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/11.1/sys/ufs/ufs/ufs_vnops.c 312072 2017-01-13 13:37:09Z kib $");
   39 
   40 #include "opt_quota.h"
   41 #include "opt_suiddir.h"
   42 #include "opt_ufs.h"
   43 #include "opt_ffs.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/filio.h>
   52 #include <sys/stat.h>
   53 #include <sys/bio.h>
   54 #include <sys/buf.h>
   55 #include <sys/mount.h>
   56 #include <sys/priv.h>
   57 #include <sys/refcount.h>
   58 #include <sys/unistd.h>
   59 #include <sys/vnode.h>
   60 #include <sys/dirent.h>
   61 #include <sys/lockf.h>
   62 #include <sys/conf.h>
   63 #include <sys/acl.h>
   64 
   65 #include <security/mac/mac_framework.h>
   66 
   67 #include <sys/file.h>           /* XXX */
   68 
   69 #include <vm/vm.h>
   70 #include <vm/vm_extern.h>
   71 
   72 #include <ufs/ufs/acl.h>
   73 #include <ufs/ufs/extattr.h>
   74 #include <ufs/ufs/quota.h>
   75 #include <ufs/ufs/inode.h>
   76 #include <ufs/ufs/dir.h>
   77 #include <ufs/ufs/ufsmount.h>
   78 #include <ufs/ufs/ufs_extern.h>
   79 #ifdef UFS_DIRHASH
   80 #include <ufs/ufs/dirhash.h>
   81 #endif
   82 #ifdef UFS_GJOURNAL
   83 #include <ufs/ufs/gjournal.h>
   84 FEATURE(ufs_gjournal, "Journaling support through GEOM for UFS");
   85 #endif
   86 
   87 #ifdef QUOTA
   88 FEATURE(ufs_quota, "UFS disk quotas support");
   89 FEATURE(ufs_quota64, "64bit UFS disk quotas support");
   90 #endif
   91 
   92 #ifdef SUIDDIR
   93 FEATURE(suiddir,
   94     "Give all new files in directory the same ownership as the directory");
   95 #endif
   96 
   97 
   98 #include <ufs/ffs/ffs_extern.h>
   99 
  100 static vop_accessx_t    ufs_accessx;
  101 static int ufs_chmod(struct vnode *, int, struct ucred *, struct thread *);
  102 static int ufs_chown(struct vnode *, uid_t, gid_t, struct ucred *, struct thread *);
  103 static vop_close_t      ufs_close;
  104 static vop_create_t     ufs_create;
  105 static vop_getattr_t    ufs_getattr;
  106 static vop_ioctl_t      ufs_ioctl;
  107 static vop_link_t       ufs_link;
  108 static int ufs_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *, const char *);
  109 static vop_markatime_t  ufs_markatime;
  110 static vop_mkdir_t      ufs_mkdir;
  111 static vop_mknod_t      ufs_mknod;
  112 static vop_open_t       ufs_open;
  113 static vop_pathconf_t   ufs_pathconf;
  114 static vop_print_t      ufs_print;
  115 static vop_readlink_t   ufs_readlink;
  116 static vop_remove_t     ufs_remove;
  117 static vop_rename_t     ufs_rename;
  118 static vop_rmdir_t      ufs_rmdir;
  119 static vop_setattr_t    ufs_setattr;
  120 static vop_strategy_t   ufs_strategy;
  121 static vop_symlink_t    ufs_symlink;
  122 static vop_whiteout_t   ufs_whiteout;
  123 static vop_close_t      ufsfifo_close;
  124 static vop_kqfilter_t   ufsfifo_kqfilter;
  125 static vop_pathconf_t   ufsfifo_pathconf;
  126 
  127 SYSCTL_NODE(_vfs, OID_AUTO, ufs, CTLFLAG_RD, 0, "UFS filesystem");
  128 
  129 /*
  130  * A virgin directory (no blushing please).
  131  */
  132 static struct dirtemplate mastertemplate = {
  133         0, 12, DT_DIR, 1, ".",
  134         0, DIRBLKSIZ - 12, DT_DIR, 2, ".."
  135 };
  136 static struct odirtemplate omastertemplate = {
  137         0, 12, 1, ".",
  138         0, DIRBLKSIZ - 12, 2, ".."
  139 };
  140 
  141 static void
  142 ufs_itimes_locked(struct vnode *vp)
  143 {
  144         struct inode *ip;
  145         struct timespec ts;
  146 
  147         ASSERT_VI_LOCKED(vp, __func__);
  148 
  149         ip = VTOI(vp);
  150         if (UFS_RDONLY(ip))
  151                 goto out;
  152         if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
  153                 return;
  154 
  155         if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp))
  156                 ip->i_flag |= IN_LAZYMOD;
  157         else if (((vp->v_mount->mnt_kern_flag &
  158                     (MNTK_SUSPENDED | MNTK_SUSPEND)) == 0) ||
  159                     (ip->i_flag & (IN_CHANGE | IN_UPDATE)))
  160                 ip->i_flag |= IN_MODIFIED;
  161         else if (ip->i_flag & IN_ACCESS)
  162                 ip->i_flag |= IN_LAZYACCESS;
  163         vfs_timestamp(&ts);
  164         if (ip->i_flag & IN_ACCESS) {
  165                 DIP_SET(ip, i_atime, ts.tv_sec);
  166                 DIP_SET(ip, i_atimensec, ts.tv_nsec);
  167         }
  168         if (ip->i_flag & IN_UPDATE) {
  169                 DIP_SET(ip, i_mtime, ts.tv_sec);
  170                 DIP_SET(ip, i_mtimensec, ts.tv_nsec);
  171         }
  172         if (ip->i_flag & IN_CHANGE) {
  173                 DIP_SET(ip, i_ctime, ts.tv_sec);
  174                 DIP_SET(ip, i_ctimensec, ts.tv_nsec);
  175                 DIP_SET(ip, i_modrev, DIP(ip, i_modrev) + 1);
  176         }
  177 
  178  out:
  179         ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
  180 }
  181 
  182 void
  183 ufs_itimes(struct vnode *vp)
  184 {
  185 
  186         VI_LOCK(vp);
  187         ufs_itimes_locked(vp);
  188         VI_UNLOCK(vp);
  189 }
  190 
  191 /*
  192  * Create a regular file
  193  */
  194 static int
  195 ufs_create(ap)
  196         struct vop_create_args /* {
  197                 struct vnode *a_dvp;
  198                 struct vnode **a_vpp;
  199                 struct componentname *a_cnp;
  200                 struct vattr *a_vap;
  201         } */ *ap;
  202 {
  203         int error;
  204 
  205         error =
  206             ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode),
  207             ap->a_dvp, ap->a_vpp, ap->a_cnp, "ufs_create");
  208         if (error != 0)
  209                 return (error);
  210         if ((ap->a_cnp->cn_flags & MAKEENTRY) != 0)
  211                 cache_enter(ap->a_dvp, *ap->a_vpp, ap->a_cnp);
  212         return (0);
  213 }
  214 
  215 /*
  216  * Mknod vnode call
  217  */
  218 /* ARGSUSED */
  219 static int
  220 ufs_mknod(ap)
  221         struct vop_mknod_args /* {
  222                 struct vnode *a_dvp;
  223                 struct vnode **a_vpp;
  224                 struct componentname *a_cnp;
  225                 struct vattr *a_vap;
  226         } */ *ap;
  227 {
  228         struct vattr *vap = ap->a_vap;
  229         struct vnode **vpp = ap->a_vpp;
  230         struct inode *ip;
  231         ino_t ino;
  232         int error;
  233 
  234         error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode),
  235             ap->a_dvp, vpp, ap->a_cnp, "ufs_mknod");
  236         if (error)
  237                 return (error);
  238         ip = VTOI(*vpp);
  239         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
  240         if (vap->va_rdev != VNOVAL) {
  241                 /*
  242                  * Want to be able to use this to make badblock
  243                  * inodes, so don't truncate the dev number.
  244                  */
  245                 DIP_SET(ip, i_rdev, vap->va_rdev);
  246         }
  247         /*
  248          * Remove inode, then reload it through VFS_VGET so it is
  249          * checked to see if it is an alias of an existing entry in
  250          * the inode cache.  XXX I don't believe this is necessary now.
  251          */
  252         (*vpp)->v_type = VNON;
  253         ino = ip->i_number;     /* Save this before vgone() invalidates ip. */
  254         vgone(*vpp);
  255         vput(*vpp);
  256         error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp);
  257         if (error) {
  258                 *vpp = NULL;
  259                 return (error);
  260         }
  261         return (0);
  262 }
  263 
  264 /*
  265  * Open called.
  266  */
  267 /* ARGSUSED */
  268 static int
  269 ufs_open(struct vop_open_args *ap)
  270 {
  271         struct vnode *vp = ap->a_vp;
  272         struct inode *ip;
  273 
  274         if (vp->v_type == VCHR || vp->v_type == VBLK)
  275                 return (EOPNOTSUPP);
  276 
  277         ip = VTOI(vp);
  278         /*
  279          * Files marked append-only must be opened for appending.
  280          */
  281         if ((ip->i_flags & APPEND) &&
  282             (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
  283                 return (EPERM);
  284         vnode_create_vobject(vp, DIP(ip, i_size), ap->a_td);
  285         return (0);
  286 }
  287 
  288 /*
  289  * Close called.
  290  *
  291  * Update the times on the inode.
  292  */
  293 /* ARGSUSED */
  294 static int
  295 ufs_close(ap)
  296         struct vop_close_args /* {
  297                 struct vnode *a_vp;
  298                 int  a_fflag;
  299                 struct ucred *a_cred;
  300                 struct thread *a_td;
  301         } */ *ap;
  302 {
  303         struct vnode *vp = ap->a_vp;
  304         int usecount;
  305 
  306         VI_LOCK(vp);
  307         usecount = vp->v_usecount;
  308         if (usecount > 1)
  309                 ufs_itimes_locked(vp);
  310         VI_UNLOCK(vp);
  311         return (0);
  312 }
  313 
  314 static int
  315 ufs_accessx(ap)
  316         struct vop_accessx_args /* {
  317                 struct vnode *a_vp;
  318                 accmode_t a_accmode;
  319                 struct ucred *a_cred;
  320                 struct thread *a_td;
  321         } */ *ap;
  322 {
  323         struct vnode *vp = ap->a_vp;
  324         struct inode *ip = VTOI(vp);
  325         accmode_t accmode = ap->a_accmode;
  326         int error;
  327 #ifdef QUOTA
  328         int relocked;
  329 #endif
  330 #ifdef UFS_ACL
  331         struct acl *acl;
  332         acl_type_t type;
  333 #endif
  334 
  335         /*
  336          * Disallow write attempts on read-only filesystems;
  337          * unless the file is a socket, fifo, or a block or
  338          * character device resident on the filesystem.
  339          */
  340         if (accmode & VMODIFY_PERMS) {
  341                 switch (vp->v_type) {
  342                 case VDIR:
  343                 case VLNK:
  344                 case VREG:
  345                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  346                                 return (EROFS);
  347 #ifdef QUOTA
  348                         /*
  349                          * Inode is accounted in the quotas only if struct
  350                          * dquot is attached to it. VOP_ACCESS() is called
  351                          * from vn_open_cred() and provides a convenient
  352                          * point to call getinoquota().
  353                          */
  354                         if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) {
  355 
  356                                 /*
  357                                  * Upgrade vnode lock, since getinoquota()
  358                                  * requires exclusive lock to modify inode.
  359                                  */
  360                                 relocked = 1;
  361                                 vhold(vp);
  362                                 vn_lock(vp, LK_UPGRADE | LK_RETRY);
  363                                 VI_LOCK(vp);
  364                                 if (vp->v_iflag & VI_DOOMED) {
  365                                         vdropl(vp);
  366                                         error = ENOENT;
  367                                         goto relock;
  368                                 }
  369                                 vdropl(vp);
  370                         } else
  371                                 relocked = 0;
  372                         error = getinoquota(ip);
  373 relock:
  374                         if (relocked)
  375                                 vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
  376                         if (error != 0)
  377                                 return (error);
  378 #endif
  379                         break;
  380                 default:
  381                         break;
  382                 }
  383         }
  384 
  385         /*
  386          * If immutable bit set, nobody gets to write it.  "& ~VADMIN_PERMS"
  387          * permits the owner of the file to remove the IMMUTABLE flag.
  388          */
  389         if ((accmode & (VMODIFY_PERMS & ~VADMIN_PERMS)) &&
  390             (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT)))
  391                 return (EPERM);
  392 
  393 #ifdef UFS_ACL
  394         if ((vp->v_mount->mnt_flag & (MNT_ACLS | MNT_NFS4ACLS)) != 0) {
  395                 if (vp->v_mount->mnt_flag & MNT_NFS4ACLS)
  396                         type = ACL_TYPE_NFS4;
  397                 else
  398                         type = ACL_TYPE_ACCESS;
  399 
  400                 acl = acl_alloc(M_WAITOK);
  401                 if (type == ACL_TYPE_NFS4)
  402                         error = ufs_getacl_nfs4_internal(vp, acl, ap->a_td);
  403                 else
  404                         error = VOP_GETACL(vp, type, acl, ap->a_cred, ap->a_td);
  405                 switch (error) {
  406                 case 0:
  407                         if (type == ACL_TYPE_NFS4) {
  408                                 error = vaccess_acl_nfs4(vp->v_type, ip->i_uid,
  409                                     ip->i_gid, acl, accmode, ap->a_cred, NULL);
  410                         } else {
  411                                 error = vfs_unixify_accmode(&accmode);
  412                                 if (error == 0)
  413                                         error = vaccess_acl_posix1e(vp->v_type, ip->i_uid,
  414                                             ip->i_gid, acl, accmode, ap->a_cred, NULL);
  415                         }
  416                         break;
  417                 default:
  418                         if (error != EOPNOTSUPP)
  419                                 printf(
  420 "ufs_accessx(): Error retrieving ACL on object (%d).\n",
  421                                     error);
  422                         /*
  423                          * XXX: Fall back until debugged.  Should
  424                          * eventually possibly log an error, and return
  425                          * EPERM for safety.
