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/ufs/ufs/ufs_vnops.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) 1982, 1986, 1989, 1993, 1995
    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  *      @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
   39  * $FreeBSD: releng/5.1/sys/ufs/ufs/ufs_vnops.c 111841 2003-03-03 19:15:40Z njl $
   40  */
   41 
   42 #include "opt_mac.h"
   43 #include "opt_quota.h"
   44 #include "opt_suiddir.h"
   45 #include "opt_ufs.h"
   46 
   47 #include <sys/param.h>
   48 #include <sys/systm.h>
   49 #include <sys/malloc.h>
   50 #include <sys/namei.h>
   51 #include <sys/kernel.h>
   52 #include <sys/fcntl.h>
   53 #include <sys/stat.h>
   54 #include <sys/bio.h>
   55 #include <sys/buf.h>
   56 #include <sys/mount.h>
   57 #include <sys/unistd.h>
   58 #include <sys/vnode.h>
   59 #include <sys/dirent.h>
   60 #include <sys/lockf.h>
   61 #include <sys/event.h>
   62 #include <sys/conf.h>
   63 #include <sys/acl.h>
   64 #include <sys/mac.h>
   65 
   66 #include <machine/mutex.h>
   67 
   68 #include <sys/file.h>           /* XXX */
   69 
   70 #include <vm/vm.h>
   71 #include <vm/vm_extern.h>
   72 
   73 #include <fs/fifofs/fifo.h>
   74 
   75 #include <ufs/ufs/acl.h>
   76 #include <ufs/ufs/extattr.h>
   77 #include <ufs/ufs/quota.h>
   78 #include <ufs/ufs/inode.h>
   79 #include <ufs/ufs/dir.h>
   80 #include <ufs/ufs/ufsmount.h>
   81 #include <ufs/ufs/ufs_extern.h>
   82 #ifdef UFS_DIRHASH
   83 #include <ufs/ufs/dirhash.h>
   84 #endif
   85 
   86 static int ufs_access(struct vop_access_args *);
   87 static int ufs_advlock(struct vop_advlock_args *);
   88 static int ufs_chmod(struct vnode *, int, struct ucred *, struct thread *);
   89 static int ufs_chown(struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
   90 static int ufs_close(struct vop_close_args *);
   91 static int ufs_create(struct vop_create_args *);
   92 static int ufs_getattr(struct vop_getattr_args *);
   93 static int ufs_link(struct vop_link_args *);
   94 static int ufs_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
   95 static int ufs_mkdir(struct vop_mkdir_args *);
   96 static int ufs_mknod(struct vop_mknod_args *);
   97 static int ufs_open(struct vop_open_args *);
   98 static int ufs_pathconf(struct vop_pathconf_args *);
   99 static int ufs_print(struct vop_print_args *);
  100 static int ufs_readlink(struct vop_readlink_args *);
  101 static int ufs_remove(struct vop_remove_args *);
  102 static int ufs_rename(struct vop_rename_args *);
  103 static int ufs_rmdir(struct vop_rmdir_args *);
  104 static int ufs_setattr(struct vop_setattr_args *);
  105 static int ufs_strategy(struct vop_strategy_args *);
  106 static int ufs_symlink(struct vop_symlink_args *);
  107 static int ufs_whiteout(struct vop_whiteout_args *);
  108 static int ufsfifo_close(struct vop_close_args *);
  109 static int ufsfifo_kqfilter(struct vop_kqfilter_args *);
  110 static int ufsfifo_read(struct vop_read_args *);
  111 static int ufsfifo_write(struct vop_write_args *);
  112 static int ufsspec_close(struct vop_close_args *);
  113 static int ufsspec_read(struct vop_read_args *);
  114 static int ufsspec_write(struct vop_write_args *);
  115 static int filt_ufsread(struct knote *kn, long hint);
  116 static int filt_ufswrite(struct knote *kn, long hint);
  117 static int filt_ufsvnode(struct knote *kn, long hint);
  118 static void filt_ufsdetach(struct knote *kn);
  119 static int ufs_kqfilter(struct vop_kqfilter_args *ap);
  120 
  121 union _qcvt {
  122         int64_t qcvt;
  123         int32_t val[2];
  124 };
  125 #define SETHIGH(q, h) { \
  126         union _qcvt tmp; \
  127         tmp.qcvt = (q); \
  128         tmp.val[_QUAD_HIGHWORD] = (h); \
  129         (q) = tmp.qcvt; \
  130 }
  131 #define SETLOW(q, l) { \
  132         union _qcvt tmp; \
  133         tmp.qcvt = (q); \
  134         tmp.val[_QUAD_LOWWORD] = (l); \
  135         (q) = tmp.qcvt; \
  136 }
  137 
  138 /*
  139  * A virgin directory (no blushing please).
  140  */
  141 static struct dirtemplate mastertemplate = {
  142         0, 12, DT_DIR, 1, ".",
  143         0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
  144 };
  145 static struct odirtemplate omastertemplate = {
  146         0, 12, 1, ".",
  147         0, DIRBLKSIZ - 12, 2, ".."
  148 };
  149 
  150 void
  151 ufs_itimes(vp)
  152         struct vnode *vp;
  153 {
  154         struct inode *ip;
  155         struct timespec ts;
  156 
  157         ip = VTOI(vp);
  158         if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
  159                 return;
  160         if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
  161                 ip->i_flag |= IN_LAZYMOD;
  162         else
  163                 ip->i_flag |= IN_MODIFIED;
  164         if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
  165                 vfs_timestamp(&ts);
  166                 if (ip->i_flag & IN_ACCESS) {
  167                         DIP(ip, i_atime) = ts.tv_sec;
  168                         DIP(ip, i_atimensec) = ts.tv_nsec;
  169                 }
  170                 if (ip->i_flag & IN_UPDATE) {
  171                         DIP(ip, i_mtime) = ts.tv_sec;
  172                         DIP(ip, i_mtimensec) = ts.tv_nsec;
  173                         ip->i_modrev++;
  174                 }
  175                 if (ip->i_flag & IN_CHANGE) {
  176                         DIP(ip, i_ctime) = ts.tv_sec;
  177                         DIP(ip, i_ctimensec) = ts.tv_nsec;
  178                 }
  179         }
  180         ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
  181 }
  182 
  183 /*
  184  * Create a regular file
  185  */
  186 static int
  187 ufs_create(ap)
  188         struct vop_create_args /* {
  189                 struct vnode *a_dvp;
  190                 struct vnode **a_vpp;
  191                 struct componentname *a_cnp;
  192                 struct vattr *a_vap;
  193         } */ *ap;
  194 {
  195         int error;
  196 
  197         error =
  198             ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
  199             ap->a_dvp, ap->a_vpp, ap->a_cnp);
  200         if (error)
  201                 return (error);
  202         VN_KNOTE(ap->a_dvp, NOTE_WRITE);
  203         return (0);
  204 }
  205 
  206 /*
  207  * Mknod vnode call
  208  */
  209 /* ARGSUSED */
  210 static int
  211 ufs_mknod(ap)
  212         struct vop_mknod_args /* {
  213                 struct vnode *a_dvp;
  214                 struct vnode **a_vpp;
  215                 struct componentname *a_cnp;
  216                 struct vattr *a_vap;
  217         } */ *ap;
  218 {
  219         struct vattr *vap = ap->a_vap;
  220         struct vnode **vpp = ap->a_vpp;
  221         struct inode *ip;
  222         ino_t ino;
  223         int error;
  224 
  225         error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
  226             ap->a_dvp, vpp, ap->a_cnp);
  227         if (error)
  228                 return (error);
  229         VN_KNOTE(ap->a_dvp, NOTE_WRITE);
  230         ip = VTOI(*vpp);
  231         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
  232         if (vap->va_rdev != VNOVAL) {
  233                 /*
  234                  * Want to be able to use this to make badblock
  235                  * inodes, so don't truncate the dev number.
  236                  */
  237                 DIP(ip, i_rdev) = vap->va_rdev;
  238         }
  239         /*
  240          * Remove inode, then reload it through VFS_VGET so it is
  241          * checked to see if it is an alias of an existing entry in
  242          * the inode cache.
  243          */
  244         vput(*vpp);
  245         (*vpp)->v_type = VNON;
  246         ino = ip->i_number;     /* Save this before vgone() invalidates ip. */
  247         vgone(*vpp);
  248         error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
  249         if (error) {
  250                 *vpp = NULL;
  251                 return (error);
  252         }
  253         return (0);
  254 }
  255 
  256 /*
  257  * Open called.
  258  *
  259  * Nothing to do.
  260  */
  261 /* ARGSUSED */
  262 static int
  263 ufs_open(ap)
  264         struct vop_open_args /* {
  265                 struct vnode *a_vp;
  266                 int  a_mode;
  267                 struct ucred *a_cred;
  268                 struct thread *a_td;
  269         } */ *ap;
  270 {
  271 
  272         /*
  273          * Files marked append-only must be opened for appending.
  274          */
  275         if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
  276             (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
  277                 return (EPERM);
  278         return (0);
  279 }
  280 
  281 /*
  282  * Close called.
  283  *
  284  * Update the times on the inode.
  285  */
  286 /* ARGSUSED */
  287 static int
  288 ufs_close(ap)
  289         struct vop_close_args /* {
  290                 struct vnode *a_vp;
  291                 int  a_fflag;
  292                 struct ucred *a_cred;
  293                 struct thread *a_td;
  294         } */ *ap;
  295 {
  296         struct vnode *vp = ap->a_vp;
  297         struct mount *mp;
  298 
  299         VI_LOCK(vp);
  300         if (vp->v_usecount > 1) {
  301                 ufs_itimes(vp);
  302                 VI_UNLOCK(vp);
  303         } else {
  304                 VI_UNLOCK(vp);
  305                 /*
  306                  * If we are closing the last reference to an unlinked
  307                  * file, then it will be freed by the inactive routine.
  308                  * Because the freeing causes a the filesystem to be
  309                  * modified, it must be held up during periods when the
  310                  * filesystem is suspended.
  311                  *
  312                  * XXX - EAGAIN is returned to prevent vn_close from
  313                  * repeating the vrele operation.
  314                  */
  315                 if (vp->v_type == VREG && VTOI(vp)->i_effnlink == 0) {
  316                         (void) vn_start_write(vp, &mp, V_WAIT);
  317                         vrele(vp);
  318                         vn_finished_write(mp);
  319                         return (EAGAIN);
  320                 }
  321         }
  322         return (0);
  323 }
  324 
  325 static int
  326 ufs_access(ap)
  327         struct vop_access_args /* {
  328                 struct vnode *a_vp;
  329                 int  a_mode;
  330                 struct ucred *a_cred;
  331                 struct thread *a_td;
  332         } */ *ap;
  333 {
  334         struct vnode *vp = ap->a_vp;
  335         struct inode *ip = VTOI(vp);
  336         mode_t mode = ap->a_mode;
  337         int error;
  338 #ifdef UFS_ACL
  339         struct acl *acl;
  340         size_t len;
  341 #endif
  342 
  343         /*
  344          * Disallow write attempts on read-only filesystems;
  345          * unless the file is a socket, fifo, or a block or
  346          * character device resident on the filesystem.
  347          */
  348         if (mode & VWRITE) {
  349                 switch (vp->v_type) {
  350                 case VDIR:
  351                 case VLNK:
  352                 case VREG:
  353                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  354                                 return (EROFS);
  355 #ifdef QUOTA
  356                         if ((error = getinoquota(ip)) != 0)
  357                                 return (error);
  358 #endif
  359                         break;
  360                 default:
  361                         break;
  362                 }
  363         }
  364 
  365         /* If immutable bit set, nobody gets to write it. */
  366         if ((mode & VWRITE) && (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT)))
  367                 return (EPERM);
  368 
  369 #ifdef UFS_ACL
  370         if ((vp->v_mount->mnt_flag & MNT_ACLS) != 0) {
  371                 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK);
  372                 len = sizeof(*acl);
  373                 error = VOP_GETACL(vp, ACL_TYPE_ACCESS, acl, ap->a_cred,
  374                     ap->a_td);
  375                 switch (error) {
  376                 case EOPNOTSUPP:
  377                         error = vaccess(vp->v_type, ip->i_mode, ip->i_uid,
  378                             ip->i_gid, ap->a_mode, ap->a_cred, NULL);
  379                         break;
  380                 case 0:
  381                         error = vaccess_acl_posix1e(vp->v_type, ip->i_uid,
  382                             ip->i_gid, acl, ap->a_mode, ap->a_cred, NULL);
  383                         break;
  384                 default:
  385                         printf(
  386 "ufs_access(): Error retrieving ACL on object (%d).\n",
  387                             error);
  388                         /*
  389                          * XXX: Fall back until debugged.  Should
  390                          * eventually possibly log an error, and return
  391                          * EPERM for safety.
