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  * 4. Neither the name of the University nor the names of its contributors
   19  *    may be used to endorse or promote products derived from this software
   20  *    without specific prior written permission.
   21  *
   22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   32  * SUCH DAMAGE.
   33  *
   34  *      @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95
   35  */
   36 
   37 #include <sys/cdefs.h>
   38 __FBSDID("$FreeBSD: releng/5.3/sys/ufs/ufs/ufs_vnops.c 134621 2004-09-02 01:12:20Z csjp $");
   39 
   40 #include "opt_mac.h"
   41 #include "opt_quota.h"
   42 #include "opt_suiddir.h"
   43 #include "opt_ufs.h"
   44 
   45 #include <sys/param.h>
   46 #include <sys/systm.h>
   47 #include <sys/malloc.h>
   48 #include <sys/namei.h>
   49 #include <sys/kernel.h>
   50 #include <sys/fcntl.h>
   51 #include <sys/stat.h>
   52 #include <sys/bio.h>
   53 #include <sys/buf.h>
   54 #include <sys/mount.h>
   55 #include <sys/unistd.h>
   56 #include <sys/vnode.h>
   57 #include <sys/dirent.h>
   58 #include <sys/lockf.h>
   59 #include <sys/event.h>
   60 #include <sys/conf.h>
   61 #include <sys/acl.h>
   62 #include <sys/mac.h>
   63 
   64 #include <machine/mutex.h>
   65 
   66 #include <sys/file.h>           /* XXX */
   67 
   68 #include <vm/vm.h>
   69 #include <vm/vm_extern.h>
   70 
   71 #include <fs/fifofs/fifo.h>
   72 
   73 #include <ufs/ufs/acl.h>
   74 #include <ufs/ufs/extattr.h>
   75 #include <ufs/ufs/quota.h>
   76 #include <ufs/ufs/inode.h>
   77 #include <ufs/ufs/dir.h>
   78 #include <ufs/ufs/ufsmount.h>
   79 #include <ufs/ufs/ufs_extern.h>
   80 #ifdef UFS_DIRHASH
   81 #include <ufs/ufs/dirhash.h>
   82 #endif
   83 
   84 static int ufs_access(struct vop_access_args *);
   85 static int ufs_advlock(struct vop_advlock_args *);
   86 static int ufs_chmod(struct vnode *, int, struct ucred *, struct thread *);
   87 static int ufs_chown(struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
   88 static int ufs_close(struct vop_close_args *);
   89 static int ufs_create(struct vop_create_args *);
   90 static int ufs_getattr(struct vop_getattr_args *);
   91 static int ufs_link(struct vop_link_args *);
   92 static int ufs_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *);
   93 static int ufs_mkdir(struct vop_mkdir_args *);
   94 static int ufs_mknod(struct vop_mknod_args *);
   95 static int ufs_open(struct vop_open_args *);
   96 static int ufs_pathconf(struct vop_pathconf_args *);
   97 static int ufs_lock(struct vop_lock_args *);
   98 static int ufs_print(struct vop_print_args *);
   99 static int ufs_readlink(struct vop_readlink_args *);
  100 static int ufs_remove(struct vop_remove_args *);
  101 static int ufs_rename(struct vop_rename_args *);
  102 static int ufs_rmdir(struct vop_rmdir_args *);
  103 static int ufs_setattr(struct vop_setattr_args *);
  104 static int ufs_strategy(struct vop_strategy_args *);
  105 static int ufs_symlink(struct vop_symlink_args *);
  106 static int ufs_whiteout(struct vop_whiteout_args *);
  107 static int ufsfifo_close(struct vop_close_args *);
  108 static int ufsfifo_kqfilter(struct vop_kqfilter_args *);
  109 static int ufsfifo_read(struct vop_read_args *);
  110 static int ufsfifo_write(struct vop_write_args *);
  111 static int ufsspec_close(struct vop_close_args *);
  112 static int ufsspec_read(struct vop_read_args *);
  113 static int ufsspec_write(struct vop_write_args *);
  114 static int filt_ufsread(struct knote *kn, long hint);
  115 static int filt_ufswrite(struct knote *kn, long hint);
  116 static int filt_ufsvnode(struct knote *kn, long hint);
  117 static void filt_ufsdetach(struct knote *kn);
  118 static int ufs_kqfilter(struct vop_kqfilter_args *ap);
  119 
  120 union _qcvt {
  121         int64_t qcvt;
  122         int32_t val[2];
  123 };
  124 #define SETHIGH(q, h) { \
  125         union _qcvt tmp; \
  126         tmp.qcvt = (q); \
  127         tmp.val[_QUAD_HIGHWORD] = (h); \
  128         (q) = tmp.qcvt; \
  129 }
  130 #define SETLOW(q, l) { \
  131         union _qcvt tmp; \
  132         tmp.qcvt = (q); \
  133         tmp.val[_QUAD_LOWWORD] = (l); \
  134         (q) = tmp.qcvt; \
  135 }
  136 
  137 /*
  138  * A virgin directory (no blushing please).
  139  */
  140 static struct dirtemplate mastertemplate = {
  141         0, 12, DT_DIR, 1, ".",
  142         0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
  143 };
  144 static struct odirtemplate omastertemplate = {
  145         0, 12, 1, ".",
  146         0, DIRBLKSIZ - 12, 2, ".."
  147 };
  148 
  149 void
  150 ufs_itimes(vp)
  151         struct vnode *vp;
  152 {
  153         struct inode *ip;
  154         struct timespec ts;
  155 
  156         ip = VTOI(vp);
  157         if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
  158                 return;
  159         if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
  160                 ip->i_flag |= IN_LAZYMOD;
  161         else
  162                 ip->i_flag |= IN_MODIFIED;
  163         if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
  164                 vfs_timestamp(&ts);
  165                 if (ip->i_flag & IN_ACCESS) {
  166                         DIP_SET(ip, i_atime, ts.tv_sec);
  167                         DIP_SET(ip, i_atimensec, ts.tv_nsec);
  168                 }
  169                 if (ip->i_flag & IN_UPDATE) {
  170                         DIP_SET(ip, i_mtime, ts.tv_sec);
  171                         DIP_SET(ip, i_mtimensec, ts.tv_nsec);
  172                         ip->i_modrev++;
  173                 }
  174                 if (ip->i_flag & IN_CHANGE) {
  175                         DIP_SET(ip, i_ctime, ts.tv_sec);
  176                         DIP_SET(ip, i_ctimensec, ts.tv_nsec);
  177                 }
  178         }
  179         ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
  180 }
  181 
  182 /*
  183  * Create a regular file
  184  */
  185 static int
  186 ufs_create(ap)
  187         struct vop_create_args /* {
  188                 struct vnode *a_dvp;
  189                 struct vnode **a_vpp;
  190                 struct componentname *a_cnp;
  191                 struct vattr *a_vap;
  192         } */ *ap;
  193 {
  194         int error;
  195 
  196         error =
  197             ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
  198             ap->a_dvp, ap->a_vpp, ap->a_cnp);
  199         if (error)
  200                 return (error);
  201         VN_KNOTE_UNLOCKED(ap->a_dvp, NOTE_WRITE);
  202         return (0);
  203 }
  204 
  205 /*
  206  * Mknod vnode call
  207  */
  208 /* ARGSUSED */
  209 static int
  210 ufs_mknod(ap)
  211         struct vop_mknod_args /* {
  212                 struct vnode *a_dvp;
  213                 struct vnode **a_vpp;
  214                 struct componentname *a_cnp;
  215                 struct vattr *a_vap;
  216         } */ *ap;
  217 {
  218         struct vattr *vap = ap->a_vap;
  219         struct vnode **vpp = ap->a_vpp;
  220         struct inode *ip;
  221         ino_t ino;
  222         int error;
  223 
  224         error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
  225             ap->a_dvp, vpp, ap->a_cnp);
  226         if (error)
  227                 return (error);
  228         VN_KNOTE_UNLOCKED(ap->a_dvp, NOTE_WRITE);
  229         ip = VTOI(*vpp);
  230         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
  231         if (vap->va_rdev != VNOVAL) {
  232                 /*
  233                  * Want to be able to use this to make badblock
  234                  * inodes, so don't truncate the dev number.
  235                  */
  236                 DIP_SET(ip, i_rdev, vap->va_rdev);
  237         }
  238         /*
  239          * Remove inode, then reload it through VFS_VGET so it is
  240          * checked to see if it is an alias of an existing entry in
  241          * the inode cache.
  242          */
  243         vput(*vpp);
  244         (*vpp)->v_type = VNON;
  245         ino = ip->i_number;     /* Save this before vgone() invalidates ip. */
  246         vgone(*vpp);
  247         error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
  248         if (error) {
  249                 *vpp = NULL;
  250                 return (error);
  251         }
  252         return (0);
  253 }
  254 
  255 /*
  256  * Open called.
  257  *
  258  * Nothing to do.
  259  */
  260 /* ARGSUSED */
  261 static int
  262 ufs_open(ap)
  263         struct vop_open_args /* {
  264                 struct vnode *a_vp;
  265                 int  a_mode;
  266                 struct ucred *a_cred;
  267                 struct thread *a_td;
  268         } */ *ap;
  269 {
  270 
  271         /*
  272          * Files marked append-only must be opened for appending.
  273          */
  274         if ((VTOI(ap->a_vp)->i_flags & APPEND) &&
  275             (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
  276                 return (EPERM);
  277         return (0);
  278 }
  279 
  280 /*
  281  * Close called.
  282  *
  283  * Update the times on the inode.
  284  */
  285 /* ARGSUSED */
  286 static int
  287 ufs_close(ap)
  288         struct vop_close_args /* {
  289                 struct vnode *a_vp;
  290                 int  a_fflag;
  291                 struct ucred *a_cred;
  292                 struct thread *a_td;
  293         } */ *ap;
  294 {
  295         struct vnode *vp = ap->a_vp;
  296         struct mount *mp;
  297 
  298         VI_LOCK(vp);
  299         if (vp->v_usecount > 1) {
  300                 ufs_itimes(vp);
  301                 VI_UNLOCK(vp);
  302         } else {
  303                 VI_UNLOCK(vp);
  304                 /*
  305                  * If we are closing the last reference to an unlinked
  306                  * file, then it will be freed by the inactive routine.
  307                  * Because the freeing causes a the filesystem to be
  308                  * modified, it must be held up during periods when the
  309                  * filesystem is suspended.
  310                  *
  311                  * XXX - EAGAIN is returned to prevent vn_close from
  312                  * repeating the vrele operation.
  313                  */
  314                 if (vp->v_type == VREG && VTOI(vp)->i_effnlink == 0) {
  315                         (void) vn_start_write(vp, &mp, V_WAIT);
  316                         vrele(vp);
  317                         vn_finished_write(mp);
  318                         return (EAGAIN);
  319                 }
  320         }
  321         return (0);
  322 }
  323 
  324 static int
  325 ufs_access(ap)
  326         struct vop_access_args /* {
  327                 struct vnode *a_vp;
  328                 int  a_mode;
  329                 struct ucred *a_cred;
  330                 struct thread *a_td;
  331         } */ *ap;
  332 {
  333         struct vnode *vp = ap->a_vp;
  334         struct inode *ip = VTOI(vp);
  335         mode_t mode = ap->a_mode;
  336         int error;
  337 #ifdef UFS_ACL
  338         struct acl *acl;
  339 #endif
  340 
  341         /*
  342          * Disallow write attempts on read-only filesystems;
  343          * unless the file is a socket, fifo, or a block or
  344          * character device resident on the filesystem.
  345          */
  346         if (mode & VWRITE) {
  347                 switch (vp->v_type) {
  348                 case VDIR:
  349                 case VLNK:
  350                 case VREG:
  351                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  352                                 return (EROFS);
  353 #ifdef QUOTA
  354                         if ((error = getinoquota(ip)) != 0)
  355                                 return (error);
  356 #endif
  357                         break;
  358                 default:
  359                         break;
  360                 }
  361         }
  362 
  363         /* If immutable bit set, nobody gets to write it. */
  364         if ((mode & VWRITE) && (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT)))
  365                 return (EPERM);
  366 
  367 #ifdef UFS_ACL
  368         if ((vp->v_mount->mnt_flag & MNT_ACLS) != 0) {
  369                 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK);
  370                 error = VOP_GETACL(vp, ACL_TYPE_ACCESS, acl, ap->a_cred,
  371                     ap->a_td);
  372                 switch (error) {
  373                 case EOPNOTSUPP:
  374                         error = vaccess(vp->v_type, ip->i_mode, ip->i_uid,
  375                             ip->i_gid, ap->a_mode, ap->a_cred, NULL);
  376                         break;
  377                 case 0:
  378                         error = vaccess_acl_posix1e(vp->v_type, ip->i_uid,
  379                             ip->i_gid, acl, ap->a_mode, ap->a_cred, NULL);
  380                         break;
  381                 default:
  382                         printf(
  383 "ufs_access(): Error retrieving ACL on object (%d).\n",
  384                             error);
  385                         /*
  386                          * XXX: Fall back until debugged.  Should
  387                          * eventually possibly log an error, and return
  388                          * EPERM for safety.