  426                          */
  427                         error = vfs_unixify_accmode(&accmode);
  428                         if (error == 0)
  429                                 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid,
  430                                     ip->i_gid, accmode, ap->a_cred, NULL);
  431                 }
  432                 acl_free(acl);
  433 
  434                 return (error);
  435         }
  436 #endif /* !UFS_ACL */
  437         error = vfs_unixify_accmode(&accmode);
  438         if (error == 0)
  439                 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid,
  440                     accmode, ap->a_cred, NULL);
  441         return (error);
  442 }
  443 
  444 /* ARGSUSED */
  445 static int
  446 ufs_getattr(ap)
  447         struct vop_getattr_args /* {
  448                 struct vnode *a_vp;
  449                 struct vattr *a_vap;
  450                 struct ucred *a_cred;
  451         } */ *ap;
  452 {
  453         struct vnode *vp = ap->a_vp;
  454         struct inode *ip = VTOI(vp);
  455         struct vattr *vap = ap->a_vap;
  456 
  457         VI_LOCK(vp);
  458         ufs_itimes_locked(vp);
  459         if (I_IS_UFS1(ip)) {
  460                 vap->va_atime.tv_sec = ip->i_din1->di_atime;
  461                 vap->va_atime.tv_nsec = ip->i_din1->di_atimensec;
  462         } else {
  463                 vap->va_atime.tv_sec = ip->i_din2->di_atime;
  464                 vap->va_atime.tv_nsec = ip->i_din2->di_atimensec;
  465         }
  466         VI_UNLOCK(vp);
  467         /*
  468          * Copy from inode table
  469          */
  470         vap->va_fsid = dev2udev(ITOUMP(ip)->um_dev);
  471         vap->va_fileid = ip->i_number;
  472         vap->va_mode = ip->i_mode & ~IFMT;
  473         vap->va_nlink = ip->i_effnlink;
  474         vap->va_uid = ip->i_uid;
  475         vap->va_gid = ip->i_gid;
  476         if (I_IS_UFS1(ip)) {
  477                 vap->va_rdev = ip->i_din1->di_rdev;
  478                 vap->va_size = ip->i_din1->di_size;
  479                 vap->va_mtime.tv_sec = ip->i_din1->di_mtime;
  480                 vap->va_mtime.tv_nsec = ip->i_din1->di_mtimensec;
  481                 vap->va_ctime.tv_sec = ip->i_din1->di_ctime;
  482                 vap->va_ctime.tv_nsec = ip->i_din1->di_ctimensec;
  483                 vap->va_bytes = dbtob((u_quad_t)ip->i_din1->di_blocks);
  484                 vap->va_filerev = ip->i_din1->di_modrev;
  485         } else {
  486                 vap->va_rdev = ip->i_din2->di_rdev;
  487                 vap->va_size = ip->i_din2->di_size;
  488                 vap->va_mtime.tv_sec = ip->i_din2->di_mtime;
  489                 vap->va_mtime.tv_nsec = ip->i_din2->di_mtimensec;
  490                 vap->va_ctime.tv_sec = ip->i_din2->di_ctime;
  491                 vap->va_ctime.tv_nsec = ip->i_din2->di_ctimensec;
  492                 vap->va_birthtime.tv_sec = ip->i_din2->di_birthtime;
  493                 vap->va_birthtime.tv_nsec = ip->i_din2->di_birthnsec;
  494                 vap->va_bytes = dbtob((u_quad_t)ip->i_din2->di_blocks);
  495                 vap->va_filerev = ip->i_din2->di_modrev;
  496         }
  497         vap->va_flags = ip->i_flags;
  498         vap->va_gen = ip->i_gen;
  499         vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize;
  500         vap->va_type = IFTOVT(ip->i_mode);
  501         return (0);
  502 }
  503 
  504 /*
  505  * Set attribute vnode op. called from several syscalls
  506  */
  507 static int
  508 ufs_setattr(ap)
  509         struct vop_setattr_args /* {
  510                 struct vnode *a_vp;
  511                 struct vattr *a_vap;
  512                 struct ucred *a_cred;
  513         } */ *ap;
  514 {
  515         struct vattr *vap = ap->a_vap;
  516         struct vnode *vp = ap->a_vp;
  517         struct inode *ip = VTOI(vp);
  518         struct ucred *cred = ap->a_cred;
  519         struct thread *td = curthread;
  520         int error;
  521 
  522         /*
  523          * Check for unsettable attributes.
  524          */
  525         if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
  526             (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
  527             (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
  528             ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
  529                 return (EINVAL);
  530         }
  531         if (vap->va_flags != VNOVAL) {
  532                 if ((vap->va_flags & ~(SF_APPEND | SF_ARCHIVED | SF_IMMUTABLE |
  533                     SF_NOUNLINK | SF_SNAPSHOT | UF_APPEND | UF_ARCHIVE |
  534                     UF_HIDDEN | UF_IMMUTABLE | UF_NODUMP | UF_NOUNLINK |
  535                     UF_OFFLINE | UF_OPAQUE | UF_READONLY | UF_REPARSE |
  536                     UF_SPARSE | UF_SYSTEM)) != 0)
  537                         return (EOPNOTSUPP);
  538                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  539                         return (EROFS);
  540                 /*
  541                  * Callers may only modify the file flags on objects they
  542                  * have VADMIN rights for.
  543                  */
  544                 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
  545                         return (error);
  546                 /*
  547                  * Unprivileged processes are not permitted to unset system
  548                  * flags, or modify flags if any system flags are set.
  549                  * Privileged non-jail processes may not modify system flags
  550                  * if securelevel > 0 and any existing system flags are set.
  551                  * Privileged jail processes behave like privileged non-jail
  552                  * processes if the security.jail.chflags_allowed sysctl is
  553                  * is non-zero; otherwise, they behave like unprivileged
  554                  * processes.
  555                  */
  556                 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
  557                         if (ip->i_flags &
  558                             (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
  559                                 error = securelevel_gt(cred, 0);
  560                                 if (error)
  561                                         return (error);
  562                         }
  563                         /* The snapshot flag cannot be toggled. */
  564                         if ((vap->va_flags ^ ip->i_flags) & SF_SNAPSHOT)
  565                                 return (EPERM);
  566                 } else {
  567                         if (ip->i_flags &
  568                             (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
  569                             ((vap->va_flags ^ ip->i_flags) & SF_SETTABLE))
  570                                 return (EPERM);
  571                 }
  572                 ip->i_flags = vap->va_flags;
  573                 DIP_SET(ip, i_flags, vap->va_flags);
  574                 ip->i_flag |= IN_CHANGE;
  575                 error = UFS_UPDATE(vp, 0);
  576                 if (ip->i_flags & (IMMUTABLE | APPEND))
  577                         return (error);
  578         }
  579         /*
  580          * If immutable or append, no one can change any of its attributes
  581          * except the ones already handled (in some cases, file flags
  582          * including the immutability flags themselves for the superuser).
  583          */
  584         if (ip->i_flags & (IMMUTABLE | APPEND))
  585                 return (EPERM);
  586         /*
  587          * Go through the fields and update iff not VNOVAL.
  588          */
  589         if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
  590                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  591                         return (EROFS);
  592                 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred,
  593                     td)) != 0)
  594                         return (error);
  595         }
  596         if (vap->va_size != VNOVAL) {
  597                 /*
  598                  * XXX most of the following special cases should be in
  599                  * callers instead of in N filesystems.  The VDIR check
  600                  * mostly already is.
  601                  */
  602                 switch (vp->v_type) {
  603                 case VDIR:
  604                         return (EISDIR);
  605                 case VLNK:
  606                 case VREG:
  607                         /*
  608                          * Truncation should have an effect in these cases.
  609                          * Disallow it if the filesystem is read-only or
  610                          * the file is being snapshotted.
  611                          */
  612                         if (vp->v_mount->mnt_flag & MNT_RDONLY)
  613                                 return (EROFS);
  614                         if ((ip->i_flags & SF_SNAPSHOT) != 0)
  615                                 return (EPERM);
  616                         break;
  617                 default:
  618                         /*
  619                          * According to POSIX, the result is unspecified
  620                          * for file types other than regular files,
  621                          * directories and shared memory objects.  We
  622                          * don't support shared memory objects in the file
  623                          * system, and have dubious support for truncating
  624                          * symlinks.  Just ignore the request in other cases.
  625                          */
  626                         return (0);
  627                 }
  628                 if ((error = UFS_TRUNCATE(vp, vap->va_size, IO_NORMAL |
  629                     ((vap->va_vaflags & VA_SYNC) != 0 ? IO_SYNC : 0),
  630                     cred)) != 0)
  631                         return (error);
  632         }
  633         if (vap->va_atime.tv_sec != VNOVAL ||
  634             vap->va_mtime.tv_sec != VNOVAL ||
  635             vap->va_birthtime.tv_sec != VNOVAL) {
  636                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  637                         return (EROFS);
  638                 if ((ip->i_flags & SF_SNAPSHOT) != 0)
  639                         return (EPERM);
  640                 error = vn_utimes_perm(vp, vap, cred, td);
  641                 if (error != 0)
  642                         return (error);
  643                 ip->i_flag |= IN_CHANGE | IN_MODIFIED;
  644                 if (vap->va_atime.tv_sec != VNOVAL) {
  645                         ip->i_flag &= ~IN_ACCESS;
  646                         DIP_SET(ip, i_atime, vap->va_atime.tv_sec);
  647                         DIP_SET(ip, i_atimensec, vap->va_atime.tv_nsec);
  648                 }
  649                 if (vap->va_mtime.tv_sec != VNOVAL) {
  650                         ip->i_flag &= ~IN_UPDATE;
  651                         DIP_SET(ip, i_mtime, vap->va_mtime.tv_sec);
  652                         DIP_SET(ip, i_mtimensec, vap->va_mtime.tv_nsec);
  653                 }
  654                 if (vap->va_birthtime.tv_sec != VNOVAL && I_IS_UFS2(ip)) {
  655                         ip->i_din2->di_birthtime = vap->va_birthtime.tv_sec;
  656                         ip->i_din2->di_birthnsec = vap->va_birthtime.tv_nsec;
  657                 }
  658                 error = UFS_UPDATE(vp, 0);
  659                 if (error)
  660                         return (error);
  661         }
  662         error = 0;
  663         if (vap->va_mode != (mode_t)VNOVAL) {
  664                 if (vp->v_mount->mnt_flag & MNT_RDONLY)
  665                         return (EROFS);
  666                 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode &
  667                    (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH)))
  668                         return (EPERM);
  669                 error = ufs_chmod(vp, (int)vap->va_mode, cred, td);
  670         }
  671         return (error);
  672 }
  673 
  674 #ifdef UFS_ACL
  675 static int
  676 ufs_update_nfs4_acl_after_mode_change(struct vnode *vp, int mode,
  677     int file_owner_id, struct ucred *cred, struct thread *td)
  678 {
  679         int error;
  680         struct acl *aclp;
  681 
  682         aclp = acl_alloc(M_WAITOK);
  683         error = ufs_getacl_nfs4_internal(vp, aclp, td);
  684         /*
  685          * We don't have to handle EOPNOTSUPP here, as the filesystem claims
  686          * it supports ACLs.
  687          */
  688         if (error)
  689                 goto out;
  690 
  691         acl_nfs4_sync_acl_from_mode(aclp, mode, file_owner_id);
  692         error = ufs_setacl_nfs4_internal(vp, aclp, td);
  693 
  694 out:
  695         acl_free(aclp);
  696         return (error);
  697 }
  698 #endif /* UFS_ACL */
  699 
  700 /*
  701  * Mark this file's access time for update for vfs_mark_atime().  This
  702  * is called from execve() and mmap().
  703  */
  704 static int
  705 ufs_markatime(ap)
  706         struct vop_markatime_args /* {
  707                 struct vnode *a_vp;
  708         } */ *ap;
  709 {
  710         struct vnode *vp = ap->a_vp;
  711         struct inode *ip = VTOI(vp);
  712 
  713         VI_LOCK(vp);
  714         ip->i_flag |= IN_ACCESS;
  715         VI_UNLOCK(vp);
  716         /*
  717          * XXXKIB No UFS_UPDATE(ap->a_vp, 0) there.
  718          */
  719         return (0);
  720 }
  721 
  722 /*
  723  * Change the mode on a file.
  724  * Inode must be locked before calling.
  725  */
  726 static int
  727 ufs_chmod(vp, mode, cred, td)
  728         struct vnode *vp;
  729         int mode;
  730         struct ucred *cred;
  731         struct thread *td;
  732 {
  733         struct inode *ip = VTOI(vp);
  734         int error;
  735 
  736         /*
  737          * To modify the permissions on a file, must possess VADMIN
  738          * for that file.
  739          */
  740         if ((error = VOP_ACCESSX(vp, VWRITE_ACL, cred, td)))
  741                 return (error);
  742         /*
  743          * Privileged processes may set the sticky bit on non-directories,
  744          * as well as set the setgid bit on a file with a group that the
  745          * process is not a member of.  Both of these are allowed in
  746          * jail(8).
  747          */
  748         if (vp->v_type != VDIR && (mode & S_ISTXT)) {
  749                 if (priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0))
  750                         return (EFTYPE);
  751         }
  752         if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) {
  753                 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
  754                 if (error)
  755                         return (error);
  756         }
  757 
  758         /*
  759          * Deny setting setuid if we are not the file owner.
