp->v_type != VDIR)
  408                 error = ENOTDIR;
  409 #ifdef MAC
  410         else if ((error = mac_check_vnode_chdir(td->td_ucred, vp)) != 0) {
  411         }
  412 #endif
  413         else
  414                 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
  415         while (!error && (mp = vp->v_mountedhere) != NULL) {
  416                 if (vfs_busy(mp, 0, 0, td))
  417                         continue;
  418                 error = VFS_ROOT(mp, &tdp);
  419                 vfs_unbusy(mp, td);
  420                 if (error)
  421                         break;
  422                 vput(vp);
  423                 vp = tdp;
  424         }
  425         if (error) {
  426                 vput(vp);
  427                 return (error);
  428         }
  429         VOP_UNLOCK(vp, 0, td);
  430         FILEDESC_LOCK(fdp);
  431         vpold = fdp->fd_cdir;
  432         fdp->fd_cdir = vp;
  433         FILEDESC_UNLOCK(fdp);
  434         vrele(vpold);
  435         return (0);
  436 }
  437 
  438 /*
  439  * Change current working directory (``.'').
  440  */
  441 #ifndef _SYS_SYSPROTO_H_
  442 struct chdir_args {
  443         char    *path;
  444 };
  445 #endif
  446 /* ARGSUSED */
  447 int
  448 chdir(td, uap)
  449         struct thread *td;
  450         struct chdir_args /* {
  451                 char *path;
  452         } */ *uap;
  453 {
  454 
  455         return (kern_chdir(td, uap->path, UIO_USERSPACE));
  456 }
  457 
  458 int
  459 kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
  460 {
  461         register struct filedesc *fdp = td->td_proc->p_fd;
  462         int error;
  463         struct nameidata nd;
  464         struct vnode *vp;
  465 
  466         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, pathseg, path, td);
  467         if ((error = change_dir(&nd, td)) != 0)
  468                 return (error);
  469         NDFREE(&nd, NDF_ONLY_PNBUF);
  470         FILEDESC_LOCK(fdp);
  471         vp = fdp->fd_cdir;
  472         fdp->fd_cdir = nd.ni_vp;
  473         FILEDESC_UNLOCK(fdp);
  474         vrele(vp);
  475         return (0);
  476 }
  477 
  478 /*
  479  * Helper function for raised chroot(2) security function:  Refuse if
  480  * any filedescriptors are open directories.
  481  */
  482 static int
  483 chroot_refuse_vdir_fds(fdp)
  484         struct filedesc *fdp;
  485 {
  486         struct vnode *vp;
  487         struct file *fp;
  488         int fd;
  489 
  490         FILEDESC_LOCK_ASSERT(fdp, MA_OWNED);
  491         for (fd = 0; fd < fdp->fd_nfiles ; fd++) {
  492                 fp = fget_locked(fdp, fd);
  493                 if (fp == NULL)
  494                         continue;
  495                 if (fp->f_type == DTYPE_VNODE) {
  496                         vp = (struct vnode *)fp->f_data;
  497                         if (vp->v_type == VDIR)
  498                                 return (EPERM);
  499                 }
  500         }
  501         return (0);
  502 }
  503 
  504 /*
  505  * This sysctl determines if we will allow a process to chroot(2) if it
  506  * has a directory open:
  507  *      0: disallowed for all processes.
  508  *      1: allowed for processes that were not already chroot(2)'ed.
  509  *      2: allowed for all processes.
  510  */
  511 
  512 static int chroot_allow_open_directories = 1;
  513 
  514 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
  515      &chroot_allow_open_directories, 0, "");
  516 
  517 /*
  518  * Change notion of root (``/'') directory.
  519  */
  520 #ifndef _SYS_SYSPROTO_H_
  521 struct chroot_args {
  522         char    *path;
  523 };
  524 #endif
  525 /* ARGSUSED */
  526 int
  527 chroot(td, uap)
  528         struct thread *td;
  529         struct chroot_args /* {
  530                 char *path;
  531         } */ *uap;
  532 {
  533         register struct filedesc *fdp = td->td_proc->p_fd;
  534         int error;
  535         struct nameidata nd;
  536         struct vnode *vp;
  537 
  538         error = suser_cred(td->td_ucred, PRISON_ROOT);
  539         if (error)
  540                 return (error);
  541         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
  542             uap->path, td);
  543         mtx_lock(&Giant);
  544         if ((error = change_dir(&nd, td)) != 0)
  545                 goto error;
  546 #ifdef MAC
  547         if ((error = mac_check_vnode_chroot(td->td_ucred, nd.ni_vp)))
  548                 goto error;
  549 #endif
  550         FILEDESC_LOCK(fdp);
  551         if (chroot_allow_open_directories == 0 ||
  552             (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
  553                 error = chroot_refuse_vdir_fds(fdp);
  554                 if (error)
  555                         goto error_unlock;
  556         }
  557         vp = fdp->fd_rdir;
  558         fdp->fd_rdir = nd.ni_vp;
  559         if (!fdp->fd_jdir) {
  560                 fdp->fd_jdir = nd.ni_vp;
  561                 VREF(fdp->fd_jdir);
  562         }
  563         FILEDESC_UNLOCK(fdp);
  564         NDFREE(&nd, NDF_ONLY_PNBUF);
  565         vrele(vp);
  566         mtx_unlock(&Giant);
  567         return (0);
  568 error_unlock:
  569         FILEDESC_UNLOCK(fdp);
  570 error:
  571         mtx_unlock(&Giant);
  572         NDFREE(&nd, 0);
  573         return (error);
  574 }
  575 
  576 /*
  577  * Common routine for chroot and chdir.
  578  */
  579 static int
  580 change_dir(ndp, td)
  581         register struct nameidata *ndp;
  582         struct thread *td;
  583 {
  584         struct vnode *vp;
  585         int error;
  586 
  587         error = namei(ndp);
  588         if (error)
  589                 return (error);
  590         vp = ndp->ni_vp;
  591         if (vp->v_type != VDIR)
  592                 error = ENOTDIR;
  593 #ifdef MAC
  594         else if ((error = mac_check_vnode_chdir(td->td_ucred, vp)) != 0) {
  595         }
  596 #endif
  597         else
  598                 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
  599         if (error)
  600                 vput(vp);
  601         else
  602                 VOP_UNLOCK(vp, 0, td);
  603         return (error);
  604 }
  605 
  606 /*
  607  * Check permissions, allocate an open file structure,
  608  * and call the device open routine if any.
  609  */
  610 #ifndef _SYS_SYSPROTO_H_
  611 struct open_args {
  612         char    *path;
  613         int     flags;
  614         int     mode;
  615 };
  616 #endif
  617 int
  618 open(td, uap)
  619         struct thread *td;
  620         register struct open_args /* {
  621                 char *path;
  622                 int flags;
  623                 int mode;
  624         } */ *uap;
  625 {
  626 
  627         return (kern_open(td, uap->path, UIO_USERSPACE, uap->flags, uap->mode));
  628 }
  629 
  630 int
  631 kern_open(struct thread *td, char *path, enum uio_seg pathseg, int flags,
  632     int mode)
  633 {
  634         struct proc *p = td->td_proc;
  635         struct filedesc *fdp = p->p_fd;
  636         struct file *fp;
  637         struct vnode *vp;
  638         struct vattr vat;
  639         struct mount *mp;
  640         int cmode, oflags;
  641         struct file *nfp;
  642         int type, indx, error;
  643         struct flock lf;
  644         struct nameidata nd;
  645 
  646         if ((flags & O_ACCMODE) == O_ACCMODE)
  647                 return (EINVAL);
  648         oflags = flags;
  649         flags = FFLAGS(flags);
  650         error = falloc(td, &nfp, &indx);
  651         if (error)
  652                 return (error);
  653         fp = nfp;
  654         FILEDESC_LOCK(fdp);
  655         cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT;
  656         FILEDESC_UNLOCK(fdp);
  657         NDINIT(&nd, LOOKUP, FOLLOW, pathseg, path, td);
  658         td->td_dupfd = -indx - 1;               /* XXX check for fdopen */
  659         /*
  660          * Bump the ref count to prevent another process from closing
  661          * the descriptor while we are blocked in vn_open()
  662          */
  663         fhold(fp);
  664         error = vn_open(&nd, &flags, cmode);
  665         if (error) {
  666                 /*
  667                  * release our own reference
  668                  */
  669                 fdrop(fp, td);
  670 
  671                 /*
  672                  * handle special fdopen() case.  bleh.  dupfdopen() is
  673                  * responsible for dropping the old contents of ofiles[indx]
  674                  * if it succeeds.
  675                  */
  676                 if ((error == ENODEV || error == ENXIO) &&
  677                     td->td_dupfd >= 0 &&                /* XXX from fdopen */
  678                     (error =
  679                         dupfdopen(td, fdp, indx, td->td_dupfd, flags, error)) == 0) {
  680                         td->td_retval[0] = indx;
  681                         return (0);
  682                 }
  683                 /*
  684                  * Clean up the descriptor, but only if another thread hadn't
  685                  * replaced or closed it.
  686                  */
  687                 FILEDESC_LOCK(fdp);
  688                 if (fdp->fd_ofiles[indx] == fp) {
  689                         fdp->fd_ofiles[indx] = NULL;
  690                         FILEDESC_UNLOCK(fdp);
  691                         fdrop(fp, td);
  692                 } else
  693                         FILEDESC_UNLOCK(fdp);
  694 
  695                 if (error == ERESTART)
  696                         error = EINTR;
  697                 return (error);
  698         }
  699         td->td_dupfd = 0;
  700         NDFREE(&nd, NDF_ONLY_PNBUF);
  701         vp = nd.ni_vp;
  702 
  703         /*
  704          * There should be 2 references on the file, one from the descriptor
  705          * table, and one for us.
  706          *
  707          * Handle the case where someone closed the file (via its file
  708          * descriptor) while we were blocked.  The end result should look
  709          * like opening the file succeeded but it was immediately closed.
  710          */
  711         FILEDESC_LOCK(fdp);
  712         FILE_LOCK(fp);
  713         if (fp->f_count == 1) {
  714                 KASSERT(fdp->fd_ofiles[indx] != fp,
  715                     ("Open file descriptor lost all refs"));
  716                 FILEDESC_UNLOCK(fdp);
  717                 FILE_UNLOCK(fp);
  718                 VOP_UNLOCK(vp, 0, td);
  719                 vn_close(vp, flags & FMASK, fp->f_cred, td);
  720                 fdrop(fp, td);
  721                 td->td_retval[0] = indx;
  722                 return 0;
  723         }
  724 
  725         /* assert that vn_open created a backing object if one is needed */
  726         KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0,
  727                 ("open: vmio vnode has no backing object after vn_open"));
  728 
  729         fp->f_data = vp;
  730         fp->f_flag = flags & FMASK;
  731         fp->f_ops = &vnops;
  732         fp->f_type = (vp->v_type == VFIFO ? DTYPE_FIFO : DTYPE_VNODE);
  733         FILEDESC_UNLOCK(fdp);
  734         FILE_UNLOCK(fp);
  735         VOP_UNLOCK(vp, 0, td);
  736         if (flags & (O_EXLOCK | O_SHLOCK)) {
  737                 lf.l_whence = SEEK_SET;
  738                 lf.l_start = 0;
  739                 lf.l_len = 0;
  740                 if (flags & O_EXLOCK)
  741                         lf.l_type = F_WRLCK;
  742                 else
  743                         lf.l_type = F_RDLCK;
  744                 type = F_FLOCK;
  745                 if ((flags & FNONBLOCK) == 0)
  746                         type |= F_WAIT;
  747                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
  748                             type)) != 0)
  749                         goto bad;
  750                 fp->f_flag |= FHASLOCK;
  751         }
  752         if (flags & O_TRUNC) {
  753                 if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
  754                         goto bad;
  755                 VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
  756                 VATTR_NULL(&vat);
  757                 vat.va_size = 0;
  758                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
  759 #ifdef MAC
  760                 error = mac_check_vnode_write(td->td_ucred, fp->f_cred, vp);
  761                 if (error == 0)
  762 #endif
  763                         error = VOP_SETATTR(vp, &vat, td->td_ucred, td);
  764                 VOP_UNLOCK(vp, 0, td);
  765                 vn_finished_write(mp);
  766                 if (error)
  767                         goto bad;
  768         }
  769         /*
  770          * Release our private reference, leaving the one associated with
  771          * the descriptor table intact.
  772          */
  773         fdrop(fp, td);
  774         td->td_retval[0] = indx;
  775         return (0);
  776 bad:
  777         FILEDESC_LOCK(fdp);
  778         if (fdp->fd_ofiles[indx] == fp) {
  779                 fdp->fd_ofiles[indx] = NULL;
  780                 FILEDESC_UNLOCK(fdp);
  781                 fdrop(fp, td);
  782         } else
  783                 FILEDESC_UNLOCK(fdp);
  784         fdrop(fp, td);
  785         return (error);
  786 }
  787 
  788 #ifdef COMPAT_43
  789 /*
  790  * Create a file.
  791  */
  792 #ifndef _SYS_SYSPROTO_H_
  793 struct ocreat_args {
  794         char    *path;
  795         int     mode;
  796 };
  797 #endif
  798 int
  799 ocreat(td, uap)
  800         struct thread *td;
  801         register struct ocreat_args /* {
  802                 char *path;
  803                 int mode;
  804         } */ *uap;
  805 {
  806         struct open_args /* {
  807                 char *path;
  808                 int flags;
  809                 int mode;
  810         } */ nuap;
  811 
  812         nuap.path = uap->path;
  813         nuap.mode = uap->mode;
  814         nuap.flags = O_WRONLY | O_CREAT | O_TRUNC;
  815         return (open(td, &nuap));
  816 }
  817 #endif /* COMPAT_43 */
  818 
  819 /*
  820  * Create a special file.
  821  */
  822 #ifndef _SYS_SYSPROTO_H_
  823 struct mknod_args {
  824         char    *path;
  825         int     mode;
  826         int     dev;
  827 };
  828 #endif
  829 /* ARGSUSED */
  830 int
  831 mknod(td, uap)
  832         struct thread *td;
  833         register struct mknod_args /* {
  834                 char *path;
  835                 int mode;
  836                 int dev;
  837         } */ *uap;
  838 {
  839 
  840         return (kern_mknod(td, uap->path, UIO_USERSPACE, uap->mode, uap->dev));
  841 }
  842 
  843 int
  844 kern_mknod(struct thread *td, char *path, enum uio_seg pathseg, int mode,
  845     int dev)
  846 {
  847         struct vnode *vp;
  848         struct mount *mp;
  849         struct vattr vattr;
  850         int error;
  851         int whiteout = 0;
  852         struct nameidata nd;
  853 
  854         switch (mode & S_IFMT) {
  855         case S_IFCHR:
  856         case S_IFBLK:
  857                 error = suser(td);
  858                 break;
  859         default:
  860                 error = suser_cred(td->td_ucred, PRISON_ROOT);
  861                 break;
  862         }
  863         if (error)
  864                 return (error);
  865 restart:
  866         bwillwrite();
  867         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME, pathseg, path, td);
  868         if ((error = namei(&nd)) != 0)
  869                 return (error);
  870         vp = nd.ni_vp;
  871         if (vp != NULL) {
  872                 vrele(vp);
  873                 error = EEXIST;
  874         } else {
  875                 VATTR_NULL(&vattr);
  876                 FILEDESC_LOCK(td->td_proc->p_fd);
  877                 vattr.va_mode = (mode & ALLPERMS) &
  878                     ~td->td_proc->p_fd->fd_cmask;
  879                 FILEDESC_UNLOCK(td->td_proc->p_fd);
  880                 vattr.va_rdev = dev;
  881                 whiteout = 0;
  882 
  883                 switch (mode & S_IFMT) {
  884                 case S_IFMT:    /* used by badsect to flag bad sectors */
  885                         vattr.va_type = VBAD;
  886                         break;
  887                 case S_IFCHR:
  888                         vattr.va_type = VCHR;
  889                         break;
  890                 case S_IFBLK:
  891                         vattr.va_type = VBLK;
  892                         break;
  893                 case S_IFWHT:
  894                         whiteout = 1;
  895                         break;
  896                 default:
  897                         error = EINVAL;
  898                         break;
  899                 }
  900         }
  901         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
  902                 NDFREE(&nd, NDF_ONLY_PNBUF);
  903                 vput(nd.ni_dvp);
  904                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
  905                         return (error);
  906                 goto restart;
  907         }
  908 #ifdef MAC
  909         if (error == 0 && !whiteout)
  910                 error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp,
  911                     &nd.ni_cnd, &vattr);
  912 #endif
  913         if (!error) {
  914                 VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
  915                 if (whiteout)
  916                         error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
  917                 else {
  918                         error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
  919                                                 &nd.ni_cnd, &vattr);
  920                         if (error == 0)
  921                                 vput(nd.ni_vp);
  922                 }
  923         }
  924         NDFREE(&nd, NDF_ONLY_PNBUF);
  925         vput(nd.ni_dvp);
  926         vn_finished_write(mp);
  927         ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mknod");
  928         ASSERT_VOP_UNLOCKED(nd.ni_vp, "mknod");
  929         return (error);
  930 }
  931 
  932 /*
  933  * Create a named pipe.