  392                          */
  393                         error = vaccess(vp->v_type, ip->i_mode, ip->i_uid,
  394                             ip->i_gid, ap->a_mode, ap->a_cred, NULL);
  395                 }
  396                 FREE(acl, M_ACL);
  397         } else
  398 #endif /* !UFS_ACL */
  399                 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
  400                     ap->a_mode, ap->a_cred, NULL);
  401         return (error);
  402 }
  403 
  404 /* ARGSUSED */
  405 static int
  406 ufs_getattr(ap)
  407         struct vop_getattr_args /* {
  408                 struct vnode *a_vp;
  409                 struct vattr *a_vap;
  410                 struct ucred *a_cred;
  411                 struct thread *a_td;
  412         } */ *ap;
  413 {
  414         struct vnode *vp = ap->a_vp;
  415         struct inode *ip = VTOI(vp);
  416         struct vattr *vap = ap->a_vap;
  417 
  418         ufs_itimes(vp);
  419         /*
  420          * Copy from inode table
  421          */
  422         vap->va_fsid = dev2udev(ip->i_dev);
  423         vap->va_fileid = ip->i_number;
  424         vap->va_mode = ip->i_mode & ~IFMT;
  425         vap->va_nlink = ip->i_effnlink;
  426         vap->va_uid = ip->i_uid;
  427         vap->va_gid = ip->i_gid;
  428         if (ip->i_ump->um_fstype == UFS1) {
  429                 vap->va_rdev = ip->i_din1->di_rdev;
  430                 vap->va_size = ip->i_din1->di_size;
  431                 vap->va_atime.tv_sec = ip->i_din1->di_atime;
  432                 vap->va_atime.tv_nsec = ip->i_din1->di_atimensec;
  433                 vap->va_mtime.tv_sec = ip->i_din1->di_mtime;
  434                 vap->va_mtime.tv_nsec = ip->i_din1->di_mtimensec;
  435                 vap->va_ctime.tv_sec = ip->i_din1->di_ctime;
  436                 vap->va_ctime.tv_nsec = ip->i_din1->di_ctimensec;
  437                 vap->va_birthtime.tv_sec = 0;
  438                 vap->va_birthtime.tv_nsec = 0;
  439                 vap->va_bytes = dbtob((u_quad_t)ip->i_din1->di_blocks);
  440         } else {
  441                 vap->va_rdev = ip->i_din2->di_rdev;
  442                 vap->va_size = ip->i_din2->di_size;
  443                 vap->va_atime.tv_sec = ip->i_din2->di_atime;
  444                 vap->va_atime.tv_nsec = ip->i_din2->di_atimensec;
  445                 vap->va_mtime.tv_sec = ip->i_din2->di_mtime;
  446                 vap->va_mtime.tv_nsec = ip->i_din2->di_mtimensec;
  447                 vap->va_ctime.tv_sec = ip->i_din2->di_ctime;
  448                 vap->va_ctime.tv_nsec = ip->i_din2->di_ctimensec;
  449                 vap->va_birthtime.tv_sec = ip->i_din2->di_birthtime;
  450                 vap->va_birthtime.tv_nsec = ip->i_din2->di_birthnsec;
  451                 vap->va_bytes = dbtob((u_quad_t)ip->i_din2->di_blocks);
  452         }
  453         vap->va_flags = ip->i_flags;
  454         vap->va_gen = ip->i_gen;
  455         vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
  456         vap->va_type = IFTOVT(ip->i_mode);
  457         vap->va_filerev = ip->i_modrev;
  458         return (0);
  459 }
  460 
  461 /*
  462  * Set attribute vnode op. called from several syscalls
  463  */
  464 static int
  465 ufs_setattr(ap)
  466         struct vop_setattr_args /* {
  467                 struct vnode *a_vp;
  468                 struct vattr *a_vap;
  469                 struct ucred *a_cred;
  470                 struct thread *a_td;
  471         } */ *ap;
  472 {
  473         struct vattr *vap = ap->a_vap;
  474         struct vnode *vp = ap->a_vp;
  475         struct inode *ip = VTOI(vp);
  476         struct ucred *cred = ap->a_cred;
  477         struct thread *td = ap->a_td;
  478         int error;
  479 
  480         /*
  481          * Check for unsettable attributes.
  482          */
  483         if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
  484             (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
  485             (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
  486             ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
  487                 return (EINVAL);
  488         }
  489         if (vap->va_flags != VNOVAL) {
  490                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  491                         return (EROFS);
  492                 /*
  493                  * Callers may only modify the file flags on objects they
  494                  * have VADMIN rights for.
  495                  */
  496                 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  497                         return (error);
  498                 /*
  499                  * Unprivileged processes and privileged processes in
  500                  * jail() are not permitted to unset system flags, or
  501                  * modify flags if any system flags are set.
  502                  * Privileged non-jail processes may not modify system flags
  503                  * if securelevel > 0 and any existing system flags are set.
  504                  */
  505                 if (!suser_cred(cred, PRISON_ROOT)) {
  506                         if (ip->i_flags
  507                             & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
  508                                 error = securelevel_gt(cred, 0);
  509                                 if (error)
  510                                         return (error);
  511                         }
  512                         /* Snapshot flag cannot be set or cleared */
  513                         if (((vap->va_flags & SF_SNAPSHOT) != 0 &&
  514                              (ip->i_flags & SF_SNAPSHOT) == 0) ||
  515                             ((vap->va_flags & SF_SNAPSHOT) == 0 &&
  516                              (ip->i_flags & SF_SNAPSHOT) != 0))
  517                                 return (EPERM);
  518                         ip->i_flags = vap->va_flags;
  519                         DIP(ip, i_flags) = vap->va_flags;
  520                 } else {
  521                         if (ip->i_flags
  522                             & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
  523                             (vap->va_flags & UF_SETTABLE) != vap->va_flags)
  524                                 return (EPERM);
  525                         ip->i_flags &= SF_SETTABLE;
  526                         ip->i_flags |= (vap->va_flags & UF_SETTABLE);
  527                         DIP(ip, i_flags) = ip->i_flags;
  528                 }
  529                 ip->i_flag |= IN_CHANGE;
  530                 if (vap->va_flags & (IMMUTABLE | APPEND))
  531                         return (0);
  532         }
  533         if (ip->i_flags & (IMMUTABLE | APPEND))
  534                 return (EPERM);
  535         /*
  536          * Go through the fields and update iff not VNOVAL.
  537          */
  538         if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
  539                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  540                         return (EROFS);
  541                 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred,
  542                     td)) != 0)
  543                         return (error);
  544         }
  545         if (vap->va_size != VNOVAL) {
  546                 /*
  547                  * Disallow write attempts on read-only filesystems;
  548                  * unless the file is a socket, fifo, or a block or
  549                  * character device resident on the filesystem.
  550                  */
  551                 switch (vp->v_type) {
  552                 case VDIR:
  553                         return (EISDIR);
  554                 case VLNK:
  555                 case VREG:
  556                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  557                                 return (EROFS);
  558                         if ((ip->i_flags & SF_SNAPSHOT) != 0)
  559                                 return (EPERM);
  560                         break;
  561                 default:
  562                         break;
  563                 }
  564                 if ((error = UFS_TRUNCATE(vp, vap->va_size, IO_NORMAL,
  565                     cred, td)) != 0)
  566                         return (error);
  567         }
  568         if (vap->va_atime.tv_sec != VNOVAL ||
  569             vap->va_mtime.tv_sec != VNOVAL ||
  570             vap->va_birthtime.tv_sec != VNOVAL) {
  571                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  572                         return (EROFS);
  573                 if ((ip->i_flags & SF_SNAPSHOT) != 0)
  574                         return (EPERM);
  575                 /*
  576                  * From utimes(2):
  577                  * If times is NULL, ... The caller must be the owner of
  578                  * the file, have permission to write the file, or be the
  579                  * super-user.
  580                  * If times is non-NULL, ... The caller must be the owner of
  581                  * the file or be the super-user.
  582                  */
  583                 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) &&
  584                     ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
  585                     (error = VOP_ACCESS(vp, VWRITE, cred, td))))
  586                         return (error);
  587                 if (vap->va_atime.tv_sec != VNOVAL)
  588                         ip->i_flag |= IN_ACCESS;
  589                 if (vap->va_mtime.tv_sec != VNOVAL)
  590                         ip->i_flag |= IN_CHANGE | IN_UPDATE;
  591                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  592                     ip->i_ump->um_fstype == UFS2)
  593                         ip->i_flag |= IN_MODIFIED;
  594                 ufs_itimes(vp);
  595                 if (vap->va_atime.tv_sec != VNOVAL) {
  596                         DIP(ip, i_atime) = vap->va_atime.tv_sec;
  597                         DIP(ip, i_atimensec) = vap->va_atime.tv_nsec;
  598                 }
  599                 if (vap->va_mtime.tv_sec != VNOVAL) {
  600                         DIP(ip, i_mtime) = vap->va_mtime.tv_sec;
  601                         DIP(ip, i_mtimensec) = vap->va_mtime.tv_nsec;
  602                 }
  603                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  604                     ip->i_ump->um_fstype == UFS2) {
  605                         ip->i_din2->di_birthtime = vap->va_birthtime.tv_sec;
  606                         ip->i_din2->di_birthnsec = vap->va_birthtime.tv_nsec;
  607                 }
  608                 error = UFS_UPDATE(vp, 0);
  609                 if (error)
  610                         return (error);
  611         }
  612         error = 0;
  613         if (vap->va_mode != (mode_t)VNOVAL) {
  614                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  615                         return (EROFS);
  616                 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode &
  617                    (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH)))
  618                         return (EPERM);
  619                 error = ufs_chmod(vp, (int)vap->va_mode, cred, td);
  620         }
  621         VN_KNOTE(vp, NOTE_ATTRIB);
  622         return (error);
  623 }
  624 
  625 /*
  626  * Change the mode on a file.
  627  * Inode must be locked before calling.
  628  */
  629 static int
  630 ufs_chmod(vp, mode, cred, td)
  631         struct vnode *vp;
  632         int mode;
  633         struct ucred *cred;
  634         struct thread *td;
  635 {
  636         struct inode *ip = VTOI(vp);
  637         int error;
  638 
  639         /*
  640          * To modify the permissions on a file, must possess VADMIN
  641          * for that file.
  642          */
  643         if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  644                 return (error);
  645         /*
  646          * Privileged processes may set the sticky bit on non-directories,
  647          * as well as set the setgid bit on a file with a group that the
  648          * process is not a member of.
  649          */
  650         if (suser_cred(cred, PRISON_ROOT)) {
  651                 if (vp->v_type != VDIR && (mode & S_ISTXT))
  652                         return (EFTYPE);
  653                 if (!groupmember(ip->i_gid, cred) && (mode & ISGID))
  654                         return (EPERM);
  655         }
  656         ip->i_mode &= ~ALLPERMS;
  657         ip->i_mode |= (mode & ALLPERMS);
  658         DIP(ip, i_mode) = ip->i_mode;
  659         ip->i_flag |= IN_CHANGE;
  660         return (0);
  661 }
  662 
  663 /*
  664  * Perform chown operation on inode ip;
  665  * inode must be locked prior to call.
  666  */
  667 static int
  668 ufs_chown(vp, uid, gid, cred, td)
  669         struct vnode *vp;
  670         uid_t uid;
  671         gid_t gid;
  672         struct ucred *cred;
  673         struct thread *td;
  674 {
  675         struct inode *ip = VTOI(vp);
  676         uid_t ouid;
  677         gid_t ogid;
  678         int error = 0;
  679 #ifdef QUOTA
  680         int i;
  681         ufs2_daddr_t change;
  682 #endif
  683 
  684         if (uid == (uid_t)VNOVAL)
  685                 uid = ip->i_uid;
  686         if (gid == (gid_t)VNOVAL)
  687                 gid = ip->i_gid;
  688         /*
  689          * To modify the ownership of a file, must possess VADMIN
  690          * for that file.
  691          */
  692         if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  693                 return (error);
  694         /*
  695          * To change the owner of a file, or change the group of a file
  696          * to a group of which we are not a member, the caller must
  697          * have privilege.