  760          */
  761         if ((mode & ISUID) && ip->i_uid != cred->cr_uid) {
  762                 error = priv_check_cred(cred, PRIV_VFS_ADMIN, 0);
  763                 if (error)
  764                         return (error);
  765         }
  766 
  767         ip->i_mode &= ~ALLPERMS;
  768         ip->i_mode |= (mode & ALLPERMS);
  769         DIP_SET(ip, i_mode, ip->i_mode);
  770         ip->i_flag |= IN_CHANGE;
  771 #ifdef UFS_ACL
  772         if ((vp->v_mount->mnt_flag & MNT_NFS4ACLS) != 0)
  773                 error = ufs_update_nfs4_acl_after_mode_change(vp, mode, ip->i_uid, cred, td);
  774 #endif
  775         if (error == 0 && (ip->i_flag & IN_CHANGE) != 0)
  776                 error = UFS_UPDATE(vp, 0);
  777 
  778         return (error);
  779 }
  780 
  781 /*
  782  * Perform chown operation on inode ip;
  783  * inode must be locked prior to call.
  784  */
  785 static int
  786 ufs_chown(vp, uid, gid, cred, td)
  787         struct vnode *vp;
  788         uid_t uid;
  789         gid_t gid;
  790         struct ucred *cred;
  791         struct thread *td;
  792 {
  793         struct inode *ip = VTOI(vp);
  794         uid_t ouid;
  795         gid_t ogid;
  796         int error = 0;
  797 #ifdef QUOTA
  798         int i;
  799         ufs2_daddr_t change;
  800 #endif
  801 
  802         if (uid == (uid_t)VNOVAL)
  803                 uid = ip->i_uid;
  804         if (gid == (gid_t)VNOVAL)
  805                 gid = ip->i_gid;
  806         /*
  807          * To modify the ownership of a file, must possess VADMIN for that
  808          * file.
  809          */
  810         if ((error = VOP_ACCESSX(vp, VWRITE_OWNER, cred, td)))
  811                 return (error);
  812         /*
  813          * To change the owner of a file, or change the group of a file to a
  814          * group of which we are not a member, the caller must have
  815          * privilege.
  816          */
  817         if (((uid != ip->i_uid && uid != cred->cr_uid) || 
  818             (gid != ip->i_gid && !groupmember(gid, cred))) &&
  819             (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0)))
  820                 return (error);
  821         ogid = ip->i_gid;
  822         ouid = ip->i_uid;
  823 #ifdef QUOTA
  824         if ((error = getinoquota(ip)) != 0)
  825                 return (error);
  826         if (ouid == uid) {
  827                 dqrele(vp, ip->i_dquot[USRQUOTA]);
  828                 ip->i_dquot[USRQUOTA] = NODQUOT;
  829         }
  830         if (ogid == gid) {
  831                 dqrele(vp, ip->i_dquot[GRPQUOTA]);
  832                 ip->i_dquot[GRPQUOTA] = NODQUOT;
  833         }
  834         change = DIP(ip, i_blocks);
  835         (void) chkdq(ip, -change, cred, CHOWN);
  836         (void) chkiq(ip, -1, cred, CHOWN);
  837         for (i = 0; i < MAXQUOTAS; i++) {
  838                 dqrele(vp, ip->i_dquot[i]);
  839                 ip->i_dquot[i] = NODQUOT;
  840         }
  841 #endif
  842         ip->i_gid = gid;
  843         DIP_SET(ip, i_gid, gid);
  844         ip->i_uid = uid;
  845         DIP_SET(ip, i_uid, uid);
  846 #ifdef QUOTA
  847         if ((error = getinoquota(ip)) == 0) {
  848                 if (ouid == uid) {
  849                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  850                         ip->i_dquot[USRQUOTA] = NODQUOT;
  851                 }
  852                 if (ogid == gid) {
  853                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  854                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  855                 }
  856                 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) {
  857                         if ((error = chkiq(ip, 1, cred, CHOWN)) == 0)
  858                                 goto good;
  859                         else
  860                                 (void) chkdq(ip, -change, cred, CHOWN|FORCE);
  861                 }
  862                 for (i = 0; i < MAXQUOTAS; i++) {
  863                         dqrele(vp, ip->i_dquot[i]);
  864                         ip->i_dquot[i] = NODQUOT;
  865                 }
  866         }
  867         ip->i_gid = ogid;
  868         DIP_SET(ip, i_gid, ogid);
  869         ip->i_uid = ouid;
  870         DIP_SET(ip, i_uid, ouid);
  871         if (getinoquota(ip) == 0) {
  872                 if (ouid == uid) {
  873                         dqrele(vp, ip->i_dquot[USRQUOTA]);
  874                         ip->i_dquot[USRQUOTA] = NODQUOT;
  875                 }
  876                 if (ogid == gid) {
  877                         dqrele(vp, ip->i_dquot[GRPQUOTA]);
  878                         ip->i_dquot[GRPQUOTA] = NODQUOT;
  879                 }
  880                 (void) chkdq(ip, change, cred, FORCE|CHOWN);
  881                 (void) chkiq(ip, 1, cred, FORCE|CHOWN);
  882                 (void) getinoquota(ip);
  883         }
  884         return (error);
  885 good:
  886         if (getinoquota(ip))
  887                 panic("ufs_chown: lost quota");
  888 #endif /* QUOTA */
  889         ip->i_flag |= IN_CHANGE;
  890         if ((ip->i_mode & (ISUID | ISGID)) && (ouid != uid || ogid != gid)) {
  891                 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0)) {
  892                         ip->i_mode &= ~(ISUID | ISGID);
  893                         DIP_SET(ip, i_mode, ip->i_mode);
  894                 }
  895         }
  896         error = UFS_UPDATE(vp, 0);
  897         return (error);
  898 }
  899 
  900 static int
  901 ufs_remove(ap)
  902         struct vop_remove_args /* {
  903                 struct vnode *a_dvp;
  904                 struct vnode *a_vp;
  905                 struct componentname *a_cnp;
  906         } */ *ap;
  907 {
  908         struct inode *ip;
  909         struct vnode *vp = ap->a_vp;
  910         struct vnode *dvp = ap->a_dvp;
  911         int error;
  912         struct thread *td;
  913 
  914         td = curthread;
  915         ip = VTOI(vp);
  916         if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
  917             (VTOI(dvp)->i_flags & APPEND)) {
  918                 error = EPERM;
  919                 goto out;
  920         }
  921 #ifdef UFS_GJOURNAL
  922         ufs_gjournal_orphan(vp);
  923 #endif
  924         error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0);
  925         if (ip->i_nlink <= 0)
  926                 vp->v_vflag |= VV_NOSYNC;
  927         if ((ip->i_flags & SF_SNAPSHOT) != 0) {
  928                 /*
  929                  * Avoid deadlock where another thread is trying to
  930                  * update the inodeblock for dvp and is waiting on
  931                  * snaplk.  Temporary unlock the vnode lock for the
  932                  * unlinked file and sync the directory.  This should
  933                  * allow vput() of the directory to not block later on
  934                  * while holding the snapshot vnode locked, assuming
  935                  * that the directory hasn't been unlinked too.
  936                  */
  937                 VOP_UNLOCK(vp, 0);
  938                 (void) VOP_FSYNC(dvp, MNT_WAIT, td);
  939                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  940         }
  941 out:
  942         return (error);
  943 }
  944 
  945 static void
  946 print_bad_link_count(const char *funcname, struct vnode *dvp)
  947 {
  948         struct inode *dip;
  949 
  950         dip = VTOI(dvp);
  951         uprintf("%s: Bad link count %d on parent inode %jd in file system %s\n",
  952             funcname, dip->i_effnlink, (intmax_t)dip->i_number,
  953             dvp->v_mount->mnt_stat.f_mntonname);
  954 }
  955 
  956 /*
  957  * link vnode call
  958  */
  959 static int
  960 ufs_link(ap)
  961         struct vop_link_args /* {
  962                 struct vnode *a_tdvp;
  963                 struct vnode *a_vp;
  964                 struct componentname *a_cnp;
  965         } */ *ap;
  966 {
  967         struct vnode *vp = ap->a_vp;
  968         struct vnode *tdvp = ap->a_tdvp;
  969         struct componentname *cnp = ap->a_cnp;
  970         struct inode *ip;
  971         struct direct newdir;
  972         int error;
  973 
  974 #ifdef INVARIANTS
  975         if ((cnp->cn_flags & HASBUF) == 0)
  976                 panic("ufs_link: no name");
  977 #endif
  978         if (VTOI(tdvp)->i_effnlink < 2) {
  979                 print_bad_link_count("ufs_link", tdvp);
  980                 error = EINVAL;
  981                 goto out;
  982         }
  983         ip = VTOI(vp);
  984         if ((nlink_t)ip->i_nlink >= LINK_MAX) {
  985                 error = EMLINK;
  986                 goto out;
  987         }
  988         /*
  989          * The file may have been removed after namei droped the original
  990          * lock.
  991          */
  992         if (ip->i_effnlink == 0) {
  993                 error = ENOENT;
  994                 goto out;
  995         }
  996         if (ip->i_flags & (IMMUTABLE | APPEND)) {
  997                 error = EPERM;
  998                 goto out;
  999         }
 1000         ip->i_effnlink++;
 1001         ip->i_nlink++;
 1002         DIP_SET(ip, i_nlink, ip->i_nlink);
 1003         ip->i_flag |= IN_CHANGE;
 1004         if (DOINGSOFTDEP(vp))
 1005                 softdep_setup_link(VTOI(tdvp), ip);
 1006         error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp)));
 1007         if (!error) {
 1008                 ufs_makedirentry(ip, cnp, &newdir);
 1009                 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL, 0);
 1010         }
 1011 
 1012         if (error) {
 1013                 ip->i_effnlink--;
 1014                 ip->i_nlink--;
 1015                 DIP_SET(ip, i_nlink, ip->i_nlink);
 1016                 ip->i_flag |= IN_CHANGE;
 1017                 if (DOINGSOFTDEP(vp))
 1018                         softdep_revert_link(VTOI(tdvp), ip);
 1019         }
 1020 out:
 1021         return (error);
 1022 }
 1023 
 1024 /*
 1025  * whiteout vnode call
 1026  */
 1027 static int
 1028 ufs_whiteout(ap)
 1029         struct vop_whiteout_args /* {
 1030                 struct vnode *a_dvp;
 1031                 struct componentname *a_cnp;
 1032                 int a_flags;
 1033         } */ *ap;
 1034 {
 1035         struct vnode *dvp = ap->a_dvp;
 1036         struct componentname *cnp = ap->a_cnp;
 1037         struct direct newdir;
 1038         int error = 0;
 1039 
 1040         switch (ap->a_flags) {
 1041         case LOOKUP:
 1042                 /* 4.4 format directories support whiteout operations */
 1043                 if (dvp->v_mount->mnt_maxsymlinklen > 0)
 1044                         return (0);
 1045                 return (EOPNOTSUPP);
 1046 
 1047         case CREATE:
 1048                 /* create a new directory whiteout */
 1049 #ifdef INVARIANTS
 1050                 if ((cnp->cn_flags & SAVENAME) == 0)
 1051                         panic("ufs_whiteout: missing name");
 1052                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
 1053                         panic("ufs_whiteout: old format filesystem");
 1054 #endif
 1055 
 1056                 newdir.d_ino = WINO;
 1057                 newdir.d_namlen = cnp->cn_namelen;
 1058                 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1);
 1059                 newdir.d_type = DT_WHT;
 1060                 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL, 0);
 1061                 break;
 1062 
 1063         case DELETE:
 1064                 /* remove an existing directory whiteout */
 1065 #ifdef INVARIANTS
 1066                 if (dvp->v_mount->mnt_maxsymlinklen <= 0)
 1067                         panic("ufs_whiteout: old format filesystem");
 1068 #endif
 1069 
 1070                 cnp->cn_flags &= ~DOWHITEOUT;
 1071                 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0);
 1072                 break;
 1073         default:
 1074                 panic("ufs_whiteout: unknown op");
 1075         }
 1076         return (error);
 1077 }
 1078 
 1079 static volatile int rename_restarts;
 1080 SYSCTL_INT(_vfs_ufs, OID_AUTO, rename_restarts, CTLFLAG_RD,
 1081     __DEVOLATILE(int *, &rename_restarts), 0,
 1082     "Times rename had to restart due to lock contention");
 1083 
 1084 /*
 1085  * Rename system call.
 1086  *      rename("foo", "bar");
 1087  * is essentially
 1088  *      unlink("bar");
 1089  *      link("foo", "bar");
 1090  *      unlink("foo");
 1091  * but ``atomically''.  Can't do full commit without saving state in the
 1092  * inode on disk which isn't feasible at this time.  Best we can do is
 1093  * always guarantee the target exists.
 1094  *
 1095  * Basic algorithm is:
 1096  *
 1097  * 1) Bump link count on source while we're linking it to the
 1098  *    target.  This also ensure the inode won't be deleted out
 1099  *    from underneath us while we work (it may be truncated by
 1100  *    a concurrent `trunc' or `open' for creation).
 1101  * 2) Link source to destination.  If destination already exists,
 1102  *    delete it first.
 1103  * 3) Unlink source reference to inode if still around. If a
 1104  *    directory was moved and the parent of the destination
 1105  *    is different from the source, patch the ".." entry in the
 1106  *    directory.