  934  */
  935 #ifndef _SYS_SYSPROTO_H_
  936 struct mkfifo_args {
  937         char    *path;
  938         int     mode;
  939 };
  940 #endif
  941 /* ARGSUSED */
  942 int
  943 mkfifo(td, uap)
  944         struct thread *td;
  945         register struct mkfifo_args /* {
  946                 char *path;
  947                 int mode;
  948         } */ *uap;
  949 {
  950 
  951         return (kern_mkfifo(td, uap->path, UIO_USERSPACE, uap->mode));
  952 }
  953 
  954 int
  955 kern_mkfifo(struct thread *td, char *path, enum uio_seg pathseg, int mode)
  956 {
  957         struct mount *mp;
  958         struct vattr vattr;
  959         int error;
  960         struct nameidata nd;
  961 
  962 restart:
  963         bwillwrite();
  964         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME, pathseg, path, td);
  965         if ((error = namei(&nd)) != 0)
  966                 return (error);
  967         if (nd.ni_vp != NULL) {
  968                 NDFREE(&nd, NDF_ONLY_PNBUF);
  969                 vrele(nd.ni_vp);
  970                 vput(nd.ni_dvp);
  971                 return (EEXIST);
  972         }
  973         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
  974                 NDFREE(&nd, NDF_ONLY_PNBUF);
  975                 vput(nd.ni_dvp);
  976                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
  977                         return (error);
  978                 goto restart;
  979         }
  980         VATTR_NULL(&vattr);
  981         vattr.va_type = VFIFO;
  982         FILEDESC_LOCK(td->td_proc->p_fd);
  983         vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
  984         FILEDESC_UNLOCK(td->td_proc->p_fd);
  985 #ifdef MAC
  986         error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
  987             &vattr);
  988         if (error)
  989                 goto out;
  990 #endif
  991         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
  992         error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
  993         if (error == 0)
  994                 vput(nd.ni_vp);
  995 #ifdef MAC
  996 out:
  997 #endif
  998         NDFREE(&nd, NDF_ONLY_PNBUF);
  999         vput(nd.ni_dvp);
 1000         vn_finished_write(mp);
 1001         return (error);
 1002 }
 1003 
 1004 /*
 1005  * Make a hard file link.
 1006  */
 1007 #ifndef _SYS_SYSPROTO_H_
 1008 struct link_args {
 1009         char    *path;
 1010         char    *link;
 1011 };
 1012 #endif
 1013 /* ARGSUSED */
 1014 int
 1015 link(td, uap)
 1016         struct thread *td;
 1017         register struct link_args /* {
 1018                 char *path;
 1019                 char *link;
 1020         } */ *uap;
 1021 {
 1022 
 1023         return (kern_link(td, uap->path, uap->link, UIO_USERSPACE));
 1024 }
 1025 
 1026 int
 1027 kern_link(struct thread *td, char *path, char *link, enum uio_seg segflg)
 1028 {
 1029         struct vnode *vp;
 1030         struct mount *mp;
 1031         struct nameidata nd;
 1032         int error;
 1033 
 1034         bwillwrite();
 1035         NDINIT(&nd, LOOKUP, FOLLOW|NOOBJ, segflg, path, td);
 1036         if ((error = namei(&nd)) != 0)
 1037                 return (error);
 1038         NDFREE(&nd, NDF_ONLY_PNBUF);
 1039         vp = nd.ni_vp;
 1040         if (vp->v_type == VDIR) {
 1041                 vrele(vp);
 1042                 return (EPERM);         /* POSIX */
 1043         }
 1044         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
 1045                 vrele(vp);
 1046                 return (error);
 1047         }
 1048         NDINIT(&nd, CREATE, LOCKPARENT | NOOBJ | SAVENAME, segflg, link, td);
 1049         if ((error = namei(&nd)) == 0) {
 1050                 if (nd.ni_vp != NULL) {
 1051                         vrele(nd.ni_vp);
 1052                         error = EEXIST;
 1053                 } else if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td))
 1054                     == 0) {
 1055                         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1056                         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 1057 #ifdef MAC
 1058                         error = mac_check_vnode_link(td->td_ucred, nd.ni_dvp,
 1059                             vp, &nd.ni_cnd);
 1060                         if (error == 0)
 1061 #endif
 1062                                 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
 1063                         VOP_UNLOCK(vp, 0, td);
 1064                 }
 1065                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1066                 vput(nd.ni_dvp);
 1067         }
 1068         vrele(vp);
 1069         vn_finished_write(mp);
 1070         ASSERT_VOP_UNLOCKED(nd.ni_dvp, "link");
 1071         ASSERT_VOP_UNLOCKED(nd.ni_vp, "link");
 1072         return (error);
 1073 }
 1074 
 1075 /*
 1076  * Make a symbolic link.
 1077  */
 1078 #ifndef _SYS_SYSPROTO_H_
 1079 struct symlink_args {
 1080         char    *path;
 1081         char    *link;
 1082 };
 1083 #endif
 1084 /* ARGSUSED */
 1085 int
 1086 symlink(td, uap)
 1087         struct thread *td;
 1088         register struct symlink_args /* {
 1089                 char *path;
 1090                 char *link;
 1091         } */ *uap;
 1092 {
 1093 
 1094         return (kern_symlink(td, uap->path, uap->link, UIO_USERSPACE));
 1095 }
 1096 
 1097 int
 1098 kern_symlink(struct thread *td, char *path, char *link, enum uio_seg segflg)
 1099 {
 1100         struct mount *mp;
 1101         struct vattr vattr;
 1102         char *syspath;
 1103         int error;
 1104         struct nameidata nd;
 1105 
 1106         if (segflg == UIO_SYSSPACE) {
 1107                 syspath = path;
 1108         } else {
 1109                 syspath = uma_zalloc(namei_zone, M_WAITOK);
 1110                 if ((error = copyinstr(path, syspath, MAXPATHLEN, NULL)) != 0)
 1111                         goto out;
 1112         }
 1113 restart:
 1114         bwillwrite();
 1115         NDINIT(&nd, CREATE, LOCKPARENT | NOOBJ | SAVENAME, segflg, link, td);
 1116         if ((error = namei(&nd)) != 0)
 1117                 goto out;
 1118         if (nd.ni_vp) {
 1119                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1120                 vrele(nd.ni_vp);
 1121                 vput(nd.ni_dvp);
 1122                 error = EEXIST;
 1123                 goto out;
 1124         }
 1125         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1126                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1127                 vput(nd.ni_dvp);
 1128                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 1129                         return (error);
 1130                 goto restart;
 1131         }
 1132         VATTR_NULL(&vattr);
 1133         FILEDESC_LOCK(td->td_proc->p_fd);
 1134         vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
 1135         FILEDESC_UNLOCK(td->td_proc->p_fd);
 1136 #ifdef MAC
 1137         vattr.va_type = VLNK;
 1138         error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
 1139             &vattr);
 1140         if (error)
 1141                 goto out2;
 1142 #endif
 1143         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1144         error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
 1145         if (error == 0)
 1146                 vput(nd.ni_vp);
 1147 #ifdef MAC
 1148 out2:
 1149 #endif
 1150         NDFREE(&nd, NDF_ONLY_PNBUF);
 1151         vput(nd.ni_dvp);
 1152         vn_finished_write(mp);
 1153         ASSERT_VOP_UNLOCKED(nd.ni_dvp, "symlink");
 1154         ASSERT_VOP_UNLOCKED(nd.ni_vp, "symlink");
 1155 out:
 1156         if (segflg != UIO_SYSSPACE)
 1157                 uma_zfree(namei_zone, syspath);
 1158         return (error);
 1159 }
 1160 
 1161 /*
 1162  * Delete a whiteout from the filesystem.
 1163  */
 1164 /* ARGSUSED */
 1165 int
 1166 undelete(td, uap)
 1167         struct thread *td;
 1168         register struct undelete_args /* {
 1169                 char *path;
 1170         } */ *uap;
 1171 {
 1172         int error;
 1173         struct mount *mp;
 1174         struct nameidata nd;
 1175 
 1176 restart:
 1177         bwillwrite();
 1178         NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE,
 1179             uap->path, td);
 1180         error = namei(&nd);
 1181         if (error)
 1182                 return (error);
 1183 
 1184         if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
 1185                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1186                 if (nd.ni_vp)
 1187                         vrele(nd.ni_vp);
 1188                 vput(nd.ni_dvp);
 1189                 return (EEXIST);
 1190         }
 1191         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1192                 NDFREE(&nd, NDF_ONLY_PNBUF);
 1193                 vput(nd.ni_dvp);
 1194                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 1195                         return (error);
 1196                 goto restart;
 1197         }
 1198         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1199         error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
 1200         NDFREE(&nd, NDF_ONLY_PNBUF);
 1201         vput(nd.ni_dvp);
 1202         vn_finished_write(mp);
 1203         ASSERT_VOP_UNLOCKED(nd.ni_dvp, "undelete");
 1204         ASSERT_VOP_UNLOCKED(nd.ni_vp, "undelete");
 1205         return (error);
 1206 }
 1207 
 1208 /*
 1209  * Delete a name from the filesystem.
 1210  */
 1211 #ifndef _SYS_SYSPROTO_H_
 1212 struct unlink_args {
 1213         char    *path;
 1214 };
 1215 #endif
 1216 /* ARGSUSED */
 1217 int
 1218 unlink(td, uap)
 1219         struct thread *td;
 1220         struct unlink_args /* {
 1221                 char *path;
 1222         } */ *uap;
 1223 {
 1224 
 1225         return (kern_unlink(td, uap->path, UIO_USERSPACE));
 1226 }
 1227 
 1228 int
 1229 kern_unlink(struct thread *td, char *path, enum uio_seg pathseg)
 1230 {
 1231         struct mount *mp;
 1232         struct vnode *vp;
 1233         int error;
 1234         struct nameidata nd;
 1235 
 1236 restart:
 1237         bwillwrite();
 1238         NDINIT(&nd, DELETE, LOCKPARENT|LOCKLEAF, pathseg, path, td);
 1239         if ((error = namei(&nd)) != 0)
 1240                 return (error);
 1241         vp = nd.ni_vp;
 1242         if (vp->v_type == VDIR)
 1243                 error = EPERM;          /* POSIX */
 1244         else {
 1245                 /*
 1246                  * The root of a mounted filesystem cannot be deleted.
 1247                  *
 1248                  * XXX: can this only be a VDIR case?
 1249                  */
 1250                 if (vp->v_vflag & VV_ROOT)
 1251                         error = EBUSY;
 1252         }
 1253         if (error == 0) {
 1254                 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 1255                         NDFREE(&nd, NDF_ONLY_PNBUF);
 1256                         if (vp == nd.ni_dvp)
 1257                                 vrele(vp);
 1258                         else
 1259                                 vput(vp);
 1260                         vput(nd.ni_dvp);
 1261                         if ((error = vn_start_write(NULL, &mp,
 1262                             V_XSLEEP | PCATCH)) != 0)
 1263                                 return (error);
 1264                         goto restart;
 1265                 }
 1266 #ifdef MAC
 1267                 error = mac_check_vnode_delete(td->td_ucred, nd.ni_dvp, vp,
 1268                     &nd.ni_cnd);
 1269                 if (error)
 1270                         goto out;
 1271 #endif
 1272                 VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 1273                 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
 1274 #ifdef MAC
 1275 out:
 1276 #endif
 1277                 vn_finished_write(mp);
 1278         }
 1279         NDFREE(&nd, NDF_ONLY_PNBUF);
 1280         if (vp == nd.ni_dvp)
 1281                 vrele(vp);
 1282         else
 1283                 vput(vp);
 1284         vput(nd.ni_dvp);
 1285         ASSERT_VOP_UNLOCKED(nd.ni_dvp, "unlink");
 1286         ASSERT_VOP_UNLOCKED(nd.ni_vp, "unlink");
 1287         return (error);
 1288 }
 1289 
 1290 /*
 1291  * Reposition read/write file offset.
 1292  */
 1293 #ifndef _SYS_SYSPROTO_H_
 1294 struct lseek_args {
 1295         int     fd;
 1296         int     pad;
 1297         off_t   offset;
 1298         int     whence;
 1299 };
 1300 #endif
 1301 int
 1302 lseek(td, uap)
 1303         struct thread *td;
 1304         register struct lseek_args /* {
 1305                 int fd;
 1306                 int pad;
 1307                 off_t offset;
 1308                 int whence;
 1309         } */ *uap;
 1310 {
 1311         struct ucred *cred = td->td_ucred;
 1312         struct file *fp;
 1313         struct vnode *vp;
 1314         struct vattr vattr;
 1315         off_t offset;
 1316         int error, noneg;
 1317 
 1318         if ((error = fget(td, uap->fd, &fp)) != 0)
 1319                 return (error);
 1320         if (fp->f_type != DTYPE_VNODE) {
 1321                 fdrop(fp, td);
 1322                 return (ESPIPE);
 1323         }
 1324         vp = (struct vnode *)fp->f_data;
 1325         noneg = (vp->v_type != VCHR);
 1326         offset = uap->offset;
 1327         switch (uap->whence) {
 1328         case L_INCR:
 1329                 if (noneg &&
 1330                     (fp->f_offset < 0 ||
 1331                     (offset > 0 && fp->f_offset > OFF_MAX - offset))) {
 1332                         error = EOVERFLOW;
 1333                         break;
 1334                 }
 1335                 offset += fp->f_offset;
 1336                 break;
 1337         case L_XTND:
 1338                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 1339                 error = VOP_GETATTR(vp, &vattr, cred, td);
 1340                 VOP_UNLOCK(vp, 0, td);
 1341                 if (error)
 1342                         break;
 1343                 if (noneg &&
 1344                     (vattr.va_size > OFF_MAX ||
 1345                     (offset > 0 && vattr.va_size > OFF_MAX - offset))) {
 1346                         error = EOVERFLOW;
 1347                         break;
 1348                 }
 1349                 offset += vattr.va_size;
 1350                 break;
 1351         case L_SET:
 1352                 break;
 1353         default:
 1354                 error = EINVAL;
 1355         }
 1356         if (error == 0 && noneg && offset < 0)
 1357                 error = EINVAL;
 1358         if (error != 0) {
 1359                 fdrop(fp, td);
 1360                 return (error);
 1361         }
 1362         fp->f_offset = offset;
 1363         *(off_t *)(td->td_retval) = fp->f_offset;
 1364         fdrop(fp, td);
 1365         return (0);
 1366 }
 1367 
 1368 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
 1369 /*
 1370  * Reposition read/write file offset.