  698          */
  699         if ((uid != ip->i_uid || 
  700             (gid != ip->i_gid && !groupmember(gid, cred))) &&
  701             (error = suser_cred(cred, PRISON_ROOT)))
  702                 return (error);
  703         ogid = ip->i_gid;
  704         ouid = ip->i_uid;
  705 #ifdef QUOTA
  706         if ((error = getinoquota(ip)) != 0)
  707                 return (error);
  708         if (ouid == uid) {
  709                 dqrele(vp, ip->i_dquot[USRQUOTA]);
  710                 ip->i_dquot[USRQUOTA] = NODQUOT;
  711         }
  712         if (ogid == gid) {
  713                 dqrele(vp, ip->i_dquot[GRPQUOTA]);
  714                 ip->i_dquot[GRPQUOTA] = NODQUOT;
  715         }
  716         change = DIP(ip, i_blocks);
  717         (void) chkdq(ip, -change, cred, CHOWN);
  718         (void) chkiq(ip, -1, cred, CHOWN);
  719         for (i = 0; i < MAXQUOTAS; i++) {
  720                 dqrele(vp, ip->i_dquot[i]);
  721                 ip->i_dquot[i] = NODQUOT;
  722         }
  723 #endif
  724         ip->i_gid = gid;
  725         DIP(ip, i_gid) = gid;
  726         ip->i_uid = uid;
  727         DIP(ip, i_uid) = uid;
  728 #ifdef QUOTA
  729         if ((error = getinoquota(ip)) == 0) {
  730                 if (ouid == uid) {
  731                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  732                         ip->i_dquot[USRQUOTA] = NODQUOT;
  733                 }
  734                 if (ogid == gid) {
  735                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  736                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  737                 }
  738                 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
  739                         if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
  740                                 goto good;
  741                         else
  742                                 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
  743                 }
  744                 for (i = 0; i < MAXQUOTAS; i++) {
  745                         dqrele(vp, ip->i_dquot[i]);
  746                         ip->i_dquot[i] = NODQUOT;
  747                 }
  748         }
  749         ip->i_gid = ogid;
  750         DIP(ip, i_gid) = ogid;
  751         ip->i_uid = ouid;
  752         DIP(ip, i_uid) = ouid;
  753         if (getinoquota(ip) == 0) {
  754                 if (ouid == uid) {
  755                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  756                         ip->i_dquot[USRQUOTA] = NODQUOT;
  757                 }
  758                 if (ogid == gid) {
  759                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  760                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  761                 }
  762                 (void) chkdq(ip, change, cred, FORCE|CHOWN);
  763                 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
  764                 (void) getinoquota(ip);
  765         }
  766         return (error);
  767 good:
  768         if (getinoquota(ip))
  769                 panic("ufs_chown: lost quota");
  770 #endif /* QUOTA */
  771         ip->i_flag |= IN_CHANGE;
  772         if (suser_cred(cred, PRISON_ROOT) && (ouid != uid || ogid != gid)) {
  773                 ip->i_mode &= ~(ISUID | ISGID);
  774                 DIP(ip, i_mode) = ip->i_mode;
  775         }
  776         return (0);
  777 }
  778 
  779 static int
  780 ufs_remove(ap)
  781         struct vop_remove_args /* {
  782                 struct vnode *a_dvp;
  783                 struct vnode *a_vp;
  784                 struct componentname *a_cnp;
  785         } */ *ap;
  786 {
  787         struct inode *ip;
  788         struct vnode *vp = ap->a_vp;
  789         struct vnode *dvp = ap->a_dvp;
  790         int error;
  791 
  792         ip = VTOI(vp);
  793         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
  794             (VTOI(dvp)->i_flags & APPEND)) {
  795                 error = EPERM;
  796                 goto out;
  797         }
  798         error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
  799         if (ip->i_nlink <= 0)
  800                 vp->v_vflag |= VV_NOSYNC;
  801         VN_KNOTE(vp, NOTE_DELETE);
  802         VN_KNOTE(dvp, NOTE_WRITE);
  803 out:
  804         return (error);
  805 }
  806 
  807 /*
  808  * link vnode call
  809  */
  810 static int
  811 ufs_link(ap)
  812         struct vop_link_args /* {
  813                 struct vnode *a_tdvp;
  814                 struct vnode *a_vp;
  815                 struct componentname *a_cnp;
  816         } */ *ap;
  817 {
  818         struct vnode *vp = ap->a_vp;
  819         struct vnode *tdvp = ap->a_tdvp;
  820         struct componentname *cnp = ap->a_cnp;
  821         struct inode *ip;
  822         struct direct newdir;
  823         int error;
  824 
  825 #ifdef DIAGNOSTIC
  826         if ((cnp->cn_flags & HASBUF) == 0)
  827                 panic("ufs_link: no name");
  828 #endif
  829         if (tdvp->v_mount != vp->v_mount) {
  830                 error = EXDEV;
  831                 goto out;
  832         }
  833         ip = VTOI(vp);
  834         if ((nlink_t)ip->i_nlink >= LINK_MAX) {
  835                 error = EMLINK;
  836                 goto out;
  837         }
  838         if (ip->i_flags & (IMMUTABLE | APPEND)) {
  839                 error = EPERM;
  840                 goto out;
  841         }
  842         ip->i_effnlink++;
  843         ip->i_nlink++;
  844         DIP(ip, i_nlink) = ip->i_nlink;
  845         ip->i_flag |= IN_CHANGE;
  846         if (DOINGSOFTDEP(vp))
  847                 softdep_change_linkcnt(ip);
  848         error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
  849         if (!error) {
  850                 ufs_makedirentry(ip, cnp, &newdir);
  851                 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
  852         }
  853 
  854         if (error) {
  855                 ip->i_effnlink--;
  856                 ip->i_nlink--;
  857                 DIP(ip, i_nlink) = ip->i_nlink;
  858                 ip->i_flag |= IN_CHANGE;
  859                 if (DOINGSOFTDEP(vp))
  860                         softdep_change_linkcnt(ip);
  861         }
  862 out:
  863         VN_KNOTE(vp, NOTE_LINK);
  864         VN_KNOTE(tdvp, NOTE_WRITE);
  865         return (error);
  866 }
  867 
  868 /*
  869  * whiteout vnode call
  870  */
  871 static int
  872 ufs_whiteout(ap)
  873         struct vop_whiteout_args /* {
  874                 struct vnode *a_dvp;
  875                 struct componentname *a_cnp;
  876                 int a_flags;
  877         } */ *ap;
  878 {
  879         struct vnode *dvp = ap->a_dvp;
  880         struct componentname *cnp = ap->a_cnp;
  881         struct direct newdir;
  882         int error = 0;
  883 
  884         switch (ap->a_flags) {
  885         case LOOKUP:
  886                 /* 4.4 format directories support whiteout operations */
  887                 if (dvp->v_mount->mnt_maxsymlinklen > 0)
  888                         return (0);
  889                 return (EOPNOTSUPP);
  890 
  891         case CREATE:
  892                 /* create a new directory whiteout */
  893 #ifdef DIAGNOSTIC
  894                 if ((cnp->cn_flags & SAVENAME) == 0)
  895                         panic("ufs_whiteout: missing name");
  896                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
  897                         panic("ufs_whiteout: old format filesystem");
  898 #endif
  899 
  900                 newdir.d_ino = WINO;
  901                 newdir.d_namlen = cnp->cn_namelen;
  902                 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
  903                 newdir.d_type = DT_WHT;
  904                 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
  905                 break;
  906 
  907         case DELETE:
  908                 /* remove an existing directory whiteout */
  909 #ifdef DIAGNOSTIC
  910                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
  911                         panic("ufs_whiteout: old format filesystem");
  912 #endif
  913 
  914                 cnp->cn_flags &= ~DOWHITEOUT;
  915                 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
  916                 break;
  917         default:
  918                 panic("ufs_whiteout: unknown op");
  919         }
  920         return (error);
  921 }
  922 
  923 /*
  924  * Rename system call.
  925  *      rename("foo", "bar");
  926  * is essentially
  927  *      unlink("bar");
  928  *      link("foo", "bar");
  929  *      unlink("foo");
  930  * but ``atomically''.  Can't do full commit without saving state in the
  931  * inode on disk which isn't feasible at this time.  Best we can do is
  932  * always guarantee the target exists.
  933  *
  934  * Basic algorithm is:
  935  *
  936  * 1) Bump link count on source while we're linking it to the
  937  *    target.  This also ensure the inode won't be deleted out
  938  *    from underneath us while we work (it may be truncated by
  939  *    a concurrent `trunc' or `open' for creation).
  940  * 2) Link source to destination.  If destination already exists,
  941  *    delete it first.
  942  * 3) Unlink source reference to inode if still around. If a
  943  *    directory was moved and the parent of the destination
  944  *    is different from the source, patch the ".." entry in the
  945  *    directory.
  946  */
  947 static int
  948 ufs_rename(ap)
  949         struct vop_rename_args  /* {
  950                 struct vnode *a_fdvp;
  951                 struct vnode *a_fvp;
  952                 struct componentname *a_fcnp;
  953                 struct vnode *a_tdvp;
  954                 struct vnode *a_tvp;
  955                 struct componentname *a_tcnp;
  956         } */ *ap;
  957 {
  958         struct vnode *tvp = ap->a_tvp;
  959         struct vnode *tdvp = ap->a_tdvp;
  960         struct vnode *fvp = ap->a_fvp;
  961         struct vnode *fdvp = ap->a_fdvp;
  962         struct componentname *tcnp = ap->a_tcnp;
  963         struct componentname *fcnp = ap->a_fcnp;
  964         struct thread *td = fcnp->cn_thread;
  965         struct inode *ip, *xp, *dp;
  966         struct direct newdir;
  967         int doingdirectory = 0, oldparent = 0, newparent = 0;
  968         int error = 0, ioflag;
  969 
  970 #ifdef DIAGNOSTIC
  971         if ((tcnp->cn_flags & HASBUF) == 0 ||
  972             (fcnp->cn_flags & HASBUF) == 0)
  973                 panic("ufs_rename: no name");
  974 #endif
  975         /*
  976          * Check for cross-device rename.
  977          */
  978         if ((fvp->v_mount != tdvp->v_mount) ||
  979             (tvp && (fvp->v_mount != tvp->v_mount))) {
  980                 error = EXDEV;
  981 abortit:
  982                 if (tdvp == tvp)
  983                         vrele(tdvp);
  984                 else
  985                         vput(tdvp);
  986                 if (tvp)
  987                         vput(tvp);
  988                 vrele(fdvp);
  989                 vrele(fvp);
  990                 return (error);
  991         }
  992 
  993         if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
  994             (VTOI(tdvp)->i_flags & APPEND))) {
  995                 error = EPERM;
  996                 goto abortit;
  997         }
  998 
  999         /*
 1000          * Renaming a file to itself has no effect.  The upper layers should
 1001          * not call us in that case.  Temporarily just warn if they do.
 1002          */
 1003         if (fvp == tvp) {
 1004                 printf("ufs_rename: fvp == tvp (can't happen)\n");
 1005                 error = 0;
 1006                 goto abortit;
 1007         }
 1008 
 1009         if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
 1010                 goto abortit;
 1011         dp = VTOI(fdvp);
 1012         ip = VTOI(fvp);
 1013         if (ip->i_nlink >= LINK_MAX) {
 1014                 VOP_UNLOCK(fvp, 0, td);
 1015                 error = EMLINK;
 1016                 goto abortit;
 1017         }
 1018         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
 1019             || (dp->i_flags & APPEND)) {
 1020                 VOP_UNLOCK(fvp, 0, td);
 1021                 error = EPERM;
 1022                 goto abortit;
 1023         }
 1024         if ((ip->i_mode & IFMT) == IFDIR) {
 1025                 /*
 1026                  * Avoid ".", "..", and aliases of "." for obvious reasons.
 1027                  */
 1028                 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
 1029                     dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
 1030                     (ip->i_flag & IN_RENAME)) {
 1031                         VOP_UNLOCK(fvp, 0, td);
 1032                         error = EINVAL;
 1033                         goto abortit;
 1034                 }
 1035                 ip->i_flag |= IN_RENAME;
 1036                 oldparent = dp->i_number;
 1037                 doingdirectory = 1;
 1038         }
 1039         VN_KNOTE(fdvp, NOTE_WRITE);             /* XXX right place? */
 1040         vrele(fdvp);
 1041 
 1042         /*
 1043          * When the target exists, both the directory
 1044          * and target vnodes are returned locked.
 1045          */
 1046         dp = VTOI(tdvp);
 1047         xp = NULL;
 1048         if (tvp)
 1049                 xp = VTOI(tvp);
 1050 
 1051         /*
 1052          * 1) Bump link count while we're moving stuff
 1053          *    around.  If we crash somewhere before
 1054          *    completing our work, the link count
 1055          *    may be wrong, but correctable.
 1056          */
 1057         ip->i_effnlink++;
 1058         ip->i_nlink++;
 1059         DIP(ip, i_nlink) = ip->i_nlink;
 1060         ip->i_flag |= IN_CHANGE;
 1061         if (DOINGSOFTDEP(fvp))
 1062                 softdep_change_linkcnt(ip);
 1063         if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
 1064                                        DOINGASYNC(fvp)))) != 0) {
 1065                 VOP_UNLOCK(fvp, 0, td);
 1066                 goto bad;
 1067         }
 1068 
 1069         /*
 1070          * If ".." must be changed (ie the directory gets a new
 1071          * parent) then the source directory must not be in the
 1072          * directory heirarchy above the target, as this would
 1073          * orphan everything below the source directory. Also
 1074          * the user must have write permission in the source so
 1075          * as to be able to change "..". We must repeat the call
 1076          * to namei, as the parent directory is unlocked by the
 1077          * call to checkpath().
 1078          */
 1079         error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
 1080         VOP_UNLOCK(fvp, 0, td);
 1081         if (oldparent != dp->i_number)
 1082                 newparent = dp->i_number;
 1083         if (doingdirectory && newparent) {
 1084                 if (error)      /* write access check above */
 1085                         goto bad;
 1086                 if (xp != NULL)
 1087                         vput(tvp);
 1088                 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
 1089                 if (error)
 1090                         goto out;
 1091                 if ((tcnp->cn_flags & SAVESTART) == 0)
 1092                         panic("ufs_rename: lost to startdir");
 1093                 VREF(tdvp);
 1094                 error = relookup(tdvp, &tvp, tcnp);
 1095                 if (error)
 1096                         goto out;
 1097                 vrele(tdvp);
 1098                 dp = VTOI(tdvp);
 1099                 xp = NULL;
 1100                 if (tvp)
 1101                         xp = VTOI(tvp);
 1102         }
 1103         /*
 1104          * 2) If target doesn't exist, link the target
 1105          *    to the source and unlink the source.
 1106          *    Otherwise, rewrite the target directory
 1107          *    entry to reference the source inode and
 1108          *    expunge the original entry's existence.