 1107  */
 1108 static int
 1109 ufs_rename(ap)
 1110         struct vop_rename_args  /* {
 1111                 struct vnode *a_fdvp;
 1112                 struct vnode *a_fvp;
 1113                 struct componentname *a_fcnp;
 1114                 struct vnode *a_tdvp;
 1115                 struct vnode *a_tvp;
 1116                 struct componentname *a_tcnp;
 1117         } */ *ap;
 1118 {
 1119         struct vnode *tvp = ap->a_tvp;
 1120         struct vnode *tdvp = ap->a_tdvp;
 1121         struct vnode *fvp = ap->a_fvp;
 1122         struct vnode *fdvp = ap->a_fdvp;
 1123         struct vnode *nvp;
 1124         struct componentname *tcnp = ap->a_tcnp;
 1125         struct componentname *fcnp = ap->a_fcnp;
 1126         struct thread *td = fcnp->cn_thread;
 1127         struct inode *fip, *tip, *tdp, *fdp;
 1128         struct direct newdir;
 1129         off_t endoff;
 1130         int doingdirectory, newparent;
 1131         int error = 0;
 1132         struct mount *mp;
 1133         ino_t ino;
 1134 
 1135 #ifdef INVARIANTS
 1136         if ((tcnp->cn_flags & HASBUF) == 0 ||
 1137             (fcnp->cn_flags & HASBUF) == 0)
 1138                 panic("ufs_rename: no name");
 1139 #endif
 1140         endoff = 0;
 1141         mp = tdvp->v_mount;
 1142         VOP_UNLOCK(tdvp, 0);
 1143         if (tvp && tvp != tdvp)
 1144                 VOP_UNLOCK(tvp, 0);
 1145         /*
 1146          * Check for cross-device rename.
 1147          */
 1148         if ((fvp->v_mount != tdvp->v_mount) ||
 1149             (tvp && (fvp->v_mount != tvp->v_mount))) {
 1150                 error = EXDEV;
 1151                 mp = NULL;
 1152                 goto releout;
 1153         }
 1154 relock:
 1155         /* 
 1156          * We need to acquire 2 to 4 locks depending on whether tvp is NULL
 1157          * and fdvp and tdvp are the same directory.  Subsequently we need
 1158          * to double-check all paths and in the directory rename case we
 1159          * need to verify that we are not creating a directory loop.  To
 1160          * handle this we acquire all but fdvp using non-blocking
 1161          * acquisitions.  If we fail to acquire any lock in the path we will
 1162          * drop all held locks, acquire the new lock in a blocking fashion,
 1163          * and then release it and restart the rename.  This acquire/release
 1164          * step ensures that we do not spin on a lock waiting for release.
 1165          */
 1166         error = vn_lock(fdvp, LK_EXCLUSIVE);
 1167         if (error)
 1168                 goto releout;
 1169         if (vn_lock(tdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
 1170                 VOP_UNLOCK(fdvp, 0);
 1171                 error = vn_lock(tdvp, LK_EXCLUSIVE);
 1172                 if (error)
 1173                         goto releout;
 1174                 VOP_UNLOCK(tdvp, 0);
 1175                 atomic_add_int(&rename_restarts, 1);
 1176                 goto relock;
 1177         }
 1178         /*
 1179          * Re-resolve fvp to be certain it still exists and fetch the
 1180          * correct vnode.
 1181          */
 1182         error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
 1183         if (error) {
 1184                 VOP_UNLOCK(fdvp, 0);
 1185                 VOP_UNLOCK(tdvp, 0);
 1186                 goto releout;
 1187         }
 1188         error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
 1189         if (error) {
 1190                 VOP_UNLOCK(fdvp, 0);
 1191                 VOP_UNLOCK(tdvp, 0);
 1192                 if (error != EBUSY)
 1193                         goto releout;
 1194                 error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
 1195                 if (error != 0)
 1196                         goto releout;
 1197                 VOP_UNLOCK(nvp, 0);
 1198                 vrele(fvp);
 1199                 fvp = nvp;
 1200                 atomic_add_int(&rename_restarts, 1);
 1201                 goto relock;
 1202         }
 1203         vrele(fvp);
 1204         fvp = nvp;
 1205         /*
 1206          * Re-resolve tvp and acquire the vnode lock if present.
 1207          */
 1208         error = ufs_lookup_ino(tdvp, NULL, tcnp, &ino);
 1209         if (error != 0 && error != EJUSTRETURN) {
 1210                 VOP_UNLOCK(fdvp, 0);
 1211                 VOP_UNLOCK(tdvp, 0);
 1212                 VOP_UNLOCK(fvp, 0);
 1213                 goto releout;
 1214         }
 1215         /*
 1216          * If tvp disappeared we just carry on.
 1217          */
 1218         if (error == EJUSTRETURN && tvp != NULL) {
 1219                 vrele(tvp);
 1220                 tvp = NULL;
 1221         }
 1222         /*
 1223          * Get the tvp ino if the lookup succeeded.  We may have to restart
 1224          * if the non-blocking acquire fails.
 1225          */
 1226         if (error == 0) {
 1227                 nvp = NULL;
 1228                 error = VFS_VGET(mp, ino, LK_EXCLUSIVE | LK_NOWAIT, &nvp);
 1229                 if (tvp)
 1230                         vrele(tvp);
 1231                 tvp = nvp;
 1232                 if (error) {
 1233                         VOP_UNLOCK(fdvp, 0);
 1234                         VOP_UNLOCK(tdvp, 0);
 1235                         VOP_UNLOCK(fvp, 0);
 1236                         if (error != EBUSY)
 1237                                 goto releout;
 1238                         error = VFS_VGET(mp, ino, LK_EXCLUSIVE, &nvp);
 1239                         if (error != 0)
 1240                                 goto releout;
 1241                         vput(nvp);
 1242                         atomic_add_int(&rename_restarts, 1);
 1243                         goto relock;
 1244                 }
 1245         }
 1246         fdp = VTOI(fdvp);
 1247         fip = VTOI(fvp);
 1248         tdp = VTOI(tdvp);
 1249         tip = NULL;
 1250         if (tvp)
 1251                 tip = VTOI(tvp);
 1252         if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
 1253             (VTOI(tdvp)->i_flags & APPEND))) {
 1254                 error = EPERM;
 1255                 goto unlockout;
 1256         }
 1257         /*
 1258          * Renaming a file to itself has no effect.  The upper layers should
 1259          * not call us in that case.  However, things could change after
 1260          * we drop the locks above.
 1261          */
 1262         if (fvp == tvp) {
 1263                 error = 0;
 1264                 goto unlockout;
 1265         }
 1266         doingdirectory = 0;
 1267         newparent = 0;
 1268         ino = fip->i_number;
 1269         if (fip->i_nlink >= LINK_MAX) {
 1270                 error = EMLINK;
 1271                 goto unlockout;
 1272         }
 1273         if ((fip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))
 1274             || (fdp->i_flags & APPEND)) {
 1275                 error = EPERM;
 1276                 goto unlockout;
 1277         }
 1278         if ((fip->i_mode & IFMT) == IFDIR) {
 1279                 /*
 1280                  * Avoid ".", "..", and aliases of "." for obvious reasons.
 1281                  */
 1282                 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
 1283                     fdp == fip ||
 1284                     (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT) {
 1285                         error = EINVAL;
 1286                         goto unlockout;
 1287                 }
 1288                 if (fdp->i_number != tdp->i_number)
 1289                         newparent = tdp->i_number;
 1290                 doingdirectory = 1;
 1291         }
 1292         if ((fvp->v_type == VDIR && fvp->v_mountedhere != NULL) ||
 1293             (tvp != NULL && tvp->v_type == VDIR &&
 1294             tvp->v_mountedhere != NULL)) {
 1295                 error = EXDEV;
 1296                 goto unlockout;
 1297         }
 1298 
 1299         /*
 1300          * If ".." must be changed (ie the directory gets a new
 1301          * parent) then the source directory must not be in the
 1302          * directory hierarchy above the target, as this would
 1303          * orphan everything below the source directory. Also
 1304          * the user must have write permission in the source so
 1305          * as to be able to change "..".
 1306          */
 1307         if (doingdirectory && newparent) {
 1308                 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
 1309                 if (error)
 1310                         goto unlockout;
 1311                 error = ufs_checkpath(ino, fdp->i_number, tdp, tcnp->cn_cred,
 1312                     &ino);
 1313                 /*
 1314                  * We encountered a lock that we have to wait for.  Unlock
 1315                  * everything else and VGET before restarting.
 1316                  */
 1317                 if (ino) {
 1318                         VOP_UNLOCK(fdvp, 0);
 1319                         VOP_UNLOCK(fvp, 0);
 1320                         VOP_UNLOCK(tdvp, 0);
 1321                         if (tvp)
 1322                                 VOP_UNLOCK(tvp, 0);
 1323                         error = VFS_VGET(mp, ino, LK_SHARED, &nvp);
 1324                         if (error == 0)
 1325                                 vput(nvp);
 1326                         atomic_add_int(&rename_restarts, 1);
 1327                         goto relock;
 1328                 }
 1329                 if (error)
 1330                         goto unlockout;
 1331                 if ((tcnp->cn_flags & SAVESTART) == 0)
 1332                         panic("ufs_rename: lost to startdir");
 1333         }
 1334         if (fip->i_effnlink == 0 || fdp->i_effnlink == 0 ||
 1335             tdp->i_effnlink == 0)
 1336                 panic("Bad effnlink fip %p, fdp %p, tdp %p", fip, fdp, tdp);
 1337 
 1338         /*
 1339          * 1) Bump link count while we're moving stuff
 1340          *    around.  If we crash somewhere before
 1341          *    completing our work, the link count
 1342          *    may be wrong, but correctable.
 1343          */
 1344         fip->i_effnlink++;
 1345         fip->i_nlink++;
 1346         DIP_SET(fip, i_nlink, fip->i_nlink);
 1347         fip->i_flag |= IN_CHANGE;
 1348         if (DOINGSOFTDEP(fvp))
 1349                 softdep_setup_link(tdp, fip);
 1350         error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) | DOINGASYNC(fvp)));
 1351         if (error)
 1352                 goto bad;
 1353 
 1354         /*
 1355          * 2) If target doesn't exist, link the target
 1356          *    to the source and unlink the source.
 1357          *    Otherwise, rewrite the target directory
 1358          *    entry to reference the source inode and
 1359          *    expunge the original entry's existence.
 1360          */
 1361         if (tip == NULL) {
 1362                 if (ITODEV(tdp) != ITODEV(fip))
 1363                         panic("ufs_rename: EXDEV");
 1364                 if (doingdirectory && newparent) {
 1365                         /*
 1366                          * Account for ".." in new directory.
 1367                          * When source and destination have the same
 1368                          * parent we don't adjust the link count.  The
 1369                          * actual link modification is completed when
 1370                          * .. is rewritten below.
 1371                          */
 1372                         if ((nlink_t)tdp->i_nlink >= LINK_MAX) {
 1373                                 error = EMLINK;
 1374                                 goto bad;
 1375                         }
 1376                 }
 1377                 ufs_makedirentry(fip, tcnp, &newdir);
 1378                 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL, 1);
 1379                 if (error)
 1380                         goto bad;
 1381                 /* Setup tdvp for directory compaction if needed. */
 1382                 if (tdp->i_count && tdp->i_endoff &&
 1383                     tdp->i_endoff < tdp->i_size)
 1384                         endoff = tdp->i_endoff;
 1385         } else {
 1386                 if (ITODEV(tip) != ITODEV(tdp) || ITODEV(tip) != ITODEV(fip))
 1387                         panic("ufs_rename: EXDEV");
 1388                 /*
 1389                  * Short circuit rename(foo, foo).
 1390                  */
 1391                 if (tip->i_number == fip->i_number)
 1392                         panic("ufs_rename: same file");
 1393                 /*
 1394                  * If the parent directory is "sticky", then the caller
 1395                  * must possess VADMIN for the parent directory, or the
 1396                  * destination of the rename.  This implements append-only
 1397                  * directories.
 1398                  */
 1399                 if ((tdp->i_mode & S_ISTXT) &&
 1400                     VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, td) &&
 1401                     VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, td)) {
 1402                         error = EPERM;
 1403                         goto bad;
 1404                 }
 1405                 /*
 1406                  * Target must be empty if a directory and have no links
 1407                  * to it. Also, ensure source and target are compatible
 1408                  * (both directories, or both not directories).
 1409                  */
 1410                 if ((tip->i_mode & IFMT) == IFDIR) {
 1411                         if ((tip->i_effnlink > 2) ||
 1412                             !ufs_dirempty(tip, tdp->i_number, tcnp->cn_cred)) {
 1413                                 error = ENOTEMPTY;
 1414                                 goto bad;
 1415                         }
 1416                         if (!doingdirectory) {
 1417                                 error = ENOTDIR;
 1418                                 goto bad;
 1419                         }
 1420                         cache_purge(tdvp);
 1421                 } else if (doingdirectory) {
 1422                         error = EISDIR;
 1423                         goto bad;
 1424                 }
 1425                 if (doingdirectory) {
 1426                         if (!newparent) {
 1427                                 tdp->i_effnlink--;
 1428                                 if (DOINGSOFTDEP(tdvp))
 1429                                         softdep_change_linkcnt(tdp);
 1430                         }
 1431                         tip->i_effnlink--;
 1432                         if (DOINGSOFTDEP(tvp))
 1433                                 softdep_change_linkcnt(tip);
 1434                 }
 1435                 error = ufs_dirrewrite(tdp, tip, fip->i_number,
 1436                     IFTODT(fip->i_mode),
 1437                     (doingdirectory && newparent) ? newparent : doingdirectory);
 1438                 if (error) {
 1439                         if (doingdirectory) {
 1440                                 if (!newparent) {
 1441                                         tdp->i_effnlink++;
 1442                                         if (DOINGSOFTDEP(tdvp))
 1443                                                 softdep_change_linkcnt(tdp);
 1444                                 }
 1445                                 tip->i_effnlink++;
 1446                                 if (DOINGSOFTDEP(tvp))
 1447                                         softdep_change_linkcnt(tip);
 1448                         }
 1449                 }
 1450                 if (doingdirectory && !DOINGSOFTDEP(tvp)) {
 1451                         /*
 1452                          * The only stuff left in the directory is "."