 1371  */
 1372 #ifndef _SYS_SYSPROTO_H_
 1373 struct olseek_args {
 1374         int     fd;
 1375         long    offset;
 1376         int     whence;
 1377 };
 1378 #endif
 1379 int
 1380 olseek(td, uap)
 1381         struct thread *td;
 1382         register struct olseek_args /* {
 1383                 int fd;
 1384                 long offset;
 1385                 int whence;
 1386         } */ *uap;
 1387 {
 1388         struct lseek_args /* {
 1389                 int fd;
 1390                 int pad;
 1391                 off_t offset;
 1392                 int whence;
 1393         } */ nuap;
 1394         int error;
 1395 
 1396         nuap.fd = uap->fd;
 1397         nuap.offset = uap->offset;
 1398         nuap.whence = uap->whence;
 1399         error = lseek(td, &nuap);
 1400         return (error);
 1401 }
 1402 #endif /* COMPAT_43 */
 1403 
 1404 /*
 1405  * Check access permissions using passed credentials.
 1406  */
 1407 static int
 1408 vn_access(vp, user_flags, cred, td)
 1409         struct vnode    *vp;
 1410         int             user_flags;
 1411         struct ucred    *cred;
 1412         struct thread   *td;
 1413 {
 1414         int error, flags;
 1415 
 1416         /* Flags == 0 means only check for existence. */
 1417         error = 0;
 1418         if (user_flags) {
 1419                 flags = 0;
 1420                 if (user_flags & R_OK)
 1421                         flags |= VREAD;
 1422                 if (user_flags & W_OK)
 1423                         flags |= VWRITE;
 1424                 if (user_flags & X_OK)
 1425                         flags |= VEXEC;
 1426 #ifdef MAC
 1427                 error = mac_check_vnode_access(cred, vp, flags);
 1428                 if (error)
 1429                         return (error);
 1430 #endif
 1431                 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
 1432                         error = VOP_ACCESS(vp, flags, cred, td);
 1433         }
 1434         return (error);
 1435 }
 1436 
 1437 /*
 1438  * Check access permissions using "real" credentials.
 1439  */
 1440 #ifndef _SYS_SYSPROTO_H_
 1441 struct access_args {
 1442         char    *path;
 1443         int     flags;
 1444 };
 1445 #endif
 1446 int
 1447 access(td, uap)
 1448         struct thread *td;
 1449         register struct access_args /* {
 1450                 char *path;
 1451                 int flags;
 1452         } */ *uap;
 1453 {
 1454 
 1455         return (kern_access(td, uap->path, UIO_USERSPACE, uap->flags));
 1456 }
 1457 
 1458 int
 1459 kern_access(struct thread *td, char *path, enum uio_seg pathseg, int flags)
 1460 {
 1461         struct ucred *cred, *tmpcred;
 1462         register struct vnode *vp;
 1463         int error;
 1464         struct nameidata nd;
 1465 
 1466         /*
 1467          * Create and modify a temporary credential instead of one that
 1468          * is potentially shared.  This could also mess up socket
 1469          * buffer accounting which can run in an interrupt context.
 1470          *
 1471          * XXX - Depending on how "threads" are finally implemented, it
 1472          * may be better to explicitly pass the credential to namei()
 1473          * rather than to modify the potentially shared process structure.
 1474          */
 1475         cred = td->td_ucred;
 1476         tmpcred = crdup(cred);
 1477         tmpcred->cr_uid = cred->cr_ruid;
 1478         tmpcred->cr_groups[0] = cred->cr_rgid;
 1479         td->td_ucred = tmpcred;
 1480         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, pathseg, path, td);
 1481         if ((error = namei(&nd)) != 0)
 1482                 goto out1;
 1483         vp = nd.ni_vp;
 1484 
 1485         error = vn_access(vp, flags, tmpcred, td);
 1486         NDFREE(&nd, NDF_ONLY_PNBUF);
 1487         vput(vp);
 1488 out1:
 1489         td->td_ucred = cred;
 1490         crfree(tmpcred);
 1491         return (error);
 1492 }
 1493 
 1494 /*
 1495  * Check access permissions using "effective" credentials.
 1496  */
 1497 #ifndef _SYS_SYSPROTO_H_
 1498 struct eaccess_args {
 1499         char    *path;
 1500         int     flags;
 1501 };
 1502 #endif
 1503 int
 1504 eaccess(td, uap)
 1505         struct thread *td;
 1506         register struct eaccess_args /* {
 1507                 char *path;
 1508                 int flags;
 1509         } */ *uap;
 1510 {
 1511         struct nameidata nd;
 1512         struct vnode *vp;
 1513         int error;
 1514 
 1515         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1516             uap->path, td);
 1517         if ((error = namei(&nd)) != 0)
 1518                 return (error);
 1519         vp = nd.ni_vp;
 1520 
 1521         error = vn_access(vp, uap->flags, td->td_ucred, td);
 1522         NDFREE(&nd, NDF_ONLY_PNBUF);
 1523         vput(vp);
 1524         return (error);
 1525 }
 1526 
 1527 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
 1528 /*
 1529  * Get file status; this version follows links.
 1530  */
 1531 #ifndef _SYS_SYSPROTO_H_
 1532 struct ostat_args {
 1533         char    *path;
 1534         struct ostat *ub;
 1535 };
 1536 #endif
 1537 /* ARGSUSED */
 1538 int
 1539 ostat(td, uap)
 1540         struct thread *td;
 1541         register struct ostat_args /* {
 1542                 char *path;
 1543                 struct ostat *ub;
 1544         } */ *uap;
 1545 {
 1546         struct stat sb;
 1547         struct ostat osb;
 1548         int error;
 1549         struct nameidata nd;
 1550 
 1551         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1552             uap->path, td);
 1553         if ((error = namei(&nd)) != 0)
 1554                 return (error);
 1555         NDFREE(&nd, NDF_ONLY_PNBUF);
 1556         error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
 1557         vput(nd.ni_vp);
 1558         if (error)
 1559                 return (error);
 1560         cvtstat(&sb, &osb);
 1561         error = copyout(&osb, uap->ub, sizeof (osb));
 1562         return (error);
 1563 }
 1564 
 1565 /*
 1566  * Get file status; this version does not follow links.
 1567  */
 1568 #ifndef _SYS_SYSPROTO_H_
 1569 struct olstat_args {
 1570         char    *path;
 1571         struct ostat *ub;
 1572 };
 1573 #endif
 1574 /* ARGSUSED */
 1575 int
 1576 olstat(td, uap)
 1577         struct thread *td;
 1578         register struct olstat_args /* {
 1579                 char *path;
 1580                 struct ostat *ub;
 1581         } */ *uap;
 1582 {
 1583         struct vnode *vp;
 1584         struct stat sb;
 1585         struct ostat osb;
 1586         int error;
 1587         struct nameidata nd;
 1588 
 1589         NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1590             uap->path, td);
 1591         if ((error = namei(&nd)) != 0)
 1592                 return (error);
 1593         vp = nd.ni_vp;
 1594         error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
 1595         NDFREE(&nd, NDF_ONLY_PNBUF);
 1596         vput(vp);
 1597         if (error)
 1598                 return (error);
 1599         cvtstat(&sb, &osb);
 1600         error = copyout(&osb, uap->ub, sizeof (osb));
 1601         return (error);
 1602 }
 1603 
 1604 /*
 1605  * Convert from an old to a new stat structure.
 1606  */
 1607 void
 1608 cvtstat(st, ost)
 1609         struct stat *st;
 1610         struct ostat *ost;
 1611 {
 1612 
 1613         ost->st_dev = st->st_dev;
 1614         ost->st_ino = st->st_ino;
 1615         ost->st_mode = st->st_mode;
 1616         ost->st_nlink = st->st_nlink;
 1617         ost->st_uid = st->st_uid;
 1618         ost->st_gid = st->st_gid;
 1619         ost->st_rdev = st->st_rdev;
 1620         if (st->st_size < (quad_t)1 << 32)
 1621                 ost->st_size = st->st_size;
 1622         else
 1623                 ost->st_size = -2;
 1624         ost->st_atime = st->st_atime;
 1625         ost->st_mtime = st->st_mtime;
 1626         ost->st_ctime = st->st_ctime;
 1627         ost->st_blksize = st->st_blksize;
 1628         ost->st_blocks = st->st_blocks;
 1629         ost->st_flags = st->st_flags;
 1630         ost->st_gen = st->st_gen;
 1631 }
 1632 #endif /* COMPAT_43 || COMPAT_SUNOS */
 1633 
 1634 /*
 1635  * Get file status; this version follows links.
 1636  */
 1637 #ifndef _SYS_SYSPROTO_H_
 1638 struct stat_args {
 1639         char    *path;
 1640         struct stat *ub;
 1641 };
 1642 #endif
 1643 /* ARGSUSED */
 1644 int
 1645 stat(td, uap)
 1646         struct thread *td;
 1647         register struct stat_args /* {
 1648                 char *path;
 1649                 struct stat *ub;
 1650         } */ *uap;
 1651 {
 1652         struct stat sb;
 1653         int error;
 1654         struct nameidata nd;
 1655 
 1656 #ifdef LOOKUP_SHARED
 1657         NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | NOOBJ,
 1658             UIO_USERSPACE, uap->path, td);
 1659 #else
 1660         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1661             uap->path, td);
 1662 #endif
 1663         if ((error = namei(&nd)) != 0)
 1664                 return (error);
 1665         error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
 1666         NDFREE(&nd, NDF_ONLY_PNBUF);
 1667         vput(nd.ni_vp);
 1668         if (error)
 1669                 return (error);
 1670         error = copyout(&sb, uap->ub, sizeof (sb));
 1671         return (error);
 1672 }
 1673 
 1674 /*
 1675  * Get file status; this version does not follow links.
 1676  */
 1677 #ifndef _SYS_SYSPROTO_H_
 1678 struct lstat_args {
 1679         char    *path;
 1680         struct stat *ub;
 1681 };
 1682 #endif
 1683 /* ARGSUSED */
 1684 int
 1685 lstat(td, uap)
 1686         struct thread *td;
 1687         register struct lstat_args /* {
 1688                 char *path;
 1689                 struct stat *ub;
 1690         } */ *uap;
 1691 {
 1692         int error;
 1693         struct vnode *vp;
 1694         struct stat sb;
 1695         struct nameidata nd;
 1696 
 1697         NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1698             uap->path, td);
 1699         if ((error = namei(&nd)) != 0)
 1700                 return (error);
 1701         vp = nd.ni_vp;
 1702         error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
 1703         NDFREE(&nd, NDF_ONLY_PNBUF);
 1704         vput(vp);
 1705         if (error)
 1706                 return (error);
 1707         error = copyout(&sb, uap->ub, sizeof (sb));
 1708         return (error);
 1709 }
 1710 
 1711 /*
 1712  * Implementation of the NetBSD stat() function.
 1713  * XXX This should probably be collapsed with the FreeBSD version,
 1714  * as the differences are only due to vn_stat() clearing spares at
 1715  * the end of the structures.  vn_stat could be split to avoid this,
 1716  * and thus collapse the following to close to zero code.
 1717  */
 1718 void
 1719 cvtnstat(sb, nsb)
 1720         struct stat *sb;
 1721         struct nstat *nsb;
 1722 {
 1723         bzero(nsb, sizeof *nsb);
 1724         nsb->st_dev = sb->st_dev;
 1725         nsb->st_ino = sb->st_ino;
 1726         nsb->st_mode = sb->st_mode;
 1727         nsb->st_nlink = sb->st_nlink;
 1728         nsb->st_uid = sb->st_uid;
 1729         nsb->st_gid = sb->st_gid;
 1730         nsb->st_rdev = sb->st_rdev;
 1731         nsb->st_atimespec = sb->st_atimespec;
 1732         nsb->st_mtimespec = sb->st_mtimespec;
 1733         nsb->st_ctimespec = sb->st_ctimespec;
 1734         nsb->st_size = sb->st_size;
 1735         nsb->st_blocks = sb->st_blocks;
 1736         nsb->st_blksize = sb->st_blksize;
 1737         nsb->st_flags = sb->st_flags;
 1738         nsb->st_gen = sb->st_gen;
 1739         nsb->st_birthtimespec = sb->st_birthtimespec;
 1740 }
 1741 
 1742 #ifndef _SYS_SYSPROTO_H_
 1743 struct nstat_args {
 1744         char    *path;
 1745         struct nstat *ub;
 1746 };
 1747 #endif
 1748 /* ARGSUSED */
 1749 int
 1750 nstat(td, uap)
 1751         struct thread *td;
 1752         register struct nstat_args /* {
 1753                 char *path;
 1754                 struct nstat *ub;
 1755         } */ *uap;
 1756 {
 1757         struct stat sb;
 1758         struct nstat nsb;
 1759         int error;
 1760         struct nameidata nd;
 1761 
 1762         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1763             uap->path, td);
 1764         if ((error = namei(&nd)) != 0)
 1765                 return (error);
 1766         NDFREE(&nd, NDF_ONLY_PNBUF);
 1767         error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
 1768         vput(nd.ni_vp);
 1769         if (error)
 1770                 return (error);
 1771         cvtnstat(&sb, &nsb);
 1772         error = copyout(&nsb, uap->ub, sizeof (nsb));
 1773         return (error);
 1774 }
 1775 
 1776 /*
 1777  * NetBSD lstat.  Get file status; this version does not follow links.
 1778  */
 1779 #ifndef _SYS_SYSPROTO_H_
 1780 struct lstat_args {
 1781         char    *path;
 1782         struct stat *ub;
 1783 };
 1784 #endif
 1785 /* ARGSUSED */
 1786 int
 1787 nlstat(td, uap)
 1788         struct thread *td;
 1789         register struct nlstat_args /* {
 1790                 char *path;
 1791                 struct nstat *ub;
 1792         } */ *uap;
 1793 {
 1794         int error;
 1795         struct vnode *vp;
 1796         struct stat sb;
 1797         struct nstat nsb;
 1798         struct nameidata nd;
 1799 
 1800         NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1801             uap->path, td);
 1802         if ((error = namei(&nd)) != 0)
 1803                 return (error);
 1804         vp = nd.ni_vp;
 1805         NDFREE(&nd, NDF_ONLY_PNBUF);
 1806         error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
 1807         vput(vp);
 1808         if (error)
 1809                 return (error);
 1810         cvtnstat(&sb, &nsb);
 1811         error = copyout(&nsb, uap->ub, sizeof (nsb));
 1812         return (error);
 1813 }
 1814 
 1815 /*
 1816  * Get configurable pathname variables.
 1817  */
 1818 #ifndef _SYS_SYSPROTO_H_
 1819 struct pathconf_args {
 1820         char    *path;
 1821         int     name;
 1822 };
 1823 #endif
 1824 /* ARGSUSED */
 1825 int
 1826 pathconf(td, uap)
 1827         struct thread *td;
 1828         register struct pathconf_args /* {
 1829                 char *path;
 1830                 int name;
 1831         } */ *uap;
 1832 {
 1833         int error;
 1834         struct nameidata nd;
 1835 
 1836         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | NOOBJ, UIO_USERSPACE,
 1837             uap->path, td);
 1838         if ((error = namei(&nd)) != 0)
 1839                 return (error);
 1840         NDFREE(&nd, NDF_ONLY_PNBUF);
 1841 
 1842         /* If asynchronous I/O is available, it works for all files. */
 1843         if (uap->name == _PC_ASYNC_IO)
 1844                 td->td_retval[0] = async_io_version;
 1845         else
 1846                 error = VOP_PATHCONF(nd.ni_vp, uap->name, td->td_retval);
 1847         vput(nd.ni_vp);
 1848         return (error);
 1849 }
 1850 
 1851 /*
 1852  * Return target name of a symbolic link.