  389                          */
  390                         error = vaccess(vp->v_type, ip->i_mode, ip->i_uid,
  391                             ip->i_gid, ap->a_mode, ap->a_cred, NULL);
  392                 }
  393                 FREE(acl, M_ACL);
  394         } else
  395 #endif /* !UFS_ACL */
  396                 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
  397                     ap->a_mode, ap->a_cred, NULL);
  398         return (error);
  399 }
  400 
  401 /* ARGSUSED */
  402 static int
  403 ufs_getattr(ap)
  404         struct vop_getattr_args /* {
  405                 struct vnode *a_vp;
  406                 struct vattr *a_vap;
  407                 struct ucred *a_cred;
  408                 struct thread *a_td;
  409         } */ *ap;
  410 {
  411         struct vnode *vp = ap->a_vp;
  412         struct inode *ip = VTOI(vp);
  413         struct vattr *vap = ap->a_vap;
  414 
  415         ufs_itimes(vp);
  416         /*
  417          * Copy from inode table
  418          */
  419         vap->va_fsid = dev2udev(ip->i_dev);
  420         vap->va_fileid = ip->i_number;
  421         vap->va_mode = ip->i_mode & ~IFMT;
  422         vap->va_nlink = ip->i_effnlink;
  423         vap->va_uid = ip->i_uid;
  424         vap->va_gid = ip->i_gid;
  425         if (ip->i_ump->um_fstype == UFS1) {
  426                 vap->va_rdev = ip->i_din1->di_rdev;
  427                 vap->va_size = ip->i_din1->di_size;
  428                 vap->va_atime.tv_sec = ip->i_din1->di_atime;
  429                 vap->va_atime.tv_nsec = ip->i_din1->di_atimensec;
  430                 vap->va_mtime.tv_sec = ip->i_din1->di_mtime;
  431                 vap->va_mtime.tv_nsec = ip->i_din1->di_mtimensec;
  432                 vap->va_ctime.tv_sec = ip->i_din1->di_ctime;
  433                 vap->va_ctime.tv_nsec = ip->i_din1->di_ctimensec;
  434                 vap->va_birthtime.tv_sec = 0;
  435                 vap->va_birthtime.tv_nsec = 0;
  436                 vap->va_bytes = dbtob((u_quad_t)ip->i_din1->di_blocks);
  437         } else {
  438                 vap->va_rdev = ip->i_din2->di_rdev;
  439                 vap->va_size = ip->i_din2->di_size;
  440                 vap->va_atime.tv_sec = ip->i_din2->di_atime;
  441                 vap->va_atime.tv_nsec = ip->i_din2->di_atimensec;
  442                 vap->va_mtime.tv_sec = ip->i_din2->di_mtime;
  443                 vap->va_mtime.tv_nsec = ip->i_din2->di_mtimensec;
  444                 vap->va_ctime.tv_sec = ip->i_din2->di_ctime;
  445                 vap->va_ctime.tv_nsec = ip->i_din2->di_ctimensec;
  446                 vap->va_birthtime.tv_sec = ip->i_din2->di_birthtime;
  447                 vap->va_birthtime.tv_nsec = ip->i_din2->di_birthnsec;
  448                 vap->va_bytes = dbtob((u_quad_t)ip->i_din2->di_blocks);
  449         }
  450         vap->va_flags = ip->i_flags;
  451         vap->va_gen = ip->i_gen;
  452         vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
  453         vap->va_type = IFTOVT(ip->i_mode);
  454         vap->va_filerev = ip->i_modrev;
  455         return (0);
  456 }
  457 
  458 /*
  459  * Set attribute vnode op. called from several syscalls
  460  */
  461 static int
  462 ufs_setattr(ap)
  463         struct vop_setattr_args /* {
  464                 struct vnode *a_vp;
  465                 struct vattr *a_vap;
  466                 struct ucred *a_cred;
  467                 struct thread *a_td;
  468         } */ *ap;
  469 {
  470         struct vattr *vap = ap->a_vap;
  471         struct vnode *vp = ap->a_vp;
  472         struct inode *ip = VTOI(vp);
  473         struct ucred *cred = ap->a_cred;
  474         struct thread *td = ap->a_td;
  475         int error;
  476 
  477         /*
  478          * Check for unsettable attributes.
  479          */
  480         if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
  481             (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
  482             (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
  483             ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
  484                 return (EINVAL);
  485         }
  486         if (vap->va_flags != VNOVAL) {
  487                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  488                         return (EROFS);
  489                 /*
  490                  * Callers may only modify the file flags on objects they
  491                  * have VADMIN rights for.
  492                  */
  493                 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  494                         return (error);
  495                 /*
  496                  * Unprivileged processes and privileged processes in
  497                  * jail() are not permitted to unset system flags, or
  498                  * modify flags if any system flags are set.
  499                  * Privileged non-jail processes may not modify system flags
  500                  * if securelevel > 0 and any existing system flags are set.
  501                  */
  502                 if (!suser_cred(cred, 0)) {
  503                         if (ip->i_flags
  504                             & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
  505                                 error = securelevel_gt(cred, 0);
  506                                 if (error)
  507                                         return (error);
  508                         }
  509                         /* Snapshot flag cannot be set or cleared */
  510                         if (((vap->va_flags & SF_SNAPSHOT) != 0 &&
  511                              (ip->i_flags & SF_SNAPSHOT) == 0) ||
  512                             ((vap->va_flags & SF_SNAPSHOT) == 0 &&
  513                              (ip->i_flags & SF_SNAPSHOT) != 0))
  514                                 return (EPERM);
  515                         ip->i_flags = vap->va_flags;
  516                         DIP_SET(ip, i_flags, vap->va_flags);
  517                 } else {
  518                         if (ip->i_flags
  519                             & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
  520                             (vap->va_flags & UF_SETTABLE) != vap->va_flags)
  521                                 return (EPERM);
  522                         ip->i_flags &= SF_SETTABLE;
  523                         ip->i_flags |= (vap->va_flags & UF_SETTABLE);
  524                         DIP_SET(ip, i_flags, ip->i_flags);
  525                 }
  526                 ip->i_flag |= IN_CHANGE;
  527                 if (vap->va_flags & (IMMUTABLE | APPEND))
  528                         return (0);
  529         }
  530         if (ip->i_flags & (IMMUTABLE | APPEND))
  531                 return (EPERM);
  532         /*
  533          * Go through the fields and update iff not VNOVAL.
  534          */
  535         if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
  536                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  537                         return (EROFS);
  538                 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred,
  539                     td)) != 0)
  540                         return (error);
  541         }
  542         if (vap->va_size != VNOVAL) {
  543                 /*
  544                  * Disallow write attempts on read-only filesystems;
  545                  * unless the file is a socket, fifo, or a block or
  546                  * character device resident on the filesystem.
  547                  */
  548                 switch (vp->v_type) {
  549                 case VDIR:
  550                         return (EISDIR);
  551                 case VLNK:
  552                 case VREG:
  553                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  554                                 return (EROFS);
  555                         if ((ip->i_flags & SF_SNAPSHOT) != 0)
  556                                 return (EPERM);
  557                         break;
  558                 default:
  559                         break;
  560                 }
  561                 if ((error = UFS_TRUNCATE(vp, vap->va_size, IO_NORMAL,
  562                     cred, td)) != 0)
  563                         return (error);
  564         }
  565         if (vap->va_atime.tv_sec != VNOVAL ||
  566             vap->va_mtime.tv_sec != VNOVAL ||
  567             vap->va_birthtime.tv_sec != VNOVAL) {
  568                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  569                         return (EROFS);
  570                 if ((ip->i_flags & SF_SNAPSHOT) != 0)
  571                         return (EPERM);
  572                 /*
  573                  * From utimes(2):
  574                  * If times is NULL, ... The caller must be the owner of
  575                  * the file, have permission to write the file, or be the
  576                  * super-user.
  577                  * If times is non-NULL, ... The caller must be the owner of
  578                  * the file or be the super-user.
  579                  */
  580                 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) &&
  581                     ((vap->va_vaflags & VA_UTIMES_NULL) == 0 ||
  582                     (error = VOP_ACCESS(vp, VWRITE, cred, td))))
  583                         return (error);
  584                 if (vap->va_atime.tv_sec != VNOVAL)
  585                         ip->i_flag |= IN_ACCESS;
  586                 if (vap->va_mtime.tv_sec != VNOVAL)
  587                         ip->i_flag |= IN_CHANGE | IN_UPDATE;
  588                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  589                     ip->i_ump->um_fstype == UFS2)
  590                         ip->i_flag |= IN_MODIFIED;
  591                 ufs_itimes(vp);
  592                 if (vap->va_atime.tv_sec != VNOVAL) {
  593                         DIP_SET(ip, i_atime, vap->va_atime.tv_sec);
  594                         DIP_SET(ip, i_atimensec, vap->va_atime.tv_nsec);
  595                 }
  596                 if (vap->va_mtime.tv_sec != VNOVAL) {
  597                         DIP_SET(ip, i_mtime, vap->va_mtime.tv_sec);
  598                         DIP_SET(ip, i_mtimensec, vap->va_mtime.tv_nsec);
  599                 }
  600                 if (vap->va_birthtime.tv_sec != VNOVAL &&
  601                     ip->i_ump->um_fstype == UFS2) {
  602                         ip->i_din2->di_birthtime = vap->va_birthtime.tv_sec;
  603                         ip->i_din2->di_birthnsec = vap->va_birthtime.tv_nsec;
  604                 }
  605                 error = UFS_UPDATE(vp, 0);
  606                 if (error)
  607                         return (error);
  608         }
  609         error = 0;
  610         if (vap->va_mode != (mode_t)VNOVAL) {
  611                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  612                         return (EROFS);
  613                 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode &
  614                    (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH)))
  615                         return (EPERM);
  616                 error = ufs_chmod(vp, (int)vap->va_mode, cred, td);
  617         }
  618         VN_KNOTE_UNLOCKED(vp, NOTE_ATTRIB);
  619         return (error);
  620 }
  621 
  622 /*
  623  * Change the mode on a file.
  624  * Inode must be locked before calling.
  625  */
  626 static int
  627 ufs_chmod(vp, mode, cred, td)
  628         struct vnode *vp;
  629         int mode;
  630         struct ucred *cred;
  631         struct thread *td;
  632 {
  633         struct inode *ip = VTOI(vp);
  634         int error;
  635 
  636         /*
  637          * To modify the permissions on a file, must possess VADMIN
  638          * for that file.
  639          */
  640         if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  641                 return (error);
  642         /*
  643          * Privileged processes may set the sticky bit on non-directories,
  644          * as well as set the setgid bit on a file with a group that the
  645          * process is not a member of.  Both of these are allowed in
  646          * jail(8).
  647          */
  648         if (vp->v_type != VDIR && (mode & S_ISTXT)) {
  649                 if (suser_cred(cred, SUSER_ALLOWJAIL))
  650                         return (EFTYPE);
  651         }
  652         if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
  653                 error = suser_cred(cred, SUSER_ALLOWJAIL);
  654                 if (error)
  655                         return (error);
  656         }
  657         ip->i_mode &= ~ALLPERMS;
  658         ip->i_mode |= (mode & ALLPERMS);
  659         DIP_SET(ip, i_mode, ip->i_mode);
  660         ip->i_flag |= IN_CHANGE;
  661         return (0);
  662 }
  663 
  664 /*
  665  * Perform chown operation on inode ip;
  666  * inode must be locked prior to call.
  667  */
  668 static int
  669 ufs_chown(vp, uid, gid, cred, td)
  670         struct vnode *vp;
  671         uid_t uid;
  672         gid_t gid;
  673         struct ucred *cred;
  674         struct thread *td;
  675 {
  676         struct inode *ip = VTOI(vp);
  677         uid_t ouid;
  678         gid_t ogid;
  679         int error = 0;
  680 #ifdef QUOTA
  681         int i;
  682         ufs2_daddr_t change;
  683 #endif
  684 
  685         if (uid == (uid_t)VNOVAL)
  686                 uid = ip->i_uid;
  687         if (gid == (gid_t)VNOVAL)
  688                 gid = ip->i_gid;
  689         /*
  690          * To modify the ownership of a file, must possess VADMIN
  691          * for that file.
  692          */
  693         if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  694                 return (error);
  695         /*
  696          * To change the owner of a file, or change the group of a file
  697          * to a group of which we are not a member, the caller must
  698          * have privilege.