 1453                          * and "..". The "." reference is inconsequential
 1454                          * since we are quashing it. We have removed the "."
 1455                          * reference and the reference in the parent directory,
 1456                          * but there may be other hard links. The soft
 1457                          * dependency code will arrange to do these operations
 1458                          * after the parent directory entry has been deleted on
 1459                          * disk, so when running with that code we avoid doing
 1460                          * them now.
 1461                          */
 1462                         if (!newparent) {
 1463                                 tdp->i_nlink--;
 1464                                 DIP_SET(tdp, i_nlink, tdp->i_nlink);
 1465                                 tdp->i_flag |= IN_CHANGE;
 1466                         }
 1467                         tip->i_nlink--;
 1468                         DIP_SET(tip, i_nlink, tip->i_nlink);
 1469                         tip->i_flag |= IN_CHANGE;
 1470                 }
 1471         }
 1472 
 1473         /*
 1474          * 3) Unlink the source.  We have to resolve the path again to
 1475          * fixup the directory offset and count for ufs_dirremove.
 1476          */
 1477         if (fdvp == tdvp) {
 1478                 error = ufs_lookup_ino(fdvp, NULL, fcnp, &ino);
 1479                 if (error)
 1480                         panic("ufs_rename: from entry went away!");
 1481                 if (ino != fip->i_number)
 1482                         panic("ufs_rename: ino mismatch %ju != %ju\n",
 1483                             (uintmax_t)ino, (uintmax_t)fip->i_number);
 1484         }
 1485         /*
 1486          * If the source is a directory with a
 1487          * new parent, the link count of the old
 1488          * parent directory must be decremented
 1489          * and ".." set to point to the new parent.
 1490          */
 1491         if (doingdirectory && newparent) {
 1492                 /*
 1493                  * If tip exists we simply use its link, otherwise we must
 1494                  * add a new one.
 1495                  */
 1496                 if (tip == NULL) {
 1497                         tdp->i_effnlink++;
 1498                         tdp->i_nlink++;
 1499                         DIP_SET(tdp, i_nlink, tdp->i_nlink);
 1500                         tdp->i_flag |= IN_CHANGE;
 1501                         if (DOINGSOFTDEP(tdvp))
 1502                                 softdep_setup_dotdot_link(tdp, fip);
 1503                         error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) |
 1504                                                    DOINGASYNC(tdvp)));
 1505                         /* Don't go to bad here as the new link exists. */
 1506                         if (error)
 1507                                 goto unlockout;
 1508                 } else if (DOINGSUJ(tdvp))
 1509                         /* Journal must account for each new link. */
 1510                         softdep_setup_dotdot_link(tdp, fip);
 1511                 fip->i_offset = mastertemplate.dot_reclen;
 1512                 ufs_dirrewrite(fip, fdp, newparent, DT_DIR, 0);
 1513                 cache_purge(fdvp);
 1514         }
 1515         error = ufs_dirremove(fdvp, fip, fcnp->cn_flags, 0);
 1516         /*
 1517          * The kern_renameat() looks up the fvp using the DELETE flag, which
 1518          * causes the removal of the name cache entry for fvp.
 1519          * As the relookup of the fvp is done in two steps:
 1520          * ufs_lookup_ino() and then VFS_VGET(), another thread might do a
 1521          * normal lookup of the from name just before the VFS_VGET() call,
 1522          * causing the cache entry to be re-instantiated.
 1523          *
 1524          * The same issue also applies to tvp if it exists as
 1525          * otherwise we may have a stale name cache entry for the new
 1526          * name that references the old i-node if it has other links
 1527          * or open file descriptors.
 1528          */
 1529         cache_purge(fvp);
 1530         if (tvp)
 1531                 cache_purge(tvp);
 1532         cache_purge_negative(tdvp);
 1533 
 1534 unlockout:
 1535         vput(fdvp);
 1536         vput(fvp);
 1537         if (tvp)
 1538                 vput(tvp);
 1539         /*
 1540          * If compaction or fsync was requested do it now that other locks
 1541          * are no longer needed.
 1542          */
 1543         if (error == 0 && endoff != 0) {
 1544                 error = UFS_TRUNCATE(tdvp, endoff, IO_NORMAL | IO_SYNC,
 1545                     tcnp->cn_cred);
 1546                 if (error != 0)
 1547                         vn_printf(tdvp, "ufs_rename: failed to truncate "
 1548                             "err %d", error);
 1549 #ifdef UFS_DIRHASH
 1550                 else if (tdp->i_dirhash != NULL)
 1551                         ufsdirhash_dirtrunc(tdp, endoff);
 1552 #endif
 1553                 /*
 1554                  * Even if the directory compaction failed, rename was
 1555                  * succesful.  Do not propagate a UFS_TRUNCATE() error
 1556                  * to the caller.
 1557                  */
 1558                 error = 0;
 1559         }
 1560         if (error == 0 && tdp->i_flag & IN_NEEDSYNC)
 1561                 error = VOP_FSYNC(tdvp, MNT_WAIT, td);
 1562         vput(tdvp);
 1563         return (error);
 1564 
 1565 bad:
 1566         fip->i_effnlink--;
 1567         fip->i_nlink--;
 1568         DIP_SET(fip, i_nlink, fip->i_nlink);
 1569         fip->i_flag |= IN_CHANGE;
 1570         if (DOINGSOFTDEP(fvp))
 1571                 softdep_revert_link(tdp, fip);
 1572         goto unlockout;
 1573 
 1574 releout:
 1575         vrele(fdvp);
 1576         vrele(fvp);
 1577         vrele(tdvp);
 1578         if (tvp)
 1579                 vrele(tvp);
 1580 
 1581         return (error);
 1582 }
 1583 
 1584 #ifdef UFS_ACL
 1585 static int
 1586 ufs_do_posix1e_acl_inheritance_dir(struct vnode *dvp, struct vnode *tvp,
 1587     mode_t dmode, struct ucred *cred, struct thread *td)
 1588 {
 1589         int error;
 1590         struct inode *ip = VTOI(tvp);
 1591         struct acl *dacl, *acl;
 1592 
 1593         acl = acl_alloc(M_WAITOK);
 1594         dacl = acl_alloc(M_WAITOK);
 1595 
 1596         /*
 1597          * Retrieve default ACL from parent, if any.
 1598          */
 1599         error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
 1600         switch (error) {
 1601         case 0:
 1602                 /*
 1603                  * Retrieved a default ACL, so merge mode and ACL if
 1604                  * necessary.  If the ACL is empty, fall through to
 1605                  * the "not defined or available" case.
 1606                  */
 1607                 if (acl->acl_cnt != 0) {
 1608                         dmode = acl_posix1e_newfilemode(dmode, acl);
 1609                         ip->i_mode = dmode;
 1610                         DIP_SET(ip, i_mode, dmode);
 1611                         *dacl = *acl;
 1612                         ufs_sync_acl_from_inode(ip, acl);
 1613                         break;
 1614                 }
 1615                 /* FALLTHROUGH */
 1616 
 1617         case EOPNOTSUPP:
 1618                 /*
 1619                  * Just use the mode as-is.
 1620                  */
 1621                 ip->i_mode = dmode;
 1622                 DIP_SET(ip, i_mode, dmode);
 1623                 error = 0;
 1624                 goto out;
 1625         
 1626         default:
 1627                 goto out;
 1628         }
 1629 
 1630         /*
 1631          * XXX: If we abort now, will Soft Updates notify the extattr
 1632          * code that the EAs for the file need to be released?
 1633          */
 1634         error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
 1635         if (error == 0)
 1636                 error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl, cred, td);
 1637         switch (error) {
 1638         case 0:
 1639                 break;
 1640 
 1641         case EOPNOTSUPP:
 1642                 /*
 1643                  * XXX: This should not happen, as EOPNOTSUPP above
 1644                  * was supposed to free acl.
 1645                  */
 1646                 printf("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()\n");
 1647                 /*
 1648                 panic("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()");
 1649                  */
 1650                 break;
 1651 
 1652         default:
 1653                 goto out;
 1654         }
 1655 
 1656 out:
 1657         acl_free(acl);
 1658         acl_free(dacl);
 1659 
 1660         return (error);
 1661 }
 1662 
 1663 static int
 1664 ufs_do_posix1e_acl_inheritance_file(struct vnode *dvp, struct vnode *tvp,
 1665     mode_t mode, struct ucred *cred, struct thread *td)
 1666 {
 1667         int error;
 1668         struct inode *ip = VTOI(tvp);
 1669         struct acl *acl;
 1670 
 1671         acl = acl_alloc(M_WAITOK);
 1672 
 1673         /*
 1674          * Retrieve default ACL for parent, if any.
 1675          */
 1676         error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cred, td);
 1677         switch (error) {
 1678         case 0:
 1679                 /*
 1680                  * Retrieved a default ACL, so merge mode and ACL if
 1681                  * necessary.
 1682                  */
 1683                 if (acl->acl_cnt != 0) {
 1684                         /*
 1685                          * Two possible ways for default ACL to not
 1686                          * be present.  First, the EA can be
 1687                          * undefined, or second, the default ACL can
 1688                          * be blank.  If it's blank, fall through to
 1689                          * the it's not defined case.
 1690                          */
 1691                         mode = acl_posix1e_newfilemode(mode, acl);
 1692                         ip->i_mode = mode;
 1693                         DIP_SET(ip, i_mode, mode);
 1694                         ufs_sync_acl_from_inode(ip, acl);
 1695                         break;
 1696                 }
 1697                 /* FALLTHROUGH */
 1698 
 1699         case EOPNOTSUPP:
 1700                 /*
 1701                  * Just use the mode as-is.
 1702                  */
 1703                 ip->i_mode = mode;
 1704                 DIP_SET(ip, i_mode, mode);
 1705                 error = 0;
 1706                 goto out;
 1707 
 1708         default:
 1709                 goto out;
 1710         }
 1711 
 1712         /*
 1713          * XXX: If we abort now, will Soft Updates notify the extattr
 1714          * code that the EAs for the file need to be released?
 1715          */
 1716         error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cred, td);
 1717         switch (error) {
 1718         case 0:
 1719                 break;
 1720 
 1721         case EOPNOTSUPP:
 1722                 /*
 1723                  * XXX: This should not happen, as EOPNOTSUPP above was
 1724                  * supposed to free acl.
 1725                  */
 1726                 printf("ufs_do_posix1e_acl_inheritance_file: VOP_GETACL() "
 1727                     "but no VOP_SETACL()\n");
 1728                 /* panic("ufs_do_posix1e_acl_inheritance_file: VOP_GETACL() "
 1729                     "but no VOP_SETACL()"); */
 1730                 break;
 1731 
 1732         default:
 1733                 goto out;
 1734         }
 1735 
 1736 out:
 1737         acl_free(acl);
 1738 
 1739         return (error);
 1740 }
 1741 
 1742 static int
 1743 ufs_do_nfs4_acl_inheritance(struct vnode *dvp, struct vnode *tvp,
 1744     mode_t child_mode, struct ucred *cred, struct thread *td)
 1745 {
 1746         int error;
 1747         struct acl *parent_aclp, *child_aclp;
 1748 
 1749         parent_aclp = acl_alloc(M_WAITOK);
 1750         child_aclp = acl_alloc(M_WAITOK | M_ZERO);
 1751 
 1752         error = ufs_getacl_nfs4_internal(dvp, parent_aclp, td);
 1753         if (error)
 1754                 goto out;
 1755         acl_nfs4_compute_inherited_acl(parent_aclp, child_aclp,
 1756             child_mode, VTOI(tvp)->i_uid, tvp->v_type == VDIR);
 1757         error = ufs_setacl_nfs4_internal(tvp, child_aclp, td);
 1758         if (error)
 1759                 goto out;
 1760 out:
 1761         acl_free(parent_aclp);
 1762         acl_free(child_aclp);
 1763 
 1764         return (error);
 1765 }
 1766 #endif
 1767 
 1768 /*
 1769  * Mkdir system call
 1770  */
 1771 static int
 1772 ufs_mkdir(ap)
 1773         struct vop_mkdir_args /* {
 1774                 struct vnode *a_dvp;
 1775                 struct vnode **a_vpp;
 1776                 struct componentname *a_cnp;
 1777                 struct vattr *a_vap;
 1778         } */ *ap;
 1779 {
 1780         struct vnode *dvp = ap->a_dvp;
 1781         struct vattr *vap = ap->a_vap;
 1782         struct componentname *cnp = ap->a_cnp;
 1783         struct inode *ip, *dp;
 1784         struct vnode *tvp;
 1785         struct buf *bp;
 1786         struct dirtemplate dirtemplate, *dtp;
 1787         struct direct newdir;
 1788         int error, dmode;
 1789         long blkoff;
 1790 
 1791 #ifdef INVARIANTS
 1792         if ((cnp->cn_flags & HASBUF) == 0)
 1793                 panic("ufs_mkdir: no name");
 1794 #endif
 1795         dp = VTOI(dvp);
 1796         if ((nlink_t)dp->i_nlink >= LINK_MAX) {
 1797                 error = EMLINK;
 1798                 goto out;
 1799         }
 1800         dmode = vap->va_mode & 0777;
 1801         dmode |= IFDIR;
 1802         /*
 1803          * Must simulate part of ufs_makeinode here to acquire the inode,
 1804          * but not have it entered in the parent directory. The entry is
 1805          * made later after writing "." and ".." entries.
 1806          */
 1807         if (dp->i_effnlink < 2) {
 1808                 print_bad_link_count("ufs_mkdir", dvp);
 1809                 error = EINVAL;
 1810                 goto out;
 1811         }
 1812         error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp);
 1813         if (error)
 1814                 goto out;
 1815         ip = VTOI(tvp);
 1816         ip->i_gid = dp->i_gid;
 1817         DIP_SET(ip, i_gid, dp->i_gid);
 1818 #ifdef SUIDDIR
 1819         {
 1820 #ifdef QUOTA
 1821                 struct ucred ucred, *ucp;
 1822                 gid_t ucred_group;
 1823                 ucp = cnp->cn_cred;
 1824 #endif
 1825                 /*
 1826                  * If we are hacking owners here, (only do this where told to)
 1827                  * and we are not giving it TO root, (would subvert quotas)
 1828                  * then go ahead and give it to the other user.