 1109          */
 1110         if (xp == NULL) {
 1111                 if (dp->i_dev != ip->i_dev)
 1112                         panic("ufs_rename: EXDEV");
 1113                 /*
 1114                  * Account for ".." in new directory.
 1115                  * When source and destination have the same
 1116                  * parent we don't fool with the link count.
 1117                  */
 1118                 if (doingdirectory && newparent) {
 1119                         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 1120                                 error = EMLINK;
 1121                                 goto bad;
 1122                         }
 1123                         dp->i_effnlink++;
 1124                         dp->i_nlink++;
 1125                         DIP(dp, i_nlink) = dp->i_nlink;
 1126                         dp->i_flag |= IN_CHANGE;
 1127                         if (DOINGSOFTDEP(tdvp))
 1128                                 softdep_change_linkcnt(dp);
 1129                         error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
 1130                                                    DOINGASYNC(tdvp)));
 1131                         if (error)
 1132                                 goto bad;
 1133                 }
 1134                 ufs_makedirentry(ip, tcnp, &newdir);
 1135                 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
 1136                 if (error) {
 1137                         if (doingdirectory && newparent) {
 1138                                 dp->i_effnlink--;
 1139                                 dp->i_nlink--;
 1140                                 DIP(dp, i_nlink) = dp->i_nlink;
 1141                                 dp->i_flag |= IN_CHANGE;
 1142                                 if (DOINGSOFTDEP(tdvp))
 1143                                         softdep_change_linkcnt(dp);
 1144                                 (void)UFS_UPDATE(tdvp, 1);
 1145                         }
 1146                         goto bad;
 1147                 }
 1148                 VN_KNOTE(tdvp, NOTE_WRITE);
 1149                 vput(tdvp);
 1150         } else {
 1151                 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
 1152                         panic("ufs_rename: EXDEV");
 1153                 /*
 1154                  * Short circuit rename(foo, foo).
 1155                  */
 1156                 if (xp->i_number == ip->i_number)
 1157                         panic("ufs_rename: same file");
 1158                 /*
 1159                  * If the parent directory is "sticky", then the caller
 1160                  * must possess VADMIN for the parent directory, or the
 1161                  * destination of the rename.  This implements append-only
 1162                  * directories.
 1163                  */
 1164                 if ((dp->i_mode & S_ISTXT) &&
 1165                     VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, td) &&
 1166                     VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, td)) {
 1167                         error = EPERM;
 1168                         goto bad;
 1169                 }
 1170                 /*
 1171                  * Target must be empty if a directory and have no links
 1172                  * to it. Also, ensure source and target are compatible
 1173                  * (both directories, or both not directories).
 1174                  */
 1175                 if ((xp->i_mode&IFMT) == IFDIR) {
 1176                         if ((xp->i_effnlink > 2) ||
 1177                             !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
 1178                                 error = ENOTEMPTY;
 1179                                 goto bad;
 1180                         }
 1181                         if (!doingdirectory) {
 1182                                 error = ENOTDIR;
 1183                                 goto bad;
 1184                         }
 1185                         cache_purge(tdvp);
 1186                 } else if (doingdirectory) {
 1187                         error = EISDIR;
 1188                         goto bad;
 1189                 }
 1190                 error = ufs_dirrewrite(dp, xp, ip->i_number,
 1191                     IFTODT(ip->i_mode),
 1192                     (doingdirectory && newparent) ? newparent : doingdirectory);
 1193                 if (error)
 1194                         goto bad;
 1195                 if (doingdirectory) {
 1196                         if (!newparent) {
 1197                                 dp->i_effnlink--;
 1198                                 if (DOINGSOFTDEP(tdvp))
 1199                                         softdep_change_linkcnt(dp);
 1200                         }
 1201                         xp->i_effnlink--;
 1202                         if (DOINGSOFTDEP(tvp))
 1203                                 softdep_change_linkcnt(xp);
 1204                 }
 1205                 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
 1206                         /*
 1207                          * Truncate inode. The only stuff left in the directory
 1208                          * is "." and "..". The "." reference is inconsequential
 1209                          * since we are quashing it. We have removed the "."
 1210                          * reference and the reference in the parent directory,
 1211                          * but there may be other hard links. The soft
 1212                          * dependency code will arrange to do these operations
 1213                          * after the parent directory entry has been deleted on
 1214                          * disk, so when running with that code we avoid doing
 1215                          * them now.
 1216                          */
 1217                         if (!newparent) {
 1218                                 dp->i_nlink--;
 1219                                 DIP(dp, i_nlink) = dp->i_nlink;
 1220                                 dp->i_flag |= IN_CHANGE;
 1221                         }
 1222                         xp->i_nlink--;
 1223                         DIP(xp, i_nlink) = xp->i_nlink;
 1224                         xp->i_flag |= IN_CHANGE;
 1225                         ioflag = IO_NORMAL;
 1226                         if (DOINGASYNC(tvp))
 1227                                 ioflag |= IO_SYNC;
 1228                         if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
 1229                             tcnp->cn_cred, tcnp->cn_thread)) != 0)
 1230                                 goto bad;
 1231                 }
 1232                 VN_KNOTE(tdvp, NOTE_WRITE);
 1233                 vput(tdvp);
 1234                 VN_KNOTE(tvp, NOTE_DELETE);
 1235                 vput(tvp);
 1236                 xp = NULL;
 1237         }
 1238 
 1239         /*
 1240          * 3) Unlink the source.
 1241          */
 1242         fcnp->cn_flags &= ~MODMASK;
 1243         fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
 1244         if ((fcnp->cn_flags & SAVESTART) == 0)
 1245                 panic("ufs_rename: lost from startdir");
 1246         VREF(fdvp);
 1247         error = relookup(fdvp, &fvp, fcnp);
 1248         if (error == 0)
 1249                 vrele(fdvp);
 1250         if (fvp != NULL) {
 1251                 xp = VTOI(fvp);
 1252                 dp = VTOI(fdvp);
 1253         } else {
 1254                 /*
 1255                  * From name has disappeared.  IN_RENAME is not sufficient
 1256                  * to protect against directory races due to timing windows,
 1257                  * so we have to remove the panic.  XXX the only real way
 1258                  * to solve this issue is at a much higher level.  By the
 1259                  * time we hit ufs_rename() it's too late.
 1260                  */
 1261 #if 0
 1262                 if (doingdirectory)
 1263                         panic("ufs_rename: lost dir entry");
 1264 #endif
 1265                 vrele(ap->a_fvp);
 1266                 return (0);
 1267         }
 1268         /*
 1269          * Ensure that the directory entry still exists and has not
 1270          * changed while the new name has been entered. If the source is
 1271          * a file then the entry may have been unlinked or renamed. In
 1272          * either case there is no further work to be done. If the source
 1273          * is a directory then it cannot have been rmdir'ed; the IN_RENAME
 1274          * flag ensures that it cannot be moved by another rename or removed
 1275          * by a rmdir.
 1276          */
 1277         if (xp != ip) {
 1278                 /*
 1279                  * From name resolves to a different inode.  IN_RENAME is
 1280                  * not sufficient protection against timing window races
 1281                  * so we can't panic here.  XXX the only real way
 1282                  * to solve this issue is at a much higher level.  By the
 1283                  * time we hit ufs_rename() it's too late.
 1284                  */
 1285 #if 0
 1286                 if (doingdirectory)
 1287                         panic("ufs_rename: lost dir entry");
 1288 #endif
 1289         } else {
 1290                 /*
 1291                  * If the source is a directory with a
 1292                  * new parent, the link count of the old
 1293                  * parent directory must be decremented
 1294                  * and ".." set to point to the new parent.
 1295                  */
 1296                 if (doingdirectory && newparent) {
 1297                         xp->i_offset = mastertemplate.dot_reclen;
 1298                         ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
 1299                         cache_purge(fdvp);
 1300                 }
 1301                 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
 1302                 xp->i_flag &= ~IN_RENAME;
 1303         }
 1304         VN_KNOTE(fvp, NOTE_RENAME);
 1305         if (dp)
 1306                 vput(fdvp);
 1307         if (xp)
 1308                 vput(fvp);
 1309         vrele(ap->a_fvp);
 1310         return (error);
 1311 
 1312 bad:
 1313         if (xp)
 1314                 vput(ITOV(xp));
 1315         vput(ITOV(dp));
 1316 out:
 1317         if (doingdirectory)
 1318                 ip->i_flag &= ~IN_RENAME;
 1319         if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
 1320                 ip->i_effnlink--;
 1321                 ip->i_nlink--;
 1322                 DIP(ip, i_nlink) = ip->i_nlink;
 1323                 ip->i_flag |= IN_CHANGE;
 1324                 ip->i_flag &= ~IN_RENAME;
 1325                 if (DOINGSOFTDEP(fvp))
 1326                         softdep_change_linkcnt(ip);
 1327                 vput(fvp);
 1328         } else
 1329                 vrele(fvp);
 1330         return (error);
 1331 }
 1332 
 1333 /*
 1334  * Mkdir system call
 1335  */
 1336 static int
 1337 ufs_mkdir(ap)
 1338         struct vop_mkdir_args /* {
 1339                 struct vnode *a_dvp;
 1340                 struct vnode **a_vpp;
 1341                 struct componentname *a_cnp;
 1342                 struct vattr *a_vap;
 1343         } */ *ap;
 1344 {
 1345         struct vnode *dvp = ap->a_dvp;
 1346         struct vattr *vap = ap->a_vap;
 1347         struct componentname *cnp = ap->a_cnp;
 1348         struct inode *ip, *dp;
 1349         struct vnode *tvp;
 1350         struct buf *bp;
 1351         struct dirtemplate dirtemplate, *dtp;
 1352         struct direct newdir;
 1353 #ifdef UFS_ACL
 1354         struct acl *acl, *dacl;
 1355 #endif
 1356         int error, dmode;
 1357         long blkoff;
 1358 
 1359 #ifdef DIAGNOSTIC
 1360         if ((cnp->cn_flags & HASBUF) == 0)
 1361                 panic("ufs_mkdir: no name");
 1362 #endif
 1363         dp = VTOI(dvp);
 1364         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 1365                 error = EMLINK;
 1366                 goto out;
 1367         }
 1368         dmode = vap->va_mode & 0777;
 1369         dmode |= IFDIR;
 1370         /*
 1371          * Must simulate part of ufs_makeinode here to acquire the inode,
 1372          * but not have it entered in the parent directory. The entry is
 1373          * made later after writing "." and ".." entries.
 1374          */
 1375         error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
 1376         if (error)
 1377                 goto out;
 1378         ip = VTOI(tvp);
 1379         ip->i_gid = dp->i_gid;
 1380         DIP(ip, i_gid) = dp->i_gid;
 1381 #ifdef SUIDDIR
 1382         {
 1383 #ifdef QUOTA
 1384                 struct ucred ucred, *ucp;
 1385                 ucp = cnp->cn_cred;
 1386 #endif
 1387                 /*
 1388                  * If we are hacking owners here, (only do this where told to)
 1389                  * and we are not giving it TO root, (would subvert quotas)
 1390                  * then go ahead and give it to the other user.
 1391                  * The new directory also inherits the SUID bit.
 1392                  * If user's UID and dir UID are the same,
 1393                  * 'give it away' so that the SUID is still forced on.
 1394                  */
 1395                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 1396                     (dp->i_mode & ISUID) && dp->i_uid) {
 1397                         dmode |= ISUID;
 1398                         ip->i_uid = dp->i_uid;
 1399                         DIP(ip, i_uid) = dp->i_uid;
 1400 #ifdef QUOTA
 1401                         if (dp->i_uid != cnp->cn_cred->cr_uid) {
 1402                                 /*
 1403                                  * Make sure the correct user gets charged
 1404                                  * for the space.
 1405                                  * Make a dummy credential for the victim.
 1406                                  * XXX This seems to never be accessed out of
 1407                                  * our context so a stack variable is ok.
 1408                                  */
 1409                                 ucred.cr_ref = 1;
 1410                                 ucred.cr_uid = ip->i_uid;
 1411                                 ucred.cr_ngroups = 1;
 1412                                 ucred.cr_groups[0] = dp->i_gid;
 1413                                 ucp = &ucred;
 1414                         }
 1415 #endif
 1416                 } else {
 1417                         ip->i_uid = cnp->cn_cred->cr_uid;
 1418                         DIP(ip, i_uid) = ip->i_uid;
 1419                 }
 1420 #ifdef QUOTA
 1421                 if ((error = getinoquota(ip)) ||
 1422                     (error = chkiq(ip, 1, ucp, 0))) {
 1423                         UFS_VFREE(tvp, ip->i_number, dmode);
 1424                         vput(tvp);
 1425                         return (error);
 1426                 }
 1427 #endif
 1428         }
 1429 #else   /* !SUIDDIR */
 1430         ip->i_uid = cnp->cn_cred->cr_uid;
 1431         DIP(ip, i_uid) = ip->i_uid;
 1432 #ifdef QUOTA
 1433         if ((error = getinoquota(ip)) ||
 1434             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 1435                 UFS_VFREE(tvp, ip->i_number, dmode);
 1436                 vput(tvp);
 1437                 return (error);
 1438         }
 1439 #endif
 1440 #endif  /* !SUIDDIR */
 1441         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 1442 #ifdef UFS_ACL
 1443         acl = dacl = NULL;
 1444         if ((dvp->v_mount->mnt_flag & MNT_ACLS) != 0) {
 1445                 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK);
 1446                 MALLOC(dacl, struct acl *, sizeof(*dacl), M_ACL, M_WAITOK);
 1447 
 1448                 /*
 1449                  * Retrieve default ACL from parent, if any.