 1853  */
 1854 #ifndef _SYS_SYSPROTO_H_
 1855 struct readlink_args {
 1856         char    *path;
 1857         char    *buf;
 1858         int     count;
 1859 };
 1860 #endif
 1861 /* ARGSUSED */
 1862 int
 1863 readlink(td, uap)
 1864         struct thread *td;
 1865         register struct readlink_args /* {
 1866                 char *path;
 1867                 char *buf;
 1868                 int count;
 1869         } */ *uap;
 1870 {
 1871 
 1872         return (kern_readlink(td, uap->path, UIO_USERSPACE, uap->buf,
 1873             UIO_USERSPACE, uap->count));
 1874 }
 1875 
 1876 int
 1877 kern_readlink(struct thread *td, char *path, enum uio_seg pathseg, char *buf,
 1878     enum uio_seg bufseg, int count)
 1879 {
 1880         register struct vnode *vp;
 1881         struct iovec aiov;
 1882         struct uio auio;
 1883         int error;
 1884         struct nameidata nd;
 1885 
 1886         NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | NOOBJ, pathseg, path, td);
 1887         if ((error = namei(&nd)) != 0)
 1888                 return (error);
 1889         NDFREE(&nd, NDF_ONLY_PNBUF);
 1890         vp = nd.ni_vp;
 1891 #ifdef MAC
 1892         error = mac_check_vnode_readlink(td->td_ucred, vp);
 1893         if (error) {
 1894                 vput(vp);
 1895                 return (error);
 1896         }
 1897 #endif
 1898         if (vp->v_type != VLNK)
 1899                 error = EINVAL;
 1900         else {
 1901                 aiov.iov_base = buf;
 1902                 aiov.iov_len = count;
 1903                 auio.uio_iov = &aiov;
 1904                 auio.uio_iovcnt = 1;
 1905                 auio.uio_offset = 0;
 1906                 auio.uio_rw = UIO_READ;
 1907                 auio.uio_segflg = bufseg;
 1908                 auio.uio_td = td;
 1909                 auio.uio_resid = count;
 1910                 error = VOP_READLINK(vp, &auio, td->td_ucred);
 1911         }
 1912         vput(vp);
 1913         td->td_retval[0] = count - auio.uio_resid;
 1914         return (error);
 1915 }
 1916 
 1917 /*
 1918  * Common implementation code for chflags() and fchflags().
 1919  */
 1920 static int
 1921 setfflags(td, vp, flags)
 1922         struct thread *td;
 1923         struct vnode *vp;
 1924         int flags;
 1925 {
 1926         int error;
 1927         struct mount *mp;
 1928         struct vattr vattr;
 1929 
 1930         /*
 1931          * Prevent non-root users from setting flags on devices.  When
 1932          * a device is reused, users can retain ownership of the device
 1933          * if they are allowed to set flags and programs assume that
 1934          * chown can't fail when done as root.
 1935          */
 1936         if (vp->v_type == VCHR || vp->v_type == VBLK) {
 1937                 error = suser_cred(td->td_ucred, PRISON_ROOT);
 1938                 if (error)
 1939                         return (error);
 1940         }
 1941 
 1942         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 1943                 return (error);
 1944         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 1945         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 1946 #ifdef MAC
 1947         error = mac_check_vnode_setflags(td->td_ucred, vp, vattr.va_flags);
 1948         if (error == 0) {
 1949 #endif
 1950                 VATTR_NULL(&vattr);
 1951                 vattr.va_flags = flags;
 1952                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 1953 #ifdef MAC
 1954         }
 1955 #endif
 1956         VOP_UNLOCK(vp, 0, td);
 1957         vn_finished_write(mp);
 1958         return (error);
 1959 }
 1960 
 1961 /*
 1962  * Change flags of a file given a path name.
 1963  */
 1964 #ifndef _SYS_SYSPROTO_H_
 1965 struct chflags_args {
 1966         char    *path;
 1967         int     flags;
 1968 };
 1969 #endif
 1970 /* ARGSUSED */
 1971 int
 1972 chflags(td, uap)
 1973         struct thread *td;
 1974         register struct chflags_args /* {
 1975                 char *path;
 1976                 int flags;
 1977         } */ *uap;
 1978 {
 1979         int error;
 1980         struct nameidata nd;
 1981 
 1982         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 1983         if ((error = namei(&nd)) != 0)
 1984                 return (error);
 1985         NDFREE(&nd, NDF_ONLY_PNBUF);
 1986         error = setfflags(td, nd.ni_vp, uap->flags);
 1987         vrele(nd.ni_vp);
 1988         return error;
 1989 }
 1990 
 1991 /*
 1992  * Same as chflags() but doesn't follow symlinks.
 1993  */
 1994 int
 1995 lchflags(td, uap)
 1996         struct thread *td;
 1997         register struct lchflags_args /* {
 1998                 char *path;
 1999                 int flags;
 2000         } */ *uap;
 2001 {
 2002         int error;
 2003         struct nameidata nd;
 2004 
 2005         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 2006         if ((error = namei(&nd)) != 0)
 2007                 return (error);
 2008         NDFREE(&nd, NDF_ONLY_PNBUF);
 2009         error = setfflags(td, nd.ni_vp, uap->flags);
 2010         vrele(nd.ni_vp);
 2011         return error;
 2012 }
 2013 
 2014 /*
 2015  * Change flags of a file given a file descriptor.
 2016  */
 2017 #ifndef _SYS_SYSPROTO_H_
 2018 struct fchflags_args {
 2019         int     fd;
 2020         int     flags;
 2021 };
 2022 #endif
 2023 /* ARGSUSED */
 2024 int
 2025 fchflags(td, uap)
 2026         struct thread *td;
 2027         register struct fchflags_args /* {
 2028                 int fd;
 2029                 int flags;
 2030         } */ *uap;
 2031 {
 2032         struct file *fp;
 2033         int error;
 2034 
 2035         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2036                 return (error);
 2037         error = setfflags(td, (struct vnode *) fp->f_data, uap->flags);
 2038         fdrop(fp, td);
 2039         return (error);
 2040 }
 2041 
 2042 /*
 2043  * Common implementation code for chmod(), lchmod() and fchmod().
 2044  */
 2045 static int
 2046 setfmode(td, vp, mode)
 2047         struct thread *td;
 2048         struct vnode *vp;
 2049         int mode;
 2050 {
 2051         int error;
 2052         struct mount *mp;
 2053         struct vattr vattr;
 2054 
 2055         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 2056                 return (error);
 2057         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2058         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2059         VATTR_NULL(&vattr);
 2060         vattr.va_mode = mode & ALLPERMS;
 2061 #ifdef MAC
 2062         error = mac_check_vnode_setmode(td->td_ucred, vp, vattr.va_mode);
 2063         if (error == 0)
 2064 #endif
 2065                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2066         VOP_UNLOCK(vp, 0, td);
 2067         vn_finished_write(mp);
 2068         return error;
 2069 }
 2070 
 2071 /*
 2072  * Change mode of a file given path name.
 2073  */
 2074 #ifndef _SYS_SYSPROTO_H_
 2075 struct chmod_args {
 2076         char    *path;
 2077         int     mode;
 2078 };
 2079 #endif
 2080 /* ARGSUSED */
 2081 int
 2082 chmod(td, uap)
 2083         struct thread *td;
 2084         register struct chmod_args /* {
 2085                 char *path;
 2086                 int mode;
 2087         } */ *uap;
 2088 {
 2089 
 2090         return (kern_chmod(td, uap->path, UIO_USERSPACE, uap->mode));
 2091 }
 2092 
 2093 int
 2094 kern_chmod(struct thread *td, char *path, enum uio_seg pathseg, int mode)
 2095 {
 2096         int error;
 2097         struct nameidata nd;
 2098 
 2099         NDINIT(&nd, LOOKUP, FOLLOW, pathseg, path, td);
 2100         if ((error = namei(&nd)) != 0)
 2101                 return (error);
 2102         NDFREE(&nd, NDF_ONLY_PNBUF);
 2103         error = setfmode(td, nd.ni_vp, mode);
 2104         vrele(nd.ni_vp);
 2105         return error;
 2106 }
 2107 
 2108 /*
 2109  * Change mode of a file given path name (don't follow links.)
 2110  */
 2111 #ifndef _SYS_SYSPROTO_H_
 2112 struct lchmod_args {
 2113         char    *path;
 2114         int     mode;
 2115 };
 2116 #endif
 2117 /* ARGSUSED */
 2118 int
 2119 lchmod(td, uap)
 2120         struct thread *td;
 2121         register struct lchmod_args /* {
 2122                 char *path;
 2123                 int mode;
 2124         } */ *uap;
 2125 {
 2126         int error;
 2127         struct nameidata nd;
 2128 
 2129         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 2130         if ((error = namei(&nd)) != 0)
 2131                 return (error);
 2132         NDFREE(&nd, NDF_ONLY_PNBUF);
 2133         error = setfmode(td, nd.ni_vp, uap->mode);
 2134         vrele(nd.ni_vp);
 2135         return error;
 2136 }
 2137 
 2138 /*
 2139  * Change mode of a file given a file descriptor.
 2140  */
 2141 #ifndef _SYS_SYSPROTO_H_
 2142 struct fchmod_args {
 2143         int     fd;
 2144         int     mode;
 2145 };
 2146 #endif
 2147 /* ARGSUSED */
 2148 int
 2149 fchmod(td, uap)
 2150         struct thread *td;
 2151         register struct fchmod_args /* {
 2152                 int fd;
 2153                 int mode;
 2154         } */ *uap;
 2155 {
 2156         struct file *fp;
 2157         struct vnode *vp;
 2158         int error;
 2159 
 2160         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2161                 return (error);
 2162         vp = (struct vnode *)fp->f_data;
 2163         error = setfmode(td, (struct vnode *)fp->f_data, uap->mode);
 2164         fdrop(fp, td);
 2165         return (error);
 2166 }
 2167 
 2168 /*
 2169  * Common implementation for chown(), lchown(), and fchown()
 2170  */
 2171 static int
 2172 setfown(td, vp, uid, gid)
 2173         struct thread *td;
 2174         struct vnode *vp;
 2175         uid_t uid;
 2176         gid_t gid;
 2177 {
 2178         int error;
 2179         struct mount *mp;
 2180         struct vattr vattr;
 2181 
 2182         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 2183                 return (error);
 2184         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2185         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2186         VATTR_NULL(&vattr);
 2187         vattr.va_uid = uid;
 2188         vattr.va_gid = gid;
 2189 #ifdef MAC
 2190         error = mac_check_vnode_setowner(td->td_ucred, vp, vattr.va_uid,
 2191             vattr.va_gid);
 2192         if (error == 0)
 2193 #endif
 2194                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2195         VOP_UNLOCK(vp, 0, td);
 2196         vn_finished_write(mp);
 2197         return error;
 2198 }
 2199 
 2200 /*
 2201  * Set ownership given a path name.
 2202  */
 2203 #ifndef _SYS_SYSPROTO_H_
 2204 struct chown_args {
 2205         char    *path;
 2206         int     uid;
 2207         int     gid;
 2208 };
 2209 #endif
 2210 /* ARGSUSED */
 2211 int
 2212 chown(td, uap)
 2213         struct thread *td;
 2214         register struct chown_args /* {
 2215                 char *path;
 2216                 int uid;
 2217                 int gid;
 2218         } */ *uap;
 2219 {
 2220 
 2221         return (kern_chown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
 2222 }
 2223 
 2224 int
 2225 kern_chown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
 2226     int gid)
 2227 {
 2228         int error;
 2229         struct nameidata nd;
 2230 
 2231         NDINIT(&nd, LOOKUP, FOLLOW, pathseg, path, td);
 2232         if ((error = namei(&nd)) != 0)
 2233                 return (error);
 2234         NDFREE(&nd, NDF_ONLY_PNBUF);
 2235         error = setfown(td, nd.ni_vp, uid, gid);
 2236         vrele(nd.ni_vp);
 2237         return (error);
 2238 }
 2239 
 2240 /*
 2241  * Set ownership given a path name, do not cross symlinks.
 2242  */
 2243 #ifndef _SYS_SYSPROTO_H_
 2244 struct lchown_args {
 2245         char    *path;
 2246         int     uid;
 2247         int     gid;
 2248 };
 2249 #endif
 2250 /* ARGSUSED */
 2251 int
 2252 lchown(td, uap)
 2253         struct thread *td;
 2254         register struct lchown_args /* {
 2255                 char *path;
 2256                 int uid;
 2257                 int gid;
 2258         } */ *uap;
 2259 {
 2260 
 2261         return (kern_lchown(td, uap->path, UIO_USERSPACE, uap->uid, uap->gid));
 2262 }
 2263 
 2264 int
 2265 kern_lchown(struct thread *td, char *path, enum uio_seg pathseg, int uid,
 2266     int gid)
 2267 {
 2268         int error;
 2269         struct nameidata nd;
 2270 
 2271         NDINIT(&nd, LOOKUP, NOFOLLOW, pathseg, path, td);
 2272         if ((error = namei(&nd)) != 0)
 2273                 return (error);
 2274         NDFREE(&nd, NDF_ONLY_PNBUF);
 2275         error = setfown(td, nd.ni_vp, uid, gid);
 2276         vrele(nd.ni_vp);
 2277         return (error);
 2278 }
 2279 
 2280 /*
 2281  * Set ownership given a file descriptor.
 2282  */
 2283 #ifndef _SYS_SYSPROTO_H_
 2284 struct fchown_args {
 2285         int     fd;
 2286         int     uid;
 2287         int     gid;
 2288 };
 2289 #endif
 2290 /* ARGSUSED */
 2291 int
 2292 fchown(td, uap)
 2293         struct thread *td;
 2294         register struct fchown_args /* {
 2295                 int fd;
 2296                 int uid;
 2297                 int gid;
 2298         } */ *uap;
 2299 {
 2300         struct file *fp;
 2301         struct vnode *vp;
 2302         int error;
 2303 
 2304         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2305                 return (error);
 2306         vp = (struct vnode *)fp->f_data;
 2307         error = setfown(td, (struct vnode *)fp->f_data,
 2308                 uap->uid, uap->gid);
 2309         fdrop(fp, td);
 2310         return (error);
 2311 }
 2312 
 2313 /*
 2314  * Common implementation code for utimes(), lutimes(), and futimes().
 2315  */
 2316 static int
 2317 getutimes(usrtvp, tvpseg, tsp)
 2318         const struct timeval *usrtvp;
 2319         enum uio_seg tvpseg;
 2320         struct timespec *tsp;
 2321 {
 2322         struct timeval tv[2];
 2323         const struct timeval *tvp;
 2324         int error;
 2325 
 2326         if (usrtvp == NULL) {
 2327                 microtime(&tv[0]);
 2328                 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]);
 2329                 tsp[1] = tsp[0];
 2330         } else {
 2331                 if (tvpseg == UIO_SYSSPACE) {
 2332                         tvp = usrtvp;
 2333                 } else {
 2334                         if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
 2335                                 return (error);
 2336                         tvp = tv;
 2337                 }
 2338 
 2339                 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
 2340                 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
 2341         }
 2342         return 0;
 2343 }
 2344 
 2345 /*
 2346  * Common implementation code for utimes(), lutimes(), and futimes().
 2347  */
 2348 static int
 2349 setutimes(td, vp, ts, numtimes, nullflag)
 2350         struct thread *td;
 2351         struct vnode *vp;
 2352         const struct timespec *ts;
 2353         int numtimes;
 2354         int nullflag;
 2355 {
 2356         int error, setbirthtime;
 2357         struct mount *mp;
 2358         struct vattr vattr;
 2359 
 2360         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 2361                 return (error);
 2362         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2363         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2364         setbirthtime = 0;
 2365         if (numtimes < 3 && VOP_GETATTR(vp, &vattr, td->td_ucred, td) == 0 &&
 2366             timespeccmp(&ts[1], &vattr.va_birthtime, < ))
 2367                 setbirthtime = 1;
 2368         VATTR_NULL(&vattr);
 2369         vattr.va_atime = ts[0];
 2370         vattr.va_mtime = ts[1];
 2371         if (setbirthtime)
 2372                 vattr.va_birthtime = ts[1];
 2373         if (numtimes > 2)
 2374                 vattr.va_birthtime = ts[2];
 2375         if (nullflag)
 2376                 vattr.va_vaflags |= VA_UTIMES_NULL;
 2377 #ifdef MAC
 2378         error = mac_check_vnode_setutimes(td->td_ucred, vp, vattr.va_atime,
 2379             vattr.va_mtime);
 2380 #endif
 2381         if (error == 0)
 2382                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2383         VOP_UNLOCK(vp, 0, td);
 2384         vn_finished_write(mp);
 2385         return error;
 2386 }
 2387 
 2388 /*
 2389  * Set the access and modification times of a file.