 1829                  * The new directory also inherits the SUID bit.
 1830                  * If user's UID and dir UID are the same,
 1831                  * 'give it away' so that the SUID is still forced on.
 1832                  */
 1833                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 1834                     (dp->i_mode & ISUID) && dp->i_uid) {
 1835                         dmode |= ISUID;
 1836                         ip->i_uid = dp->i_uid;
 1837                         DIP_SET(ip, i_uid, dp->i_uid);
 1838 #ifdef QUOTA
 1839                         if (dp->i_uid != cnp->cn_cred->cr_uid) {
 1840                                 /*
 1841                                  * Make sure the correct user gets charged
 1842                                  * for the space.
 1843                                  * Make a dummy credential for the victim.
 1844                                  * XXX This seems to never be accessed out of
 1845                                  * our context so a stack variable is ok.
 1846                                  */
 1847                                 refcount_init(&ucred.cr_ref, 1);
 1848                                 ucred.cr_uid = ip->i_uid;
 1849                                 ucred.cr_ngroups = 1;
 1850                                 ucred.cr_groups = &ucred_group;
 1851                                 ucred.cr_groups[0] = dp->i_gid;
 1852                                 ucp = &ucred;
 1853                         }
 1854 #endif
 1855                 } else {
 1856                         ip->i_uid = cnp->cn_cred->cr_uid;
 1857                         DIP_SET(ip, i_uid, ip->i_uid);
 1858                 }
 1859 #ifdef QUOTA
 1860                 if ((error = getinoquota(ip)) ||
 1861                     (error = chkiq(ip, 1, ucp, 0))) {
 1862                         if (DOINGSOFTDEP(tvp))
 1863                                 softdep_revert_link(dp, ip);
 1864                         UFS_VFREE(tvp, ip->i_number, dmode);
 1865                         vput(tvp);
 1866                         return (error);
 1867                 }
 1868 #endif
 1869         }
 1870 #else   /* !SUIDDIR */
 1871         ip->i_uid = cnp->cn_cred->cr_uid;
 1872         DIP_SET(ip, i_uid, ip->i_uid);
 1873 #ifdef QUOTA
 1874         if ((error = getinoquota(ip)) ||
 1875             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 1876                 if (DOINGSOFTDEP(tvp))
 1877                         softdep_revert_link(dp, ip);
 1878                 UFS_VFREE(tvp, ip->i_number, dmode);
 1879                 vput(tvp);
 1880                 return (error);
 1881         }
 1882 #endif
 1883 #endif  /* !SUIDDIR */
 1884         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 1885         ip->i_mode = dmode;
 1886         DIP_SET(ip, i_mode, dmode);
 1887         tvp->v_type = VDIR;     /* Rest init'd in getnewvnode(). */
 1888         ip->i_effnlink = 2;
 1889         ip->i_nlink = 2;
 1890         DIP_SET(ip, i_nlink, 2);
 1891 
 1892         if (cnp->cn_flags & ISWHITEOUT) {
 1893                 ip->i_flags |= UF_OPAQUE;
 1894                 DIP_SET(ip, i_flags, ip->i_flags);
 1895         }
 1896 
 1897         /*
 1898          * Bump link count in parent directory to reflect work done below.
 1899          * Should be done before reference is created so cleanup is
 1900          * possible if we crash.
 1901          */
 1902         dp->i_effnlink++;
 1903         dp->i_nlink++;
 1904         DIP_SET(dp, i_nlink, dp->i_nlink);
 1905         dp->i_flag |= IN_CHANGE;
 1906         if (DOINGSOFTDEP(dvp))
 1907                 softdep_setup_mkdir(dp, ip);
 1908         error = UFS_UPDATE(dvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp)));
 1909         if (error)
 1910                 goto bad;
 1911 #ifdef MAC
 1912         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 1913                 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
 1914                     dvp, tvp, cnp);
 1915                 if (error)
 1916                         goto bad;
 1917         }
 1918 #endif
 1919 #ifdef UFS_ACL
 1920         if (dvp->v_mount->mnt_flag & MNT_ACLS) {
 1921                 error = ufs_do_posix1e_acl_inheritance_dir(dvp, tvp, dmode,
 1922                     cnp->cn_cred, cnp->cn_thread);
 1923                 if (error)
 1924                         goto bad;
 1925         } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
 1926                 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, dmode,
 1927                     cnp->cn_cred, cnp->cn_thread);
 1928                 if (error)
 1929                         goto bad;
 1930         }
 1931 #endif /* !UFS_ACL */
 1932 
 1933         /*
 1934          * Initialize directory with "." and ".." from static template.
 1935          */
 1936         if (dvp->v_mount->mnt_maxsymlinklen > 0)
 1937                 dtp = &mastertemplate;
 1938         else
 1939                 dtp = (struct dirtemplate *)&omastertemplate;
 1940         dirtemplate = *dtp;
 1941         dirtemplate.dot_ino = ip->i_number;
 1942         dirtemplate.dotdot_ino = dp->i_number;
 1943         vnode_pager_setsize(tvp, DIRBLKSIZ);
 1944         if ((error = UFS_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred,
 1945             BA_CLRBUF, &bp)) != 0)
 1946                 goto bad;
 1947         ip->i_size = DIRBLKSIZ;
 1948         DIP_SET(ip, i_size, DIRBLKSIZ);
 1949         ip->i_flag |= IN_CHANGE | IN_UPDATE;
 1950         bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate);
 1951         if (DOINGSOFTDEP(tvp)) {
 1952                 /*
 1953                  * Ensure that the entire newly allocated block is a
 1954                  * valid directory so that future growth within the
 1955                  * block does not have to ensure that the block is
 1956                  * written before the inode.
 1957                  */
 1958                 blkoff = DIRBLKSIZ;
 1959                 while (blkoff < bp->b_bcount) {
 1960                         ((struct direct *)
 1961                            (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ;
 1962                         blkoff += DIRBLKSIZ;
 1963                 }
 1964         }
 1965         if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) |
 1966                                        DOINGASYNC(tvp)))) != 0) {
 1967                 (void)bwrite(bp);
 1968                 goto bad;
 1969         }
 1970         /*
 1971          * Directory set up, now install its entry in the parent directory.
 1972          *
 1973          * If we are not doing soft dependencies, then we must write out the
 1974          * buffer containing the new directory body before entering the new 
 1975          * name in the parent. If we are doing soft dependencies, then the
 1976          * buffer containing the new directory body will be passed to and
 1977          * released in the soft dependency code after the code has attached
 1978          * an appropriate ordering dependency to the buffer which ensures that
 1979          * the buffer is written before the new name is written in the parent.
 1980          */
 1981         if (DOINGASYNC(dvp))
 1982                 bdwrite(bp);
 1983         else if (!DOINGSOFTDEP(dvp) && ((error = bwrite(bp))))
 1984                 goto bad;
 1985         ufs_makedirentry(ip, cnp, &newdir);
 1986         error = ufs_direnter(dvp, tvp, &newdir, cnp, bp, 0);
 1987         
 1988 bad:
 1989         if (error == 0) {
 1990                 *ap->a_vpp = tvp;
 1991         } else {
 1992                 dp->i_effnlink--;
 1993                 dp->i_nlink--;
 1994                 DIP_SET(dp, i_nlink, dp->i_nlink);
 1995                 dp->i_flag |= IN_CHANGE;
 1996                 /*
 1997                  * No need to do an explicit VOP_TRUNCATE here, vrele will
 1998                  * do this for us because we set the link count to 0.
 1999                  */
 2000                 ip->i_effnlink = 0;
 2001                 ip->i_nlink = 0;
 2002                 DIP_SET(ip, i_nlink, 0);
 2003                 ip->i_flag |= IN_CHANGE;
 2004                 if (DOINGSOFTDEP(tvp))
 2005                         softdep_revert_mkdir(dp, ip);
 2006 
 2007                 vput(tvp);
 2008         }
 2009 out:
 2010         return (error);
 2011 }
 2012 
 2013 /*
 2014  * Rmdir system call.
 2015  */
 2016 static int
 2017 ufs_rmdir(ap)
 2018         struct vop_rmdir_args /* {
 2019                 struct vnode *a_dvp;
 2020                 struct vnode *a_vp;
 2021                 struct componentname *a_cnp;
 2022         } */ *ap;
 2023 {
 2024         struct vnode *vp = ap->a_vp;
 2025         struct vnode *dvp = ap->a_dvp;
 2026         struct componentname *cnp = ap->a_cnp;
 2027         struct inode *ip, *dp;
 2028         int error;
 2029 
 2030         ip = VTOI(vp);
 2031         dp = VTOI(dvp);
 2032 
 2033         /*
 2034          * Do not remove a directory that is in the process of being renamed.
 2035          * Verify the directory is empty (and valid). Rmdir ".." will not be
 2036          * valid since ".." will contain a reference to the current directory
 2037          * and thus be non-empty. Do not allow the removal of mounted on
 2038          * directories (this can happen when an NFS exported filesystem
 2039          * tries to remove a locally mounted on directory).
 2040          */
 2041         error = 0;
 2042         if (dp->i_effnlink <= 2) {
 2043                 if (dp->i_effnlink == 2)
 2044                         print_bad_link_count("ufs_rmdir", dvp);
 2045                 error = EINVAL;
 2046                 goto out;
 2047         }
 2048         if (!ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) {
 2049                 error = ENOTEMPTY;
 2050                 goto out;
 2051         }
 2052         if ((dp->i_flags & APPEND)
 2053             || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
 2054                 error = EPERM;
 2055                 goto out;
 2056         }
 2057         if (vp->v_mountedhere != 0) {
 2058                 error = EINVAL;
 2059                 goto out;
 2060         }
 2061 #ifdef UFS_GJOURNAL
 2062         ufs_gjournal_orphan(vp);
 2063 #endif
 2064         /*
 2065          * Delete reference to directory before purging
 2066          * inode.  If we crash in between, the directory
 2067          * will be reattached to lost+found,
 2068          */
 2069         dp->i_effnlink--;
 2070         ip->i_effnlink--;
 2071         if (DOINGSOFTDEP(vp))
 2072                 softdep_setup_rmdir(dp, ip);
 2073         error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1);
 2074         if (error) {
 2075                 dp->i_effnlink++;
 2076                 ip->i_effnlink++;
 2077                 if (DOINGSOFTDEP(vp))
 2078                         softdep_revert_rmdir(dp, ip);
 2079                 goto out;
 2080         }
 2081         cache_purge(dvp);
 2082         /*
 2083          * The only stuff left in the directory is "." and "..". The "."
 2084          * reference is inconsequential since we are quashing it. The soft
 2085          * dependency code will arrange to do these operations after
 2086          * the parent directory entry has been deleted on disk, so
 2087          * when running with that code we avoid doing them now.
 2088          */
 2089         if (!DOINGSOFTDEP(vp)) {
 2090                 dp->i_nlink--;
 2091                 DIP_SET(dp, i_nlink, dp->i_nlink);
 2092                 dp->i_flag |= IN_CHANGE;
 2093                 error = UFS_UPDATE(dvp, 0);
 2094                 ip->i_nlink--;
 2095                 DIP_SET(ip, i_nlink, ip->i_nlink);
 2096                 ip->i_flag |= IN_CHANGE;
 2097         }
 2098         cache_purge(vp);
 2099 #ifdef UFS_DIRHASH
 2100         /* Kill any active hash; i_effnlink == 0, so it will not come back. */
 2101         if (ip->i_dirhash != NULL)
 2102                 ufsdirhash_free(ip);
 2103 #endif
 2104 out:
 2105         return (error);
 2106 }
 2107 
 2108 /*
 2109  * symlink -- make a symbolic link
 2110  */
 2111 static int
 2112 ufs_symlink(ap)
 2113         struct vop_symlink_args /* {
 2114                 struct vnode *a_dvp;
 2115                 struct vnode **a_vpp;
 2116                 struct componentname *a_cnp;
 2117                 struct vattr *a_vap;
 2118                 char *a_target;
 2119         } */ *ap;
 2120 {
 2121         struct vnode *vp, **vpp = ap->a_vpp;
 2122         struct inode *ip;
 2123         int len, error;
 2124 
 2125         error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp,
 2126             vpp, ap->a_cnp, "ufs_symlink");
 2127         if (error)
 2128                 return (error);
 2129         vp = *vpp;
 2130         len = strlen(ap->a_target);
 2131         if (len < vp->v_mount->mnt_maxsymlinklen) {
 2132                 ip = VTOI(vp);
 2133                 bcopy(ap->a_target, SHORTLINK(ip), len);
 2134                 ip->i_size = len;
 2135                 DIP_SET(ip, i_size, len);
 2136                 ip->i_flag |= IN_CHANGE | IN_UPDATE;
 2137                 error = UFS_UPDATE(vp, 0);
 2138         } else
 2139                 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0,
 2140                     UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
 2141                     ap->a_cnp->cn_cred, NOCRED, NULL, NULL);
 2142         if (error)
 2143                 vput(vp);
 2144         return (error);
 2145 }
 2146 
 2147 /*
 2148  * Vnode op for reading directories.