 1450                  */
 1451                 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cnp->cn_cred,
 1452                     cnp->cn_thread);
 1453                 switch (error) {
 1454                 case 0:
 1455                         /*
 1456                          * Retrieved a default ACL, so merge mode and ACL if
 1457                          * necessary.
 1458                          */
 1459                         if (acl->acl_cnt != 0) {
 1460                                 /*
 1461                                  * Two possible ways for default ACL to not
 1462                                  * be present.  First, the EA can be
 1463                                  * undefined, or second, the default ACL can
 1464                                  * be blank.  If it's blank, fall through to
 1465                                  * the it's not defined case.
 1466                                  */
 1467                                 ip->i_mode = dmode;
 1468                                 DIP(ip, i_mode) = dmode;
 1469                                 *dacl = *acl;
 1470                                 ufs_sync_acl_from_inode(ip, acl);
 1471                                 break;
 1472                         }
 1473                         /* FALLTHROUGH */
 1474         
 1475                 case EOPNOTSUPP:
 1476                         /*
 1477                          * Just use the mode as-is.
 1478                          */
 1479                         ip->i_mode = dmode;
 1480                         DIP(ip, i_mode) = dmode;
 1481                         FREE(acl, M_ACL);
 1482                         FREE(dacl, M_ACL);
 1483                         dacl = acl = NULL;
 1484                         break;
 1485                 
 1486                 default:
 1487                         UFS_VFREE(tvp, ip->i_number, dmode);
 1488                         vput(tvp);
 1489                         FREE(acl, M_ACL);
 1490                                 FREE(dacl, M_ACL);
 1491                         return (error);
 1492                 }
 1493         } else {
 1494 #endif /* !UFS_ACL */
 1495                 ip->i_mode = dmode;
 1496                 DIP(ip, i_mode) = dmode;
 1497 #ifdef UFS_ACL
 1498         }
 1499 #endif
 1500         tvp->v_type = VDIR;     /* Rest init'd in getnewvnode(). */
 1501         ip->i_effnlink = 2;
 1502         ip->i_nlink = 2;
 1503         DIP(ip, i_nlink) = 2;
 1504         if (DOINGSOFTDEP(tvp))
 1505                 softdep_change_linkcnt(ip);
 1506         if (cnp->cn_flags & ISWHITEOUT) {
 1507                 ip->i_flags |= UF_OPAQUE;
 1508                 DIP(ip, i_flags) = ip->i_flags;
 1509         }
 1510 
 1511         /*
 1512          * Bump link count in parent directory to reflect work done below.
 1513          * Should be done before reference is created so cleanup is
 1514          * possible if we crash.
 1515          */
 1516         dp->i_effnlink++;
 1517         dp->i_nlink++;
 1518         DIP(dp, i_nlink) = dp->i_nlink;
 1519         dp->i_flag |= IN_CHANGE;
 1520         if (DOINGSOFTDEP(dvp))
 1521                 softdep_change_linkcnt(dp);
 1522         error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
 1523         if (error)
 1524                 goto bad;
 1525 #ifdef MAC
 1526         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 1527                 error = mac_create_vnode_extattr(cnp->cn_cred, dvp->v_mount,
 1528                     dvp, tvp, cnp);
 1529                 if (error)
 1530                         goto bad;
 1531         }
 1532 #endif
 1533 #ifdef UFS_ACL
 1534         if (acl != NULL) {
 1535                 /*
 1536                  * XXX: If we abort now, will Soft Updates notify the extattr
 1537                  * code that the EAs for the file need to be released?
 1538                  */
 1539                 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cnp->cn_cred,
 1540                     cnp->cn_thread);
 1541                 if (error == 0)
 1542                         error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl,
 1543                             cnp->cn_cred, cnp->cn_thread);
 1544                 switch (error) {
 1545                 case 0:
 1546                         break;
 1547 
 1548                 case EOPNOTSUPP:
 1549                         /*
 1550                          * XXX: This should not happen, as EOPNOTSUPP above
 1551                          * was supposed to free acl.
 1552                          */
 1553                         printf("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
 1554                         /*
 1555                         panic("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()");
 1556                          */
 1557                         break;
 1558 
 1559                 default:
 1560                         FREE(acl, M_ACL);
 1561                         FREE(dacl, M_ACL);
 1562                         goto bad;
 1563                 }
 1564                 FREE(acl, M_ACL);
 1565                 FREE(dacl, M_ACL);
 1566         }
 1567 #endif /* !UFS_ACL */
 1568 
 1569         /*
 1570          * Initialize directory with "." and ".." from static template.
 1571          */
 1572         if (dvp->v_mount->mnt_maxsymlinklen > 0
 1573         )
 1574                 dtp = &mastertemplate;
 1575         else
 1576                 dtp = (struct dirtemplate *)&omastertemplate;
 1577         dirtemplate = *dtp;
 1578         dirtemplate.dot_ino = ip->i_number;
 1579         dirtemplate.dotdot_ino = dp->i_number;
 1580         if ((error = UFS_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
 1581             BA_CLRBUF, &bp)) != 0)
 1582                 goto bad;
 1583         ip->i_size = DIRBLKSIZ;
 1584         DIP(ip, i_size) = DIRBLKSIZ;
 1585         ip->i_flag |= IN_CHANGE | IN_UPDATE;
 1586         vnode_pager_setsize(tvp, (u_long)ip->i_size);
 1587         bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
 1588         if (DOINGSOFTDEP(tvp)) {
 1589                 /*
 1590                  * Ensure that the entire newly allocated block is a
 1591                  * valid directory so that future growth within the
 1592                  * block does not have to ensure that the block is
 1593                  * written before the inode.
 1594                  */
 1595                 blkoff = DIRBLKSIZ;
 1596                 while (blkoff < bp->b_bcount) {
 1597                         ((struct direct *)
 1598                            (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
 1599                         blkoff += DIRBLKSIZ;
 1600                 }
 1601         }
 1602         if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
 1603                                        DOINGASYNC(tvp)))) != 0) {
 1604                 (void)BUF_WRITE(bp);
 1605                 goto bad;
 1606         }
 1607         /*
 1608          * Directory set up, now install its entry in the parent directory.
 1609          *
 1610          * If we are not doing soft dependencies, then we must write out the
 1611          * buffer containing the new directory body before entering the new 
 1612          * name in the parent. If we are doing soft dependencies, then the
 1613          * buffer containing the new directory body will be passed to and
 1614          * released in the soft dependency code after the code has attached
 1615          * an appropriate ordering dependency to the buffer which ensures that
 1616          * the buffer is written before the new name is written in the parent.
 1617          */
 1618         if (DOINGASYNC(dvp))
 1619                 bdwrite(bp);
 1620         else if (!DOINGSOFTDEP(dvp) && ((error = BUF_WRITE(bp))))
 1621                 goto bad;
 1622         ufs_makedirentry(ip, cnp, &newdir);
 1623         error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
 1624         
 1625 bad:
 1626         if (error == 0) {
 1627                 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
 1628                 *ap->a_vpp = tvp;
 1629         } else {
 1630                 dp->i_effnlink--;
 1631                 dp->i_nlink--;
 1632                 DIP(dp, i_nlink) = dp->i_nlink;
 1633                 dp->i_flag |= IN_CHANGE;
 1634                 if (DOINGSOFTDEP(dvp))
 1635                         softdep_change_linkcnt(dp);
 1636                 /*
 1637                  * No need to do an explicit VOP_TRUNCATE here, vrele will
 1638                  * do this for us because we set the link count to 0.
 1639                  */
 1640                 ip->i_effnlink = 0;
 1641                 ip->i_nlink = 0;
 1642                 DIP(ip, i_nlink) = 0;
 1643                 ip->i_flag |= IN_CHANGE;
 1644                 if (DOINGSOFTDEP(tvp))
 1645                         softdep_change_linkcnt(ip);
 1646                 vput(tvp);
 1647         }
 1648 out:
 1649         return (error);
 1650 }
 1651 
 1652 /*
 1653  * Rmdir system call.
 1654  */
 1655 static int
 1656 ufs_rmdir(ap)
 1657         struct vop_rmdir_args /* {
 1658                 struct vnode *a_dvp;
 1659                 struct vnode *a_vp;
 1660                 struct componentname *a_cnp;
 1661         } */ *ap;
 1662 {
 1663         struct vnode *vp = ap->a_vp;
 1664         struct vnode *dvp = ap->a_dvp;
 1665         struct componentname *cnp = ap->a_cnp;
 1666         struct inode *ip, *dp;
 1667         int error, ioflag;
 1668 
 1669         ip = VTOI(vp);
 1670         dp = VTOI(dvp);
 1671 
 1672         /*
 1673          * Do not remove a directory that is in the process of being renamed.
 1674          * Verify the directory is empty (and valid). Rmdir ".." will not be
 1675          * valid since ".." will contain a reference to the current directory
 1676          * and thus be non-empty. Do not allow the removal of mounted on
 1677          * directories (this can happen when an NFS exported filesystem
 1678          * tries to remove a locally mounted on directory).
 1679          */
 1680         error = 0;
 1681         if (ip->i_flag & IN_RENAME) {
 1682                 error = EINVAL;
 1683                 goto out;
 1684         }
 1685         if (ip->i_effnlink != 2 ||
 1686             !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
 1687                 error = ENOTEMPTY;
 1688                 goto out;
 1689         }
 1690         if ((dp->i_flags & APPEND)
 1691             || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
 1692                 error = EPERM;
 1693                 goto out;
 1694         }
 1695         if (vp->v_mountedhere != 0) {
 1696                 error = EINVAL;
 1697                 goto out;
 1698         }
 1699         /*
 1700          * Delete reference to directory before purging
 1701          * inode.  If we crash in between, the directory
 1702          * will be reattached to lost+found,
 1703          */
 1704         dp->i_effnlink--;
 1705         ip->i_effnlink--;
 1706         if (DOINGSOFTDEP(vp)) {
 1707                 softdep_change_linkcnt(dp);
 1708                 softdep_change_linkcnt(ip);
 1709         }
 1710         error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
 1711         if (error) {
 1712                 dp->i_effnlink++;
 1713                 ip->i_effnlink++;
 1714                 if (DOINGSOFTDEP(vp)) {
 1715                         softdep_change_linkcnt(dp);
 1716                         softdep_change_linkcnt(ip);
 1717                 }
 1718                 goto out;
 1719         }
 1720         VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK);
 1721         cache_purge(dvp);
 1722         /*
 1723          * Truncate inode. The only stuff left in the directory is "." and
 1724          * "..". The "." reference is inconsequential since we are quashing
 1725          * it. The soft dependency code will arrange to do these operations
 1726          * after the parent directory entry has been deleted on disk, so
 1727          * when running with that code we avoid doing them now.
 1728          */
 1729         if (!DOINGSOFTDEP(vp)) {
 1730                 dp->i_nlink--;
 1731                 DIP(dp, i_nlink) = dp->i_nlink;
 1732                 dp->i_flag |= IN_CHANGE;
 1733                 ip->i_nlink--;
 1734                 DIP(ip, i_nlink) = ip->i_nlink;
 1735                 ip->i_flag |= IN_CHANGE;
 1736                 ioflag = IO_NORMAL;
 1737                 if (DOINGASYNC(vp))
 1738                         ioflag |= IO_SYNC;
 1739                 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
 1740                     cnp->cn_thread);
 1741         }
 1742         cache_purge(vp);
 1743 #ifdef UFS_DIRHASH
 1744         /* Kill any active hash; i_effnlink == 0, so it will not come back. */
 1745         if (ip->i_dirhash != NULL)
 1746                 ufsdirhash_free(ip);
 1747 #endif
 1748 out:
 1749         VN_KNOTE(vp, NOTE_DELETE);
 1750         return (error);
 1751 }
 1752 
 1753 /*
 1754  * symlink -- make a symbolic link
 1755  */
 1756 static int
 1757 ufs_symlink(ap)
 1758         struct vop_symlink_args /* {
 1759                 struct vnode *a_dvp;
 1760                 struct vnode **a_vpp;
 1761                 struct componentname *a_cnp;
 1762                 struct vattr *a_vap;
 1763                 char *a_target;
 1764         } */ *ap;
 1765 {
 1766         struct vnode *vp, **vpp = ap->a_vpp;
 1767         struct inode *ip;
 1768         int len, error;
 1769 
 1770         error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
 1771             vpp, ap->a_cnp);
 1772         if (error)
 1773                 return (error);
 1774         VN_KNOTE(ap->a_dvp, NOTE_WRITE);
 1775         vp = *vpp;
 1776         len = strlen(ap->a_target);
 1777         if (len < vp->v_mount->mnt_maxsymlinklen) {
 1778                 ip = VTOI(vp);
 1779                 bcopy(ap->a_target, SHORTLINK(ip), len);
 1780                 ip->i_size = len;
 1781                 DIP(ip, i_size) = len;
 1782                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 1783         } else
 1784                 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
 1785                     UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
 1786                     ap->a_cnp->cn_cred, NOCRED, (int *)0, (struct thread *)0);
 1787         if (error)
 1788                 vput(vp);
 1789         return (error);
 1790 }
 1791 
 1792 /*
 1793  * Vnode op for reading directories.