  699          */
  700         if ((uid != ip->i_uid || 
  701             (gid != ip->i_gid && !groupmember(gid, cred))) &&
  702             (error = suser_cred(cred, SUSER_ALLOWJAIL)))
  703                 return (error);
  704         ogid = ip->i_gid;
  705         ouid = ip->i_uid;
  706 #ifdef QUOTA
  707         if ((error = getinoquota(ip)) != 0)
  708                 return (error);
  709         if (ouid == uid) {
  710                 dqrele(vp, ip->i_dquot[USRQUOTA]);
  711                 ip->i_dquot[USRQUOTA] = NODQUOT;
  712         }
  713         if (ogid == gid) {
  714                 dqrele(vp, ip->i_dquot[GRPQUOTA]);
  715                 ip->i_dquot[GRPQUOTA] = NODQUOT;
  716         }
  717         change = DIP(ip, i_blocks);
  718         (void) chkdq(ip, -change, cred, CHOWN);
  719         (void) chkiq(ip, -1, cred, CHOWN);
  720         for (i = 0; i < MAXQUOTAS; i++) {
  721                 dqrele(vp, ip->i_dquot[i]);
  722                 ip->i_dquot[i] = NODQUOT;
  723         }
  724 #endif
  725         ip->i_gid = gid;
  726         DIP_SET(ip, i_gid, gid);
  727         ip->i_uid = uid;
  728         DIP_SET(ip, i_uid, uid);
  729 #ifdef QUOTA
  730         if ((error = getinoquota(ip)) == 0) {
  731                 if (ouid == uid) {
  732                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  733                         ip->i_dquot[USRQUOTA] = NODQUOT;
  734                 }
  735                 if (ogid == gid) {
  736                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  737                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  738                 }
  739                 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
  740                         if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
  741                                 goto good;
  742                         else
  743                                 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
  744                 }
  745                 for (i = 0; i < MAXQUOTAS; i++) {
  746                         dqrele(vp, ip->i_dquot[i]);
  747                         ip->i_dquot[i] = NODQUOT;
  748                 }
  749         }
  750         ip->i_gid = ogid;
  751         DIP_SET(ip, i_gid, ogid);
  752         ip->i_uid = ouid;
  753         DIP_SET(ip, i_uid, ouid);
  754         if (getinoquota(ip) == 0) {
  755                 if (ouid == uid) {
  756                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  757                         ip->i_dquot[USRQUOTA] = NODQUOT;
  758                 }
  759                 if (ogid == gid) {
  760                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  761                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  762                 }
  763                 (void) chkdq(ip, change, cred, FORCE|CHOWN);
  764                 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
  765                 (void) getinoquota(ip);
  766         }
  767         return (error);
  768 good:
  769         if (getinoquota(ip))
  770                 panic("ufs_chown: lost quota");
  771 #endif /* QUOTA */
  772         ip->i_flag |= IN_CHANGE;
  773         if (suser_cred(cred, SUSER_ALLOWJAIL) && (ouid != uid || ogid != gid)) {
  774                 ip->i_mode &= ~(ISUID | ISGID);
  775                 DIP_SET(ip, i_mode, ip->i_mode);
  776         }
  777         return (0);
  778 }
  779 
  780 static int
  781 ufs_remove(ap)
  782         struct vop_remove_args /* {
  783                 struct vnode *a_dvp;
  784                 struct vnode *a_vp;
  785                 struct componentname *a_cnp;
  786         } */ *ap;
  787 {
  788         struct inode *ip;
  789         struct vnode *vp = ap->a_vp;
  790         struct vnode *dvp = ap->a_dvp;
  791         int error;
  792 
  793         ip = VTOI(vp);
  794         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
  795             (VTOI(dvp)->i_flags & APPEND)) {
  796                 error = EPERM;
  797                 goto out;
  798         }
  799         error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
  800         if (ip->i_nlink <= 0)
  801                 vp->v_vflag |= VV_NOSYNC;
  802         VN_KNOTE_UNLOCKED(vp, NOTE_DELETE);
  803         VN_KNOTE_UNLOCKED(dvp, NOTE_WRITE);
  804 out:
  805         return (error);
  806 }
  807 
  808 /*
  809  * link vnode call
  810  */
  811 static int
  812 ufs_link(ap)
  813         struct vop_link_args /* {
  814                 struct vnode *a_tdvp;
  815                 struct vnode *a_vp;
  816                 struct componentname *a_cnp;
  817         } */ *ap;
  818 {
  819         struct vnode *vp = ap->a_vp;
  820         struct vnode *tdvp = ap->a_tdvp;
  821         struct componentname *cnp = ap->a_cnp;
  822         struct inode *ip;
  823         struct direct newdir;
  824         int error;
  825 
  826 #ifdef DIAGNOSTIC
  827         if ((cnp->cn_flags & HASBUF) == 0)
  828                 panic("ufs_link: no name");
  829 #endif
  830         if (tdvp->v_mount != vp->v_mount) {
  831                 error = EXDEV;
  832                 goto out;
  833         }
  834         ip = VTOI(vp);
  835         if ((nlink_t)ip->i_nlink >= LINK_MAX) {
  836                 error = EMLINK;
  837                 goto out;
  838         }
  839         if (ip->i_flags & (IMMUTABLE | APPEND)) {
  840                 error = EPERM;
  841                 goto out;
  842         }
  843         ip->i_effnlink++;
  844         ip->i_nlink++;
  845         DIP_SET(ip, i_nlink, ip->i_nlink);
  846         ip->i_flag |= IN_CHANGE;
  847         if (DOINGSOFTDEP(vp))
  848                 softdep_change_linkcnt(ip);
  849         error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
  850         if (!error) {
  851                 ufs_makedirentry(ip, cnp, &newdir);
  852                 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL);
  853         }
  854 
  855         if (error) {
  856                 ip->i_effnlink--;
  857                 ip->i_nlink--;
  858                 DIP_SET(ip, i_nlink, ip->i_nlink);
  859                 ip->i_flag |= IN_CHANGE;
  860                 if (DOINGSOFTDEP(vp))
  861                         softdep_change_linkcnt(ip);
  862         }
  863 out:
  864         VN_KNOTE_UNLOCKED(vp, NOTE_LINK);
  865         VN_KNOTE_UNLOCKED(tdvp, NOTE_WRITE);
  866         return (error);
  867 }
  868 
  869 /*
  870  * whiteout vnode call
  871  */
  872 static int
  873 ufs_whiteout(ap)
  874         struct vop_whiteout_args /* {
  875                 struct vnode *a_dvp;
  876                 struct componentname *a_cnp;
  877                 int a_flags;
  878         } */ *ap;
  879 {
  880         struct vnode *dvp = ap->a_dvp;
  881         struct componentname *cnp = ap->a_cnp;
  882         struct direct newdir;
  883         int error = 0;
  884 
  885         switch (ap->a_flags) {
  886         case LOOKUP:
  887                 /* 4.4 format directories support whiteout operations */
  888                 if (dvp->v_mount->mnt_maxsymlinklen > 0)
  889                         return (0);
  890                 return (EOPNOTSUPP);
  891 
  892         case CREATE:
  893                 /* create a new directory whiteout */
  894 #ifdef DIAGNOSTIC
  895                 if ((cnp->cn_flags & SAVENAME) == 0)
  896                         panic("ufs_whiteout: missing name");
  897                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
  898                         panic("ufs_whiteout: old format filesystem");
  899 #endif
  900 
  901                 newdir.d_ino = WINO;
  902                 newdir.d_namlen = cnp->cn_namelen;
  903                 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
  904                 newdir.d_type = DT_WHT;
  905                 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL);
  906                 break;
  907 
  908         case DELETE:
  909                 /* remove an existing directory whiteout */
  910 #ifdef DIAGNOSTIC
  911                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
  912                         panic("ufs_whiteout: old format filesystem");
  913 #endif
  914 
  915                 cnp->cn_flags &= ~DOWHITEOUT;
  916                 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
  917                 break;
  918         default:
  919                 panic("ufs_whiteout: unknown op");
  920         }
  921         return (error);
  922 }
  923 
  924 /*
  925  * Rename system call.
  926  *      rename("foo", "bar");
  927  * is essentially
  928  *      unlink("bar");
  929  *      link("foo", "bar");
  930  *      unlink("foo");
  931  * but ``atomically''.  Can't do full commit without saving state in the
  932  * inode on disk which isn't feasible at this time.  Best we can do is
  933  * always guarantee the target exists.
  934  *
  935  * Basic algorithm is:
  936  *
  937  * 1) Bump link count on source while we're linking it to the
  938  *    target.  This also ensure the inode won't be deleted out
  939  *    from underneath us while we work (it may be truncated by
  940  *    a concurrent `trunc' or `open' for creation).
  941  * 2) Link source to destination.  If destination already exists,
  942  *    delete it first.
  943  * 3) Unlink source reference to inode if still around. If a
  944  *    directory was moved and the parent of the destination
  945  *    is different from the source, patch the ".." entry in the
  946  *    directory.
  947  */
  948 static int
  949 ufs_rename(ap)
  950         struct vop_rename_args  /* {
  951                 struct vnode *a_fdvp;
  952                 struct vnode *a_fvp;
  953                 struct componentname *a_fcnp;
  954                 struct vnode *a_tdvp;
  955                 struct vnode *a_tvp;
  956                 struct componentname *a_tcnp;
  957         } */ *ap;
  958 {
  959         struct vnode *tvp = ap->a_tvp;
  960         struct vnode *tdvp = ap->a_tdvp;
  961         struct vnode *fvp = ap->a_fvp;
  962         struct vnode *fdvp = ap->a_fdvp;
  963         struct componentname *tcnp = ap->a_tcnp;
  964         struct componentname *fcnp = ap->a_fcnp;
  965         struct thread *td = fcnp->cn_thread;
  966         struct inode *ip, *xp, *dp;
  967         struct direct newdir;
  968         int doingdirectory = 0, oldparent = 0, newparent = 0;
  969         int error = 0, ioflag;
  970 
  971 #ifdef DIAGNOSTIC
  972         if ((tcnp->cn_flags & HASBUF) == 0 ||
  973             (fcnp->cn_flags & HASBUF) == 0)
  974                 panic("ufs_rename: no name");
  975 #endif
  976         /*
  977          * Check for cross-device rename.
  978          */
  979         if ((fvp->v_mount != tdvp->v_mount) ||
  980             (tvp && (fvp->v_mount != tvp->v_mount))) {
  981                 error = EXDEV;
  982 abortit:
  983                 if (tdvp == tvp)
  984                         vrele(tdvp);
  985                 else
  986                         vput(tdvp);
  987                 if (tvp)
  988                         vput(tvp);
  989                 vrele(fdvp);
  990                 vrele(fvp);
  991                 return (error);
  992         }
  993 
  994         if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
  995             (VTOI(tdvp)->i_flags & APPEND))) {
  996                 error = EPERM;
  997                 goto abortit;
  998         }
  999 
 1000         /*
 1001          * Renaming a file to itself has no effect.  The upper layers should
 1002          * not call us in that case.  Temporarily just warn if they do.
 1003          */
 1004         if (fvp == tvp) {
 1005                 printf("ufs_rename: fvp == tvp (can't happen)\n");
 1006                 error = 0;
 1007                 goto abortit;
 1008         }
 1009 
 1010         if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0)
 1011                 goto abortit;
 1012         dp = VTOI(fdvp);
 1013         ip = VTOI(fvp);
 1014         if (ip->i_nlink >= LINK_MAX) {
 1015                 VOP_UNLOCK(fvp, 0, td);
 1016                 error = EMLINK;
 1017                 goto abortit;
 1018         }
 1019         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
 1020             || (dp->i_flags & APPEND)) {
 1021                 VOP_UNLOCK(fvp, 0, td);
 1022                 error = EPERM;
 1023                 goto abortit;
 1024         }
 1025         if ((ip->i_mode & IFMT) == IFDIR) {
 1026                 /*
 1027                  * Avoid ".", "..", and aliases of "." for obvious reasons.
 1028                  */
 1029                 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
 1030                     dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT ||
 1031                     (ip->i_flag & IN_RENAME)) {
 1032                         VOP_UNLOCK(fvp, 0, td);
 1033                         error = EINVAL;
 1034                         goto abortit;
 1035                 }
 1036                 ip->i_flag |= IN_RENAME;
 1037                 oldparent = dp->i_number;
 1038                 doingdirectory = 1;
 1039         }
 1040         VN_KNOTE_UNLOCKED(fdvp, NOTE_WRITE);            /* XXX right place? */
 1041         vrele(fdvp);
 1042 
 1043         /*
 1044          * When the target exists, both the directory
 1045          * and target vnodes are returned locked.
 1046          */
 1047         dp = VTOI(tdvp);
 1048         xp = NULL;
 1049         if (tvp)
 1050                 xp = VTOI(tvp);
 1051 
 1052         /*
 1053          * 1) Bump link count while we're moving stuff
 1054          *    around.  If we crash somewhere before
 1055          *    completing our work, the link count
 1056          *    may be wrong, but correctable.
 1057          */
 1058         ip->i_effnlink++;
 1059         ip->i_nlink++;
 1060         DIP_SET(ip, i_nlink, ip->i_nlink);
 1061         ip->i_flag |= IN_CHANGE;
 1062         if (DOINGSOFTDEP(fvp))
 1063                 softdep_change_linkcnt(ip);
 1064         if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) |
 1065                                        DOINGASYNC(fvp)))) != 0) {
 1066                 VOP_UNLOCK(fvp, 0, td);
 1067                 goto bad;
 1068         }
 1069 
 1070         /*
 1071          * If ".." must be changed (ie the directory gets a new
 1072          * parent) then the source directory must not be in the
 1073          * directory heirarchy above the target, as this would
 1074          * orphan everything below the source directory. Also
 1075          * the user must have write permission in the source so
 1076          * as to be able to change "..". We must repeat the call
 1077          * to namei, as the parent directory is unlocked by the
 1078          * call to checkpath().
 1079          */
 1080         error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
 1081         VOP_UNLOCK(fvp, 0, td);
 1082         if (oldparent != dp->i_number)
 1083                 newparent = dp->i_number;
 1084         if (doingdirectory && newparent) {
 1085                 if (error)      /* write access check above */
 1086                         goto bad;
 1087                 if (xp != NULL)
 1088                         vput(tvp);
 1089                 error = ufs_checkpath(ip, dp, tcnp->cn_cred);
 1090                 if (error)
 1091                         goto out;
 1092                 if ((tcnp->cn_flags & SAVESTART) == 0)
 1093                         panic("ufs_rename: lost to startdir");
 1094                 VREF(tdvp);
 1095                 error = relookup(tdvp, &tvp, tcnp);
 1096                 if (error)
 1097                         goto out;
 1098                 vrele(tdvp);
 1099                 dp = VTOI(tdvp);
 1100                 xp = NULL;
 1101                 if (tvp)
 1102                         xp = VTOI(tvp);
 1103         }
 1104         /*
 1105          * 2) If target doesn't exist, link the target
 1106          *    to the source and unlink the source.
 1107          *    Otherwise, rewrite the target directory
 1108          *    entry to reference the source inode and
 1109          *    expunge the original entry's existence.
 1110          */
 1111         if (xp == NULL) {
 1112                 if (dp->i_dev != ip->i_dev)
 1113                         panic("ufs_rename: EXDEV");
 1114                 /*
 1115                  * Account for ".." in new directory.
 1116                  * When source and destination have the same
 1117                  * parent we don't fool with the link count.