 2149  */
 2150 int
 2151 ufs_readdir(ap)
 2152         struct vop_readdir_args /* {
 2153                 struct vnode *a_vp;
 2154                 struct uio *a_uio;
 2155                 struct ucred *a_cred;
 2156                 int *a_eofflag;
 2157                 int *a_ncookies;
 2158                 u_long **a_cookies;
 2159         } */ *ap;
 2160 {
 2161         struct vnode *vp = ap->a_vp;
 2162         struct uio *uio = ap->a_uio;
 2163         struct buf *bp;
 2164         struct inode *ip;
 2165         struct direct *dp, *edp;
 2166         u_long *cookies;
 2167         struct dirent dstdp;
 2168         off_t offset, startoffset;
 2169         size_t readcnt, skipcnt;
 2170         ssize_t startresid;
 2171         int ncookies;
 2172         int error;
 2173 
 2174         if (uio->uio_offset < 0)
 2175                 return (EINVAL);
 2176         ip = VTOI(vp);
 2177         if (ip->i_effnlink == 0)
 2178                 return (0);
 2179         if (ap->a_ncookies != NULL) {
 2180                 ncookies = uio->uio_resid;
 2181                 if (uio->uio_offset >= ip->i_size)
 2182                         ncookies = 0;
 2183                 else if (ip->i_size - uio->uio_offset < ncookies)
 2184                         ncookies = ip->i_size - uio->uio_offset;
 2185                 ncookies = ncookies / (offsetof(struct direct, d_name) + 4) + 1;
 2186                 cookies = malloc(ncookies * sizeof(*cookies), M_TEMP, M_WAITOK);
 2187                 *ap->a_ncookies = ncookies;
 2188                 *ap->a_cookies = cookies;
 2189         } else {
 2190                 ncookies = 0;
 2191                 cookies = NULL;
 2192         }
 2193         offset = startoffset = uio->uio_offset;
 2194         startresid = uio->uio_resid;
 2195         error = 0;
 2196         while (error == 0 && uio->uio_resid > 0 &&
 2197             uio->uio_offset < ip->i_size) {
 2198                 error = ffs_blkatoff(vp, uio->uio_offset, NULL, &bp);
 2199                 if (error)
 2200                         break;
 2201                 if (bp->b_offset + bp->b_bcount > ip->i_size)
 2202                         readcnt = ip->i_size - bp->b_offset;
 2203                 else
 2204                         readcnt = bp->b_bcount;
 2205                 skipcnt = (size_t)(uio->uio_offset - bp->b_offset) &
 2206                     ~(size_t)(DIRBLKSIZ - 1);
 2207                 offset = bp->b_offset + skipcnt;
 2208                 dp = (struct direct *)&bp->b_data[skipcnt];
 2209                 edp = (struct direct *)&bp->b_data[readcnt];
 2210                 while (error == 0 && uio->uio_resid > 0 && dp < edp) {
 2211                         if (dp->d_reclen <= offsetof(struct direct, d_name) ||
 2212                             (caddr_t)dp + dp->d_reclen > (caddr_t)edp) {
 2213                                 error = EIO;
 2214                                 break;
 2215                         }
 2216 #if BYTE_ORDER == LITTLE_ENDIAN
 2217                         /* Old filesystem format. */
 2218                         if (vp->v_mount->mnt_maxsymlinklen <= 0) {
 2219                                 dstdp.d_namlen = dp->d_type;
 2220                                 dstdp.d_type = dp->d_namlen;
 2221                         } else
 2222 #endif
 2223                         {
 2224                                 dstdp.d_namlen = dp->d_namlen;
 2225                                 dstdp.d_type = dp->d_type;
 2226                         }
 2227                         if (offsetof(struct direct, d_name) + dstdp.d_namlen >
 2228                             dp->d_reclen) {
 2229                                 error = EIO;
 2230                                 break;
 2231                         }
 2232                         if (offset < startoffset || dp->d_ino == 0)
 2233                                 goto nextentry;
 2234                         dstdp.d_fileno = dp->d_ino;
 2235                         dstdp.d_reclen = GENERIC_DIRSIZ(&dstdp);
 2236                         bcopy(dp->d_name, dstdp.d_name, dstdp.d_namlen);
 2237                         dstdp.d_name[dstdp.d_namlen] = '\0';
 2238                         if (dstdp.d_reclen > uio->uio_resid) {
 2239                                 if (uio->uio_resid == startresid)
 2240                                         error = EINVAL;
 2241                                 else
 2242                                         error = EJUSTRETURN;
 2243                                 break;
 2244                         }
 2245                         /* Advance dp. */
 2246                         error = uiomove((caddr_t)&dstdp, dstdp.d_reclen, uio);
 2247                         if (error)
 2248                                 break;
 2249                         if (cookies != NULL) {
 2250                                 KASSERT(ncookies > 0,
 2251                                     ("ufs_readdir: cookies buffer too small"));
 2252                                 *cookies = offset + dp->d_reclen;
 2253                                 cookies++;
 2254                                 ncookies--;
 2255                         }
 2256 nextentry:
 2257                         offset += dp->d_reclen;
 2258                         dp = (struct direct *)((caddr_t)dp + dp->d_reclen);
 2259                 }
 2260                 bqrelse(bp);
 2261                 uio->uio_offset = offset;
 2262         }
 2263         /* We need to correct uio_offset. */
 2264         uio->uio_offset = offset;
 2265         if (error == EJUSTRETURN)
 2266                 error = 0;
 2267         if (ap->a_ncookies != NULL) {
 2268                 if (error == 0) {
 2269                         ap->a_ncookies -= ncookies;
 2270                 } else {
 2271                         free(*ap->a_cookies, M_TEMP);
 2272                         *ap->a_ncookies = 0;
 2273                         *ap->a_cookies = NULL;
 2274                 }
 2275         }
 2276         if (error == 0 && ap->a_eofflag)
 2277                 *ap->a_eofflag = ip->i_size <= uio->uio_offset;
 2278         return (error);
 2279 }
 2280 
 2281 /*
 2282  * Return target name of a symbolic link
 2283  */
 2284 static int
 2285 ufs_readlink(ap)
 2286         struct vop_readlink_args /* {
 2287                 struct vnode *a_vp;
 2288                 struct uio *a_uio;
 2289                 struct ucred *a_cred;
 2290         } */ *ap;
 2291 {
 2292         struct vnode *vp = ap->a_vp;
 2293         struct inode *ip = VTOI(vp);
 2294         doff_t isize;
 2295 
 2296         isize = ip->i_size;
 2297         if ((isize < vp->v_mount->mnt_maxsymlinklen) ||
 2298             DIP(ip, i_blocks) == 0) { /* XXX - for old fastlink support */
 2299                 return (uiomove(SHORTLINK(ip), isize, ap->a_uio));
 2300         }
 2301         return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
 2302 }
 2303 
 2304 /*
 2305  * Calculate the logical to physical mapping if not done already,
 2306  * then call the device strategy routine.
 2307  *
 2308  * In order to be able to swap to a file, the ufs_bmaparray() operation may not
 2309  * deadlock on memory.  See ufs_bmap() for details.
 2310  */
 2311 static int
 2312 ufs_strategy(ap)
 2313         struct vop_strategy_args /* {
 2314                 struct vnode *a_vp;
 2315                 struct buf *a_bp;
 2316         } */ *ap;
 2317 {
 2318         struct buf *bp = ap->a_bp;
 2319         struct vnode *vp = ap->a_vp;
 2320         ufs2_daddr_t blkno;
 2321         int error;
 2322 
 2323         if (bp->b_blkno == bp->b_lblkno) {
 2324                 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, bp, NULL, NULL);
 2325                 bp->b_blkno = blkno;
 2326                 if (error) {
 2327                         bp->b_error = error;
 2328                         bp->b_ioflags |= BIO_ERROR;
 2329                         bufdone(bp);
 2330                         return (0);
 2331                 }
 2332                 if ((long)bp->b_blkno == -1)
 2333                         vfs_bio_clrbuf(bp);
 2334         }
 2335         if ((long)bp->b_blkno == -1) {
 2336                 bufdone(bp);
 2337                 return (0);
 2338         }
 2339         bp->b_iooffset = dbtob(bp->b_blkno);
 2340         BO_STRATEGY(VFSTOUFS(vp->v_mount)->um_bo, bp);
 2341         return (0);
 2342 }
 2343 
 2344 /*
 2345  * Print out the contents of an inode.
 2346  */
 2347 static int
 2348 ufs_print(ap)
 2349         struct vop_print_args /* {
 2350                 struct vnode *a_vp;
 2351         } */ *ap;
 2352 {
 2353         struct vnode *vp = ap->a_vp;
 2354         struct inode *ip = VTOI(vp);
 2355 
 2356         printf("\tino %lu, on dev %s", (u_long)ip->i_number,
 2357             devtoname(ITODEV(ip)));
 2358         if (vp->v_type == VFIFO)
 2359                 fifo_printinfo(vp);
 2360         printf("\n");
 2361         return (0);
 2362 }
 2363 
 2364 /*
 2365  * Close wrapper for fifos.
 2366  *
 2367  * Update the times on the inode then do device close.
 2368  */
 2369 static int
 2370 ufsfifo_close(ap)
 2371         struct vop_close_args /* {
 2372                 struct vnode *a_vp;
 2373                 int  a_fflag;
 2374                 struct ucred *a_cred;
 2375                 struct thread *a_td;
 2376         } */ *ap;
 2377 {
 2378         struct vnode *vp = ap->a_vp;
 2379         int usecount;
 2380 
 2381         VI_LOCK(vp);
 2382         usecount = vp->v_usecount;
 2383         if (usecount > 1)
 2384                 ufs_itimes_locked(vp);
 2385         VI_UNLOCK(vp);
 2386         return (fifo_specops.vop_close(ap));
 2387 }
 2388 
 2389 /*
 2390  * Kqfilter wrapper for fifos.
 2391  *
 2392  * Fall through to ufs kqfilter routines if needed 
 2393  */
 2394 static int
 2395 ufsfifo_kqfilter(ap)
 2396         struct vop_kqfilter_args *ap;
 2397 {
 2398         int error;
 2399 
 2400         error = fifo_specops.vop_kqfilter(ap);
 2401         if (error)
 2402                 error = vfs_kqfilter(ap);
 2403         return (error);
 2404 }
 2405 
 2406 /*
 2407  * Return POSIX pathconf information applicable to fifos.
 2408  */
 2409 static int
 2410 ufsfifo_pathconf(ap)
 2411         struct vop_pathconf_args /* {
 2412                 struct vnode *a_vp;
 2413                 int a_name;
 2414                 int *a_retval;
 2415         } */ *ap;
 2416 {
 2417 
 2418         switch (ap->a_name) {
 2419         case _PC_ACL_EXTENDED:
 2420         case _PC_ACL_NFS4:
 2421         case _PC_ACL_PATH_MAX:
 2422         case _PC_MAC_PRESENT:
 2423                 return (ufs_pathconf(ap));
 2424         default:
 2425                 return (fifo_specops.vop_pathconf(ap));
 2426         }
 2427         /* NOTREACHED */
 2428 }
 2429 
 2430 /*
 2431  * Return POSIX pathconf information applicable to ufs filesystems.
 2432  */
 2433 static int
 2434 ufs_pathconf(ap)
 2435         struct vop_pathconf_args /* {
 2436                 struct vnode *a_vp;
 2437                 int a_name;
 2438                 int *a_retval;
 2439         } */ *ap;
 2440 {
 2441         int error;
 2442 
 2443         error = 0;
 2444         switch (ap->a_name) {
 2445         case _PC_LINK_MAX:
 2446                 *ap->a_retval = LINK_MAX;
 2447                 break;
 2448         case _PC_NAME_MAX:
 2449                 *ap->a_retval = NAME_MAX;
 2450                 break;
 2451         case _PC_PATH_MAX:
 2452                 *ap->a_retval = PATH_MAX;
 2453                 break;
 2454         case _PC_PIPE_BUF:
 2455                 *ap->a_retval = PIPE_BUF;
 2456                 break;
 2457         case _PC_CHOWN_RESTRICTED:
 2458                 *ap->a_retval = 1;
 2459                 break;
 2460         case _PC_NO_TRUNC:
 2461                 *ap->a_retval = 1;
 2462                 break;
 2463         case _PC_ACL_EXTENDED:
 2464 #ifdef UFS_ACL
 2465                 if (ap->a_vp->v_mount->mnt_flag & MNT_ACLS)
 2466                         *ap->a_retval = 1;
 2467                 else
 2468                         *ap->a_retval = 0;
 2469 #else
 2470                 *ap->a_retval = 0;
 2471 #endif
 2472                 break;
 2473 
 2474         case _PC_ACL_NFS4:
 2475 #ifdef UFS_ACL
 2476                 if (ap->a_vp->v_mount->mnt_flag & MNT_NFS4ACLS)
 2477                         *ap->a_retval = 1;
 2478                 else
 2479                         *ap->a_retval = 0;
 2480 #else
 2481                 *ap->a_retval = 0;
 2482 #endif
 2483                 break;
 2484 
 2485         case _PC_ACL_PATH_MAX:
 2486 #ifdef UFS_ACL
 2487                 if (ap->a_vp->v_mount->mnt_flag & (MNT_ACLS | MNT_NFS4ACLS))
 2488                         *ap->a_retval = ACL_MAX_ENTRIES;
 2489                 else
 2490                         *ap->a_retval = 3;
 2491 #else
 2492                 *ap->a_retval = 3;
 2493 #endif
 2494                 break;
 2495         case _PC_MAC_PRESENT:
 2496 #ifdef MAC
 2497                 if (ap->a_vp->v_mount->mnt_flag & MNT_MULTILABEL)
 2498                         *ap->a_retval = 1;
 2499                 else
 2500                         *ap->a_retval = 0;
 2501 #else
 2502                 *ap->a_retval = 0;
 2503 #endif
 2504                 break;
 2505         case _PC_MIN_HOLE_SIZE:
 2506                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2507                 break;
 2508         case _PC_ASYNC_IO:
 2509                 /* _PC_ASYNC_IO should have been handled by upper layers. */
 2510                 KASSERT(0, ("_PC_ASYNC_IO should not get here"));
 2511                 error = EINVAL;
 2512                 break;
 2513         case _PC_PRIO_IO:
 2514                 *ap->a_retval = 0;
 2515                 break;
 2516         case _PC_SYNC_IO:
 2517                 *ap->a_retval = 0;
 2518                 break;
 2519         case _PC_ALLOC_SIZE_MIN:
 2520                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
 2521                 break;
 2522         case _PC_FILESIZEBITS:
 2523                 *ap->a_retval = 64;
 2524                 break;
 2525         case _PC_REC_INCR_XFER_SIZE:
 2526                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2527                 break;
 2528         case _PC_REC_MAX_XFER_SIZE:
 2529                 *ap->a_retval = -1; /* means ``unlimited'' */
 2530                 break;
 2531         case _PC_REC_MIN_XFER_SIZE:
 2532                 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
 2533                 break;
 2534         case _PC_REC_XFER_ALIGN:
 2535                 *ap->a_retval = PAGE_SIZE;
 2536                 break;
 2537         case _PC_SYMLINK_MAX:
 2538                 *ap->a_retval = MAXPATHLEN;
 2539                 break;
 2540 
 2541         default:
 2542                 error = EINVAL;
 2543                 break;
 2544         }
 2545         return (error);
 2546 }
 2547 
 2548 /*
 2549  * Initialize the vnode associated with a new inode, handle aliased
 2550  * vnodes.