 1794  *
 1795  * The routine below assumes that the on-disk format of a directory
 1796  * is the same as that defined by <sys/dirent.h>. If the on-disk
 1797  * format changes, then it will be necessary to do a conversion
 1798  * from the on-disk format that read returns to the format defined
 1799  * by <sys/dirent.h>.
 1800  */
 1801 int
 1802 ufs_readdir(ap)
 1803         struct vop_readdir_args /* {
 1804                 struct vnode *a_vp;
 1805                 struct uio *a_uio;
 1806                 struct ucred *a_cred;
 1807                 int *a_eofflag;
 1808                 int *ncookies;
 1809                 u_long **a_cookies;
 1810         } */ *ap;
 1811 {
 1812         struct uio *uio = ap->a_uio;
 1813         int error;
 1814         size_t count, lost;
 1815         off_t off;
 1816 
 1817         if (ap->a_ncookies != NULL)
 1818                 /*
 1819                  * Ensure that the block is aligned.  The caller can use
 1820                  * the cookies to determine where in the block to start.
 1821                  */
 1822                 uio->uio_offset &= ~(DIRBLKSIZ - 1);
 1823         off = uio->uio_offset;
 1824         count = uio->uio_resid;
 1825         /* Make sure we don't return partial entries. */
 1826         if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1)))
 1827                 return (EINVAL);
 1828         count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
 1829         lost = uio->uio_resid - count;
 1830         uio->uio_resid = count;
 1831         uio->uio_iov->iov_len = count;
 1832 #       if (BYTE_ORDER == LITTLE_ENDIAN)
 1833                 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
 1834                         error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
 1835                 } else {
 1836                         struct dirent *dp, *edp;
 1837                         struct uio auio;
 1838                         struct iovec aiov;
 1839                         caddr_t dirbuf;
 1840                         int readcnt;
 1841                         u_char tmp;
 1842 
 1843                         auio = *uio;
 1844                         auio.uio_iov = &aiov;
 1845                         auio.uio_iovcnt = 1;
 1846                         auio.uio_segflg = UIO_SYSSPACE;
 1847                         aiov.iov_len = count;
 1848                         MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
 1849                         aiov.iov_base = dirbuf;
 1850                         error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
 1851                         if (error == 0) {
 1852                                 readcnt = count - auio.uio_resid;
 1853                                 edp = (struct dirent *)&dirbuf[readcnt];
 1854                                 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
 1855                                         tmp = dp->d_namlen;
 1856                                         dp->d_namlen = dp->d_type;
 1857                                         dp->d_type = tmp;
 1858                                         if (dp->d_reclen > 0) {
 1859                                                 dp = (struct dirent *)
 1860                                                     ((char *)dp + dp->d_reclen);
 1861                                         } else {
 1862                                                 error = EIO;
 1863                                                 break;
 1864                                         }
 1865                                 }
 1866                                 if (dp >= edp)
 1867                                         error = uiomove(dirbuf, readcnt, uio);
 1868                         }
 1869                         FREE(dirbuf, M_TEMP);
 1870                 }
 1871 #       else
 1872                 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
 1873 #       endif
 1874         if (!error && ap->a_ncookies != NULL) {
 1875                 struct dirent* dpStart;
 1876                 struct dirent* dpEnd;
 1877                 struct dirent* dp;
 1878                 int ncookies;
 1879                 u_long *cookies;
 1880                 u_long *cookiep;
 1881 
 1882                 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
 1883                         panic("ufs_readdir: unexpected uio from NFS server");
 1884                 dpStart = (struct dirent *)
 1885                     ((char *)uio->uio_iov->iov_base - (uio->uio_offset - off));
 1886                 dpEnd = (struct dirent *) uio->uio_iov->iov_base;
 1887                 for (dp = dpStart, ncookies = 0;
 1888                      dp < dpEnd;
 1889                      dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
 1890                         ncookies++;
 1891                 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
 1892                     M_WAITOK);
 1893                 for (dp = dpStart, cookiep = cookies;
 1894                      dp < dpEnd;
 1895                      dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
 1896                         off += dp->d_reclen;
 1897                         *cookiep++ = (u_long) off;
 1898                 }
 1899                 *ap->a_ncookies = ncookies;
 1900                 *ap->a_cookies = cookies;
 1901         }
 1902         uio->uio_resid += lost;
 1903         if (ap->a_eofflag)
 1904             *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
 1905         return (error);
 1906 }
 1907 
 1908 /*
 1909  * Return target name of a symbolic link
 1910  */
 1911 static int
 1912 ufs_readlink(ap)
 1913         struct vop_readlink_args /* {
 1914                 struct vnode *a_vp;
 1915                 struct uio *a_uio;
 1916                 struct ucred *a_cred;
 1917         } */ *ap;
 1918 {
 1919         struct vnode *vp = ap->a_vp;
 1920         struct inode *ip = VTOI(vp);
 1921         doff_t isize;
 1922 
 1923         isize = ip->i_size;
 1924         if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
 1925             DIP(ip, i_blocks) == 0) { /* XXX - for old fastlink support */
 1926                 uiomove(SHORTLINK(ip), isize, ap->a_uio);
 1927                 return (0);
 1928         }
 1929         return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
 1930 }
 1931 
 1932 /*
 1933  * Calculate the logical to physical mapping if not done already,
 1934  * then call the device strategy routine.
 1935  *
 1936  * In order to be able to swap to a file, the ufs_bmaparray() operation may not
 1937  * deadlock on memory.  See ufs_bmap() for details.
 1938  */
 1939 static int
 1940 ufs_strategy(ap)
 1941         struct vop_strategy_args /* {
 1942                 struct vnode *a_vp;
 1943                 struct buf *a_bp;
 1944         } */ *ap;
 1945 {
 1946         struct buf *bp = ap->a_bp;
 1947         struct vnode *vp = ap->a_vp;
 1948         struct inode *ip;
 1949         ufs2_daddr_t blkno;
 1950         int error;
 1951 
 1952         ip = VTOI(vp);
 1953         if (bp->b_blkno == bp->b_lblkno) {
 1954                 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, bp, NULL, NULL);
 1955                 bp->b_blkno = blkno;
 1956                 if (error) {
 1957                         bp->b_error = error;
 1958                         bp->b_ioflags |= BIO_ERROR;
 1959                         bufdone(bp);
 1960                         return (error);
 1961                 }
 1962                 if ((long)bp->b_blkno == -1)
 1963                         vfs_bio_clrbuf(bp);
 1964         }
 1965         if ((long)bp->b_blkno == -1) {
 1966                 bufdone(bp);
 1967                 return (0);
 1968         }
 1969         vp = ip->i_devvp;
 1970         bp->b_dev = vp->v_rdev;
 1971         VOP_SPECSTRATEGY(vp, bp);
 1972         return (0);
 1973 }
 1974 
 1975 /*
 1976  * Print out the contents of an inode.
 1977  */
 1978 static int
 1979 ufs_print(ap)
 1980         struct vop_print_args /* {
 1981                 struct vnode *a_vp;
 1982         } */ *ap;
 1983 {
 1984         struct vnode *vp = ap->a_vp;
 1985         struct inode *ip = VTOI(vp);
 1986 
 1987         printf("\tino %lu, on dev %s (%d, %d)", (u_long)ip->i_number,
 1988             devtoname(ip->i_dev), major(ip->i_dev), minor(ip->i_dev));
 1989         if (vp->v_type == VFIFO)
 1990                 fifo_printinfo(vp);
 1991         printf("\n");
 1992         return (0);
 1993 }
 1994 
 1995 /*
 1996  * Read wrapper for special devices.
 1997  */
 1998 static int
 1999 ufsspec_read(ap)
 2000         struct vop_read_args /* {
 2001                 struct vnode *a_vp;
 2002                 struct uio *a_uio;
 2003                 int  a_ioflag;
 2004                 struct ucred *a_cred;
 2005         } */ *ap;
 2006 {
 2007         int error, resid;
 2008         struct inode *ip;
 2009         struct uio *uio;
 2010 
 2011         uio = ap->a_uio;
 2012         resid = uio->uio_resid;
 2013         error = VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap);
 2014         /*
 2015          * The inode may have been revoked during the call, so it must not
 2016          * be accessed blindly here or in the other wrapper functions.
 2017          */
 2018         ip = VTOI(ap->a_vp);
 2019         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2020                 ip->i_flag |= IN_ACCESS;
 2021         return (error);
 2022 }
 2023 
 2024 /*
 2025  * Write wrapper for special devices.
 2026  */
 2027 static int
 2028 ufsspec_write(ap)
 2029         struct vop_write_args /* {
 2030                 struct vnode *a_vp;
 2031                 struct uio *a_uio;
 2032                 int  a_ioflag;
 2033                 struct ucred *a_cred;
 2034         } */ *ap;
 2035 {
 2036         int error, resid;
 2037         struct inode *ip;
 2038         struct uio *uio;
 2039 
 2040         uio = ap->a_uio;
 2041         resid = uio->uio_resid;
 2042         error = VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap);
 2043         ip = VTOI(ap->a_vp);
 2044         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2045                 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
 2046         return (error);
 2047 }
 2048 
 2049 /*
 2050  * Close wrapper for special devices.
 2051  *
 2052  * Update the times on the inode then do device close.
 2053  */
 2054 static int
 2055 ufsspec_close(ap)
 2056         struct vop_close_args /* {
 2057                 struct vnode *a_vp;
 2058                 int  a_fflag;
 2059                 struct ucred *a_cred;
 2060                 struct thread *a_td;
 2061         } */ *ap;
 2062 {
 2063         struct vnode *vp = ap->a_vp;
 2064 
 2065         VI_LOCK(vp);
 2066         if (vp->v_usecount > 1)
 2067                 ufs_itimes(vp);
 2068         VI_UNLOCK(vp);
 2069         return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap));
 2070 }
 2071 
 2072 /*
 2073  * Read wrapper for fifos.
 2074  */
 2075 static int
 2076 ufsfifo_read(ap)
 2077         struct vop_read_args /* {
 2078                 struct vnode *a_vp;
 2079                 struct uio *a_uio;
 2080                 int  a_ioflag;
 2081                 struct ucred *a_cred;
 2082         } */ *ap;
 2083 {
 2084         int error, resid;
 2085         struct inode *ip;
 2086         struct uio *uio;
 2087 
 2088         uio = ap->a_uio;
 2089         resid = uio->uio_resid;
 2090         error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap);
 2091         ip = VTOI(ap->a_vp);
 2092         if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
 2093             (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2094                 VTOI(ap->a_vp)->i_flag |= IN_ACCESS;
 2095         return (error);
 2096 }
 2097 
 2098 /*
 2099  * Write wrapper for fifos.
 2100  */
 2101 static int
 2102 ufsfifo_write(ap)
 2103         struct vop_write_args /* {
 2104                 struct vnode *a_vp;
 2105                 struct uio *a_uio;
 2106                 int  a_ioflag;
 2107                 struct ucred *a_cred;
 2108         } */ *ap;
 2109 {
 2110         int error, resid;
 2111         struct inode *ip;
 2112         struct uio *uio;
 2113 
 2114         uio = ap->a_uio;
 2115         resid = uio->uio_resid;
 2116         error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap);
 2117         ip = VTOI(ap->a_vp);
 2118         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2119                 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
 2120         return (error);
 2121 }
 2122 
 2123 /*
 2124  * Close wrapper for fifos.
 2125  *
 2126  * Update the times on the inode then do device close.
 2127  */
 2128 static int
 2129 ufsfifo_close(ap)
 2130         struct vop_close_args /* {
 2131                 struct vnode *a_vp;
 2132                 int  a_fflag;
 2133                 struct ucred *a_cred;
 2134                 struct thread *a_td;
 2135         } */ *ap;
 2136 {
 2137         struct vnode *vp = ap->a_vp;
 2138 
 2139         VI_LOCK(vp);
 2140         if (vp->v_usecount > 1)
 2141                 ufs_itimes(vp);
 2142         VI_UNLOCK(vp);
 2143         return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap));
 2144 }
 2145 
 2146 /*
 2147  * Kqfilter wrapper for fifos.
 2148  *
 2149  * Fall through to ufs kqfilter routines if needed 
 2150  */
 2151 static int
 2152 ufsfifo_kqfilter(ap)
 2153         struct vop_kqfilter_args *ap;
 2154 {
 2155         int error;
 2156 
 2157         error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_kqfilter), ap);
 2158         if (error)
 2159                 error = ufs_kqfilter(ap);
 2160         return (error);
 2161 }
 2162 
 2163 /*
 2164  * Return POSIX pathconf information applicable to ufs filesystems.