 1118                  */
 1119                 if (doingdirectory && newparent) {
 1120                         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 1121                                 error = EMLINK;
 1122                                 goto bad;
 1123                         }
 1124                         dp->i_effnlink++;
 1125                         dp->i_nlink++;
 1126                         DIP_SET(dp, i_nlink, dp->i_nlink);
 1127                         dp->i_flag |= IN_CHANGE;
 1128                         if (DOINGSOFTDEP(tdvp))
 1129                                 softdep_change_linkcnt(dp);
 1130                         error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
 1131                                                    DOINGASYNC(tdvp)));
 1132                         if (error)
 1133                                 goto bad;
 1134                 }
 1135                 ufs_makedirentry(ip, tcnp, &newdir);
 1136                 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL);
 1137                 if (error) {
 1138                         if (doingdirectory && newparent) {
 1139                                 dp->i_effnlink--;
 1140                                 dp->i_nlink--;
 1141                                 DIP_SET(dp, i_nlink, dp->i_nlink);
 1142                                 dp->i_flag |= IN_CHANGE;
 1143                                 if (DOINGSOFTDEP(tdvp))
 1144                                         softdep_change_linkcnt(dp);
 1145                                 (void)UFS_UPDATE(tdvp, 1);
 1146                         }
 1147                         goto bad;
 1148                 }
 1149                 VN_KNOTE_UNLOCKED(tdvp, NOTE_WRITE);
 1150                 vput(tdvp);
 1151         } else {
 1152                 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
 1153                         panic("ufs_rename: EXDEV");
 1154                 /*
 1155                  * Short circuit rename(foo, foo).
 1156                  */
 1157                 if (xp->i_number == ip->i_number)
 1158                         panic("ufs_rename: same file");
 1159                 /*
 1160                  * If the parent directory is "sticky", then the caller
 1161                  * must possess VADMIN for the parent directory, or the
 1162                  * destination of the rename.  This implements append-only
 1163                  * directories.
 1164                  */
 1165                 if ((dp->i_mode & S_ISTXT) &&
 1166                     VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, td) &&
 1167                     VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, td)) {
 1168                         error = EPERM;
 1169                         goto bad;
 1170                 }
 1171                 /*
 1172                  * Target must be empty if a directory and have no links
 1173                  * to it. Also, ensure source and target are compatible
 1174                  * (both directories, or both not directories).
 1175                  */
 1176                 if ((xp->i_mode&IFMT) == IFDIR) {
 1177                         if ((xp->i_effnlink > 2) ||
 1178                             !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) {
 1179                                 error = ENOTEMPTY;
 1180                                 goto bad;
 1181                         }
 1182                         if (!doingdirectory) {
 1183                                 error = ENOTDIR;
 1184                                 goto bad;
 1185                         }
 1186                         cache_purge(tdvp);
 1187                 } else if (doingdirectory) {
 1188                         error = EISDIR;
 1189                         goto bad;
 1190                 }
 1191                 error = ufs_dirrewrite(dp, xp, ip->i_number,
 1192                     IFTODT(ip->i_mode),
 1193                     (doingdirectory && newparent) ? newparent : doingdirectory);
 1194                 if (error)
 1195                         goto bad;
 1196                 if (doingdirectory) {
 1197                         if (!newparent) {
 1198                                 dp->i_effnlink--;
 1199                                 if (DOINGSOFTDEP(tdvp))
 1200                                         softdep_change_linkcnt(dp);
 1201                         }
 1202                         xp->i_effnlink--;
 1203                         if (DOINGSOFTDEP(tvp))
 1204                                 softdep_change_linkcnt(xp);
 1205                 }
 1206                 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
 1207                         /*
 1208                          * Truncate inode. The only stuff left in the directory
 1209                          * is "." and "..". The "." reference is inconsequential
 1210                          * since we are quashing it. We have removed the "."
 1211                          * reference and the reference in the parent directory,
 1212                          * but there may be other hard links. The soft
 1213                          * dependency code will arrange to do these operations
 1214                          * after the parent directory entry has been deleted on
 1215                          * disk, so when running with that code we avoid doing
 1216                          * them now.
 1217                          */
 1218                         if (!newparent) {
 1219                                 dp->i_nlink--;
 1220                                 DIP_SET(dp, i_nlink, dp->i_nlink);
 1221                                 dp->i_flag |= IN_CHANGE;
 1222                         }
 1223                         xp->i_nlink--;
 1224                         DIP_SET(xp, i_nlink, xp->i_nlink);
 1225                         xp->i_flag |= IN_CHANGE;
 1226                         ioflag = IO_NORMAL;
 1227                         if (DOINGASYNC(tvp))
 1228                                 ioflag |= IO_SYNC;
 1229                         if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag,
 1230                             tcnp->cn_cred, tcnp->cn_thread)) != 0)
 1231                                 goto bad;
 1232                 }
 1233                 VN_KNOTE_UNLOCKED(tdvp, NOTE_WRITE);
 1234                 vput(tdvp);
 1235                 VN_KNOTE_UNLOCKED(tvp, NOTE_DELETE);
 1236                 vput(tvp);
 1237                 xp = NULL;
 1238         }
 1239 
 1240         /*
 1241          * 3) Unlink the source.
 1242          */
 1243         fcnp->cn_flags &= ~MODMASK;
 1244         fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
 1245         if ((fcnp->cn_flags & SAVESTART) == 0)
 1246                 panic("ufs_rename: lost from startdir");
 1247         VREF(fdvp);
 1248         error = relookup(fdvp, &fvp, fcnp);
 1249         if (error == 0)
 1250                 vrele(fdvp);
 1251         if (fvp != NULL) {
 1252                 xp = VTOI(fvp);
 1253                 dp = VTOI(fdvp);
 1254         } else {
 1255                 /*
 1256                  * From name has disappeared.  IN_RENAME is not sufficient
 1257                  * to protect against directory races due to timing windows,
 1258                  * so we have to remove the panic.  XXX the only real way
 1259                  * to solve this issue is at a much higher level.  By the
 1260                  * time we hit ufs_rename() it's too late.
 1261                  */
 1262 #if 0
 1263                 if (doingdirectory)
 1264                         panic("ufs_rename: lost dir entry");
 1265 #endif
 1266                 vrele(ap->a_fvp);
 1267                 return (0);
 1268         }
 1269         /*
 1270          * Ensure that the directory entry still exists and has not
 1271          * changed while the new name has been entered. If the source is
 1272          * a file then the entry may have been unlinked or renamed. In
 1273          * either case there is no further work to be done. If the source
 1274          * is a directory then it cannot have been rmdir'ed; the IN_RENAME
 1275          * flag ensures that it cannot be moved by another rename or removed
 1276          * by a rmdir.
 1277          */
 1278         if (xp != ip) {
 1279                 /*
 1280                  * From name resolves to a different inode.  IN_RENAME is
 1281                  * not sufficient protection against timing window races
 1282                  * so we can't panic here.  XXX the only real way
 1283                  * to solve this issue is at a much higher level.  By the
 1284                  * time we hit ufs_rename() it's too late.
 1285                  */
 1286 #if 0
 1287                 if (doingdirectory)
 1288                         panic("ufs_rename: lost dir entry");
 1289 #endif
 1290         } else {
 1291                 /*
 1292                  * If the source is a directory with a
 1293                  * new parent, the link count of the old
 1294                  * parent directory must be decremented
 1295                  * and ".." set to point to the new parent.
 1296                  */
 1297                 if (doingdirectory && newparent) {
 1298                         xp->i_offset = mastertemplate.dot_reclen;
 1299                         ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0);
 1300                         cache_purge(fdvp);
 1301                 }
 1302                 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0);
 1303                 xp->i_flag &= ~IN_RENAME;
 1304         }
 1305         VN_KNOTE_UNLOCKED(fvp, NOTE_RENAME);
 1306         if (dp)
 1307                 vput(fdvp);
 1308         if (xp)
 1309                 vput(fvp);
 1310         vrele(ap->a_fvp);
 1311         return (error);
 1312 
 1313 bad:
 1314         if (xp)
 1315                 vput(ITOV(xp));
 1316         vput(ITOV(dp));
 1317 out:
 1318         if (doingdirectory)
 1319                 ip->i_flag &= ~IN_RENAME;
 1320         if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) {
 1321                 ip->i_effnlink--;
 1322                 ip->i_nlink--;
 1323                 DIP_SET(ip, i_nlink, ip->i_nlink);
 1324                 ip->i_flag |= IN_CHANGE;
 1325                 ip->i_flag &= ~IN_RENAME;
 1326                 if (DOINGSOFTDEP(fvp))
 1327                         softdep_change_linkcnt(ip);
 1328                 vput(fvp);
 1329         } else
 1330                 vrele(fvp);
 1331         return (error);
 1332 }
 1333 
 1334 /*
 1335  * Mkdir system call
 1336  */
 1337 static int
 1338 ufs_mkdir(ap)
 1339         struct vop_mkdir_args /* {
 1340                 struct vnode *a_dvp;
 1341                 struct vnode **a_vpp;
 1342                 struct componentname *a_cnp;
 1343                 struct vattr *a_vap;
 1344         } */ *ap;
 1345 {
 1346         struct vnode *dvp = ap->a_dvp;
 1347         struct vattr *vap = ap->a_vap;
 1348         struct componentname *cnp = ap->a_cnp;
 1349         struct inode *ip, *dp;
 1350         struct vnode *tvp;
 1351         struct buf *bp;
 1352         struct dirtemplate dirtemplate, *dtp;
 1353         struct direct newdir;
 1354 #ifdef UFS_ACL
 1355         struct acl *acl, *dacl;
 1356 #endif
 1357         int error, dmode;
 1358         long blkoff;
 1359 
 1360 #ifdef DIAGNOSTIC
 1361         if ((cnp->cn_flags & HASBUF) == 0)
 1362                 panic("ufs_mkdir: no name");
 1363 #endif
 1364         dp = VTOI(dvp);
 1365         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 1366                 error = EMLINK;
 1367                 goto out;
 1368         }
 1369         dmode = vap->va_mode & 0777;
 1370         dmode |= IFDIR;
 1371         /*
 1372          * Must simulate part of ufs_makeinode here to acquire the inode,
 1373          * but not have it entered in the parent directory. The entry is
 1374          * made later after writing "." and ".." entries.
 1375          */
 1376         error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
 1377         if (error)
 1378                 goto out;
 1379         ip = VTOI(tvp);
 1380         ip->i_gid = dp->i_gid;
 1381         DIP_SET(ip, i_gid, dp->i_gid);
 1382 #ifdef SUIDDIR
 1383         {
 1384 #ifdef QUOTA
 1385                 struct ucred ucred, *ucp;
 1386                 ucp = cnp->cn_cred;
 1387 #endif
 1388                 /*
 1389                  * If we are hacking owners here, (only do this where told to)
 1390                  * and we are not giving it TO root, (would subvert quotas)
 1391                  * then go ahead and give it to the other user.
 1392                  * The new directory also inherits the SUID bit.
 1393                  * If user's UID and dir UID are the same,
 1394                  * 'give it away' so that the SUID is still forced on.
 1395                  */
 1396                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 1397                     (dp->i_mode & ISUID) && dp->i_uid) {
 1398                         dmode |= ISUID;
 1399                         ip->i_uid = dp->i_uid;
 1400                         DIP_SET(ip, i_uid, dp->i_uid);
 1401 #ifdef QUOTA
 1402                         if (dp->i_uid != cnp->cn_cred->cr_uid) {
 1403                                 /*
 1404                                  * Make sure the correct user gets charged
 1405                                  * for the space.
 1406                                  * Make a dummy credential for the victim.
 1407                                  * XXX This seems to never be accessed out of
 1408                                  * our context so a stack variable is ok.
 1409                                  */
 1410                                 ucred.cr_ref = 1;
 1411                                 ucred.cr_uid = ip->i_uid;
 1412                                 ucred.cr_ngroups = 1;
 1413                                 ucred.cr_groups[0] = dp->i_gid;
 1414                                 ucp = &ucred;
 1415                         }
 1416 #endif
 1417                 } else {
 1418                         ip->i_uid = cnp->cn_cred->cr_uid;
 1419                         DIP_SET(ip, i_uid, ip->i_uid);
 1420                 }
 1421 #ifdef QUOTA
 1422                 if ((error = getinoquota(ip)) ||
 1423                     (error = chkiq(ip, 1, ucp, 0))) {
 1424                         UFS_VFREE(tvp, ip->i_number, dmode);
 1425                         vput(tvp);
 1426                         return (error);
 1427                 }
 1428 #endif
 1429         }
 1430 #else   /* !SUIDDIR */
 1431         ip->i_uid = cnp->cn_cred->cr_uid;
 1432         DIP_SET(ip, i_uid, ip->i_uid);
 1433 #ifdef QUOTA
 1434         if ((error = getinoquota(ip)) ||
 1435             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 1436                 UFS_VFREE(tvp, ip->i_number, dmode);
 1437                 vput(tvp);
 1438                 return (error);
 1439         }
 1440 #endif
 1441 #endif  /* !SUIDDIR */
 1442         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 1443 #ifdef UFS_ACL
 1444         acl = dacl = NULL;
 1445         if ((dvp->v_mount->mnt_flag & MNT_ACLS) != 0) {
 1446                 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK);
 1447                 MALLOC(dacl, struct acl *, sizeof(*dacl), M_ACL, M_WAITOK);
 1448 
 1449                 /*
 1450                  * Retrieve default ACL from parent, if any.
 1451                  */
 1452                 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cnp->cn_cred,
 1453                     cnp->cn_thread);
 1454                 switch (error) {
 1455                 case 0:
 1456                         /*
 1457                          * Retrieved a default ACL, so merge mode and ACL if
 1458                          * necessary.  If the ACL is empty, fall through to
 1459                          * the "not defined or available" case.