 2551  */
 2552 int
 2553 ufs_vinit(mntp, fifoops, vpp)
 2554         struct mount *mntp;
 2555         struct vop_vector *fifoops;
 2556         struct vnode **vpp;
 2557 {
 2558         struct inode *ip;
 2559         struct vnode *vp;
 2560 
 2561         vp = *vpp;
 2562         ip = VTOI(vp);
 2563         vp->v_type = IFTOVT(ip->i_mode);
 2564         if (vp->v_type == VFIFO)
 2565                 vp->v_op = fifoops;
 2566         ASSERT_VOP_LOCKED(vp, "ufs_vinit");
 2567         if (ip->i_number == ROOTINO)
 2568                 vp->v_vflag |= VV_ROOT;
 2569         *vpp = vp;
 2570         return (0);
 2571 }
 2572 
 2573 /*
 2574  * Allocate a new inode.
 2575  * Vnode dvp must be locked.
 2576  */
 2577 static int
 2578 ufs_makeinode(mode, dvp, vpp, cnp, callfunc)
 2579         int mode;
 2580         struct vnode *dvp;
 2581         struct vnode **vpp;
 2582         struct componentname *cnp;
 2583         const char *callfunc;
 2584 {
 2585         struct inode *ip, *pdir;
 2586         struct direct newdir;
 2587         struct vnode *tvp;
 2588         int error;
 2589 
 2590         pdir = VTOI(dvp);
 2591 #ifdef INVARIANTS
 2592         if ((cnp->cn_flags & HASBUF) == 0)
 2593                 panic("%s: no name", callfunc);
 2594 #endif
 2595         *vpp = NULL;
 2596         if ((mode & IFMT) == 0)
 2597                 mode |= IFREG;
 2598 
 2599         if (pdir->i_effnlink < 2) {
 2600                 print_bad_link_count(callfunc, dvp);
 2601                 return (EINVAL);
 2602         }
 2603         error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp);
 2604         if (error)
 2605                 return (error);
 2606         ip = VTOI(tvp);
 2607         ip->i_gid = pdir->i_gid;
 2608         DIP_SET(ip, i_gid, pdir->i_gid);
 2609 #ifdef SUIDDIR
 2610         {
 2611 #ifdef QUOTA
 2612                 struct ucred ucred, *ucp;
 2613                 gid_t ucred_group;
 2614                 ucp = cnp->cn_cred;
 2615 #endif
 2616                 /*
 2617                  * If we are not the owner of the directory,
 2618                  * and we are hacking owners here, (only do this where told to)
 2619                  * and we are not giving it TO root, (would subvert quotas)
 2620                  * then go ahead and give it to the other user.
 2621                  * Note that this drops off the execute bits for security.
 2622                  */
 2623                 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) &&
 2624                     (pdir->i_mode & ISUID) &&
 2625                     (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) {
 2626                         ip->i_uid = pdir->i_uid;
 2627                         DIP_SET(ip, i_uid, ip->i_uid);
 2628                         mode &= ~07111;
 2629 #ifdef QUOTA
 2630                         /*
 2631                          * Make sure the correct user gets charged
 2632                          * for the space.
 2633                          * Quickly knock up a dummy credential for the victim.
 2634                          * XXX This seems to never be accessed out of our
 2635                          * context so a stack variable is ok.
 2636                          */
 2637                         refcount_init(&ucred.cr_ref, 1);
 2638                         ucred.cr_uid = ip->i_uid;
 2639                         ucred.cr_ngroups = 1;
 2640                         ucred.cr_groups = &ucred_group;
 2641                         ucred.cr_groups[0] = pdir->i_gid;
 2642                         ucp = &ucred;
 2643 #endif
 2644                 } else {
 2645                         ip->i_uid = cnp->cn_cred->cr_uid;
 2646                         DIP_SET(ip, i_uid, ip->i_uid);
 2647                 }
 2648 
 2649 #ifdef QUOTA
 2650                 if ((error = getinoquota(ip)) ||
 2651                     (error = chkiq(ip, 1, ucp, 0))) {
 2652                         if (DOINGSOFTDEP(tvp))
 2653                                 softdep_revert_link(pdir, ip);
 2654                         UFS_VFREE(tvp, ip->i_number, mode);
 2655                         vput(tvp);
 2656                         return (error);
 2657                 }
 2658 #endif
 2659         }
 2660 #else   /* !SUIDDIR */
 2661         ip->i_uid = cnp->cn_cred->cr_uid;
 2662         DIP_SET(ip, i_uid, ip->i_uid);
 2663 #ifdef QUOTA
 2664         if ((error = getinoquota(ip)) ||
 2665             (error = chkiq(ip, 1, cnp->cn_cred, 0))) {
 2666                 if (DOINGSOFTDEP(tvp))
 2667                         softdep_revert_link(pdir, ip);
 2668                 UFS_VFREE(tvp, ip->i_number, mode);
 2669                 vput(tvp);
 2670                 return (error);
 2671         }
 2672 #endif
 2673 #endif  /* !SUIDDIR */
 2674         ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
 2675         ip->i_mode = mode;
 2676         DIP_SET(ip, i_mode, mode);
 2677         tvp->v_type = IFTOVT(mode);     /* Rest init'd in getnewvnode(). */
 2678         ip->i_effnlink = 1;
 2679         ip->i_nlink = 1;
 2680         DIP_SET(ip, i_nlink, 1);
 2681         if (DOINGSOFTDEP(tvp))
 2682                 softdep_setup_create(VTOI(dvp), ip);
 2683         if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) &&
 2684             priv_check_cred(cnp->cn_cred, PRIV_VFS_SETGID, 0)) {
 2685                 ip->i_mode &= ~ISGID;
 2686                 DIP_SET(ip, i_mode, ip->i_mode);
 2687         }
 2688 
 2689         if (cnp->cn_flags & ISWHITEOUT) {
 2690                 ip->i_flags |= UF_OPAQUE;
 2691                 DIP_SET(ip, i_flags, ip->i_flags);
 2692         }
 2693 
 2694         /*
 2695          * Make sure inode goes to disk before directory entry.
 2696          */
 2697         error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp)));
 2698         if (error)
 2699                 goto bad;
 2700 #ifdef MAC
 2701         if (dvp->v_mount->mnt_flag & MNT_MULTILABEL) {
 2702                 error = mac_vnode_create_extattr(cnp->cn_cred, dvp->v_mount,
 2703                     dvp, tvp, cnp);
 2704                 if (error)
 2705                         goto bad;
 2706         }
 2707 #endif
 2708 #ifdef UFS_ACL
 2709         if (dvp->v_mount->mnt_flag & MNT_ACLS) {
 2710                 error = ufs_do_posix1e_acl_inheritance_file(dvp, tvp, mode,
 2711                     cnp->cn_cred, cnp->cn_thread);
 2712                 if (error)
 2713                         goto bad;
 2714         } else if (dvp->v_mount->mnt_flag & MNT_NFS4ACLS) {
 2715                 error = ufs_do_nfs4_acl_inheritance(dvp, tvp, mode,
 2716                     cnp->cn_cred, cnp->cn_thread);
 2717                 if (error)
 2718                         goto bad;
 2719         }
 2720 #endif /* !UFS_ACL */
 2721         ufs_makedirentry(ip, cnp, &newdir);
 2722         error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL, 0);
 2723         if (error)
 2724                 goto bad;
 2725         *vpp = tvp;
 2726         return (0);
 2727 
 2728 bad:
 2729         /*
 2730          * Write error occurred trying to update the inode
 2731          * or the directory so must deallocate the inode.
 2732          */
 2733         ip->i_effnlink = 0;
 2734         ip->i_nlink = 0;
 2735         DIP_SET(ip, i_nlink, 0);
 2736         ip->i_flag |= IN_CHANGE;
 2737         if (DOINGSOFTDEP(tvp))
 2738                 softdep_revert_create(VTOI(dvp), ip);
 2739         vput(tvp);
 2740         return (error);
 2741 }
 2742 
 2743 static int
 2744 ufs_ioctl(struct vop_ioctl_args *ap)
 2745 {
 2746 
 2747         switch (ap->a_command) {
 2748         case FIOSEEKDATA:
 2749         case FIOSEEKHOLE:
 2750                 return (vn_bmap_seekhole(ap->a_vp, ap->a_command,
 2751                     (off_t *)ap->a_data, ap->a_cred));
 2752         default:
 2753                 return (ENOTTY);
 2754         }
 2755 }
 2756 
 2757 /* Global vfs data structures for ufs. */
 2758 struct vop_vector ufs_vnodeops = {
 2759         .vop_default =          &default_vnodeops,
 2760         .vop_fsync =            VOP_PANIC,
 2761         .vop_read =             VOP_PANIC,
 2762         .vop_reallocblks =      VOP_PANIC,
 2763         .vop_write =            VOP_PANIC,
 2764         .vop_accessx =          ufs_accessx,
 2765         .vop_bmap =             ufs_bmap,
 2766         .vop_cachedlookup =     ufs_lookup,
 2767         .vop_close =            ufs_close,
 2768         .vop_create =           ufs_create,
 2769         .vop_getattr =          ufs_getattr,
 2770         .vop_inactive =         ufs_inactive,
 2771         .vop_ioctl =            ufs_ioctl,
 2772         .vop_link =             ufs_link,
 2773         .vop_lookup =           vfs_cache_lookup,
 2774         .vop_markatime =        ufs_markatime,
 2775         .vop_mkdir =            ufs_mkdir,
 2776         .vop_mknod =            ufs_mknod,
 2777         .vop_open =             ufs_open,
 2778         .vop_pathconf =         ufs_pathconf,
 2779         .vop_poll =             vop_stdpoll,
 2780         .vop_print =            ufs_print,
 2781         .vop_readdir =          ufs_readdir,
 2782         .vop_readlink =         ufs_readlink,
 2783         .vop_reclaim =          ufs_reclaim,
 2784         .vop_remove =           ufs_remove,
 2785         .vop_rename =           ufs_rename,
 2786         .vop_rmdir =            ufs_rmdir,
 2787         .vop_setattr =          ufs_setattr,
 2788 #ifdef MAC
 2789         .vop_setlabel =         vop_stdsetlabel_ea,
 2790 #endif
 2791         .vop_strategy =         ufs_strategy,
 2792         .vop_symlink =          ufs_symlink,
 2793         .vop_whiteout =         ufs_whiteout,
 2794 #ifdef UFS_EXTATTR
 2795         .vop_getextattr =       ufs_getextattr,
 2796         .vop_deleteextattr =    ufs_deleteextattr,
 2797         .vop_setextattr =       ufs_setextattr,
 2798 #endif
 2799 #ifdef UFS_ACL
 2800         .vop_getacl =           ufs_getacl,
 2801         .vop_setacl =           ufs_setacl,
 2802         .vop_aclcheck =         ufs_aclcheck,
 2803 #endif
 2804 };
 2805 
 2806 struct vop_vector ufs_fifoops = {
 2807         .vop_default =          &fifo_specops,
 2808         .vop_fsync =            VOP_PANIC,
 2809         .vop_accessx =          ufs_accessx,
 2810         .vop_close =            ufsfifo_close,
 2811         .vop_getattr =          ufs_getattr,
 2812         .vop_inactive =         ufs_inactive,
 2813         .vop_kqfilter =         ufsfifo_kqfilter,
 2814         .vop_markatime =        ufs_markatime,
 2815         .vop_pathconf =         ufsfifo_pathconf,
 2816         .vop_print =            ufs_print,
 2817         .vop_read =             VOP_PANIC,
 2818         .vop_reclaim =          ufs_reclaim,
 2819         .vop_setattr =          ufs_setattr,
 2820 #ifdef MAC
 2821         .vop_setlabel =         vop_stdsetlabel_ea,
 2822 #endif
 2823         .vop_write =            VOP_PANIC,
 2824 #ifdef UFS_EXTATTR
 2825         .vop_getextattr =       ufs_getextattr,
 2826         .vop_deleteextattr =    ufs_deleteextattr,
 2827         .vop_setextattr =       ufs_setextattr,
 2828 #endif
 2829 #ifdef UFS_ACL
 2830         .vop_getacl =           ufs_getacl,
 2831         .vop_setacl =           ufs_setacl,
 2832         .vop_aclcheck =         ufs_aclcheck,
 2833 #endif
 2834 };

Cache object: 6317402b8ee387cc8f7f712202bcbf39


[ 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.