 2165  */
 2166 static int
 2167 ufs_pathconf(ap)
 2168         struct vop_pathconf_args /* {
 2169                 struct vnode *a_vp;
 2170                 int a_name;
 2171                 int *a_retval;
 2172         } */ *ap;
 2173 {
 2174         int error;
 2175 
 2176         error = 0;
 2177         switch (ap->a_name) {
 2178         case _PC_LINK_MAX:
 2179                 *ap->a_retval = LINK_MAX;
 2180                 break;
 2181         case _PC_NAME_MAX:
 2182                 *ap->a_retval = NAME_MAX;
 2183                 break;
 2184         case _PC_PATH_MAX:
 2185                 *ap->a_retval = PATH_MAX;
 2186                 break;
 2187         case _PC_PIPE_BUF:
 2188                 *ap->a_retval = PIPE_BUF;
 2189                 break;
 2190         case _PC_CHOWN_RESTRICTED:
 2191                 *ap->a_retval = 1;
 2192                 break;
 2193         case _PC_NO_TRUNC:
 2194                 *ap->a_retval = 1;
 2195                 break;
 2196         case _PC_ACL_EXTENDED:
 2197 #ifdef UFS_ACL
 2198                 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
 2199                         *ap->a_retval = 1;
 2200                 else
 2201                         *ap->a_retval = 0;
 2202 #else
 2203                 *ap->a_retval = 0;
 2204 #endif
 2205                 break;
 2206         case _PC_ACL_PATH_MAX:
 2207 #ifdef UFS_ACL
 2208                 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
 2209                         *ap->a_retval = ACL_MAX_ENTRIES;
 2210                 else
 2211                         *ap->a_retval = 3;
 2212 #else
 2213                 *ap->a_retval = 3;
 2214 #endif
 2215                 break;
 2216         case _PC_MAC_PRESENT:
 2217 #ifdef MAC
 2218                 if (ap->a_vp->v_mount->mnt_flag & MNT_MULTILABEL)
 2219                         *ap->a_retval = 1;
 2220                 else
 2221                         *ap->a_retval = 0;
 2222 #else
 2223                 *ap->a_retval = 0;
 2224 #endif
 2225                 break;
 2226         case _PC_ASYNC_IO:
 2227                 /* _PC_ASYNC_IO should have been handled by upper layers. */
 2228                 KASSERT(0, ("_PC_ASYNC_IO should not get here"));
 2229                 error = EINVAL;
 2230                 break;
 2231         case _PC_PRIO_IO:
 2232                 *ap->a_retval = 0;
 2233                 break;
 2234         case _PC_SYNC_IO:
 2235                 *ap->a_retval = 0;
 2236                 break;
 2237         case _PC_ALLOC_SIZE_MIN:
 2238                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
 2239                 break;
 2240         case _PC_FILESIZEBITS:
 2241                 *ap->a_retval = 64;
 2242                 break;
 2243         case _PC_REC_INCR_XFER_SIZE:
 2244                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2245                 break;
 2246         case _PC_REC_MAX_XFER_SIZE:
 2247                 *ap->a_retval = -1; /* means ``unlimited'' */
 2248                 break;
 2249         case _PC_REC_MIN_XFER_SIZE:
 2250                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2251                 break;
 2252         case _PC_REC_XFER_ALIGN:
 2253                 *ap->a_retval = PAGE_SIZE;
 2254                 break;
 2255         case _PC_SYMLINK_MAX:
 2256                 *ap->a_retval = MAXPATHLEN;
 2257                 break;
 2258 
 2259         default:
 2260                 error = EINVAL;
 2261                 break;
 2262         }
 2263         return (error);
 2264 }
 2265 
 2266 /*
 2267  * Advisory record locking support
 2268  */
 2269 static int
 2270 ufs_advlock(ap)
 2271         struct vop_advlock_args /* {
 2272                 struct vnode *a_vp;
 2273                 caddr_t  a_id;
 2274                 int  a_op;
 2275                 struct flock *a_fl;
 2276                 int  a_flags;
 2277         } */ *ap;
 2278 {
 2279         struct inode *ip = VTOI(ap->a_vp);
 2280 
 2281         return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
 2282 }
 2283 
 2284 /*
 2285  * Initialize the vnode associated with a new inode, handle aliased
 2286  * vnodes.
 2287  */
 2288 int
 2289 ufs_vinit(mntp, specops, fifoops, vpp)
 2290         struct mount *mntp;
 2291         vop_t **specops;
 2292         vop_t **fifoops;
 2293         struct vnode **vpp;
 2294 {
 2295         struct inode *ip;
 2296         struct vnode *vp;
 2297         struct timeval tv;
 2298 
 2299         vp = *vpp;
 2300         ip = VTOI(vp);
 2301         switch(vp->v_type = IFTOVT(ip->i_mode)) {
 2302         case VCHR:
 2303         case VBLK:
 2304                 vp->v_op = specops;
 2305                 vp = addaliasu(vp, DIP(ip, i_rdev));
 2306                 ip->i_vnode = vp;
 2307                 break;
 2308         case VFIFO:
 2309                 vp->v_op = fifoops;
 2310                 break;
 2311         default:
 2312                 break;
 2313 
 2314         }
 2315         ASSERT_VOP_LOCKED(vp, "ufs_vinit");
 2316         if (ip->i_number == ROOTINO)
 2317                 vp->v_vflag |= VV_ROOT;
 2318         /*
 2319          * Initialize modrev times
 2320          */
 2321         getmicrouptime(&tv);
 2322         SETHIGH(ip->i_modrev, tv.tv_sec);
 2323         SETLOW(ip->i_modrev, tv.tv_usec * 4294);
 2324         *vpp = vp;
 2325         return (0);
 2326 }
 2327 
 2328 /*
 2329  * Allocate a new inode.
 2330  * Vnode dvp must be locked.
 2331  */
 2332 static int
 2333 ufs_makeinode(mode, dvp, vpp, cnp)
 2334         int mode;
 2335         struct vnode *dvp;
 2336         struct vnode **vpp;
 2337         struct componentname *cnp;
 2338 {
 2339         struct inode *ip, *pdir;
 2340         struct direct newdir;
 2341         struct vnode *tvp;
 2342 #ifdef UFS_ACL
 2343         struct acl *acl;
 2344 #endif
 2345         int error;
 2346 
 2347         pdir = VTOI(dvp);
 2348 #ifdef DIAGNOSTIC
 2349         if ((cnp->cn_flags & HASBUF) == 0)
 2350                 panic("ufs_makeinode: no name");
 2351 #endif
 2352         *vpp = NULL;
 2353         if ((mode & IFMT) == 0)
 2354                 mode |= IFREG;
 2355 
 2356         error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
 2357         if (error)
 2358                 return (error);
 2359         ip = VTOI(tvp);
 2360         ip->i_gid = pdir->i_gid;
 2361         DIP(ip, i_gid) = pdir->i_gid;
 2362 #ifdef SUIDDIR
 2363         {
 2364 #ifdef QUOTA
 2365                 struct ucred ucred, *ucp;
 2366                 ucp = cnp->cn_cred;
 2367 #endif
 2368                 /*
 2369                  * If we are not the owner of the directory,
 2370                  * and we are hacking owners here, (only do this where told to)
 2371                  * and we are not giving it TO root, (would subvert quotas)
 2372                  * then go ahead and give it to the other user.
 2373                  * Note that this drops off the execute bits for security.
 2374                  */
 2375                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 2376                     (pdir->i_mode & ISUID) &&
 2377                     (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
 2378                         ip->i_uid = pdir->i_uid;
 2379                         DIP(ip, i_uid) = ip->i_uid;
 2380                         mode &= ~07111;
 2381 #ifdef QUOTA
 2382                         /*
 2383                          * Make sure the correct user gets charged
 2384                          * for the space.
 2385                          * Quickly knock up a dummy credential for the victim.
 2386                          * XXX This seems to never be accessed out of our
 2387                          * context so a stack variable is ok.
 2388                          */
 2389                         ucred.cr_ref = 1;
 2390                         ucred.cr_uid = ip->i_uid;
 2391                         ucred.cr_ngroups = 1;
 2392                         ucred.cr_groups[0] = pdir->i_gid;
 2393                         ucp = &ucred;
 2394 #endif
 2395                 } else {
 2396                         ip->i_uid = cnp->cn_cred->cr_uid;
 2397                         DIP(ip, i_uid) = ip->i_uid;
 2398                 }
 2399 
 2400 #ifdef QUOTA
 2401                 if ((error = getinoquota(ip)) ||
 2402                     (error = chkiq(ip, 1, ucp, 0))) {
 2403                         UFS_VFREE(tvp, ip->i_number, mode);
 2404                         vput(tvp);
 2405                         return (error);
 2406                 }
 2407 #endif
 2408         }
 2409 #else   /* !SUIDDIR */
 2410         ip->i_uid = cnp->cn_cred->cr_uid;
 2411         DIP(ip, i_uid) = ip->i_uid;
 2412 #ifdef QUOTA
 2413         if ((error = getinoquota(ip)) ||
 2414             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 2415                 UFS_VFREE(tvp, ip->i_number, mode);
 2416                 vput(tvp);
 2417                 return (error);
 2418         }
 2419 #endif
 2420 #endif  /* !SUIDDIR */
 2421         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 2422 #ifdef UFS_ACL
 2423         acl = NULL;
 2424         if ((dvp->v_mount->mnt_flag & MNT_ACLS) != 0) {
 2425                 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK);
 2426 
 2427                 /*
 2428                  * Retrieve default ACL for parent, if any.
 2429                  */
 2430                 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cnp->cn_cred,
 2431                     cnp->cn_thread);
 2432                 switch (error) {
 2433                 case 0:
 2434                         /*
 2435                          * Retrieved a default ACL, so merge mode and ACL if
 2436                          * necessary.
 2437                          */
 2438                         if (acl->acl_cnt != 0) {
 2439                                 /*
 2440                                  * Two possible ways for default ACL to not
 2441                                  * be present.  First, the EA can be
 2442                                  * undefined, or second, the default ACL can
 2443                                  * be blank.  If it's blank, fall through to
 2444                                  * the it's not defined case.
 2445                                  */
 2446                                 ip->i_mode = mode;
 2447                                 DIP(ip, i_mode) = mode;
 2448                                 ufs_sync_acl_from_inode(ip, acl);
 2449                                 break;
 2450                         }
 2451                         /* FALLTHROUGH */
 2452         
 2453                 case EOPNOTSUPP:
 2454                         /*
 2455                          * Just use the mode as-is.
 2456                          */
 2457                         ip->i_mode = mode;
 2458                         DIP(ip, i_mode) = mode;
 2459                         FREE(acl, M_ACL);
 2460                         acl = NULL;
 2461                         break;
 2462         
 2463                 default:
 2464                         UFS_VFREE(tvp, ip->i_number, mode);
 2465                         vput(tvp);
 2466                         FREE(acl, M_ACL);
 2467                         acl = NULL;
 2468                         return (error);
 2469                 }
 2470         } else {
 2471 #endif
 2472                 ip->i_mode = mode;
 2473                 DIP(ip, i_mode) = mode;
 2474 #ifdef UFS_ACL
 2475         }
 2476 #endif
 2477         tvp->v_type = IFTOVT(mode);     /* Rest init'd in getnewvnode(). */
 2478         ip->i_effnlink = 1;
 2479         ip->i_nlink = 1;
 2480         DIP(ip, i_nlink) = 1;
 2481         if (DOINGSOFTDEP(tvp))
 2482                 softdep_change_linkcnt(ip);
 2483         if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
 2484             suser_cred(cnp->cn_cred, PRISON_ROOT)) {
 2485                 ip->i_mode &= ~ISGID;
 2486                 DIP(ip, i_mode) = ip->i_mode;
 2487         }
 2488 
 2489         if (cnp->cn_flags & ISWHITEOUT) {
 2490                 ip->i_flags |= UF_OPAQUE;
 2491                 DIP(ip, i_flags) = ip->i_flags;
 2492         }
 2493 
 2494         /*
 2495          * Make sure inode goes to disk before directory entry.
 2496          */
 2497         error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
 2498         if (error)
 2499                 goto bad;
 2500 #ifdef MAC
 2501         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 2502                 error = mac_create_vnode_extattr(cnp->cn_cred, dvp->v_mount,
 2503                     dvp, tvp, cnp);
 2504                 if (error)
 2505                         goto bad;
 2506         }
 2507 #endif
 2508 #ifdef UFS_ACL
 2509         if (acl != NULL) {
 2510                 /*
 2511                  * XXX: If we abort now, will Soft Updates notify the extattr
 2512                  * code that the EAs for the file need to be released?
 2513                  */
 2514                 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cnp->cn_cred,
 2515                     cnp->cn_thread);
 2516                 switch (error) {
 2517                 case 0:
 2518                         break;
 2519 
 2520                 case EOPNOTSUPP:
 2521                         /*
 2522                          * XXX: This should not happen, as EOPNOTSUPP above was
 2523                          * supposed to free acl.
 2524                          */
 2525                         printf("ufs_makeinode: VOP_GETACL() but no "
 2526                             "VOP_SETACL()\n");
 2527                         /* panic("ufs_makeinode: VOP_GETACL() but no "
 2528                             "VOP_SETACL()"); */
 2529                         break;
 2530 
 2531                 default:
 2532                         FREE(acl, M_ACL);
 2533                         goto bad;
 2534                 }
 2535                 FREE(acl, M_ACL);
 2536         }
 2537 #endif /* !UFS_ACL */
 2538         ufs_makedirentry(ip, cnp, &newdir);
 2539         error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
 2540         if (error)
 2541                 goto bad;
 2542         *vpp = tvp;
 2543         return (0);
 2544 
 2545 bad:
 2546         /*
 2547          * Write error occurred trying to update the inode
 2548          * or the directory so must deallocate the inode.