 1460                          */
 1461                         if (acl->acl_cnt != 0) {
 1462                                 dmode = acl_posix1e_newfilemode(dmode, acl);
 1463                                 ip->i_mode = dmode;
 1464                                 DIP_SET(ip, i_mode, dmode);
 1465                                 *dacl = *acl;
 1466                                 ufs_sync_acl_from_inode(ip, acl);
 1467                                 break;
 1468                         }
 1469                         /* FALLTHROUGH */
 1470         
 1471                 case EOPNOTSUPP:
 1472                         /*
 1473                          * Just use the mode as-is.
 1474                          */
 1475                         ip->i_mode = dmode;
 1476                         DIP_SET(ip, i_mode, dmode);
 1477                         FREE(acl, M_ACL);
 1478                         FREE(dacl, M_ACL);
 1479                         dacl = acl = NULL;
 1480                         break;
 1481                 
 1482                 default:
 1483                         UFS_VFREE(tvp, ip->i_number, dmode);
 1484                         vput(tvp);
 1485                         FREE(acl, M_ACL);
 1486                                 FREE(dacl, M_ACL);
 1487                         return (error);
 1488                 }
 1489         } else {
 1490 #endif /* !UFS_ACL */
 1491                 ip->i_mode = dmode;
 1492                 DIP_SET(ip, i_mode, dmode);
 1493 #ifdef UFS_ACL
 1494         }
 1495 #endif
 1496         tvp->v_type = VDIR;     /* Rest init'd in getnewvnode(). */
 1497         ip->i_effnlink = 2;
 1498         ip->i_nlink = 2;
 1499         DIP_SET(ip, i_nlink, 2);
 1500         if (DOINGSOFTDEP(tvp))
 1501                 softdep_change_linkcnt(ip);
 1502         if (cnp->cn_flags & ISWHITEOUT) {
 1503                 ip->i_flags |= UF_OPAQUE;
 1504                 DIP_SET(ip, i_flags, ip->i_flags);
 1505         }
 1506 
 1507         /*
 1508          * Bump link count in parent directory to reflect work done below.
 1509          * Should be done before reference is created so cleanup is
 1510          * possible if we crash.
 1511          */
 1512         dp->i_effnlink++;
 1513         dp->i_nlink++;
 1514         DIP_SET(dp, i_nlink, dp->i_nlink);
 1515         dp->i_flag |= IN_CHANGE;
 1516         if (DOINGSOFTDEP(dvp))
 1517                 softdep_change_linkcnt(dp);
 1518         error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
 1519         if (error)
 1520                 goto bad;
 1521 #ifdef MAC
 1522         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 1523                 error = mac_create_vnode_extattr(cnp->cn_cred, dvp->v_mount,
 1524                     dvp, tvp, cnp);
 1525                 if (error)
 1526                         goto bad;
 1527         }
 1528 #endif
 1529 #ifdef UFS_ACL
 1530         if (acl != NULL) {
 1531                 /*
 1532                  * XXX: If we abort now, will Soft Updates notify the extattr
 1533                  * code that the EAs for the file need to be released?
 1534                  */
 1535                 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cnp->cn_cred,
 1536                     cnp->cn_thread);
 1537                 if (error == 0)
 1538                         error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl,
 1539                             cnp->cn_cred, cnp->cn_thread);
 1540                 switch (error) {
 1541                 case 0:
 1542                         break;
 1543 
 1544                 case EOPNOTSUPP:
 1545                         /*
 1546                          * XXX: This should not happen, as EOPNOTSUPP above
 1547                          * was supposed to free acl.
 1548                          */
 1549                         printf("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
 1550                         /*
 1551                         panic("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()");
 1552                          */
 1553                         break;
 1554 
 1555                 default:
 1556                         FREE(acl, M_ACL);
 1557                         FREE(dacl, M_ACL);
 1558                         goto bad;
 1559                 }
 1560                 FREE(acl, M_ACL);
 1561                 FREE(dacl, M_ACL);
 1562         }
 1563 #endif /* !UFS_ACL */
 1564 
 1565         /*
 1566          * Initialize directory with "." and ".." from static template.
 1567          */
 1568         if (dvp->v_mount->mnt_maxsymlinklen > 0
 1569         )
 1570                 dtp = &mastertemplate;
 1571         else
 1572                 dtp = (struct dirtemplate *)&omastertemplate;
 1573         dirtemplate = *dtp;
 1574         dirtemplate.dot_ino = ip->i_number;
 1575         dirtemplate.dotdot_ino = dp->i_number;
 1576         if ((error = UFS_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
 1577             BA_CLRBUF, &bp)) != 0)
 1578                 goto bad;
 1579         ip->i_size = DIRBLKSIZ;
 1580         DIP_SET(ip, i_size, DIRBLKSIZ);
 1581         ip->i_flag |= IN_CHANGE | IN_UPDATE;
 1582         vnode_pager_setsize(tvp, (u_long)ip->i_size);
 1583         bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
 1584         if (DOINGSOFTDEP(tvp)) {
 1585                 /*
 1586                  * Ensure that the entire newly allocated block is a
 1587                  * valid directory so that future growth within the
 1588                  * block does not have to ensure that the block is
 1589                  * written before the inode.
 1590                  */
 1591                 blkoff = DIRBLKSIZ;
 1592                 while (blkoff < bp->b_bcount) {
 1593                         ((struct direct *)
 1594                            (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
 1595                         blkoff += DIRBLKSIZ;
 1596                 }
 1597         }
 1598         if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
 1599                                        DOINGASYNC(tvp)))) != 0) {
 1600                 (void)bwrite(bp);
 1601                 goto bad;
 1602         }
 1603         /*
 1604          * Directory set up, now install its entry in the parent directory.
 1605          *
 1606          * If we are not doing soft dependencies, then we must write out the
 1607          * buffer containing the new directory body before entering the new 
 1608          * name in the parent. If we are doing soft dependencies, then the
 1609          * buffer containing the new directory body will be passed to and
 1610          * released in the soft dependency code after the code has attached
 1611          * an appropriate ordering dependency to the buffer which ensures that
 1612          * the buffer is written before the new name is written in the parent.
 1613          */
 1614         if (DOINGASYNC(dvp))
 1615                 bdwrite(bp);
 1616         else if (!DOINGSOFTDEP(dvp) && ((error = bwrite(bp))))
 1617                 goto bad;
 1618         ufs_makedirentry(ip, cnp, &newdir);
 1619         error = ufs_direnter(dvp, tvp, &newdir, cnp, bp);
 1620         
 1621 bad:
 1622         if (error == 0) {
 1623                 VN_KNOTE_UNLOCKED(dvp, NOTE_WRITE | NOTE_LINK);
 1624                 *ap->a_vpp = tvp;
 1625         } else {
 1626                 dp->i_effnlink--;
 1627                 dp->i_nlink--;
 1628                 DIP_SET(dp, i_nlink, dp->i_nlink);
 1629                 dp->i_flag |= IN_CHANGE;
 1630                 if (DOINGSOFTDEP(dvp))
 1631                         softdep_change_linkcnt(dp);
 1632                 /*
 1633                  * No need to do an explicit VOP_TRUNCATE here, vrele will
 1634                  * do this for us because we set the link count to 0.
 1635                  */
 1636                 ip->i_effnlink = 0;
 1637                 ip->i_nlink = 0;
 1638                 DIP_SET(ip, i_nlink, 0);
 1639                 ip->i_flag |= IN_CHANGE;
 1640                 if (DOINGSOFTDEP(tvp))
 1641                         softdep_change_linkcnt(ip);
 1642                 vput(tvp);
 1643         }
 1644 out:
 1645         return (error);
 1646 }
 1647 
 1648 /*
 1649  * Rmdir system call.
 1650  */
 1651 static int
 1652 ufs_rmdir(ap)
 1653         struct vop_rmdir_args /* {
 1654                 struct vnode *a_dvp;
 1655                 struct vnode *a_vp;
 1656                 struct componentname *a_cnp;
 1657         } */ *ap;
 1658 {
 1659         struct vnode *vp = ap->a_vp;
 1660         struct vnode *dvp = ap->a_dvp;
 1661         struct componentname *cnp = ap->a_cnp;
 1662         struct inode *ip, *dp;
 1663         int error, ioflag;
 1664 
 1665         ip = VTOI(vp);
 1666         dp = VTOI(dvp);
 1667 
 1668         /*
 1669          * Do not remove a directory that is in the process of being renamed.
 1670          * Verify the directory is empty (and valid). Rmdir ".." will not be
 1671          * valid since ".." will contain a reference to the current directory
 1672          * and thus be non-empty. Do not allow the removal of mounted on
 1673          * directories (this can happen when an NFS exported filesystem
 1674          * tries to remove a locally mounted on directory).
 1675          */
 1676         error = 0;
 1677         if (ip->i_flag & IN_RENAME) {
 1678                 error = EINVAL;
 1679                 goto out;
 1680         }
 1681         if (ip->i_effnlink != 2 ||
 1682             !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
 1683                 error = ENOTEMPTY;
 1684                 goto out;
 1685         }
 1686         if ((dp->i_flags & APPEND)
 1687             || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
 1688                 error = EPERM;
 1689                 goto out;
 1690         }
 1691         if (vp->v_mountedhere != 0) {
 1692                 error = EINVAL;
 1693                 goto out;
 1694         }
 1695         /*
 1696          * Delete reference to directory before purging
 1697          * inode.  If we crash in between, the directory
 1698          * will be reattached to lost+found,
 1699          */
 1700         dp->i_effnlink--;
 1701         ip->i_effnlink--;
 1702         if (DOINGSOFTDEP(vp)) {
 1703                 softdep_change_linkcnt(dp);
 1704                 softdep_change_linkcnt(ip);
 1705         }
 1706         error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
 1707         if (error) {
 1708                 dp->i_effnlink++;
 1709                 ip->i_effnlink++;
 1710                 if (DOINGSOFTDEP(vp)) {
 1711                         softdep_change_linkcnt(dp);
 1712                         softdep_change_linkcnt(ip);
 1713                 }
 1714                 goto out;
 1715         }
 1716         VN_KNOTE_UNLOCKED(dvp, NOTE_WRITE | NOTE_LINK);
 1717         cache_purge(dvp);
 1718         /*
 1719          * Truncate inode. The only stuff left in the directory is "." and
 1720          * "..". The "." reference is inconsequential since we are quashing
 1721          * it. The soft dependency code will arrange to do these operations
 1722          * after the parent directory entry has been deleted on disk, so
 1723          * when running with that code we avoid doing them now.
 1724          */
 1725         if (!DOINGSOFTDEP(vp)) {
 1726                 dp->i_nlink--;
 1727                 DIP_SET(dp, i_nlink, dp->i_nlink);
 1728                 dp->i_flag |= IN_CHANGE;
 1729                 ip->i_nlink--;
 1730                 DIP_SET(ip, i_nlink, ip->i_nlink);
 1731                 ip->i_flag |= IN_CHANGE;
 1732                 ioflag = IO_NORMAL;
 1733                 if (DOINGASYNC(vp))
 1734                         ioflag |= IO_SYNC;
 1735                 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred,
 1736                     cnp->cn_thread);
 1737         }
 1738         cache_purge(vp);
 1739 #ifdef UFS_DIRHASH
 1740         /* Kill any active hash; i_effnlink == 0, so it will not come back. */
 1741         if (ip->i_dirhash != NULL)
 1742                 ufsdirhash_free(ip);
 1743 #endif
 1744 out:
 1745         VN_KNOTE_UNLOCKED(vp, NOTE_DELETE);
 1746         return (error);
 1747 }
 1748 
 1749 /*
 1750  * symlink -- make a symbolic link
 1751  */
 1752 static int
 1753 ufs_symlink(ap)
 1754         struct vop_symlink_args /* {
 1755                 struct vnode *a_dvp;
 1756                 struct vnode **a_vpp;
 1757                 struct componentname *a_cnp;
 1758                 struct vattr *a_vap;
 1759                 char *a_target;
 1760         } */ *ap;
 1761 {
 1762         struct vnode *vp, **vpp = ap->a_vpp;
 1763         struct inode *ip;
 1764         int len, error;
 1765 
 1766         error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
 1767             vpp, ap->a_cnp);
 1768         if (error)
 1769                 return (error);
 1770         VN_KNOTE_UNLOCKED(ap->a_dvp, NOTE_WRITE);
 1771         vp = *vpp;
 1772         len = strlen(ap->a_target);
 1773         if (len < vp->v_mount->mnt_maxsymlinklen) {
 1774                 ip = VTOI(vp);
 1775                 bcopy(ap->a_target, SHORTLINK(ip), len);
 1776                 ip->i_size = len;
 1777                 DIP_SET(ip, i_size, len);
 1778                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 1779         } else
 1780                 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
 1781                     UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
 1782                     ap->a_cnp->cn_cred, NOCRED, (int *)0, (struct thread *)0);
 1783         if (error)
 1784                 vput(vp);
 1785         return (error);
 1786 }
 1787 
 1788 /*
 1789  * Vnode op for reading directories.
 1790  *
 1791  * The routine below assumes that the on-disk format of a directory
 1792  * is the same as that defined by <sys/dirent.h>. If the on-disk
 1793  * format changes, then it will be necessary to do a conversion
 1794  * from the on-disk format that read returns to the format defined
 1795  * by <sys/dirent.h>.