 2390  */
 2391 #ifndef _SYS_SYSPROTO_H_
 2392 struct utimes_args {
 2393         char    *path;
 2394         struct  timeval *tptr;
 2395 };
 2396 #endif
 2397 /* ARGSUSED */
 2398 int
 2399 utimes(td, uap)
 2400         struct thread *td;
 2401         register struct utimes_args /* {
 2402                 char *path;
 2403                 struct timeval *tptr;
 2404         } */ *uap;
 2405 {
 2406 
 2407         return (kern_utimes(td, uap->path, UIO_USERSPACE, uap->tptr,
 2408             UIO_USERSPACE));
 2409 }
 2410 
 2411 int
 2412 kern_utimes(struct thread *td, char *path, enum uio_seg pathseg,
 2413     struct timeval *tptr, enum uio_seg tptrseg)
 2414 {
 2415         struct timespec ts[2];
 2416         int error;
 2417         struct nameidata nd;
 2418 
 2419         if ((error = getutimes(tptr, tptrseg, ts)) != 0)
 2420                 return (error);
 2421         NDINIT(&nd, LOOKUP, FOLLOW, pathseg, path, td);
 2422         if ((error = namei(&nd)) != 0)
 2423                 return (error);
 2424         NDFREE(&nd, NDF_ONLY_PNBUF);
 2425         error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
 2426         vrele(nd.ni_vp);
 2427         return (error);
 2428 }
 2429 
 2430 /*
 2431  * Set the access and modification times of a file.
 2432  */
 2433 #ifndef _SYS_SYSPROTO_H_
 2434 struct lutimes_args {
 2435         char    *path;
 2436         struct  timeval *tptr;
 2437 };
 2438 #endif
 2439 /* ARGSUSED */
 2440 int
 2441 lutimes(td, uap)
 2442         struct thread *td;
 2443         register struct lutimes_args /* {
 2444                 char *path;
 2445                 struct timeval *tptr;
 2446         } */ *uap;
 2447 {
 2448 
 2449         return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
 2450             UIO_USERSPACE));
 2451 }
 2452 
 2453 int
 2454 kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
 2455     struct timeval *tptr, enum uio_seg tptrseg)
 2456 {
 2457         struct timespec ts[2];
 2458         int error;
 2459         struct nameidata nd;
 2460 
 2461         if ((error = getutimes(tptr, tptrseg, ts)) != 0)
 2462                 return (error);
 2463         NDINIT(&nd, LOOKUP, NOFOLLOW, pathseg, path, td);
 2464         if ((error = namei(&nd)) != 0)
 2465                 return (error);
 2466         NDFREE(&nd, NDF_ONLY_PNBUF);
 2467         error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
 2468         vrele(nd.ni_vp);
 2469         return (error);
 2470 }
 2471 
 2472 /*
 2473  * Set the access and modification times of a file.
 2474  */
 2475 #ifndef _SYS_SYSPROTO_H_
 2476 struct futimes_args {
 2477         int     fd;
 2478         struct  timeval *tptr;
 2479 };
 2480 #endif
 2481 /* ARGSUSED */
 2482 int
 2483 futimes(td, uap)
 2484         struct thread *td;
 2485         register struct futimes_args /* {
 2486                 int  fd;
 2487                 struct timeval *tptr;
 2488         } */ *uap;
 2489 {
 2490 
 2491         return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
 2492 }
 2493 
 2494 int
 2495 kern_futimes(struct thread *td, int fd, struct timeval *tptr,
 2496     enum uio_seg tptrseg)
 2497 {
 2498         struct timespec ts[2];
 2499         struct file *fp;
 2500         int error;
 2501 
 2502         if ((error = getutimes(tptr, tptrseg, ts)) != 0)
 2503                 return (error);
 2504         if ((error = getvnode(td->td_proc->p_fd, fd, &fp)) != 0)
 2505                 return (error);
 2506         error = setutimes(td, (struct vnode *)fp->f_data, ts, 2, tptr == NULL);
 2507         fdrop(fp, td);
 2508         return (error);
 2509 }
 2510 
 2511 /*
 2512  * Truncate a file given its path name.
 2513  */
 2514 #ifndef _SYS_SYSPROTO_H_
 2515 struct truncate_args {
 2516         char    *path;
 2517         int     pad;
 2518         off_t   length;
 2519 };
 2520 #endif
 2521 /* ARGSUSED */
 2522 int
 2523 truncate(td, uap)
 2524         struct thread *td;
 2525         register struct truncate_args /* {
 2526                 char *path;
 2527                 int pad;
 2528                 off_t length;
 2529         } */ *uap;
 2530 {
 2531 
 2532         return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
 2533 }
 2534 
 2535 int
 2536 kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
 2537 {
 2538         struct mount *mp;
 2539         struct vnode *vp;
 2540         struct vattr vattr;
 2541         int error;
 2542         struct nameidata nd;
 2543 
 2544         if (length < 0)
 2545                 return(EINVAL);
 2546         NDINIT(&nd, LOOKUP, FOLLOW, pathseg, path, td);
 2547         if ((error = namei(&nd)) != 0)
 2548                 return (error);
 2549         vp = nd.ni_vp;
 2550         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
 2551                 vrele(vp);
 2552                 return (error);
 2553         }
 2554         NDFREE(&nd, NDF_ONLY_PNBUF);
 2555         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2556         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2557         if (vp->v_type == VDIR)
 2558                 error = EISDIR;
 2559 #ifdef MAC
 2560         else if ((error = mac_check_vnode_write(td->td_ucred, NOCRED, vp))) {
 2561         }
 2562 #endif
 2563         else if ((error = vn_writechk(vp)) == 0 &&
 2564             (error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
 2565                 VATTR_NULL(&vattr);
 2566                 vattr.va_size = length;
 2567                 error = VOP_SETATTR(vp, &vattr, td->td_ucred, td);
 2568         }
 2569         vput(vp);
 2570         vn_finished_write(mp);
 2571         return (error);
 2572 }
 2573 
 2574 /*
 2575  * Truncate a file given a file descriptor.
 2576  */
 2577 #ifndef _SYS_SYSPROTO_H_
 2578 struct ftruncate_args {
 2579         int     fd;
 2580         int     pad;
 2581         off_t   length;
 2582 };
 2583 #endif
 2584 /* ARGSUSED */
 2585 int
 2586 ftruncate(td, uap)
 2587         struct thread *td;
 2588         register struct ftruncate_args /* {
 2589                 int fd;
 2590                 int pad;
 2591                 off_t length;
 2592         } */ *uap;
 2593 {
 2594         struct mount *mp;
 2595         struct vattr vattr;
 2596         struct vnode *vp;
 2597         struct file *fp;
 2598         int error;
 2599 
 2600         if (uap->length < 0)
 2601                 return(EINVAL);
 2602         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2603                 return (error);
 2604         if ((fp->f_flag & FWRITE) == 0) {
 2605                 fdrop(fp, td);
 2606                 return (EINVAL);
 2607         }
 2608         vp = (struct vnode *)fp->f_data;
 2609         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
 2610                 fdrop(fp, td);
 2611                 return (error);
 2612         }
 2613         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 2614         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2615         if (vp->v_type == VDIR)
 2616                 error = EISDIR;
 2617 #ifdef MAC
 2618         else if ((error = mac_check_vnode_write(td->td_ucred, fp->f_cred,
 2619             vp))) {
 2620         }
 2621 #endif
 2622         else if ((error = vn_writechk(vp)) == 0) {
 2623                 VATTR_NULL(&vattr);
 2624                 vattr.va_size = uap->length;
 2625                 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td);
 2626         }
 2627         VOP_UNLOCK(vp, 0, td);
 2628         vn_finished_write(mp);
 2629         fdrop(fp, td);
 2630         return (error);
 2631 }
 2632 
 2633 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
 2634 /*
 2635  * Truncate a file given its path name.
 2636  */
 2637 #ifndef _SYS_SYSPROTO_H_
 2638 struct otruncate_args {
 2639         char    *path;
 2640         long    length;
 2641 };
 2642 #endif
 2643 /* ARGSUSED */
 2644 int
 2645 otruncate(td, uap)
 2646         struct thread *td;
 2647         register struct otruncate_args /* {
 2648                 char *path;
 2649                 long length;
 2650         } */ *uap;
 2651 {
 2652         struct truncate_args /* {
 2653                 char *path;
 2654                 int pad;
 2655                 off_t length;
 2656         } */ nuap;
 2657 
 2658         nuap.path = uap->path;
 2659         nuap.length = uap->length;
 2660         return (truncate(td, &nuap));
 2661 }
 2662 
 2663 /*
 2664  * Truncate a file given a file descriptor.
 2665  */
 2666 #ifndef _SYS_SYSPROTO_H_
 2667 struct oftruncate_args {
 2668         int     fd;
 2669         long    length;
 2670 };
 2671 #endif
 2672 /* ARGSUSED */
 2673 int
 2674 oftruncate(td, uap)
 2675         struct thread *td;
 2676         register struct oftruncate_args /* {
 2677                 int fd;
 2678                 long length;
 2679         } */ *uap;
 2680 {
 2681         struct ftruncate_args /* {
 2682                 int fd;
 2683                 int pad;
 2684                 off_t length;
 2685         } */ nuap;
 2686 
 2687         nuap.fd = uap->fd;
 2688         nuap.length = uap->length;
 2689         return (ftruncate(td, &nuap));
 2690 }
 2691 #endif /* COMPAT_43 || COMPAT_SUNOS */
 2692 
 2693 /*
 2694  * Sync an open file.
 2695  */
 2696 #ifndef _SYS_SYSPROTO_H_
 2697 struct fsync_args {
 2698         int     fd;
 2699 };
 2700 #endif
 2701 /* ARGSUSED */
 2702 int
 2703 fsync(td, uap)
 2704         struct thread *td;
 2705         struct fsync_args /* {
 2706                 int fd;
 2707         } */ *uap;
 2708 {
 2709         struct vnode *vp;
 2710         struct mount *mp;
 2711         struct file *fp;
 2712         vm_object_t obj;
 2713         int error;
 2714 
 2715         GIANT_REQUIRED;
 2716 
 2717         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 2718                 return (error);
 2719         vp = (struct vnode *)fp->f_data;
 2720         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
 2721                 fdrop(fp, td);
 2722                 return (error);
 2723         }
 2724         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 2725         if (VOP_GETVOBJECT(vp, &obj) == 0) {
 2726                 vm_object_page_clean(obj, 0, 0, 0);
 2727         }
 2728         error = VOP_FSYNC(vp, fp->f_cred, MNT_WAIT, td);
 2729         if (error == 0 && vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)
 2730             && softdep_fsync_hook != NULL)
 2731                 error = (*softdep_fsync_hook)(vp);
 2732 
 2733         VOP_UNLOCK(vp, 0, td);
 2734         vn_finished_write(mp);
 2735         fdrop(fp, td);
 2736         return (error);
 2737 }
 2738 
 2739 /*
 2740  * Rename files.  Source and destination must either both be directories,
 2741  * or both not be directories.  If target is a directory, it must be empty.
 2742  */
 2743 #ifndef _SYS_SYSPROTO_H_
 2744 struct rename_args {
 2745         char    *from;
 2746         char    *to;
 2747 };
 2748 #endif
 2749 /* ARGSUSED */
 2750 int
 2751 rename(td, uap)
 2752         struct thread *td;
 2753         register struct rename_args /* {
 2754                 char *from;
 2755                 char *to;
 2756         } */ *uap;
 2757 {
 2758 
 2759         return (kern_rename(td, uap->from, uap->to, UIO_USERSPACE));
 2760 }
 2761 
 2762 int
 2763 kern_rename(struct thread *td, char *from, char *to, enum uio_seg pathseg)
 2764 {
 2765         struct mount *mp = NULL;
 2766         struct vnode *tvp, *fvp, *tdvp;
 2767         struct nameidata fromnd, tond;
 2768         int error;
 2769 
 2770         bwillwrite();
 2771 #ifdef MAC
 2772         NDINIT(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART, pathseg,
 2773             from, td);
 2774 #else
 2775         NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, pathseg, from, td);
 2776 #endif
 2777         if ((error = namei(&fromnd)) != 0)
 2778                 return (error);
 2779 #ifdef MAC
 2780         error = mac_check_vnode_rename_from(td->td_ucred, fromnd.ni_dvp,
 2781             fromnd.ni_vp, &fromnd.ni_cnd);
 2782         VOP_UNLOCK(fromnd.ni_dvp, 0, td);
 2783         VOP_UNLOCK(fromnd.ni_vp, 0, td);
 2784 #endif
 2785         fvp = fromnd.ni_vp;
 2786         if (error == 0)
 2787                 error = vn_start_write(fvp, &mp, V_WAIT | PCATCH);
 2788         if (error != 0) {
 2789                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 2790                 vrele(fromnd.ni_dvp);
 2791                 vrele(fvp);
 2792                 goto out1;
 2793         }
 2794         NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART |
 2795             NOOBJ, pathseg, to, td);
 2796         if (fromnd.ni_vp->v_type == VDIR)
 2797                 tond.ni_cnd.cn_flags |= WILLBEDIR;
 2798         if ((error = namei(&tond)) != 0) {
 2799                 /* Translate error code for rename("dir1", "dir2/."). */
 2800                 if (error == EISDIR && fvp->v_type == VDIR)
 2801                         error = EINVAL;
 2802                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 2803                 vrele(fromnd.ni_dvp);
 2804                 vrele(fvp);
 2805                 goto out1;
 2806         }
 2807         tdvp = tond.ni_dvp;
 2808         tvp = tond.ni_vp;
 2809         if (tvp != NULL) {
 2810                 if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
 2811                         error = ENOTDIR;
 2812                         goto out;
 2813                 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
 2814                         error = EISDIR;
 2815                         goto out;
 2816                 }
 2817         }
 2818         if (fvp == tdvp)
 2819                 error = EINVAL;
 2820         /*
 2821          * If the source is the same as the destination (that is, if they
 2822          * are links to the same vnode), then there is nothing to do.
 2823          */
 2824         if (fvp == tvp)
 2825                 error = -1;
 2826 #ifdef MAC
 2827         else
 2828                 error = mac_check_vnode_rename_to(td->td_ucred, tdvp,
 2829                     tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
 2830 #endif
 2831 out:
 2832         if (!error) {
 2833                 VOP_LEASE(tdvp, td, td->td_ucred, LEASE_WRITE);
 2834                 if (fromnd.ni_dvp != tdvp) {
 2835                         VOP_LEASE(fromnd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 2836                 }
 2837                 if (tvp) {
 2838                         VOP_LEASE(tvp, td, td->td_ucred, LEASE_WRITE);
 2839                 }
 2840                 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
 2841                                    tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
 2842                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 2843                 NDFREE(&tond, NDF_ONLY_PNBUF);
 2844         } else {
 2845                 NDFREE(&fromnd, NDF_ONLY_PNBUF);
 2846                 NDFREE(&tond, NDF_ONLY_PNBUF);
 2847                 if (tdvp == tvp)
 2848                         vrele(tdvp);
 2849                 else
 2850                         vput(tdvp);
 2851                 if (tvp)
 2852                         vput(tvp);
 2853                 vrele(fromnd.ni_dvp);
 2854                 vrele(fvp);
 2855         }
 2856         vrele(tond.ni_startdir);
 2857         ASSERT_VOP_UNLOCKED(fromnd.ni_dvp, "rename");
 2858         ASSERT_VOP_UNLOCKED(fromnd.ni_vp, "rename");
 2859         ASSERT_VOP_UNLOCKED(tond.ni_dvp, "rename");
 2860         ASSERT_VOP_UNLOCKED(tond.ni_vp, "rename");
 2861 out1:
 2862         vn_finished_write(mp);
 2863         if (fromnd.ni_startdir)
 2864                 vrele(fromnd.ni_startdir);
 2865         if (error == -1)
 2866                 return (0);
 2867         return (error);
 2868 }
 2869 
 2870 /*
 2871  * Make a directory file.