 2549          */
 2550         ip->i_effnlink = 0;
 2551         ip->i_nlink = 0;
 2552         DIP(ip, i_nlink) = 0;
 2553         ip->i_flag |= IN_CHANGE;
 2554         if (DOINGSOFTDEP(tvp))
 2555                 softdep_change_linkcnt(ip);
 2556         vput(tvp);
 2557         return (error);
 2558 }
 2559 
 2560 static struct filterops ufsread_filtops = 
 2561         { 1, NULL, filt_ufsdetach, filt_ufsread };
 2562 static struct filterops ufswrite_filtops = 
 2563         { 1, NULL, filt_ufsdetach, filt_ufswrite };
 2564 static struct filterops ufsvnode_filtops = 
 2565         { 1, NULL, filt_ufsdetach, filt_ufsvnode };
 2566 
 2567 static int
 2568 ufs_kqfilter(ap)
 2569         struct vop_kqfilter_args /* {
 2570                 struct vnode *a_vp;
 2571                 struct knote *a_kn;
 2572         } */ *ap;
 2573 {
 2574         struct vnode *vp = ap->a_vp;
 2575         struct knote *kn = ap->a_kn;
 2576 
 2577         switch (kn->kn_filter) {
 2578         case EVFILT_READ:
 2579                 kn->kn_fop = &ufsread_filtops;
 2580                 break;
 2581         case EVFILT_WRITE:
 2582                 kn->kn_fop = &ufswrite_filtops;
 2583                 break;
 2584         case EVFILT_VNODE:
 2585                 kn->kn_fop = &ufsvnode_filtops;
 2586                 break;
 2587         default:
 2588                 return (1);
 2589         }
 2590 
 2591         kn->kn_hook = (caddr_t)vp;
 2592 
 2593         if (vp->v_pollinfo == NULL)
 2594                 v_addpollinfo(vp);
 2595         mtx_lock(&vp->v_pollinfo->vpi_lock);
 2596         SLIST_INSERT_HEAD(&vp->v_pollinfo->vpi_selinfo.si_note, kn, kn_selnext);
 2597         mtx_unlock(&vp->v_pollinfo->vpi_lock);
 2598 
 2599         return (0);
 2600 }
 2601 
 2602 static void
 2603 filt_ufsdetach(struct knote *kn)
 2604 {
 2605         struct vnode *vp = (struct vnode *)kn->kn_hook;
 2606 
 2607         KASSERT(vp->v_pollinfo != NULL, ("Mising v_pollinfo"));
 2608         mtx_lock(&vp->v_pollinfo->vpi_lock);
 2609         SLIST_REMOVE(&vp->v_pollinfo->vpi_selinfo.si_note,
 2610             kn, knote, kn_selnext);
 2611         mtx_unlock(&vp->v_pollinfo->vpi_lock);
 2612 }
 2613 
 2614 /*ARGSUSED*/
 2615 static int
 2616 filt_ufsread(struct knote *kn, long hint)
 2617 {
 2618         struct vnode *vp = (struct vnode *)kn->kn_hook;
 2619         struct inode *ip = VTOI(vp);
 2620 
 2621         /*
 2622          * filesystem is gone, so set the EOF flag and schedule 
 2623          * the knote for deletion.
 2624          */
 2625         if (hint == NOTE_REVOKE) {
 2626                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
 2627                 return (1);
 2628         }
 2629 
 2630         kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
 2631         return (kn->kn_data != 0);
 2632 }
 2633 
 2634 /*ARGSUSED*/
 2635 static int
 2636 filt_ufswrite(struct knote *kn, long hint)
 2637 {
 2638 
 2639         /*
 2640          * filesystem is gone, so set the EOF flag and schedule 
 2641          * the knote for deletion.
 2642          */
 2643         if (hint == NOTE_REVOKE)
 2644                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
 2645 
 2646         kn->kn_data = 0;
 2647         return (1);
 2648 }
 2649 
 2650 static int
 2651 filt_ufsvnode(struct knote *kn, long hint)
 2652 {
 2653 
 2654         if (kn->kn_sfflags & hint)
 2655                 kn->kn_fflags |= hint;
 2656         if (hint == NOTE_REVOKE) {
 2657                 kn->kn_flags |= EV_EOF;
 2658                 return (1);
 2659         }
 2660         return (kn->kn_fflags != 0);
 2661 }
 2662 
 2663 /* Global vfs data structures for ufs. */
 2664 static vop_t **ufs_vnodeop_p;
 2665 static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
 2666         { &vop_default_desc,            (vop_t *) vop_defaultop },
 2667         { &vop_fsync_desc,              (vop_t *) vop_panic },
 2668         { &vop_read_desc,               (vop_t *) vop_panic },
 2669         { &vop_reallocblks_desc,        (vop_t *) vop_panic },
 2670         { &vop_write_desc,              (vop_t *) vop_panic },
 2671         { &vop_access_desc,             (vop_t *) ufs_access },
 2672         { &vop_advlock_desc,            (vop_t *) ufs_advlock },
 2673         { &vop_bmap_desc,               (vop_t *) ufs_bmap },
 2674         { &vop_cachedlookup_desc,       (vop_t *) ufs_lookup },
 2675         { &vop_close_desc,              (vop_t *) ufs_close },
 2676         { &vop_create_desc,             (vop_t *) ufs_create },
 2677         { &vop_getattr_desc,            (vop_t *) ufs_getattr },
 2678         { &vop_inactive_desc,           (vop_t *) ufs_inactive },
 2679         { &vop_link_desc,               (vop_t *) ufs_link },
 2680         { &vop_lookup_desc,             (vop_t *) vfs_cache_lookup },
 2681         { &vop_mkdir_desc,              (vop_t *) ufs_mkdir },
 2682         { &vop_mknod_desc,              (vop_t *) ufs_mknod },
 2683         { &vop_open_desc,               (vop_t *) ufs_open },
 2684         { &vop_pathconf_desc,           (vop_t *) ufs_pathconf },
 2685         { &vop_poll_desc,               (vop_t *) vop_stdpoll },
 2686         { &vop_kqfilter_desc,           (vop_t *) ufs_kqfilter },
 2687         { &vop_getwritemount_desc,      (vop_t *) vop_stdgetwritemount },
 2688         { &vop_print_desc,              (vop_t *) ufs_print },
 2689         { &vop_readdir_desc,            (vop_t *) ufs_readdir },
 2690         { &vop_readlink_desc,           (vop_t *) ufs_readlink },
 2691         { &vop_reclaim_desc,            (vop_t *) ufs_reclaim },
 2692         { &vop_remove_desc,             (vop_t *) ufs_remove },
 2693         { &vop_rename_desc,             (vop_t *) ufs_rename },
 2694         { &vop_rmdir_desc,              (vop_t *) ufs_rmdir },
 2695         { &vop_setattr_desc,            (vop_t *) ufs_setattr },
 2696 #ifdef MAC
 2697         { &vop_setlabel_desc,           (vop_t *) vop_stdsetlabel_ea },
 2698 #endif
 2699         { &vop_strategy_desc,           (vop_t *) ufs_strategy },
 2700         { &vop_symlink_desc,            (vop_t *) ufs_symlink },
 2701         { &vop_whiteout_desc,           (vop_t *) ufs_whiteout },
 2702 #ifdef UFS_EXTATTR
 2703         { &vop_getextattr_desc,         (vop_t *) ufs_getextattr },
 2704         { &vop_setextattr_desc,         (vop_t *) ufs_setextattr },
 2705 #endif
 2706 #ifdef UFS_ACL
 2707         { &vop_getacl_desc,             (vop_t *) ufs_getacl },
 2708         { &vop_setacl_desc,             (vop_t *) ufs_setacl },
 2709         { &vop_aclcheck_desc,           (vop_t *) ufs_aclcheck },
 2710 #endif
 2711         { NULL, NULL }
 2712 };
 2713 static struct vnodeopv_desc ufs_vnodeop_opv_desc =
 2714         { &ufs_vnodeop_p, ufs_vnodeop_entries };
 2715 
 2716 static vop_t **ufs_specop_p;
 2717 static struct vnodeopv_entry_desc ufs_specop_entries[] = {
 2718         { &vop_default_desc,            (vop_t *) spec_vnoperate },
 2719         { &vop_fsync_desc,              (vop_t *) vop_panic },
 2720         { &vop_access_desc,             (vop_t *) ufs_access },
 2721         { &vop_close_desc,              (vop_t *) ufsspec_close },
 2722         { &vop_getattr_desc,            (vop_t *) ufs_getattr },
 2723         { &vop_inactive_desc,           (vop_t *) ufs_inactive },
 2724         { &vop_print_desc,              (vop_t *) ufs_print },
 2725         { &vop_read_desc,               (vop_t *) ufsspec_read },
 2726         { &vop_reclaim_desc,            (vop_t *) ufs_reclaim },
 2727         { &vop_setattr_desc,            (vop_t *) ufs_setattr },
 2728 #ifdef MAC
 2729         { &vop_setlabel_desc,           (vop_t *) vop_stdsetlabel_ea },
 2730 #endif
 2731         { &vop_write_desc,              (vop_t *) ufsspec_write },
 2732 #ifdef UFS_EXTATTR
 2733         { &vop_getextattr_desc,         (vop_t *) ufs_getextattr },
 2734         { &vop_setextattr_desc,         (vop_t *) ufs_setextattr },
 2735 #endif
 2736 #ifdef UFS_ACL
 2737         { &vop_getacl_desc,             (vop_t *) ufs_getacl },
 2738         { &vop_setacl_desc,             (vop_t *) ufs_setacl },
 2739         { &vop_aclcheck_desc,           (vop_t *) ufs_aclcheck },
 2740 #endif
 2741         {NULL, NULL}
 2742 };
 2743 static struct vnodeopv_desc ufs_specop_opv_desc =
 2744         { &ufs_specop_p, ufs_specop_entries };
 2745 
 2746 static vop_t **ufs_fifoop_p;
 2747 static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
 2748         { &vop_default_desc,            (vop_t *) fifo_vnoperate },
 2749         { &vop_fsync_desc,              (vop_t *) vop_panic },
 2750         { &vop_access_desc,             (vop_t *) ufs_access },
 2751         { &vop_close_desc,              (vop_t *) ufsfifo_close },
 2752         { &vop_getattr_desc,            (vop_t *) ufs_getattr },
 2753         { &vop_inactive_desc,           (vop_t *) ufs_inactive },
 2754         { &vop_kqfilter_desc,           (vop_t *) ufsfifo_kqfilter },
 2755         { &vop_print_desc,              (vop_t *) ufs_print },
 2756         { &vop_read_desc,               (vop_t *) ufsfifo_read },
 2757         { &vop_reclaim_desc,            (vop_t *) ufs_reclaim },
 2758         { &vop_setattr_desc,            (vop_t *) ufs_setattr },
 2759 #ifdef MAC
 2760         { &vop_setlabel_desc,           (vop_t *) vop_stdsetlabel_ea },
 2761 #endif
 2762         { &vop_write_desc,              (vop_t *) ufsfifo_write },
 2763 #ifdef UFS_EXTATTR
 2764         { &vop_getextattr_desc,         (vop_t *) ufs_getextattr },
 2765         { &vop_setextattr_desc,         (vop_t *) ufs_setextattr },
 2766 #endif
 2767 #ifdef UFS_ACL
 2768         { &vop_getacl_desc,             (vop_t *) ufs_getacl },
 2769         { &vop_setacl_desc,             (vop_t *) ufs_setacl },
 2770         { &vop_aclcheck_desc,           (vop_t *) ufs_aclcheck },
 2771 #endif
 2772         { NULL, NULL }
 2773 };
 2774 static struct vnodeopv_desc ufs_fifoop_opv_desc =
 2775         { &ufs_fifoop_p, ufs_fifoop_entries };
 2776 
 2777 VNODEOP_SET(ufs_vnodeop_opv_desc);
 2778 VNODEOP_SET(ufs_specop_opv_desc);
 2779 VNODEOP_SET(ufs_fifoop_opv_desc);
 2780 
 2781 int
 2782 ufs_vnoperate(ap)
 2783         struct vop_generic_args /* {
 2784                 struct vnodeop_desc *a_desc;
 2785         } */ *ap;
 2786 {
 2787         return (VOCALL(ufs_vnodeop_p, ap->a_desc->vdesc_offset, ap));
 2788 }
 2789 
 2790 int
 2791 ufs_vnoperatefifo(ap)
 2792         struct vop_generic_args /* {
 2793                 struct vnodeop_desc *a_desc;
 2794         } */ *ap;
 2795 {
 2796         return (VOCALL(ufs_fifoop_p, ap->a_desc->vdesc_offset, ap));
 2797 }
 2798 
 2799 int
 2800 ufs_vnoperatespec(ap)
 2801         struct vop_generic_args /* {
 2802                 struct vnodeop_desc *a_desc;
 2803         } */ *ap;
 2804 {
 2805         return (VOCALL(ufs_specop_p, ap->a_desc->vdesc_offset, ap));
 2806 }

Cache object: 7f6b7e1771db6bee46589a137748e612


[ 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.