 1796  */
 1797 int
 1798 ufs_readdir(ap)
 1799         struct vop_readdir_args /* {
 1800                 struct vnode *a_vp;
 1801                 struct uio *a_uio;
 1802                 struct ucred *a_cred;
 1803                 int *a_eofflag;
 1804                 int *ncookies;
 1805                 u_long **a_cookies;
 1806         } */ *ap;
 1807 {
 1808         struct uio *uio = ap->a_uio;
 1809         int error;
 1810         size_t count, lost;
 1811         off_t off;
 1812 
 1813         if (ap->a_ncookies != NULL)
 1814                 /*
 1815                  * Ensure that the block is aligned.  The caller can use
 1816                  * the cookies to determine where in the block to start.
 1817                  */
 1818                 uio->uio_offset &= ~(DIRBLKSIZ - 1);
 1819         off = uio->uio_offset;
 1820         count = uio->uio_resid;
 1821         /* Make sure we don't return partial entries. */
 1822         if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1)))
 1823                 return (EINVAL);
 1824         count -= (uio->uio_offset + count) & (DIRBLKSIZ -1);
 1825         lost = uio->uio_resid - count;
 1826         uio->uio_resid = count;
 1827         uio->uio_iov->iov_len = count;
 1828 #       if (BYTE_ORDER == LITTLE_ENDIAN)
 1829                 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) {
 1830                         error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
 1831                 } else {
 1832                         struct dirent *dp, *edp;
 1833                         struct uio auio;
 1834                         struct iovec aiov;
 1835                         caddr_t dirbuf;
 1836                         int readcnt;
 1837                         u_char tmp;
 1838 
 1839                         auio = *uio;
 1840                         auio.uio_iov = &aiov;
 1841                         auio.uio_iovcnt = 1;
 1842                         auio.uio_segflg = UIO_SYSSPACE;
 1843                         aiov.iov_len = count;
 1844                         MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK);
 1845                         aiov.iov_base = dirbuf;
 1846                         error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred);
 1847                         if (error == 0) {
 1848                                 readcnt = count - auio.uio_resid;
 1849                                 edp = (struct dirent *)&dirbuf[readcnt];
 1850                                 for (dp = (struct dirent *)dirbuf; dp < edp; ) {
 1851                                         tmp = dp->d_namlen;
 1852                                         dp->d_namlen = dp->d_type;
 1853                                         dp->d_type = tmp;
 1854                                         if (dp->d_reclen > 0) {
 1855                                                 dp = (struct dirent *)
 1856                                                     ((char *)dp + dp->d_reclen);
 1857                                         } else {
 1858                                                 error = EIO;
 1859                                                 break;
 1860                                         }
 1861                                 }
 1862                                 if (dp >= edp)
 1863                                         error = uiomove(dirbuf, readcnt, uio);
 1864                         }
 1865                         FREE(dirbuf, M_TEMP);
 1866                 }
 1867 #       else
 1868                 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred);
 1869 #       endif
 1870         if (!error && ap->a_ncookies != NULL) {
 1871                 struct dirent* dpStart;
 1872                 struct dirent* dpEnd;
 1873                 struct dirent* dp;
 1874                 int ncookies;
 1875                 u_long *cookies;
 1876                 u_long *cookiep;
 1877 
 1878                 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1)
 1879                         panic("ufs_readdir: unexpected uio from NFS server");
 1880                 dpStart = (struct dirent *)
 1881                     ((char *)uio->uio_iov->iov_base - (uio->uio_offset - off));
 1882                 dpEnd = (struct dirent *) uio->uio_iov->iov_base;
 1883                 for (dp = dpStart, ncookies = 0;
 1884                      dp < dpEnd;
 1885                      dp = (struct dirent *)((caddr_t) dp + dp->d_reclen))
 1886                         ncookies++;
 1887                 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP,
 1888                     M_WAITOK);
 1889                 for (dp = dpStart, cookiep = cookies;
 1890                      dp < dpEnd;
 1891                      dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) {
 1892                         off += dp->d_reclen;
 1893                         *cookiep++ = (u_long) off;
 1894                 }
 1895                 *ap->a_ncookies = ncookies;
 1896                 *ap->a_cookies = cookies;
 1897         }
 1898         uio->uio_resid += lost;
 1899         if (ap->a_eofflag)
 1900             *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset;
 1901         return (error);
 1902 }
 1903 
 1904 /*
 1905  * Return target name of a symbolic link
 1906  */
 1907 static int
 1908 ufs_readlink(ap)
 1909         struct vop_readlink_args /* {
 1910                 struct vnode *a_vp;
 1911                 struct uio *a_uio;
 1912                 struct ucred *a_cred;
 1913         } */ *ap;
 1914 {
 1915         struct vnode *vp = ap->a_vp;
 1916         struct inode *ip = VTOI(vp);
 1917         doff_t isize;
 1918 
 1919         isize = ip->i_size;
 1920         if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
 1921             DIP(ip, i_blocks) == 0) { /* XXX - for old fastlink support */
 1922                 uiomove(SHORTLINK(ip), isize, ap->a_uio);
 1923                 return (0);
 1924         }
 1925         return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
 1926 }
 1927 
 1928 /*
 1929  * Calculate the logical to physical mapping if not done already,
 1930  * then call the device strategy routine.
 1931  *
 1932  * In order to be able to swap to a file, the ufs_bmaparray() operation may not
 1933  * deadlock on memory.  See ufs_bmap() for details.
 1934  */
 1935 static int
 1936 ufs_strategy(ap)
 1937         struct vop_strategy_args /* {
 1938                 struct vnode *a_vp;
 1939                 struct buf *a_bp;
 1940         } */ *ap;
 1941 {
 1942         struct buf *bp = ap->a_bp;
 1943         struct vnode *vp = ap->a_vp;
 1944         struct inode *ip;
 1945         ufs2_daddr_t blkno;
 1946         int error;
 1947 
 1948         KASSERT(ap->a_vp == ap->a_bp->b_vp, ("%s(%p != %p)",
 1949             __func__, ap->a_vp, ap->a_bp->b_vp));
 1950         ip = VTOI(vp);
 1951         if (bp->b_blkno == bp->b_lblkno) {
 1952                 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, bp, NULL, NULL);
 1953                 bp->b_blkno = blkno;
 1954                 if (error) {
 1955                         bp->b_error = error;
 1956                         bp->b_ioflags |= BIO_ERROR;
 1957                         bufdone(bp);
 1958                         return (error);
 1959                 }
 1960                 if ((long)bp->b_blkno == -1)
 1961                         vfs_bio_clrbuf(bp);
 1962         }
 1963         if ((long)bp->b_blkno == -1) {
 1964                 bufdone(bp);
 1965                 return (0);
 1966         }
 1967         vp = ip->i_devvp;
 1968         bp->b_dev = vp->v_rdev;
 1969         bp->b_iooffset = dbtob(bp->b_blkno);
 1970         if (!buf_prewrite(vp, bp)) 
 1971                 VOP_SPECSTRATEGY(vp, bp);
 1972         return (0);
 1973 }
 1974 
 1975 /*
 1976  * Snapshots require all lock requests to be exclusive.
 1977  */
 1978 static int
 1979 ufs_lock(ap)
 1980         struct vop_lock_args /* {
 1981                 struct vnode *a_vp;
 1982                 int a_flags;
 1983                 struct thread *a_td;
 1984         } */ *ap;
 1985 {
 1986         struct vnode *vp = ap->a_vp;
 1987         int flags = ap->a_flags;
 1988 
 1989         if ((VTOI(vp)->i_flags & SF_SNAPSHOT) &&
 1990             ((flags & LK_TYPE_MASK) == LK_SHARED)) {
 1991                 flags &= ~LK_TYPE_MASK;
 1992                 flags |= LK_EXCLUSIVE;
 1993         }
 1994 #ifndef DEBUG_LOCKS
 1995         return (lockmgr(vp->v_vnlock, flags, VI_MTX(vp), ap->a_td));
 1996 #else
 1997         return (debuglockmgr(vp->v_vnlock, flags, VI_MTX(vp),
 1998             ap->a_td, "vop_stdlock", vp->filename, vp->line));
 1999 #endif
 2000 }
 2001 
 2002 /*
 2003  * Print out the contents of an inode.
 2004  */
 2005 static int
 2006 ufs_print(ap)
 2007         struct vop_print_args /* {
 2008                 struct vnode *a_vp;
 2009         } */ *ap;
 2010 {
 2011         struct vnode *vp = ap->a_vp;
 2012         struct inode *ip = VTOI(vp);
 2013 
 2014         printf("\tino %lu, on dev %s (%d, %d)", (u_long)ip->i_number,
 2015             devtoname(ip->i_dev), major(ip->i_dev), minor(ip->i_dev));
 2016         if (vp->v_type == VFIFO)
 2017                 fifo_printinfo(vp);
 2018         printf("\n");
 2019         return (0);
 2020 }
 2021 
 2022 /*
 2023  * Read wrapper for special devices.
 2024  */
 2025 static int
 2026 ufsspec_read(ap)
 2027         struct vop_read_args /* {
 2028                 struct vnode *a_vp;
 2029                 struct uio *a_uio;
 2030                 int  a_ioflag;
 2031                 struct ucred *a_cred;
 2032         } */ *ap;
 2033 {
 2034         int error, resid;
 2035         struct inode *ip;
 2036         struct uio *uio;
 2037 
 2038         uio = ap->a_uio;
 2039         resid = uio->uio_resid;
 2040         error = VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap);
 2041         /*
 2042          * The inode may have been revoked during the call, so it must not
 2043          * be accessed blindly here or in the other wrapper functions.
 2044          */
 2045         ip = VTOI(ap->a_vp);
 2046         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2047                 ip->i_flag |= IN_ACCESS;
 2048         return (error);
 2049 }
 2050 
 2051 /*
 2052  * Write wrapper for special devices.
 2053  */
 2054 static int
 2055 ufsspec_write(ap)
 2056         struct vop_write_args /* {
 2057                 struct vnode *a_vp;
 2058                 struct uio *a_uio;
 2059                 int  a_ioflag;
 2060                 struct ucred *a_cred;
 2061         } */ *ap;
 2062 {
 2063         int error, resid;
 2064         struct inode *ip;
 2065         struct uio *uio;
 2066 
 2067         uio = ap->a_uio;
 2068         resid = uio->uio_resid;
 2069         error = VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap);
 2070         ip = VTOI(ap->a_vp);
 2071         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2072                 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE;
 2073         return (error);
 2074 }
 2075 
 2076 /*
 2077  * Close wrapper for special devices.
 2078  *
 2079  * Update the times on the inode then do device close.
 2080  */
 2081 static int
 2082 ufsspec_close(ap)
 2083         struct vop_close_args /* {
 2084                 struct vnode *a_vp;
 2085                 int  a_fflag;
 2086                 struct ucred *a_cred;
 2087                 struct thread *a_td;
 2088         } */ *ap;
 2089 {
 2090         struct vnode *vp = ap->a_vp;
 2091 
 2092         VI_LOCK(vp);
 2093         if (vp->v_usecount > 1)
 2094                 ufs_itimes(vp);
 2095         VI_UNLOCK(vp);
 2096         return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap));
 2097 }
 2098 
 2099 /*
 2100  * Read wrapper for fifos.
 2101  */
 2102 static int
 2103 ufsfifo_read(ap)
 2104         struct vop_read_args /* {
 2105                 struct vnode *a_vp;
 2106                 struct uio *a_uio;
 2107                 int  a_ioflag;
 2108                 struct ucred *a_cred;
 2109         } */ *ap;
 2110 {
 2111         int error, resid;
 2112         struct inode *ip;
 2113         struct uio *uio;
 2114 
 2115         uio = ap->a_uio;
 2116         resid = uio->uio_resid;
 2117         error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap);
 2118         ip = VTOI(ap->a_vp);
 2119         if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL &&
 2120             (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2121                 ip->i_flag |= IN_ACCESS;
 2122         return (error);
 2123 }
 2124 
 2125 /*
 2126  * Write wrapper for fifos.
 2127  */
 2128 static int
 2129 ufsfifo_write(ap)
 2130         struct vop_write_args /* {
 2131                 struct vnode *a_vp;
 2132                 struct uio *a_uio;
 2133                 int  a_ioflag;
 2134                 struct ucred *a_cred;
 2135         } */ *ap;
 2136 {
 2137         int error, resid;
 2138         struct inode *ip;
 2139         struct uio *uio;
 2140 
 2141         uio = ap->a_uio;
 2142         resid = uio->uio_resid;
 2143         error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap);
 2144         ip = VTOI(ap->a_vp);
 2145         if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0)))
 2146                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 2147         return (error);
 2148 }
 2149 
 2150 /*
 2151  * Close wrapper for fifos.
 2152  *
 2153  * Update the times on the inode then do device close.
 2154  */
 2155 static int
 2156 ufsfifo_close(ap)
 2157         struct vop_close_args /* {
 2158                 struct vnode *a_vp;
 2159                 int  a_fflag;
 2160                 struct ucred *a_cred;
 2161                 struct thread *a_td;
 2162         } */ *ap;
 2163 {
 2164         struct vnode *vp = ap->a_vp;
 2165 
 2166         VI_LOCK(vp);
 2167         if (vp->v_usecount > 1)
 2168                 ufs_itimes(vp);
 2169         VI_UNLOCK(vp);
 2170         return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap));
 2171 }
 2172 
 2173 /*
 2174  * Kqfilter wrapper for fifos.
 2175  *
 2176  * Fall through to ufs kqfilter routines if needed 
 2177  */
 2178 static int
 2179 ufsfifo_kqfilter(ap)
 2180         struct vop_kqfilter_args *ap;
 2181 {
 2182         int error;
 2183 
 2184         error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_kqfilter), ap);
 2185         if (error)
 2186                 error = ufs_kqfilter(ap);
 2187         return (error);
 2188 }
 2189 
 2190 /*
 2191  * Return POSIX pathconf information applicable to ufs filesystems.