 2872  */
 2873 #ifndef _SYS_SYSPROTO_H_
 2874 struct mkdir_args {
 2875         char    *path;
 2876         int     mode;
 2877 };
 2878 #endif
 2879 /* ARGSUSED */
 2880 int
 2881 mkdir(td, uap)
 2882         struct thread *td;
 2883         register struct mkdir_args /* {
 2884                 char *path;
 2885                 int mode;
 2886         } */ *uap;
 2887 {
 2888 
 2889         return (kern_mkdir(td, uap->path, UIO_USERSPACE, uap->mode));
 2890 }
 2891 
 2892 int
 2893 kern_mkdir(struct thread *td, char *path, enum uio_seg segflg, int mode)
 2894 {
 2895         struct mount *mp;
 2896         struct vnode *vp;
 2897         struct vattr vattr;
 2898         int error;
 2899         struct nameidata nd;
 2900 
 2901 restart:
 2902         bwillwrite();
 2903         NDINIT(&nd, CREATE, LOCKPARENT | SAVENAME, segflg, path, td);
 2904         nd.ni_cnd.cn_flags |= WILLBEDIR;
 2905         if ((error = namei(&nd)) != 0)
 2906                 return (error);
 2907         vp = nd.ni_vp;
 2908         if (vp != NULL) {
 2909                 NDFREE(&nd, NDF_ONLY_PNBUF);
 2910                 vrele(vp);
 2911                 /*
 2912                  * XXX namei called with LOCKPARENT but not LOCKLEAF has
 2913                  * the strange behaviour of leaving the vnode unlocked
 2914                  * if the target is the same vnode as the parent.
 2915                  */
 2916                 if (vp == nd.ni_dvp)
 2917                         vrele(nd.ni_dvp);
 2918                 else
 2919                         vput(nd.ni_dvp);
 2920                 return (EEXIST);
 2921         }
 2922         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 2923                 NDFREE(&nd, NDF_ONLY_PNBUF);
 2924                 vput(nd.ni_dvp);
 2925                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 2926                         return (error);
 2927                 goto restart;
 2928         }
 2929         VATTR_NULL(&vattr);
 2930         vattr.va_type = VDIR;
 2931         FILEDESC_LOCK(td->td_proc->p_fd);
 2932         vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
 2933         FILEDESC_UNLOCK(td->td_proc->p_fd);
 2934 #ifdef MAC
 2935         error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
 2936             &vattr);
 2937         if (error)
 2938                 goto out;
 2939 #endif
 2940         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 2941         error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
 2942 #ifdef MAC
 2943 out:
 2944 #endif
 2945         NDFREE(&nd, NDF_ONLY_PNBUF);
 2946         vput(nd.ni_dvp);
 2947         if (!error)
 2948                 vput(nd.ni_vp);
 2949         vn_finished_write(mp);
 2950         ASSERT_VOP_UNLOCKED(nd.ni_dvp, "mkdir");
 2951         ASSERT_VOP_UNLOCKED(nd.ni_vp, "mkdir");
 2952         return (error);
 2953 }
 2954 
 2955 /*
 2956  * Remove a directory file.
 2957  */
 2958 #ifndef _SYS_SYSPROTO_H_
 2959 struct rmdir_args {
 2960         char    *path;
 2961 };
 2962 #endif
 2963 /* ARGSUSED */
 2964 int
 2965 rmdir(td, uap)
 2966         struct thread *td;
 2967         struct rmdir_args /* {
 2968                 char *path;
 2969         } */ *uap;
 2970 {
 2971 
 2972         return (kern_rmdir(td, uap->path, UIO_USERSPACE));
 2973 }
 2974 
 2975 int
 2976 kern_rmdir(struct thread *td, char *path, enum uio_seg pathseg)
 2977 {
 2978         struct mount *mp;
 2979         struct vnode *vp;
 2980         int error;
 2981         struct nameidata nd;
 2982 
 2983 restart:
 2984         bwillwrite();
 2985         NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, pathseg, path, td);
 2986         if ((error = namei(&nd)) != 0)
 2987                 return (error);
 2988         vp = nd.ni_vp;
 2989         if (vp->v_type != VDIR) {
 2990                 error = ENOTDIR;
 2991                 goto out;
 2992         }
 2993         /*
 2994          * No rmdir "." please.
 2995          */
 2996         if (nd.ni_dvp == vp) {
 2997                 error = EINVAL;
 2998                 goto out;
 2999         }
 3000         /*
 3001          * The root of a mounted filesystem cannot be deleted.
 3002          */
 3003         if (vp->v_vflag & VV_ROOT) {
 3004                 error = EBUSY;
 3005                 goto out;
 3006         }
 3007 #ifdef MAC
 3008         error = mac_check_vnode_delete(td->td_ucred, nd.ni_dvp, vp,
 3009             &nd.ni_cnd);
 3010         if (error)
 3011                 goto out;
 3012 #endif
 3013         if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
 3014                 NDFREE(&nd, NDF_ONLY_PNBUF);
 3015                 if (nd.ni_dvp == vp)
 3016                         vrele(nd.ni_dvp);
 3017                 else
 3018                         vput(nd.ni_dvp);
 3019                 vput(vp);
 3020                 if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
 3021                         return (error);
 3022                 goto restart;
 3023         }
 3024         VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE);
 3025         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 3026         error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
 3027         vn_finished_write(mp);
 3028 out:
 3029         NDFREE(&nd, NDF_ONLY_PNBUF);
 3030         if (nd.ni_dvp == vp)
 3031                 vrele(nd.ni_dvp);
 3032         else
 3033                 vput(nd.ni_dvp);
 3034         vput(vp);
 3035         ASSERT_VOP_UNLOCKED(nd.ni_dvp, "rmdir");
 3036         ASSERT_VOP_UNLOCKED(nd.ni_vp, "rmdir");
 3037         return (error);
 3038 }
 3039 
 3040 #ifdef COMPAT_43
 3041 /*
 3042  * Read a block of directory entries in a filesystem independent format.
 3043  */
 3044 #ifndef _SYS_SYSPROTO_H_
 3045 struct ogetdirentries_args {
 3046         int     fd;
 3047         char    *buf;
 3048         u_int   count;
 3049         long    *basep;
 3050 };
 3051 #endif
 3052 int
 3053 ogetdirentries(td, uap)
 3054         struct thread *td;
 3055         register struct ogetdirentries_args /* {
 3056                 int fd;
 3057                 char *buf;
 3058                 u_int count;
 3059                 long *basep;
 3060         } */ *uap;
 3061 {
 3062         struct vnode *vp;
 3063         struct file *fp;
 3064         struct uio auio, kuio;
 3065         struct iovec aiov, kiov;
 3066         struct dirent *dp, *edp;
 3067         caddr_t dirbuf;
 3068         int error, eofflag, readcnt;
 3069         long loff;
 3070 
 3071         /* XXX arbitrary sanity limit on `count'. */
 3072         if (uap->count > 64 * 1024)
 3073                 return (EINVAL);
 3074         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 3075                 return (error);
 3076         if ((fp->f_flag & FREAD) == 0) {
 3077                 fdrop(fp, td);
 3078                 return (EBADF);
 3079         }
 3080         vp = (struct vnode *)fp->f_data;
 3081 unionread:
 3082         if (vp->v_type != VDIR) {
 3083                 fdrop(fp, td);
 3084                 return (EINVAL);
 3085         }
 3086         aiov.iov_base = uap->buf;
 3087         aiov.iov_len = uap->count;
 3088         auio.uio_iov = &aiov;
 3089         auio.uio_iovcnt = 1;
 3090         auio.uio_rw = UIO_READ;
 3091         auio.uio_segflg = UIO_USERSPACE;
 3092         auio.uio_td = td;
 3093         auio.uio_resid = uap->count;
 3094         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3095         loff = auio.uio_offset = fp->f_offset;
 3096 #ifdef MAC
 3097         error = mac_check_vnode_readdir(td->td_ucred, vp);
 3098         if (error) {
 3099                 VOP_UNLOCK(vp, 0, td);
 3100                 fdrop(fp, td);
 3101                 return (error);
 3102         }
 3103 #endif
 3104 #       if (BYTE_ORDER != LITTLE_ENDIAN)
 3105                 if (vp->v_mount->mnt_maxsymlinklen <= 0) {
 3106                         error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
 3107                             NULL, NULL);
 3108                         fp->f_offset = auio.uio_offset;
 3109                 } else
 3110 #       endif
 3111         {
 3112                 kuio = auio;
 3113                 kuio.uio_iov = &kiov;
 3114                 kuio.uio_segflg = UIO_SYSSPACE;
 3115                 kiov.iov_len = uap->count;
 3116                 MALLOC(dirbuf, caddr_t, uap->count, M_TEMP, M_WAITOK);
 3117                 kiov.iov_base = dirbuf;
 3118                 error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
 3119                             NULL, NULL);
 3120                 fp->f_offset = kuio.uio_offset;
 3121                 if (error == 0) {
 3122                         readcnt = uap->count - kuio.uio_resid;
 3123                         edp = (struct dirent *)&dirbuf[readcnt];
 3124                         for (dp = (struct dirent *)dirbuf; dp < edp; ) {
 3125 #                               if (BYTE_ORDER == LITTLE_ENDIAN)
 3126                                         /*
 3127                                          * The expected low byte of
 3128                                          * dp->d_namlen is our dp->d_type.
 3129                                          * The high MBZ byte of dp->d_namlen
 3130                                          * is our dp->d_namlen.
 3131                                          */
 3132                                         dp->d_type = dp->d_namlen;
 3133                                         dp->d_namlen = 0;
 3134 #                               else
 3135                                         /*
 3136                                          * The dp->d_type is the high byte
 3137                                          * of the expected dp->d_namlen,
 3138                                          * so must be zero'ed.
 3139                                          */
 3140                                         dp->d_type = 0;
 3141 #                               endif
 3142                                 if (dp->d_reclen > 0) {
 3143                                         dp = (struct dirent *)
 3144                                             ((char *)dp + dp->d_reclen);
 3145                                 } else {
 3146                                         error = EIO;
 3147                                         break;
 3148                                 }
 3149                         }
 3150                         if (dp >= edp)
 3151                                 error = uiomove(dirbuf, readcnt, &auio);
 3152                 }
 3153                 FREE(dirbuf, M_TEMP);
 3154         }
 3155         VOP_UNLOCK(vp, 0, td);
 3156         if (error) {
 3157                 fdrop(fp, td);
 3158                 return (error);
 3159         }
 3160         if (uap->count == auio.uio_resid) {
 3161                 if (union_dircheckp) {
 3162                         error = union_dircheckp(td, &vp, fp);
 3163                         if (error == -1)
 3164                                 goto unionread;
 3165                         if (error) {
 3166                                 fdrop(fp, td);
 3167                                 return (error);
 3168                         }
 3169                 }
 3170                 /*
 3171                  * XXX We could delay dropping the lock above but
 3172                  * union_dircheckp complicates things.
 3173                  */
 3174                 vn_lock(vp, LK_EXCLUSIVE|LK_RETRY, td);
 3175                 if ((vp->v_vflag & VV_ROOT) &&
 3176                     (vp->v_mount->mnt_flag & MNT_UNION)) {
 3177                         struct vnode *tvp = vp;
 3178                         vp = vp->v_mount->mnt_vnodecovered;
 3179                         VREF(vp);
 3180                         fp->f_data = vp;
 3181                         fp->f_offset = 0;
 3182                         vput(tvp);
 3183                         goto unionread;
 3184                 }
 3185                 VOP_UNLOCK(vp, 0, td);
 3186         }
 3187         error = copyout(&loff, uap->basep, sizeof(long));
 3188         fdrop(fp, td);
 3189         td->td_retval[0] = uap->count - auio.uio_resid;
 3190         return (error);
 3191 }
 3192 #endif /* COMPAT_43 */
 3193 
 3194 /*
 3195  * Read a block of directory entries in a filesystem independent format.
 3196  */
 3197 #ifndef _SYS_SYSPROTO_H_
 3198 struct getdirentries_args {
 3199         int     fd;
 3200         char    *buf;
 3201         u_int   count;
 3202         long    *basep;
 3203 };
 3204 #endif
 3205 int
 3206 getdirentries(td, uap)
 3207         struct thread *td;
 3208         register struct getdirentries_args /* {
 3209                 int fd;
 3210                 char *buf;
 3211                 u_int count;
 3212                 long *basep;
 3213         } */ *uap;
 3214 {
 3215         struct vnode *vp;
 3216         struct file *fp;
 3217         struct uio auio;
 3218         struct iovec aiov;
 3219         long loff;
 3220         int error, eofflag;
 3221 
 3222         if ((error = getvnode(td->td_proc->p_fd, uap->fd, &fp)) != 0)
 3223                 return (error);
 3224         if ((fp->f_flag & FREAD) == 0) {
 3225                 fdrop(fp, td);
 3226                 return (EBADF);
 3227         }
 3228         vp = (struct vnode *)fp->f_data;
 3229 unionread:
 3230         if (vp->v_type != VDIR) {
 3231                 fdrop(fp, td);
 3232                 return (EINVAL);
 3233         }
 3234         aiov.iov_base = uap->buf;
 3235         aiov.iov_len = uap->count;
 3236         auio.uio_iov = &aiov;
 3237         auio.uio_iovcnt = 1;
 3238         auio.uio_rw = UIO_READ;
 3239         auio.uio_segflg = UIO_USERSPACE;
 3240         auio.uio_td = td;
 3241         auio.uio_resid = uap->count;
 3242         /* vn_lock(vp, LK_SHARED | LK_RETRY, td); */
 3243         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3244         loff = auio.uio_offset = fp->f_offset;
 3245 #ifdef MAC
 3246         error = mac_check_vnode_readdir(td->td_ucred, vp);
 3247         if (error == 0)
 3248 #endif
 3249                 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
 3250                     NULL);
 3251         fp->f_offset = auio.uio_offset;
 3252         VOP_UNLOCK(vp, 0, td);
 3253         if (error) {
 3254                 fdrop(fp, td);
 3255                 return (error);
 3256         }
 3257         if (uap->count == auio.uio_resid) {
 3258                 if (union_dircheckp) {
 3259                         error = union_dircheckp(td, &vp, fp);
 3260                         if (error == -1)
 3261                                 goto unionread;
 3262                         if (error) {
 3263                                 fdrop(fp, td);
 3264                                 return (error);
 3265                         }
 3266                 }
 3267                 /*
 3268                  * XXX We could delay dropping the lock above but
 3269                  * union_dircheckp complicates things.
 3270                  */
 3271                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3272                 if ((vp->v_vflag & VV_ROOT) &&
 3273                     (vp->v_mount->mnt_flag & MNT_UNION)) {
 3274                         struct vnode *tvp = vp;
 3275                         vp = vp->v_mount->mnt_vnodecovered;
 3276                         VREF(vp);
 3277                         fp->f_data = vp;
 3278                         fp->f_offset = 0;
 3279                         vput(tvp);
 3280                         goto unionread;
 3281                 }
 3282                 VOP_UNLOCK(vp, 0, td);
 3283         }
 3284         if (uap->basep != NULL) {
 3285                 error = copyout(&loff, uap->basep, sizeof(long));
 3286         }
 3287         td->td_retval[0] = uap->count - auio.uio_resid;
 3288         fdrop(fp, td);
 3289         return (error);
 3290 }
 3291 #ifndef _SYS_SYSPROTO_H_
 3292 struct getdents_args {
 3293         int fd;
 3294         char *buf;
 3295         size_t count;
 3296 };
 3297 #endif
 3298 int
 3299 getdents(td, uap)
 3300         struct thread *td;
 3301         register struct getdents_args /* {
 3302                 int fd;
 3303                 char *buf;
 3304                 u_int count;
 3305         } */ *uap;
 3306 {
 3307         struct getdirentries_args ap;
 3308         ap.fd = uap->fd;
 3309         ap.buf = uap->buf;
 3310         ap.count = uap->count;
 3311         ap.basep = NULL;
 3312         return getdirentries(td, &ap);
 3313 }
 3314 
 3315 /*
 3316  * Set the mode mask for creation of filesystem nodes.