 2192  */
 2193 static int
 2194 ufs_pathconf(ap)
 2195         struct vop_pathconf_args /* {
 2196                 struct vnode *a_vp;
 2197                 int a_name;
 2198                 int *a_retval;
 2199         } */ *ap;
 2200 {
 2201         int error;
 2202 
 2203         error = 0;
 2204         switch (ap->a_name) {
 2205         case _PC_LINK_MAX:
 2206                 *ap->a_retval = LINK_MAX;
 2207                 break;
 2208         case _PC_NAME_MAX:
 2209                 *ap->a_retval = NAME_MAX;
 2210                 break;
 2211         case _PC_PATH_MAX:
 2212                 *ap->a_retval = PATH_MAX;
 2213                 break;
 2214         case _PC_PIPE_BUF:
 2215                 *ap->a_retval = PIPE_BUF;
 2216                 break;
 2217         case _PC_CHOWN_RESTRICTED:
 2218                 *ap->a_retval = 1;
 2219                 break;
 2220         case _PC_NO_TRUNC:
 2221                 *ap->a_retval = 1;
 2222                 break;
 2223         case _PC_ACL_EXTENDED:
 2224 #ifdef UFS_ACL
 2225                 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
 2226                         *ap->a_retval = 1;
 2227                 else
 2228                         *ap->a_retval = 0;
 2229 #else
 2230                 *ap->a_retval = 0;
 2231 #endif
 2232                 break;
 2233         case _PC_ACL_PATH_MAX:
 2234 #ifdef UFS_ACL
 2235                 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
 2236                         *ap->a_retval = ACL_MAX_ENTRIES;
 2237                 else
 2238                         *ap->a_retval = 3;
 2239 #else
 2240                 *ap->a_retval = 3;
 2241 #endif
 2242                 break;
 2243         case _PC_MAC_PRESENT:
 2244 #ifdef MAC
 2245                 if (ap->a_vp->v_mount->mnt_flag & MNT_MULTILABEL)
 2246                         *ap->a_retval = 1;
 2247                 else
 2248                         *ap->a_retval = 0;
 2249 #else
 2250                 *ap->a_retval = 0;
 2251 #endif
 2252                 break;
 2253         case _PC_ASYNC_IO:
 2254                 /* _PC_ASYNC_IO should have been handled by upper layers. */
 2255                 KASSERT(0, ("_PC_ASYNC_IO should not get here"));
 2256                 error = EINVAL;
 2257                 break;
 2258         case _PC_PRIO_IO:
 2259                 *ap->a_retval = 0;
 2260                 break;
 2261         case _PC_SYNC_IO:
 2262                 *ap->a_retval = 0;
 2263                 break;
 2264         case _PC_ALLOC_SIZE_MIN:
 2265                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
 2266                 break;
 2267         case _PC_FILESIZEBITS:
 2268                 *ap->a_retval = 64;
 2269                 break;
 2270         case _PC_REC_INCR_XFER_SIZE:
 2271                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2272                 break;
 2273         case _PC_REC_MAX_XFER_SIZE:
 2274                 *ap->a_retval = -1; /* means ``unlimited'' */
 2275                 break;
 2276         case _PC_REC_MIN_XFER_SIZE:
 2277                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2278                 break;
 2279         case _PC_REC_XFER_ALIGN:
 2280                 *ap->a_retval = PAGE_SIZE;
 2281                 break;
 2282         case _PC_SYMLINK_MAX:
 2283                 *ap->a_retval = MAXPATHLEN;
 2284                 break;
 2285 
 2286         default:
 2287                 error = EINVAL;
 2288                 break;
 2289         }
 2290         return (error);
 2291 }
 2292 
 2293 /*
 2294  * Advisory record locking support
 2295  */
 2296 static int
 2297 ufs_advlock(ap)
 2298         struct vop_advlock_args /* {
 2299                 struct vnode *a_vp;
 2300                 caddr_t  a_id;
 2301                 int  a_op;
 2302                 struct flock *a_fl;
 2303                 int  a_flags;
 2304         } */ *ap;
 2305 {
 2306         struct inode *ip = VTOI(ap->a_vp);
 2307 
 2308         return (lf_advlock(ap, &(ip->i_lockf), ip->i_size));
 2309 }
 2310 
 2311 /*
 2312  * Initialize the vnode associated with a new inode, handle aliased
 2313  * vnodes.
 2314  */
 2315 int
 2316 ufs_vinit(mntp, specops, fifoops, vpp)
 2317         struct mount *mntp;
 2318         vop_t **specops;
 2319         vop_t **fifoops;
 2320         struct vnode **vpp;
 2321 {
 2322         struct inode *ip;
 2323         struct vnode *vp;
 2324         struct timeval tv;
 2325 
 2326         vp = *vpp;
 2327         ip = VTOI(vp);
 2328         switch(vp->v_type = IFTOVT(ip->i_mode)) {
 2329         case VCHR:
 2330         case VBLK:
 2331                 vp->v_op = specops;
 2332                 vp = addaliasu(vp, DIP(ip, i_rdev));
 2333                 ip->i_vnode = vp;
 2334                 break;
 2335         case VFIFO:
 2336                 vp->v_op = fifoops;
 2337                 break;
 2338         default:
 2339                 break;
 2340 
 2341         }
 2342         ASSERT_VOP_LOCKED(vp, "ufs_vinit");
 2343         if (ip->i_number == ROOTINO)
 2344                 vp->v_vflag |= VV_ROOT;
 2345         /*
 2346          * Initialize modrev times
 2347          */
 2348         getmicrouptime(&tv);
 2349         SETHIGH(ip->i_modrev, tv.tv_sec);
 2350         SETLOW(ip->i_modrev, tv.tv_usec * 4294);
 2351         *vpp = vp;
 2352         return (0);
 2353 }
 2354 
 2355 /*
 2356  * Allocate a new inode.
 2357  * Vnode dvp must be locked.
 2358  */
 2359 static int
 2360 ufs_makeinode(mode, dvp, vpp, cnp)
 2361         int mode;
 2362         struct vnode *dvp;
 2363         struct vnode **vpp;
 2364         struct componentname *cnp;
 2365 {
 2366         struct inode *ip, *pdir;
 2367         struct direct newdir;
 2368         struct vnode *tvp;
 2369 #ifdef UFS_ACL
 2370         struct acl *acl;
 2371 #endif
 2372         int error;
 2373 
 2374         pdir = VTOI(dvp);
 2375 #ifdef DIAGNOSTIC
 2376         if ((cnp->cn_flags & HASBUF) == 0)
 2377                 panic("ufs_makeinode: no name");
 2378 #endif
 2379         *vpp = NULL;
 2380         if ((mode & IFMT) == 0)
 2381                 mode |= IFREG;
 2382 
 2383         error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
 2384         if (error)
 2385                 return (error);
 2386         ip = VTOI(tvp);
 2387         ip->i_gid = pdir->i_gid;
 2388         DIP_SET(ip, i_gid, pdir->i_gid);
 2389 #ifdef SUIDDIR
 2390         {
 2391 #ifdef QUOTA
 2392                 struct ucred ucred, *ucp;
 2393                 ucp = cnp->cn_cred;
 2394 #endif
 2395                 /*
 2396                  * If we are not the owner of the directory,
 2397                  * and we are hacking owners here, (only do this where told to)
 2398                  * and we are not giving it TO root, (would subvert quotas)
 2399                  * then go ahead and give it to the other user.
 2400                  * Note that this drops off the execute bits for security.
 2401                  */
 2402                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 2403                     (pdir->i_mode & ISUID) &&
 2404                     (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
 2405                         ip->i_uid = pdir->i_uid;
 2406                         DIP_SET(ip, i_uid, ip->i_uid);
 2407                         mode &= ~07111;
 2408 #ifdef QUOTA
 2409                         /*
 2410                          * Make sure the correct user gets charged
 2411                          * for the space.
 2412                          * Quickly knock up a dummy credential for the victim.
 2413                          * XXX This seems to never be accessed out of our
 2414                          * context so a stack variable is ok.
 2415                          */
 2416                         ucred.cr_ref = 1;
 2417                         ucred.cr_uid = ip->i_uid;
 2418                         ucred.cr_ngroups = 1;
 2419                         ucred.cr_groups[0] = pdir->i_gid;
 2420                         ucp = &ucred;
 2421 #endif
 2422                 } else {
 2423                         ip->i_uid = cnp->cn_cred->cr_uid;
 2424                         DIP_SET(ip, i_uid, ip->i_uid);
 2425                 }
 2426 
 2427 #ifdef QUOTA
 2428                 if ((error = getinoquota(ip)) ||
 2429                     (error = chkiq(ip, 1, ucp, 0))) {
 2430                         UFS_VFREE(tvp, ip->i_number, mode);
 2431                         vput(tvp);
 2432                         return (error);
 2433                 }
 2434 #endif
 2435         }
 2436 #else   /* !SUIDDIR */
 2437         ip->i_uid = cnp->cn_cred->cr_uid;
 2438         DIP_SET(ip, i_uid, ip->i_uid);
 2439 #ifdef QUOTA
 2440         if ((error = getinoquota(ip)) ||
 2441             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 2442                 UFS_VFREE(tvp, ip->i_number, mode);
 2443                 vput(tvp);
 2444                 return (error);
 2445         }
 2446 #endif
 2447 #endif  /* !SUIDDIR */
 2448         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 2449 #ifdef UFS_ACL
 2450         acl = NULL;
 2451         if ((dvp->v_mount->mnt_flag & MNT_ACLS) != 0) {
 2452                 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK);
 2453 
 2454                 /*
 2455                  * Retrieve default ACL for parent, if any.
 2456                  */
 2457                 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cnp->cn_cred,
 2458                     cnp->cn_thread);
 2459                 switch (error) {
 2460                 case 0:
 2461                         /*
 2462                          * Retrieved a default ACL, so merge mode and ACL if
 2463                          * necessary.
 2464                          */
 2465                         if (acl->acl_cnt != 0) {
 2466                                 /*
 2467                                  * Two possible ways for default ACL to not
 2468                                  * be present.  First, the EA can be
 2469                                  * undefined, or second, the default ACL can
 2470                                  * be blank.  If it's blank, fall through to
 2471                                  * the it's not defined case.
 2472                                  */
 2473                                 mode = acl_posix1e_newfilemode(mode, acl);
 2474                                 ip->i_mode = mode;
 2475                                 DIP_SET(ip, i_mode, mode);
 2476                                 ufs_sync_acl_from_inode(ip, acl);
 2477                                 break;
 2478                         }
 2479                         /* FALLTHROUGH */
 2480         
 2481                 case EOPNOTSUPP:
 2482                         /*
 2483                          * Just use the mode as-is.
 2484                          */
 2485                         ip->i_mode = mode;
 2486                         DIP_SET(ip, i_mode, mode);
 2487                         FREE(acl, M_ACL);
 2488                         acl = NULL;
 2489                         break;
 2490         
 2491                 default:
 2492                         UFS_VFREE(tvp, ip->i_number, mode);
 2493                         vput(tvp);
 2494                         FREE(acl, M_ACL);
 2495                         acl = NULL;
 2496                         return (error);
 2497                 }
 2498         } else {
 2499 #endif
 2500                 ip->i_mode = mode;
 2501                 DIP_SET(ip, i_mode, mode);
 2502 #ifdef UFS_ACL
 2503         }
 2504 #endif
 2505         tvp->v_type = IFTOVT(mode);     /* Rest init'd in getnewvnode(). */
 2506         ip->i_effnlink = 1;
 2507         ip->i_nlink = 1;
 2508         DIP_SET(ip, i_nlink, 1);
 2509         if (DOINGSOFTDEP(tvp))
 2510                 softdep_change_linkcnt(ip);
 2511         if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
 2512             suser_cred(cnp->cn_cred, SUSER_ALLOWJAIL)) {
 2513                 ip->i_mode &= ~ISGID;
 2514                 DIP_SET(ip, i_mode, ip->i_mode);
 2515         }
 2516 
 2517         if (cnp->cn_flags & ISWHITEOUT) {
 2518                 ip->i_flags |= UF_OPAQUE;
 2519                 DIP_SET(ip, i_flags, ip->i_flags);
 2520         }
 2521 
 2522         /*
 2523          * Make sure inode goes to disk before directory entry.
 2524          */
 2525         error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
 2526         if (error)
 2527                 goto bad;
 2528 #ifdef MAC
 2529         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 2530                 error = mac_create_vnode_extattr(cnp->cn_cred, dvp->v_mount,
 2531                     dvp, tvp, cnp);
 2532                 if (error)
 2533                         goto bad;
 2534         }
 2535 #endif
 2536 #ifdef UFS_ACL
 2537         if (acl != NULL) {
 2538                 /*
 2539                  * XXX: If we abort now, will Soft Updates notify the extattr
 2540                  * code that the EAs for the file need to be released?
 2541                  */
 2542                 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cnp->cn_cred,
 2543                     cnp->cn_thread);
 2544                 switch (error) {
 2545                 case 0:
 2546                         break;
 2547 
 2548                 case EOPNOTSUPP:
 2549                         /*
 2550                          * XXX: This should not happen, as EOPNOTSUPP above was
 2551                          * supposed to free acl.
 2552                          */
 2553                         printf("ufs_makeinode: VOP_GETACL() but no "
 2554                             "VOP_SETACL()\n");
 2555                         /* panic("ufs_makeinode: VOP_GETACL() but no "
 2556                             "VOP_SETACL()"); */
 2557                         break;
 2558 
 2559                 default:
 2560                         FREE(acl, M_ACL);
 2561                         goto bad;
 2562                 }
 2563                 FREE(acl, M_ACL);
 2564         }
 2565 #endif /* !UFS_ACL */
 2566         ufs_makedirentry(ip, cnp, &newdir);
 2567         error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL);
 2568         if (error)
 2569                 goto bad;
 2570         *vpp = tvp;
 2571         return (0);
 2572 
 2573 bad:
 2574         /*
 2575          * Write error occurred trying to update the inode
 2576          * or the directory so must deallocate the inode.
 2577          */
 2578         ip->i_effnlink = 0;
 2579         ip->i_nlink = 0;
 2580         DIP_SET(ip, i_nlink, 0);
 2581         ip->i_flag |= IN_CHANGE;
 2582         if (DOINGSOFTDEP(tvp))
 2583                 softdep_change_linkcnt(ip);
 2584         vput(tvp);
 2585         return (error);
 2586 }
 2587 
 2588 static struct filterops ufsread_filtops = 
 2589         { 1, NULL, filt_ufsdetach, filt_ufsread };
 2590 static struct filterops ufswrite_filtops = 
 2591         { 1, NULL, filt_ufsdetach, filt_ufswrite };
 2592 static struct filterops ufsvnode_filtops = 
 2593         { 1, NULL, filt_ufsdetach, filt_ufsvnode };
 2594 
 2595 static int
 2596 ufs_kqfilter(ap)
 2597         struct vop_kqfilter_args /* {
 2598                 struct vnode *a_vp;
 2599                 struct knote *a_kn;
 2600         } */ *ap;
 2601 {
 2602         struct vnode *vp = ap->a_vp;
 2603         struct knote *kn = ap->a_kn;
 2604 
 2605         switch (kn->kn_filter) {
 2606         case EVFILT_READ:
 2607                 kn->kn_fop = &ufsread_filtops;
 2608                 break;
 2609         case EVFILT_WRITE:
 2610                 kn->kn_fop = &ufswrite_filtops;
 2611                 break;
 2612         case EVFILT_VNODE:
 2613                 kn->kn_fop = &ufsvnode_filtops;
 2614                 break;
 2615         default:
 2616                 return (1);
 2617         }
 2618 
 2619         kn->kn_hook = (caddr_t)vp;
 2620 
 2621         if (vp->v_pollinfo == NULL)
 2622                 v_addpollinfo(vp);
 2623         if (vp->v_pollinfo == NULL)
 2624                 return ENOMEM;
 2625         knlist_add(&vp->v_pollinfo->vpi_selinfo.si_note, kn, 0);
 2626 
 2627         return (0);
 2628 }
 2629 
 2630 static void
 2631 filt_ufsdetach(struct knote *kn)
 2632 {
 2633         struct vnode *vp = (struct vnode *)kn->kn_hook;
 2634 
 2635         KASSERT(vp->v_pollinfo != NULL, ("Mising v_pollinfo"));
 2636         knlist_remove(&vp->v_pollinfo->vpi_selinfo.si_note, kn, 0);
 2637 }
 2638 
 2639 /*ARGSUSED*/
 2640 static int
 2641 filt_ufsread(struct knote *kn, long hint)
 2642 {
 2643         struct vnode *vp = (struct vnode *)kn->kn_hook;
 2644         struct inode *ip = VTOI(vp);
 2645 
 2646         /*
 2647          * filesystem is gone, so set the EOF flag and schedule 
 2648          * the knote for deletion.
 2649          */
 2650         if (hint == NOTE_REVOKE) {
 2651                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
 2652                 return (1);
 2653         }
 2654 
 2655         kn->kn_data = ip->i_size - kn->kn_fp->f_offset;
 2656         return (kn->kn_data != 0);
 2657 }
 2658 
 2659 /*ARGSUSED*/
 2660 static int
 2661 filt_ufswrite(struct knote *kn, long hint)
 2662 {
 2663 
 2664         /*
 2665          * filesystem is gone, so set the EOF flag and schedule 
 2666          * the knote for deletion.
 2667          */
 2668         if (hint == NOTE_REVOKE)
 2669                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
 2670 
 2671         kn->kn_data = 0;
 2672         return (1);
 2673 }
 2674 
 2675 static int
 2676 filt_ufsvnode(struct knote *kn, long hint)
 2677 {
 2678 
 2679         if (kn->kn_sfflags & hint)
 2680                 kn->kn_fflags |= hint;
 2681         if (hint == NOTE_REVOKE) {
 2682                 kn->kn_flags |= EV_EOF;
 2683                 return (1);
 2684         }
 2685         return (kn->kn_fflags != 0);
 2686 }
 2687 
 2688 /* Global vfs data structures for ufs. */
 2689 static vop_t **ufs_vnodeop_p;
 2690 static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = {
 2691         { &vop_default_desc,            (vop_t *) vop_defaultop },
 2692         { &vop_fsync_desc,              (vop_t *) vop_panic },
 2693         { &vop_read_desc,               (vop_t *) vop_panic },
 2694         { &vop_reallocblks_desc,        (vop_t *) vop_panic },
 2695         { &vop_write_desc,              (vop_t *) vop_panic },
 2696         { &vop_access_desc,             (vop_t *) ufs_access },
 2697         { &vop_advlock_desc,            (vop_t *) ufs_advlock },
 2698         { &vop_lock_desc,               (vop_t *) ufs_lock },
 2699         { &vop_bmap_desc,               (vop_t *) ufs_bmap },
 2700         { &vop_cachedlookup_desc,       (vop_t *) ufs_lookup },
 2701         { &vop_close_desc,              (vop_t *) ufs_close },
 2702         { &vop_create_desc,             (vop_t *) ufs_create },
 2703         { &vop_getattr_desc,            (vop_t *) ufs_getattr },
 2704         { &vop_inactive_desc,           (vop_t *) ufs_inactive },
 2705         { &vop_link_desc,               (vop_t *) ufs_link },
 2706         { &vop_lookup_desc,             (vop_t *) vfs_cache_lookup },
 2707         { &vop_mkdir_desc,              (vop_t *) ufs_mkdir },
 2708         { &vop_mknod_desc,              (vop_t *) ufs_mknod },
 2709         { &vop_open_desc,               (vop_t *) ufs_open },
 2710         { &vop_pathconf_desc,           (vop_t *) ufs_pathconf },
 2711         { &vop_poll_desc,               (vop_t *) vop_stdpoll },
 2712         { &vop_kqfilter_desc,           (vop_t *) ufs_kqfilter },
 2713         { &vop_getwritemount_desc,      (vop_t *) vop_stdgetwritemount },
 2714         { &vop_print_desc,              (vop_t *) ufs_print },
 2715         { &vop_readdir_desc,            (vop_t *) ufs_readdir },
 2716         { &vop_readlink_desc,           (vop_t *) ufs_readlink },
 2717         { &vop_reclaim_desc,            (vop_t *) ufs_reclaim },
 2718         { &vop_remove_desc,             (vop_t *) ufs_remove },
 2719         { &vop_rename_desc,             (vop_t *) ufs_rename },
 2720         { &vop_rmdir_desc,              (vop_t *) ufs_rmdir },
 2721         { &vop_setattr_desc,            (vop_t *) ufs_setattr },
 2722 #ifdef MAC
 2723         { &vop_setlabel_desc,           (vop_t *) vop_stdsetlabel_ea },
 2724 #endif
 2725         { &vop_strategy_desc,           (vop_t *) ufs_strategy },
 2726         { &vop_symlink_desc,            (vop_t *) ufs_symlink },
 2727         { &vop_whiteout_desc,           (vop_t *) ufs_whiteout },
 2728 #ifdef UFS_EXTATTR
 2729         { &vop_getextattr_desc,         (vop_t *) ufs_getextattr },
 2730         { &vop_deleteextattr_desc,              (vop_t *) ufs_deleteextattr },
 2731         { &vop_setextattr_desc,         (vop_t *) ufs_setextattr },
 2732 #endif
 2733 #ifdef UFS_ACL
 2734         { &vop_getacl_desc,             (vop_t *) ufs_getacl },
 2735         { &vop_setacl_desc,             (vop_t *) ufs_setacl },
 2736         { &vop_aclcheck_desc,           (vop_t *) ufs_aclcheck },
 2737 #endif
 2738         { NULL, NULL }
 2739 };
 2740 static struct vnodeopv_desc ufs_vnodeop_opv_desc =
 2741         { &ufs_vnodeop_p, ufs_vnodeop_entries };
 2742 
 2743 static vop_t **ufs_specop_p;
 2744 static struct vnodeopv_entry_desc ufs_specop_entries[] = {
 2745         { &vop_default_desc,            (vop_t *) spec_vnoperate },
 2746         { &vop_fsync_desc,              (vop_t *) vop_panic },
 2747         { &vop_access_desc,             (vop_t *) ufs_access },
 2748         { &vop_close_desc,              (vop_t *) ufsspec_close },
 2749         { &vop_getattr_desc,            (vop_t *) ufs_getattr },
 2750         { &vop_inactive_desc,           (vop_t *) ufs_inactive },
 2751         { &vop_print_desc,              (vop_t *) ufs_print },
 2752         { &vop_read_desc,               (vop_t *) ufsspec_read },
 2753         { &vop_reclaim_desc,            (vop_t *) ufs_reclaim },
 2754         { &vop_setattr_desc,            (vop_t *) ufs_setattr },
 2755 #ifdef MAC
 2756         { &vop_setlabel_desc,           (vop_t *) vop_stdsetlabel_ea },
 2757 #endif
 2758         { &vop_write_desc,              (vop_t *) ufsspec_write },
 2759 #ifdef UFS_EXTATTR
 2760         { &vop_getextattr_desc,         (vop_t *) ufs_getextattr },
 2761         { &vop_deleteextattr_desc,              (vop_t *) ufs_deleteextattr },
 2762         { &vop_setextattr_desc,         (vop_t *) ufs_setextattr },
 2763 #endif
 2764 #ifdef UFS_ACL
 2765         { &vop_getacl_desc,             (vop_t *) ufs_getacl },
 2766         { &vop_setacl_desc,             (vop_t *) ufs_setacl },
 2767         { &vop_aclcheck_desc,           (vop_t *) ufs_aclcheck },
 2768 #endif
 2769         {NULL, NULL}
 2770 };
 2771 static struct vnodeopv_desc ufs_specop_opv_desc =
 2772         { &ufs_specop_p, ufs_specop_entries };
 2773 
 2774 static vop_t **ufs_fifoop_p;
 2775 static struct vnodeopv_entry_desc ufs_fifoop_entries[] = {
 2776         { &vop_default_desc,            (vop_t *) fifo_vnoperate },
 2777         { &vop_fsync_desc,              (vop_t *) vop_panic },
 2778         { &vop_access_desc,             (vop_t *) ufs_access },
 2779         { &vop_close_desc,              (vop_t *) ufsfifo_close },
 2780         { &vop_getattr_desc,            (vop_t *) ufs_getattr },
 2781         { &vop_inactive_desc,           (vop_t *) ufs_inactive },
 2782         { &vop_kqfilter_desc,           (vop_t *) ufsfifo_kqfilter },
 2783         { &vop_print_desc,              (vop_t *) ufs_print },
 2784         { &vop_read_desc,               (vop_t *) ufsfifo_read },
 2785         { &vop_reclaim_desc,            (vop_t *) ufs_reclaim },
 2786         { &vop_setattr_desc,            (vop_t *) ufs_setattr },
 2787 #ifdef MAC
 2788         { &vop_setlabel_desc,           (vop_t *) vop_stdsetlabel_ea },
 2789 #endif
 2790         { &vop_write_desc,              (vop_t *) ufsfifo_write },
 2791 #ifdef UFS_EXTATTR
 2792         { &vop_getextattr_desc,         (vop_t *) ufs_getextattr },
 2793         { &vop_deleteextattr_desc,              (vop_t *) ufs_deleteextattr },
 2794         { &vop_setextattr_desc,         (vop_t *) ufs_setextattr },
 2795 #endif
 2796 #ifdef UFS_ACL
 2797         { &vop_getacl_desc,             (vop_t *) ufs_getacl },
 2798         { &vop_setacl_desc,             (vop_t *) ufs_setacl },
 2799         { &vop_aclcheck_desc,           (vop_t *) ufs_aclcheck },
 2800 #endif
 2801         { NULL, NULL }
 2802 };
 2803 static struct vnodeopv_desc ufs_fifoop_opv_desc =
 2804         { &ufs_fifoop_p, ufs_fifoop_entries };
 2805 
 2806 VNODEOP_SET(ufs_vnodeop_opv_desc);
 2807 VNODEOP_SET(ufs_specop_opv_desc);
 2808 VNODEOP_SET(ufs_fifoop_opv_desc);
 2809 
 2810 int
 2811 ufs_vnoperate(ap)
 2812         struct vop_generic_args /* {
 2813                 struct vnodeop_desc *a_desc;
 2814         } */ *ap;
 2815 {
 2816         return (VOCALL(ufs_vnodeop_p, ap->a_desc->vdesc_offset, ap));
 2817 }
 2818 
 2819 int
 2820 ufs_vnoperatefifo(ap)
 2821         struct vop_generic_args /* {
 2822                 struct vnodeop_desc *a_desc;
 2823         } */ *ap;
 2824 {
 2825         return (VOCALL(ufs_fifoop_p, ap->a_desc->vdesc_offset, ap));
 2826 }
 2827 
 2828 int
 2829 ufs_vnoperatespec(ap)
 2830         struct vop_generic_args /* {
 2831                 struct vnodeop_desc *a_desc;
 2832         } */ *ap;
 2833 {
 2834         return (VOCALL(ufs_specop_p, ap->a_desc->vdesc_offset, ap));
 2835 }

Cache object: c95af28a31181eda778beb4ec7042e9f


[ 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.