 3317  *
 3318  * MP SAFE
 3319  */
 3320 #ifndef _SYS_SYSPROTO_H_
 3321 struct umask_args {
 3322         int     newmask;
 3323 };
 3324 #endif
 3325 int
 3326 umask(td, uap)
 3327         struct thread *td;
 3328         struct umask_args /* {
 3329                 int newmask;
 3330         } */ *uap;
 3331 {
 3332         register struct filedesc *fdp;
 3333 
 3334         FILEDESC_LOCK(td->td_proc->p_fd);
 3335         fdp = td->td_proc->p_fd;
 3336         td->td_retval[0] = fdp->fd_cmask;
 3337         fdp->fd_cmask = uap->newmask & ALLPERMS;
 3338         FILEDESC_UNLOCK(td->td_proc->p_fd);
 3339         return (0);
 3340 }
 3341 
 3342 /*
 3343  * Void all references to file by ripping underlying filesystem
 3344  * away from vnode.
 3345  */
 3346 #ifndef _SYS_SYSPROTO_H_
 3347 struct revoke_args {
 3348         char    *path;
 3349 };
 3350 #endif
 3351 /* ARGSUSED */
 3352 int
 3353 revoke(td, uap)
 3354         struct thread *td;
 3355         register struct revoke_args /* {
 3356                 char *path;
 3357         } */ *uap;
 3358 {
 3359         struct mount *mp;
 3360         struct vnode *vp;
 3361         struct vattr vattr;
 3362         int error;
 3363         struct nameidata nd;
 3364 
 3365         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->path,
 3366             td);
 3367         if ((error = namei(&nd)) != 0)
 3368                 return (error);
 3369         vp = nd.ni_vp;
 3370         NDFREE(&nd, NDF_ONLY_PNBUF);
 3371         if (vp->v_type != VCHR) {
 3372                 vput(vp);
 3373                 return (EINVAL);
 3374         }
 3375 #ifdef MAC
 3376         error = mac_check_vnode_revoke(td->td_ucred, vp);
 3377         if (error) {
 3378                 vput(vp);
 3379                 return (error);
 3380         }
 3381 #endif
 3382         error = VOP_GETATTR(vp, &vattr, td->td_ucred, td);
 3383         if (error) {
 3384                 vput(vp);
 3385                 return (error);
 3386         }
 3387         VOP_UNLOCK(vp, 0, td);
 3388         if (td->td_ucred->cr_uid != vattr.va_uid) {
 3389                 error = suser_cred(td->td_ucred, PRISON_ROOT);
 3390                 if (error)
 3391                         goto out;
 3392         }
 3393         if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
 3394                 goto out;
 3395         if (vcount(vp) > 1)
 3396                 VOP_REVOKE(vp, REVOKEALL);
 3397         vn_finished_write(mp);
 3398 out:
 3399         vrele(vp);
 3400         return (error);
 3401 }
 3402 
 3403 /*
 3404  * Convert a user file descriptor to a kernel file entry.
 3405  * The file entry is locked upon returning.
 3406  */
 3407 int
 3408 getvnode(fdp, fd, fpp)
 3409         struct filedesc *fdp;
 3410         int fd;
 3411         struct file **fpp;
 3412 {
 3413         int error;
 3414         struct file *fp;
 3415 
 3416         fp = NULL;
 3417         if (fdp == NULL)
 3418                 error = EBADF;
 3419         else {
 3420                 FILEDESC_LOCK(fdp);
 3421                 if ((u_int)fd >= fdp->fd_nfiles ||
 3422                     (fp = fdp->fd_ofiles[fd]) == NULL)
 3423                         error = EBADF;
 3424                 else if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO) {
 3425                         fp = NULL;
 3426                         error = EINVAL;
 3427                 } else {
 3428                         fhold(fp);
 3429                         error = 0;
 3430                 }
 3431                 FILEDESC_UNLOCK(fdp);
 3432         }
 3433         *fpp = fp;
 3434         return (error);
 3435 }
 3436 /*
 3437  * Get (NFS) file handle
 3438  */
 3439 #ifndef _SYS_SYSPROTO_H_
 3440 struct getfh_args {
 3441         char    *fname;
 3442         fhandle_t *fhp;
 3443 };
 3444 #endif
 3445 int
 3446 getfh(td, uap)
 3447         struct thread *td;
 3448         register struct getfh_args *uap;
 3449 {
 3450         struct nameidata nd;
 3451         fhandle_t fh;
 3452         register struct vnode *vp;
 3453         int error;
 3454 
 3455         /*
 3456          * Must be super user
 3457          */
 3458         error = suser(td);
 3459         if (error)
 3460                 return (error);
 3461         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, uap->fname, td);
 3462         error = namei(&nd);
 3463         if (error)
 3464                 return (error);
 3465         NDFREE(&nd, NDF_ONLY_PNBUF);
 3466         vp = nd.ni_vp;
 3467         bzero(&fh, sizeof(fh));
 3468         fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
 3469         error = VFS_VPTOFH(vp, &fh.fh_fid);
 3470         vput(vp);
 3471         if (error)
 3472                 return (error);
 3473         error = copyout(&fh, uap->fhp, sizeof (fh));
 3474         return (error);
 3475 }
 3476 
 3477 /*
 3478  * syscall for the rpc.lockd to use to translate a NFS file handle into
 3479  * an open descriptor.
 3480  *
 3481  * warning: do not remove the suser() call or this becomes one giant
 3482  * security hole.
 3483  */
 3484 #ifndef _SYS_SYSPROTO_H_
 3485 struct fhopen_args {
 3486         const struct fhandle *u_fhp;
 3487         int flags;
 3488 };
 3489 #endif
 3490 int
 3491 fhopen(td, uap)
 3492         struct thread *td;
 3493         struct fhopen_args /* {
 3494                 const struct fhandle *u_fhp;
 3495                 int flags;
 3496         } */ *uap;
 3497 {
 3498         struct proc *p = td->td_proc;
 3499         struct mount *mp;
 3500         struct vnode *vp;
 3501         struct fhandle fhp;
 3502         struct vattr vat;
 3503         struct vattr *vap = &vat;
 3504         struct flock lf;
 3505         struct file *fp;
 3506         register struct filedesc *fdp = p->p_fd;
 3507         int fmode, mode, error, type;
 3508         struct file *nfp; 
 3509         int indx;
 3510 
 3511         /*
 3512          * Must be super user
 3513          */
 3514         error = suser(td);
 3515         if (error)
 3516                 return (error);
 3517 
 3518         fmode = FFLAGS(uap->flags);
 3519         /* why not allow a non-read/write open for our lockd? */
 3520         if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
 3521                 return (EINVAL);
 3522         error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
 3523         if (error)
 3524                 return(error);
 3525         /* find the mount point */
 3526         mp = vfs_getvfs(&fhp.fh_fsid);
 3527         if (mp == NULL)
 3528                 return (ESTALE);
 3529         /* now give me my vnode, it gets returned to me locked */
 3530         error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp);
 3531         if (error)
 3532                 return (error);
 3533         /*
 3534          * from now on we have to make sure not
 3535          * to forget about the vnode
 3536          * any error that causes an abort must vput(vp) 
 3537          * just set error = err and 'goto bad;'.
 3538          */
 3539 
 3540         /* 
 3541          * from vn_open 
 3542          */
 3543         if (vp->v_type == VLNK) {
 3544                 error = EMLINK;
 3545                 goto bad;
 3546         }
 3547         if (vp->v_type == VSOCK) {
 3548                 error = EOPNOTSUPP;
 3549                 goto bad;
 3550         }
 3551         mode = 0;
 3552         if (fmode & (FWRITE | O_TRUNC)) {
 3553                 if (vp->v_type == VDIR) {
 3554                         error = EISDIR;
 3555                         goto bad;
 3556                 }
 3557                 error = vn_writechk(vp);
 3558                 if (error)
 3559                         goto bad;
 3560                 mode |= VWRITE;
 3561         }
 3562         if (fmode & FREAD)
 3563                 mode |= VREAD;
 3564         if (fmode & O_APPEND)
 3565                 mode |= VAPPEND;
 3566 #ifdef MAC
 3567         error = mac_check_vnode_open(td->td_ucred, vp, mode);
 3568         if (error)
 3569                 goto bad;
 3570 #endif
 3571         if (mode) {
 3572                 error = VOP_ACCESS(vp, mode, td->td_ucred, td);
 3573                 if (error)
 3574                         goto bad;
 3575         }
 3576         if (fmode & O_TRUNC) {
 3577                 VOP_UNLOCK(vp, 0, td);                          /* XXX */
 3578                 if ((error = vn_start_write(NULL, &mp, V_WAIT | PCATCH)) != 0) {
 3579                         vrele(vp);
 3580                         return (error);
 3581                 }
 3582                 VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 3583                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);       /* XXX */
 3584 #ifdef MAC
 3585                 /*
 3586                  * We don't yet have fp->f_cred, so use td->td_ucred, which
 3587                  * should be right.
 3588                  */
 3589                 error = mac_check_vnode_write(td->td_ucred, td->td_ucred, vp);
 3590                 if (error == 0) {
 3591 #endif
 3592                         VATTR_NULL(vap);
 3593                         vap->va_size = 0;
 3594                         error = VOP_SETATTR(vp, vap, td->td_ucred, td);
 3595 #ifdef MAC
 3596                 }
 3597 #endif
 3598                 vn_finished_write(mp);
 3599                 if (error)
 3600                         goto bad;
 3601         }
 3602         error = VOP_OPEN(vp, fmode, td->td_ucred, td);
 3603         if (error)
 3604                 goto bad;
 3605         /*
 3606          * Make sure that a VM object is created for VMIO support.
 3607          */
 3608         if (vn_canvmio(vp) == TRUE) {
 3609                 if ((error = vfs_object_create(vp, td, td->td_ucred)) != 0)
 3610                         goto bad;
 3611         }
 3612         if (fmode & FWRITE)
 3613                 vp->v_writecount++;
 3614 
 3615         /*
 3616          * end of vn_open code 
 3617          */
 3618 
 3619         if ((error = falloc(td, &nfp, &indx)) != 0) {
 3620                 if (fmode & FWRITE)
 3621                         vp->v_writecount--;
 3622                 goto bad;
 3623         }
 3624         fp = nfp;       
 3625 
 3626         /*
 3627          * Hold an extra reference to avoid having fp ripped out 
 3628          * from under us while we block in the lock op
 3629          */
 3630         fhold(fp);
 3631         nfp->f_data = vp;
 3632         nfp->f_flag = fmode & FMASK;
 3633         nfp->f_ops = &vnops;
 3634         nfp->f_type = DTYPE_VNODE;
 3635         if (fmode & (O_EXLOCK | O_SHLOCK)) {
 3636                 lf.l_whence = SEEK_SET;
 3637                 lf.l_start = 0;
 3638                 lf.l_len = 0;
 3639                 if (fmode & O_EXLOCK)
 3640                         lf.l_type = F_WRLCK;
 3641                 else
 3642                         lf.l_type = F_RDLCK;
 3643                 type = F_FLOCK;
 3644                 if ((fmode & FNONBLOCK) == 0)
 3645                         type |= F_WAIT;
 3646                 VOP_UNLOCK(vp, 0, td);
 3647                 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
 3648                             type)) != 0) {
 3649                         /*
 3650                          * The lock request failed.  Normally close the
 3651                          * descriptor but handle the case where someone might
 3652                          * have dup()d or close()d it when we weren't looking.
 3653                          */
 3654                         FILEDESC_LOCK(fdp);
 3655                         if (fdp->fd_ofiles[indx] == fp) {
 3656                                 fdp->fd_ofiles[indx] = NULL;
 3657                                 FILEDESC_UNLOCK(fdp);
 3658                                 fdrop(fp, td);
 3659                         } else
 3660                                 FILEDESC_UNLOCK(fdp);
 3661                         /*
 3662                          * release our private reference
 3663                          */
 3664                         fdrop(fp, td);
 3665                         return(error);
 3666                 }
 3667                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3668                 fp->f_flag |= FHASLOCK;
 3669         }
 3670         if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0))
 3671                 vfs_object_create(vp, td, td->td_ucred);
 3672 
 3673         VOP_UNLOCK(vp, 0, td);
 3674         fdrop(fp, td);
 3675         td->td_retval[0] = indx;
 3676         return (0);
 3677 
 3678 bad:
 3679         vput(vp);
 3680         return (error);
 3681 }
 3682 
 3683 /*
 3684  * Stat an (NFS) file handle.
 3685  */
 3686 #ifndef _SYS_SYSPROTO_H_
 3687 struct fhstat_args {
 3688         struct fhandle *u_fhp;
 3689         struct stat *sb;
 3690 };
 3691 #endif
 3692 int
 3693 fhstat(td, uap)
 3694         struct thread *td;
 3695         register struct fhstat_args /* {
 3696                 struct fhandle *u_fhp;
 3697                 struct stat *sb;
 3698         } */ *uap;
 3699 {
 3700         struct stat sb;
 3701         fhandle_t fh;
 3702         struct mount *mp;
 3703         struct vnode *vp;
 3704         int error;
 3705 
 3706         /*
 3707          * Must be super user
 3708          */
 3709         error = suser(td);
 3710         if (error)
 3711                 return (error);
 3712         
 3713         error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
 3714         if (error)
 3715                 return (error);
 3716 
 3717         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
 3718                 return (ESTALE);
 3719         if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
 3720                 return (error);
 3721         error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td);
 3722         vput(vp);
 3723         if (error)
 3724                 return (error);
 3725         error = copyout(&sb, uap->sb, sizeof(sb));
 3726         return (error);
 3727 }
 3728 
 3729 /*
 3730  * Implement fstatfs() for (NFS) file handles.
 3731  */
 3732 #ifndef _SYS_SYSPROTO_H_
 3733 struct fhstatfs_args {
 3734         struct fhandle *u_fhp;
 3735         struct statfs *buf;
 3736 };
 3737 #endif
 3738 int
 3739 fhstatfs(td, uap)
 3740         struct thread *td;
 3741         struct fhstatfs_args /* {
 3742                 struct fhandle *u_fhp;
 3743                 struct statfs *buf;
 3744         } */ *uap;
 3745 {
 3746         struct statfs *sp;
 3747         struct mount *mp;
 3748         struct vnode *vp;
 3749         struct statfs sb;
 3750         fhandle_t fh;
 3751         int error;
 3752 
 3753         /*
 3754          * Must be super user
 3755          */
 3756         error = suser(td);
 3757         if (error)
 3758                 return (error);
 3759 
 3760         if ((error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t))) != 0)
 3761                 return (error);
 3762 
 3763         if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL)
 3764                 return (ESTALE);
 3765         if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)))
 3766                 return (error);
 3767         mp = vp->v_mount;
 3768         sp = &mp->mnt_stat;
 3769         vput(vp);
 3770 #ifdef MAC
 3771         error = mac_check_mount_stat(td->td_ucred, mp);
 3772         if (error)
 3773                 return (error);
 3774 #endif
 3775         if ((error = VFS_STATFS(mp, sp, td)) != 0)
 3776                 return (error);
 3777         sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
 3778         if (suser(td)) {
 3779                 bcopy(sp, &sb, sizeof(sb));
 3780                 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
 3781                 sp = &sb;
 3782         }
 3783         return (copyout(sp, uap->buf, sizeof(*sp)));
 3784 }
 3785 
 3786 /*
 3787  * Syscall to push extended attribute configuration information into the
 3788  * VFS.  Accepts a path, which it converts to a mountpoint, as well as
 3789  * a command (int cmd), and attribute name and misc data.  For now, the
 3790  * attribute name is left in userspace for consumption by the VFS_op.
 3791  * It will probably be changed to be copied into sysspace by the
 3792  * syscall in the future, once issues with various consumers of the
 3793  * attribute code have raised their hands.
 3794  *
 3795  * Currently this is used only by UFS Extended Attributes.
 3796  */
 3797 int
 3798 extattrctl(td, uap)
 3799         struct thread *td;
 3800         struct extattrctl_args /* {
 3801                 const char *path;
 3802                 int cmd;
 3803                 const char *filename;
 3804                 int attrnamespace;
 3805                 const char *attrname;
 3806         } */ *uap;
 3807 {
 3808         struct vnode *filename_vp;
 3809         struct nameidata nd;
 3810         struct mount *mp, *mp_writable;
 3811         char attrname[EXTATTR_MAXNAMELEN];
 3812         int error;
 3813 
 3814         /*
 3815          * uap->attrname is not always defined.  We check again later when we
 3816          * invoke the VFS call so as to pass in NULL there if needed.
 3817          */
 3818         if (uap->attrname != NULL) {
 3819                 error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN,
 3820                     NULL);
 3821                 if (error)
 3822                         return (error);
 3823         }
 3824 
 3825         /*
 3826          * uap->filename is not always defined.  If it is, grab a vnode lock,
 3827          * which VFS_EXTATTRCTL() will later release.
 3828          */
 3829         filename_vp = NULL;
 3830         if (uap->filename != NULL) {
 3831                 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE,
 3832                     uap->filename, td);
 3833                 error = namei(&nd);
 3834                 if (error)
 3835                         return (error);
 3836                 filename_vp = nd.ni_vp;
 3837                 NDFREE(&nd, NDF_NO_VP_RELE | NDF_NO_VP_UNLOCK);
 3838         }
 3839 
 3840         /* uap->path is always defined. */
 3841         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 3842         error = namei(&nd);
 3843         if (error) {
 3844                 if (filename_vp != NULL)
 3845                         vput(filename_vp);
 3846                 return (error);
 3847         }
 3848         mp = nd.ni_vp->v_mount;
 3849         error = vn_start_write(nd.ni_vp, &mp_writable, V_WAIT | PCATCH);
 3850         NDFREE(&nd, 0);
 3851         if (error) {
 3852                 if (filename_vp != NULL)
 3853                         vput(filename_vp);
 3854                 return (error);
 3855         }
 3856 
 3857         error = VFS_EXTATTRCTL(mp, uap->cmd, filename_vp, uap->attrnamespace,
 3858             uap->attrname != NULL ? attrname : NULL, td);
 3859 
 3860         vn_finished_write(mp_writable);
 3861         /*
 3862          * VFS_EXTATTRCTL will have unlocked, but not de-ref'd,
 3863          * filename_vp, so vrele it if it is defined.
 3864          */
 3865         if (filename_vp != NULL)
 3866                 vrele(filename_vp);
 3867         return (error);
 3868 }
 3869 
 3870 /*-
 3871  * Set a named extended attribute on a file or directory
 3872  * 
 3873  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
 3874  *            kernelspace string pointer "attrname", userspace buffer
 3875  *            pointer "data", buffer length "nbytes", thread "td".
 3876  * Returns: 0 on success, an error number otherwise
 3877  * Locks: none
 3878  * References: vp must be a valid reference for the duration of the call
 3879  */
 3880 static int
 3881 extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname,
 3882     void *data, size_t nbytes, struct thread *td)
 3883 {
 3884         struct mount *mp;
 3885         struct uio auio;
 3886         struct iovec aiov;
 3887         ssize_t cnt;
 3888         int error;
 3889 
 3890         error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
 3891         if (error)
 3892                 return (error);
 3893         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 3894         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 3895 
 3896         aiov.iov_base = data;
 3897         aiov.iov_len = nbytes;
 3898         auio.uio_iov = &aiov;
 3899         auio.uio_iovcnt = 1;
 3900         auio.uio_offset = 0;
 3901         if (nbytes > INT_MAX) {
 3902                 error = EINVAL;
 3903                 goto done;
 3904         }
 3905         auio.uio_resid = nbytes;
 3906         auio.uio_rw = UIO_WRITE;
 3907         auio.uio_segflg = UIO_USERSPACE;
 3908         auio.uio_td = td;
 3909         cnt = nbytes;
 3910 
 3911 #ifdef MAC
 3912         error = mac_check_vnode_setextattr(td->td_ucred, vp, attrnamespace,
 3913             attrname, &auio);
 3914         if (error)
 3915                 goto done;
 3916 #endif
 3917 
 3918         error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio,
 3919             td->td_ucred, td);
 3920         cnt -= auio.uio_resid;
 3921         td->td_retval[0] = cnt;
 3922 
 3923 done:
 3924         VOP_UNLOCK(vp, 0, td);
 3925         vn_finished_write(mp);
 3926         return (error);
 3927 }
 3928 
 3929 int
 3930 extattr_set_fd(td, uap)
 3931         struct thread *td;
 3932         struct extattr_set_fd_args /* {
 3933                 int fd;
 3934                 int attrnamespace;
 3935                 const char *attrname;
 3936                 void *data;
 3937                 size_t nbytes;
 3938         } */ *uap;
 3939 {
 3940         struct file *fp;
 3941         char attrname[EXTATTR_MAXNAMELEN];
 3942         int error;
 3943 
 3944         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 3945         if (error)
 3946                 return (error);
 3947 
 3948         error = getvnode(td->td_proc->p_fd, uap->fd, &fp);
 3949         if (error)
 3950                 return (error);
 3951 
 3952         error = extattr_set_vp((struct vnode *)fp->f_data, uap->attrnamespace,
 3953             attrname, uap->data, uap->nbytes, td);
 3954         fdrop(fp, td);
 3955 
 3956         return (error);
 3957 }
 3958 
 3959 int
 3960 extattr_set_file(td, uap)
 3961         struct thread *td;
 3962         struct extattr_set_file_args /* {
 3963                 const char *path;
 3964                 int attrnamespace;
 3965                 const char *attrname;
 3966                 void *data;
 3967                 size_t nbytes;
 3968         } */ *uap;
 3969 {
 3970         struct nameidata nd;
 3971         char attrname[EXTATTR_MAXNAMELEN];
 3972         int error;
 3973 
 3974         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 3975         if (error)
 3976                 return (error);
 3977 
 3978         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 3979         error = namei(&nd);
 3980         if (error)
 3981                 return (error);
 3982         NDFREE(&nd, NDF_ONLY_PNBUF);
 3983 
 3984         error = extattr_set_vp(nd.ni_vp, uap->attrnamespace, attrname,
 3985             uap->data, uap->nbytes, td);
 3986 
 3987         vrele(nd.ni_vp);
 3988         return (error);
 3989 }
 3990 
 3991 int
 3992 extattr_set_link(td, uap)
 3993         struct thread *td;
 3994         struct extattr_set_link_args /* {
 3995                 const char *path;
 3996                 int attrnamespace;
 3997                 const char *attrname;
 3998                 void *data;
 3999                 size_t nbytes;
 4000         } */ *uap;
 4001 {
 4002         struct nameidata nd;
 4003         char attrname[EXTATTR_MAXNAMELEN];
 4004         int error;
 4005 
 4006         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4007         if (error)
 4008                 return (error);
 4009 
 4010         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 4011         error = namei(&nd);
 4012         if (error)
 4013                 return (error);
 4014         NDFREE(&nd, NDF_ONLY_PNBUF);
 4015 
 4016         error = extattr_set_vp(nd.ni_vp, uap->attrnamespace, attrname,
 4017             uap->data, uap->nbytes, td);
 4018 
 4019         vrele(nd.ni_vp);
 4020         return (error);
 4021 }
 4022 
 4023 /*-
 4024  * Get a named extended attribute on a file or directory
 4025  * 
 4026  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
 4027  *            kernelspace string pointer "attrname", userspace buffer
 4028  *            pointer "data", buffer length "nbytes", thread "td".
 4029  * Returns: 0 on success, an error number otherwise
 4030  * Locks: none
 4031  * References: vp must be a valid reference for the duration of the call
 4032  */
 4033 static int
 4034 extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname,
 4035     void *data, size_t nbytes, struct thread *td)
 4036 {
 4037         struct uio auio, *auiop;
 4038         struct iovec aiov;
 4039         ssize_t cnt;
 4040         size_t size, *sizep;
 4041         int error;
 4042 
 4043         VOP_LEASE(vp, td, td->td_ucred, LEASE_READ);
 4044         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 4045 
 4046         /*
 4047          * Slightly unusual semantics: if the user provides a NULL data
 4048          * pointer, they don't want to receive the data, just the
 4049          * maximum read length.
 4050          */
 4051         auiop = NULL;
 4052         sizep = NULL;
 4053         cnt = 0;
 4054         if (data != NULL) {
 4055                 aiov.iov_base = data;
 4056                 aiov.iov_len = nbytes;
 4057                 auio.uio_iov = &aiov;
 4058                 auio.uio_offset = 0;
 4059                 if (nbytes > INT_MAX) {
 4060                         error = EINVAL;
 4061                         goto done;
 4062                 }
 4063                 auio.uio_resid = nbytes;
 4064                 auio.uio_rw = UIO_READ;
 4065                 auio.uio_segflg = UIO_USERSPACE;
 4066                 auio.uio_td = td;
 4067                 auiop = &auio;
 4068                 cnt = nbytes;
 4069         } else
 4070                 sizep = &size;
 4071 
 4072 #ifdef MAC
 4073         error = mac_check_vnode_getextattr(td->td_ucred, vp, attrnamespace,
 4074             attrname, &auio);
 4075         if (error)
 4076                 goto done;
 4077 #endif
 4078 
 4079         error = VOP_GETEXTATTR(vp, attrnamespace, attrname, auiop, sizep,
 4080             td->td_ucred, td);
 4081 
 4082         if (auiop != NULL) {
 4083                 cnt -= auio.uio_resid;
 4084                 td->td_retval[0] = cnt;
 4085         } else
 4086                 td->td_retval[0] = size;
 4087 
 4088 done:
 4089         VOP_UNLOCK(vp, 0, td);
 4090         return (error);
 4091 }
 4092 
 4093 int
 4094 extattr_get_fd(td, uap)
 4095         struct thread *td;
 4096         struct extattr_get_fd_args /* {
 4097                 int fd;
 4098                 int attrnamespace;
 4099                 const char *attrname;
 4100                 void *data;
 4101                 size_t nbytes;
 4102         } */ *uap;
 4103 {
 4104         struct file *fp;
 4105         char attrname[EXTATTR_MAXNAMELEN];
 4106         int error;
 4107 
 4108         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4109         if (error)
 4110                 return (error);
 4111 
 4112         error = getvnode(td->td_proc->p_fd, uap->fd, &fp);
 4113         if (error)
 4114                 return (error);
 4115 
 4116         error = extattr_get_vp((struct vnode *)fp->f_data, uap->attrnamespace,
 4117             attrname, uap->data, uap->nbytes, td);
 4118 
 4119         fdrop(fp, td);
 4120         return (error);
 4121 }
 4122 
 4123 int
 4124 extattr_get_file(td, uap)
 4125         struct thread *td;
 4126         struct extattr_get_file_args /* {
 4127                 const char *path;
 4128                 int attrnamespace;
 4129                 const char *attrname;
 4130                 void *data;
 4131                 size_t nbytes;
 4132         } */ *uap;
 4133 {
 4134         struct nameidata nd;
 4135         char attrname[EXTATTR_MAXNAMELEN];
 4136         int error;
 4137 
 4138         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4139         if (error)
 4140                 return (error);
 4141 
 4142         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 4143         error = namei(&nd);
 4144         if (error)
 4145                 return (error);
 4146         NDFREE(&nd, NDF_ONLY_PNBUF);
 4147 
 4148         error = extattr_get_vp(nd.ni_vp, uap->attrnamespace, attrname,
 4149             uap->data, uap->nbytes, td);
 4150 
 4151         vrele(nd.ni_vp);
 4152         return (error);
 4153 }
 4154 
 4155 int
 4156 extattr_get_link(td, uap)
 4157         struct thread *td;
 4158         struct extattr_get_link_args /* {
 4159                 const char *path;
 4160                 int attrnamespace;
 4161                 const char *attrname;
 4162                 void *data;
 4163                 size_t nbytes;
 4164         } */ *uap;
 4165 {
 4166         struct nameidata nd;
 4167         char attrname[EXTATTR_MAXNAMELEN];
 4168         int error;
 4169 
 4170         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4171         if (error)
 4172                 return (error);
 4173 
 4174         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 4175         error = namei(&nd);
 4176         if (error)
 4177                 return (error);
 4178         NDFREE(&nd, NDF_ONLY_PNBUF);
 4179 
 4180         error = extattr_get_vp(nd.ni_vp, uap->attrnamespace, attrname,
 4181             uap->data, uap->nbytes, td);
 4182 
 4183         vrele(nd.ni_vp);
 4184         return (error);
 4185 }
 4186 
 4187 /*
 4188  * extattr_delete_vp(): Delete a named extended attribute on a file or
 4189  *                      directory
 4190  * 
 4191  * Arguments: unlocked vnode "vp", attribute namespace "attrnamespace",
 4192  *            kernelspace string pointer "attrname", proc "p"
 4193  * Returns: 0 on success, an error number otherwise
 4194  * Locks: none
 4195  * References: vp must be a valid reference for the duration of the call
 4196  */
 4197 static int
 4198 extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname,
 4199     struct thread *td)
 4200 {
 4201         struct mount *mp;
 4202         int error;
 4203 
 4204         error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
 4205         if (error)
 4206                 return (error);
 4207         VOP_LEASE(vp, td, td->td_ucred, LEASE_WRITE);
 4208         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
 4209 
 4210 #ifdef MAC
 4211         error = mac_check_vnode_setextattr(td->td_ucred, vp, attrnamespace,
 4212             attrname, NULL);
 4213         if (error)
 4214                 goto done;
 4215 #endif
 4216 
 4217         error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, td->td_ucred,
 4218             td);
 4219 #ifdef MAC
 4220 done:
 4221 #endif
 4222         VOP_UNLOCK(vp, 0, td);
 4223         vn_finished_write(mp);
 4224         return (error);
 4225 }
 4226 
 4227 int
 4228 extattr_delete_fd(td, uap)
 4229         struct thread *td;
 4230         struct extattr_delete_fd_args /* {
 4231                 int fd;
 4232                 int attrnamespace;
 4233                 const char *attrname;
 4234         } */ *uap;
 4235 {
 4236         struct file *fp;
 4237         struct vnode *vp;
 4238         char attrname[EXTATTR_MAXNAMELEN];
 4239         int error;
 4240 
 4241         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4242         if (error)
 4243                 return (error);
 4244 
 4245         error = getvnode(td->td_proc->p_fd, uap->fd, &fp);
 4246         if (error)
 4247                 return (error);
 4248         vp = (struct vnode *)fp->f_data;
 4249 
 4250         error = extattr_delete_vp(vp, uap->attrnamespace, attrname, td);
 4251         fdrop(fp, td);
 4252         return (error);
 4253 }
 4254 
 4255 int
 4256 extattr_delete_file(td, uap)
 4257         struct thread *td;
 4258         struct extattr_delete_file_args /* {
 4259                 const char *path;
 4260                 int attrnamespace;
 4261                 const char *attrname;
 4262         } */ *uap;
 4263 {
 4264         struct nameidata nd;
 4265         char attrname[EXTATTR_MAXNAMELEN];
 4266         int error;
 4267 
 4268         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4269         if (error)
 4270                 return(error);
 4271 
 4272         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->path, td);
 4273         error = namei(&nd);
 4274         if (error)
 4275                 return(error);
 4276         NDFREE(&nd, NDF_ONLY_PNBUF);
 4277 
 4278         error = extattr_delete_vp(nd.ni_vp, uap->attrnamespace, attrname, td);
 4279         vrele(nd.ni_vp);
 4280         return(error);
 4281 }
 4282 
 4283 int
 4284 extattr_delete_link(td, uap)
 4285         struct thread *td;
 4286         struct extattr_delete_link_args /* {
 4287                 const char *path;
 4288                 int attrnamespace;
 4289                 const char *attrname;
 4290         } */ *uap;
 4291 {
 4292         struct nameidata nd;
 4293         char attrname[EXTATTR_MAXNAMELEN];
 4294         int error;
 4295 
 4296         error = copyinstr(uap->attrname, attrname, EXTATTR_MAXNAMELEN, NULL);
 4297         if (error)
 4298                 return(error);
 4299 
 4300         NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
 4301         error = namei(&nd);
 4302         if (error)
 4303                 return(error);
 4304         NDFREE(&nd, NDF_ONLY_PNBUF);
 4305 
 4306         error = extattr_delete_vp(nd.ni_vp, uap->attrnamespace, attrname, td);
 4307         vrele(nd.ni_vp);
 4308         return(error);
 4309 }

Cache object: 3a99423c41a9c635abca8b574281a099


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.